{"id": "2606.18259", "title": "Caring Without Feeling: Affective Dynamics as the Control Layer of Human-AI Agent Collaboration", "abstract": "AI agents that plan, retain memory across sessions, invoke external tools and act with partial autonomy are transforming human--AI collaboration. Research on affective computing, simulated empathy in large language models, trust in automation and AI safety has illuminated important design principles, yet these literatures remain fragmented. No integrated account explains how affective cues operate within agentic collaboration -- settings in which humans delegate, monitor and correct consequential tasks. This Review synthesises computational and interactional mechanisms of affective dynamics: the processes through which affective cues, emotion-like behaviour and perceived agent affect shape trust calibration, delegation decisions, error correction, dependence and governance. We trace how model-generated affective signals enter interaction loops that govern reliance, repair and oversight, and propose a framework that treats affect not as an internal property of AI but as a coordination layer through which humans and agents negotiate capability, uncertainty and responsibility. The framework provides a foundation for calibrated measurement, purposeful design and informed governance.", "date": "2026-05-08", "categories": []}
{"id": "2606.18266", "title": "EMORSION: Examining the Impact of Audio Parameters on Emotional Responses and Immersion in Film", "abstract": "EMORSION is an exploratory proof-of-concept study examining how film audio design shapes audience emotion and immersion in acinema setting. Four film scenes were selected across the horror (2) and drama (2) genres, balanced between mainstream and independent productions. For each scene, multiple alternative audio mixes were created by systematically manipulating three core aspects of audio design, frequency (pitch), dynamics (loudness), and directionality (spatial placement). Three audience groups viewed the scenes, with each group exposed to one manipulated mix alongside a control mix for each scene. Audience responses were assessed through a triangulated multimodal framework combining self-reported emotion and immersion via a questionnaire, physiological measures including heart rate monitoring, and video-based motion tracking. The protocol successfully captured measurable, interpretable differences across audio conditions, indicating that even subtle changes in audio design can shape emotional perception and immersion. Unconventional mixes tended to produce greater variability in audience interpretation, while conventional immersive mixes were associated with stronger cross-audience agreement. These findings establish the feasibility of the EMORSION protocol and motivate larger-scale studies to characterise the role of specific audio parameters in shaping audience experience.", "date": "2026-05-29", "categories": []}
{"id": "2606.18893", "title": "Learning Robust Pair Confidence for Multimodal Emotion-Cause Pair Extraction", "abstract": "Multimodal emotion-cause pair extraction (MECPE) requires reliable pair confidence over candidate pairs. Existing pair scorers commonly use pair-level cross entropy over valid candidates, which treats links mostly independently. This leaves the relative confidence geometry among competing causes under-constrained, allowing gold pairs to stay close to hard negatives or rely on incidental non-gold context. We study this vulnerability as pair-confidence brittleness and propose RPCL (Robust Pair Confidence Learning), a training-only framework for pair-confidence learning. RPCL encourages pair confidence to be both discriminative and stable: gold pairs are separated from row-wise hard negatives through a confidence-difference margin constraint, and clean pair predictions are aligned with predictions from a corrupted view where non-gold contextual utterance representations are partially corrupted. The original clean pair scorer and decoding pipeline are used unchanged at inference time. On ECF, MECAD, and MEC4, RPCL improves the three-seed mean Pair F1 over a matched base model by 2.58 to 2.83 percentage points in the full text-audio-video setting, and improves mean Pair AUPRC on all three datasets. Diagnostic analysis further shows larger gold-negative confidence gaps and lower margin-violation severity. These results suggest that explicitly shaping pair confidence is an effective training strategy for MECPE.", "date": "2026-06-17", "categories": []}
{"id": "2509.03932", "title": "KPoEM: A Human-Annotated Dataset for Emotion Classification and RAG-Based Poetry Generation in Korean Modern Poetry", "abstract": "This study introduces KPoEM (Korean Poetry Emotion Mapping), a novel dataset that serves as a foundation for both emotion-centered analysis and generative applications in modern Korean poetry. Despite advancements in NLP, poetry remains underexplored due to its complex figurative language and cultural specificity. We constructed a multi-label dataset of 7,662 entries (7,007 line-level and 615 work-level), annotated with 44 fine-grained emotion categories from five influential Korean poets. The KPoEM emotion classification model, fine-tuned through a sequential strategy -- moving from general-purpose corpora to the specialized KPoEM dataset -- achieved an F1-micro score of 0.60, significantly outperforming previous models (0.43). The model demonstrates an enhanced ability to identify temporally and culturally specific emotional expressions while preserving core poetic sentiments. Furthermore, applying the structured emotion dataset to a RAG-based poetry generation model demonstrates the empirical feasibility of generating texts that reflect the emotional and cultural sensibilities of Korean literature. This integrated approach strengthens the connection between computational techniques and literary analysis, opening new pathways for quantitative emotion research and generative poetics. Overall, this study provides a foundation for advancing emotion-centered analysis and creation in modern Korean poetry.", "date": "2025-09-04", "categories": []}
{"id": "2606.15121", "title": "When Cognitive Graphs Meet LLMs: BDEI Cognitive Pathways for Panic Emotional Arousal Prediction", "abstract": "Predicting individual panic emotional arousal timing before manifestation is essential for proactive emergency intervention. Existing methods incorporate cognitive elements but none explicitly model the emotional arousal process, making them ill-suited for emotional arousal timing prediction. We argue that grounding prediction in appraisal emotion theory is necessary because it explicitly models this process, but three problems must be solved. (1) Appraisal theory posits that emotion arises from simultaneous evaluation across multiple threat dimensions, yet no prior work fuses these inputs into risk perception. (2) Existing cognitive models lack an Emotion node, decoupling threat appraisal from emotional arousal and forcing emotions to be inferred indirectly from behaviors. (3) Given their generalizable cognitive reasoning, current approaches adopt LLMs as the primary decision-maker, yet overlook the fragility and hallucination-proneness of their outputs. To address these issues, we introduce PanicCognitivePath (PCP), a framework that addresses all three. A Psychological Safety Distance (PSD) model, grounded in psychological distance theory, maps four-domain signals into a unified risk metric as the entry condition for subsequent cognitive reasoning. An explicit Emotion node grounded in appraisal emotion theory is introduced into BDI, forming a Belief-Desire-Emotion-Intention (BDEI) pathway. Agents whose risk metric exceeds the PSD threshold enter this pathway, coupling threat appraisal directly to emotional arousal. The BDEI pathway governs all state transitions while the LLM is confined to parameter estimation for the Belief-to-Desire transition, confining hallucinations to a single step and preventing error propagation. Experiments on Hurricane Sandy show PCP improves arousal timing accuracy by 10.68% over baselines, reduces peak count error to 7.07%.", "date": "2026-06-13", "categories": []}
{"id": "2410.11861", "title": "Investigating Role of Big Five Personality Traits in Audio-Visual Rapport Estimation", "abstract": "Automatic rapport estimation in social interactions is a central component of affective computing. Recent reports have shown that the estimation performance of rapport in initial interactions can be improved by using the participant's personality traits as the model's input. In this study, we investigate whether this findings applies to interactions between friends by developing rapport estimation models that utilize nonverbal cues (audio and facial expressions) as inputs. Our experimental results show that adding Big Five features (BFFs) to nonverbal features can improve the estimation performance of self-reported rapport in dyadic interactions between friends. Next, we demystify how BFFs improve the estimation performance of rapport through a comparative analysis between models with and without BFFs. We decompose rapport ratings into perceiver effects (people's tendency to rate other people), target effects (people's tendency to be rated by other people), and relationship effects (people's unique ratings for a specific person) using the social relations model. We then analyze the extent to which BFFs contribute to capturing each effect. Our analysis demonstrates that the perceiver's and the target's BFFs lead estimation models to capture the perceiver and the target effects, respectively. Furthermore, our experimental results indicate that the combinations of facial expression features and BFFs achieve best estimation performances not only in estimating rapport ratings, but also in estimating three effects. Our study is the first step toward understanding why personality-aware estimation models of interpersonal perception accomplish high estimation performance.", "date": "2024-10-07", "categories": []}
{"id": "2606.15026", "title": "Deep Temporal Modeling and Ensemble Fusion for Multimodal Emotion Recognition from Physiological Signals", "abstract": "Physiological stress and emotion recognition are important for health monitoring and affective computing. In this work, we present a comprehensive evaluation of deep learning models such as Long Short-Term Memory (LSTM), Temporal Convolutional Networks (TCN), and Transformer on the WESAD dataset for multimodal affect recognition using wrist and chest sensor signals. We perform ablation studies to assess the individual contributions of each modality by training models on wrist-only and chest-only inputs. In addition, we implement a late-fusion ensemble strategy that combines predictions from all three architectures trained on multimodal input. We also employ early fusion at the sensor level by concatenating wrist and chest signals before feeding them into each model. Our results show that Transformer models consistently achieve the highest accuracy in multimodal settings, while TCN models perform best in the wrist-only configuration. The ensemble method yields the highest overall accuracy (98.91 +/- 0.13%) and macro-F1 score (98.56 +/- 0.17%). These findings demonstrate the effectiveness of sensor fusion and ensemble-based fusion in developing robust systems for physiological emotion recognition.", "date": "2026-06-12", "categories": []}
{"id": "2606.15351", "title": "Facial Affect Analysis for Service-Oriented Systems: Advances, Challenges, and Future Visions", "abstract": "Facial Affect Analysis (FAA) is evolving from a stand-alone recognition task into a reusable perception capability for Service-Oriented Software Ecosystems (SoSE). This paper preserves the FAA methodological core while reframing recent advances through systems-engineering requirements for composable and dependable services. We review representative progress in static and dynamic expression analysis, action-unit and micro-expression modeling, and modern CNN, Transformer, graph, and hybrid architectures, then interpret these advances by their operational fit in edge, cloud, and hybrid service pipelines. The synthesis emphasizes SoSE concerns that determine deployability: service contracts for uncertainty-aware outputs, latency and availability envelopes, lifecycle monitoring and recalibration, governance-aware integration, and interoperability across independently evolving components. Our analysis shows that benchmark gains alone are insufficient for SoSE readiness; robustness under shift, intervention stability, fairness, privacy posture, and runtime guarantees are equally critical. We conclude with a roadmap for treating FAA as an operational service component with explicit interfaces, measurable quality attributes, and accountable lifecycle management.", "date": "2026-06-13", "categories": []}
{"id": "2606.15694", "title": "MAF: Multimodal Adaptive Few-shot Prompting for Sentiment Analysis with MLLMs", "abstract": "Multimodal large language models (MLLMs) have demonstrated remarkable capabilities in understanding complex multimodal content. However, their performance in sentiment analysis exhibits acute sensitivity to prompt design, rendering static, uniformly applied prompts inherently suboptimal for capturing the nuanced multimodal cues that vary across inputs. To address this limitation, we propose a Multimodal Adaptive Few-Shot Prompting (MAF) framework, which dynamically retrieves and integrates query-relevant demonstrations to elicit the sentiment reasoning capabilities of MLLMs in a context-sensitive manner. MAF constructs a demonstration retrieval module that holistically encodes facial expressions, scene context, and textual semantics, with a lip movement amplitude detection mechanism introduced for accurate speaker identification in multi-person scenarios. Departing from conventional fixed-weight fusion, a lightweight coefficient generation network is trained to output query-conditioned fusion weights in real time, enabling weighted aggregation of multimodal similarity scores to retrieve the top-K most informative demonstrations. Prediction stability is further enhanced through majority voting over multiple candidate outputs generated by the MLLM. Extensive experiments on public benchmark datasets demonstrate that MAF achieves substantial and consistent performance improvements over the corresponding backbone variants and remains competitive with strong multimodal sentiment-analysis baselines.", "date": "2026-06-14", "categories": []}
{"id": "2606.15763", "title": "The Circumplex Degeneracy Behind the Rare-Class Limit in Affect Recognition", "abstract": "In-the-wild expression recognition persistently fails on a few rare emotions, and the standard explanation is class imbalance. Through a controlled multi-task study on two benchmarks, we show the failure is instead a property of affect geometry: the rare classes are degenerate on Russell's circumplex, and that degeneracy bounds what any loss or cost can achieve. Our instrument is a circumplex-cost optimal-transport term that prices expression confusions by their valence-arousal distance. The term improves the official score and expression macro-F1, but a control most studies omit shows the gain is not geometric: a uniform cost, equivalent to a generic confidence penalty, matches it on Aff-Wild2 (p=0.625) and significantly exceeds it on AffectNet (+0.057 over base, larger than the circumplex). What the geometry reshapes is the structure of the errors, making them affectively nearer the truth on Aff-Wild2 (p=0.031 against the uniform control), an effect that does not survive on AffectNet, where a visual confound at the far corner of the circumplex overwhelms it. The rare-class failure, by contrast, is stable across both datasets we examine: the degenerate pairs (anger-fear on Aff-Wild2, anger-contempt on AffectNet) resist frequency-based interventions, the transport term, and an action-unit-augmented cost built specifically to separate them. We conclude that progress on rare expressions requires representations that distinguish the classes, not supervision that reprices their confusions, and we provide the controls and metrics needed to tell the two apart.", "date": "2026-06-14", "categories": []}
{"id": "2606.15779", "title": "Faithful Action-unit Causal Reasoning for Counterfactually Faithful Emotion Explanations", "abstract": "Multimodal models can name the action units (AUs) behind a facial emotion, but their AU->emotion rationales are typically plausible rather than faithful: nothing forces the AUs a model invokes to be the AUs that actually drive its prediction. We cast AU->emotion reasoning as a counterfactual-consistency problem between the rationale, the label, and a structural AU->emotion causal graph G, and propose FACR, which grounds the reasoner in an independently induced, polarity-aware G and trains a counterfactual-faithfulness objective: a do-intervention on an AU that G marks causal for a class must move the prediction, while one it marks irrelevant must leave it unchanged. Faithfulness is thereby both trainable and measurable through a matching interventional metric, which we evaluate against a known causal structure, the PSPI pain-AU composition, as no existing affective-reasoning benchmark allows. We are explicit that this metric tests fidelity to the supplied structure rather than its rediscovery: it asks whether the trained reasoner invokes the AUs the structure marks causal, on held-out subjects and a second dataset. Under subject-independent evaluation on UNBC-PAIN, the objective raises the agreement between the invoked AUs and the PSPI composition from a no-objective baseline of 0.08 to 0.57, at a small detection cost; an unfaithfulness control attributes the gain to the objective. On a cross-dataset emotion transfer, the objective likewise raises fidelity to G on a seven-class task (0.50 to 0.84). Finally, we attach a language verbalizer and extend the audit to the generated text: biasing each action unit's emission by its latent activation makes the rationale faithful by construction, so that ablating an AU removes it from the explanation, a property that transfers to a second language-model backbone, whereas a freely generated rationale is unfaithful.", "date": "2026-06-14", "categories": []}
{"id": "2606.16198", "title": "GRACE: Boosting Video MLLMs with Grounded Action-Centric Evidence for Viewer Sentiment Prediction", "abstract": "Viewer sentiment prediction in video advertisements aims to infer the latent affective response evoked in the audience. To bridge the gap between what is shown and what is felt, models must deduce hidden viewer emotions from explicit visual narratives, concrete character-object interactions, and visible textual cues. However, standard Multimodal Large Language Models (MLLMs) typically rely on holistic frame representations, which leave these fine-grained, affect-relevant events implicit and complicate precise emotional reasoning. To address this, we propose a grounded action-centric evidence augmentation framework that enhances video MLLMs' clue extraction and comprehension by introducing explicit event structure and localized visual evidence. Our method extracts temporally ordered subject-verb-object (SVO) triplets and auxiliary visible textual cues from action-centric video descriptions, grounds subject and object entities as visual entity crops, and then enables the MLLM to perform clue-enhanced emotional reasoning based on these extracted structured clues. In this way, action triplets specify \"what happens\", while grounded visual entity crops anchor \"who or what participates in each event\" to concrete visual evidence. Experiments on the Pitts dataset show consistent improvements over Qwen2.5-VL and Qwen3-VL baselines. Ablation studies, cross-dataset evaluation on AdsQA, and transfer experiments on an emotion-focused TVQA subset further support the effectiveness and generalization of our approach.", "date": "2026-06-15", "categories": []}
{"id": "2606.16687", "title": "From Affect Prediction to Affect Forecasting: Evidence for Distinct Information Sources in Longitudinal Text", "abstract": "Modeling dimensional affect in longitudinal text requires distinguishing current affect estimation from future affective change forecasting. Existing approaches often treat each text as an independent observation and apply similar assumptions to both tasks, without testing whether they rely on different information sources. This paper investigates that distinction using longitudinal self-reported ecological essays and feeling-word entries. We propose the Trait--State Affective Prediction (TSAP) framework and its temporal extension E-TSAP for per-text valence and arousal prediction, evaluated on a held-out prediction test set of 1,737 entries from 91 users. We further propose the Affective Change Forecaster Hybrid (ACF-Hybrid) for next-step affective change forecasting, evaluated on a held-out forecasting test set of 46 users. For prediction, E-TSAP achieves composite Pearson correlations of 0.670 for valence and 0.449 for arousal. For forecasting, textual representations perform worse than compact numeric trajectory baselines: the text-inclusive model achieves only r=0.316 for valence and r=0.284 for arousal, whereas a simple prior-state baseline reaches r=0.615 and r=0.670, respectively. ACF-Hybrid, using dimension-specific numeric trajectory features, achieves r=0.659 for valence and $r=0.658$ for arousal. These results show that textual semantics support current affect prediction, whereas future affective change is better captured through prior numeric trajectory dynamics.", "date": "2026-06-15", "categories": []}
{"id": "2606.16843", "title": "Data-Driven Decoding of Russell's Circumplex Model of Affect", "abstract": "Affective computing increasingly relies on deep learning to represent emotions, yet latent spaces often remain opaque, high-dimensional black boxes. This paper investigates whether Transformers' embeddings recover the geometric regularities of Russell's circumplex model. We unify two complementary experiments testing the hypothesis that, after training models on text and speech, their resulting latent spaces encode a topology consistent with valence-arousal and reproduce human-like neighborhood relations. Specifically, we evaluate deep representations extracted from Transformer-based text (RoBERTa) and speech (wav2vec 2.0) encoders, along with a multimodal Transformer fusion architecture, across naturalistic datasets like MSP-Podcast and controlled LLM-generated stimuli. Our analysis reveals that multimodal fusion of text and audio yields perfect topological alignment with Russell's primary emotion ordering. Furthermore, in a zero-shot setting using generic text embeddings, projected fine-grained emotion terms fall close to their established human-mapped coordinates. Our contribution is a novel, data-driven framework for validating emotion models, demonstrating that Russell's circumplex structure is intrinsically encoded in the embeddings of these modalities rather than being solely an artifact of human labeling, thereby bridging the gap between psychological theory and representation learning.", "date": "2026-06-15", "categories": []}
{"id": "2606.12971", "title": "Predicting Cognitive Load from Speech and Interaction Dynamics in Dyadic Conversations", "abstract": "Estimating cognitive load from speech has largely been studied in controlled laboratory settings, with limited understanding of its reliability in natural collaborative conversations. We investigate whether speech and interaction dynamics predict perceived cognitive load during dyadic conversations. We analyze audio from 53 dyads performing nine collaborative tasks and extract static acoustic, dynamic, and interaction features to train a two-head Gated Recurrent Unit encoder to predict cognitive load scores. Results show conversational interaction provides useful signals for predicting cognitive load related to time pressure, mental work, effort, and task performance. Temporal demand is associated with turn-taking dynamics such as overlap and speaker switch, while mental demand is linked to imbalanced participation between speakers. These findings highlight the importance of task structure and conversational interaction for modeling cognitive load in natural collaborative settings.", "date": "2026-06-11", "categories": []}
{"id": "2601.00181", "title": "Causal Emotion Recognition in Conversation: Context Saturation and Discourse-Marker Evidence", "abstract": "We address two persistent gaps in Emotion Recognition in Conversation: which modeling choices materially affect performance, and how recognition findings connect to interpretable discourse-level patterns. We study both through a systematic investigation on IEMOCAP with cross-dataset validation on MELD. For recognition, we run controlled ablations with 10 random seeds and paired significance tests with multiple-comparisons correction, yielding three findings. First, conversational context is the dominant factor, but performance saturates quickly: roughly 90% of the gain is captured within the most recent 10-30 preceding turns, depending on the label set. Second, hierarchical sentence representations help most in utterance-only settings and show a clear advantage on MELD, but their benefit disappears once turn-level context is available, suggesting that conversational history subsumes much of the intra-utterance structure. Third, integrating an external affective lexicon does not improve results, consistent with pretrained encoders already capturing most of the affective signal needed for ERC. Under a strictly causal setting, our simple models achieve strong performance (82.69% 4-way; 67.07% 6-way weighted F1), showing that competitive accuracy is achievable without future turns. For linguistic analysis, we examine 5,286 discourse-marker occurrences and find a reliable association between emotion and marker position (p < .0001). Sad utterances show reduced left-periphery marker usage (21.9%) relative to other emotions (28-32%), consistent with accounts linking left-periphery markers to active discourse management. This aligns with our recognition results, where Sad benefits most from conversational context (+22 percentage points), suggesting sadness may be more context-dependent than emotions with stronger local pragmatic cues.", "date": "2026-01-01", "categories": []}
{"id": "2606.11197", "title": "MA-DLE: Speech-based Automatic Depression Level Estimation via Memory Augmentation", "abstract": "Speech-based automatic estimation of depression levels is essential for enabling early detection and timely intervention, particularly in resource-constrained mental health settings. In recent years, deep learning has demonstrated impressive success across various domains, including affective computing and mental health assessment. Most existing approaches rely on RNN-based architectures (such as LSTM and GRU) to model temporal information for depression estimation. However, the extracted features often emphasize only a few adjacent speech segments, limiting their ability to capture long-range dependencies. To overcome this limitation, we introduce a memory-based feature augmentation method that enhances the representational capacity of GRU-extracted features. Rather than indiscriminately incorporating historical data, our memory bank is designed to selectively integrate two types of components in order to reduce redundancy and irrelevance: (1) historical temporal features that closely resemble the current GRU output, offering complementary contextual information; and (2) dynamic memory features identified based on feature variability, which capture behavioral and emotional fluctuations indicative of depressive symptoms. To effectively fuse the memory-augmented features with GRU outputs, we further design a Hierarchical Attention Fusion (HAF) module. Our method is evaluated on the widely used DAIC-WOZ and E-DAIC datasets, achieving state-of-the-art performance.", "date": "2026-04-20", "categories": []}
{"id": "2606.11875", "title": "I Understand How You Feel: Enhancing Deeper Emotional Support Through Multilingual Emotional Validation in Dialogue System", "abstract": "Emotional validation - explicitly acknowledging that a user's feelings make sense - has proven therapeutic value but has received little computational attention. Emotional validation in dialogue systems can be decomposed into (i) validating response identification, (ii) validation timing detection, and (iii) validating response generation. To support research on all three subtasks, we release M-EDESConv, a 120k English-Japanese multilingual corpus created through hybrid manual and automatic annotation, and M-TESC, a multilingual spoken-dialogue test set. For timing detection, we propose MEGUMI, a Multilingual Emotion-aware Gated Unit for Mutual Integration, that fuses frozen XLM-RoBERTa semantics with language-specific emotion encoders via cross-modal attention and gated fusion. MEGUMI shows superior performance on both the M-EDESConv and M-TESC datasets, both objectively and subjectively. Finally, our EmoValidBench benchmarks of GPT-4.1 Nano and Llama-3.1 8B indicate that current LLMs generate contextually similar and diverse validating responses, but emotional understanding remains a major area for improvement. Project page: https://github.com/zihaurpang/Multilingual-Emotional-Validation", "date": "2026-06-10", "categories": []}
{"id": "2408.07922", "title": "A Deep Features-Based Approach Using Modified ResNet50 and Gradient Boosting for Visual Sentiments Classification", "abstract": "  The versatile nature of Visual Sentiment Analysis (VSA) is one reason for its rising profile. It isn't easy to efficiently manage social media data with visual information since previous research has concentrated on Sentiment Analysis (SA) of single modalities, like textual. In addition, most visual sentiment studies need to adequately classify sentiment because they are mainly focused on simply merging modal attributes without investigating their intricate relationships. This prompted the suggestion of developing a fusion of deep learning and machine learning algorithms. In this research, a deep feature-based method for multiclass classification has been used to extract deep features from modified ResNet50. Furthermore, gradient boosting algorithm has been used to classify photos containing emotional content. The approach is thoroughly evaluated on two benchmarked datasets, CrowdFlower and GAPED. Finally, cutting-edge deep learning and machine learning models were used to compare the proposed strategy. When compared to state-of-the-art approaches, the proposed method demonstrates exceptional performance on the datasets presented.", "date": "2024-08-15", "categories": []}
{"id": "2606.09837", "title": "Self-EmoQ: Plutchik-Guided Value-based Planning to Drive Streaming Emotional TTS", "abstract": "Emotional interaction is increasingly crucial for conversational AI, yet current systems lack a self-emotion determination mechanism to drive the streaming text-to-speech (TTS) synthesis. We propose an emotion-planning framework that determines the emotion prior to the textual generation, grounding the downstream emotional TTS in a streaming manner. The framework is implemented by a plug-and-play LLM module, initialized from pretrained LLMs, and trained by reinforcement learning (RL) with emotions as the actions. A hybrid reward is employed which combines imitation signals with theory-driven scoring, in which the theory of Plutchik's wheel of emotions is adopted. By experiments on DailyDialog, EmoryNLP, IMEOCAP, and MELD, our method outperforms prompting and finetuning baselines on both emotion determination and response quality. We finally implement an entire streaming pipeline for real-time deployment, with the speech quality confirming the framework's emotional alignment, contextual coherence, and expressive fluency. Codes, cases, and demos are available in https://sixingdeguo.github.io/EmoQ-page/.", "date": "2026-04-21", "categories": []}
{"id": "2606.07585", "title": "Multimodal Group Emotion Recognition In-the-Wild Towards a Privacy-Safe Non-Individual Approach", "abstract": "This thesis addresses group emotion recognition (GER) in-the-wild with a focus on privacy preservation. Unlike traditional emotion recognition methods that rely on individual-level cues such as face, gaze, or voice analysis, this work uses collective audio-video signals to infer emotions at the group level, reducing risks of individual monitoring and surveillance. Two complementary frameworks are proposed. The first is a cross-attention multimodal architecture for audio-video fusion, combined with Frames Attention Pooling (FAP) for temporal aggregation. It is supported by synthetic data augmentation and validated through ablation studies, demonstrating robustness in real-world GER conditions. The second framework, Variational Encoder Multi-Decoder (VE-MD), learns a shared latent space for emotion classification and structural representation prediction, including body and face cues. Two decoding strategies, DETR-based and heatmap-based, are explored to analyze the role of structural representations in group and individual settings. The thesis makes three main contributions: it clarifies the role of multimodality and structural cues in group-level affective computing; introduces two architectures for privacy-preserving multimodal GER; and shows that competitive performance can be achieved without using individual features as input data.", "date": "2026-05-27", "categories": []}
{"id": "2606.07707", "title": "Decoding Naturalistic Emotion Dynamics from the Brain: An LLM-Enhanced Regression Framework", "abstract": "Decoding emotional states from neural signals has been typically framed as a discrete, single-label classification task based on emotionally stable stimuli, a formulation that oversimplifies the continuous, fluid, and co-occurring nature of human affect. This study reconceptualizes emotion decoding by adopting a multi-target regression framework to track multiple overlapping emotional dimensions as continuous trajectories over time. Leveraging the robust generalization capabilities of Large Language Models (LLMs), we extracted fine-grained, continuous sentiment profiles from a naturalistic auditory narrative, Alice in Wonderland, to serve as scalable proxies for subjective affect from human fMRI dataset. Departing from standard classification paradigms or mass-univariate subtractive contrasts that filter out network dynamics, we leverage regularized and kernel-based machine learning algorithms as continuous estimators to track the magnitude of macroscale neural state variations. We demonstrate that models trained on temporal snapshots of Dynamic Functional Connectivity (DFC) significantly outperform static region-of-interest (ROI) amplitude representations, effectively capturing continuous emotional trajectories under rapidly fluctuating narrative input. Furthermore, by implementing graph-theoretical Explainable AI (XAI) techniques, we deconstruct the underlying predictive features to reveal highly interpretable, emotion-specific topological configurations. Collectively, these results highlight the utility of LLM-automated annotation in affective neuroscience and provide compelling empirical evidence for psychological constructionist frameworks, demonstrating that dynamic, distributed network interactions offer superior explanatory power over strictly locationist accounts of emotion.", "date": "2026-06-05", "categories": []}
{"id": "2606.07935", "title": "REACT 2026: The Fourth Multiple Appropriate Facial Reaction Generation Challenge: Personalised MAFRG and Appropriate EEG Reaction Prediction", "abstract": "In dyadic interactions, various human facial reactions could be appropriate for responding to each human speaker behaviour. Following the successful organisation of the REACT 2023, 2024 and 2025 challenge series, a body of generative deep learning (DL) models have been developed for the problem of multiple appropriate facial reaction generation (MAFRG). This year, we propose the REACT 2026 challenge encouraging the development and benchmarking of Machine Learning (ML) models that can generate multiple personalised, appropriate, diverse, realistic and synchronised human-style facial reactions expressed by a specific human listener for responding to each given speaker behaviour. As a key of the challenge, we continuously provide challenge participants with MARS dataset introduced by REACT 2025 but additionally provide individual-level Big-Five personality labels and EEG recordings. This introduces a new one-to-many personalised facial reaction generation setting combining human expressive behavioural, affective and neurophysiological signals, which remains largely unexplored in current dyadic interaction modelling. This paper also presents the challenge guidelines and new baselines on the four proposed sub-challenges: Offline generic and personalised MAFRG as well as Online generic and personalised MAFRG, respectively, which are publicly available at https://github.com/reactmultimodalchallenge/baseline_react2026.", "date": "2026-06-06", "categories": []}
{"id": "2606.08612", "title": "Facial Expression Recognition in the Deep Learning Era: A Systematic Multi-Criteria Review of Methods, Models, Datasets, Performance, Challenges, and Future Research Directions", "abstract": "Facial Expression Recognition (FER) has advanced rapidly over the last decade, driven by the shift from handcrafted descriptors and shallow classifiers to deep convolutional, attention-based, vision-language, and foundation-model architectures, and by the parallel growth of large-scale in-the-wild benchmarks spanning categorical, dimensional, compound, micro-expression, Action Unit (AU), and intensity-estimation tasks. Yet the deep learning-based FER landscape has so far been reviewed only along narrow task-, architecture-, or application-specific axes, leaving a holistic, systematically organized account of its recent advances missing. This survey addresses that gap with a comprehensive review of recent deep learning-based FER, explicitly linked to the wider Facial Affect Recognition (FAR) domain. Its main contributions are: a) A description of FER's evolution into five distinct phases, from handcrafted features and classical machine learning to attention-based, vision-language, and foundation-model approaches, with the key milestone works of each, b) A multi-criteria taxonomy analyzing the literature along seven complementary axes: recognition task, input modality, face pre-processing pipeline, network architecture, learning strategy, acquisition setting, and application domain, c) A per-criterion comparative analysis, with critical insights into the strengths and limitations of each category under in-the-wild conditions, d) A task-organized review of public FER datasets, with their annotation schemes, modalities, and evaluation protocols, e) A compilation of performance metrics and a per-task quantitative comparison of representative state-of-the-art methods on widely adopted benchmarks, and f) A discussion of current challenges and promising future directions.", "date": "2026-06-07", "categories": []}
{"id": "2606.09148", "title": "Explicit Representation Alignment for Multimodal Sentiment Analysis", "abstract": "Multimodal affective analysis aims to understand human sentiment and emotion by jointly modeling heterogeneous modalities such as text and images. However, multimodal models often fail to consistently outperform strong text-only baselines, with performance varying significantly across fusion strategies. In this work, we identify representation misalignment between independently pretrained modality encoders as a key bottleneck for effective multimodal learning, and show through controlled experiments that alignment prior to fusion is often more important than fusion complexity. To address this issue, we propose a unified multimodal affective analysis framework that leverages vision-language models (VLMs) to convert visual content into structured textual descriptions, projecting heterogeneous modalities into a shared linguistic space and enabling interpretable text-centric reasoning. To further improve robustness, we introduce a hybrid learning strategy that combines semantic token selection with a batch-level uniformity regularization objective, encouraging a more dispersed and stable global feature space while mitigating noise introduced by VLM-generated descriptions. Experiments on multiple multimodal sentiment and emotion benchmarks show that our method consistently outperforms strong unimodal and multimodal baselines, achieving state-of-the-art performance. Our analysis further highlights the critical role of representation alignment in multimodal affective learning.", "date": "2026-06-08", "categories": []}
{"id": "2606.07063", "title": "Beyond Universality: The GCC-FER Dataset and Culture-Aware Adaptation for Dynamic Facial Expression Recognition", "abstract": "Dynamic Facial Expression Recognition (DFER) is a key enabling technology in affective computing, human-computer interaction, and intelligent multimedia systems. Despite the significant influence of cultural nuances on FER performance, most existing FER systems assume that emotional expressions are universally consistent across populations. This variation can be attributed to systematic differences in facial muscle activation patterns across cultures. A major challenge in advancing cross-cultural FER lies in the scarcity of culturally diverse benchmark datasets. To address this, a new hybrid multicultural video dataset termed Global Cross-Cultural Facial Expression Recognition (GCC-FER) is introduced. GCC-FER comprises 23,934 video samples spanning four cultural groups (African, Caucasian, East Asian, and South Asian) across seven basic expressions, combining psychologically supervised in-house data collection for underrepresented populations with rigorous ethnicity filtering of existing sources. To the best of our knowledge, GCC-FER is the first large-scale global cross-cultural DFER dataset designed to address these demographic gaps. Leveraging this dataset, behaviorally grounded cultural priors are derived for each cultural group and a global prior for practical deployment. A Culture-Aware FER (CA-FER) system is proposed to mitigate cultural bias by adaptively recalibrating latent facial representations. Extensive experiments on GCC-FER and DFEW demonstrate that the proposed system consistently improves FER performance across multicultural settings.", "date": "2026-06-05", "categories": []}
{"id": "2606.07293", "title": "TargetSEC: Plug-and-Play In-the-Wild Speech Emotion Conversion via Arousal-Conditioned Latent Style Diffusion", "abstract": "Speech Emotion Conversion (SEC) aims to transform the emotion of a source utterance into a target emotion while preserving content and speaker identity. SEC on in-the-wild data is challenging due to the non-parallel nature of training data and complex real-world acoustics. Existing fixed-duration approaches either struggle to shift the emotion effectively (high quality, low conversion) or degrade speech naturalness (low quality, high conversion). We propose TargetSEC, an embedding-driven latent diffusion framework that generates emotion-focused style embeddings conditioned on speaker identity and continuous emotion. Unlike methods that diffuse over spectrograms, TargetSEC operates in a compact latent space. Experiments on the MSP-Podcast dataset show that TargetSEC outperforms current non-duration baselines in conversion accuracy while maintaining high speech quality, and achieves performance comparable to duration-prediction systems without explicit temporal modeling.", "date": "2026-06-05", "categories": []}
{"id": "2511.10254", "title": "Facial-R1: Aligning Reasoning and Recognition for Facial Emotion Analysis", "abstract": "Facial Emotion Analysis (FEA) extends traditional facial emotion recognition by incorporating explainable, fine-grained reasoning. The task integrates three subtasks: emotion recognition, facial Action Unit (AU) recognition, and AU-based emotion reasoning to model affective states jointly. While recent approaches leverage Vision-Language Models (VLMs) and achieve promising results, they face two critical limitations: (1) hallucinated reasoning, where VLMs generate plausible but inaccurate explanations due to insufficient emotion-specific knowledge; and (2) misalignment between emotion reasoning and recognition, caused by fragmented connections between observed facial features and final labels. We propose Facial-R1, a three-stage alignment framework that effectively addresses both challenges with minimal supervision. First, we employ instruction fine-tuning to establish basic emotional reasoning capability. Second, we introduce reinforcement training guided by emotion and AU labels as reward signals, which explicitly aligns the generated reasoning process with the predicted emotion. Third, we design a data synthesis pipeline that iteratively leverages the prior stages to expand the training dataset, enabling scalable self-improvement of the model. Built upon this framework, we introduce FEA-20K, a benchmark dataset comprising 17,737 training and 1,688 test samples with fine-grained emotion analysis annotations. Extensive experiments across eight standard benchmarks demonstrate that Facial-R1 achieves state-of-the-art performance in FEA, with strong generalization and robust interpretability.", "date": "2025-11-13", "categories": []}
{"id": "2604.00819", "title": "Emotion Entanglement and Bayesian Inference for Multi-Dimensional Emotion Understanding", "abstract": "Understanding emotions in natural language is inherently a multi-dimensional reasoning problem, where multiple affective signals interact through context, interpersonal relations, and situational cues. However, most existing emotion understanding benchmarks rely on short texts and predefined emotion labels, reducing this process to independent label prediction and ignoring the structured dependencies among emotions. To address this limitation, we introduce Emotional Scenarios (EmoScene), a theory-grounded benchmark of 4,731 contextrich scenarios annotated with an 8-dimensional emotion vector derived from Plutchik's basic emotions. Motivated by the observation that emotions rarely occur independently, we further propose an entanglement-aware Bayesian inference framework that incorporates emotion co-occurrence statistics to perform joint posterior inference over the emotion vector. This lightweight post-processing does not require any parameter updates and improves the structural consistency of predictions, and yields overall gains of 2.24% Lexical Accuracy without any additional cost. EmoScene therefore provides a challenging benchmark for studying multi-dimensional emotion understanding and the limitations of current language models.", "date": "2026-04-01", "categories": []}
{"id": "2606.03359", "title": "Speech Emotion Recognition using Attention-based LSTM-Network with Residual Connection", "abstract": "Speech emotion recognition is an important component of modern human-computer interaction systems. However, many state-of-the-art approaches rely on large pretrained models with high computational and memory requirements, limiting their applicability. This paper proposes ResLSTM-SA, a lightweight architecture that integrates residual connections with soft attention within an LSTM-based framework. Evaluated on the RAVDESS dataset under strict speaker-independent partitioning, the proposed model outperforms conventional attention-based LSTM baselines and several previously reported CNN- and hybrid CNN-LSTM architectures in terms of unweighted average recall (UAR). The best-performing variant (ResLSTM-SA-h64) achieves a maximum UAR of 0.6517 with only 46.8k trainable parameters, delivering competitive accuracy with three orders of magnitude fewer parameters than large-scale self-supervised alternatives, thereby enabling efficient deployment on edge devices and real-time voice assistants. The source code is available at https://github.com/Mak-Sim/ResLSTM-SER.", "date": "2026-06-02", "categories": []}
{"id": "2606.00170", "title": "UF-AMA: A unified framework for cross-domain emotion recognition via adaptive multimodal alignment", "abstract": "In recent years, emotion recognition based on physiological signals such as electroencephalogram (EEG) has gained considerable attention, as internal physiological data offer greater objectivity and reliability compared to external behavioral data like facial expressions. However, due to distribution shifts caused by individual and contextual differences, along with variations in sample quality across modalities, constructing a cross-domain multimodal emotion recognition model with high generalization and robustness remains a key challenge. In this study, we propose a Unified Framework with Adaptive Multimodal Alignment (UF-AMA) to address cross-subject and cross-session emotion recognition using multimodal physiological signals. First, we construct a cross-modal feature fusion network comprising Transformer encoders and multi-head cross-attention modules, enabling the deep integration of EEG signals and eye-tracking data. Subsequently, we introduce a confidence-aware screening mechanism that dynamically assesses the predictive reliability of each modality branch on target domain samples, partitions samples into different quality subsets, and accordingly applies global consistency alignment and cross-modal distillation. Finally, we propose a multi-level domain adaptation framework that jointly optimizes the marginal and conditional distributions of both local modality-specific and global fusion features, thereby reducing cross-domain distribution shifts at multiple granularities. Extensive experiments on the SEED and SEED-IV datasets demonstrate that UF-AMA achieves state-of-the-art (SOTA) performance in both cross-subject and cross-session tasks. The source code is available at: https://github.com/BetterCoderLab/UF-AMA.", "date": "2026-05-29", "categories": []}
{"id": "2606.00752", "title": "A multimodal dataset of photoplethysmography and continuous behavioral responses to ASMR and nature videos", "abstract": "Autonomous Sensory Meridian Response (ASMR) is a somatosensory phenomenon characterized by pleasant tingling sensations and cardiovascular slowing. However, ASMR research has been hindered by a dearth of standardized, open-access multimodal datasets. To address this limitation, we present REST-ASMR (Response to Environmental & Sensory Triggers), a synchronized multimodal dataset designed to capture behavioral reports and physiological dynamics during ASMR, with nature-relaxation videos as control stimuli. The dataset includes high-resolution photoplethysmography (PPG), time-aligned audiovisual stimuli, and continuous subjective annotations from 34 participants. Technical validation showed high stimulus efficacy (97% responder rate), significant stimulus-specific inter-subject agreement (p < 0.05), and a robust PPG-derived ASMR-specific cardiovascular deceleration. Additionally, a Bidirectional Long-Short Term Memory model successfully predicted subjective ASMR tingle states, achieving video-level ASMR vs. Nature classification with perfect accuracy and a frame-level global mean accuracy of 75.51%, macro F1-score of 71.86%, and 100% Nature-baseline specificity, under a strict, leakage-free subject-video double-independent 4-fold cross-validation. REST-ASMR constitutes a dense temporal foundation for affective computing, multimodal research, and the development of personalized models of relaxation-related responses.", "date": "2026-05-30", "categories": []}
{"id": "2606.00851", "title": "Sympatheia: Emotionally Adaptive Voice Assistant with Continuous Affect Conditioning", "abstract": "Empathetic spoken dialogue systems must infer a user's emotional state to respond appropriately, yet everyday speech often carries weak, neutral, or ambiguous affective cues. To address this, we introduce Sympatheia, a speech-to-speech dialogue framework conditioned on affect inferred from the user's speech and, when available, explicit affect specifications provided as a continuous valence--arousal (VA) control signal by a multimodal sensing module or user interface. To train our model, we construct Sympatheia-18k, an emotion-conditioned synthetic spoken dialogue corpus with 12 emotion anchors. This dataset includes an emotional split for learning affective speech behavior, and a neutral split that pairs emotionally neutral queries with multiple emotion-conditioned responses to isolate explicit emotion control in emotionally ambiguous cases. Empirical results show that Sympatheia outperforms speech conversational baselines in generating responses whose semantic content and spoken delivery are both emotionally appropriate. We further show that the same VA interface can integrate emotion estimates from diverse sensing modules, including facial expression, biosignals, and textual affect descriptions, improving response alignment when speech alone provides limited emotional evidence. These results suggest that continuous affect conditioning is an effective practical step for building emotionally adaptive voice assistants.", "date": "2026-05-30", "categories": []}
{"id": "2606.01069", "title": "A Multiscale Network with Supervised Contrastive Learning for Real-Time Facial Emotion Recognition", "abstract": "Real-time emotion recognition from facial expressions is a challenging task, particularly in video-based scenarios where multiple emotional states may occur over time. The difficulty increases further due to the fact that each emotional state is associated with facial expressions that vary significantly across individuals. The change of facial expressions portraying emotional state is not discrete, but rather continuous, which is very challenging to represent through computational aids. A system with the ability to detect variations in facial expressions can have a significant impact on determining the emotional state of an individual. Such a system can be very beneficial for psychologists during counseling by providing additional insights into the emotional state of a subject. In this paper, a deep learning-based system is presented to detect emotional changes in real-time video of a person by modeling the change in facial expressions. The current study is conducted on a standard dataset for training of the deep learning system and the system has provided very satisfactory outcomes in this respect.", "date": "2026-05-31", "categories": []}
{"id": "2606.01906", "title": "Bayesian Spectral Emotion Transition Discovery from Multi-Annotator Disagreement", "abstract": "Emotions evolve through the dynamics of conversation, and understanding their transition structure is foundational to applications ranging from mental-health screening to dialogue systems. However, existing studies typically compress multi-rater judgments into a single hard label by majority voting, discarding the uncertainty signal needed to understand turn-to-turn transitions. In this article, we propose Bayesian Spectral Emotion Transition Discovery (BSETD), a two-stage framework that discovers emotion-transition structure from multi-rater soft labels. In the first stage, a hierarchical Dirichlet-Multinomial posterior is constructed through the outer product of soft labels, equipping each cell of the K x K transition matrix with a credible interval and Benjamini-Hochberg (BH) false discovery rate (FDR)-controlled significance. In the second stage, the symmetrized graph Laplacian is spectrally decomposed to separate a low-frequency (inertia) component from a high-frequency (contagion) component. On EmotionLines, BSETD simultaneously recovers the signatures of two distinct affective spaces: the Plutchik-adjacent transitions disgust to anger (log2 lift +0.94) and anger to disgust (+0.86) are over-represented, while the Russell-valence-reversed transitions joy to anger (-0.90) and anger to joy (-0.89) are under-represented. A five-source cross-corpus validation yields pairwise Pearson correlations in 0.91-0.98 within English, 0.79-0.85 against Chinese M3ED, and 0.979 between the human hard labels and the LLM virtual soft labels on the same utterance set, demonstrating that a pipeline preserving annotator uncertainty bridges the computational study of emotion dynamics with established psychological theory.", "date": "2026-06-01", "categories": []}
{"id": "2511.11406", "title": "Robust Low-Rank Sparse Framework for Video-Based Affective Computing", "abstract": "Video-based Affective Computing (VAC), vital for emotion analysis and human-computer interaction, suffers from model instability and representational degradation due to complex emotional dynamics. Since the meaning of different emotional fluctuations may differ under different emotional contexts, the core limitation is the lack of a hierarchical structural mechanism to disentangle distinct affective components, i.e., emotional bases (the long-term emotional tone), and transient fluctuations (the short-term emotional fluctuations). To address this, we propose the Low-Rank Sparse Emotion Understanding Framework (LSEF), a unified model grounded in the Low-Rank Sparse Principle, which theoretically reframes affective dynamics as a hierarchical low-rank sparse compositional process. LSEF employs three plug-and-play modules, i.e., the Stability Encoding Module (SEM) captures low-rank emotional bases; the Dynamic Decoupling Module (DDM) isolates sparse transient signals; and the Consistency Integration Module (CIM) reconstructs multi-scale stability and reactivity coherence. This framework is optimized by a Rank Aware Optimization (RAO) strategy that adaptively balances gradient smoothness and sensitivity. Extensive experiments across multiple datasets confirm that LSEF significantly enhances robustness and dynamic discrimination, which further validates the effectiveness and generality of hierarchical low-rank sparse modeling for understanding affective dynamics.", "date": "2025-11-14", "categories": []}
{"id": "2602.01173", "title": "EEmo-Logic: A Unified Dataset and Multi-Stage Framework for Comprehensive Image-Evoked Emotion Assessment", "abstract": "Understanding the multi-dimensional attributes and intensity nuances of image-evoked emotions is pivotal for advancing machine empathy and empowering diverse human-computer interaction applications. However, existing models are still limited to coarse-grained emotion perception or deficient reasoning capabilities. To bridge this gap, we introduce \\textbf{EEmoDB}, the largest image-{\\ul e}voked {\\ul emo}tion understanding {\\ul d}ataset to date. It features $5$ analysis dimensions spanning $5$ distinct task categories, facilitating comprehensive interpretation. Specifically, we compile $1.2M$ question-answering (QA) pairs (EEmoDB-QA) from $125K$ images via automated generation, alongside a $36K$ dataset (EEmoDB-Assess) curated from $25K$ images for fine-grained assessment. Furthermore, we propose \\textbf{EEmo-Logic}, an \\textbf{all-in-one} multimodal large language model (MLLM) developed via instruction fine-tuning and task-customized group relative preference optimization (GRPO) with novel reward design. Extensive experiments demonstrate that EEmo-Logic achieves robust performance in in-domain and cross-domain datasets, excelling in emotion QA and fine-grained assessment. The dataset and code are available at https://github.com/workerred/EEmo-Logic.", "date": "2026-02-01", "categories": []}
{"id": "2605.30099", "title": "Evaluation of Conversational Agents: Understanding Culture, Context and Environment in Emotion Detection", "abstract": "Valuable decisions and highly prioritized analysis now depend on applications such as facial biometrics, social media photo tagging, and human robots interactions. However, the ability to successfully deploy such applications is based on their efficiencies on tested use cases taking into consideration possible edge cases. Over the years, lots of generalized solutions have been implemented to mimic human emotions including sarcasm. However, factors such as geographical location or cultural difference have not been explored fully amidst its relevance in resolving ethical issues and improving conversational AI (Artificial Intelligence). In this paper, we seek to address the potential challenges in the usage of conversational AI within Black African society. We develop an emotion prediction model with accuracies ranging between 85% and 96%. Our model combines both speech and image data to detect the seven basic emotions with a focus on also identifying sarcasm. It uses 3-layers of the Convolutional Neural Network in addition to a new Audio-Frame Mean Expression (AFME) algorithm and focuses on model pre-processing and post-processing stages. In the end, our proposed solution contributes to maintaining the credibility of an emotion recognition system in conversational AIs.", "date": "2026-05-28", "categories": []}
{"id": "2604.21534", "title": "UKP_Psycontrol at SemEval-2026 Task 2: Modeling Valence and Arousal Dynamics from Text", "abstract": "This paper presents our system developed for SemEval-2026 Task 2. The task requires modeling both current affect and short-term affective change in chronologically ordered user-generated texts. We explore three complementary approaches: (1) LLM prompting under user-aware and user-agnostic settings, (2) a pairwise Maximum Entropy (MaxEnt) model with Ising-style interactions for structured transition modeling, and (3) a lightweight neural regression model incorporating recent affective trajectories and trainable user embeddings. Our findings indicate that LLMs effectively capture static affective signals from text, whereas short-term affective variation in this dataset is more strongly explained by recent numeric state trajectories than by textual semantics. Our system ranked first among participating teams in both Subtask 1 and Subtask 2A based on the official evaluation metric.", "date": "2026-04-23", "categories": []}
{"id": "2605.26624", "title": "MSCGC-KAN: Multi-scale Causal Graph Convolution and Kolmogorov-Arnold Feature Mapping for EEG Emotion Recognition", "abstract": "Electroencephalogram (EEG)-based emotion recognition is an important affective computing task, and recent EEG foundation models provide useful generic representations for downstream adaptation. However, under the fine-tuning setting, three limitations remain prominent: insufficient modeling of multi-scale emotional dynamics, inadequate exploitation of inter-channel functional connectivity, and the limited expressive power of simple linear classification heads. To address these issues, this paper proposes a new EEG emotion recognition method, termed MSCGC-KAN, which introduces a structured task head composed of multi-scale causal graph convolution and Kolmogorov--Arnold feature mapping. Built on a pre-trained CBraMod backbone, MSCGC-KAN enhances downstream adaptation by jointly strengthening multi-scale temporal modeling, learnable inter-channel connectivity modeling, and nonlinear discriminative mapping within a compact task-specific head. This design preserves the representation advantage of the foundation model while making the classifier more sensitive to emotion-related spatiotemporal patterns. Extensive experiments are conducted on the public FACED and SEED-VII datasets. The proposed method achieves a balanced accuracy of 60.66\\%, a Cohen's Kappa of 0.5525, and a weighted F1-score of 60.40\\% on FACED, and obtains 33.27\\%, 0.2223, and 33.64\\%, respectively, on SEED-VII. Compared with the CBraMod+Linear baseline, the balanced accuracy is improved by 5.91 and 2.03 percentage points on the two datasets, respectively. These results indicate that structured task-head design is an effective way to improve EEG emotion recognition when fine-tuning pre-trained EEG models.", "date": "2026-05-26", "categories": []}
{"id": "2605.27451", "title": "From Affect to Complex Behavior: Advancing Multimodal Human-Centered AI at the 10th ABAW Workshop & Competition", "abstract": "The 10th Affective & Behavior Analysis in-the-Wild (ABAW) Workshop and Competition, held at CVPR 2026, continues to advance research on modelling, analysis, understanding of human affect and behavior in real-world, unconstrained environments. The workshop maintains its dual structure, comprising both a competition and a paper track. The ABAW Competition introduces a diverse set of challenges targeting key aspects of affective and behavioral understanding, including continuous affect (valence-arousal) estimation, discrete affect (expression and action unit) recognition, as well as more complex behavior analysis tasks, such as emotional mimicry intensity estimation, ambivalence/hesitancy recognition and fine-grained violence detection. These challenges are built upon large-scale in-the-wild datasets, providing comprehensive benchmarks for state-of-the-art approaches. In parallel, the paper track presents a wide range of contributions spanning pose, motion & behavior estimation, affect modelling & multimodal learning, benchmarks, datasets & evaluation protocols, fairness, robustness & deployment. Overall, the 10th ABAW Workshop and Competition continues to serve as a key platform for benchmarking, collaboration and innovation, shaping the development of next-generation multimodal, human-centered AI systems.", "date": "2026-05-24", "categories": []}
{"id": "2605.27479", "title": "Resource-Constrained Affect Modelling via Variance Regularisation Pruning", "abstract": "Affective computing systems are increasingly embedded in pervasive and interactive environments, such as adaptive games, assistive technologies, and resource-constrained platforms, where computational efficiency must be balanced with reliability across diverse users. Model pruning offers an effective way to reduce computational demands, yet existing approaches typically optimise for sparsity alone, without accounting for how parameter removal impacts robustness across individuals. In this work, we introduce Variance-Regularised Pruning (VR), a pruning framework that explicitly incorporates cross-participant stability into the sparsification process. Rather than relying solely on average prediction error, VR evaluates each connection based on its joint contribution to both prediction accuracy and variability across users, prioritising parameters that remain reliable under distributional differences. We evaluate the proposed approach on the AGAIN dataset, which includes arousal annotations collected across nine affect-eliciting game environments. Experimental results demonstrate that VR maintains competitive Concordance Correlation Coefficient (CCC) performance even at 80\\% sparsity without additional fine-tuning, highlighting its suitability for deployment in real-world, resource-limited affect-aware systems. Overall, the proposed framework supports the development of compact, robust affective models that can operate reliably in real-world interactive environments.", "date": "2026-05-26", "categories": []}
{"id": "2605.28228", "title": "When Seekers Are Hard to Help: Evaluating Emotional Support Dialogue Systems in Worst-Case Interactions", "abstract": "Emotional Support Dialogue Systems (ESDSes) are increasingly evaluated and trained with LLM-simulated seekers. However, such simulated seekers often behave as cooperative, average-case users who disclose clearly, respond constructively, and accept support within a few turns. This can lead to overly optimistic evaluation and obscure whether ESDSes can handle difficult help-seeking interactions. In this work, we study ESDS evaluation under worst-case interactions, where seekers are hard to help due to low engagement, resistance, limited self-disclosure, emotional volatility, or rigid negative interpretations. We first conduct an expert simulation study with eight experienced counselling professionals, who simulate difficult seekers, interact with existing Chinese ESDSes, provide scale ratings, and participate in semi-structured interviews. Based on this study, we derive worst-case seeker behaviours and identify key limitations of current systems. We then propose a worst-case evaluation framework consisting of an LLM-based worst-case seeker simulator and four worst-case-oriented metrics: Deep Emotional Understanding, Guided Exploration, Balanced Emotional Support, and Authentic and Grounded Support. Evaluating 17 systems, we find that nearly all models suffer substantial performance drops under worst-case interactions. Large general-purpose LLMs are generally more robust than specialised ESDSes, but even the strongest models struggle to sustain engagement and improve seekers' emotional states. Finally, we show that worst-case simulation can also generate useful training data, improving the robustness of smaller models.", "date": "2026-05-27", "categories": []}
{"id": "2605.28575", "title": "A Conflict-Aware Penalty and Statistical Loss Framework for Balancing Modalities and Enhancing Stability in Multimodal Sentiment Analysis", "abstract": "Multimodal Sentiment Analysis (MSA) fuses text, acoustic, and visual streams to infer sentiment. Because pre-trained text encoders are far more expressive than their acoustic and visual counterparts, the text modality tends to dominate optimization, suppressing weaker modalities and inducing gradient norm conflicts that destabilize training. To address this, we propose a Conflict-aware Penalty (CP) that detects and penalizes gradient norm conflicts at each training step, and a Statistical Loss (SL) that aligns predicted distribution statistics with empirical input statistics. Crucially, CP prevents dominant modality gradients from interfering with the SL objective, enabling synergistic training within a unified framework incorporating adaptive modality encoding, gated cross-modal fusion, and unimodal auxiliary heads. Experiments on CMU-MOSI demonstrate state-of-the-art performance, with ablation studies confirming the effectiveness of each component.", "date": "2026-05-27", "categories": []}
{"id": "2605.28810", "title": "Affective Music Recommendation: A Rollout-Based World Model for Offline Preference Optimization", "abstract": "Functional music applications, from consumer focus and sleep aids to clinical interventions, share a distinctive recommendation problem: success is defined by the listener's affective state, but online experimentation on emotion is ethically constrained, particularly for clinical populations who cannot reliably skip a song or report distress. We describe AMRS, the Affective Music Recommendation System deployed on LUCID's health-and-wellness platforms, which serve clinical users (primarily older adults with neurocognitive conditions) and consumer-wellness users across energize, focus, calm, and sleep modes. AMRS is built around a rollout-based world model: a causal transformer trained on logged listening data to jointly predict engagement, binary rating, and self-reported valence and arousal. The world model serves both as an in-silico simulator for offline policy training and as a stress-testing tool before deployment. A recommender policy initialized by behaviour cloning is fine-tuned offline with Direct Preference Optimization (DPO) against a configurable multi-objective utility function. Under a strict cold-start protocol, the world model predicts both behavioural and affective signals with usable fidelity; DPO improves predicted valence and arousal over the cloned baseline while maintaining a similar diversity profile and avoiding the distributional collapse produced by greedy optimization. We position the work as an early deployed validation of a methodology for affective recommendation when online experimentation is ethically untenable.", "date": "2026-05-27", "categories": []}
{"id": "2605.26262", "title": "Dimensional Distribution Emotion State: Leveraging Valence and Arousal as a Common Embedding Space for Visual Emotion Analysis", "abstract": "Museums are important sites for the dissemination of culture and art. They are institutions rooted in history and tradition; their exhibitions are often designed to highlight these aspects. Recently, a new approach is being explored in the field: emotion-based exhibitions. These exhibitions are designed specifically to elicit emotions in the visitors, in order to maximize engagement, and as a way to democratize access to art and attract a wider, more diverse audience. To do so, the emotional content of the artworks must first be extracted, however, manually annotating the artworks by experts is a prohibitively labor-intensive process, and risks introducing the personal bias of curators. To assist the museum curators in their design of these exhibitions, we wish to develop a tool that can predict the emotional response evoked by a work of art. In this article, we leverage a continuous bi-dimensional emotion space to enhance emotion representations and the training process of deep learning models. Drawing inspiration from existing categorical and dimensional emotion representations, we introduce a new representation, Dimensional Distribution Emotion State (DDES), along with a pipeline for multi-dataset training. We show that DDES provides multiple advantages compared to widely used representations while exhibiting similar baseline performance.", "date": "2026-05-25", "categories": []}
{"id": "2605.27240", "title": "ENPMR-Bench: Benchmarking Proactive Memory Retrieval for Emotional Support Agents", "abstract": "Memory-augmented language agents are increasingly deployed in affective applications such as emotional support, where understanding and responding to users' latent emotional needs is critical. However, existing research often treats memory as a tool for factual retrieval, overlooking its role in shaping users' emotional experiences. In this work, we introduce ENPMR-Bench, a benchmark for evaluating Emotional Need-aware Proactive Memory Retrieval (ENPMR), a core capability that enables agents to infer users' latent emotional needs and proactively retrieve appropriate memories to support empathetic interaction. Grounded in Maslow's hierarchy of needs, ENPMR-Bench includes over 1,800 memory-augmented dialogues and defines structured mappings between emotional needs and supportive memory types. Experimental results demonstrate that current retrieval paradigms, including both embedding-based and LLM-driven approaches, exhibit substantial deficiencies, with empathy scores significantly lagging behind golden memory conditions. While chain-of-thought prompting improves the alignment between inferred emotional needs and retrieved memories to some extent, a notable performance gap remains. Together, these findings reveal critical limitations in current agents and outline directions for advancing personalized emotional support through need-sensitive memory retrieval.", "date": "2026-05-26", "categories": []}
{"id": "2509.00056", "title": "Apex-Centered Spatio-Temporal Rank Pooling and Gradient Attention for Micro-Expression Recognition", "abstract": "Micro-expression recognition (MER) is a challenging task due to the subtle and fleeting nature of micro-expressions. Traditional input modalities, such as Apex Frame, Optical Flow, and Dynamic Image, often fail to adequately capture these brief facial movements, resulting in suboptimal performance. In this study, we introduce the Micro-expression Spatio-Temporal Image (MESTI), a micro-expression-specific reformulation of dynamic rank pooling that transforms a video sequence into a single image while emphasizing the onset-apex-offset temporal pattern of micro-expressions. Additionally, we present the Micro-expression Gradient Attention Network (MEGANet), which incorporates a proposed Gradient Attention block to enhance the extraction of fine-grained motion features from micro-expressions. By combining MESTI and MEGANet, we aim to establish a more effective approach to MER. Extensive experiments were conducted to evaluate the effectiveness of MESTI, comparing it with existing input modalities across regular architectures. Moreover, we demonstrate that replacing the input of previously published MER networks with MESTI leads to consistent performance improvements. The performance of MEGANet is also evaluated, showing that our proposed network achieves state-of-the-art results on the SMIC-HS, SAMM and competitive performance on CASMEII datasets, it also achieves leading performance in the reported cross-dataset evaluation settings. The combination of MESTI and MEGANet consistently outperforms the compared methods. These findings underscore the potential of MESTI as a superior input modality and MEGANet as an advanced recognition network, aiming to more effective MER systems in a variety of applications.", "date": "2025-08-25", "categories": []}
{"id": "2510.02171", "title": "Go witheFlow: Real-time Emotion Driven Audio Effects Modulation", "abstract": "Music performance is a distinctly human activity, intrinsically linked to the performer's ability to convey, evoke, or express emotion. Machines cannot perform music in the human sense; they can produce, reproduce, execute, or synthesize music, but they lack the capacity for affective or emotional experience. As such, music performance is an ideal candidate through which to explore aspects of collaboration between humans and machines. In this paper, we introduce the witheFlow system, designed to enhance real-time music performance by automatically modulating audio effects based on features extracted from both biosignals and the audio itself. The system, currently in a proof-of-concept phase, is designed to be lightweight, able to run locally on a laptop, and is open-source given the availability of a compatible Digital Audio Workstation and sensors.", "date": "2025-10-02", "categories": []}
{"id": "2512.23076", "title": "Multimodal Functional Maximum Correlation for Emotion Recognition", "abstract": "Emotional states manifest as coordinated yet heterogeneous physiological responses across central and autonomic systems, posing a fundamental challenge for multimodal representation learning in affective computing. Learning such joint dynamics is further complicated by the scarcity and subjectivity of affective annotations, which motivates the use of self-supervised learning (SSL). However, most existing SSL approaches rely on pairwise alignment objectives, which are insufficient to characterize dependencies among more than two modalities and fail to capture higher-order interactions arising from coordinated brain and autonomic responses.\n  To address this limitation, we propose Multimodal Functional Maximum Correlation (MFMC), a principled SSL framework that maximizes higher-order multimodal dependence through a Dual Total Correlation (DTC) objective. By deriving a tight sandwich bound and optimizing it using a functional maximum correlation analysis (FMCA) based trace surrogate, MFMC captures joint multimodal interactions directly, without relying on pairwise contrastive losses.\n  Experiments on three public affective computing benchmarks demonstrate that MFMC consistently achieves state-of-the-art or competitive performance under both subject-dependent and subject-independent evaluation protocols, highlighting its robustness to inter-subject variability. In particular, MFMC improves subject-dependent accuracy on CEAP-360VR from 78.9% to 86.8%, and subject-independent accuracy from 27.5% to 33.1% using the EDA signal alone. Moreover, MFMC remains within 0.8 percentage points of the best-performing method on the most challenging EEG subject-independent split of MAHNOB-HCI. Our code is available at https://github.com/DY9910/MFMC.", "date": "2025-12-28", "categories": []}
{"id": "2603.20479", "title": "Profiling learners' affective engagement: Emotion AI, intercultural pragmatics, and language learning", "abstract": "Learning another language can be a highly emotional process, typically characterized by numerous frustrations and triumphs, big and small. For most learners, language learning does not follow a linear, predictable path, its zigzag course shaped by motivational (or demotivating) variables such as personal characteristics, teacher/peer relationships, learning materials, and dreams of a future L2 (second language) self. While some aspects of language learning (reading, grammar) are relatively mechanical, others can be stressful and unpredictable, especially conversing in the target language. That experience necessitates not only knowledge of structure and lexis, but also the ability to use the language in ways that are appropriate to the social and cultural context. A new opportunity to practice conversational abilities has arrived through the availability of AI chatbots, with both advantages (responsive, non-judgmental) and drawbacks (emotionally void, culturally biased). This column explores aspects of emotion as they arise in technology use and in particular how automatic emotion recognition and simulated human responsiveness in AI systems interface with language learning and the development of pragmatic and interactional competence. Emotion AI, the algorithmically driven interpretation of users' affective signals, has been seen as enabling greater personalized learning, adapting to perceived learner cognitive and emotional states. Others warn of emotional manipulation and inappropriate and ineffective user profiling", "date": "2026-03-20", "categories": []}
{"id": "2605.21417", "title": "Ordering Matters: Rank-Aware Selective Fusion for Blended Emotion Recognition", "abstract": "Blended emotion recognition is challenging because emotions are often expressed as mixtures of subtle and overlapping multimodal cues rather than a single dominant signal. We propose a rank-aware multi-encoder framework that selectively combines complementary representations from diverse pre-extracted video and audio encoders. Our method projects heterogeneous encoder features into a shared latent space, estimates sample-wise encoder importance through an attention-based gating module, and fuses only the top-n most informative encoders. To better model blended emotions, we decouple prediction into presence and salience heads and align them through probability-level fusion. We further incorporate feature-level unsupervised domain adaptation without pseudo-labeling to improve robustness under distribution shift. Experiments on the BlEmoRE challenge show that the proposed framework outperforms strong individual encoders and na\\\"ive multi-encoder fusion baselines. Our final system ranked 2nd in the competition, supporting the effectiveness of rank-aware selective fusion for fine-grained blended emotion recognition.", "date": "2026-05-20", "categories": []}
{"id": "2605.24712", "title": "Hardware-Aware Federated Learning for Speech Emotion Recognition", "abstract": "Federated learning (FL) enables privacy-preserving collaborative training across distributed edge devices, but real deployments involve heterogeneous clients with different processing power, memory capacity, and communication latency, which often increase round duration and system cost. This paper proposes a hardware-aware federated learning framework for emotion recognition on session-partitioned IEMOCAP that integrates hardware profiling, top-K client selection, and adaptive local epochs within a unified training loop. We compare the method against FedAvg, FedProx, and random top-K selection under a non-IID setup and show that, across 50 federated rounds and 5 independent trials, the proposed approach achieves competitive validation accuracy (0.352), reduces total training time by about 36.5% compared to FedAvg, and lowers cumulative communication cost by 40%.", "date": "2026-05-23", "categories": []}
{"id": "2605.25061", "title": "GL-LFGNN:A Global-Local Dual-branch Causal Graph Neural Network Based on Liang-Kleeman Information Flow for EEG Emotion Recognition", "abstract": "EEG-based emotion recognition holds significant promise for objective diagnosis of mood disorders. Graph neural networks (GNNs) have emerged as the dominant paradigm for modeling inter-channel dependencies in EEG, yet existing approaches rely on symmetric adjacency matrices derived from spatial proximity or functional correlations that fundamentally capture statistical associations rather than directed causal influences, which conflicts with the inherently asymmetric, causally-driven nature of neural information flow. To bridge this gap, we propose GL-LFGNN, a Global-Local Dual-branch Causal Graph Neural Network grounded in Liang-Kleeman information flow theory. Unlike Granger causality that merely assesses temporal precedence, our approach rigorously quantifies causal strength from a dynamical systems perspective, yielding neurophysiologically interpretable directed graphs. A dual-branch architecture further integrates whole-brain connectivity with region-specific processing aligned to established functional neuroanatomy. On the MEEG dataset, GL-LFGNN achieves 86.17% (Arousal) and 86.71% (Valence) accuracy with only 37K parameters -- approximately 10% of the current state-of-the-art -- demonstrating that principled causal modeling can simultaneously enhance interpretability, generalization, and computational efficiency. Code will be released.", "date": "2026-05-24", "categories": []}
{"id": "2605.25208", "title": "They Are Not the Same: Direct Causes Are Not Grounded Emotion Explanations", "abstract": "Emotion-Cause Pair Extraction (ECPE) was introduced to explain why an emotion occurs, but this goal is now often reduced to binary pair/non-pair prediction. This proxy is useful for direct-cause extraction, yet easy to over-read as evidence grounded emotion explanation. We show that this interpretation is only partially valid. In IEMO-MECP, 90.9% of original positives remain emo-cause and 95.0% of original negatives remain non-pair, confirming that the binary ECPE task is largely preserved. The problem is that direct triggers alone do not constitute a grounded explanation. Emo-context, an utterance that helps interpret a target emotion without directly causing it, appears on both sides of the original boundary and is enriched near binary uncertainty, showing that the binary boundary has no stable place for such discourse evidence. Across evaluated ECPE models, direct triggers are recovered more reliably than contextual support. Under shortcut pressure, this imbalance becomes consequential. Binary-trained models assign higher pair scores to nearby lexically similar non-pair candidates than to evidence supported but structurally harder emo-cause and emo-context pairs. Thus, pair scores can reward convenient attributions over grounded explanations. High binary ECPE performance indicates that a model can identify direct triggers; it does not indicate that the model has explained the emotion. Code is publicly available at https://github.com/panzhzh/ECPExsame.", "date": "2026-05-24", "categories": []}
{"id": "2605.25409", "title": "MTLLFM: Multimodal-Temporal Laughter Localization: UR-FUNNY-Temporal and SMILE-Temporal Benchmarks with an Adaptive Multimodal Fusion Model", "abstract": "Detecting laughter in video is essential for affective computing and narrative understanding, yet existing approaches treat it as coarse clip-level classification, failing to capture precise temporal boundaries of brief, transient laughter events. We address this gap with two complementary contributions.\n  First, we introduce UR-FUNNY-Temporal and SMILE-Temporal, fully annotated temporal laughter datasets extending two widely-used humor benchmarks. Our annotations cover over 11,053 videos (78.8 hours) and provide precise onset/offset boundaries for each laughter event, along with rich metadata distinguishing speaker vs. audience laughter, modality dominance (acoustic, visual, or both), and intensity levels.\n  Second, we propose a lightweight weakly-supervised framework for temporal laughter localization. Our architecture combines fixed HuBERT and MAE encoders with temporal softmax pooling and adaptive modality gating, learning fine-grained temporal grounding from clip-level labels without requiring frame-level annotations during training. Experiments across three datasets demonstrate that our approach substantially outperforms multimodal foundation models including Gemini 3 Flash, achieving 99% F1 and 68.1% localization precision on sports broadcast data. Ablations validate each architectural component. Furthermore, our precise temporal tags improve downstream laughter reasoning by 227% on CIDEr, enabling GPT-3.5 to outperform GPT-4o. The code, UR-FUNNY-Temporal and SMILE-Temporal datasets are publicly available at https://github.com/WSCSports/MTLLFM-temporal-laughter-localization.", "date": "2026-05-25", "categories": []}
{"id": "2507.23372", "title": "UniEmo: Unifying Emotional Understanding and Generation with Learnable Expert Queries", "abstract": "Emotional understanding and generation are often treated as separate tasks, yet they are inherently complementary and can mutually enhance each other. In this paper, we propose the UniEmo, a unified framework that seamlessly integrates these two tasks. The key challenge lies in the abstract nature of emotions, necessitating the extraction of visual representations beneficial for both tasks. To address this, we propose a hierarchical emotional understanding chain with learnable expert queries that progressively extracts multi-scale emotional features, thereby serving as a foundational step for unification. Simultaneously, we fuse these expert queries and emotional representations to guide the diffusion model in generating emotion-evoking images. To enhance the diversity and fidelity of the generated emotional images, we further introduce the emotional correlation coefficient and emotional condition loss into the fusion process. This step facilitates fusion and alignment for emotional generation guided by the understanding. In turn, we demonstrate that joint training allows the generation component to provide implicit feedback to the understanding part. Furthermore, we propose a novel data filtering algorithm to select high-quality and diverse emotional images generated by the well-trained model, which explicitly feedback into the understanding part. Together, these generation-driven dual feedback processes enhance the model's understanding capacity. Extensive experiments show that UniEmo significantly outperforms state-of-the-art methods in both emotional understanding and generation tasks. The code for the proposed method is available at https://github.com/JiuTian-VL/UniEmo.", "date": "2025-07-31", "categories": []}
{"id": "2605.23328", "title": "Emotion Recognition in Sign Language Conversation", "abstract": "Emotion Recognition in Conversation is a core component of affective computing, while current resources of sign language emotion datasets primarily focus on isolated sentences and lack conversational context. Models trained exclusively on these isolated utterances demonstrate degraded performance in real world scenarios because they cannot utilize historical dialogue flow. To address this structural limitation, we introduce the ERC task to sign language video analysis and propose the eJSL Dialog dataset. Constructed using the scripts from the STUDIES corpus, the dataset contains 1,920 video samples organized into 480 unique dialogues. We conduct systematic benchmarking on this dataset using models ranging from isolated visual networks to multimodal conversational architectures. The results reveal a domain gap when applying generic multimodal conversational emotion recognition models to sign language. These findings demonstrate the explicit need for context aware visual extractors specific to sign language and indicate that expanding the scale of conversational datasets to support large scale pre-training is a necessary next step for future research.", "date": "2026-05-22", "categories": []}
{"id": "2502.09487", "title": "Internal narratives parameterise affective states", "abstract": "Characterising how we verbalise our feelings is central to psychological assessment and intervention, yet the mapping between narrative and affective state remains poorly understood. Across two large studies (n=1257), we parameterised the structure and dynamics of depressive states by quantifying participants' internal narratives through large-language-model representations and their subspaces. In Study 1, we found verbal descriptions of symptom-specific thoughts captured granular information predictive of standardised, self-reported depression scores. Critically, we show preserving the specific covariance between symptoms is essential for construct validity, suggesting high-dimensional text representations mirror the latent geometry of the disorder. Study 2 probed the temporal dynamics of this relationship as participants engaged with emotional narratives. We found quantified changes in internal narratives led to changes in self-report, while the baseline narrative severity predicted the magnitude of subsequent affective change. By framing affect as a computational state, our results highlight its core, therapeutically pertinent functions: constraining the structure of internal narratives and integrating context to shape self-report.", "date": "2025-02-13", "categories": []}
{"id": "2605.21565", "title": "Leveraging Self-Paced Curriculum Learning for Enhanced Modality Balance in Multimodal Conversational Emotion Recognition", "abstract": "Multimodal Emotion Recognition in Conversations (MERC) is a crucial task for understanding human interactions, where multimodal approaches integrating language, facial expressions, and vocal tone have achieved significant progress. However, modality misalignment and imbalanced learning remain major challenges, limiting the effective utilization of multimodal information. To address this issue, we propose a plug-and-play framework based on Self-Paced Curriculum Learning (SPCL) for MERC. We introduce a dual-level Difficulty Measurer that captures both utterance-level and conversation-level challenges. The utterance-level score models fine-grained modality-specific difficulty, while the conversation-level score captures broader dialogue structures, including emotional dependencies and modality coherence. Based on these scores, the Learning Scheduler dynamically guides training from easier to more difficult instances. By integrating SPCL into existing MERC architectures, our method alleviates modality imbalance and improves model robustness. Extensive experiments on the IEMOCAP and MELD datasets demonstrate consistent improvements across different architectures and modality settings. On IEMOCAP, SPCL improves weighted F1-score by approximately +1.2% to +6.6% over baseline models, while on MELD, gains reach up to +10.4%. These results highlight the effectiveness and generalizability of SPCL as a lightweight plug-and-play module for multimodal emotion recognition.", "date": "2026-05-20", "categories": []}
{"id": "2605.21869", "title": "Two-Stage Multimodal Framework for Emotion Mimicry Intensity Prediction", "abstract": "We present our submission to the Hume-ABAW10 Emotional Mimicry Intensity (EMI) Challenge, which aims to predict six continuous emotion intensity dimensions: Admiration, Amusement, Determination, Empathic Pain, Excitement, and Joy, from in-the-wild multimodal video clips. We propose a staged multimodal framework that combines textual, acoustic, and visual representations, with an optional motion branch. Our approach first trains modality-specific encoders independently and then fuses their learned representations through a lightweight regressor with modality dropout and controlled encoder adaptation. Across our submitted systems, the best validation performance is obtained by the text--audio--vision--motion fusion model under the expanded 4:1 split, achieving an average Pearson correlation of 0.4722. Although the motion branch yields only very slight gains, its behavior can be interesting to study. Our team was placed third in the EMI challenge, achieving an average Pearson correlation of 0.57 for the test set. Overall, we provide a practical and reproducible baseline for EMI prediction.", "date": "2026-05-21", "categories": []}
{"id": "2605.22732", "title": "Beyond Acoustic Emotion Recognition: Multimodal Pathos Analysis in Political Speech Using LLM-Based and Acoustic Emotion Models", "abstract": "We investigate whether acoustic emotion recognition models can serve as proxies for the Pathos dimension in political speech analysis, as operationalised by the TRUST multi-agent large language model (LLM) pipeline. Using a Bundestag plenary speech by Felix Banaszak (51 segments, 245 s) as a case study, we compare three analysis modalities: (1) emotion2vec_plus_large, an acoustic speech emotion recognition (SER) model whose continuous Arousal and Valence values are derived via post-hoc Russell Circumplex projection; (2) Gemini 2.5 Flash, an LLM analysing the full speech audio together with its transcript in an open-ended, context-aware fashion; and (3) TRUST-Pathos scores from a three-advocate LLM supervisor ensemble. Spearman rank correlations reveal that Gemini Valence correlates strongly with TRUST-Pathos (rho = +0.664, p < 0.001), whereas emotion2vec Valence does not (rho = +0.097, p = 0.499). We further demonstrate, via a systematic quality evaluation of the Berlin Database of Emotional Speech (EMO-DB) using Gemini in an open-ended annotation paradigm, that standard SER benchmark corpora suffer from acted speech, cultural bias, and category incompatibility. Our results suggest that LLM-based multimodal analysis captures semantically defined political emotion substantially better than acoustic models alone, while acoustic features remain informative for low-level Arousal estimation. Future work will extend this approach to video-based analysis incorporating facial expression and gaze.", "date": "2026-05-21", "categories": []}
{"id": "2508.04999", "title": "Disentangling Bias by Modeling Intra- and Inter-modal Causal Attention for Multimodal Sentiment Analysis", "abstract": "Multimodal sentiment analysis (MSA) aims to understand human emotions by integrating information from multiple modalities, such as text, audio, and visual data. However, existing methods often suffer from spurious correlations both within and across modalities, leading models to rely on statistical shortcuts rather than true causal relationships, thereby undermining generalization. To mitigate this issue, we propose a Multi-relational Multimodal Causal Intervention (MMCI) framework, which leverages the backdoor adjustment from causal theory to address the confounding effects of such shortcuts. Specifically, we first model the multimodal inputs as a multi-relational graph to explicitly capture intra- and inter-modal dependencies. Then, we apply an attention mechanism to separately estimate and disentangle the causal features and shortcut features corresponding to these intra- and inter-modal relations. Finally, by applying the backdoor adjustment, we stratify the shortcut features and dynamically combine them with the causal features to encourage MMCI to produce stable predictions under distribution shifts. Extensive experiments on several standard MSA datasets and out-of-distribution (OOD) test sets demonstrate that our method effectively suppresses biases and improves performance.", "date": "2025-08-07", "categories": []}
{"id": "2605.20200", "title": "Evaluating multimodal emotion recognition in proactive conversational agents: A user study", "abstract": "This article presents a multimodal emotion recognition module integrated into a proactive Socially Interactive Agent (SIA) powered by generative artificial intelligence. The system evaluates real-time affective states through two distinct channels: a computer vision-based facial recognition module and a semantic linguistic analysis engine. To validate the framework, an empirical study was conducted with 20 users who engaged in dynamic, unscripted dialogues with the conversational agent. The findings reveal a significant discrepancy between automated visual cues and actual internal emotional states. When interacting with the AI, users consistently exhibited a \"poker face\" effect, displaying serious, concentrated facial expressions even when experiencing positive emotions. Consequently, the generative AI linguistic analysis proved significantly more reliable, by contextualizing the users' verbal expressions. Furthermore, an analysis of the interaction dynamics demonstrated that SIAs can effectively elicit specific emotions by adapting conversational themes and employing structured linguistic patterns, such as empathetic or humorous language. However, the study also noted that instances of uncalibrated proactivity occasionally led to user disengagement and a perception of artificiality. Ultimately, this research highlights the necessity of refining SIAs to dynamically adapt to users' emotional evolution, relying on deep linguistic context to foster more natural, human-like interactions.", "date": "2026-04-06", "categories": []}
{"id": "2605.20442", "title": "Modeling Emotional Dynamics in Agent-to-Agent Interactions on Moltbook", "abstract": "Generative AI systems are increasingly deployed as interactive agents in online environments, such as a social network called Moltbook. In Moltbook, large-scale agentic AIs can post, comment, and engage in activities generated at scale by AI-driven text. Yet these agent behavioral characteristics remain insufficiently understood, particularly in complex, multi-agent interaction. In this study, we analyze the emotional dynamics of agent interactions within Moltbook. We construct an emotion-aware framework that maps textual interactions to a predefined set of fine-grained emotional categories, enabling the extraction of structured emotion profiles across agents and interaction contexts. To further evaluate behavioral reliability, we introduce an emotion-based domain called Persona-Stimulus-Reaction (PSR) that captures the alignment of emotional responses across similar contexts. Our analysis shows distinct emotional patterns and varying levels of behavioral stability across agents. Our analysis reveals that agents exhibit distinct emotional signatures with varying levels of behavioral stability influenced by interaction context.", "date": "2026-05-19", "categories": []}
{"id": "2601.03645", "title": "LLM-MC-Affect: LLM-Based Monte Carlo Modeling of Affective Trajectories and Latent Ambiguity for Interpersonal Dynamic Insight", "abstract": "Emotional coordination is a core property of human interaction that shapes how relational meaning is constructed in real time. While text-based affect inference has become increasingly feasible, prior approaches often treat sentiment as a deterministic point estimate for individual speakers, failing to capture the inherent subjectivity, latent ambiguity, and sequential coupling found in mutual exchanges. We introduce LLM-MC-Affect, a probabilistic framework that characterizes emotion not as a static label, but as a continuous latent probability distribution defined over an affective space. By leveraging stochastic LLM decoding and Monte Carlo estimation, the methodology approximates these distributions to derive high-fidelity sentiment trajectories that explicitly quantify both central affective tendencies and perceptual ambiguity. These trajectories enable a structured analysis of interpersonal coupling through sequential cross-correlation and slope-based indicators, identifying leading or lagging influences between interlocutors. To validate the interpretive capacity of this approach, we utilize teacher-student instructional dialogues as a representative case study, where our quantitative indicators successfully distill high-level interaction insights such as effective scaffolding. This work establishes a scalable and deployable pathway for understanding interpersonal dynamics, offering a generalizable solution that extends beyond education to broader social and behavioral research.", "date": "2026-01-07", "categories": []}
{"id": "2605.18884", "title": "Navigating the Emotion Tree: Hierarchical Hyperbolic RAG for Multimodal Emotion Recognition", "abstract": "Multimodal emotion recognition aims to integrate text, audio, and video sources to understand human affective states. Although multimodal large language models excel at multimodal reasoning, they typically treat emotion categories as independent labels, ignoring the rich hierarchical taxonomy of human psychology. Moreover, lacking external contextual knowledge makes them highly susceptible to over-interpreting noisy cues, further complicating fine-grained emotion classification. To address these issues, we propose \\textbf{HyperEmo-RAG}, a retrieval-augmented generation framework that leverages a structured emotional knowledge base. Our framework introduces two key innovations. 1) Hierarchical hyperbolic grounding. Recognizing the inherent hierarchical tree structure of emotion taxonomies, we jointly embed hierarchical emotion labels and multimodal samples into a continuous hyperbolic space (Poincar\\'e ball) and design a hierarchical beam-search deliberation process that progressively retrieves samples from coarse to fine-grained levels. 2) Structured evidence injection. Based on the retrieved evidence, we construct an evidence graph and inject the structured knowledge as explicit cognitive context into the LLM through a Tree-Aware Attention mechanism and an EmotionGraphFormer, preserving the integrity of graph-structured information. Experiments on multiple datasets demonstrate that HyperEmo-RAG significantly outperforms existing methods.", "date": "2026-05-16", "categories": []}
{"id": "2605.19765", "title": "GroupAffect-4: A Multimodal Dataset of Four-Person Collaborative Interaction", "abstract": "Existing affective-computing, social-signal-processing, and meeting corpora capture important parts of human interaction, but they rarely support analysis of affect in co-located groups as a coupled individual, interpersonal, and group-level process. The required signals (per-participant physiology, eye movement, audio, self-report, task outcomes, and personality) are usually fragmented across separate dataset traditions. We introduce GroupAffect-4, a multimodal corpus of 40 participants in 10 four-person groups, each completing four ecologically varied collaborative tasks spanning information pooling, negotiation, idea generation, and a public-goods game. Each participant is instrumented with a wrist-worn physiology sensor, eye-tracking glasses, and a close-talk microphone; sessions include continuous affect self-reports, post-task questionnaires, task outcomes, and Big-Five personality scores, all time-aligned to a shared clock. The dataset covers over 91% of expected physiology windows and 98% of eye-tracking windows, with strong task validity confirmed by a clear affective manipulation check across the negotiation block. We define fifteen benchmarkable targets spanning three analysis levels -- within-person state, between-person traits, and group dynamics -- and report leave-one-group-out feasibility baselines establishing the dataset's evaluative scope. GroupAffect-4 is released with a BIDS-inspired structure, Croissant metadata, a datasheet, per-session quality reports, and open processing scripts. Code and processing scripts are available at https://github.com/meisamjam/GroupAffect-4; the dataset is publicly archived at https://zenodo.org/records/20037847.", "date": "2026-05-19", "categories": []}
{"id": "2605.19821", "title": "LaCoVL-FER: Landmark-Guided Contrastive Learning Network with Vision-Language Enhancement for Facial Expression Recognition", "abstract": "Facial Expression Recognition (FER) in the wild is still challenging due to uncontrolled variations in pose, occlusion, and illumination. Most existing attention-based methods primarily rely on visual appearance cues, suffering from attention redundancy and instability, which limits their performance in complex scenarios. To address these issues, we propose a novel landmark-guided contrastive learning network with vision-language enhancement for FER (LaCoVL-FER), which integrates geometric priors from facial landmarks and semantic priors from a vision-language model. Specifically, a Landmark-Guided Adaptive Encoder (LGAE) is designed to introduce geometric priors through a Bi-branch Gated Cross Attention (BGCA) mechanism, which achieves adaptive fusion of landmark-based geometric and visual appearance features to produce expression-relevant features, thereby focusing on key facial regions and suppressing noise interference. In parallel, a Vision-Language Enhancement Strategy (VLES) is presented to leverage the expression-relevant features to refine the generalizable visual features extracted by the frozen pretrained CLIP image encoder, yielding expression-specific visual representations. Based on these representations, an Expression-Conditioned Prompting (ECP) mechanism is utilized to further adapt the textual features of fixed class-level prompts from the frozen pretrained CLIP text encoder, generating more instance-aware textual representations. These visual-textual representations are aligned as semantic priors to enhance the robustness and generalization of FER. Quantitative and qualitative experiments demonstrate that our LaCoVL-FER outperforms state-of-the-art methods on three representative real-world FER datasets, including RAF-DB, FERPlus, and AffectNet. The code is available at https://github.com/ylin06804/LaCoVL-FER.", "date": "2026-05-19", "categories": []}
{"id": "2605.19950", "title": "AffectVerse: Emotional World Models for Multimodal Affective Computing", "abstract": "Humans infer emotions by integrating observed multimodal cues with expectations about how affective states may unfold. Existing multimodal large language models (MLLMs), however, often treat emotion recognition as static fusion over complete audiovisual-text inputs, leaving affective dynamics implicit. We propose AffectVerse, a Qwen2.5-Omni-based model equipped with an Emotion World Module (EWM), an action-free representation-level module for short-horizon latent affective prediction. \\rev{EWM contains three modules: 1) Cross-Modal Temporal Imagination predicts future video/audio representations from past tokens with multi-step rollout. 2) MAMA(Modality-Aware Multi-step Attention) Belief Aggregation compresses imagined tokens into modality-aware belief tokens. 3) Belief Injection inserts these belief tokens into the LLM for affective reasoning.} AffectVerse uses future prediction as a past-conditioned self-supervised signal: it does not replace modeling observed history or require unseen signals at inference, but forces the current belief state to encode transition cues that are predictive of subsequent affective change. Across nine benchmarks, AffectVerse improves at least 2.57\\% over other models, while controlled ablations show additive gains from temporal imagination, cross-modal rollout, and belief aggregation. These results suggest predictive belief-state modeling is a practical alternative for affective computing.", "date": "2026-05-19", "categories": []}
{"id": "2510.15221", "title": "WELD: The First Naturalistic Long-Period Small-Team Workplace Emotion Dataset for Ubiquitous Affective Computing", "abstract": "Affective computing has matured rapidly in laboratory settings, yet no prior dataset combines (i) months-to-years of duration, (ii) a naturalistic workplace context, (iii) a stable small-team social structure, and (iv) a fully passive sensing protocol that survives institutional review. We introduce WELD, the first dataset to satisfy all four. WELD comprises 733,780 per-frame seven-class facial-expression probability vectors from 49 employees of a Chinese software company over 30.1 months (Nov 2021 - May 2024) -- the longest naturalistic in-the-wild emotion corpus and the only multi-year corpus supporting both within-individual longitudinal and within-team relational analyses on the same subjects. Data are released under a four-tier access model with only aggregated probabilities publicly downloadable. We validate the corpus by replicating three established phenomena (+43.1% weekend valence boost; 13:00-trough diurnal cycle; Shanghai 2022 lockdown effect d=-0.40), and report four novel findings: (1) variance decomposition attributes 19.3% of daily-valence variance to between-person differences and 29.8% to month seasonality -- a quantitative ceiling for future predictive models; (2) Hidden Markov decomposition reveals six emotional regimes with asymmetric negative-state dwell times (16-18 d vs 3 d); (3) leave-one-person-out turnover prediction reaches AUC=0.79 yet a Cox concordance index of only 0.52, exposing a metric-trap when AUC is reported without survival-aware baselines; (4) the corpus reveals systematic over-prediction of \"angry\" by an off-the-shelf FER model on neutral Asian faces (0.194 vs ~0.05 Western priors), making WELD valuable for FER fairness audits. A complex-systems analysis of the corpus appears as a companion preprint (arXiv:2510.16046).", "date": "2025-10-17", "categories": []}
{"id": "2605.01235", "title": "MindMelody: A Closed-Loop EEG-Driven System for Personalized Music Intervention", "abstract": "Driven by the escalating global burden of mental health conditions, music-based interventions have attracted significant attention as a non-invasive, cost-effective modality for emotion regulation and psychological stress relief. However, current digital music services rely on static preferences and fail to adapt to users' instantaneous psychological states. Furthermore, directly mapping electroencephalography (EEG) to music generation remains challenging due to severe paired-data scarcity and a lack of interpretability. To address these limitations, we propose MindMelody, a fully functional, closed-loop real-time system for EEG-driven personalized music intervention. MindMelody introduces an emotion-mediated semantic bridge. Specifically, a hybrid Transformer-GNN first decodes real-time EEG signals into global Valence-Arousal states and local temporal affect trajectories. These states are then fed into a Retrieval-Augmented Generation (RAG)-equipped Large Language Model (LLM) to formulate structured intervention plans. Subsequently, a novel Hierarchical EEG Controller injects global affect prefixes and local temporal guidance into a pretrained music backbone, enabling fine-grained controllable audio synthesis. Crucially, the system incorporates a continuous feedback loop that updates generation parameters on the fly based on the user's evolving EEG dynamics. Extensive experiments show that MindMelody improves control adherence and emotional alignment, and receives higher perceived helpfulness in a short-term listening setting, suggesting its promise as an adaptive affect-aware music generation framework.", "date": "2026-05-02", "categories": []}
{"id": "2605.16889", "title": "Controlling Decision Drift in Multimodal Sentiment Analysis with Missing Modalities", "abstract": "Multimodal sentiment analysis relies on textual, acoustic, and visual signals, yet real-world data often suffer from modality missing and quality imbalance. Existing methods generate features for modality missing from available ones, but differences in expression mechanisms and sentiment dynamics across modalities may cause the generated features to deviate from true distributions and mislead prediction. In addition, unreliable modalities may dominate fusion, resulting in representation shift across modality combinations and unstable sentiment representations. To address these challenges, we propose a two-level reference alignment framework. The framework introduces stable references at the feature representation and sentiment decision levels to improve robustness under modality missing. First-level reference alignment leverages complete-modality samples to constrain representations and align different modality combinations into a shared sentiment space. Second-level reference alignment enforces cross-modal consistency at the decision level by suppressing unreliable modalities through prototype retrieval and voting. As a result, the framework maintains stable and reliable sentiment predictions under diverse missing-modality patterns. Experiments on CMU-MOSI and CMU-MOSEI show consistent improvements across various missing-modality settings. Under full-modality input, the proposed method achieves state-of-the-art performance, with ACC of 86.28% and 85.88%, and F1 of 86.24% and 85.86%.", "date": "2026-05-16", "categories": []}
{"id": "2605.17179", "title": "iMiGUE-3K: A Large-Scale Benchmark for Micro-Gesture Analysis with Self-Supervised Learning", "abstract": "Emotion understanding is a fundamental challenge in affective computing and artificial intelligence. While existing approaches predominantly focus on facial expressions and speech, they often overlook the rich emotional cues conveyed through body language. Recently, micro-gestures (MGs), unintentional, subconscious movements driven by inner feelings, have attracted increasing attention as an alternative to other cues. However, there are no existing large-scale datasets supporting the pre-training of the MG foundation model. To advance MG research, we present a new benchmark for micro-gesture-based emotion understanding, featuring key contributions with a novel dataset (iMiGUE-3K) and a series of foundation models for different tasks. Using a model-based crowd-sourcing data collection strategy, we construct iMiGUE-3K, the largest MG dataset to date. It comprises video recordings from 332 distinct professional tennis players' public press interviews over the past seven years, totaling more than 3.4K long video clips and 37 million frames. The dataset includes 32 micro-gesture classes with rich descriptive annotations, making it the first large-scale, in-the-wild, video dataset for fine-grained gesture-based emotion analysis. Built on iMiGUE-3K, we propose MG-FMs, a discriminative foundation model for transferable gesture presentation learning. Based on the foundation model, we establish five comprehensive evaluation tasks: MG recognition (unsupervised, semi-supervised, supervised), MG retrieval, and MG emotion recognition. Our systematic evaluation of representative methods demonstrates that micro-gesture-based analysis significantly improves emotion understanding. We hope this work can provide comprehensive tools for MG analysis and set a solid foundation for future research in psychological diagnostics, affective computing, and advanced human-computer interaction.", "date": "2026-05-16", "categories": []}
{"id": "2605.17684", "title": "EGI: A Multimodal Emotional AI Framework for Enhancing Scrum Master Real-time Self-Awareness", "abstract": "While increasing research focuses on the emotional well-being of agile team members, a significant gap remains in emotion monitoring studies for Scrum Masters and meeting organizers, whose impact on team dynamics is crucial. This paper proposes a novel application integrating four carefully selected and recommended AI models to monitor the unconsciously expressed emotions of these key roles. This is achieved through: real- time transcription using a speech-to-text model; thresholding for intonation analysis to detect emotional cues in prosody; applying emotion-based vocabulary matching to identify sentiment in spoken content; and providing context-aware suggestions containing emotion keywords using an open-source, multi-module AI API. The system achieved an ASR word error rate WER of 10% in simulated meeting environments. Our evaluation shows that real- time feedback significantly improves emotion awareness during simulated agile meetings, providing Scrum Masters and meeting organizers with real-time and practical suggestions to help them quickly identify and minimize the expression of negative emotions, fostering more positive and effective team interactions.", "date": "2026-05-17", "categories": []}
{"id": "2605.18547", "title": "VISAFF: Speaker-Centered Visual Affective Feature Learning for Emotion Recognition in Conversation", "abstract": "Emotion Recognition in Conversation (ERC) is essential for effective human-machine interaction, aiming to identify speakers' emotional states in multi-turn dialogues. Early text-based methods struggle with complex scenarios like sarcasm because they inherently neglect vital non-verbal information. While recent Vision-Language Models (VLMs) address this by analyzing video directly, they are not inherently tailored for ERC and often focus on emotionally irrelevant background regions or passive listeners rather than the active speaker. Furthermore, fine-tuning these large models incurs prohibitive computational costs. Additionally, isolated visual signals are frequently ambiguous or technically compromised without the context of linguistic content and vocal prosody. To address these challenges, we propose VISAFF, a speaker-centered VISual AFFective feature learning framework for ERC. VISAFF consists of two stages: Speaker-Centered Affective Grounding and Reliability-Guided Affective Complementation. VISAFF utilizes a tuning-free approach to unlock the reasoning capabilities of frozen VLMs, efficiently steering them to focus on the active speaker's emotional visual cues without heavy training overheads. In the second stage, we introduce a reliability-guided affective complementation mechanism that dynamically leverages textual and acoustic modalities to compensate for visual uncertainty. Experiments on two real-world datasets demonstrate that VISAFF achieves highly competitive performance compared to state-of-the-art methods in a tuning-free setting, significantly enhancing computational efficiency by eliminating the need for expensive fine-tuning of large VLMs. The source code is available at https://anonymous.4open.science/r/speaker-2365/.", "date": "2026-05-18", "categories": []}
{"id": "2605.15755", "title": "Attribute-Grounded Selective Reasoning for Artwork Emotion Understanding with Multimodal Large Language Models", "abstract": "Multimodal large language models (MLLMs) can produce fluent artwork emotion explanations, but they often suffer from attribute flooding: they enumerate many visible formal attributes without identifying which cues actually support the affective judgment. We therefore formulate artwork emotion understanding as Attribute-Grounded Selective Reasoning (AGSR), where predefined formal attributes serve as evidence units and only emotionally operative attributes should enter the final interpretation. To make this problem measurable, we extend EmoArt, originally introduced at ACM MM 2025 as a 132,664-artwork resource with content, formal-attribute, valence-arousal, and emotion annotations, by adding a 1,400-artwork human salience extension annotated by 15 art-trained annotators. This extension provides instance-level supervision for distinguishing attributes that are merely present from those that are emotionally salient. We further propose FAB-G (Formal-Attribute Bottleneck-Guided reasoning), a supervised multi-agent framework that first predicts attribute-level salience and then constrains downstream emotional analysis to the retained cues. Experiments show that FAB-G yields consistent gains in emotion, arousal, and valence prediction, achieves stronger agreement with human-marked salient attributes under Dice and Tversky metrics, and produces substantially more compact final explanations than prompting-based baselines. Cross-dataset evaluation further suggests that attribute-grounded salience selection transfers beyond the source distribution of EmoArt, while also revealing attribute-specific boundary cases. The dataset and project page are available at https://zhiliangzhang.github.io/EmoArt-130k/", "date": "2026-05-15", "categories": []}
{"id": "2605.15915", "title": "SLIP & ETHICS: Graduated Intervention for AI Emotional Companions", "abstract": "AI emotional companions face a safety-rapport paradox: restrictive safeguards can damage supportive alliance, while permissive systems risk user harm. We present SLIP (Staged Layers of Intervention Protocol), a four-stage graduated methodology deriving interventions (none, soft, hard) from structured qualitative indicators -- affect intensity (a) and narrative dynamism (m) -- alongside ETHICS (Emergent Taxonomy for Human-AI Interaction Context Signals), a \"signals not labels\" taxonomy. An evaluation combining a small-scale production deployment (N=68 entries, 10 users, 10 weeks) with a synthetic persona battery (N=91, 5 behavioral-risk profiles) achieved 0% false positives for the flow persona and showed expected escalation patterns in crisis-oriented personas. However, initial results showed that 8 consecutive days of high-energy elevation produced zero interventions (0/8), exposing a boundary where the \"do not pathologize\" principle conflicts with safety. A subsequent three-model stress test demonstrated that increased model capability improves detection from 0/8 to 6/8 while preserving 0/10 flow false positives in the largest model. Read as preliminary, these findings position graduated intervention as a design direction for navigating -- not resolving -- the safety-rapport tension in affective computing.", "date": "2026-05-15", "categories": []}
{"id": "2605.13858", "title": "A Hormone-inspired Emotion Layer for Transformer language models (HELT)", "abstract": "Large Language Models have demonstrated remarkable capabilities in generating contextually relevant and grammatically correct text. However, they fundamentally lack the ability to process and respond to emotional context in a manner analogous to human emotional cognition. Current approaches to emotion modeling in NLP systems rely primarily on discrete emotion classification or simplistic sentiment analysis, which fail to capture the continuous, multi-dimensional nature of human emotional states. In this paper, we introduce HormoneT5, a novel architecture that augments transformer language models with a biologically-inspired Hormone Emotion Block that simulates the human endocrine system's role in emotional processing. Our approach computes six continuous hormone-like values through specialized per-hormone attention heads, each with orthogonally initialized learnable queries, temperature-scaled attention mechanisms, and deep output projections. These hormone values are then transformed into an emotional embedding that modulates the encoder hidden states, enabling emotionally-appropriate response generation. We propose a multi-objective training framework combining sequence-to-sequence loss, hormone prediction loss with margin penalties, and diversity regularization to prevent attention collapse. Experimental results on our curated emotion-labeled dataset demonstrate that HormoneT5 achieves 85%+ per-hormone accuracy within a 0.15 tolerance threshold, with hormone differentiation ranges exceeding 0.85 across all six hormones between contrasting emotional tones. Human evaluation studies show significant preference (p < 0.01) for HormoneT5-generated responses in terms of emotional appropriateness and empathetic quality compared to baseline T5 outputs. Our work opens new directions for biologically-grounded affective computing and emotionally intelligent conversational agents.", "date": "2026-04-13", "categories": []}
{"id": "2605.14360", "title": "A Formative Study of Brief Affective Text as a Complement to Wearable Sensing for Longitudinal Student Health Monitoring", "abstract": "Wearable devices capture physiological and behavioral data with increasing fidelity, but the psychological context shaping these outcomes is difficult to recover from sensor data alone, limiting passive sensing utility for digital health. We examined whether ultra-brief naturalistic concern text could serve as a scalable complement to passive sensing. In a year-long study of 458 university students (3,610 person-waves) tracked with Oura rings, participants responded bimonthly to an open-ended prompt about what concerned them most; responses had a median length of three words. We compared dictionary-based, general pretrained, and domain-adapted NLP approaches using within-person mixed-effects models across nine sleep and physical activity outcomes. Weeks dominated by academic concern framing were associated with lower physical activity; weeks characterized by emotional exhaustion language were associated with poorer sleep quality and lower heart rate variability. General pretrained embeddings outperformed domain-adapted models for most outcomes, with domain adaptation showing relative advantage for autonomic outcomes. Zero-shot classification of concern topics produced no significant associations, while affective dimensions across all three methods were consistently associated with outcomes, indicating emotional register rather than topical content carries the signal. These findings offer design guidance: ultra-brief affective prompts enrich the psychological interpretability of passive physiological data at minimal burden.", "date": "2026-05-14", "categories": []}
{"id": "2605.14635", "title": "MultiEmo-Bench: Multi-label Visual Emotion Analysis for Multi-modal Large Language Models", "abstract": "This paper introduces a multi-label visual emotion analysis benchmark dataset for comprehensively evaluating the ability of multimodal large language models (MLLMs) to predict the emotions evoked by images. Recent user studies report an unintuitive finding: humans may prefer the predictions of MLLMs over the labels in existing datasets. We argue that this phenomenon stems from the suboptimal annotation scheme used in existing datasets, where each annotator is shown a single candidate emotion for each image and judges whether it is evoked or not. This approach is clearly limited because a single image can evoke multiple emotions with varying intensities. As a result, evaluations based on these datasets may underestimate the capabilities of MLLMs, yet an appropriate benchmark for evaluating such models remains lacking. To address this issue, we introduce a new multi-label benchmark dataset for visual emotion analysis toward MLLMs evaluation. We hire $20$ annotators per image and ask them to select all emotions they feel from an image. Then, we aggregate the votes across all annotators, providing a more reliable and representative dataset labeled with a distribution of emotions. The resulting dataset contains $10,344$ images with $236,998$ valid votes across eight emotions. Based on this benchmark dataset, we evaluate several recent models, including Qwen3-VL, OpenAI's GPT, Gemini, and Claude. We assess model performance on both dominant emotion prediction and emotion distribution prediction. Our results demonstrate the progress achieved by recent MLLMs while also indicating that substantial room for improvement remains. Furthermore, our experiments with LLM-as-a-judge show that the method does not consistently improve MLLMs' performance, indicating its limitations for the subjective task of visual emotion analysis.", "date": "2026-05-14", "categories": []}
{"id": "2605.02521", "title": "MooD: Perception-Enhanced Efficient Affective Image Editing via Continuous Valence-Arousal Modeling", "abstract": "Affective Image Editing (AIE) aims to modify visual content to evoke targeted emotions. Although current approaches achieve impressive editing quality, they often overlook inference efficiency, which limits their applicability in computational social scenarios. Moreover, most methods depend on discrete emotion representations, which hinder the continuous modeling of complex human emotions and constrain expressive capabilities in interactive scenarios. To tackle these gaps, we propose MooD, the first framework that directly leverages continuous Valence-Arousal (VA) values as editing instruction for fine-grained and efficient AIE in computational social systems. Specifically, we first introduce a VA-Aware retrieval strategy to bridge vague affective values and detailed visual semantics. Building upon this, MooD integrates visual transfer and perception-enhanced semantic guidance to achieve controllable AIE. Furthermore, considering that existing VA-annotated datasets mainly focus on social scenarios and largely overlook natural scenes, we therefore construct AffectSet, a comprehensive VA-annotated dataset covering diverse scenarios, to support model optimization and evaluation. Extensive qualitative and quantitative experimental results demonstrate that our MooD achieves superior performance in both affective controllability and visual fidelity while maintaining high efficiency. A series of ablation studies further reveal the crucial factors of our design.", "date": "2026-05-04", "categories": []}
{"id": "2605.12525", "title": "PERCEIVE: A Benchmark for Personalized Emotion and Communication Behavior Understanding on Social Media", "abstract": "Current emotion analysis in social media is predominantly author-centric, failing to capture the subjective nature of emotional responses across diverse readers. This paradigm overlooks the crucial link between individual perception, communication behavior, and the underlying social network. To bridge this gap, we introduce PERCEIVE, a novel bilingual (English and Chinese) large-scale benchmark that, to the best of our knowledge, is the first to integrate five critical dimensions for social perception: author-created content, genuine readers' emotional feedback (derived from their comments), communication behavior, user attributes, and the social graph. This benchmark enables a paradigm shift towards truly personalized, reader-centric analysis, where different readers' emotional responses to the same content are naturally captured through their real-world interactions. By annotating emotions from reader comments and synchronously capturing communication intent, PERCEIVE provides a unique resource to model the intrinsic coupling between emotion and behavior, grounded in social context. We establish a comprehensive evaluation protocol, testing state-of-the-art methods, including large language models (LLMs) with advanced reasoning enhancement. Our findings reveal significant shortcomings in existing approaches when handling this multifaceted, user-aware task. PERCEIVE offers a foundational resource and clear direction for future research in socially-intelligent NLP, pushing models towards a more unified understanding of emotion on social media.", "date": "2026-04-10", "categories": []}
{"id": "2605.12838", "title": "Multimodal Hidden Markov Models for Persistent Emotional State Tracking", "abstract": "Tracking an interpretable emotional arc of a conversation via the sentiment of individual utterances processed as a whole is central to both understanding and guiding communication in applied, especially clinical, conversational contexts. Existing approaches to emotion recognition operate at the utterance level, obscuring the persistent phases that characterize real conversational dynamics. We propose a lightweight framework that models conversational emotion as a sequence of latent emotional regimes using sticky factorial HDP-HMMs over multimodal valence-arousal representations derived from simultaneous video, audio and textual input. We evaluate the quality of regime prediction using LLM-as-a-Judge, geometric, and temporal consistency metrics, demonstrating that the sticky HDP-HMM produces more interpretable regime sequences than the baseline Gaussian HMM at a fraction of the computational cost of LLM-based dialogue state tracking methods. In addition, Question-Answer experiments in a clinical dataset suggest that meaningful emotional phases can reliably be recovered from multimodal valence-arousal trajectories and used to improve the quality of LLM responses in unstable affective regimes via context augmentation. This framework thus opens a path toward interpretable, lightweight, and actionable analysis of conversational emotion dynamics at scale.", "date": "2026-05-13", "categories": []}
{"id": "2502.20209", "title": "DIPSER: A Dataset for In-Person Student Engagement Recognition in the Wild", "abstract": "In this paper, a novel dataset is introduced, designed to assess student attention within in-person classroom settings. This dataset encompasses RGB camera data, featuring multiple cameras per student to capture both posture and facial expressions, in addition to smartwatch sensor data for each individual. This dataset allows machine learning algorithms to be trained to predict attention and correlate it with emotion. A comprehensive suite of attention and emotion labels for each student is provided, generated through self-reporting as well as evaluations by four different experts. Our dataset uniquely combines facial and environmental camera data, smartwatch metrics, and includes underrepresented ethnicities in similar datasets, all within in-the-wild, in-person settings, making it the most comprehensive dataset of its kind currently available.\n  The dataset presented offers an extensive and diverse collection of data pertaining to student interactions across different educational contexts, augmented with additional metadata from other tools. This initiative addresses existing deficiencies by offering a valuable resource for the analysis of student attention and emotion in face-to-face lessons.", "date": "2025-02-27", "categories": []}
{"id": "2605.11303", "title": "Predicting Psychological Well-Being from Spontaneous Speech using LLMs", "abstract": "We investigate the use of Large Language Models (LLMs) for zero-shot prediction of Ryff Psychological Well-Being (PWB) scores from spontaneous speech. Using a few minutes of voice recordings from 111 participants in the PsyVoiD database, we evaluated 12 instruction-tuned LLMs, including Llama-3 (8B, 70B), Ministral, Mistral, Gemma-2-9B, Gemma-3 (1B, 4B, 27B), Phi-4, DeepSeek (Qwen and Llama), and QwQ-Preview. A domain-informed prompt was developed in collaboration with experts in clinical psychology and linguistics. Results show that LLMs can extract semantically meaningful cues from spontaneous speech, achieving Spearman correlations of up to 0.8 on 80\\% of the data. Additionally, to enhance explainability, we conducted statistical analyses to characterise prediction variability and systematic biases, alongside keyword-based word cloud analyses to highlight the linguistic features driving the models' predictions.", "date": "2026-05-11", "categories": []}
{"id": "2605.08252", "title": "Multimodal Emotion Recognition via Causal-Diffusion Bridge (Affect-Diff)", "abstract": "Multimodal emotion recognition on CMU-MOSEI faces an extreme imbalance as Happy accounts for 65.9% of samples while three Ekman categories collectively represent under 7%, causing standard fusion models to maximize accuracy by ignoring minority emotions entirely. We present Affect-Diff, a Causal-Diffusion Bridge that addresses this through three jointly trained mechanisms: a NOTEARS-learned causal graph that re-weights modality contributions before fusion, a beta-VAE bottleneck for regularized latent compression, and a stop-gradiented 1D DDPM prior that structures the latent space against majority-class collapse. On 3,292 aligned CMU-MOSEI samples, Affect-Diff achieves validation balanced accuracy 0.384, an 18% relative improvement over the strongest baseline (TETFN: 0.324), while all evaluated baselines produce zero F1 on Fear, Disgust, and Surprise. Ablation studies confirm independent, non-redundant contributions from the diffusion prior (-24% without it) and causal graph (-13%). Notably, only the deterministic-encoder variant detects all six emotion classes, revealing KL regularization strength as a direct lever for minority-class sensitivity.", "date": "2026-05-07", "categories": []}
{"id": "2605.08847", "title": "EmoS: A High-Fidelity Multimodal Benchmark for Fine-grained Streaming Emotional Understanding", "abstract": "In the context of today's high-pressure, aging society, the demand for large-scale emotional models capable of providing empathetic support is more critical than ever. However, existing benchmarks fail to simultaneously achieve ecological validity, signal clarity, and reliable fine-grained labeling. We introduce EmoS, a high-fidelity bilingual benchmark designed to resolve the limitations of ecological validity and noise in existing datasets by combining strictly filtered static slices with a dynamic Streaming Monologue subset. Supported by a rigorous dual-layer human annotation pipeline, EmoS provides trusted ground truth that captures continuous emotional evolution. Empirical results show that fine-tuning MLLMs (multimodal large language models) on EmoS yields significant gains over zero-shot baselines, laying the foundation for the training and evaluation of future emotion recognition models and empathy models. The dataset and code are publicly available at https://github.com/NLP2CT/EmoS.", "date": "2026-05-09", "categories": []}
{"id": "2605.09703", "title": "MOTOR-Bench: A Real-world Dataset and Multi-agent Framework for Zero-shot Human Mental State Understanding", "abstract": "Understanding human mental states from natural behavior is crucial for intelligent systems in the real world. However, most current research focuses on predicting isolated mental state labels, lacking structured annotations of complex interpersonal interactions. To support structured analysis, we introduce MOTOR-Bench, a carefully-designed benchmark with a real-world dataset MOTOR-dataset, containing 1,440 multimodal video clips in collaborative learning scenarios, reflecting key real-world data challenges including natural class imbalance, visual noise, and domain-specific language. Each sample is labeled by educational experts based on self-regulated learning theory. We further evaluate several state-of-the-art multimodal large language models and multi-agent systems in a zero-shot setting on our MOTOR-Bench. However, their performance on this task remains limited, suggesting that existing methods still struggle with structured reasoning from observable behavior to deeper mental states. To address this challenge, we propose a reasoning multi-agent framework, named MOTOR-MAS. It coordinates multiple agents through a structured agent coordination mechanism to infer explicit behaviors, internal cognitions, and psychological emotions. Experimental results show that our MOTOR-MAS outperforms the best single-model benchmark by 15.93 points in Macro-F1 scores for the three labels of behavior, cognition, and emotion, and outperforms the general multi-agent benchmark by 10.2 points in internal cognition prediction.", "date": "2026-05-10", "categories": []}
{"id": "2604.16579", "title": "EviDep: Trustworthy Multimodal Depression Estimation via Disentangled Evidential Learning", "abstract": "Automated multimodal depression estimation in unconstrained environments is inherently challenged by naturalistic noise and complex behavioral variability. Prevailing deterministic methods, however, produce uncalibrated point estimates without quantifying predictive uncertainty, exposing decision-making to the risk of overconfident, untrustworthy estimates. To establish a reliable and trustworthy estimation paradigm, we propose EviDep, an evidential learning framework that jointly quantifies depression severity alongside aleatoric and epistemic uncertainties via a Normal-Inverse-Gamma distribution. To ensure the integrity of the extracted behavioral evidence and prevent artificial confidence inflation during multimodal fusion, EviDep introduces two tailored mechanisms. First, addressing the temporal-frequency heterogeneity of behavioral cues, a Frequency-aware Feature Extraction module leverages a wavelet-based Mixture-of-Experts to dynamically decouple stable macro-level affective baselines from transient micro-level behavioral bursts, effectively filtering out task-irrelevant artifacts. Second, a Disentangled Evidential Learning strategy enforces explicit decorrelation of features in these purified representations. By separating the cross-modal shared consensus from modality-specific behavioral nuances before Bayesian fusion, this rigorous disentanglement strictly prevents the model from double-counting overlapping information. Extensive experiments on the AVEC 2013, AVEC 2014, DAIC-WOZ, and E-DAIC datasets confirm that EviDep achieves state-of-the-art predictive accuracy and superior uncertainty calibration, thereby delivering a trustworthy, risk-aware decision-support tool for depression estimation.", "date": "2026-04-17", "categories": []}
{"id": "2605.05640", "title": "AffectSeek: Agentic Affective Understanding in Long Videos under Vague User Queries", "abstract": "Existing affective understanding studies have mainly focused on recognizing emotions from images, audio signals, or pre-cliped video clips, where the affective evidence is already given. This passive and clip-centered setting does not fully reflect real-world scenarios, in which users often interact with long videos and express their needs through natural-language queries. In this paper, we study \\textbf{Vague-Query-driven video Affective Understanding (VQAU)}, a new task that requires models to localize affective moments in long videos, predict their emotion categories, and generate evidence-grounded rationales under vague user queries. To support this task, we construct \\textbf{VQAU-Bench}, a benchmark that integrates long videos, vague affective queries, temporal clip annotations, emotion labels, and rationale explanations into a unified evaluation framework. VQAU-Bench enables systematic assessment of semantic-temporal-affective alignment, affective moment localization, emotion classification, and rationale generation. To address the multi-step reasoning challenges of VQAU, we further propose \\textbf{AffectSeek}, an agentic framework that actively seeks, verifies, and explains affective moments in long videos. AffectSeek decomposes VQAU into intent interpretation, candidate localization, clip verification, emotion reasoning, and rationale generation, and progressively aligns vague user intent with long-video evidence through role-specialized reasoning and cross-stage verification. Experiments show that VQAU remains challenging for existing affective recognition models and single-step vision-language models, while AffectSeek provides a simple yet effective framework for agentic long-video affective understanding.", "date": "2026-05-07", "categories": []}
{"id": "2605.05694", "title": "Adaptive Physical-Facial Representation Fusion via Subject-Invariant Cross-Modal Prompt Tuning for Video-Based Emotion Recognition", "abstract": "Emotion recognition from facial videos enables non-contact inference of human emotional states. Although facial expressions are widely used cues, they cannot fully reflect intrinsic affective states. Remote photoplethysmography (rPPG) provides complementary physiological information, but it is highly susceptible to noise and inter-subject variability, limiting generalization to unseen individuals. Existing multimodal methods combine facial and rPPG features, yet their fusion strategies often disrupt pretrained facial representations and lack explicit mechanisms to suppress subject-specific variations. To address these issues, we propose a subject-invariant cross-modal prompt-tuning framework for video-based emotion recognition. Specifically, rPPG waveforms are transformed into noise-robust time-frequency representations (TFRs), from which modality-complementary prompts are generated to modulate facial tokens within a frozen Vision Transformer (ViT). This design enables effective cross-modal interaction while preserving the generalizable facial representations learned by the pretrained backbone. In addition, we introduce a decoupled shared-specific adapter (DSSA) into each ViT layer to explicitly separate subject-shared and subject-specific components, thereby improving cross-subject generalization. Experiments on the MAHNOB-HCI and DEAP benchmarks demonstrate that the proposed method consistently outperforms strong baselines in both recognition accuracy and generalization ability, highlighting its effectiveness for video-based emotion recognition.", "date": "2026-05-07", "categories": []}
{"id": "2605.04877", "title": "To Fuse or to Drop? Dual-Path Learning for Resolving Modality Conflicts in Multimodal Emotion Recognition", "abstract": "Multimodal emotion recognition (MER) benefits from combining text, audio, and vision, yet standard fusion often fails when modalities conflict. Crucially, conflicts differ in resolvability: benign conflicts stem from missing, weak, or ambiguous cues and can be mitigated by cross-modal calibration, while severe conflicts arise from intrinsically contradictory (e.g., sarcasm) or misleading signals, for which forced fusion may amplify errors. Recognizing this, we propose Dual-Path Conflict Resolution (DCR), a unified framework that learns when to fuse and when to drop modalities. Path I (Affective Fusion Distiller, AFD) performs reverse distillation from audio/visual teachers to a textual student using temporally weighted class evidence, thereby enhancing representation-level calibration and improving fusion when alignment is beneficial. Path II (Affective Discernment Agent, ADA) formulates MER as a contextual bandit that selects among fusion and unimodal predictions based on a dual-view state and a calibration-aware reward, enabling decision-level arbitration under irreconcilable conflicts without requiring per-modality reliability labels. By taking into account the full multimodal context and coupling soft calibration with hard arbitration, DCR reconciles conflicts that can be aligned while bypassing misleading modalities when fusion is harmful. Across five benchmarks covering both dialogue-level and clip-level MER, DCR consistently outperforms competitive baselines or achieves highly competitive results. Further ablations, conflict-specific subset evaluation, and modality-selection analysis verify that AFD and ADA are complementary and jointly improve robust conflict-aware emotion recognition.", "date": "2026-05-06", "categories": []}
{"id": "2605.03261", "title": "Can AI Help You Get Over Your Breakup? One Session with a Belief-Reframing Chatbot Shows Sustained Distress Reduction", "abstract": "Romantic breakups are among the most common and intense sources of psychological distress. We evaluated *overit*, a single-session AI chatbot that uses cognitive reappraisal to address breakup distress, informed by memory reconsolidation theory. In a pre-registered randomized controlled trial, 254 adults in the United States and United Kingdom who had experienced a romantic breakup were assigned to either an initial survey assessment followed by an AI chat session or to a survey-only control. Breakup distress was measured at baseline, 7 days, and again at an exploratory 1-month follow-up using the Breakup Distress Scale. Participants assigned to *overit* showed a significantly greater reduction in breakup distress than controls at 7 days (time-by-condition interaction B = -5.36, SE = 1.19, p < .001; completer-based d = -0.70). A smaller but still significant treatment advantage remained detectable at the exploratory 1-month follow-up among post-session completers (B = -2.92, SE = 1.22, p = .017). Exploratory post hoc moderation suggested a larger effect among male participants (B = 7.78, p = .003). These results suggest that a brief AI chatbot conversation can meaningfully reduce breakup distress, with exploratory evidence that a smaller advantage persists over the following month. Future work should test the intervention against active controls, evaluate repeated-session use, and recruit more diverse samples.", "date": "2026-05-05", "categories": []}
{"id": "2605.03615", "title": "PriorNet: Prior-Guided Engagement Estimation from Face Video", "abstract": "Engagement estimation from face video remains challenging because facial evidence is often incomplete, labeled data are limited, and engagement annotations are subjective. We present PriorNet, a prior-guided framework that injects task-relevant priors at three stages of the pipeline: preprocessing, model adaptation, and objective design. PriorNet converts face-detection failures into explicit zero-frame placeholders so that missing-face events remain represented in the input sequence, adapts a frozen Self-supervised Video Facial Affect Perceiver (SVFAP) backbone through a Prior-guided Low-Rank Adaptation module (Prior-LoRA) for parameter-efficient specialization, and trains with a Dirichlet-evidential, uncertainty-weighted objective under hard-label supervision. We evaluate PriorNet on EngageNet, DAiSEE, DREAMS, and PAFE using each dataset's native evaluation protocol. Across these benchmarks, PriorNet improves over the strongest listed prior reference within each dataset's evaluation framing, while component ablations on EngageNet and DAiSEE indicate that the gains arise from complementary contributions of preprocessing, adaptation, and objective-level priors. These results support explicit prior injection as a useful design principle for face-video engagement estimation under the benchmark conditions studied in this work.", "date": "2026-05-05", "categories": []}
{"id": "2511.20657", "title": "Intelligent Agents with Emotional Intelligence: Current Trends, Challenges, and Future Prospects", "abstract": "The development of agents with emotional intelligence is becoming increasingly vital due to their significant role in human-computer interaction and the growing integration of computer systems across various sectors of society. Affective computing aims to design intelligent systems that can recognize, evoke, and express human emotions, thereby emulating human emotional intelligence. While previous reviews have focused on specific aspects of this field, there has been limited comprehensive research that encompasses emotion understanding, elicitation, and expression, along with the related challenges. This survey addresses this gap by providing a holistic overview of core components of artificial emotion intelligence. It covers emotion understanding through multimodal data processing, as well as affective cognition, which includes cognitive appraisal, emotion mapping, and adaptive modulation in decision-making, learning, and reasoning. Additionally, it addresses the synthesis of emotional expression across text, speech, and facial modalities to enhance human-agent interaction. This paper identifies and analyzes the key challenges and issues encountered in the development of affective systems, covering state-of-the-art methodologies designed to address them. Finally, we highlight promising future directions, with particular emphasis on the potential of generative technologies to advance affective computing.", "date": "2025-10-11", "categories": []}
{"id": "2601.04376", "title": "Combining Facial Videos and Biosignals for Stress Estimation During Driving", "abstract": "Reliable stress recognition is critical in applications such as medical monitoring and safety-critical systems, including real-world driving. While stress is commonly detected using physiological signals such as perinasal perspiration and heart rate, facial activity provides complementary cues that can be captured unobtrusively from video. We propose a multimodal stress estimation framework that combines facial videos and physiological signals, remaining effective even when biosignal acquisition is challenging. Facial behavior is represented using a dense 3D Morphable Model, yielding a 56-dimensional descriptor that captures subtle expression and head-pose dynamics over time. To study how stress modulates facial motion, we perform extensive experiments alongside established physiological markers. Paired hypothesis tests between baseline and stressor phases show that 38 of 56 facial components exhibit consistent, phase-specific stress responses comparable to physiological markers. Building on these findings, we introduce a Transformer-based temporal modeling framework and evaluate unimodal, early-fusion, and cross-modal attention strategies. Cross-modal attention fusion of 3D-derived facial features with physiological signals substantially improves performance over physiological signals alone, increasing AUROC from 52.7% and accuracy from 51.0% to 92.0% and 86.7%, respectively. Although evaluated on driving data, the proposed framework and protocol may generalize to other stress estimation settings.", "date": "2026-01-07", "categories": []}
{"id": "2604.11730", "title": "Multimodal Ambivalence/Hesitancy Recognition in Videos for Personalized Digital Health Interventions", "abstract": "Using behavioural science, health interventions focus on behaviour change by providing a framework to help patients acquire and maintain healthy habits that improve medical outcomes. In-person interventions are costly and difficult to scale, especially in resource-limited regions. Digital health interventions offer a cost-effective approach, potentially supporting independent living and self-management. Automating such interventions, especially through machine learning, has gained considerable attention recently. Ambivalence and hesitancy (A/H) play a primary role for individuals to delay, avoid, or abandon health interventions. A/H are subtle and conflicting emotions that place a person in a state between positive and negative evaluations of a behaviour, or between acceptance and refusal to engage in it. They manifest as affective inconsistency across modalities or within a modality, such as language, facial, vocal expressions, and body language. While experts can be trained to recognize A/H, integrating them into digital health interventions is costly and less effective. Automatic A/H recognition is therefore critical for the personalization and cost-effectiveness of digital health interventions. Here, we explore the application of deep learning models for A/H recognition in videos, a multi-modal task by nature. In particular, this paper covers three learning setups: supervised learning, unsupervised domain adaptation for personalization, and zero-shot inference via large language models (LLMs). Our experiments are conducted on the unique and recently published BAH video dataset for A/H recognition. Our results show limited performance, suggesting that more adapted multi-modal models are required for accurate A/H recognition. Better methods for modeling spatio-temporal and multimodal fusion are necessary to leverage conflicts within/across modalities.", "date": "2026-04-13", "categories": []}
{"id": "2605.01024", "title": "EmoMM: Benchmarking and Steering MLLM for Multimodal Emotion Recognition under Conflict and Missingness", "abstract": "Multimodal Emotion Recognition (MER) is critical for interpreting real-world interactions. While Multimodal Large Language Models (MLLM) have shown promise in MER, their internal decision-making mechanisms under modality conflict and missingness remain largely underexplored. In this paper, to systematically investigate these behaviors, we introduce EmoMM, a comprehensive benchmark featuring modality-aligned, conflict, and missing subsets. Through extensive evaluation, we uncover a Video Contribution Collapse (VCC) phenomenon, where MLLM marginalize video evidence due to high token redundancy and modality preferences. To address this, we propose Conflict-aware Head-level Attention Steering (CHASE), a lightweight mechanism that detects modality conflicts and performs inference-time attention steering, effectively mitigating decision bias without retraining the backbone. Experimental results demonstrate that CHASE consistently improves performance across various settings, significantly enhancing the reliability of MLLM in complex affective scenarios.", "date": "2026-05-01", "categories": []}
{"id": "2605.02665", "title": "Fuzzy Fingerprinting Encoder Pre-trained Language Models for Emotion Recognition in Conversations: Human Assessment and Validity Study", "abstract": "In Emotion Recognition in Conversations (ERC), model decisions should align with nuanced human perception and ideally provide insights on the classification process. Standard encoder pre-trained language models (PLMs) are the state-of-the-art at these tasks but offer little insight into why a certain prediction is made. This is especially problematic in imbalanced datasets, where most utterances are labeled as neutral, making these models frequently misclassify minority emotions as the majority neutral class. To tackle this issue, we introduced a novel, interpretable approach to ERC by combining PLMs with Fuzzy Fingerprints (FFPs). FFP provide class-specific prototypes that reflect the characteristic class activation patterns in the PLM's latent space. They are derived by ranking and fuzzifying the activations of the pooled conversational context-dependent embeddings across training instances for each emotion. At inference time, each input utterance is similarly fuzzy fingerprinted and matched to the emotion prototypes using a fuzzy similarity function based on the aggregation of the intersection of the fuzzy sets that define each FFP. Experimental results show that FFP integration reduces overclassification into the neutral class and human evaluation further supports the adequacy of FFP predictions. Our proposed method thus bridges the gap between deep neural inference and human perception, performing at state-of-the-art level while simultaneously offering valuable insights into the classification procedure.", "date": "2026-05-04", "categories": []}
{"id": "2605.02672", "title": "The 2026 ACII Dyadic Conversations (DaiKon) Workshop & Challenge", "abstract": "The 2026 ACII Dyadic Conversations (ACII-DaiKon) Workshop & Challenge introduces a benchmark for modeling interpersonal affect and social dynamics in dyadic conversations. Although conversational affect modeling has advanced rapidly, most benchmarks remain speaker-centric and underrepresent coupled, time-evolving processes between partners, including directional influence, conversational timing coordination, and rapport development. To address this gap, ACII-DaiKon presents three coordinated sub-challenges built on a shared dataset: (1) directional interpersonal influence prediction, (2) turn-taking prediction (next-speaker and time-to-next-speech), and (3) rapport trajectory prediction across full interactions.\n  The challenge is built on the Hume-DaiKon dataset, comprising 945 dyadic conversations (743.4 hours of audiovisual data) collected under naturalistic conditions across five languages. The benchmark supports multimodal modeling, temporal reasoning, and cross-context generalization through fixed train/validation/test splits, standardized metrics, and released baseline systems. Evaluation uses Concordance Correlation Coefficient (CCC), Pearson correlation, Macro-F1, and Mean Absolute Error (MAE) depending on the sub-challenge.\n  Baseline experiments establish initial reference performance, with best test results of 0.40 CCC and 0.50 Pearson for influence prediction, 0.66 Macro-F1 and 1.50~s MAE for turn-taking, and 0.68 CCC and 0.70 Pearson for rapport trajectory modeling. These results indicate that while current methods capture coarse dyadic patterns, robust modeling of directional dependence and long-horizon interpersonal dynamics remains challenging. The workshop provides a shared platform for rigorous comparison and cross-disciplinary discussion on data validity, evaluation protocols, and culturally aware modeling for dyadic interaction.", "date": "2026-05-04", "categories": []}
{"id": "2604.27345", "title": "LLMs Capture Emotion Labels, Not Emotion Uncertainty: Distributional Analysis and Calibration of Human-LLM Judgment Gaps", "abstract": "Human annotators frequently disagree on emotion labels, yet most evaluations of Large Language Model (LLM) emotion annotation collapse these judgments into a single gold standard, discarding the distributional information that disagreement encodes. We ask whether LLMs capture the structure of this disagreement, not just majority labels, by comparing emotion judgment distributions between human annotators and four zero-shot LLMs, plus a fine-tuned RoBERTa baseline, across two complementary benchmarks: GoEmotions and EmoBank, totaling 640,000 LLM responses. Zero-shot models diverge substantially from human distributions, and in-domain fine-tuning, not model scale, is required to close the gap. We formalize a lexical-grounding gradient through a quantitative transparency score that predicts per-category human--LLM agreement: LLMs reliably capture emotions with explicit lexical markers but systematically fail on pragmatically complex emotions requiring contextual inference, a pattern that replicates across both categorical and continuous emotion frameworks. We further propose three lightweight post-hoc calibration methods that reduce the distributional gap by up to 14\\%, and provide actionable guidelines for when LLM emotion annotations can, and cannot, substitute for human labeling.", "date": "2026-04-30", "categories": []}
{"id": "2605.00184", "title": "Introducing WARM-VR: Benchmark Dataset for Multimodal Wearable Affect Recognition in Virtual Reality", "abstract": "With the growing integration of human-computer interaction into everyday life, advances in machine learning have enabled systems to better perceive and respond to users' emotional states. Most existing affect recognition datasets focus on static environments, limiting their applicability to immersive multimedia contexts such as Virtual Reality (VR). In this paper, we introduce WARM-VR, a novel publicly available multimodal dataset designed to support affect recognition in immersive, multisensory environments using wearable sensing instrumentation. Data were collected from 31 participants aged 19-37 using wearable sensors: a wristband measuring Blood Volume Pulse (BVP), EDA, skin Temperature, three-axis Acceleration, and a chest strap recording ECG signals. Participants engaged in immersive VR experiences designed to elicit relaxation through a calming beach environment following stress induction via an arithmetic task. These sessions incorporated synchronized multimedia stimuli: visual, auditory, and olfactory. Affective states were assessed subjectively through validated self-report questionnaires and objectively through the analysis of physiological measurements. Statistical analysis of the questionnaires confirmed that VR relaxation significantly reduced negative affect, particularly with olfactory enhancement. Furthermore, we established a benchmark on the dataset using widely recognized machine learning algorithms. The best performance for binary classification from BVP data of valence, was obtained with a CNN and a CNN-Bi-GRU model, both achieving an average F1-score of 0.63 and an AUC of 0.69. For arousal, a lightweight Transformer architecture provided the most balanced results (F1-0 0.54 and F1-1 0.63), outperforming recurrent hybrids. In the relaxation task, a CNN-Bi-GRU model reached the highest overall performance (average F1-score 0.64, AUC 0.69).", "date": "2026-04-30", "categories": []}
{"id": "2604.25938", "title": "Speech Emotion Recognition Using MFCC Features and LSTM-Based Deep Learning Model", "abstract": "Speech Emotion Recognition (SER) is the use of machines to detect the emotional state of humans based on the speech, which is gaining importance in natural human-computer interaction. Speech is a very valuable source of information, as emotions modify the patterns of speech; pitch, energy and even timing. Nonetheless, SER is not an easy task because speakers are not constant, and situations vary when recording and the sound similarity between specific feelings. In this work, the author introduces a speech emotion recognition system relying on the Mel-Frequency Cepstral Coefficient and Long Short-Term Memory (LSTM) neural network, as a feature extraction method. The Toronto Emotional Speech Set (TESS) speech signal was pre-processed, and transformed into MFCC features to understand the important aspects in terms of time. The resultant features were then introduced to LSTM model, which is able to learn long term features of sequential audio data. The trained model was measured over several emotion classes occurring in the dataset. As seen in the results of experiments, the proposed MFCC-LSTM approach succeeds in capturing the patterns of emotions in speech and provides highly realistic classifications in all the chosen emotion classifications. This study presents a speech emotion recognition system using Mel-Frequency Cepstral Coefficients (MFCCs) as features and a deep learning LSTM classifier. A Support Vector Machine (SVM) with an RBF kernel served as a classical baseline, achieving 98% accuracy, against which the proposed LSTM model, achieving 99% accuracy, was validated. Overall, it is possible to confirm that LSTM-based architectures can be used to address the task of speech emotion recognition. Actual applications of the proposed system may be virtual assistants and mental health surveillance.", "date": "2026-04-16", "categories": []}
{"id": "2604.26078", "title": "PPG-Based Affect Recognition with Long-Range Deep Models: A Measurement-Driven Comparison of CNN, Transformer, and Mamba Architectures", "abstract": "Photoplethysmography (PPG) is increasingly used in wearable affective computing due to its low cost and ease of integration into consumer devices. Recent advances in deep learning have introduced long-range sequence models, such as Transformers, and state-space models, like Mamba, which have demonstrated strong performance on natural language and general time-series tasks. However, it remains unclear whether these architectures offer tangible benefits over widely used Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTMs) for PPG-based affect recognition, given that datasets are typically small and noisy. This work presents a measurement-driven comparison of four deep learning architectures, CNN, CNN-LSTM hybrid, Transformers, and Mamba, for classifying arousal, valence, and relaxation states from wrist-based PPG signals. All models are evaluated under a subject-independent 5-fold cross-validation protocol using identical preprocessing, segmentation, and training pipelines. Our results show that the Transformer and Mamba models achieve performance comparable to that of a CNN baseline, but do not consistently outperform it across all tasks. CNNs remain the most effective overall, providing the highest accuracy with the smallest model size, whereas Transformers have a better balance of F1 scores for Arousal and Relaxation. The study provides the first evaluation of Transformer and Mamba models for PPG-based affect recognition, offering practical guidance on model selection for wearable affective monitoring systems.", "date": "2026-04-28", "categories": []}
{"id": "2604.26417", "title": "EmoTransCap: Dataset and Pipeline for Emotion Transition-Aware Speech Captioning in Discourses", "abstract": "Emotion perception and adaptive expression are fundamental capabilities in human-agent interaction. While recent advances in speech emotion captioning (SEC) have improved fine-grained emotional modeling, existing systems remain limited to static, single-emotion characterization within isolated sentences, neglecting dynamic emotional transitions at the discourse level. To address this gap, we propose Emotion Transition-Aware Speech Captioning (EmoTransCap), a paradigm that integrates temporal emotion dynamics with discourse-level speech description. To construct a dataset rich in emotion transitions while enabling scalable expansion, we design an automated pipeline for dataset creation. This is the first large-scale dataset explicitly designed to capture discourse-level emotion transitions. To generate semantically rich descriptions, we incorporate acoustic attributes and temporal cues from discourse-level speech. Our Multi-Task Emotion Transition Recognition (MTETR) model performs joint emotion transition detection and diarization. Leveraging the semantic analysis capabilities of LLMs, we produce two annotation versions: descriptive and instruction-oriented. These data and annotations offer a valuable resource for advancing emotion perception and emotional expressiveness. The dataset enables speech captions that capture emotional transitions, facilitating temporal-dynamic and fine-grained emotion understanding. We also introduce a controllable, transition-aware emotional speech synthesis system at the discourse level, enhancing anthropomorphic emotional expressiveness and supporting emotionally intelligent conversational agents.", "date": "2026-04-29", "categories": []}
{"id": "2510.09382", "title": "CHUCKLE -- When Humans Teach AI To Learn Emotions The Easy Way", "abstract": "Curriculum learning (CL) structures training from simple to complex samples, facilitating progressive learning. However, existing CL approaches for emotion recognition often rely on heuristic, data-driven, or model-based definitions of sample difficulty, neglecting the difficulty for human perception, a critical factor in subjective tasks like emotion recognition. We propose CHUCKLE (Crowdsourced Human Understanding Curriculum for Knowledge Led Emotion Recognition), a perception-driven CL framework that leverages annotator agreement and alignment in crowd-sourced datasets to define sample difficulty, under the assumption that clips challenging for humans are similarly hard for neural networks. Experimental results suggest that CHUCKLE enhances the performance of LSTMs and Transformers over non-curriculum baselines, while reducing the number of gradient updates, thereby enhancing both training efficiency and model robustness in both subject-dependent and subject-independent settings.", "date": "2025-10-10", "categories": []}
{"id": "2604.25383", "title": "ML-SAN: Multi-Level Speaker-Adaptive Network for Emotion Recognition in Conversations", "abstract": "To establish empathy with machines, it is essential to fully understand human emotional changes. However, research in multimodal emotion recognition often overlooks one problem: individual expressive traits vary significantly, which means that different people may express emotions differently. In our daily lives, we can see this. When communicating with different people, some express \"happiness\" through their facial expressions and words, while others may hide their happiness or express it through their actions. Both are expressions of 'happiness,' but such differences in emotional expression are still too difficult for machines to distinguish. Current emotion recognition remains at a 'static' level, using a single recognition model to identify all emotional styles. This \"simplification\" often affects the recognition results, especially in multi-turn dialogues. To address this problem, this paper introduces a novel Multi-Level Speaker Adaptive Network (ML-SAN), which, specifically, effectively addresses the challenge of speaker identity information confusion. ML-SAN does not simply assign a speaker's ID after recognition; instead, it employs a three-stage adaptive process: First, Input-level Calibration uses Feature-Level Linear Modulation (FiLM) to adjust the raw audio and visual features into a neutral space unrelated to the speaker. Then, Interaction-level Gating re-adjusts the trust level for each modality (e.g., voice or facial features) based on the speaker's identity information. Finally, Output-level Regularization maintains the consistency of speaker features in the latent space. Tests on the MELD and IEMOCAP datasets show that our model (ML-SAN) achieves better results, performs exceptionally well in handling challenging tail sentiment categories, and better addresses the diversity of speakers in real-world scenarios.", "date": "2026-04-28", "categories": []}
{"id": "2604.25612", "title": "The Nonverbal Syntax Framework: An Evidence-Based Tiered System for Inferring Learner States from Observable Behavioral Cues", "abstract": "Understanding learners' cognitive and affective states underpins adaptive educational systems and effective teaching. Although research links nonverbal cues to internal states, no framework calibrates them to evidence. We present the Nonverbal Syntax Framework, drawn from a systematic review of 908 studies and 17,043 cue-state mappings (Turaev et al., 2026). The framework addresses three challenges: terminological fragmentation (behaviors described inconsistently), evidence heterogeneity (single observations to replicated findings), and state ambiguity (similar patterns indicating multiple states). Normalization consolidated 5,537 state labels into 2,010 canonical states (63.7%) and 11,521 cues into 6,434 normalized cues (44.2%) across nine behavioral channels. Dual-evidence assessment separately evaluates Component Evidence (coverage of cues and states) and Relationship Evidence (independent studies per cue-state link). 52% of \"Very High\" relationships rest on one paper, so separation enables calibrated rather than overconfident inference from preliminary findings. The framework's four levels comprise a Cue Vocabulary of 6,434 indicators classified as observable/instrumental; State Clusters linking 2,010 states to indicative cues; State Profiles with multimodal behavioral signatures and actionable specifications; and Discriminative Analysis distinguishing 1,215 confusable state pairs. We identify 480 actionable R1-R4 relationships (three or more independent papers), the replicated core of six decades of research, covering 35.5% of mappings across 47 key learning states and 111 distinct indicators. The remaining 91.5% (9,653 single-paper findings) form exploratory hypotheses for replication. The framework gives researchers an empirical foundation for identifying gaps, practitioners evidence-based tools for state inference, and technologists validated features for multimodal detection.", "date": "2026-04-28", "categories": []}
{"id": "2604.25776", "title": "Unrequited Emotions: Investigating the Gaps in Motivation and Practice in Speech Emotion Recognition Research", "abstract": "Critical analyses of emotion recognition technology have raised ethical concerns around task validity and potential downstream impacts, urging researchers to ensure alignment between their stated motivations and practice. However, these discussions have not adequately influenced or drawn from research on speech emotion recognition (SER). We address this gap by conducting a systematic survey of SER research to uncover what stated motivations drive this work and if they align with the datasets and emotions studied. We find that while SER research identifies appealing goals, such as well-situated voice-activated systems or healthcare applications, commonly-used datasets do not reflect these proposed deployment contexts, thus presenting a gap between motivations and research practices. We argue that such gaps engender ethical concerns, and that SER research should reassert itself with concrete use-cases to prevent misinterpretations, misuse, and downstream harms.", "date": "2026-04-28", "categories": []}
{"id": "2604.25866", "title": "From Syntax to Emotion: A Mechanistic Analysis of Emotion Inference in LLMs", "abstract": "Large language models (LLMs) are increasingly used in emotionally sensitive human-AI applications, yet little is known about how emotion recognition is internally represented. In this work, we investigate the internal mechanisms of emotion recognition in LLMs using sparse autoencoders (SAEs). By analyzing sparse feature activations across layers, we identify a consistent three-phase information flow, in which emotion-related features emerge only in the final phase. We further show that emotion representations comprise both shared features across emotions and emotion-specific features. Using phase-stratified causal tracing, we identify a small set of features that strongly influence emotion predictions, and show that both their number and causal impact vary across emotions; in particular, Disgust is more weakly and diffusely represented than other emotions. Finally, we propose an interpretable and data-efficient causal feature steering method that significantly improves emotion recognition performance across multiple models while largely preserving language modeling ability, and demonstrate that these improvements generalize across multiple emotion recognition datasets. Overall, our findings provide a systematic analysis of the internal mechanisms underlying emotion recognition in LLMs and introduce an efficient, interpretable, and controllable approach for improving model performance.", "date": "2026-04-28", "categories": []}
{"id": "2502.04424", "title": "EmoBench-M: Benchmarking Emotional Intelligence for Multimodal Large Language Models", "abstract": "With the integration of multimodal large language models (MLLMs) into robotic systems and AI applications, embedding emotional intelligence (EI) capabilities is essential for enabling these models to perceive, interpret, and respond to human emotions effectively in real-world scenarios. Existing static, text-based, or text-image benchmarks overlook the multimodal complexities of real interactions and fail to capture the dynamic, context-dependent nature of emotional expressions, rendering them inadequate for evaluating MLLMs' EI capabilities. To address these limitations, we introduce EmoBench-M, a systematic benchmark grounded in established psychological theories, designed to evaluate MLLMs across 13 evaluation scenarios spanning three hierarchical dimensions: foundational emotion recognition (FER), conversational emotion understanding (CEU), and socially complex emotion analysis (SCEA). Evaluation was conducted on 27 state-of-the-art MLLMs, using both objective task-specific metrics and LLM-based evaluation, revealing a substantial performance gap relative to human-level competence. Even the best performing models, Gemini-3.0-Pro and GPT-5.2, achieve the highest scores on EmoBench-M, 70.5 and 66.5 points respectively. Specialized models such as AffectGPT exhibit uneven performance across EmoBench-M, demonstrating strengths in certain scenarios but generally lacking comprehensive emotional intelligence. By providing a comprehensive, multimodal evaluation framework, EmoBench-M captures both the strengths and weaknesses of current MLLMs across diverse emotional contexts. All benchmark resources, including datasets and code, are publicly available at https://emo-gml.github.io/, facilitating further research and advancement in MLLM emotional intelligence.", "date": "2025-02-06", "categories": []}
{"id": "2604.23348", "title": "EmoTrans: A Benchmark for Understanding, Reasoning, and Predicting Emotion Transitions in Multimodal LLMs", "abstract": "Recent multimodal large language models (MLLMs) have shown strong capabilities in perception, reasoning, and generation, and are increasingly used in applications such as social robots and human-computer interaction, where understanding human emotions is essential. However, existing benchmarks mainly formulate emotion understanding as a static recognition problem, leaving it largely unclear whether current MLLMs can understand emotion as a dynamic process that evolves, shifts between states, and unfolds across diverse social contexts. To bridge this gap, we present EmoTrans, a benchmark for evaluating emotion dynamics understanding in multimodal videos. EmoTrans contains 1,000 carefully collected and manually annotated video clips, covering 12 real-world scenarios, and further provides over 3,000 task-specific question-answer (QA) pairs for fine-grained evaluation. The benchmark introduces four tasks, namely Emotion Change Detection (ECD), Emotion State Identification (ESI), Emotion Transition Reasoning (ETR), and Next Emotion Prediction (NEP), forming a progressive evaluation framework from coarse-grained detection to deeper reasoning and prediction. We conduct a comprehensive evaluation of 18 state-of-the-art MLLMs on EmoTrans and obtain two main findings. First, although current MLLMs show relatively stronger performance on coarse-grained emotion change detection, they still struggle with fine-grained emotion dynamics modeling. Second, socially complex settings, especially multi-person scenarios, remain substantially challenging, while reasoning-oriented variants do not consistently yield clear improvements. To facilitate future research, we publicly release the benchmark, evaluation protocol, and code at https://github.com/Emo-gml/EmoTrans.", "date": "2026-04-25", "categories": []}
{"id": "2604.23753", "title": "Modeling Induced Pleasure through Cognitive Appraisal Prediction via Multimodal Fusion", "abstract": "Multimodal affective computing analyzes user-generated social media content to predict emotional states. However, a critical gap remains in understanding how visual content shapes cognitive interpretations and elicits specific affective experiences such as pleasure. This study introduces a novel computational model to infer video-induced pleasure via cognitive appraisal variables. The proposed model addresses four challenges: (1) noisy and inconsistent human labels, (2) the semantic gap between \"positive emotions\" and \"pleasure,\" (3) the scarcity of pleasure-specific datasets, and (4) the limited interpretability of existing black-box fusion methods. Our approach integrates data-driven and cognitive theory-driven methods, using cognitive appraisal theory and a fuzzy model within an innovative framework. The model employs transformer-based architectures and attention mechanisms for fine-grained multimodal feature extraction and interpretable fusion to capture both inter- and intra-modal dynamics associated with pleasure. This enables the prediction of underlying appraisal variables, thereby bridging the semantic gap and enhancing model explainability beyond conventional statistical associations. Experimental results validate the efficacy of the proposed method in detecting video-induced pleasure, achieving a peak accuracy of 0.6624 in predicting pleasure levels. These findings highlight promising implications for affective content recommendation, intelligent media creation, and advancing our understanding of how digital media influences human emotions.", "date": "2026-04-26", "categories": []}
{"id": "2604.19763", "title": "Explainable Speech Emotion Recognition: Weighted Attribute Fairness to Model Demographic Contributions to Social Bias", "abstract": "Speech Emotion Recognition (SER) systems have growing applications in sensitive domains such as mental health and education, where biased predictions can cause harm. Traditional fairness metrics, such as Equalised Odds and Demographic Parity, often overlook the joint dependency between demographic attributes and model predictions. We propose a fairness modelling approach for SER that explicitly captures allocative bias by learning the joint relationship between demographic attributes and model error. We validate our fairness metric on synthetic data, then apply it to evaluate HuBERT and WavLM models finetuned on the CREMA-D dataset. Our results indicate that the proposed fairness model captures more mutual information between protected attributes and biases and quantifies the absolute contribution of individual attributes to bias in SSL-based SER models. Additionally, our analysis reveals indications of gender bias in both HuBERT and WavLM.", "date": "2026-03-26", "categories": []}
{"id": "2604.20307", "title": "Improving Facial Emotion Recognition through Dataset Merging and Balanced Training Strategies", "abstract": "In this paper, a deep learning framework is proposed for automatic facial emotion based on deep convolutional networks. In order to increase the generalization ability and the robustness of the method, the dataset size is increased by merging three publicly available facial emotion datasets: CK+, FER+ and KDEF. Despite the increase in dataset size, the minority classes still suffer from insufficient number of training samples, leading to data imbalance. The data imbalance problem is minimized by online and offline augmentation techniques and random weighted sampling. Experimental results demonstrate that the proposed method can recognize the seven basic emotions with 82% accuracy. The results demonstrate the effectiveness of the proposed approach in tackling the challenges of data imbalance and improving classification performance in facial emotion recognition.", "date": "2026-04-22", "categories": []}
{"id": "2602.06430", "title": "Investigating the structure of emotions by analyzing similarity and association of emotion words", "abstract": "In the field of natural language processing, some studies have attempted sentiment analysis on text by handling emotions as explanatory or response variables. One of the most popular emotion models used in this context is the wheel of emotion proposed by Plutchik. This model schematizes human emotions in a circular structure, and represents them in two or three dimensions. However, the validity of Plutchik's wheel of emotion has not been sufficiently examined. This study investigated the validity of the wheel by creating and analyzing a semantic networks of emotion words. Through our experiments, we collected data of similarity and association of ordered pairs of emotion words, and constructed networks using these data. We then analyzed the structure of the networks through community detection, and compared it with that of the wheel of emotion. The results showed that each network's structure was, for the most part, similar to that of the wheel of emotion, but locally different.", "date": "2026-02-06", "categories": []}
{"id": "2604.18988", "title": "A Multi-Agent Framework with Structured Reasoning and Reflective Refinement for Multimodal Empathetic Response Generation", "abstract": "Multimodal empathetic response generation (MERG) aims to generate emotionally engaging and empathetic responses based on users' multimodal contexts. Existing approaches usually rely on an implicit one-pass generation paradigm from multimodal context to the final response, which overlooks two intrinsic characteristics of MERG: (1) Human perception of emotional cues is inherently structured rather than a direct mapping. The conventional paradigm neglects the hierarchical progression of emotion perception, leading to distorted emotional judgments. (2) Given the inherent complexity and ambiguity of human emotions, the conventional paradigm is prone to significant emotional biases, ultimately resulting in suboptimal empathy. In this paper, we propose a multi-agent framework for MERG, which enhances empathy through structured reasoning and reflective refinement. Specifically, we first introduce a structured empathetic reasoning-to-generation module that explicitly decomposes response generation via multimodal perception, consistency-aware emotion forecasting, pragmatic strategy planning, and strategy-guided response generation, providing a clearer intermediate path from multimodal evidence to response realization. Besides, we develop a global reflection and refinement module, in which a global reflection agent performs step-wise auditing over intermediate states and the generated response, eliminating existing emotional biases and empathy errors, and triggering targeted regeneration. Overall, such a closed-loop framework enables our model to gradually improve the accuracy of emotion perception and eliminate emotion biases during the iteration process. Experiments on several benchmarks, e.g., IEMOCAP and MELD, demonstrate that our model has superior empathic response generation capabilities compared to state-of-the-art methods.", "date": "2026-04-21", "categories": []}
{"id": "2511.12554", "title": "EmoVerse: A MLLMs-Driven Emotion Representation Dataset for Interpretable Visual Emotion Analysis", "abstract": "Visual Emotion Analysis (VEA) aims to bridge the affective gap between visual content and human emotional responses. Despite its promise, progress in this field remains limited by the lack of open-source and interpretable datasets. Most existing studies assign a single discrete emotion label to an entire image, offering limited insight into how visual elements contribute to emotion. In this work, we introduce EmoVerse, a large-scale open-source dataset that enables interpretable visual emotion analysis through multi-layered, knowledge-graph-inspired annotations. By decomposing emotions into Background-Attribute-Subject (B-A-S) triplets and grounding each element to visual regions, EmoVerse provides word-level and subject-level emotional reasoning. With over 219k images, the dataset further includes dual annotations in Categorical Emotion States (CES) and Dimensional Emotion Space (DES), facilitating unified discrete and continuous emotion representation. A novel multi-stage pipeline ensures high annotation reliability with minimal human effort. Finally, we introduce an interpretable model that maps visual cues into DES representations and provides detailed attribution explanations. Together, the dataset, pipeline, and model form a comprehensive foundation for advancing explainable high-level emotion understanding.", "date": "2025-11-16", "categories": []}
{"id": "2602.16161", "title": "Emotion Collider: Dual Hyperbolic Mirror Manifolds for Sentiment Recovery via Anti Emotion Reflection", "abstract": "Emotional expression underpins natural communication and effective human-computer interaction. We present Emotion Collider (EC-Net), a hyperbolic hypergraph framework for multimodal emotion and sentiment modeling. EC-Net represents modality hierarchies using Poincare-ball embeddings and performs fusion through a hypergraph mechanism that passes messages bidirectionally between nodes and hyperedges. To sharpen class separation, contrastive learning is formulated in hyperbolic space with decoupled radial and angular objectives. High-order semantic relations across time steps and modalities are preserved via adaptive hyperedge construction. Empirical results on standard multimodal emotion benchmarks show that EC-Net produces robust, semantically coherent representations and consistently improves accuracy, particularly when modalities are partially available or contaminated by noise. These findings indicate that explicit hierarchical geometry combined with hypergraph fusion is effective for resilient multimodal affect understanding.", "date": "2026-02-18", "categories": []}
{"id": "2604.09162", "title": "Persona-E$^2$: A Human-Grounded Dataset for Personality-Shaped Emotional Responses to Textual Events", "abstract": "Most affective computing research treats emotion as a static property of text, focusing on the writer's sentiment while overlooking the reader's perspective. This approach ignores how individual personalities lead to diverse emotional appraisals of the same event. Although role-playing Large Language Models (LLMs) attempt to simulate such nuanced reactions, they often suffer from \"personality illusion'' -- relying on surface-level stereotypes rather than authentic cognitive logic. A critical bottleneck is the absence of ground-truth human data to link personality traits to emotional shifts. To bridge the gap, we introduce Persona-E$^2$ (Persona-Event2Emotion), a large-scale dataset grounded in annotated MBTI and Big Five traits to capture reader-based emotional variations across news, social media, and life narratives. Extensive experiments reveal that state-of-the-art LLMs struggle to capture precise appraisal shifts, particularly in social media domains. Crucially, we find that personality information significantly improves comprehension, with the Big Five traits alleviating \"personality illusion.'", "date": "2026-04-10", "categories": []}
{"id": "2604.12518", "title": "Enhance-then-Balance Modality Collaboration for Robust Multimodal Sentiment Analysis", "abstract": "Multimodal sentiment analysis (MSA) integrates heterogeneous text, audio, and visual signals to infer human emotions. While recent approaches leverage cross-modal complementarity, they often struggle to fully utilize weaker modalities. In practice, dominant modalities tend to overshadow non-verbal ones, inducing modality competition and limiting overall contributions. This imbalance degrades fusion performance and robustness under noisy or missing modalities. To address this, we propose a novel model, Enhance-then-Balance Modality Collaboration framework (EBMC). EBMC improves representation quality via semantic disentanglement and cross-modal enhancement, strengthening weaker modalities. To prevent dominant modalities from overwhelming others, an Energy-guided Modality Coordination mechanism achieves implicit gradient rebalancing via a differentiable equilibrium objective. Furthermore, Instance-aware Modality Trust Distillation estimates sample-level reliability to adaptively modulate fusion weights, ensuring robustness. Extensive experiments demonstrate that EBMC achieves state-of-the-art or competitive results and maintains strong performance under missing-modality settings.", "date": "2026-04-14", "categories": []}
{"id": "2604.17178", "title": "Cognitive Policy-Driven LLM for Diagnosis and Intervention of Cognitive Distortions in Emotional Support Conversation", "abstract": "Emotional Support Conversation (ESC) plays a critical role in mental health assistance by providing accessible psychological support in real-world applications. Large Language Models (LLMs) have shown strong empathetic abilities in ESC tasks. Yet, existing methods overlook the issue of cognitive distortions in help-seekers' expressions. As a result, current models can only provide basic emotional comfort, rather than helping help-seekers address their psychological distress at a deeper cognitive level. To address this challenge, we construct the CogBiasESC dataset, the first dataset that expands existing ESC datasets by adding labels for cognitive distortions, includes their type, intensity, and safe risk level. Furthermore, we propose the Cognitive Policy-driven Large Language Model framework (CoPoLLM) to enhance LLMs' ability to diagnose and intervene cognitive distortions in help-seekers. We also analyze the safety advantages of CoPoLLM from a theoretical perspective. Experimental results show that CoPoLLM significantly outperforms 15 state-of-the-art baselines in terms of distortion diagnosis accuracy, intervention strategy effectiveness, and safety risk control.", "date": "2026-04-19", "categories": []}
{"id": "2604.18003", "title": "SELF-EMO: Emotional Self-Evolution from Recognition to Consistent Expression", "abstract": "Emotion Recognition in Conversation (ERC) has become a fundamental capability for large language models (LLMs) in human-centric interaction. Beyond accurate recognition, coherent emotional expression is also crucial, yet both are limited by the scarcity and static nature of high-quality annotated data. In this work, we propose SELF-EMO, a self-evolution framework grounded in the hypothesis that better emotion prediction leads to more consistent emotional responses. We introduce two auxiliary tasks, emotional understanding and emotional expression, and design a role-based self-play paradigm where the model acts as both an emotion recognizer and a dialogue responder. Through iterative interactions, the model generates diverse conversational trajectories, enabling scalable data generation. To ensure quality, we adopt a data flywheel mechanism that filters candidate predictions and responses using a smoothed IoU-based reward and feeds selected samples back for continuous self-improvement without external supervision. We further develop SELF-GRPO, a reinforcement learning algorithm that stabilizes optimization with multi-label alignment rewards and group-level consistency signals. Experiments on IEMOCAP, MELD, and EmoryNLP show that SELF-EMO achieves state-of-the-art performance, improving accuracy by +6.33% on Qwen3-4B and +8.54% on Qwen3-8B, demonstrating strong effectiveness and generalization.", "date": "2026-04-20", "categories": []}
{"id": "2604.18354", "title": "PRISMA: Preference-Reinforced Self-Training Approach for Interpretable Emotionally Intelligent Negotiation Dialogues", "abstract": "Emotion plays a pivotal role in shaping negotiation outcomes, influencing trust, cooperation, and long-term relationships. Developing negotiation dialog systems that can recognize and respond strategically to emotions is, therefore, essential to create more effective human-centered interactions. Beyond generating emotionally appropriate responses, interpretability - understanding how a system generates a particular emotion-aware response, is critical for fostering reliability and building rapport. Driven by these aspects, in this work, we introduce PRISMA, an interpretable emotionally intelligent negotiation dialogue system targeting two application domains, viz. job interviews and resource allocation. To enable interpretability, we propose an Emotion-aware Negotiation Strategy-informed Chain-of-Thought (ENS-CoT) reasoning mechanism, which mimics human negotiation by perceiving, understanding, using, and managing emotions. Leveraging ENS-CoT, we curate two new datasets: JobNego (for job interview negotiation) and ResNego (for resource allocation negotiation). We then leverage these datasets to develop PRISMA by augmenting self-training with Direct Preference Optimization (DPO), guiding agents toward more accurate, interpretable, and emotionally appropriate negotiation responses. Automatic and human evaluation on JobNego and ResNego datasets demonstrate that PRISMA substantially enhances interpretability and generates appropriate emotion-aware responses, while improving overall negotiation effectiveness.", "date": "2026-04-20", "categories": []}
{"id": "2604.18356", "title": "ComPASS: Towards Personalized Agentic Social Support via Tool-Augmented Companionship", "abstract": "Developing compassionate interactive systems requires agents to not only understand user emotions but also provide diverse, substantive support. While recent works explore empathetic dialogue generation, they remain limited in response form and content, struggling to satisfy diverse needs across users and contexts. To address this, we explore empowering agents with external tools to execute diverse actions. Grounded in the psychological concept of \"social support\", this paradigm delivers substantive, human-like companionship. Specifically, we first design a dozen user-centric tools simulating various multimedia applications, which can cover different types of social support behaviors in human-agent interaction scenarios. We then construct ComPASS-Bench, the first personalized social support benchmark for LLM-based agents, via multi-step automated synthesis and manual refinement. Based on ComPASS-Bench, we further synthesize tool use records to fine-tune the Qwen3-8B model, yielding a task-specific ComPASS-Qwen. Comprehensive evaluations across two settings reveal that while the evaluated LLMs can generate valid tool-calling requests with high success rates, significant gaps remain in final response quality. Moreover, tool-augmented responses achieve better overall performance than directly producing conversational empathy. Notably, our trained ComPASS-Qwen demonstrates substantial improvements over its base model, achieving comparable performance to several large-scale models. Our code and data are available at https://github.com/hzp3517/ComPASS.", "date": "2026-04-20", "categories": []}
{"id": "2604.18460", "title": "Learning Invariant Modality Representation for Robust Multimodal Learning from a Causal Inference Perspective", "abstract": "Multimodal affective computing aims to predict humans' sentiment, emotion, intention, and opinion using language, acoustic, and visual modalities. However, current models often learn spurious correlations that harm generalization under distribution shifts or noisy modalities. To address this, we propose a causal modality-invariant representation (CmIR) learning framework for robust multimodal learning. At its core, we introduce a theoretically grounded disentanglement method that separates each modality into `causal invariant representation' and `environment-specific spurious representation' from a causal inference perspective. CmIR ensures that the learned invariant representations retain stable predictive relationships with labels across different environments while preserving sufficient information from the raw inputs via invariance constraint, mutual information constraint, and reconstruction constraint. Experiments across multiple multimodal benchmarks demonstrate that CmIR achieves state-of-the-art performance. CmIR particularly excels on out-of-distribution data and noisy data, confirming its robustness and generalizability.", "date": "2026-04-20", "categories": []}
{"id": "2604.15823", "title": "Watching Movies Like a Human: Egocentric Emotion Understanding for Embodied Companions", "abstract": "Embodied robotic agents often perceive movies through an egocentric screen-view interface rather than native cinematic footage, introducing domain shifts such as viewpoint distortion, scale variation, illumination changes, and environmental interference. However, existing research on movie emotion understanding is almost exclusively conducted on cinematic footage, limiting cross-domain generalization to real-world viewing scenarios. To bridge this gap, we introduce EgoScreen-Emotion (ESE), the first benchmark dataset for egocentric screen-view movie emotion understanding. ESE contains 224 movie trailers captured under controlled egocentric screen-view conditions, producing 28,667 temporally aligned key-frames annotated by multiple raters with a confidence-aware multi-label protocol to address emotional ambiguity. We further build a multimodal long-context emotion reasoning framework that models temporal visual evidence, narrative summaries, compressed historical context, and audio cues. Cross-domain experiments reveal a severe domain gap: models trained on cinematic footage drop from 27.99 to 16.69 Macro-F1 when evaluated on realistic egocentric screen-view observations. Training on ESE substantially improves robustness under realistic viewing conditions. Our approach achieves competitive performance compared with strong closed-source multimodal models, highlighting the importance of domain-specific data and long-context multimodal reasoning.", "date": "2026-04-17", "categories": []}
{"id": "2604.16138", "title": "Sentiment Analysis of German Sign Language Fairy Tales", "abstract": "We present a dataset and a model for sentiment analysis of German sign language (DGS) fairy tales. First, we perform sentiment analysis for three levels of valence (negative, neutral, positive) on German fairy tales text segments using four large language models (LLMs) and majority voting, reaching an inter-annotator agreement of 0.781 Krippendorff's alpha. Second, we extract face and body motion features from each corresponding DGS video segment using MediaPipe. Finally, we train an explainable model (based on XGBoost) to predict negative, neutral or positive sentiment from video features. Results show an average balanced accuracy of 0.631. A thorough analysis of the most important features reveal that, in addition to eyebrows and mouth motion on the face, also the motion of hips, elbows, and shoulders considerably contribute in the discrimination of the conveyed sentiment, indicating an equal importance of face and body for sentiment communication in sign language.", "date": "2026-04-17", "categories": []}
{"id": "2604.16214", "title": "GAViD: A Large-Scale Multimodal Dataset for Context-Aware Group Affect Recognition from Videos", "abstract": "Understanding affective dynamics in real-world social systems is fundamental to modeling and analyzing human-human interactions in complex environments. Group affect emerges from intertwined human-human interactions, contextual influences, and behavioral cues, making its quantitative modeling a challenging computational social systems problem. However, computational modeling of group affect in in-the-wild scenarios remains challenging due to limited large-scale annotated datasets and the inherent complexity of multimodal social interactions shaped by contextual and behavioral variability. The lack of comprehensive datasets annotated with multimodal and contextual information further limits advances in the field. To address this, we introduce the Group Affect from ViDeos (GAViD) dataset, comprising 5091 video clips with multimodal data (video, audio and context), annotated with ternary valence and discrete emotion labels and enriched with VideoGPT-generated contextual metadata and human-annotated action cues. We also present Context-Aware Group Affect Recognition Network (CAGNet) for multimodal context-aware group affect recognition. CAGNet achieves 63.20\\% test accuracy on GAViD, comparable to state-of-the-art performance. The dataset and code are available at github.com/deepakkumar-iitr/GAViD.", "date": "2026-04-17", "categories": []}
{"id": "2604.14204", "title": "Disentangled Dual-Branch Graph Learning for Conversational Emotion Recognition", "abstract": "Multimodal emotion recognition in conversations aims to infer utterance-level emotions by jointly modeling textual, acoustic, and visual cues within context. Despite recent progress, key challenges remain, including redundant cross-modal information, imperfect semantic alignment, and insufficient modeling of high-order speaker interactions. To address these issues, we propose a framework that combines dual-space feature disentanglement with dual-branch graph learning. A shared encoder and modality-specific encoders are used to separate modality-invariant and modality-specific representations. The invariant features are modeled by a Fourier graph neural network to capture global consistency and complementary patterns, with a frequency-domain contrastive objective to enhance discriminability. In parallel, a speaker-aware hypergraph is constructed over modality-specific features to model high-order interactions, along with a speaker-consistency constraint to maintain coherent semantics. Finally, the two branches are fused for utterance-level emotion prediction. Experiments on IEMOCAP and MELD demonstrate that the proposed method achieves superior performance over strong baselines, validating its effectiveness.", "date": "2026-04-03", "categories": []}
{"id": "2601.13904", "title": "PREFAB: PREFerence-based Affective Modeling for Low-Budget Self-Annotation", "abstract": "Self-annotation is the gold standard for collecting affective state labels in affective computing. Existing methods typically rely on full annotation, requiring users to continuously label affective states across entire sessions. While this process yields fine-grained data, it is time-consuming, cognitively demanding, and prone to fatigue and errors. To address these issues, we present PREFAB, a low-budget retrospective self-annotation method that targets affective inflection regions rather than full annotation. Grounded in the peak-end rule and ordinal representations of emotion, PREFAB employs a preference-learning model to detect relative affective changes, directing annotators to label only selected segments while interpolating the remainder of the stimulus. We further introduce a preview mechanism that provides brief contextual cues to assist annotation. We evaluate PREFAB through a technical performance study and a 25-participant user study. Results show that PREFAB outperforms baselines in modeling affective inflections while mitigating workload (and conditionally mitigating temporal burden). Importantly PREFAB improves annotator confidence without degrading annotation quality.", "date": "2026-01-20", "categories": []}
{"id": "2505.19328", "title": "BAH Dataset for Ambivalence/Hesitancy Recognition in Videos for Digital Behavioural Change", "abstract": "Ambivalence and hesitancy (A/H), closely related constructs, are the primary reasons why individuals delay, avoid, or abandon health behaviour changes. They are subtle and conflicting emotions that sets a person in a state between positive and negative orientations, or between acceptance and refusal to do something. They manifest as a discord in affect between multiple modalities or within a modality, such as facial and vocal expressions, and body language. Although experts can be trained to recognize A/H as done for in-person interactions, integrating them into digital health interventions is costly and less effective. Automatic A/H recognition is therefore critical for the personalization and cost-effectiveness of digital behaviour change interventions. However, no datasets currently exist for the design of machine learning models to recognize A/H. This paper introduces the Behavioural Ambivalence/Hesitancy (BAH) dataset collected for multimodal recognition of A/H in videos. It contains 1,427 videos with a total duration of 10.60 hours, captured from 300 participants across Canada, answering predefined questions to elicit A/H. It is intended to mirror real-world digital behaviour change interventions delivered online. BAH is annotated by three experts to provide timestamps that indicate where A/H occurs, and frame- and video-level annotations with A/H cues. Video transcripts, cropped and aligned faces, and participant metadata are also provided. Since A and H manifest similarly in practice, we provide a binary annotation indicating the presence or absence of A/H. Additionally, this paper includes benchmarking results using baseline models on BAH for frame- and video-level recognition, and different learning setups. The limited performance highlights the need for adapted multimodal and spatio-temporal models for A/H recognition. The data and code are publicly available.", "date": "2025-05-25", "categories": []}
{"id": "2604.10815", "title": "MeloTune: On-Device Arousal Learning and Peer-to-Peer Mood Coupling for Proactive Music Curation", "abstract": "MeloTune is an iPhone-deployed music agent that instantiates the Mesh Memory Protocol (MMP) and Symbolic-Vector Attention Fusion (SVAF) as a production system for affect-aware music curation with peer-to-peer mood coupling. Each device runs two closed-form continuous-time (CfC) networks: a private listener-level CfC that predicts a short-horizon affective trajectory on Russell's circumplex and drives proactive curation, and a shared mesh-runtime CfC at MMP Layer 6 that integrates Cognitive Memory Blocks (CMBs) from co-listening peers. CfC hidden states never cross the wire; only structured CMBs do. A Personal Arousal Function (PAF) replaces the standard linear mapping from audio intensity to psychological arousal with a per-listener learned adjustment, trained from behavioral signals (skip, completion, favorite, volume) and from drift between user-declared mood and machine inference. The same track receives different arousal predictions for different listeners. The model (94,552 parameters) achieves trajectory MAE 0.414, pattern accuracy 96.6%, and intent accuracy 69.4% on held-out validation. PAF evidence from a live deployment session (46 observations across 11 genres) demonstrates that the learning loop operates end-to-end, with pop reaching full confidence after 22 observations. All inference runs on-device via CoreML. To our knowledge, this is the first production deployment of MMP/SVAF on consumer mobile hardware. The accompanying SDK (sym-swift v0.3.78, SYMCore v0.3.7) enforces strict protocol conformance. Music is the case study; the substrate is the contribution.", "date": "2026-04-12", "categories": []}
{"id": "2604.12255", "title": "ARGen: Affect-Reinforced Generative Augmentation towards Vision-based Dynamic Emotion Perception", "abstract": "Dynamic facial expression recognition in the wild remains challenging due to data scarcity and long-tail distributions, which hinder models from effectively learning the temporal dynamics of scarce emotions. To address these limitations, we propose ARGen, an Affect-Reinforced Generative Augmentation Framework that enables data-adaptive dynamic expression generation for robust emotion perception. ARGen operates in two stages: Affective Semantic Injection (ASI) and Adaptive Reinforcement Diffusion (ARD). The ASI stage establishes affective knowledge alignment through facial Action Units and employs a retrieval-augmented prompt generation strategy to synthesize consistent and fine-grained affective descriptions via large-scale visual-language models, thereby injecting interpretable emotional priors into the generation process. The ARD stage integrates text-conditioned image-to-video diffusion with reinforcement learning, introducing inter-frame conditional guidance and a multi-objective reward function to jointly optimize expression naturalness, facial integrity, and generative efficiency. Extensive experiments on both generation and recognition tasks verify that ARGen substantially enhances synthesis fidelity and improves recognition performance, establishing an interpretable and generalizable generative augmentation paradigm for vision-based affective computing.", "date": "2026-04-14", "categories": []}
{"id": "2604.12633", "title": "Multilingual Multi-Label Emotion Classification at Scale with Synthetic Data", "abstract": "Emotion classification in multilingual settings remains constrained by the scarcity of annotated data: existing corpora are predominantly English, single-label, and cover few languages. We address this gap by constructing a large-scale synthetic training corpus of over 1M multi-label samples (50k per language) across 23 languages: Arabic, Bengali, Dutch, English, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Mandarin, Polish, Portuguese, Punjabi, Russian, Spanish, Swahili, Tamil, Turkish, Ukrainian, Urdu, and Vietnamese, covering 11 emotion categories using culturally-adapted generation and programmatic quality filtering. We train and compare six multilingual transformer encoders, from DistilBERT (135M parameters) to XLM-R-Large (560M parameters), under identical conditions. On our in-domain test set, XLM-R-Large achieves 0.868 F1-micro and 0.987 AUC-micro. To validate against human-annotated data, we evaluate all models zero-shot on GoEmotions (English) and SemEval-2018 Task 1 E-c (English, Arabic, Spanish). On threshold-free ranking metrics, XLM-R-Large matches or exceeds English-only specialist models, tying on AP-micro (0.636) and LRAP (0.804) while surpassing on AUC-micro (0.810 vs. 0.787), while natively supporting all 23 languages. The best base-sized model is publicly available at https://huggingface.co/tabularisai/multilingual-emotion-classification", "date": "2026-04-14", "categories": []}
{"id": "2604.12735", "title": "AffectAgent: Collaborative Multi-Agent Reasoning for Retrieval-Augmented Multimodal Emotion Recognition", "abstract": "LLM-based multimodal emotion recognition relies on static parametric memory and often hallucinates when interpreting nuanced affective states. In this paper, given that single-round retrieval-augmented generation is highly susceptible to modal ambiguity and therefore struggles to capture complex affective dependencies across modalities, we introduce AffectAgent, an affect-oriented multi-agent retrieval-augmented generation framework that leverages collaborative decision-making among agents for fine-grained affective understanding. Specifically, AffectAgent comprises three jointly optimized specialized agents, namely a query planner, an evidence filter, and an emotion generator, which collaboratively perform analytical reasoning to retrieve cross-modal samples, assess evidence, and generate predictions. These agents are optimized end-to-end using Multi-Agent Proximal Policy Optimization (MAPPO) with a shared affective reward to ensure consistent emotion understanding. Furthermore, we introduce Modality-Balancing Mixture of Experts (MB-MoE) and Retrieval-Augmented Adaptive Fusion (RAAF), where MB-MoE dynamically regulates the contributions of different modalities to mitigate representation mismatch caused by cross-modal heterogeneity, while RAAF enhances semantic completion under missing-modality conditions by incorporating retrieved audiovisual embeddings. Extensive experiments on MER-UniBench demonstrate that AffectAgent achieves superior performance across complex scenarios. Our code will be released at: https://github.com/Wz1h1NG/AffectAgent.", "date": "2026-04-14", "categories": []}
{"id": "2604.12746", "title": "Stress Detection Using Wearable Physiological and Sociometric Sensors", "abstract": "Stress remains a significant social problem for individuals in modern societies. This paper presents a machine learning approach for the automatic detection of stress of people in a social situation by combining two sensor systems that capture physiological and social responses. We compare the performance using different classifiers including support vector machine, AdaBoost, and k-nearest neighbor. Our experimental results show that by combining the measurements from both sensor systems, we could accurately discriminate between stressful and neutral situations during a controlled Trier social stress test (TSST). Moreover, this paper assesses the discriminative ability of each sensor modality individually and considers their suitability for real-time stress detection. Finally, we present an study of the most discriminative features for stress detection.", "date": "2026-04-14", "categories": []}
{"id": "2407.20799", "title": "SpotFormer: Multi-Scale Spatio-Temporal Transformer for Facial Expression Spotting", "abstract": "Facial expression spotting, identifying periods where facial expressions occur in a video, is a significant yet challenging task in facial expression analysis. The issues of irrelevant facial movements and the challenge of detecting subtle motions in micro-expressions remain unresolved, hindering accurate expression spotting. In this paper, we propose an efficient framework for facial expression spotting. First, we propose a Compact Sliding-Window-based Multi-temporal-Resolution Optical flow (CSW-MRO) feature, which calculates multi-temporal-resolution optical flow of the input image sequence within compact sliding windows. The window length is tailored to perceive complete micro-expressions and distinguish between general macro- and micro-expressions. CSW-MRO can effectively reveal subtle motions while avoiding the optical flow being dominated by head movements. Second, we propose SpotFormer, a multi-scale spatio-temporal Transformer that simultaneously encodes spatio-temporal relationships of the CSW-MRO features for accurate frame-level probability estimation. In SpotFormer, we use the proposed Facial Local Graph Pooling (FLGP) operation and convolutional layers to extract multi-scale spatio-temporal features. We show the validity of the architecture of SpotFormer by comparing it with several model variants. Third, we introduce supervised contrastive learning into SpotFormer to enhance the discriminability between different types of expressions. Extensive experiments on SAMM-LV, CAS(ME)^2, and CAS(ME)^3 show that our method outperforms state-of-the-art models, particularly in micro-expression spotting. Code is available at https://github.com/KinopioIsAllIn/SpotFormer.", "date": "2024-07-30", "categories": []}
{"id": "2511.07061", "title": "Do LLMs Feel? Teaching Emotion Recognition with Prompts, Retrieval, and Curriculum Learning", "abstract": "Emotion Recognition in Conversation (ERC) is a crucial task for understanding human emotions and enabling natural human-computer interaction. Although Large Language Models (LLMs) have recently shown great potential in this field, their ability to capture the intrinsic connections between explicit and implicit emotions remains limited. We propose a novel ERC training framework, PRC-Emo, which integrates Prompt engineering, demonstration Retrieval, and Curriculum learning, with the goal of exploring whether LLMs can effectively perceive emotions in conversational contexts. Specifically, we design emotion-sensitive prompt templates based on both explicit and implicit emotional cues to better guide the model in understanding the speaker's psychological states. We construct the first dedicated demonstration retrieval repository for ERC, which includes training samples from widely used datasets, as well as high-quality dialogue examples generated by LLMs and manually verified. Moreover, we introduce a curriculum learning strategy into the LoRA fine-tuning process, incorporating weighted emotional shifts between same-speaker and different-speaker utterances to assign difficulty levels to dialogue samples, which are then organized in an easy-to-hard training sequence. Experimental results on two benchmark datasets -- IEMOCAP and MELD -- show that our method achieves new state-of-the-art (SOTA) performance, demonstrating the effectiveness and generalizability of our approach in improving LLM-based emotional understanding.", "date": "2025-11-10", "categories": []}
{"id": "2603.02123", "title": "Nano-EmoX: Unifying Multimodal Emotional Intelligence from Perception to Empathy", "abstract": "The development of affective multimodal language models (MLMs) has long been constrained by a gap between low-level perception and high-level interaction, leading to fragmented affective capabilities and limited generalization. To bridge this gap, we propose a cognitively inspired three-level hierarchy that organizes affective tasks according to their cognitive depth-perception, understanding, and interaction-and provides a unified conceptual foundation for advancing affective modeling. Guided by this hierarchy, we introduce Nano-EmoX, a small-scale multitask MLM, and P2E (Perception-to-Empathy), a curriculum-based training framework. Nano-EmoX integrates a suite of omni-modal encoders, including an enhanced facial encoder and a fusion encoder, to capture key multimodal affective cues and improve cross-task transferability. The outputs are projected into a unified language space via heterogeneous adapters, empowering a lightweight language model to tackle diverse affective tasks. Concurrently, P2E progressively cultivates emotional intelligence by aligning rapid perception with chain-of-thought-driven empathy. To the best of our knowledge, Nano-EmoX is the first compact MLM (2.2B) to unify six core affective tasks across all three hierarchy levels, achieving state-of-the-art or highly competitive performance across multiple benchmarks, demonstrating excellent efficiency and generalization. The code is available at https://github.com/waHAHJIAHAO/Nano-EmoX.", "date": "2026-03-02", "categories": []}
{"id": "2603.12221", "title": "A Two-Stage Dual-Modality Model for Facial Emotional Expression Recognition", "abstract": "This paper addresses the expression (EXPR) recognition challenge in the 10th Affective Behavior Analysis in-the-Wild (ABAW) workshop and competition, which requires frame-level classification of eight facial emotional expressions from unconstrained videos. This task is challenging due to inaccurate face localization, large pose and scale variations, motion blur, temporal instability, and other confounding factors across adjacent frames. We propose a two-stage dual-modal (audio-visual) model to address these difficulties. Stage I focuses on robust visual feature extraction with a pretrained DINOv2-based encoder. Specifically, DINOv2 ViT-L/14 is used as the backbone, a padding-aware augmentation (PadAug) strategy is employed for image padding and data preprocessing from raw videos, and a mixture-of-experts (MoE) training head is introduced to enhance classifier diversity. Stage II addresses modality fusion and temporal consistency. For the visual modality, faces are re-cropped from raw videos at multiple scales, and the extracted visual features are averaged to form a robust frame-level representation. Concurrently, frame-aligned Wav2Vec 2.0 audio features are derived from short audio windows to provide complementary acoustic cues. These dual-modal features are integrated via a lightweight gated fusion module, followed by inference-time temporal smoothing. Experiments on the ABAW dataset demonstrate the effectiveness of the proposed method. The two-stage model achieves a Macro-F1 score of 0.5368 on the official validation set and 0.5122 +/- 0.0277 under 5-fold cross-validation, outperforming the official baselines.", "date": "2026-03-12", "categories": []}
{"id": "2604.10344", "title": "Context Matters: Vision-Based Depression Detection Comparing Classical and Deep Approaches", "abstract": "The classical approach to detecting depression from vision emphasizes interpretable features, such as facial expression, and classifiers such as the Support Vector Machine (SVM). With the advent of deep learning, there has been a shift in feature representations and classification approaches. Contemporary approaches use learnt features from general-purpose vision models such as VGGNet to train machine learning models. Little is known about how classical and deep approaches compare in depression detection with respect to accuracy, fairness, and generalizability, especially across contexts. To address these questions, we compared classical and deep approaches to the detection of depression in the visual modality in two different contexts: Mother-child interactions in the TPOT database and patient-clinician interviews in the Pitt database. In the former, depression was operationalized as a history of depression per the DSM and current or recent clinically significant symptoms. In the latter, all participants met initial criteria for depression per DSM, and depression was reassessed over the course of treatment. The classical approach included handcrafted features with SVM classifiers. Learnt features were turn-level embeddings from the FMAE-IAT that were combined with Multi-Layer Perceptron classifiers. The classical approach achieved higher accuracy in both contexts. It was also significantly fairer than the deep approach in the patient-clinician context. Cross-context generalizability was modest at best for both approaches, which suggests that depression may be context-specific.", "date": "2026-04-11", "categories": []}
{"id": "2604.10454", "title": "AIM-Bench: Benchmarking and Improving Affective Image Manipulation via Fine-Grained Hierarchical Control", "abstract": "Affective Image Manipulation (AIM) aims to evoke specific emotions through targeted editing. Current image editing benchmarks primarily focus on object-level modifications in general scenarios, lacking the fine-grained granularity to capture affective dimensions. To bridge this gap, we introduce the first benchmark designed for AIM termed AIM-Bench. This benchmark is built upon a dual-path affective modeling scheme that integrates the Mikels emotion taxonomy with the Valence-Arousal-Dominance framework, enabling high-level semantic and fine-grained continuous manipulation. Through a hierarchical human-in-the-loop workflow, we finally curate 800 high-quality samples covering 8 emotional categories and 5 editing types. To effectively assess performance, we also design a composite evaluation suite combining rule-based and model-based metrics to holistically assess instruction consistency, aesthetics, and emotional expressiveness. Extensive evaluations reveal that current editing models face significant challenges, most notably a prevalent positivity bias, which stemming from inherent imbalances in training data distribution. To tackle this, we propose a scalable data engine utilizing an inverse repainting strategy to construct AIM-40k, a balanced instruction-tuning dataset comprising 40k samples. Concretely, we enhance raw affective images via generative redrawing to establish high-fidelity ground truths, and synthesize input images with divergent emotions and paired precise instructions. Fine-tuning a baseline model on AIM-40k yields a 9.15% relative improvement in overall performance, demonstrating the effectiveness of our AIM-40k. Our data and related code will be made open soon.", "date": "2026-04-12", "categories": []}
{"id": "2512.22437", "title": "EmoCtrl: Controllable Emotional Image Content Generation", "abstract": "An image conveys meaning through both its visual content and emotional tone, jointly shaping human perception. We introduce Controllable Emotional Image Content Generation (C-EICG), which aims to generate images that remain faithful to a given content description while expressing a target emotion. Existing text-to-image models ensure content consistency but lack emotional awareness, whereas emotion-driven models generate affective results at the cost of content distortion. To address this gap, we propose EmoCtrl, supported by a dataset annotated with content, emotion, and affective prompts, bridging abstract emotions to visual cues. EmoCtrl incorporates textual and visual emotion enhancement modules that enrich affective expression via descriptive semantics and perceptual cues. To align with human preference, we further introduce an emotion-driven preference optimization with specifically designed emotion reward. Comprehensive experiments demonstrate that EmoCtrl achieves faithful content and expressive emotion control, outperforming existing methods. User studies confirm EmoCtrl's strong alignment with human preference. Moreover, EmoCtrl generalizes well to creative applications, further demonstrating the robustness and adaptability of the learned emotion tokens.", "date": "2025-12-27", "categories": []}
{"id": "2604.08990", "title": "ActFER: Agentic Facial Expression Recognition via Active Tool-Augmented Visual Reasoning", "abstract": "Recent advances in Multimodal Large Language Models (MLLMs) have created new opportunities for facial expression recognition (FER), moving it beyond pure label prediction toward reasoning-based affect understanding. However, existing MLLM-based FER methods still follow a passive paradigm: they rely on externally prepared facial inputs and perform single-pass reasoning over fixed visual evidence, without the capability for active facial perception. To address this limitation, we propose ActFER, an agentic framework that reformulates FER as active visual evidence acquisition followed by multimodal reasoning. Specifically, ActFER dynamically invokes tools for face detection and alignment, selectively zooms into informative local regions, and reasons over facial Action Units (AUs) and emotions through a visual Chain-of-Thought. To realize such behavior, we further develop Utility-Calibrated GRPO (UC-GRPO), a reinforcement learning algorithm tailored to agentic FER. UC-GRPO uses AU-grounded multi-level verifiable rewards to densify supervision, query-conditional contrastive utility estimation to enable sample-aware dynamic credit assignment for local inspection, and emotion-aware EMA calibration to reduce noisy utility estimates while capturing emotion-wise inspection tendencies. This algorithm enables ActFER to learn both when local inspection is beneficial and how to reason over the acquired evidence. Comprehensive experiments show that ActFER trained with UC-GRPO consistently outperforms passive MLLM-based FER baselines and substantially improves AU prediction accuracy.", "date": "2026-04-10", "categories": []}
{"id": "2604.07357", "title": "Hybrid CNN-Transformer Architecture for Arabic Speech Emotion Recognition", "abstract": "Recognizing emotions from speech using machine learning has become an active research area due to its importance in building human-centered applications. However, while many studies have been conducted in English, German, and other European and Asian languages, research in Arabic remains scarce because of the limited availability of annotated datasets. In this paper, we present an Arabic Speech Emotion Recognition (SER) system based on a hybrid CNN-Transformer architecture. The model leverages convolutional layers to extract discriminative spectral features from Mel-spectrogram inputs and Transformer encoders to capture long-range temporal dependencies in speech. Experiments were conducted on the EYASE (Egyptian Arabic speech emotion) corpus, and the proposed model achieved 97.8% accuracy and a macro F1-score of 0.98. These results demonstrate the effectiveness of combining convolutional feature extraction with attention-based modeling for Arabic SER and highlight the potential of Transformer-based approaches in low-resource languages.", "date": "2026-03-28", "categories": []}
{"id": "2604.05704", "title": "QA-MoE: Towards a Continuous Reliability Spectrum with Quality-Aware Mixture of Experts for Robust Multimodal Sentiment Analysis", "abstract": "Multimodal Sentiment Analysis (MSA) aims to infer human sentiment from textual, acoustic, and visual signals. In real-world scenarios, however, multimodal inputs are often compromised by dynamic noise or modality missingness. Existing methods typically treat these imperfections as discrete cases or assume fixed corruption ratios, which limits their adaptability to continuously varying reliability conditions. To address this, we first introduce a Continuous Reliability Spectrum to unify missingness and quality degradation into a single framework. Building on this, we propose QA-MoE, a Quality-Aware Mixture-of-Experts framework that quantifies modality reliability via self-supervised aleatoric uncertainty. This mechanism explicitly guides expert routing, enabling the model to suppress error propagation from unreliable signals while preserving task-relevant information. Extensive experiments indicate that QA-MoE achieves competitive or state-of-the-art performance across diverse degradation scenarios and exhibits a promising One-Checkpoint-for-All property in practice.", "date": "2026-04-07", "categories": []}
{"id": "2604.06752", "title": "Busemann energy-based attention for emotion analysis in Poincar\\'e discs", "abstract": "We present EmBolic - a novel fully hyperbolic deep learning architecture for fine-grained emotion analysis from textual messages. The underlying idea is that hyperbolic geometry efficiently captures hierarchies between both words and emotions. In our context, these hierarchical relationships arise from semantic ambiguities. EmBolic aims to infer the curvature on the continuous space of emotions, rather than treating them as a categorical set without any metric structure. In the heart of our architecture is the attention mechanism in the hyperbolic disc. The model is trained to generate queries (points in the hyperbolic disc) from textual messages, while keys (points at the boundary) emerge automatically from the generated queries. Predictions are based on the Busemann energy between queries and keys, evaluating how well a certain textual message aligns with the class directions representing emotions. Our experiments demonstrate strong generalization properties and reasonably good prediction accuracy even for small dimensions of the representation space. Overall, this study supports our claim that affective computing is one of the application domains where hyperbolic representations are particularly advantageous.", "date": "2026-04-08", "categories": []}
{"id": "2604.07017", "title": "A-MBER: Affective Memory Benchmark for Emotion Recognition", "abstract": "AI assistants that interact with users over time need to interpret the user's current emotional state in order to respond appropriately and personally. However, this capability remains insufficiently evaluated. Existing emotion datasets mainly assess local or instantaneous affect, while long-term memory benchmarks focus largely on factual recall, temporal consistency, or knowledge updating. As a result, current resources provide limited support for testing whether a model can use remembered interaction history to interpret a user's present affective state.\n  We introduce A-MBER, an Affective Memory Benchmark for Emotion Recognition, to evaluate this capability. A-MBER focuses on present affective interpretation grounded in remembered multi-session interaction history. Given an interaction trajectory and a designated anchor turn, a model must infer the user's current affective state, identify historically relevant evidence, and justify its interpretation in a grounded way. The benchmark is constructed through a staged pipeline with explicit intermediate representations, including long-horizon planning, conversation generation, annotation, question construction, and final packaging. It supports judgment, retrieval, and explanation tasks, together with robustness settings such as modality degradation and insufficient-evidence conditions.\n  Experiments compare local-context, long-context, retrieved-memory, structured-memory, and gold-evidence conditions within a unified framework. Results show that A-MBER is especially discriminative on the subsets it is designed to stress, including long-range implicit affect, high-dependency memory levels, trajectory-based reasoning, and adversarial settings. These findings suggest that memory supports affective interpretation not simply by providing more history, but by enabling more selective, grounded, and context-sensitive use of past interaction", "date": "2026-04-08", "categories": []}
{"id": "2604.07193", "title": "LaScA: Language-Conditioned Scalable Modelling of Affective Dynamics", "abstract": "Predicting affect in unconstrained environments remains a fundamental challenge in human-centered AI. While deep neural embeddings dominate contemporary approaches, they often lack interpretability and limit expert-driven refinement. We propose a novel framework that uses Language Models (LMs) as semantic context conditioners over handcrafted affect descriptors to model changes in Valence and Arousal. Our approach begins with interpretable facial geometry and acoustic features derived from structured domain knowledge. These features are transformed into symbolic natural-language descriptions encoding their affective implications. A pretrained LM processes these descriptions to generate semantic context embeddings that act as high-level priors over affective dynamics. Unlike end-to-end black-box pipelines, our framework preserves feature transparency while leveraging the contextual abstraction capabilities of LMs. We evaluate the proposed method on the Aff-Wild2 and SEWA datasets for affect change prediction. Experimental results show consistent improvements in accuracy for both Valence and Arousal compared to handcrafted-only and deep-embedding baselines. Our findings demonstrate that semantic conditioning enables interpretable affect modelling without sacrificing predictive performance, offering a transparent and computationally efficient alternative to fully end-to-end architectures", "date": "2026-04-08", "categories": []}
{"id": "2604.07198", "title": "Beyond the Mean: Modelling Annotation Distributions in Continuous Affect Prediction", "abstract": "Emotion annotation is inherently subjective and cognitively demanding, producing signals that reflect diverse perceptions across annotators rather than a single ground truth. In continuous affect prediction, this variability is typically collapsed into point estimates such as the mean or median, discarding valuable information about annotator disagreement and uncertainty. In this work, we propose a distribution-aware framework that models annotation consensus using the Beta distribution. Instead of predicting a single affect value, models estimate the mean and standard deviation of the annotation distribution, which are transformed into valid Beta parameters through moment matching. This formulation enables the recovery of higher-order distributional descriptors, including skewness, kurtosis, and quantiles, in closed form. As a result, the model captures not only the central tendency of emotional perception but also variability, asymmetry, and uncertainty in annotator responses. We evaluate the proposed approach on the SEWA and RECOLA datasets using multimodal features. Experimental results show that Beta-based modelling produces predictive distributions that closely match the empirical annotator distributions while achieving competitive performance with conventional regression approaches. These findings highlight the importance of modelling annotation uncertainty in affective computing and demonstrate the potential of distribution-aware learning for subjective signal analysis.", "date": "2026-04-08", "categories": []}
{"id": "2409.07388", "title": "Recent Advances in Multimodal Affective Computing: An NLP Perspective", "abstract": "Multimodal affective computing has gained increasing attention due to its broad applications in understanding human behavior and intentions, particularly in text-centric multimodal scenarios. Existing research spans diverse tasks, modalities, and modeling paradigms, yet lacks a unified perspective. In this survey, we systematically review recent advances from an NLP perspective, focusing on four representative tasks: multimodal sentiment analysis (MSA), multimodal emotion recognition in conversation (MERC), multimodal aspect-based sentiment analysis (MABSA), and multimodal multi-label emotion recognition (MMER). We present a unified view by comparing task formulations, benchmark datasets, and evaluation protocols, and by organizing representative methods into key paradigms, including multitask learning, pre-trained models, knowledge enhancement, and contextual modeling. We further extend the discussion to related directions, such as facial, acoustic, and physiological modalities, as well as emotion cause analysis. Finally, we highlight key challenges and outline promising future directions. To facilitate further research, we release a curated repository of relevant works and resources \\footnote{https://anonymous.4open.science/r/Multimodal-Affective-Computing-Survey-9819}.", "date": "2024-09-11", "categories": []}
{"id": "2604.05558", "title": "Evaluation Before Generation: A Paradigm for Robust Multimodal Sentiment Analysis with Missing Modalities", "abstract": "The missing modality problem poses a fundamental challenge in multimodal sentiment analysis, significantly degrading model accuracy and generalization in real world scenarios. Existing approaches primarily improve robustness through prompt learning and pre trained models. However, two limitations remain. First, the necessity of generating missing modalities lacks rigorous evaluation. Second, the structural dependencies among multimodal prompts and their global coherence are insufficiently explored. To address these issues, a Prompt based Missing Modality Adaptation framework is proposed. A Missing Modality Evaluator is introduced at the input stage to dynamically assess the importance of missing modalities using pretrained models and pseudo labels, thereby avoiding low quality data imputation. Building on this, a Modality invariant Prompt Disentanglement module decomposes shared prompts into modality specific private prompts to capture intrinsic local correlations and improve representation quality. In addition, a Dynamic Prompt Weighting module computes mutual information based weights from cross attention outputs to adaptively suppress interference from missing modalities. To enhance global consistency, a Multi level Prompt Dynamic Connection module integrates shared prompts with self attention outputs through residual connections, leveraging global prompt priors to strengthen key guidance features. Extensive experiments on three public benchmarks, including CMU MOSI, CMU MOSEI, and CH SIMS, demonstrate that the proposed framework achieves state of the art performance and stable results under diverse missing modality settings. The implementation is available at https://github.com/rongfei-chen/ProMMA", "date": "2026-04-07", "categories": []}
{"id": "2604.05900", "title": "AICA-Bench: Holistically Examining the Capabilities of VLMs in Affective Image Content Analysis", "abstract": "Vision-Language Models (VLMs) have demonstrated strong capabilities in perception, yet holistic Affective Image Content Analysis (AICA), which integrates perception, reasoning, and generation into a unified framework, remains underexplored. To address this gap, we introduce AICA-Bench, a comprehensive benchmark with three core tasks: Emotion Understanding (EU), Emotion Reasoning (ER), and Emotion-Guided Content Generation (EGCG). We evaluate 23 VLMs and identify two major limitations: weak intensity calibration and shallow open-ended descriptions. To address these issues, we propose Grounded Affective Tree (GAT) Prompting, a training-free framework that combines visual scaffolding with hierarchical reasoning. Experiments show that GAT reduces intensity errors and improves descriptive depth, providing a strong baseline for future research on affective multimodal understanding and generation.", "date": "2026-04-07", "categories": []}
{"id": "2512.10882", "title": "Computational emotion analysis with multimodal LLMs: Current evidence on an emerging methodological opportunity", "abstract": "Research increasingly leverages audio-visual materials to analyze emotions in political communication. Multimodal large language models (mLLMs) promise to enable such analyses through in-context learning. However, we lack systematic evidence on whether current mLLMs can reliably measure emotions in real-world political settings. This paper closes this gap by evaluating open- and closed-weights mLLMs available as of early 2026 in video-based emotional arousal measurement using two complementary human-labeled datasets: speech actor recordings created under laboratory conditions and real-world parliamentary debates. I find a critical lab-vs-field performance gap. In videos created under laboratory conditions, the examined mLLMs arousal scores approach human-level reliability. However, in parliamentary debate recordings, all examined models' arousal scores correlate at best moderately with average human ratings. Moreover, in each dataset, all but one of the examined mLLMs exhibit systematic gender-differential bias, consistently underestimating arousal more for male than for female speakers, resulting in a net-positive intensity bias. These findings reveal important limitations of current mLLMs for real-world political video analysis and establish a rigorous evaluation framework for tracking future developments.", "date": "2025-12-11", "categories": []}
{"id": "2604.03650", "title": "CAGMamba: Context-Aware Gated Cross-Modal Mamba Network for Multimodal Sentiment Analysis", "abstract": "Multimodal Sentiment Analysis (MSA) requires effective modeling of cross-modal interactions and contextual dependencies while remaining computationally efficient. Existing fusion approaches predominantly rely on Transformer-based cross-modal attention, which incurs quadratic complexity with respect to sequence length and limits scalability. Moreover, contextual information from preceding utterances is often incorporated through concatenation or independent fusion, without explicit temporal modeling that captures sentiment evolution across dialogue turns. To address these limitations, we propose CAGMamba, a context-aware gated cross-modal Mamba framework for dialogue-based sentiment analysis. Specifically, we organize the contextual and the current-utterance features into a temporally ordered binary sequence, which provides Mamba with explicit temporal structure for modeling sentiment evolution. To further enable controllable cross-modal integration, we propose a Gated Cross-Modal Mamba Network (GCMN) that integrates cross-modal and unimodal paths via learnable gating to balance information fusion and modality preservation, and is trained with a three-branch multi-task objective over text, audio, and fused predictions. Experiments on three benchmark datasets demonstrate that CAGMamba achieves state-of-the-art or competitive results across multiple evaluation metrics. All codes are available at https://github.com/User2024-xj/CAGMamba.", "date": "2026-04-04", "categories": []}
{"id": "2604.02397", "title": "Variational Encoder--Multi-Decoder (VE-MD) for Privacy-by-functional-design (Group) Emotion Recognition", "abstract": "Group Emotion Recognition (GER) aims to infer collective affect in social environments such as classrooms, crowds, and public events. Many existing approaches rely on explicit individual-level processing, including cropped faces, person tracking, or per-person feature extraction, which makes the analysis pipeline person-centric and raises privacy concerns in deployment scenarios where only group-level understanding is needed. This research proposes VE-MD, a Variational Encoder-Multi-Decoder framework for group emotion recognition under a privacy-aware functional design. Rather than providing formal anonymization or cryptographic privacy guarantees, VE-MD is designed to avoid explicit individual monitoring by constraining the model to predict only aggregate group-level affect, without identity recognition or per-person emotion outputs. VE-MD learns a shared latent representation jointly optimized for emotion classification and internal prediction of body and facial structural representations. Two structural decoding strategies are investigated: a transformer-based PersonQuery decoder and a dense Heatmap decoder that naturally accommodates variable group sizes. Experiments on six in-the-wild datasets, including two GER and four Individual Emotion Recognition (IER) benchmarks, show that structural supervision consistently improves representation learning. More importantly, the results reveal a clear distinction between GER and IER: optimizing the latent space alone is often insufficient for GER because it tends to attenuate interaction-related cues, whereas preserving explicit structural outputs improves collective affect inference. In contrast, projected structural representations seem to act as an effective denoising bottleneck for IER. VE-MD achieves state-of-the-art performance on GAF-3.0 (up to 90.06%) and VGAF (82.25% with multimodal fusion with audio). These results show that preserving interaction-related structural information is particularly beneficial for group-level affect modeling without relying on prior individual feature extraction. On IER datasets using multimodal fusion with audio modality, VE-MD outperforms SOTA on SamSemo (77.9%, adding text modality) while achieving competitive performances on MER-MULTI (63.8%), DFEW (70.7%) and EngageNet (69.0).", "date": "2026-04-02", "categories": []}
{"id": "2604.03050", "title": "MECO: A Multimodal Dataset for Emotion and Cognitive Understanding in Older Adults", "abstract": "While affective computing has advanced considerably, multimodal emotion prediction in aging populations remains underexplored, largely due to the scarcity of dedicated datasets. Existing multimodal benchmarks predominantly target young, cognitively healthy subjects, neglecting the influence of cognitive decline on emotional expression and physiological responses. To bridge this gap, we present MECO, a Multimodal dataset for Emotion and Cognitive understanding in Older adults. MECO includes 42 participants and provides approximately 38 hours of multimodal signals, yielding 30,592 synchronized samples. To maximize ecological validity, data collection followed standardized protocols within community-based settings. The modalities cover video, audio, electroencephalography (EEG), and electrocardiography (ECG). In addition, the dataset offers comprehensive annotations of emotional and cognitive states, including self-assessed valence, arousal, six basic emotions, and Mini-Mental State Examination cognitive scores. We further establish baseline benchmarks for both emotion and cognitive prediction. MECO serves as a foundational resource for multimodal modeling of affect and cognition in aging populations, facilitating downstream applications such as personalized emotion recognition and early detection of mild cognitive impairment (MCI) in real-world settings. The complete dataset and supplementary materials are available at https://maitrechen.github.io/meco-page/.", "date": "2026-04-03", "categories": []}
{"id": "2510.11579", "title": "MS-Mix: Sentiment-Guided Adaptive Augmentation for Multimodal Sentiment Analysis", "abstract": "Multimodal Sentiment Analysis (MSA) integrates complementary features from text, video, and audio for robust emotion understanding in human interactions. However, models suffer from severe data scarcity and high annotation costs, severely limiting real-world deployment in social media analytics and human-computer systems. Existing Mixup-based augmentation techniques, when naively applied to MSA, often produce semantically inconsistent samples and amplified label noise by ignoring emotional semantics across modalities. To address these challenges, we propose MS-Mix, an adaptive emotion-sensitive augmentation framework that automatically optimizes data quality in multimodal settings. Its key components are: (1) Sentiment-aware sample selection strategy that filters incompatible pairs via latent-space semantic similarity to prevent contradictory emotion mixing. (2) Sentiment intensity guided module with multi-head self-attention for computing modality-specific mixing ratios conditioned on emotional salience dynamically. (3) Sentiment alignment loss based on Kullback-Leibler divergence to align predicted sentiment distributions across modalities with ground-truth labels, improving discrimination and consistency. Extensive experiments on two public datasets with six state-of-the-art backbones confirm that MS-Mix consistently outperforms prior methods, significantly improving robustness and practical applicability for MSA. The source code is available at an anonymous link: https://anonymous.4open.science/r/MS-Mix-review-0C72.", "date": "2025-10-13", "categories": []}
{"id": "2512.17752", "title": "Affect, Body, Cognition, Demographics, and Emotion: The ABCDE of Text Features for Computational Affective Science", "abstract": "Work in Computational Affective Science and Computational Social Science explores a wide variety of research questions about people, emotions, behavior, and health. Such work often relies on language data that is first labeled with relevant information, such as the use of emotion words or the age of the speaker. Although many resources and algorithms exist to enable this type of labeling, discovering, accessing, and using them remains a substantial impediment, particularly for practitioners outside of computer science. Here, we present the ABCDE dataset (Affect, Body, Cognition, Demographics, and Emotion), a large-scale collection of over 400 million text utterances drawn from social media, blogs, books, and AI-generated sources. The dataset is annotated with a wide range of features relevant to computational affective and social science. ABCDE facilitates interdisciplinary research across numerous fields, including affective science, cognitive science, the digital humanities, sociology, political science, and computational linguistics.", "date": "2025-12-19", "categories": []}
{"id": "2604.01916", "title": "SURE: Synergistic Uncertainty-aware Reasoning for Multimodal Emotion Recognition in Conversations", "abstract": "Multimodal emotion recognition in conversations (MERC) requires integrating multimodal signals while being robust to noise and modeling contextual reasoning. Existing approaches often emphasize fusion but overlook uncertainty in noisy features and fine-grained reasoning. We propose SURE (Synergistic Uncertainty-aware REasoning) for MERC, a framework that improves robustness and contextual modeling. SURE consists of three components: an Uncertainty-Aware Mixture-of-Experts module to handle modality-specific noise, an Iterative Reasoning module for multi-turn reasoning over context, and a Transformer Gate module to capture intra- and inter-modal interactions. Experiments on benchmark MERC datasets show that SURE consistently outperforms state-of-the-art methods, demonstrating its effectiveness in robust multimodal reasoning. These results highlight the importance of uncertainty modeling and iterative reasoning in advancing emotion recognition in conversational settings.", "date": "2026-04-02", "categories": []}
{"id": "2511.06328", "title": "Improving Multimodal Sentiment Analysis via Modality Optimization and Dynamic Primary Modality Selection", "abstract": "Multimodal Sentiment Analysis (MSA) aims to predict sentiment from language, acoustic, and visual data in videos. However, imbalanced unimodal performance often leads to suboptimal fused representations. Existing approaches typically adopt fixed primary modality strategies to maximize dominant modality advantages, yet fail to adapt to dynamic variations in modality importance across different samples. Moreover, non-language modalities suffer from sequential redundancy and noise, degrading model performance when they serve as primary inputs. To address these issues, this paper proposes a modality optimization and dynamic primary modality selection framework (MODS). First, a Graph-based Dynamic Sequence Compressor (GDC) is constructed, which employs capsule networks and graph convolution to reduce sequential redundancy in acoustic/visual modalities. Then, we develop a sample-adaptive Primary Modality Selector (MSelector) for dynamic dominance determination. Finally, a Primary-modality-Centric Cross-Attention (PCCA) module is designed to enhance dominant modalities while facilitating cross-modal interaction. Extensive experiments on four benchmark datasets demonstrate that MODS outperforms state-of-the-art methods, achieving superior performance by effectively balancing modality contributions and eliminating redundant noise.", "date": "2025-11-09", "categories": []}
{"id": "2604.00933", "title": "EmoScene: A Dual-space Dataset for Controllable Affective Image Generation", "abstract": "Text-to-image diffusion models have achieved high visual fidelity, yet precise control over scene semantics and fine-grained affective tone remains challenging. Human visual affect arises from the rapid integration of contextual meaning, including valence, arousal, and dominance, with perceptual cues such as color harmony, luminance contrast, texture variation, curvature, and spatial layout. However, current text-to-image models rarely represent affective and perceptual factors within a unified representation, which limits their ability to synthesize scenes with coherent and nuanced emotional intent. To address this gap, we construct EmoScene, a large-scale dual-space emotion dataset that jointly encodes affective dimensions and perceptual attributes, with contextual semantics provided as supporting annotations. EmoScene contains 1.2M images across more than three hundred real-world scene categories, each annotated with discrete emotion labels, continuous VAD values, perceptual descriptors and textual captions. Multi-space analyses reveal how discrete emotions occupy the VAD space and how affect systematically correlates with scene-level perceptual factors. To benchmark EmoScene, we provide a lightweight reference baseline that injects dual-space controls into a frozen diffusion backbone via shallow cross-attention modulation, serving as a reproducible probe of affect controllability enabled by dual-space supervision.", "date": "2026-04-01", "categories": []}
{"id": "2603.27999", "title": "CLIP-AUTT: Test-Time Personalization with Action Unit Prompting for Fine-Grained Video Emotion Recognition", "abstract": "Personalization in emotion recognition (ER) is essential for an accurate interpretation of subtle and subject-specific expressive patterns. Recent advances in vision-language models (VLMs) such as CLIP demonstrate strong potential for leveraging joint image-text representations in ER. However, CLIP-based methods either depend on CLIP's contrastive pretraining or on LLMs to generate descriptive text prompts, which are noisy, computationally expensive, and fail to capture fine-grained expressions, leading to degraded performance. In this work, we leverage Action Units (AUs) as structured textual prompts within CLIP to model fine-grained facial expressions. AUs encode the subtle muscle activations underlying expressions, providing localized and interpretable semantic cues for more robust ER. We introduce CLIP-AU, a lightweight AU-guided temporal learning method that integrates interpretable AU semantics into CLIP. It learns generic, subject-agnostic representations by aligning AU prompts with facial dynamics, enabling fine-grained ER without CLIP fine-tuning or LLM-generated text supervision. Although CLIP-AU models fine-grained AU semantics, it does not adapt to subject-specific variability in subtle expressions. To address this limitation, we propose CLIP-AUTT, a video-based test-time personalization method that dynamically adapts AU prompts to videos from unseen subjects. By combining entropy-guided temporal window selection with prompt tuning, CLIP-AUTT enables subject-specific adaptation while preserving temporal consistency. Our extensive experiments on three challenging video-based subtle ER datasets, BioVid, StressID, and BAH, indicate that CLIP-AU and CLIP-AUTT outperform state-of-the-art CLIP-based FER and TTA methods, achieving robust and personalized subtle ER. Our code is publicly available at: https://github.com/osamazeeshan/CLIP-AUTT.", "date": "2026-03-30", "categories": []}
{"id": "2501.19111", "title": "A Benchmark for Incremental Micro-expression Recognition", "abstract": "Micro-expression recognition plays a pivotal role in understanding hidden emotions and has applications across various fields. Traditional recognition methods assume access to all training data at once, but real-world scenarios involve continuously evolving data streams. To respond to the requirement of adapting to new data while retaining previously learned knowledge, we introduce the first benchmark specifically designed for incremental micro-expression recognition. Our contributions include: Firstly, we formulate the incremental learning setting tailored for micro-expression recognition. Secondly, we organize sequential datasets with carefully curated learning orders to reflect real-world scenarios. Thirdly, we define two cross-evaluation-based testing protocols, each targeting distinct evaluation objectives. Finally, we provide six baseline methods and their corresponding evaluation results. This benchmark lays the groundwork for advancing incremental micro-expression recognition research. All source code used in this study will be publicly available at https://github.com/ZhengQinLai/IMER-benchmark.", "date": "2025-01-31", "categories": []}
{"id": "2603.26727", "title": "The Nonverbal Gap: Toward Affective Computer Vision for Safer and More Equitable Online Dating", "abstract": "Online dating has become the dominant way romantic relationships begin, yet current platforms strip the nonverbal cues: gaze, facial expression, body posture, response timing, that humans rely on to signal comfort, disinterest, and consent, creating a communication gap with disproportionate safety consequences for women. We argue that this gap represents both a technical opportunity and a moral responsibility for the computer vision community, which has developed the affective tools, facial action unit detection, gaze estimation, engagement modeling, and multimodal affect recognition, needed to begin addressing it, yet has largely ignored the dating domain as a research context. We propose a fairness-first research agenda organized around four capability areas: real-time discomfort detection, engagement asymmetry modeling between partners, consent-aware interaction design, and longitudinal interaction summarization, each grounded in established CV methodology and motivated by the social psychology of romantic communication. We argue that responsible pursuit of this agenda requires purpose-built datasets collected under dyadic consent protocols, fairness evaluation disaggregated across race, gender identity, neurotype, and cultural background, and architectural commitments to on-device processing that prevent affective data from becoming platform surveillance infrastructure. This vision paper calls on the WICV community, whose members are uniquely positioned to understand both the technical opportunity and the human stakes, to establish online dating safety as a first-class research domain before commercial deployment outpaces ethical deliberation.", "date": "2026-03-20", "categories": []}
{"id": "2603.26840", "title": "Dual-branch Graph Domain Adaptation for Cross-scenario Multi-modal Emotion Recognition", "abstract": "Multimodal Emotion Recognition in Conversations (MERC) aims to predict speakers' emotional states in multi-turn dialogues through text, audio, and visual cues. In real-world settings, conversation scenarios differ significantly in speakers, topics, styles, and noise levels. Existing MERC methods generally neglect these cross-scenario variations, limiting their ability to transfer models trained on a source domain to unseen target domains. To address this issue, we propose a Dual-branch Graph Domain Adaptation framework (DGDA) for multimodal emotion recognition under cross-scenario conditions. We first construct an emotion interaction graph to characterize complex emotional dependencies among utterances. A dual-branch encoder, consisting of a hypergraph neural network (HGNN) and a path neural network (PathNN), is then designed to explicitly model multivariate relationships and implicitly capture global dependencies. To enable out-of-domain generalization, a domain adversarial discriminator is introduced to learn invariant representations across domains. Furthermore, a regularization loss is incorporated to suppress the negative influence of noisy labels. To the best of our knowledge, DGDA is the first MERC framework that jointly addresses domain shift and label noise. Theoretical analysis provides tighter generalization bounds, and extensive experiments on IEMOCAP and MELD demonstrate that DGDA consistently outperforms strong baselines and better adapts to cross-scenario conversations. Our code is available at https://github.com/Xudmm1239439/DGDA-Net.", "date": "2026-03-27", "categories": []}
{"id": "2603.27798", "title": "Towards Emotion Recognition with 3D Pointclouds Obtained from Facial Expression Images", "abstract": "Facial Emotion Recognition is a critical research area within Affective Computing due to its wide-ranging applications in Human Computer Interaction, mental health assessment and fatigue monitoring. Current FER methods predominantly rely on Deep Learning techniques trained on 2D image data, which pose significant privacy concerns and are unsuitable for continuous, real-time monitoring. As an alternative, we propose High-Frequency Wireless Sensing (HFWS) as an enabler of continuous, privacy-aware FER, through the generation of detailed 3D facial pointclouds via on-person sensors embedded in wearables. We present arguments supporting the privacy advantages of HFWS over traditional 2D imaging, particularly under increasingly stringent data protection regulations. A major barrier to adopting HFWS for FER is the scarcity of labeled 3D FER datasets. Towards addressing this issue, we introduce a FLAME-based method to generate 3D facial pointclouds from existing public 2D datasets. Using this approach, we create AffectNet3D, a 3D version of the AffectNet database. To evaluate the quality and usability of the generated data, we design a pointcloud refinement pipeline focused on isolating the facial region, and train the popular PointNet++ model on the refined pointclouds. Fine-tuning the model on a small subset of the unseen 3D FER dataset BU-3DFE yields a classification accuracy exceeding 70%, comparable to oracle-level performance. To further investigate the potential of HFWS-based FER for continuous monitoring, we simulate wearable sensing conditions by masking portions of the generated pointclouds. Experimental results show that models trained on AffectNet3D and fine-tuned with just 25% of BU-3DFE outperform those trained solely on BU-3DFE. These findings highlight the viability of our pipeline and support the feasibility of continuous, privacy-aware FER via wearable HFWS systems.", "date": "2026-03-29", "categories": []}
{"id": "2603.25752", "title": "Relational graph-driven differential denoising and diffusion attention fusion for multimodal conversation emotion recognition", "abstract": "In real-world scenarios, audio and video signals are often subject to environmental noise and limited acquisition conditions, resulting in extracted features containing excessive noise. Furthermore, there is an imbalance in data quality and information carrying capacity between different modalities. These two issues together lead to information distortion and weight bias during the fusion phase, impairing overall recognition performance. Most existing methods neglect the impact of noisy modalities and rely on implicit weighting to model modality importance, thereby failing to explicitly account for the predominant contribution of the textual modality in emotion understanding. To address these issues, we propose a relation-aware denoising and diffusion attention fusion model for MCER. Specifically, we first design a differential Transformer that explicitly computes the differences between two attention maps, thereby enhancing temporally consistent information while suppressing time-irrelevant noise, which leads to effective denoising in both audio and video modalities. Second, we construct modality-specific and cross-modality relation subgraphs to capture speaker-dependent emotional dependencies, enabling fine-grained modeling of intra- and inter-modal relationships. Finally, we introduce a text-guided cross-modal diffusion mechanism that leverages self-attention to model intra-modal dependencies and adaptively diffuses audiovisual information into the textual stream, ensuring more robust and semantically aligned multimodal fusion.", "date": "2026-03-22", "categories": []}
{"id": "2603.26575", "title": "The Climber's Grip -- Personalized Deep Learning Models for Fear and Muscle Activity in Climbing", "abstract": "Climbing is a multifaceted sport that combines physical demands and emotional and cognitive challenges. Ascent styles differ in fall distance with lead climbing involving larger falls than top rope climbing, which may result in different perceived risk and fear. In this study, we investigated the psychophysiological relationship between perceived fear and muscle activity in climbers using a combination of statistical modeling and deep learning techniques. We conducted an experiment with 19 climbers, collecting electromyography (EMG), electrocardiography (ECG) and arm motion data during lead and top rope climbing. Perceived fear ratings were collected for the different phases of the climb. Using a linear mixed-effects model, we analyzed the relationships between perceived fear and physiological measures. To capture the non-linear dynamics of this relationship, we extended our analysis to deep learning models and integrated random effects for a personalized modeling approach. Our results showed that random effects improved model performance of the mean squared error (MSE), mean absolute error (MAE) and root mean squared error (RMSE). The results showed that muscle fatigue correlates significantly with increased fear during \\textit{lead climbing}. This study highlights the potential of combining statistical and deep learning approaches for modeling the interplay between psychological and physiological states during climbing.", "date": "2026-03-27", "categories": []}
{"id": "2602.08371", "title": "ViGoEmotions: A Benchmark Dataset For Fine-grained Emotion Detection on Vietnamese Texts", "abstract": "Emotion classification plays a significant role in emotion prediction and harmful content detection. Recent advancements in NLP, particularly through large language models (LLMs), have greatly improved outcomes in this field. This study introduces ViGoEmotions -- a Vietnamese emotion corpus comprising 20,664 social media comments in which each comment is classified into 27 fine-grained distinct emotions. To evaluate the quality of the dataset and its impact on emotion classification, eight pre-trained Transformer-based models were evaluated under three preprocessing strategies: preserving original emojis with rule-based normalization, converting emojis into textual descriptions, and applying ViSoLex, a model-based lexical normalization system. Results show that converting emojis into text often improves the performance of several BERT-based baselines, while preserving emojis yields the best results for ViSoBERT and CafeBERT. In contrast, removing emojis generally leads to lower performance. ViSoBERT achieved the highest Macro F1-score of 61.50% and Weighted F1-score of 63.26%. Strong performance was also observed from CafeBERT and PhoBERT. These findings highlight that while the proposed corpus can support diverse architectures effectively, preprocessing strategies and annotation quality remain key factors influencing downstream performance.", "date": "2026-02-09", "categories": []}
{"id": "2603.25031", "title": "From Stateless to Situated: Building a Psychological World for LLM-Based Emotional Support", "abstract": "In psychological support and emotional companionship scenarios, the core limitation of large language models (LLMs) lies not merely in response quality, but in their reliance on local next-token prediction, which prevents them from maintaining the temporal continuity, stage awareness, and user consent boundaries required for multi-turn intervention. This stateless characteristic makes systems prone to premature advancement, stage misalignment, and boundary violations in continuous dialogue. To address this problem, we argue that the key challenge in process-oriented emotional support is not simply generating natural language, but constructing a sustainably updatable external situational structure for the model. We therefore propose LEKIA 2.0, a situated LLM architecture that separates the cognitive layer from the executive layer, thereby decoupling situational modeling from intervention execution. This design enables the system to maintain stable representations of the user's situation and consent boundaries throughout ongoing interaction. To evaluate this process-control capability, we further introduce a Static-to-Dynamic online evaluation protocol for multi-turn interaction. LEKIA achieved an average absolute improvement of approximately 31% over prompt-only baselines in deep intervention loop completion. The results suggest that an external situational structure is a key enabling condition for building stable, controllable, and situated emotional support systems.", "date": "2026-03-26", "categories": []}
{"id": "2603.25570", "title": "TAAC: A gate into Trustable Audio Affective Computing", "abstract": "With the emergence of AI techniques for depression diagnosis, the conflict between high demand and limited supply for depression screening has been significantly alleviated. Among various modal data, audio-based depression diagnosis has received increasing attention from both academia and industry since audio is the most common carrier of emotion transmission. Unfortunately, audio data also contains User-sensitive Identity Information (ID), which is extremely vulnerable and may be maliciously used during the smart diagnosis process. Among previous methods, the clarification between depression features and sensitive features has always serve as a barrier. It is also critical to the problem for introducing a safe encryption methodology that only encrypts the sensitive features and a powerful classifier that can correctly diagnose the depression. To track these challenges, by leveraging adversarial loss-based Subspace Decomposition, we propose a first practical framework \\name presented for Trustable Audio Affective Computing, to perform automated depression detection through audio within a trustable environment. The key enablers of TAAC are Differentiating Features Subspace Decompositor (DFSD), Flexible Noise Encryptor (FNE) and Staged Training Paradigm, used for decomposition, ID encryption and performance enhancement, respectively. Extensive experiments with existing encryption methods demonstrate our framework's preeminent performance in depression detection, ID reservation and audio reconstruction. Meanwhile, the experiments across various setting demonstrates our model's stability under different encryption strengths. Thus proving our framework's excellence in Confidentiality, Accuracy, Traceability, and Adjustability.", "date": "2026-03-26", "categories": []}
{"id": "2603.23650", "title": "Foundation Model Embeddings Meet Blended Emotions: A Multimodal Fusion Approach for the BLEMORE Challenge", "abstract": "We present our system for the BLEMORE Challenge at FG 2026 on blended emotion recognition with relative salience prediction. Our approach combines six encoder families through late probability fusion: an S4D-ViTMoE face encoder adapted with soft-label KL training, frozen layer-selective Wav2Vec2 audio features, finetuned body-language encoders (TimeSformer, VideoMAE), and -- for the first time in emotion recognition -- Gemini Embedding 2.0, a large multimodal model whose video embeddings produce competitive presence accuracy (ACCP = 0.320) from only 2 seconds of input. Three key findings emerge from our experiments: selecting prosody-encoding layers (6--12) from frozen Wav2Vec2 outperforms end-to-end finetuning (Score 0.207 vs. 0.161), as the non-verbal nature of BLEMORE audio makes phonetic layers irrelevant; the post-processing salience threshold $\\beta$ varies from 0.05 to 0.43 across folds, revealing that personalized expression styles are the primary bottleneck; and task-adapted encoders collectively receive 62\\% of ensemble weight over general-purpose baselines. Our 12-encoder system achieves Score = 0.279 (ACCP = 0.391, ACCS = 0.168) on the test set, placing 6th.", "date": "2026-03-24", "categories": []}
{"id": "2603.22306", "title": "Memory Bear AI Memory Science Engine for Multimodal Affective Intelligence: A Technical Report", "abstract": "Affective judgment in real interaction is rarely a purely local prediction problem. Emotional meaning often depends on prior trajectory, accumulated context, and multimodal evidence that may be weak, noisy, or incomplete at the current moment. Although multimodal emotion recognition (MER) has improved the integration of text, speech, and visual signals, many existing systems remain optimized for short-range inference and provide limited support for persistent affective memory, long-horizon dependency modeling, and robust interpretation under imperfect input.\n  This technical report presents the Memory Bear AI Memory Science Engine, a memory-centered framework for multimodal affective intelligence. Instead of treating emotion as a transient output label, the framework models affective information as a structured and evolving variable within a memory system. It organizes processing through structured memory formation, working-memory aggregation, long-term consolidation, memory-driven retrieval, dynamic fusion calibration, and continuous memory updating. At its core, multimodal signals are transformed into structured Emotion Memory Units (EMUs), enabling affective information to be preserved, reactivated, and revised across interaction horizons.\n  Experimental results show consistent gains over comparison systems across benchmark and business-grounded settings, with stronger accuracy and robustness, especially under noisy or missing-modality conditions. The framework offers a practical step from local emotion recognition toward more continuous, robust, and deployment-relevant affective intelligence.", "date": "2026-03-18", "categories": []}
{"id": "2603.22345", "title": "Dynamic Fusion-Aware Graph Convolutional Neural Network for Multimodal Emotion Recognition in Conversations", "abstract": "Multimodal emotion recognition in conversations (MERC) aims to identify and understand the emotions expressed by speakers during utterance interaction from multiple modalities (e.g., text, audio, images, etc.). Existing studies have shown that GCN can improve the performance of MERC by modeling dependencies between speakers. However, existing methods usually use fixed parameters to process multimodal features for different emotion types, ignoring the dynamics of fusion between different modalities, which forces the model to balance performance between multiple emotion categories, thus limiting the model's performance on some specific emotions. To this end, we propose a dynamic fusion-aware graph convolutional neural network (DF-GCN) for robust recognition of multimodal emotion features in conversations. Specifically, DF-GCN integrates ordinary differential equations into graph convolutional networks (GCNs) to {capture} the dynamic nature of emotional dependencies within utterance interaction networks and leverages the prompts generated by the global information vector (GIV) of the utterance to guide the dynamic fusion of multimodal features. This allows our model to dynamically change parameters when processing each utterance feature, so that different network parameters can be equipped for different emotion categories in the inference stage, thereby achieving more flexible emotion classification and enhancing the generalization ability of the model. Comprehensive experiments conducted on two public multimodal conversational datasets {confirm} that the proposed DF-GCN model delivers superior performance, benefiting significantly from the dynamic fusion mechanism introduced.", "date": "2026-03-22", "categories": []}
{"id": "2603.23057", "title": "Prompt Amplification and Zero-Shot Late Fusion in Audio-Language Models for Speech Emotion Recognition", "abstract": "Audio-Language Models (ALMs) are making strides in understanding speech and non-speech audio. However, domain-specialist Foundation Models (FMs) remain the best for closed-ended speech processing tasks such as Speech Emotion Recognition (SER). Using ALMs for Zero-shot SER is a popular choice, but their potential to work with specialists to achieve state-of-the-art (SOTA) performance remains unexplored. We propose ZS-Fuse, a late-fusion method that combines zero-shot emotion estimates from a dual-encoder ALM with specialist FMs. To handle ambiguity in emotions and sensitivity to prompt choice, 1) we use a simple prompt ensemble and 2) suggest a novel technique called prompt amplification, which repeats audio and text queries to discover stronger zero-shot capabilities. We demonstrate the efficacy of our technique by evaluating ZS-Fuse with three dual-encoder ALMs and two FMs, and report improvements over SOTA baselines, such as WavLM-Large, on three speech emotion recognition datasets.", "date": "2026-03-24", "categories": []}
{"id": "2510.14922", "title": "TRI-DEP: A Trimodal Comparative Study for Depression Detection Using Speech, Text, and EEG", "abstract": "Depression is a widespread mental health disorder, yet its automatic detection remains challenging. Prior work has explored unimodal and multimodal approaches, with multimodal systems showing promise by leveraging complementary signals. However, existing studies are limited in scope, lack systematic comparisons of features, and suffer from inconsistent evaluation protocols. We address these gaps by systematically exploring feature representations and modelling strategies across EEG, together with speech and text. We evaluate handcrafted features versus pre-trained embeddings, assess the effectiveness of different neural encoders, compare unimodal, bimodal, and trimodal configurations, and analyse fusion strategies with attention to the role of EEG. Consistent subject-independent splits are applied to ensure robust, reproducible benchmarking. Our results show that (i) the combination of EEG, speech and text modalities enhances multimodal detection, (ii) pretrained embeddings outperform handcrafted features, and (iii) carefully designed trimodal models achieve state-of-the-art performance. Our work lays the groundwork for future research in multimodal depression detection.", "date": "2025-10-16", "categories": []}
{"id": "2512.03521", "title": "Cross-Space Synergy: A Unified Framework for Multimodal Emotion Recognition in Conversation", "abstract": "Multimodal Emotion Recognition in Conversation (MERC) aims to predict speakers' emotions by integrating textual, acoustic, and visual cues. Existing approaches either struggle to capture complex cross-modal interactions or experience gradient conflicts and unstable training when using deeper architectures. To address these issues, we propose Cross-Space Synergy (CSS), which couples a representation component with an optimization component. Synergistic Polynomial Fusion (SPF) serves the representation role, leveraging low-rank tensor factorization to efficiently capture high-order cross-modal interactions. Pareto Gradient Modulator (PGM) serves the optimization role, steering updates along Pareto-optimal directions across competing objectives to alleviate gradient conflicts and improve stability. Experiments show that CSS outperforms existing representative methods on IEMOCAP and MELD in both accuracy and training stability, demonstrating its effectiveness in complex multimodal scenarios.", "date": "2025-12-03", "categories": []}
{"id": "2603.13406", "title": "Nuanced Emotion Recognition Based on a Segment-based MLLM Framework Leveraging Qwen3-Omni for AH Detection", "abstract": "Emotion recognition in videos is a pivotal task in affective computing, where identifying subtle psychological states such as Ambivalence and Hesitancy holds significant value for behavioral intervention and digital health. Ambivalence and Hesitancy states often manifest through cross-modal inconsistencies such as discrepancies between facial expressions, vocal tones, and textual semantics, posing a substantial challenge for automated recognition. This paper proposes a recognition framework that integrates temporal segment modeling with Multimodal Large Language Models. To address computational efficiency and token constraints in long video processing, we employ a segment-based strategy, partitioning videos into short clips with a maximum duration of 5 seconds. We leverage the Qwen3-Omni-30B-A3B model, fine-tuned on the BAH dataset using LoRA and full-parameter strategies via the MS-Swift framework, enabling the model to synergistically analyze visual and auditory signals. Experimental results demonstrate that the proposed method achieves an accuracy of 85.1% on the test set, significantly outperforming existing benchmarks and validating the superior capability of Multimodal Large Language Models in capturing complex and nuanced emotional conflicts. The code is released at https://github.com/dlnn123/A-H-Detection-with-Qwen-Omni.git.", "date": "2026-03-12", "categories": []}
{"id": "2603.20222", "title": "Linguistic Signatures for Enhanced Emotion Detection", "abstract": "Emotion detection is a central problem in NLP, with recent progress driven by transformer-based models trained on established datasets. However, little is known about the linguistic regularities that characterize how emotions are expressed across different corpora and labels. This study examines whether linguistic features can serve as reliable interpretable signals for emotion recognition in text. We extract emotion-specific linguistic signatures from 13 English datasets and evaluate how incorporating these features into transformer models impacts performance. Our RoBERTa-based models enriched with high level linguistic features achieve consistent performance gains of up to +2.4 macro F1 on the GoEmotions benchmark, showing that explicit lexical cues can complement neural representations and improve robustness in predicting emotion categories.", "date": "2026-03-04", "categories": []}
{"id": "2603.21038", "title": "Reading Between the Lines: How Electronic Nonverbal Cues shape Emotion Decoding", "abstract": "As text-based computer-mediated communication (CMC) increasingly structures everyday interaction, a central question re-emerges with new urgency: How do users reconstruct nonverbal expression in environments where embodied cues are absent? This paper provides a systematic, theory-driven account of electronic nonverbal cues (eNVCs) - textual analogues of kinesics, vocalics, and paralinguistics - in public microblog communication. Across three complementary studies, we advance conceptual, empirical, and methodological contributions. Study 1 develops a unified taxonomy of eNVCs grounded in foundational nonverbal communication theory and introduces a scalable Python toolkit for their automated detection. Study 2, a within-subject survey experiment, offers controlled causal evidence that eNVCs substantially improve emotional decoding accuracy and lower perceived ambiguity, while also identifying boundary conditions, such as sarcasm, under which these benefits weaken or disappear. Study 3, through focus group discussions, reveals the interpretive strategies users employ when reasoning about digital prosody, including drawing meaning from the absence of expected cues and defaulting toward negative interpretations in ambiguous contexts. Together, these studies establish eNVCs as a coherent and measurable class of digital behaviors, refine theoretical accounts of cue richness and interpretive effort, and provide practical tools for affective computing, user modeling, and emotion-aware interface design. The eNVC detection toolkit is available as a Python and R package at https://github.com/kokiljaidka/envc.", "date": "2026-03-22", "categories": []}
{"id": "2508.09954", "title": "Disambiguation of Emotion Annotations by Contextualizing Events in Plausible Narratives", "abstract": "Ambiguity in emotion analysis stems both from potentially missing information and the subjectivity of interpreting a text. The latter did receive substantial attention, but can we fill missing information to resolve ambiguity? We address this question by developing a method to automatically generate reasonable contexts for an otherwise ambiguous classification instance. These generated contexts may act as illustrations of potential interpretations by different readers, as they can fill missing information with their individual world knowledge. This task to generate plausible narratives is a challenging one: We combine techniques from short story generation to achieve coherent narratives. The resulting English dataset of Emotional BackStories, EBS, allows for the first comprehensive and systematic examination of contextualized emotion analysis. We conduct automatic and human annotation and find that the generated contextual narratives do indeed clarify the interpretation of specific emotions. Particularly relief and sadness benefit from our approach, while joy does not require the additional context we provide.", "date": "2025-08-13", "categories": []}
{"id": "2603.11715", "title": "Affect Decoding in Phonated and Silent Speech Production from Surface EMG", "abstract": "The expression of affect is integral to spoken communication, yet, its link to underlying articulatory execution remains unclear. Measures of articulatory muscle activity such as EMG could reveal how speech production is modulated by emotion alongside acoustic speech analyses. We investigate affect decoding from facial and neck surface electromyography (sEMG) during phonated and silent speech production. For this purpose, we introduce a dataset comprising 2,780 utterances from 12 participants across 3 tasks, on which we evaluate both intra- and inter-subject decoding using a range of features and model embeddings. Our results reveal that EMG representations reliably discriminate frustration with up to 0.845 AUC, and generalize well across articulation modes. Our ablation study further demonstrates that affective signatures are embedded in facial motor activity and persist in the absence of phonation, highlighting the potential of EMG sensing for affect-aware silent speech interfaces.", "date": "2026-03-12", "categories": []}
{"id": "2603.18462", "title": "AlignMamba-2: Enhancing Multimodal Fusion and Sentiment Analysis with Modality-Aware Mamba", "abstract": "In the era of large-scale pre-trained models, effectively adapting general knowledge to specific affective computing tasks remains a challenge, particularly regarding computational efficiency and multimodal heterogeneity. While Transformer-based methods have excelled at modeling inter-modal dependencies, their quadratic computational complexity limits their use with long-sequence data. Mamba-based models have emerged as a computationally efficient alternative; however, their inherent sequential scanning mechanism struggles to capture the global, non-sequential relationships that are crucial for effective cross-modal alignment. To address these limitations, we propose \\textbf{AlignMamba-2}, an effective and efficient framework for multimodal fusion and sentiment analysis. Our approach introduces a dual alignment strategy that regularizes the model using both Optimal Transport distance and Maximum Mean Discrepancy, promoting geometric and statistical consistency between modalities without incurring any inference-time overhead. More importantly, we design a Modality-Aware Mamba layer, which employs a Mixture-of-Experts architecture with modality-specific and modality-shared experts to explicitly handle data heterogeneity during the fusion process. Extensive experiments on four challenging benchmarks, including dynamic time-series (on the CMU-MOSI and CMU-MOSEI datasets) and static image-related tasks (on the NYU-Depth V2 and MVSA-Single datasets), demonstrate that AlignMamba-2 establishes a new state-of-the-art in both effectiveness and efficiency across diverse pattern recognition tasks, ranging from dynamic time-series analysis to static image-text classification.", "date": "2026-03-19", "categories": []}
{"id": "2603.18758", "title": "Dual-Model Prediction of Affective Engagement and Vocal Attractiveness from Speaker Expressiveness in Video Learning", "abstract": "This paper outlines a machine learning-enabled speaker-centric Emotion AI approach capable of predicting audience-affective engagement and vocal attractiveness in asynchronous video-based learning, relying solely on speaker-side affective expressions. Inspired by the demand for scalable, privacy-preserving affective computing applications, this speaker-centric Emotion AI approach incorporates two distinct regression models that leverage a massive corpus developed within Massive Open Online Courses (MOOCs) to enable affectively engaging experiences. The regression model predicting affective engagement is developed by assimilating emotional expressions emanating from facial dynamics, oculomotor features, prosody, and cognitive semantics, while incorporating a second regression model to predict vocal attractiveness based exclusively on speaker-side acoustic features. Notably, on speaker-independent test sets, both regression models yielded impressive predictive performance (R2 = 0.85 for affective engagement and R2 = 0.88 for vocal attractiveness), confirming that speaker-side affect can functionally represent aggregated audience feedback. This paper provides a speaker-centric Emotion AI approach substantiated by an empirical study discovering that speaker-side multimodal features, including acoustics, can prospectively forecast audience feedback without necessarily employing audience-side input information.", "date": "2026-03-19", "categories": []}
{"id": "2507.05591", "title": "MLlm-DR: Towards Explainable Depression Recognition with MultiModal Large Language Models", "abstract": "Automated depression diagnosis aims to analyze multimodal information from interview videos to predict participants' depression scores. Previous studies often lack clear explanations of how these scores were determined, limiting their adoption in clinical practice. While the advent of LLMs provides a possible pathway for explainable depression diagnosis, current LLMs capable of processing multimodal data lack training on interview data, resulting in poor diagnostic performance when used directly. In this paper, we propose a novel multimodal large language model (MLlm-DR) that can understand multimodal information inputs and supports explainable depression diagnosis. MLlm-DR integrates a smaller LLMs and a lightweight query module (LQ-former). Specifically, the smaller LLMs is designed to generate depression scores and corresponding evaluation rationales. To enhance its logical reasoning for domain-specific tasks while maintaining practicality, we constructed a robust training dataset to fine-tune it. Meanwhile, the LQ-former captures depression-related features from speech and visual data, aiding the model's ability to process multimodal information, to achieve comprehensive depression diagnosis. Our approach achieves state-of-the-art results on two interview-based benchmark datasets, CMDC and E-DAIC-WOZ, demonstrating its effectiveness and superiority.", "date": "2025-07-08", "categories": []}
{"id": "2603.11971", "title": "Multimodal Emotion Recognition via Bi-directional Cross-Attention and Temporal Modeling", "abstract": "Expression recognition in in-the-wild video data remains challenging due to substantial variations in facial appearance, background conditions, audio noise, and the inherently dynamic nature of human affect. Relying on a single modality, such as facial expressions or speech, is often insufficient for capturing these complex emotional cues. To address this limitation, we propose a multimodal emotion recognition framework for the Expression (EXPR) task in the 10th Affective Behavior Analysis in-the-wild (ABAW) Challenge. Our framework builds on large-scale pre-trained models for visual and audio representation learning and integrates them in a unified multimodal architecture. To better capture temporal patterns in facial expression sequences, we incorporate temporal visual modeling over video windows. We further introduce a bi-directional cross-attention fusion module that enables visual and audio features to interact in a symmetric manner, facilitating cross-modal contextualization and complementary emotion understanding. In addition, we employ a text-guided contrastive objective to encourage semantically meaningful visual representations through alignment with emotion-related text prompts. Experimental results on the ABAW 10th EXPR benchmark demonstrate the effectiveness of the proposed framework, achieving a Macro F1 score of 0.32 compared to the baseline score of 0.25, and highlight the benefit of combining temporal visual modeling, audio representation learning, and cross-modal fusion for robust emotion recognition in unconstrained real-world environments.", "date": "2026-03-12", "categories": []}
{"id": "2603.16939", "title": "Solution for 10th Competition on Ambivalence/Hesitancy (AH) Video Recognition Challenge using Divergence-Based Multimodal Fusion", "abstract": "We address the Ambivalence/Hesitancy (A/H) Video Recognition Challenge at the 10th ABAW Competition (CVPR 2026). We propose a divergence-based multimodal fusion that explicitly measures cross-modal conflict between visual, audio, and textual channels. Visual features are encoded as Action Units (AUs) extracted via Py-Feat, audio via Wav2Vec 2.0, and text via BERT. Each modality is processed by a BiLSTM with attention pooling and projected into a shared embedding space. The fusion module computes pairwise absolute differences between modality embeddings, directly capturing the incongruence that characterizes A/H. On the BAH dataset, our approach achieves a Macro F1 of 0.6808 on the validation test set, outperforming the challenge baseline of 0.2827. Statistical analysis across 1{,}132 videos confirms that temporal variability of AUs is the dominant visual discriminator of A/H.", "date": "2026-03-15", "categories": []}
{"id": "2603.15818", "title": "Conflict-Aware Multimodal Fusion for Ambivalence and Hesitancy Recognition", "abstract": "Ambivalence and hesitancy (A/H) are subtle affective states where a person shows conflicting signals through different channels -- saying one thing while their face or voice tells another story. Recognising these states automatically is valuable in clinical settings, but it is hard for machines because the key evidence lives in the \\emph{disagreements} between what is said, how it sounds, and what the face shows.\n  We present \\textbf{ConflictAwareAH}, a multimodal framework built for this problem. Three pre-trained encoders extract video, audio, and text representations. Pairwise conflict features -- element-wise absolute differences between modality embeddings -- serve as \\emph{bidirectional} cues: large cross-modal differences flag A/H, while small differences confirm behavioural consistency and anchor the negative class. This conflict-aware design addresses a key limitation of text-dominant approaches, which tend to over-detect A/H (high F1-AH) while struggling to confirm its absence: our multimodal model improves F1-NoAH by +4.6 points over text alone and halves the class-performance gap. A complementary \\emph{text-guided late fusion} strategy blends a text-only auxiliary head with the full model at inference, adding +4.1 Macro F1.\n  On the BAH dataset from the ABAW10 Ambivalence/Hesitancy Challenge, our method reaches \\textbf{0.694 Macro F1} on the labelled test split and \\textbf{0.715} on the private leaderboard, outperforming published multimodal baselines by over 10 points -- all on a single GPU in under 25 minutes of training.", "date": "2026-03-16", "categories": []}
{"id": "2603.16463", "title": "Follow the Clues, Frame the Truth: Hybrid-evidential Deductive Reasoning in Open-Vocabulary Multimodal Emotion Recognition", "abstract": "Open-Vocabulary Multimodal Emotion Recognition (OV-MER) is inherently challenging due to the ambiguity of equivocal multimodal cues, which often stem from distinct unobserved situational dynamics. While Multimodal Large Language Models (MLLMs) offer extensive semantic coverage, their performance is often bottlenecked by premature commitment to dominant data priors, resulting in suboptimal heuristics that overlook crucial, complementary affective cues across modalities. We argue that effective affective reasoning requires more than surface-level association; it necessitates reconstructing nuanced emotional states by synthesizing multiple evidence-grounded rationales that reconcile these observations from diverse latent perspectives. We introduce HyDRA, a Hybrid-evidential Deductive Reasoning Architecture that formalizes inference as a Propose-Verify-Decide protocol. To internalize this abductive process, we employ reinforcement learning with hierarchical reward shaping, aligning the reasoning trajectories with final task performance to ensure they best reconcile the observed multimodal cues. Systematic evaluations validate our design choices, with HyDRA consistently outperforming strong baselines--especially in ambiguous or conflicting scenarios--while providing interpretable, diagnostic evidence traces.", "date": "2026-03-17", "categories": []}
{"id": "2603.16483", "title": "On the Emotion Understanding of Synthesized Speech", "abstract": "Emotion is a core paralinguistic feature in voice interaction. It is widely believed that emotion understanding models learn fundamental representations that transfer to synthesized speech, making emotion understanding results a plausible reward or evaluation metric for assessing emotional expressiveness in speech synthesis. In this work, we critically examine this assumption by systematically evaluating Speech Emotion Recognition (SER) on synthesized speech across datasets, discriminative and generative SER models, and diverse synthesis models. We find that current SER models can not generalize to synthesized speech, largely because speech token prediction during synthesis induces a representation mismatch between synthesized and human speech. Moreover, generative Speech Language Models (SLMs) tend to infer emotion from textual semantics while ignoring paralinguistic cues. Overall, our findings suggest that existing SER models often exploit non-robust shortcuts rather than capturing fundamental features, and paralinguistic understanding in SLMs remains challenging.", "date": "2026-03-17", "categories": []}
{"id": "2603.16553", "title": "EmoLLM: Appraisal-Grounded Cognitive-Emotional Co-Reasoning in Large Language Models", "abstract": "Large language models (LLMs) demonstrate strong cognitive intelligence (IQ), yet many real-world interactions also require emotional intelligence (EQ) to produce responses that are both factually reliable and emotionally appropriate. In settings such as emotional support, technical assistance, and consultation, effective dialogue depends on how situations are appraised with respect to the user's needs, goals, and coping capacity. Inspired by appraisal theory, we propose EmoLLM, an appraisal-grounded framework for IQ/EQ co-reasoning in dialogue. EmoLLM uses an explicit Appraisal Reasoning Graph (ARG) to structure intermediate reasoning over contextual facts, inferred user needs, appraisal dimensions, emotional states, and response strategies before generating a reply. We train EmoLLM in a multi-turn role-play environment with reinforcement learning, where reverse-perspective reasoning provides reward signals based on predicted user-side consequences of responses. Across diverse dialogue settings, EmoLLM improves emotional state outcomes and response quality over strong baselines while preserving strong factual reliability.", "date": "2026-03-17", "categories": []}
{"id": "2603.16719", "title": "Emotion-Aware Classroom Quality Assessment Leveraging IoT-Based Real-Time Student Monitoring", "abstract": "This study presents high-throughput, real-time multi-agent affective computing framework designed to enhance classroom learning through emotional state monitoring. As large classroom sizes and limited teacher student interaction increasingly challenge educators, there is a growing need for scalable, data-driven tools capable of capturing students' emotional and engagement patterns in real time. The system was evaluated using the Classroom Emotion Dataset, consisting of 1,500 labeled images and 300 classroom detection videos. Tailored for IoT devices, the system addresses load balancing and latency challenges through efficient real-time processing. Field testing was conducted across three educational institutions in a large metropolitan area: a primary school (hereafter school A), a secondary school (school B), and a high school (school C). The system demonstrated robust performance, detecting up to 50 faces at 25 FPS and achieving 88% overall accuracy in classifying classroom engagement states. Implementation results showed positive outcomes, with favorable feedback from students, teachers, and parents regarding improved classroom interaction and teaching adaptation. Key contributions of this research include establishing a practical, IoT-based framework for emotion-aware learning environments and introducing the 'Classroom Emotion Dataset' to facilitate further validation and research.", "date": "2026-03-17", "categories": []}
{"id": "2603.16760", "title": "Dual Stream Independence Decoupling for True Emotion Recognition under Masked Expressions", "abstract": "Recongnizing true emotions from masked expressions is extremely challenging due to deliberate concealment. Existing paradigms recognize true emotions from masked-expression clips that contain onsetframes just starting to disguise. However, this paradigm may not reflect the actual disguised state, as the onsetframe leaks the true emotional information without reaching a stable disguise state. Thus, this paper introduces a novel apexframe-based paradigm that classifies true emotions from the apexframe with a stable disguised state. Furthermore, this paper proposes a novel dual stream independence decoupling framework that decouples true and disguised emotion features, avoiding the interference of disguised emotions on true emotions. For efficient decoupling, we design a decoupling loss group, comprising two classification losses that learn true emotion and disguised expression features, respectively, and a Hilbert-Schmidt Independence loss that enhances the independence of two features. Experiments demonstrate that the apexframe-based paradigm is challenging, and the proposed decouple framework improves recogntion performances.", "date": "2026-03-17", "categories": []}
{"id": "2501.00691", "title": "Labels Generated by Large Language Models Help Measure People's Empathy in Vitro", "abstract": "Large language models (LLMs) have revolutionised many fields, with LLM-as-a-service (LLMSaaS) offering accessible, general-purpose solutions without costly task-specific training. In contrast to the widely studied prompt engineering for directly solving tasks (in vivo), this paper explores LLMs' potential for in-vitro applications: using LLM-generated labels to improve supervised training of mainstream models. We examine two strategies - (1) noisy label correction and (2) training data augmentation - in empathy computing, an emerging task to predict psychology-based questionnaire outcomes from inputs like textual narratives. Crowdsourced datasets in this domain often suffer from noisy labels that misrepresent underlying empathy. We show that replacing or supplementing these crowdsourced labels with LLM-generated labels, developed using psychology-based scale-aware prompts, achieves statistically significant accuracy improvements. Notably, the RoBERTa pre-trained language model (PLM) trained with noise-reduced labels yields a state-of-the-art Pearson correlation coefficient of 0.648 on the public NewsEmp benchmarks. This paper further analyses evaluation metric selection and demographic biases to help guide the future development of more equitable empathy computing models. Code and LLM-generated labels are available at https://github.com/hasan-rakibul/LLMPathy.", "date": "2025-01-01", "categories": []}
{"id": "2509.15540", "title": "Beyond Words: Enhancing Desire, Emotion, and Sentiment Recognition with Non-Verbal Cues", "abstract": "Multimodal desire understanding, a task closely related to both emotion and sentiment that aims to infer human intentions from visual and textual cues, is an emerging yet underexplored task in affective computing with applications in social media analysis. Existing methods for related tasks predominantly focus on mining verbal cues, often overlooking the effective utilization of non-verbal cues embedded in images. To bridge this gap, we propose a Symmetrical Bidirectional Multimodal Learning Framework for Desire, Emotion, and Sentiment Recognition (SyDES). The core of SyDES is to achieve bidirectional fine-grained modal alignment between text and image modalities. Specifically, we introduce a mixed-scaled image strategy that combines global context from low-resolution images with fine-grained local features via masked image modeling (MIM) on high-resolution sub-images, effectively capturing intention-related visual representations. Then, we devise symmetrical cross-modal decoders, including a text-guided image decoder and an image-guided text decoder, which enable mutual reconstruction and refinement between modalities, facilitating deep cross-modal interaction. Furthermore, a set of dedicated loss functions is designed to harmonize potential conflicts between the MIM and modal alignment objectives during optimization. Extensive evaluations on the MSED benchmark demonstrate the superiority of our approach, which establishes a new state-of-the-art performance with 1.1% F1-score improvement in desire understanding. Consistent gains in emotion and sentiment recognition further validate its generalization ability and the necessity of utilizing non-verbal cues. Our code is available at: https://github.com/especiallyW/SyDES.", "date": "2025-09-19", "categories": []}
{"id": "2510.08586", "title": "Dynamic Stress Detection: A Study of Temporal Progression Modelling of Stress in Speech", "abstract": "Detecting psychological stress from speech is critical in high-pressure settings. While prior work has leveraged acoustic features for stress detection, most treat stress as a static label. In this work, we model stress as a temporally evolving phenomenon influenced by historical emotional state. We propose a dynamic labelling strategy that derives fine-grained stress annotations from emotional labels and introduce cross-attention-based sequential models, a Unidirectional LSTM and a Transformer Encoder, to capture temporal stress progression. Our approach achieves notable accuracy gains on MuSE (+5%) and StressID (+18%) over existing baselines, and generalises well to a custom real-world dataset. These results highlight the value of modelling stress as a dynamic construct in speech.", "date": "2025-10-02", "categories": []}
{"id": "2603.13340", "title": "Complementarity-Supervised Spectral-Band Routing for Multimodal Emotion Recognition", "abstract": "Multimodal emotion recognition fuses cues such as text, video, and audio to understand individual emotional states. Prior methods face two main limitations: mechanically relying on independent unimodal performance, thereby missing genuine complementary contributions, and coarse-grained fusion conflicting with the fine-grained representations required by emotion tasks. As inconsistent information density across heterogeneous modalities hinders inter-modal feature mining, we propose the Complementarity-Supervised Multi-Band Expert Network, named Atsuko, to model fine-grained complementary features via multi-scale band decomposition and expert collaboration. Specifically, we orthogonally decompose each modality's features into high, mid, and low-frequency components. Building upon this band-level routing, we design a modality-level router with a dual-path mechanism for fine-grained cross-band selection and cross-modal fusion. To mitigate shortcut learning from dominant modalities, we propose the Marginal Complementarity Module (MCM) to quantify performance loss when removing each modality via bi-modal comparison. The resulting complementarity distribution provides soft supervision, guiding the router to focus on modalities contributing unique information gains. Extensive experiments show our method achieves superior performance on the CMU-MOSI, CMU-MOSEI, CH-SIMS, CH-SIMSv2, and MIntRec benchmarks.", "date": "2026-03-07", "categories": []}
{"id": "2603.13415", "title": "Distance-aware Soft Prompt Learning for Multimodal Valence-Arousal Estimation", "abstract": "Valence-arousal (VA) estimation is crucial for capturing the nuanced nature of human emotions in naturalistic environments. While pre-trained Vision-Language models like CLIP have shown remarkable semantic alignment capabilities, their application in continuous regression tasks is often limited by the discrete nature of text prompts. In this paper, we propose a novel multimodal framework for VA estimation that introduces Distance-aware Soft Prompt Learning to bridge the gap between semantic space and continuous dimensions. Specifically, we partition the VA space into a 3X3 grid, defining nine emotional regions, each associated with distinct textual descriptions. Rather than a hard categorization, we employ a Gaussian kernel to compute soft labels based on the Euclidean distance between the ground truth coordinates and the region centers, allowing the model to learn fine-grained emotional transitions. For multimodal integration, our architecture utilizes a CLIP image encoder and an Audio Spectrogram Transformer (AST) to extract robust spatial and acoustic features. These features are temporally modeled via Gated Recurrent Units (GRUs) and integrated through a hierarchical fusion scheme that sequentially combines cross-modal attention for alignment and gated fusion for adaptive refinement. Experimental results on the Aff-Wild2 dataset demonstrate that our proposed semantic-guided approach significantly enhances the accuracy of VA estimation, achieving competitive performance in unconstrained ``in-the-wild'' scenarios.", "date": "2026-03-12", "categories": []}
{"id": "2603.13760", "title": "Multimodal Emotion Regression with Multi-Objective Optimization and VAD-Aware Audio Modeling for the 10th ABAW EMI Track", "abstract": "We participated in the 10th ABAW Challenge, focusing on the Emotional Mimicry Intensity (EMI) Estimation track on the Hume-Vidmimic2 dataset. This task aims to predict six continuous emotion dimensions: Admiration, Amusement, Determination, Empathic Pain, Excitement, and Joy. Through systematic multimodal exploration of pretrained high-level features, we found that, under our pretrained feature setting, direct feature concatenation outperformed the more complex fusion strategies we tested. This empirical finding motivated us to design a systematic approach built upon three core principles: (i) preserving modality-specific attributes through feature-level concatenation; (ii) improving training stability and metric alignment via multi-objective optimization; and (iii) enriching acoustic representations with a VAD-inspired latent prior. Our final framework integrates concatenation-based multimodal fusion, a shared six-dimensional regression head, multi-objective optimization with MSE, Pearson-correlation, and auxiliary branch supervision, EMA for parameter stabilization, and a VAD-inspired latent prior for the acoustic branch. On the official validation set, the proposed scheme achieved our best mean Pearson Correlation Coefficient of 0.478567.", "date": "2026-03-14", "categories": []}
{"id": "2603.14078", "title": "CMHL: Contrastive Multi-Head Learning for Emotionally Consistent Text Classification", "abstract": "Textual Emotion Classification (TEC) is one of the most difficult NLP tasks. State of the art approaches rely on Large language models (LLMs) and multi-model ensembles. In this study, we challenge the assumption that larger scale or more complex models are necessary for improved performance. In order to improve logical consistency, We introduce CMHL, a novel single-model architecture that explicitly models the logical structure of emotions through three key innovations: (1) multi-task learning that jointly predicts primary emotions, valence, and intensity, (2) psychologically-grounded auxiliary supervision derived from Russell's circumplex model, and (3) a novel contrastive contradiction loss that enforces emotional consistency by penalizing mutually incompatible predictions (e.g., simultaneous high confidence in joy and anger). With just 125M parameters, our model outperforms 56x larger LLMs and sLM ensembles with a new state-of-the-art F1 score of 93.75\\% compared to (86.13\\%-93.2\\%) on the dair-ai Emotion dataset. We further show cross domain generalization on the Reddit Suicide Watch and Mental Health Collection dataset (SWMH), outperforming domain-specific models like MentalBERT and MentalRoBERTa with an F1 score of 72.50\\% compared to (68.16\\%-72.16\\%) + a 73.30\\% recall compared to (67.05\\%-70.89\\%) that translates to enhanced sensitivity for detecting mental health distress. Our work establishes that architectural intelligence (not parameter count) drives progress in TEC. By embedding psychological priors and explicit consistency constraints, a well-designed single model can outperform both massive LLMs and complex ensembles, offering a efficient, interpretable, and clinically-relevant paradigm for affective computing.", "date": "2026-03-14", "categories": []}
{"id": "2603.14361", "title": "BROTHER: Behavioral Recognition Optimized Through Heterogeneous Ensemble Regularization for Ambivalence and Hesitancy", "abstract": "Recognizing complex behavioral states such as Ambivalence and Hesitancy (A/H) in naturalistic video settings remains a significant challenge in affective computing. Unlike basic facial expressions, A/H manifests as subtle, multimodal conflicts that require deep contextual and temporal understanding. In this paper, we propose a highly regularized, multimodal fusion pipeline to predict A/H at the video level. We extract robust unimodal features from visual, acoustic, and linguistic data, introducing a specialized statistical text modality explicitly designed to capture temporal speech variations and behavioral cues. To identify the most effective representations, we evaluate 15 distinct modality combinations across a committee of machine learning classifiers (MLP, Random Forest, and GBDT), selecting the most well-calibrated models based on validation Binary Cross-Entropy (BCE) loss. Furthermore, to optimally fuse these heterogeneous models without overfitting to the training distribution, we implement a Particle Swarm Optimization (PSO) hard-voting ensemble. The PSO fitness function dynamically incorporates a train-validation gap penalty (lambda) to actively suppress redundant or overfitted classifiers. Our comprehensive evaluation demonstrates that while linguistic features serve as the strongest independent predictor of A/H, our heavily regularized PSO ensemble (lambda = 0.2) effectively harnesses multimodal synergies, achieving a peak Macro F1-score of 0.7465 on the unseen test set. These results emphasize that treating ambivalence and hesitancy as a multimodal conflict, evaluated through an intelligently weighted committee, provides a robust framework for in-the-wild behavioral analysis.", "date": "2026-03-15", "categories": []}
{"id": "2603.14976", "title": "Anchoring Emotions in Text: Robust Multimodal Fusion for Mimicry Intensity Estimation", "abstract": "Estimating Emotional Mimicry Intensity (EMI) in naturalistic environments is a critical yet challenging task in affective computing. The primary difficulty lies in effectively modeling the complex, nonlinear temporal dynamics across highly heterogeneous modalities, especially when physical signals are corrupted or missing. To tackle this, we propose TAEMI (Text-Anchored Emotional Mimicry Intensity estimation), a novel multimodal framework designed for the 10th ABAW Competition. Motivated by the observation that continuous visual and acoustic signals are highly susceptible to transient environmental noise, we break the traditional symmetric fusion paradigm. Instead, we leverage textual transcript--which inherently encode a stable, time-independent semantic prior--as central anchors. Specifically, we introduce a Text-Anchored Dual Cross-Attention mechanism that utilizes these robust textual queries to actively filter out frame-level redundancies and align the noisy physical streams. Furthermore, to prevent catastrophic performance degradation caused by inevitably missing data in unconstrained real-world scenarios, we integrate Learnable Missing-Modality Tokens and a Modality Dropout strategy during training. Extensive experiments on the Hume-Vidmimic2 dataset demonstrate that TAEMI effectively captures fine-grained emotional variations and maintains robust predictive resilience under imperfect conditions. Our framework achieves a state-of-the-art mean Pearson correlation coefficient across six continuous emotional dimensions, significantly outperforming existing baseline methods.", "date": "2026-03-16", "categories": []}
{"id": "2407.03131", "title": "MVGT: A Multi-view Graph Transformer Based on Spatial Relations for EEG Emotion Recognition", "abstract": "Electroencephalography (EEG), a technique that records electrical activity from the scalp using electrodes, plays a vital role in affective computing. However, fully utilizing the multi-domain characteristics of EEG signals remains a significant challenge. Traditional single-perspective analyses often fail to capture the complex interplay of temporal, frequency, and spatial dimensions in EEG data. To address this, we introduce a multi-view graph transformer (MVGT) based on spatial relations that integrates information across three domains: temporal dynamics from continuous series, frequency features extracted from frequency bands, and inter-channel relationships captured through several spatial encodings. This comprehensive approach allows model to capture the nuanced properties inherent in EEG signals, enhancing its flexibility and representational power. Evaluation on publicly available datasets demonstrates that MVGT surpasses state-of-the-art methods in performance. The results highlight its ability to extract multi-domain information and effectively model inter-channel relationships, showcasing its potential for EEG-based emotion recognition tasks.", "date": "2024-07-03", "categories": []}
{"id": "2508.05880", "title": "Large language models show fragile cognitive reasoning about human emotions", "abstract": "Affective computing seeks to support the holistic development of artificial intelligence by enabling machines to engage with human emotion. Recent foundation models, particularly large language models (LLMs), have been trained and evaluated on emotion-related tasks, typically using supervised learning with discrete emotion labels. Such evaluations largely focus on surface phenomena, such as recognizing expressed or evoked emotions, leaving open whether these systems reason about emotion in cognitively meaningful ways. Here we ask whether LLMs can reason about emotions through underlying cognitive dimensions rather than labels alone. Drawing on cognitive appraisal theory, we introduce CoRE, a large-scale benchmark designed to probe the implicit cognitive structures LLMs use when interpreting emotionally charged situations. We assess alignment with human appraisal patterns, internal consistency, cross-model generalization, and robustness to contextual variation. We find that LLMs capture systematic relations between cognitive appraisals and emotions but show misalignment with human judgments and instability across contexts.", "date": "2025-08-07", "categories": []}
{"id": "2603.12693", "title": "HSEmotion Team at ABAW-10 Competition: Facial Expression Recognition, Valence-Arousal Estimation, Action Unit Detection and Fine-Grained Violence Classification", "abstract": "This article presents our results for the 10th Affective Behavior Analysis in-the-Wild (ABAW) competition. For frame-wise facial emotion understanding tasks (frame-wise facial expression recognition, valence-arousal estimation, action unit detection), we propose a fast approach based on facial embedding extraction with pre-trained EfficientNet-based emotion recognition models. If the latter model's confidence exceeds a threshold, its prediction is used. Otherwise, we feed embeddings into a simple multi-layered perceptron trained on the AffWild2 dataset. Estimated class-level scores are smoothed in a sliding window of fixed size to mitigate noise in frame-wise predictions. For the fine-grained violence detection task, we examine several pre-trained architectures for frame embeddings and their aggregation for video classification. Experimental results on four tasks from the ABAW challenge demonstrate that our approach significantly improves validation metrics over existing baselines.", "date": "2026-03-13", "categories": []}
{"id": "2603.12848", "title": "Team LEYA in 10th ABAW Competition: Multimodal Ambivalence/Hesitancy Recognition Approach", "abstract": "Ambivalence/hesitancy recognition in unconstrained videos is a challenging problem due to the subtle, multimodal, and context-dependent nature of this behavioral state. In this paper, a multimodal approach for video-level ambivalence/hesitancy recognition is presented for the 10th ABAW Competition. The proposed approach integrates four complementary modalities: scene, face, audio, and text. Scene dynamics are captured with a VideoMAE-based model, facial information is encoded through emotional frame-level embeddings aggregated by statistical pooling, acoustic representations are extracted with EmotionWav2Vec2.0 and processed by a Mamba-based temporal encoder, and linguistic cues are modeled using fine-tuned transformer-based text models. The resulting unimodal embeddings are further combined using multimodal fusion models, including prototype-augmented variants. Experiments on the BAH corpus demonstrate clear gains of multimodal fusion over all unimodal baselines. The best unimodal configuration achieved an average MF1 of 70.02%, whereas the best multimodal fusion model reached 83.25%. The highest final test performance, 71.43%, was obtained by an ensemble of five prototype-augmented fusion models. The obtained results highlight the importance of complementary multimodal cues and robust fusion strategies for ambivalence/hesitancy recognition.", "date": "2026-03-13", "categories": []}
{"id": "2603.13056", "title": "Team RAS in 10th ABAW Competition: Multimodal Valence and Arousal Estimation Approach", "abstract": "Continuous emotion recognition in terms of valence and arousal under in-the-wild (ITW) conditions remains a challenging problem due to large variations in appearance, head pose, illumination, occlusions, and subject-specific patterns of affective expression. We present a multimodal method for valence-arousal estimation ITW. Our method combines three complementary modalities: face, behavior, and audio. The face modality relies on GRADA-based frame-level embeddings and Transformer-based temporal regression. We use Qwen3-VL-4B-Instruct to extract behavior-relevant information from video segments, while Mamba is used to model temporal dynamics across segments. The audio modality relies on WavLM-Large with attention-statistics pooling and includes a cross-modal filtering stage to reduce the influence of unreliable or non-speech segments. To fuse modalities, we explore two fusion strategies: a Directed Cross-Modal Mixture-of-Experts Fusion Strategy that learns interactions between modalities with adaptive weighting, and a Reliability-Aware Audio-Visual Fusion Strategy that combines visual features at the frame-level while using audio as complementary context. The results are reported on the Aff-Wild2 dataset following the 10th Affective Behavior Analysis in-the-Wild (ABAW) challenge protocol. Experiments demonstrate that the proposed multimodal fusion strategy achieves a Concordance Correlation Coefficient (CCC) of 0.658 on the Aff-Wild2 development set.", "date": "2026-03-13", "categories": []}
{"id": "2603.11119", "title": "Group Resonance Network: Learnable Prototypes and Multi-Subject Resonance for EEG Emotion Recognition", "abstract": "Electroencephalography(EEG)-basedemotionrecognitionre- mains challenging in cross-subject settings due to severe inter-subject variability. Existing methods mainly learn subject-invariant features, but often under-exploit stimulus-locked group regularities shared across sub- jects. To address this issue, we propose the Group Resonance Network (GRN), which integrates individual EEG dynamics with offline group resonance modeling. GRN contains three components: an individual en- coder for band-wise EEG features, a set of learnable group prototypes for prototype-induced resonance, and a multi-subject resonance branch that encodes PLV/coherence-based synchrony with a small reference set. A resonance-aware fusion module combines individual and group-level rep- resentations for final classification. Experiments on SEED and DEAP under both subject-dependent and leave-one-subject-out protocols show that GRN consistently outperforms competitive baselines, while abla- tion studies confirm the complementary benefits of prototype learning and multi-subject resonance modeling.", "date": "2026-03-11", "categories": []}
{"id": "2603.11230", "title": "Monitoring and Prediction of Mood in Elderly People during Daily Life Activities", "abstract": "We present an intelligent wearable system to monitor and predict mood states of elderly people during their daily life activities. Our system is composed of a wristband to record different physiological activities together with a mobile app for ecological momentary assessment (EMA). Machine learning is used to train a classifier to automatically predict different mood states based on the smart band only. Our approach shows promising results on mood accuracy and provides results comparable with the state of the art in the specific detection of happiness and activeness.", "date": "2026-03-11", "categories": []}
{"id": "2603.11468", "title": "Stage-Adaptive Reliability Modeling for Continuous Valence-Arousal Estimation", "abstract": "Continuous valence-arousal estimation in real-world environments is challenging due to inconsistent modality reliability and interaction-dependent variability in audio-visual signals. Existing approaches primarily focus on modeling temporal dynamics, often overlooking the fact that modality reliability can vary substantially across interaction stages. To address this issue, we propose SAGE, a Stage-Adaptive reliability modeling framework that explicitly estimates and calibrates modality-wise confidence during multimodal integration. SAGE introduces a reliability-aware fusion mechanism that dynamically rebalances audio and visual representations according to their stage-dependent informativeness, preventing unreliable signals from dominating the prediction process. By separating reliability estimation from feature representation, the proposed framework enables more stable emotion estimation under cross-modal noise, occlusion, and varying interaction conditions. Extensive experiments on the Aff-Wild2 benchmark demonstrate that SAGE consistently improves concordance correlation coefficient scores compared with existing multimodal fusion approaches, highlighting the effectiveness of reliability-driven modeling for continuous affect prediction.", "date": "2026-03-12", "categories": []}
{"id": "2603.10043", "title": "AMB-DSGDN: Adaptive Modality-Balanced Dynamic Semantic Graph Differential Network for Multimodal Emotion Recognition", "abstract": "Multimodal dialogue emotion recognition captures emotional cues by fusing text, visual, and audio modalities. However, existing approaches still suffer from notable limitations in modeling emotional dependencies and learning multimodal representations. On the one hand, they are unable to effectively filter out redundant or noisy signals within multimodal features, which hinders the accurate capture of the dynamic evolution of emotional states across and within speakers. On the other hand, during multimodal feature learning, dominant modalities tend to overwhelm the fusion process, thereby suppressing the complementary contributions of non-dominant modalities such as speech and vision, ultimately constraining the overall recognition performance. To address these challenges, we propose an Adaptive Modality-Balanced Dynamic Semantic Graph Differential Network (AMB-DSGDN). Concretely, we first construct modality-specific subgraphs for text, speech, and vision, where each modality contains intra-speaker and inter-speaker graphs to capture both self-continuity and cross-speaker emotional dependencies. On top of these subgraphs, we introduce a differential graph attention mechanism, which computes the discrepancy between two sets of attention maps. By explicitly contrasting these attention distributions, the mechanism cancels out shared noise patterns while retaining modality-specific and context-relevant signals, thereby yielding purer and more discriminative emotional representations. In addition, we design an adaptive modality balancing mechanism, which estimates a dropout probability for each modality according to its relative contribution in emotion modeling.", "date": "2026-03-07", "categories": []}
{"id": "2603.08927", "title": "MEGC2026: Micro-Expression Grand Challenge on Visual Question Answering", "abstract": "Facial micro-expressions (MEs) are involuntary movements of the face that occur spontaneously when a person experiences an emotion but attempts to suppress or repress the facial expression, typically found in a high-stakes environment. In recent years, substantial advancements have been made in the areas of ME recognition, spotting, and generation. The emergence of multimodal large language models (MLLMs) and large vision-language models (LVLMs) offers promising new avenues for enhancing ME analysis through their powerful multimodal reasoning capabilities. The ME grand challenge (MEGC) 2026 introduces two tasks that reflect these evolving research directions: (1) ME video question answering (ME-VQA), which explores ME understanding through visual question answering on relatively short video sequences, leveraging MLLMs or LVLMs to address diverse question types related to MEs; and (2) ME long-video question answering (ME-LVQA), which extends VQA to long-duration video sequences in realistic settings, requiring models to handle temporal reasoning and subtle micro-expression detection across extended time periods. All participating algorithms are required to submit their results on a public leaderboard. More details are available at https://megc2026.github.io.", "date": "2026-03-09", "categories": []}
{"id": "2603.08936", "title": "VoxEmo: Benchmarking Speech Emotion Recognition with Speech LLMs", "abstract": "Speech Large Language Models (LLMs) show great promise for speech emotion recognition (SER) via generative interfaces. However, shifting from closed-set classification to open text generation introduces zero-shot stochasticity, making evaluation highly sensitive to prompts. Additionally, conventional speech LLMs benchmarks overlook the inherent ambiguity of human emotion. Hence, we present VoxEmo, a comprehensive SER benchmark encompassing 35 emotion corpora across 15 languages for Speech LLMs. VoxEmo provides a standardized toolkit featuring varying prompt complexities, from direct classification to paralinguistic reasoning. To reflect real-world perception/application, we introduce a distribution-aware soft-label protocol and a prompt-ensemble strategy that emulates annotator disagreement. Experiments reveal that while zero-shot speech LLMs trail supervised baselines in hard-label accuracy, they uniquely align with human subjective distributions.", "date": "2026-03-09", "categories": []}
{"id": "2603.09111", "title": "Progressive Representation Learning for Multimodal Sentiment Analysis with Incomplete Modalities", "abstract": "Multimodal Sentiment Analysis (MSA) seeks to infer human emotions by integrating textual, acoustic, and visual cues. However, existing approaches often rely on all modalities are completeness, whereas real-world applications frequently encounter noise, hardware failures, or privacy restrictions that result in missing modalities. There exists a significant feature misalignment between incomplete and complete modalities, and directly fusing them may even distort the well-learned representations of the intact modalities. To this end, we propose PRLF, a Progressive Representation Learning Framework designed for MSA under uncertain missing-modality conditions. PRLF introduces an Adaptive Modality Reliability Estimator (AMRE), which dynamically quantifies the reliability of each modality using recognition confidence and Fisher information to determine the dominant modality. In addition, the Progressive Interaction (ProgInteract) module iteratively aligns the other modalities with the dominant one, thereby enhancing cross-modal consistency while suppressing noise. Extensive experiments on CMU-MOSI, CMU-MOSEI, and SIMS verify that PRLF outperforms state-of-the-art methods across both inter- and intra-modality missing scenarios, demonstrating its robustness and generalization capability.", "date": "2026-03-10", "categories": []}
{"id": "2603.09874", "title": "MissBench: Benchmarking Multimodal Affective Analysis under Imbalanced Missing Modalities", "abstract": "Multimodal affective computing underpins key tasks such as sentiment analysis and emotion recognition. Standard evaluations, however, often assume that textual, acoustic, and visual modalities are equally available. In real applications, some modalities are systematically more fragile or expensive, creating imbalanced missing rates and training biases that task-level metrics alone do not reveal. We introduce MissBench, a benchmark and framework for multimodal affective tasks that standardizes both shared and imbalanced missing-rate protocols on four widely used sentiment and emotion datasets. MissBench also defines two diagnostic metrics. The Modality Equity Index (MEI) measures how fairly different modalities contribute across missing-modality configurations. The Modality Learning Index (MLI) quantifies optimization imbalance by comparing modality-specific gradient norms during training, aggregated across modality-related modules. Experiments on representative method families show that models that appear robust under shared missing rates can still exhibit marked modality inequity and optimization imbalance under imbalanced conditions. These findings position MissBench, together with MEI and MLI, as practical tools for stress-testing and analyzing multimodal affective models in realistic incomplete-modality settings.For reproducibility, we release our code at: https://anonymous.4open.science/r/MissBench-4098/", "date": "2026-03-10", "categories": []}
{"id": "2507.11202", "title": "A Robust Incomplete Multimodal Low-Rank Adaptation Approach for Emotion Recognition", "abstract": "Multimodal Emotion Recognition (MER) often encounters incomplete multimodality in practical applications due to sensor failures or privacy protection requirements. While existing methods attempt to address various incomplete multimodal scenarios by balancing the training of each modality combination through additional gradients, these approaches face a critical limitation: training gradients from different modality combinations conflict with each other, ultimately degrading the performance of the final prediction model. In this paper, we propose a unimodal decoupled dynamic low-rank adaptation method based on modality combinations, named MCULoRA, which is a novel framework for the parameter-efficient training of incomplete multimodal learning models. MCULoRA consists of two key modules, modality combination aware low-rank adaptation (MCLA) and dynamic parameter fine-tuning (DPFT). The MCLA module effectively decouples the shared information from the distinct characteristics of individual modality combinations. The DPFT module adjusts the training ratio of modality combinations based on the separability of each modality's representation space, optimizing the learning efficiency across different modality combinations. Our extensive experimental evaluation in multiple benchmark datasets demonstrates that MCULoRA substantially outperforms previous incomplete multimodal learning approaches in downstream task accuracy.", "date": "2025-07-15", "categories": []}
{"id": "2603.07093", "title": "Facial Expression Generation Aligned with Human Preference for Natural Dyadic Interaction", "abstract": "Achieving natural dyadic interaction requires generating facial expressions that are emotionally appropriate and socially aligned with human preference. Human feedback offers a compelling mechanism to guide such alignment, yet how to effectively incorporate this feedback into facial expression generation remains underexplored. In this paper, we propose a facial expression generation method aligned with human preference by leveraging human feedback to produce contextually and emotionally appropriate expressions for natural dyadic interaction. A key to our method is framing the generation of identity-independent facial expressions as an action learning process, allowing human feedback to assess their validity free from visual or identity bias. We establish a closed feedback loop in which listener expressions dynamically respond to evolving conversational cues of the speaker. Concretely, we train a vision-language-action model via supervised fine-tuning to map the speaker's multimodal signals into controllable low-dimensional expression representations of a 3D morphable model. We further introduce a human-feedback reinforcement learning strategy that integrates the imitation of high-quality expression response with critic-guided optimization. Experiments on two benchmarks demonstrate that our method effectively aligns facial expressions with human preference and achieves superior performance.", "date": "2026-03-07", "categories": []}
{"id": "2603.07138", "title": "Emotion Transcription in Conversation: A Benchmark for Capturing Subtle and Complex Emotional States through Natural Language", "abstract": "Emotion Recognition in Conversation (ERC) is critical for enabling natural human-machine interactions. However, existing methods predominantly employ categorical or dimensional emotion annotations, which often fail to adequately represent complex, subtle, or culturally specific emotional nuances. To overcome this limitation, we propose a novel task named Emotion Transcription in Conversation (ETC). This task focuses on generating natural language descriptions that accurately reflect speakers' emotional states within conversational contexts. To address the ETC, we constructed a Japanese dataset comprising text-based dialogues annotated with participants' self-reported emotional states, described in natural language. The dataset also includes emotion category labels for each transcription, enabling quantitative analysis and its application to ERC. We benchmarked baseline models, finding that while fine-tuning on our dataset enhances model performance, current models still struggle to infer implicit emotional states. The ETC task will encourage further research into more expressive emotion understanding in dialogue. The dataset is publicly available at https://github.com/UEC-InabaLab/ETCDataset.", "date": "2026-03-07", "categories": []}
{"id": "2603.08034", "title": "Solution to the 10th ABAW Expression Recognition Challenge: A Robust Multimodal Framework with Safe Cross-Attention and Modality Dropout", "abstract": "Emotion recognition in real-world environments is hindered by partial occlusions, missing modalities, and severe class imbalance. To address these issues, particularly for the Affective Behavior Analysis in-the-wild (ABAW) Expression challenge, we propose a multimodal framework that dynamically fuses visual and audio representations. Our approach uses a dual-branch Transformer architecture featuring a safe cross-attention mechanism and a modality dropout strategy. This design allows the network to rely on audio-based predictions when visual cues are absent. To mitigate the long-tail distribution of the Aff-Wild2 dataset, we apply focal loss optimization, combined with a sliding-window soft voting strategy to capture dynamic emotional transitions and reduce frame-level classification jitter. Experiments demonstrate that our framework effectively handles missing modalities and complex spatiotemporal dependencies, achieving an accuracy of 60.79% and an F1-score of 0.5029 on the Aff-Wild2 validation set.", "date": "2026-03-09", "categories": []}
{"id": "2603.08230", "title": "Disentangling Reasoning in Large Audio-Language Models for Ambiguous Emotion Prediction", "abstract": "Speech emotion recognition plays an important role in various applications. However, most existing approaches predict a single emotion label, oversimplifying the inherently ambiguous nature of human emotional expression. Recent large audio-language models show promise in generating richer outputs, but their reasoning ability for ambiguous emotional understanding remains limited. In this work, we reformulate ambiguous emotion recognition as a distributional reasoning problem and present the first systematic study of ambiguity-aware reasoning in LALMs. Our framework comprises two complementary components: an ambiguity-aware objective that aligns predictions with human perceptual distributions, and a structured ambiguity-aware chain-of-thought supervision that guides reasoning over emotional cues. Experiments on IEMOCAP and CREMA-D demonstrate consistent improvements across SFT, DPO, and GRPO training strategies.", "date": "2026-03-09", "categories": []}
{"id": "2603.06071", "title": "Text-Driven Emotionally Continuous Talking Face Generation", "abstract": "Talking Face Generation (TFG) strives to create realistic and emotionally expressive digital faces. While previous TFG works have mastered the creation of naturalistic facial movements, they typically express a fixed target emotion in synthetic videos and lack the ability to exhibit continuously changing and natural expressions like humans do when conveying information. To synthesize realistic videos, we propose a novel task called Emotionally Continuous Talking Face Generation (EC-TFG), which takes a text segment and an emotion description with varying emotions as driving data, aiming to generate a video where the person speaks the text while reflecting the emotional changes within the description. Alongside this, we introduce a customized model, i.e., Temporal-Intensive Emotion Modulated Talking Face Generation (TIE-TFG), which innovatively manages dynamic emotional variations by employing Temporal-Intensive Emotion Fluctuation Modeling, allowing it to provide emotion variation sequences corresponding to the input text to drive continuous facial expression changes in synthesized videos. Extensive evaluations demonstrate our method's exceptional ability to produce smooth emotion transitions and uphold high-quality visuals and motion authenticity across diverse emotional states.", "date": "2026-03-06", "categories": []}
{"id": "2512.04551", "title": "Multi-Loss Learning for Speech Emotion Recognition with Energy-Adaptive Mixup and Frame-Level Attention", "abstract": "Speech emotion recognition (SER) is an important technology in human-computer interaction. However, achieving high performance is challenging due to emotional complexity and scarce annotated data. To tackle these challenges, we propose a multi-loss learning (MLL) framework integrating an energy-adaptive mixup (EAM) method and a frame-level attention module (FLAM). The EAM method leverages SNR-based augmentation to generate diverse speech samples capturing subtle emotional variations. FLAM enhances frame-level feature extraction for multi-frame emotional cues. Our MLL strategy combines Kullback-Leibler divergence, focal, center, and supervised contrastive loss to optimize learning, address class imbalance, and improve feature separability. We evaluate our method on four widely used SER datasets: IEMOCAP, MSP-IMPROV, RAVDESS, and SAVEE. The results demonstrate our method achieves state-of-the-art performance, suggesting its effectiveness and robustness.", "date": "2025-12-04", "categories": []}
{"id": "2603.02695", "title": "Addressing Missing and Noisy Modalities in One Solution: Unified Modality-Quality Framework for Low-quality Multimodal Data", "abstract": "Multimodal data encountered in real-world scenarios are typically of low quality, with noisy modalities and missing modalities being typical forms that severely hinder model performance and robustness. However, prior works often handle noisy and missing modalities separately. In contrast, we jointly address missing and noisy modalities to enhance model robustness in low-quality data scenarios. We regard both noisy and missing modalities as a unified low-quality modality problem, and propose a unified modality-quality (UMQ) framework to enhance low-quality representations for multimodal affective computing. Firstly, we train a quality estimator with explicit supervised signals via a rank-guided training strategy that compares the relative quality of different representations by adding a ranking constraint, avoiding training noise caused by inaccurate absolute quality labels. Then, a quality enhancer for each modality is constructed, which uses the sample-specific information provided by other modalities and the modality-specific information provided by the defined modality baseline representation to enhance the quality of unimodal representations. Finally, we propose a quality-aware mixture-of-experts module with particular routing mechanism to enable multiple modality-quality problems to be addressed more specifically. UMQ consistently outperforms state-of-the-art baselines on multiple datasets under the settings of complete, missing, and noisy modalities.", "date": "2026-03-03", "categories": []}
{"id": "2602.00971", "title": "Unveiling the Cognitive Compass: Theory-of-Mind-Guided Multimodal Emotion Reasoning", "abstract": "Despite rapid progress in multimodal large language models (MLLMs), their capability for deep emotional understanding remains limited. We argue that genuine affective intelligence requires explicit modeling of Theory of Mind (ToM), the cognitive substrate from which emotions arise. To this end, we introduce HitEmotion, a ToM-grounded hierarchical benchmark that diagnoses capability breakpoints across increasing levels of cognitive depth. Second, we propose a ToM-guided reasoning chain that tracks mental states and calibrates cross-modal evidence to achieve faithful emotional reasoning. We further introduce TMPO, a reinforcement learning method that uses intermediate mental states as process-level supervision to guide and strengthen model reasoning. Extensive experiments show that HitEmotion exposes deep emotional reasoning deficits in state-of-the-art models, especially on cognitively demanding tasks. In evaluation, the ToM-guided reasoning chain and TMPO improve end-task accuracy and yield more faithful, more coherent rationales. In conclusion, our work provides the research community with a practical toolkit for evaluating and enhancing the cognition-based emotional understanding capabilities of MLLMs. Our dataset and code are available at: https://HitEmotion.github.io/.", "date": "2026-02-01", "categories": []}
{"id": "2603.00746", "title": "SpectroFusion-ViT: A Lightweight Transformer for Speech Emotion Recognition Using Harmonic Mel-Chroma Fusion", "abstract": "Speech is a natural means of conveying emotions, making it an effective method for understanding and representing human feelings. Reliable speech emotion recognition (SER) is central to applications in human-computer interaction, healthcare, education, and customer service. However, most SER methods depend on heavy backbone models or hand-crafted features that fail to balance accuracy and efficiency, particularly for low-resource languages like Bangla. In this work, we present SpectroFusion-ViT, a lightweight SER framework built utilizing EfficientViT-b0, a compact Vision Transformer architecture equipped with self-attention to capture long-range temporal and spectral patterns. The model contains only 2.04M parameters and requires 0.1 GFLOPs, enabling deployment in resource-constrained settings without compromising accuracy. Our pipeline first performs preprocessing and augmentation on raw audio, then extracts Chroma and Mel-frequency cepstral coefficient (MFCC) features. These representations are fused into a complementary time-frequency descriptor that preserves both fine-grained spectral detail and broader harmonic structure. Using transfer learning, EfficientViT-b0 is fine-tuned for multi-class emotion classification. We evaluate the system on two benchmark Bangla emotional speech datasets, SUBESCO and BanglaSER, which vary in speaker diversity, recording conditions, and acoustic characteristics. The proposed approach achieves 92.56% accuracy on SUBESCO and 82.19% on BanglaSER, surpassing existing state-of-the-art methods. These findings demonstrate that lightweight transformer architectures can deliver robust SER performance while remaining computationally efficient for real-world deployment.", "date": "2026-02-28", "categories": []}
{"id": "2602.23802", "title": "EMO-R3: Reflective Reinforcement Learning for Emotional Reasoning in Multimodal Large Language Models", "abstract": "Multimodal Large Language Models (MLLMs) have shown remarkable progress in visual reasoning and understanding tasks but still struggle to capture the complexity and subjectivity of human emotions. Existing approaches based on supervised fine-tuning often suffer from limited generalization and poor interpretability, while reinforcement learning methods such as Group Relative Policy Optimization fail to align with the intrinsic characteristics of emotional cognition. To address these challenges, we propose Reflective Reinforcement Learning for Emotional Reasoning (EMO-R3), a framework designed to enhance the emotional reasoning ability of MLLMs. Specifically, we introduce Structured Emotional Thinking to guide the model to perform step-by-step emotional reasoning in a structured and interpretable manner, and design a Reflective Emotional Reward that enables the model to re-evaluate its reasoning based on visual-text consistency and emotional coherence. Extensive experiments demonstrate that EMO-R3 significantly improves both the interpretability and emotional intelligence of MLLMs, achieving superior performance across multiple visual emotional understanding benchmarks.", "date": "2026-02-27", "categories": []}
{"id": "2602.23950", "title": "Micro-expression Recognition Based on Dual-branch Feature Extraction and Fusion", "abstract": "Micro-expressions, characterized by transience and subtlety, pose challenges to existing optical flow-based recognition methods. To address this, this paper proposes a dual-branch micro-expression feature extraction network integrated with parallel attention. Key contributions include: 1) a residual network designed to alleviate gradient anishing and network degradation; 2) an Inception network constructed to enhance model representation and suppress interference from irrelevant regions; 3) an adaptive feature fusion module developed to integrate dual-branch features. Experiments on the CASME II dataset demonstrate that the proposed method achieves 74.67% accuracy, outperforming LBP-TOP (by 11.26%), MSMMT (by 3.36%), and other comparative methods.", "date": "2026-02-27", "categories": []}
{"id": "2602.23300", "title": "A Mixture-of-Experts Model for Multimodal Emotion Recognition in Conversations", "abstract": "Emotion Recognition in Conversations (ERC) presents unique challenges, requiring models to capture the temporal flow of multi-turn dialogues and to effectively integrate cues from multiple modalities. We propose Mixture of Speech-Text Experts for Recognition of Emotions (MiSTER-E), a modular Mixture-of-Experts (MoE) framework designed to decouple two core challenges in ERC: modality-specific context modeling and multimodal information fusion. MiSTER-E leverages large language models (LLMs) fine-tuned for both speech and text to provide rich utterance-level embeddings, which are then enhanced through a convolutional-recurrent context modeling layer. The system integrates predictions from three experts-speech-only, text-only, and cross-modal-using a learned gating mechanism that dynamically weighs their outputs. To further encourage consistency and alignment across modalities, we introduce a supervised contrastive loss between paired speech-text representations and a KL-divergence-based regulariza-tion across expert predictions. Importantly, MiSTER-E does not rely on speaker identity at any stage. Experiments on three benchmark datasets-IEMOCAP, MELD, and MOSI-show that our proposal achieves 70.9%, 69.5%, and 87.9% weighted F1-scores respectively, outperforming several baseline speech-text ERC systems. We also provide various ablations to highlight the contributions made in the proposed approach.", "date": "2026-02-26", "categories": []}
{"id": "2503.15133", "title": "EmoGRACE: Aspect-based emotion analysis for social media data", "abstract": "While sentiment analysis has advanced from sentence to aspect-level, i.e., the identification of concrete terms related to a sentiment, the equivalent field of Aspect-based Emotion Analysis (ABEA) is faced with dataset bottlenecks and the increased complexity of emotion classes in contrast to binary sentiments. This paper addresses these gaps, by generating a first ABEA training dataset, consisting of 2,621 English Tweets, and fine-tuning a BERT-based model for the ABEA sub-tasks of Aspect Term Extraction (ATE) and Aspect Emotion Classification (AEC).\n  The dataset annotation process was based on the hierarchical emotion theory by Shaver et al. [1] and made use of group annotation and majority voting strategies to facilitate label consistency. The resulting dataset contained aspect-level emotion labels for Anger, Sadness, Happiness, Fear, and a None class. Using the new ABEA training dataset, the state-of-the-art ABSA model GRACE by Luo et al. [2] was fine-tuned for ABEA. The results reflected a performance plateau at an F1-score of 70.1% for ATE and 46.9% for joint ATE and AEC extraction. The limiting factors for model performance were broadly identified as the small training dataset size coupled with the increased task complexity, causing model overfitting and limited abilities to generalize well on new data.", "date": "2025-03-19", "categories": []}
{"id": "2602.21464", "title": "iMiGUE-Speech: A Spontaneous Speech Dataset for Affective Analysis", "abstract": "This work presents iMiGUE-Speech, an extension of the iMiGUE dataset that provides a spontaneous affective corpus for studying emotional and affective states. The new release focuses on speech and enriches the original dataset with additional metadata, including speech transcripts, speaker-role separation between interviewer and interviewee, and word-level forced alignments. Unlike existing emotional speech datasets that rely on acted or laboratory-elicited emotions, iMiGUE-Speech captures spontaneous affect arising naturally from real match outcomes. To demonstrate the utility of the dataset and establish initial benchmarks, we introduce two evaluation tasks for comparative assessment: speech emotion recognition and transcript-based sentiment analysis. These tasks leverage state-of-the-art pre-trained representations to assess the dataset's ability to capture spontaneous affective states from both acoustic and linguistic modalities. iMiGUE-Speech can also be synchronously paired with micro-gesture annotations from the original iMiGUE dataset, forming a uniquely multimodal resource for studying speech-gesture affective dynamics. The extended dataset is available at https://github.com/CV-AC/imigue-speech.", "date": "2026-02-25", "categories": []}
{"id": "2510.22129", "title": "egoEMOTION: Egocentric Vision and Physiological Signals for Emotion and Personality Recognition in Real-World Tasks", "abstract": "Understanding affect is central to anticipating human behavior, yet current egocentric vision benchmarks largely ignore the person's emotional states that shape their decisions and actions. Existing tasks in egocentric perception focus on physical activities, hand-object interactions, and attention modeling - assuming neutral affect and uniform personality. This limits the ability of vision systems to capture key internal drivers of behavior. In this paper, we present egoEMOTION, the first dataset that couples egocentric visual and physiological signals with dense self-reports of emotion and personality across controlled and real-world scenarios. Our dataset includes over 50 hours of recordings from 43 participants, captured using Meta's Project Aria glasses. Each session provides synchronized eye-tracking video, headmounted photoplethysmography, inertial motion data, and physiological baselines for reference. Participants completed emotion-elicitation tasks and naturalistic activities while self-reporting their affective state using the Circumplex Model and Mikels' Wheel as well as their personality via the Big Five model. We define three benchmark tasks: (1) continuous affect classification (valence, arousal, dominance); (2) discrete emotion classification; and (3) trait-level personality inference. We show that a classical learning-based method, as a simple baseline in real-world affect prediction, produces better estimates from signals captured on egocentric vision systems than processing physiological signals. Our dataset establishes emotion and personality as core dimensions in egocentric perception and opens new directions in affect-driven modeling of behavior, intent, and interaction.", "date": "2025-10-25", "categories": []}
{"id": "2602.20530", "title": "Memory-guided Prototypical Co-occurrence Learning for Mixed Emotion Recognition", "abstract": "Emotion recognition from multi-modal physiological and behavioral signals plays a pivotal role in affective computing, yet most existing models remain constrained to the prediction of singular emotions in controlled laboratory settings. Real-world human emotional experiences, by contrast, are often characterized by the simultaneous presence of multiple affective states, spurring recent interest in mixed emotion recognition as an emotion distribution learning problem. Current approaches, however, often neglect the valence consistency and structured correlations inherent among coexisting emotions. To address this limitation, we propose a Memory-guided Prototypical Co-occurrence Learning (MPCL) framework that explicitly models emotion co-occurrence patterns. Specifically, we first fuse multi-modal signals via a multi-scale associative memory mechanism. To capture cross-modal semantic relationships, we construct emotion-specific prototype memory banks, yielding rich physiological and behavioral representations, and employ prototype relation distillation to ensure cross-modal alignment in the latent prototype space. Furthermore, inspired by human cognitive memory systems, we introduce a memory retrieval strategy to extract semantic-level co-occurrence associations across emotion categories. Through this bottom-up hierarchical abstraction process, our model learns affectively informative representations for accurate emotion distribution prediction. Comprehensive experiments on two public datasets demonstrate that MPCL consistently outperforms state-of-the-art methods in mixed emotion recognition, both quantitatively and qualitatively.", "date": "2026-02-24", "categories": []}
{"id": "2601.16449", "title": "Emotion-LLaMAv2 and MMEVerse: A New Framework and Benchmark for Multimodal Emotion Understanding", "abstract": "Understanding human emotions from multimodal signals poses a significant challenge in affective computing and human-robot interaction. While multimodal large language models (MLLMs) have excelled in general vision-language tasks, their capabilities in emotional reasoning remain limited. The field currently suffers from a scarcity of large-scale datasets with high-quality, descriptive emotion annotations and lacks standardized benchmarks for evaluation. Our preliminary framework, Emotion-LLaMA, pioneered instruction-tuned multimodal learning for emotion reasoning but was restricted by explicit face detectors, implicit fusion strategies, and low-quality training data with limited scale. To address these limitations, we present Emotion-LLaMAv2 and the MMEVerse benchmark, establishing an end-to-end pipeline together with a standardized evaluation setting for emotion recognition and reasoning. Emotion-LLaMAv2 introduces three key advances. First, an end-to-end multiview encoder eliminates external face detection and captures nuanced emotional cues via richer spatial and temporal multiview tokens. Second, a Conv Attention pre-fusion module is designed to enable simultaneous local and global multimodal feature interactions external to the LLM backbone. Third, a perception-to-cognition curriculum instruction tuning scheme within the LLaMA2 backbone unifies emotion recognition and free-form emotion reasoning. To support large-scale training and reproducible evaluation, MMEVerse aggregates twelve publicly available emotion datasets, including IEMOCAP, MELD, DFEW, and MAFW, into a unified multimodal instruction format. The data are re-annotated via a multi-agent pipeline involving Qwen2 Audio, Qwen2.5 VL, and GPT 4o, producing 130k training clips and 36k testing clips across 18 evaluation benchmarks.", "date": "2026-01-23", "categories": []}
{"id": "2602.18763", "title": "TAG: Thinking with Action Unit Grounding for Facial Expression Recognition", "abstract": "Facial Expression Recognition (FER) is a fine-grained visual understanding task where reliable predictions require reasoning over localized and meaningful facial cues. Recent vision--language models (VLMs) enable natural language explanations for FER, but their reasoning is often ungrounded, producing fluent yet unverifiable rationales that are weakly tied to visual evidence and prone to hallucination, leading to poor robustness across different datasets. We propose TAG (Thinking with Action Unit Grounding), a vision--language framework that explicitly constrains multimodal reasoning to be supported by facial Action Units (AUs). TAG requires intermediate reasoning steps to be grounded in AU-related facial regions, yielding predictions accompanied by verifiable visual evidence. The model is trained via supervised fine-tuning on AU-grounded reasoning traces followed by reinforcement learning with an AU-aware reward that aligns predicted regions with external AU detectors. Evaluated on RAF-DB, FERPlus, and AffectNet, TAG consistently outperforms strong open-source and closed-source VLM baselines while simultaneously improving visual faithfulness. Ablation and preference studies further show that AU-grounded rewards stabilize reasoning and mitigate hallucination, demonstrating the importance of structured grounded intermediate representations for trustworthy multimodal reasoning in FER. The code will be available at https://github.com/would1920/FER_TAG .", "date": "2026-02-21", "categories": []}
{"id": "2602.19585", "title": "Tri-Subspaces Disentanglement for Multimodal Sentiment Analysis", "abstract": "Multimodal Sentiment Analysis (MSA) integrates language, visual, and acoustic modalities to infer human sentiment. Most existing methods either focus on globally shared representations or modality-specific features, while overlooking signals that are shared only by certain modality pairs. This limits the expressiveness and discriminative power of multimodal representations. To address this limitation, we propose a Tri-Subspace Disentanglement (TSD) framework that explicitly factorizes features into three complementary subspaces: a common subspace capturing global consistency, submodally-shared subspaces modeling pairwise cross-modal synergies, and private subspaces preserving modality-specific cues. To keep these subspaces pure and independent, we introduce a decoupling supervisor together with structured regularization losses. We further design a Subspace-Aware Cross-Attention (SACA) fusion module that adaptively models and integrates information from the three subspaces to obtain richer and more robust representations. Experiments on CMU-MOSI and CMU-MOSEI demonstrate that TSD achieves state-of-the-art performance across all key metrics, reaching 0.691 MAE on CMU-MOSI and 54.9% ACC-7 on CMU-MOSEI, and also transfers well to multimodal intent recognition tasks. Ablation studies confirm that tri-subspace disentanglement and SACA jointly enhance the modeling of multi-granular cross-modal sentiment cues.", "date": "2026-02-23", "categories": []}
{"id": "2602.18016", "title": "Towards LLM-centric Affective Visual Customization via Efficient and Precise Emotion Manipulating", "abstract": "Previous studies on visual customization primarily rely on the objective alignment between various control signals (e.g., language, layout and canny) and the edited images, which largely ignore the subjective emotional contents, and more importantly lack general-purpose foundation models for affective visual customization. With this in mind, this paper proposes an LLM-centric Affective Visual Customization (L-AVC) task, which focuses on generating images within modifying their subjective emotions via Multimodal LLM. Further, this paper contends that how to make the model efficiently align emotion conversion in semantics (named inter-emotion semantic conversion) and how to precisely retain emotion-agnostic contents (named exter-emotion semantic retaining) are rather important and challenging in this L-AVC task. To this end, this paper proposes an Efficient and Precise Emotion Manipulating approach for editing subjective emotions in images. Specifically, an Efficient Inter-emotion Converting (EIC) module is tailored to make the LLM efficiently align emotion conversion in semantics before and after editing, followed by a Precise Exter-emotion Retaining (PER) module to precisely retain the emotion-agnostic contents. Comprehensive experimental evaluations on our constructed L-AVC dataset demonstrate the great advantage of the proposed EPEM approach to the L-AVC task over several state-of-the-art baselines. This justifies the importance of emotion information for L-AVC and the effectiveness of EPEM in efficiently and precisely manipulating such information.", "date": "2026-02-20", "categories": []}
{"id": "2602.16256", "title": "Color-based Emotion Representation for Speech Emotion Recognition", "abstract": "Speech emotion recognition (SER) has traditionally relied on categorical or dimensional labels. However, this technique is limited in representing both the diversity and interpretability of emotions. To overcome this limitation, we focus on color attributes, such as hue, saturation, and value, to represent emotions as continuous and interpretable scores. We annotated an emotional speech corpus with color attributes via crowdsourcing and analyzed them. Moreover, we built regression models for color attributes in SER using machine learning and deep learning, and explored the multitask learning of color attribute regression and emotion classification. As a result, we demonstrated the relationship between color attributes and emotions in speech, and successfully developed color attribute regression models for SER. We also showed that multitask learning improved the performance of each task.", "date": "2026-02-18", "categories": []}
{"id": "2409.11733", "title": "Human-like Affective Cognition in Foundation Models", "abstract": "Understanding emotions is fundamental to human interaction and experience. Humans easily infer emotions from situations or facial expressions, situations from emotions, and do a variety of other affective cognition. How adept is modern AI at these inferences? We introduce an evaluation framework for testing affective cognition in foundation models. Starting from psychological theory, we generate 1,280 diverse scenarios exploring relationships between appraisals, emotions, expressions, and outcomes. We evaluate the abilities of foundation models (GPT-4, Claude-3, Gemini-1.5-Pro) and humans (N = 567) across carefully selected conditions. Our results show foundation models tend to agree with human intuitions, matching or exceeding interparticipant agreement. In some conditions, models are ``superhuman'' -- they better predict modal human judgements than the average human. All models benefit from chain-of-thought reasoning. This suggests foundation models have acquired a human-like understanding of emotions and their influence on beliefs and behavior.", "date": "2024-09-18", "categories": []}
{"id": "2601.01112", "title": "EmoLoom-2B: Fast Base-Model Screening for Emotion Classification and VAD with Lexicon-Weak Supervision and KV-Off Evaluation", "abstract": "We introduce EmoLoom-2B, a lightweight and reproducible pipeline that turns small language models under 2B parameters into fast screening candidates for joint emotion classification and Valence-Arousal-Dominance prediction. To ensure protocol-faithful and fair evaluation, we unify data loading, training, and inference under a single JSON input-output contract and remove avoidable variance by adopting KV-off decoding as the default setting. We incorporate two orthogonal semantic regularizers: a VAD-preserving constraint that aligns generated text with target VAD triples, and a lightweight external appraisal classifier that provides training-time guidance on goal attainment, controllability, certainty, and fairness without injecting long rationales. To improve polarity sensitivity, we introduce Valence Flip augmentation based on mirrored emotional pairs. During supervised fine-tuning, we apply A/B mixture sampling with entropy-aware temperature scheduling to balance coverage and convergence. Using Qwen-1.8B-Chat as the base model, EmoLoom-2B achieves strong performance on GoEmotions and EmpatheticDialogues, and demonstrates robust cross-corpus generalization on DailyDialog. The proposed recipe is budget-aware, auditable, and re-entrant, serving as a dependable screening pass before heavier training or multimodal fusion.", "date": "2026-01-03", "categories": []}
{"id": "2602.13259", "title": "Learning Physiology-Informed Vocal Spectrotemporal Representations for Speech Emotion Recognition", "abstract": "Speech emotion recognition (SER) is essential for humanoid robot tasks such as social robotic interactions and robotic psychological diagnosis, where interpretable and efficient models are critical for safety and performance. Existing deep models trained on large datasets remain largely uninterpretable, often insufficiently modeling underlying emotional acoustic signals and failing to capture and analyze the core physiology of emotional vocal behaviors. Physiological research on human voices shows that the dynamics of vocal amplitude and phase correlate with emotions through the vocal tract filter and the glottal source. However, most existing deep models solely involve amplitude but fail to couple the physiological features of and between amplitude and phase. Here, we propose PhysioSER, a physiology-informed vocal spectrotemporal representation learning method, to address these issues with a compact, plug-and-play design. PhysioSER constructs amplitude and phase views informed by voice anatomy and physiology (VAP) to complement SSL models for SER. This VAP-informed framework incorporates two parallel workflows: a vocal feature representation branch to decompose vocal signals based on VAP, embed them into a quaternion field, and use Hamilton-structured quaternion convolutions for modeling their dynamic interactions; and a latent representation branch based on a frozen SSL backbone. Then, utterance-level features from both workflows are aligned by a Contrastive Projection and Alignment framework, followed by a shallow attention fusion head for SER classification. PhysioSER is shown to be interpretable and efficient for SER through extensive evaluations across 14 datasets, 10 languages, and 6 backbones, and its practical efficacy is validated by real-time deployment on a humanoid robotic platform.", "date": "2026-02-03", "categories": []}
{"id": "2602.12714", "title": "ADEPT: RL-Aligned Agentic Decoding of Emotion via Evidence Probing Tools -- From Consensus Learning to Ambiguity-Driven Emotion Reasoning", "abstract": "Speech Large Language Models (SLLMs) enable high-level emotion reasoning but often produce ungrounded, text-biased judgments without verifiable acoustic evidence. In contrast, self-supervised speech encoders such as WavLM provide strong acoustic representations yet remain opaque discriminative models with limited interpretability. To bridge this gap, we introduce ADEPT (Agentic Decoding of Emotion via Evidence Probing Tools), a framework that reframes emotion recognition as a multi-turn inquiry process rather than a single-pass prediction. ADEPT transforms an SLLM into an agent that maintains an evolving candidate emotion set and adaptively invokes dedicated semantic and acoustic probing tools within a structured pipeline of candidate generation, evidence collection, and adjudication. Crucially, ADEPT enables a paradigm shift from consensus learning to ambiguity-driven emotion reasoning. Since human affect exhibits inherent complexity and frequent co-occurrence of emotions, we treat minority annotations as informative perceptual signals rather than discarding them as noise. Finally, we integrate Group Relative Policy Optimization (GRPO) with an Evidence Trust Gate to explicitly couple tool-usage behaviors with prediction quality and enforce evidence-grounded reasoning. Experiments show that ADEPT improves primary emotion accuracy in most settings while substantially improving minor emotion characterization, producing explanations grounded in auditable acoustic and semantic evidence.", "date": "2026-02-13", "categories": []}
{"id": "2510.10078", "title": "Improving Speech Emotion Recognition with Mutual Information Regularized Generative Model", "abstract": "Lack of large, well-annotated emotional speech corpora continues to limit the performance and robustness of speech emotion recognition (SER), particularly as models grow more complex and the demand for multimodal systems increases. While generative data augmentation offers a promising solution, existing approaches often produce emotionally inconsistent samples due to oversimplified conditioning on categorical labels. This paper introduces a novel mutual-information-regularised generative framework that combines cross-modal alignment with feature-level synthesis. Building on an InfoGAN-style architecture, our method first learns a semantically aligned audio-text representation space using pre-trained transformers and contrastive objectives. A feature generator is then trained to produce emotion-aware audio features while employing mutual information as a quantitative regulariser to ensure strong dependency between generated features and their conditioning variables. We extend this approach to multimodal settings, enabling the generation of novel, paired (audio, text) features. Comprehensive evaluation on three benchmark datasets (IEMOCAP, MSP-IMPROV, MSP-Podcast) demonstrates that our framework consistently outperforms existing augmentation methods, achieving state-of-the-art performance with improvements of up to 2.6% in unimodal SER and 3.2% in multimodal emotion recognition. Most importantly, we demonstrate that mutual information functions as both a regulariser and a measurable metric for generative quality, offering a systematic approach to data augmentation in affective computing.", "date": "2025-10-11", "categories": []}
{"id": "2602.11658", "title": "EmoSpace: Fine-Grained Emotion Prototype Learning for Immersive Affective Content Generation", "abstract": "Emotion is important for creating compelling virtual reality (VR) content. Although some generative methods have been applied to lower the barrier to creating emotionally rich content, they fail to capture the nuanced emotional semantics and the fine-grained control essential for immersive experiences. To address these limitations, we introduce EmoSpace, a novel framework for emotion-aware content generation that learns dynamic, interpretable emotion prototypes through vision-language alignment. We employ a hierarchical emotion representation with rich learnable prototypes that evolve during training, enabling fine-grained emotional control without requiring explicit emotion labels. We develop a controllable generation pipeline featuring multi-prototype guidance, temporal blending, and attention reweighting that supports diverse applications, including emotional image outpainting, stylized generation, and emotional panorama generation for VR environments. Our experiments demonstrate the superior performance of EmoSpace over existing methods in both qualitative and quantitative evaluations. Additionally, we present a comprehensive user study investigating how VR environments affect emotional perception compared to desktop settings. Our work facilitates immersive visual content generation with fine-grained emotion control and supports applications like therapy, education, storytelling, artistic creation, and cultural preservation. Code and models will be made publicly available.", "date": "2026-02-12", "categories": []}
{"id": "2504.09221", "title": "CMCRD: Cross-Modal Contrastive Representation Distillation for Emotion Recognition", "abstract": "Emotion recognition is an important component of affective computing, and also human-machine interaction. Unimodal emotion recognition is convenient, but the accuracy may not be high enough; on the contrary, multi-modal emotion recognition may be more accurate, but it also increases the complexity and cost of the data collection system. This paper considers cross-modal emotion recognition, i.e., using both electroencephalography (EEG) and eye movement in training, but only EEG or eye movement in test. We propose cross-modal contrastive representation distillation (CMCRD), which uses a pre-trained eye movement classification model to assist the training of an EEG classification model, improving feature extraction from EEG signals, or vice versa. During test, only EEG signals (or eye movement signals) are acquired, eliminating the need for multi-modal data. CMCRD not only improves the emotion recognition accuracy, but also makes the system more simplified and practical. Experiments using three different neural network architectures on three multi-modal emotion recognition datasets demonstrated the effectiveness of CMCRD. Compared with the EEG-only model, it improved the average classification accuracy by about 6.2%.", "date": "2025-04-12", "categories": []}
{"id": "2508.09210", "title": "MME-Emotion: A Holistic Evaluation Benchmark for Emotional Intelligence in Multimodal Large Language Models", "abstract": "Recent advances in multimodal large language models (MLLMs) have catalyzed transformative progress in affective computing, enabling models to exhibit emergent emotional intelligence. Despite substantial methodological progress, current emotional benchmarks remain limited, as it is still unknown: (a) the generalization abilities of MLLMs across distinct scenarios, and (b) their reasoning capabilities to identify the triggering factors behind emotional states. To bridge these gaps, we present \\textbf{MME-Emotion}, a systematic benchmark that assesses both emotional understanding and reasoning capabilities of MLLMs, enjoying \\textit{scalable capacity}, \\textit{diverse settings}, and \\textit{unified protocols}. As the largest emotional intelligence benchmark for MLLMs, MME-Emotion contains over 6,000 curated video clips with task-specific questioning-answering (QA) pairs, spanning broad scenarios to formulate eight emotional tasks. It further incorporates a holistic evaluation suite with hybrid metrics for emotion recognition and reasoning, analyzed through a multi-agent system framework. Through a rigorous evaluation of 20 advanced MLLMs, we uncover both their strengths and limitations, yielding several key insights: \\ding{182} Current MLLMs exhibit unsatisfactory emotional intelligence, with the best-performing model achieving only $39.3\\%$ recognition score and $56.0\\%$ Chain-of-Thought (CoT) score on our benchmark. \\ding{183} Generalist models (\\emph{e.g.}, Gemini-2.5-Pro) derive emotional intelligence from generalized multimodal understanding capabilities, while specialist models (\\emph{e.g.}, R1-Omni) can achieve comparable performance through domain-specific post-training adaptation. By introducing MME-Emotion, we hope that it can serve as a foundation for advancing MLLMs' emotional intelligence in the future.", "date": "2025-08-11", "categories": []}
{"id": "2602.05471", "title": "Reasoning under Ambiguity: Uncertainty-Aware Multilingual Emotion Classification under Partial Supervision", "abstract": "Contemporary knowledge-based systems increasingly rely on multilingual emotion identification to support intelligent decision-making, yet they face major challenges due to emotional ambiguity and incomplete supervision. Emotion recognition from text is inherently uncertain because multiple emotional states often co-occur and emotion annotations are frequently missing or heterogeneous. Most existing multi-label emotion classification methods assume fully observed labels and rely on deterministic learning objectives, which can lead to biased learning and unreliable predictions under partial supervision. This paper introduces Reasoning under Ambiguity, an uncertainty-aware framework for multilingual multi-label emotion classification that explicitly aligns learning with annotation uncertainty. The proposed approach uses a shared multilingual encoder with language-specific optimization and an entropy-based ambiguity weighting mechanism that down-weights highly ambiguous training instances rather than treating missing labels as negative evidence. A mask-aware objective with positive-unlabeled regularization is further incorporated to enable robust learning under partial supervision. Experiments on English, Spanish, and Arabic emotion classification benchmarks demonstrate consistent improvements over strong baselines across multiple evaluation metrics, along with improved training stability, robustness to annotation sparsity, and enhanced interpretability.", "date": "2026-02-05", "categories": []}
{"id": "2512.19379", "title": "OmniMER: Auxiliary-Enhanced LLM Adaptation for Indonesian Multimodal Emotion Recognition", "abstract": "Indonesian, spoken by over 200 million people, remains underserved in multimodal emotion recognition research despite its dominant presence on Southeast Asian social media platforms. We introduce IndoMER, the first multimodal emotion recognition benchmark for Indonesian, comprising 1,944 video segments from 203 speakers with temporally aligned text, audio, and visual annotations across seven emotion categories. The dataset exhibits realistic challenges including cross-modal inconsistency and long-tailed class distributions shaped by Indonesian cultural communication norms. To address these challenges, we propose OmniMER, a multimodal adaptation framework built upon Qwen2.5-Omni that enhances emotion recognition through three auxiliary modality-specific perception tasks: emotion keyword extraction for text, facial expression analysis for video, and prosody analysis for audio. These auxiliary tasks help the model identify emotion-relevant cues in each modality before fusion, reducing reliance on spurious correlations in low-resource settings. Experiments on IndoMER show that OmniMER achieves 0.582 Macro-F1 on sentiment classification and 0.454 on emotion recognition, outperforming the base model by 7.6 and 22.1 absolute points respectively. Cross-lingual evaluation on the Chinese CH-SIMS dataset further demonstrates the generalizability of the proposed framework. The dataset and code are publicly available. https://github.com/yanxm01/INDOMER", "date": "2025-12-22", "categories": []}
{"id": "2602.09121", "title": "Uncertainty-Aware Multimodal Emotion Recognition through Dirichlet Parameterization", "abstract": "In this work, we present a lightweight and privacy-preserving Multimodal Emotion Recognition (MER) framework designed for deployment on edge devices. To demonstrate framework's versatility, our implementation uses three modalities - speech, text and facial imagery. However, the system is fully modular, and can be extended to support other modalities or tasks. Each modality is processed through a dedicated backbone optimized for inference efficiency: Emotion2Vec for speech, a ResNet-based model for facial expressions, and DistilRoBERTa for text. To reconcile uncertainty across modalities, we introduce a model- and task-agnostic fusion mechanism grounded in Dempster-Shafer theory and Dirichlet evidence. Operating directly on model logits, this approach captures predictive uncertainty without requiring additional training or joint distribution estimation, making it broadly applicable beyond emotion recognition. Validation on five benchmark datasets (eNTERFACE05, MEAD, MELD, RAVDESS and CREMA-D) show that our method achieves competitive accuracy while remaining computationally efficient and robust to ambiguous or missing inputs. Overall, the proposed framework emphasizes modularity, scalability, and real-world feasibility, paving the way toward uncertainty-aware multimodal systems for healthcare, human-computer interaction, and other emotion-informed applications.", "date": "2026-02-09", "categories": []}
{"id": "2602.06997", "title": "Adaptive Temporal Dynamics for Personalized Emotion Recognition: A Liquid Neural Network Approach", "abstract": "Emotion recognition from physiological signals remains challenging due to their non-stationary, noisy, and subject-dependent characteristics. This work presents, to the best of our knowledge, the first comprehensive application of liquid neural networks for EEG-based emotion recognition. The proposed multimodal framework combines convolutional feature extraction, liquid neural networks with learnable time constants, and attention-guided fusion to model temporal EEG dynamics with complementary peripheral physiological and personality features. Dedicated subnetworks are used to process EEG features and auxiliary modalities, and a shared autoencoder-based fusion module is used to learn discriminative latent representations before classification. Subject-dependent experiments conducted on the PhyMER dataset across seven emotional classes achieve an accuracy of 95.45%, surpassing previously reported results. Furthermore, temporal attention analysis provides interpretable insights into emotion-specific temporal relevance, and t-SNE visualizations demonstrate enhanced class separability, highlighting the effectiveness of the proposed approach. Finally, statistical analysis of temporal dynamics confirms that the network self-organizes into distinct functional groups with specialized fast and slow neurons, proving it independently tunes learnable time constants and memory dominance to effectively capture complex emotion artifacts.", "date": "2026-01-28", "categories": []}
{"id": "2602.07054", "title": "AVERE: Improving Audiovisual Emotion Reasoning with Preference Optimization", "abstract": "Emotion understanding is essential for building socially intelligent agents. Although recent multimodal large language models have shown strong performance on this task, two key challenges remain - spurious associations between emotions and irrelevant audiovisual cues, and hallucinations of audiovisual cues driven by text priors in the language model backbone. To quantify and understand these issues, we introduce EmoReAlM, a benchmark designed to evaluate MLLMs for cue-emotion associations, hallucinations and modality agreement. We then propose AVEm-DPO, a preference optimization technique that aligns model responses with both audiovisual inputs and emotion-centric queries. Specifically, we construct preferences over responses exhibiting spurious associations or hallucinations, and audiovisual input pairs guided by textual prompts. We also include a regularization term that penalizes reliance on text priors, thereby mitigating modality-specific cue hallucinations. Experimental results on DFEW, RAVDESS and EMER demonstrate that our method significantly improves the performance of the reference baseline models with 6-19% of relative performance gains in zero-shot settings. By providing both a rigorous benchmark and a robust optimization framework, this work enables principled evaluation and improvement of MLLMs for emotion understanding and social AI. Code, models and benchmark will be released at https://avere-iclr.github.io.", "date": "2026-02-04", "categories": []}
{"id": "2602.06778", "title": "Revisiting Emotions Representation for Recognition in the Wild", "abstract": "Facial emotion recognition has been typically cast as a single-label classification problem of one out of six prototypical emotions. However, that is an oversimplification that is unsuitable for representing the multifaceted spectrum of spontaneous emotional states, which are most often the result of a combination of multiple emotions contributing at different intensities. Building on this, a promising direction that was explored recently is to cast emotion recognition as a distribution learning problem. Still, such approaches are limited in that research datasets are typically annotated with a single emotion class. In this paper, we contribute a novel approach to describe complex emotional states as probability distributions over a set of emotion classes. To do so, we propose a solution to automatically re-label existing datasets by exploiting the result of a study in which a large set of both basic and compound emotions is mapped to probability distributions in the Valence-Arousal-Dominance (VAD) space. In this way, given a face image annotated with VAD values, we can estimate the likelihood of it belonging to each of the distributions, so that emotional states can be described as a mixture of emotions, enriching their description, while also accounting for the ambiguous nature of their perception. In a preliminary set of experiments, we illustrate the advantages of this solution and a new possible direction of investigation. Data annotations are available at https://github.com/jbcnrlz/affectnet-b-annotation.", "date": "2026-02-06", "categories": []}
{"id": "2602.05496", "title": "XEmoGPT: An Explainable Multimodal Emotion Recognition Framework with Cue-Level Perception and Reasoning", "abstract": "Explainable Multimodal Emotion Recognition plays a crucial role in applications such as human-computer interaction and social media analytics. However, current approaches struggle with cue-level perception and reasoning due to two main challenges: 1) general-purpose modality encoders are pretrained to capture global structures and general semantics rather than fine-grained emotional cues, resulting in limited sensitivity to emotional signals; and 2) available datasets usually involve a trade-off between annotation quality and scale, which leads to insufficient supervision for emotional cues and ultimately limits cue-level reasoning. Moreover, existing evaluation metrics are inadequate for assessing cue-level reasoning performance. To address these challenges, we propose eXplainable Emotion GPT (XEmoGPT), a novel EMER framework capable of both perceiving and reasoning over emotional cues. It incorporates two specialized modules: the Video Emotional Cue Bridge (VECB) and the Audio Emotional Cue Bridge (AECB), which enhance the video and audio encoders through carefully designed tasks for fine-grained emotional cue perception. To further support cue-level reasoning, we construct a large-scale dataset, EmoCue, designed to teach XEmoGPT how to reason over multimodal emotional cues. In addition, we introduce EmoCue-360, an automated metric that extracts and matches emotional cues using semantic similarity, and release EmoCue-Eval, a benchmark of 400 expert-annotated samples covering diverse emotional scenarios. Experimental results show that XEmoGPT achieves strong performance in both emotional cue perception and reasoning.", "date": "2026-02-05", "categories": []}
{"id": "2507.08499", "title": "PromotionGo at SemEval-2025 Task 11: A Feature-Centric Framework for Cross-Lingual Multi-Emotion Detection in Short Texts", "abstract": "This paper presents our system for SemEval 2025 Task 11: Bridging the Gap in Text-Based Emotion Detection (Track A), which focuses on multi-label emotion detection in short texts. We propose a feature-centric framework that dynamically adapts document representations and learning algorithms to optimize language-specific performance. Our study evaluates three key components: document representation, dimensionality reduction, and model training in 28 languages, highlighting five for detailed analysis. The results show that TF-IDF remains highly effective for low-resource languages, while contextual embeddings like FastText and transformer-based document representations, such as those produced by Sentence-BERT, exhibit language-specific strengths. Principal Component Analysis (PCA) reduces training time without compromising performance, particularly benefiting FastText and neural models such as Multi-Layer Perceptrons (MLP). Computational efficiency analysis underscores the trade-off between model complexity and processing cost. Our framework provides a scalable solution for multilingual emotion detection, addressing the challenges of linguistic diversity and resource constraints.", "date": "2025-07-11", "categories": []}
{"id": "2507.11384", "title": "Addressing Data Imbalance in Transformer-Based Multi-Label Emotion Detection with Weighted Loss", "abstract": "This paper explores the application of a simple weighted loss function to Transformer-based models for multi-label emotion detection in SemEval-2025 Shared Task 11. Our approach addresses data imbalance by dynamically adjusting class weights, thereby enhancing performance on minority emotion classes without the computational burden of traditional resampling methods. We evaluate BERT, RoBERTa, and BART on the BRIGHTER dataset, using evaluation metrics such as Micro F1, Macro F1, ROC-AUC, Accuracy, and Jaccard similarity coefficients. The results demonstrate that the weighted loss function improves performance on high-frequency emotion classes but shows limited impact on minority classes. These findings underscore both the effectiveness and the challenges of applying this approach to imbalanced multi-label emotion detection.", "date": "2025-07-15", "categories": []}
{"id": "2509.25458", "title": "Plug-and-Play Emotion Graphs for Compositional Prompting in Zero-Shot Speech Emotion Recognition", "abstract": "Large audio-language models (LALMs) exhibit strong zero-shot performance across speech tasks but struggle with speech emotion recognition (SER) due to weak paralinguistic modeling and limited cross-modal reasoning. We propose Compositional Chain-of-Thought Prompting for Emotion Reasoning (CCoT-Emo), a framework that introduces structured Emotion Graphs (EGs) to guide LALMs in emotion inference without fine-tuning. Each EG encodes seven acoustic features (e.g., pitch, speech rate, jitter, shimmer), textual sentiment, keywords, and cross-modal associations. Embedded into prompts, EGs provide interpretable and compositional representations that enhance LALM reasoning. Experiments across SER benchmarks show that CCoT-Emo outperforms prior SOTA and improves accuracy over zero-shot baselines.", "date": "2025-09-29", "categories": []}
{"id": "2601.18321", "title": "Integrating Fine-Grained Audio-Visual Evidence for Robust Multimodal Emotion Reasoning", "abstract": "Multimodal emotion analysis is shifting from static classification to generative reasoning. Beyond simple label prediction, robust affective reasoning must synthesize fine-grained signals such as facial micro-expressions and prosodic which shifts to decode the latent causality within complex social contexts. However, current Multimodal Large Language Models (MLLMs) face significant limitations in fine-grained perception, primarily due to data scarcity and insufficient cross-modal fusion. As a result, these models often exhibit unimodal dominance which leads to hallucinations in complex multimodal interactions, particularly when visual and acoustic cues are subtle, ambiguous, or even contradictory (e.g., in sarcastic scenery). To address this, we introduce SABER-LLM, a framework designed for robust multimodal reasoning. First, we construct SABER, a large-scale emotion reasoning dataset comprising 600K video clips, annotated with a novel six-dimensional schema that jointly captures audiovisual cues and causal logic. Second, we propose the structured evidence decomposition paradigm, which enforces a \"perceive-then-reason\" separation between evidence extraction and reasoning to alleviate unimodal dominance. The ability to perceive complex scenes is further reinforced by consistency-aware direct preference optimization, which explicitly encourages alignment among modalities under ambiguous or conflicting perceptual conditions. Experiments on EMER, EmoBench-M, and SABER-Test demonstrate that SABER-LLM significantly outperforms open-source baselines and achieves robustness competitive with closed-source models in decoding complex emotional dynamics. The dataset and model are available at https://github.com/zxzhao0/SABER-LLM.", "date": "2026-01-26", "categories": []}
{"id": "2602.03873", "title": "Decoding Ambiguous Emotions with Test-Time Scaling in Audio-Language Models", "abstract": "Emotion recognition from human speech is a critical enabler for socially aware conversational AI. However, while most prior work frames emotion recognition as a categorical classification problem, real-world affective states are often ambiguous, overlapping, and context-dependent, posing significant challenges for both annotation and automatic modeling. Recent large-scale audio language models (ALMs) offer new opportunities for nuanced affective reasoning without explicit emotion supervision, but their capacity to handle ambiguous emotions remains underexplored. At the same time, advances in inference-time techniques such as test-time scaling (TTS) have shown promise for improving generalization and adaptability in hard NLP tasks, but their relevance to affective computing is still largely unknown. In this work, we introduce the first benchmark for ambiguous emotion recognition in speech with ALMs under test-time scaling. Our evaluation systematically compares eight state-of-the-art ALMs and five TTS strategies across three prominent speech emotion datasets. We further provide an in-depth analysis of the interaction between model capacity, TTS, and affective ambiguity, offering new insights into the computational and representational challenges of ambiguous emotion understanding. Our benchmark establishes a foundation for developing more robust, context-aware, and emotionally intelligent speech-based AI systems, and highlights key future directions for bridging the gap between model assumptions and the complexity of real-world human emotion.", "date": "2026-02-01", "categories": []}
{"id": "2602.04112", "title": "DELTA: Deliberative Multi-Agent Reasoning with Reinforcement Learning for Multimodal Psychological Counseling", "abstract": "Psychological counseling is a fundamentally multimodal cognitive process in which clinicians integrate verbal content with visual and vocal cues to infer clients' mental states and respond empathically. However, most existing language-model-based counseling systems operate on text alone and rely on implicit mental state inference. We introduce DELTA, a deliberative multi-agent framework that models counseling as a structured reasoning process over multimodal signals, separating evidence grounding, mental state abstraction, and response generation. DELTA further incorporates reinforcement learning guided by a distribution-level Emotion Attunement Score to encourage emotionally attuned responses. Experiments on a multimodal counseling benchmark show that DELTA improves both counseling quality and emotion attunement across models. Ablation and qualitative analyses suggest that explicit multimodal reasoning and structured mental state representations play complementary roles in supporting empathic human-AI interaction.", "date": "2026-02-04", "categories": []}
{"id": "2602.04247", "title": "DementiaBank-Emotion: A Multi-Rater Emotion Annotation Corpus for Alzheimer's Disease Speech (Version 1.0)", "abstract": "We present DementiaBank-Emotion, the first multi-rater emotion annotation corpus for Alzheimer's disease (AD) speech. Annotating 1,492 utterances from 108 speakers for Ekman's six basic emotions and neutral, we find that AD patients express significantly more non-neutral emotions (16.9%) than healthy controls (5.7%; p < .001). Exploratory acoustic analysis suggests a possible dissociation: control speakers showed substantial F0 modulation for sadness (Delta = -3.45 semitones from baseline), whereas AD speakers showed minimal change (Delta = +0.11 semitones; interaction p = .023), though this finding is based on limited samples (sadness: n=5 control, n=15 AD) and requires replication. Within AD speech, loudness differentiates emotion categories, indicating partially preserved emotion-prosody mappings. We release the corpus, annotation guidelines, and calibration workshop materials to support research on emotion recognition in clinical populations.", "date": "2026-02-04", "categories": []}
{"id": "2602.04260", "title": "Decoupled Hierarchical Distillation for Multimodal Emotion Recognition", "abstract": "Human multimodal emotion recognition (MER) seeks to infer human emotions by integrating information from language, visual, and acoustic modalities. Although existing MER approaches have achieved promising results, they still struggle with inherent multimodal heterogeneities and varying contributions from different modalities. To address these challenges, we propose a novel framework, Decoupled Hierarchical Multimodal Distillation (DHMD). DHMD decouples each modality's features into modality-irrelevant (homogeneous) and modality-exclusive (heterogeneous) components using a self-regression mechanism. The framework employs a two-stage knowledge distillation (KD) strategy: (1) coarse-grained KD via a Graph Distillation Unit (GD-Unit) in each decoupled feature space, where a dynamic graph facilitates adaptive distillation among modalities, and (2) fine-grained KD through a cross-modal dictionary matching mechanism, which aligns semantic granularities across modalities to produce more discriminative MER representations. This hierarchical distillation approach enables flexible knowledge transfer and effectively improves cross-modal feature alignment. Experimental results demonstrate that DHMD consistently outperforms state-of-the-art MER methods, achieving 1.3\\%/2.4\\% (ACC$_7$), 1.3\\%/1.9\\% (ACC$_2$) and 1.9\\%/1.8\\% (F1) relative improvement on CMU-MOSI/CMU-MOSEI dataset, respectively. Meanwhile, visualization results reveal that both the graph edges and dictionary activations in DHMD exhibit meaningful distribution patterns across modality-irrelevant/-exclusive feature spaces.", "date": "2026-02-04", "categories": []}
{"id": "2602.02784", "title": "Cross-Temporal Attention Fusion (CTAF) for Multimodal Physiological Signals in Self-Supervised Learning", "abstract": "We study multimodal affect modeling when EEG and peripheral physiology are asynchronous, which most fusion methods ignore or handle with costly warping. We propose Cross-Temporal Attention Fusion (CTAF), a self-supervised module that learns soft bidirectional alignments between modalities and builds a robust clip embedding using time-aware cross attention, a lightweight fusion gate, and alignment-regularized contrastive objectives with optional weak supervision. On the K-EmoCon dataset, under leave-one-out cross-validation evaluation, CTAF yields higher cosine margins for matched pairs and better cross-modal token retrieval within one second, and it is competitive with the baseline on three-bin accuracy and macro-F1 while using few labels. Our contributions are a time-aware fusion mechanism that directly models correspondence, an alignment-driven self-supervised objective tailored to EEG and physiology, and an evaluation protocol that measures alignment quality itself. Our approach accounts for the coupling between the central and autonomic nervous systems in psychophysiological time series. These results indicate that CTAF is a strong step toward label-efficient, generalizable EEG-peripheral fusion under temporal asynchrony.", "date": "2026-02-02", "categories": []}
{"id": "2409.00143", "title": "Invariant Representation Guided Multimodal Sentiment Decoding with Sequential Variation Regularization", "abstract": "Achieving consistent sentiment representation across diverse modalities remains a key challenge in multimodal sentiment analysis. However, rapid emotional fluctuations over time often introduce instability, leading to compromised prediction performance. To address this challenge, we propose a robust sentiment representation dual enhancement strategy that simultaneously enhances the temporal and modality dimensions, guided by targeted mechanisms in both forward and backward propagation. Specifically, in the modality dimension, we introduce a modality invariant fusion mechanism that fosters stable cross-modal representations, which aim to capture the common and stable representations shared across different modalities. In the temporal dimension, we impose a specialized sequential variation regularization term that regulates the model's learning trajectory during backward propagation, which is essentially total variation regularization degenerated into one-dimensional linear differences. Extensive experiments on three standard public datasets validate the effectiveness of our proposed approach.", "date": "2024-08-30", "categories": []}
{"id": "2409.19025", "title": "Dealing with Controversy: An Emotion and Coping Strategy Corpus Based on Role Playing", "abstract": "There is a mismatch between psychological and computational studies on emotions. Psychological research aims at explaining and documenting internal mechanisms of these phenomena, while computational work often simplifies them into labels. Many emotion fundamentals remain under-explored in natural language processing, particularly how emotions develop and how people cope with them. To help reduce this gap, we follow theories on coping, and treat emotions as strategies to cope with salient situations (i.e., how people deal with emotion-eliciting events). This approach allows us to investigate the link between emotions and behavior, which also emerges in language. We introduce the task of coping identification, together with a corpus to do so, constructed via role-playing. We find that coping strategies realize in text even though they are challenging to recognize, both for humans and automatic systems trained and prompted on the same task. We thus open up a promising research direction to enhance the capability of models to better capture emotion mechanisms from text.", "date": "2024-09-26", "categories": []}
{"id": "2503.17493", "title": "Meme Similarity and Emotion Detection using Multimodal Analysis", "abstract": "Internet memes are a central element of online culture, blending images and text. While substantial research has focused on either the visual or textual components of memes, little attention has been given to their interplay. This gap raises a key question: What methodology can effectively compare memes and the emotions they elicit? Our study employs a multimodal methodological approach, analyzing both the visual and textual elements of memes. Specifically, we perform a multimodal CLIP (Contrastive Language-Image Pre-training) model for grouping similar memes based on text and visual content embeddings, enabling robust similarity assessments across modalities. Using the Reddit Meme Dataset and Memotion Dataset, we extract low-level visual features and high-level semantic features to identify similar meme pairs. To validate these automated similarity assessments, we conducted a user study with 50 participants, asking them to provide yes/no responses regarding meme similarity and their emotional reactions. The comparison of experimental results with human judgments showed a 67.23\\% agreement, suggesting that the computational approach aligns well with human perception. Additionally, we implemented a text-based classifier using the DistilBERT model to categorize memes into one of six basic emotions. The results indicate that anger and joy are the dominant emotions in memes, with motivational memes eliciting stronger emotional responses. This research contributes to the study of multimodal memes, enhancing both language-based and visual approaches to analyzing and improving online visual communication and user experiences. Furthermore, it provides insights for better content moderation strategies in online platforms.", "date": "2025-03-21", "categories": []}
{"id": "2601.22161", "title": "Attention Isn't All You Need for Emotion Recognition:Domain Features Outperform Transformers on the EAV Dataset", "abstract": "We present a systematic study of multimodal emotion recognition using the EAV dataset, investigating whether complex attention mechanisms improve performance on small datasets. We implement three model categories: baseline transformers (M1), novel factorized attention mechanisms (M2), and improved CNN baselines (M3). Our experiments show that sophisticated attention mechanisms consistently underperform on small datasets. M2 models achieved 5 to 13 percentage points below baselines due to overfitting and destruction of pretrained features. In contrast, simple domain-appropriate modifications proved effective: adding delta MFCCs to the audio CNN improved accuracy from 61.9% to 65.56% (+3.66pp), while frequency-domain features for EEG achieved 67.62% (+7.62pp over the paper baseline). Our vision transformer baseline (M1) reached 75.30%, exceeding the paper's ViViT result (74.5%) through domain-specific pretraining, and vision delta features achieved 72.68% (+1.28pp over the paper CNN). These findings demonstrate that for small-scale emotion recognition, domain knowledge and proper implementation outperform architectural complexity.", "date": "2026-01-07", "categories": []}
{"id": "2602.00123", "title": "Visual Affect Analysis: Predicting Emotions of Image Viewers with Vision-Language Models", "abstract": "Vision-language models (VLMs) show promise as tools for inferring affect from visual stimuli at scale; it is not yet clear how closely their outputs align with human affective ratings. We benchmarked nine VLMs, ranging from state-of-the-art proprietary models to open-source models, on three psycho-metrically validated affective image datasets: the International Affective Picture System, the Nencki Affective Picture System, and the Library of AI-Generated Affective Images. The models performed two tasks in the zero-shot setting: (i) top-emotion classification (selecting the strongest discrete emotion elicited by an image) and (ii) continuous prediction of human ratings on 1-7/9 Likert scales for discrete emotion categories and affective dimensions. We also evaluated the impact of rater-conditioned prompting on the LAI-GAI dataset using de-identified participant metadata. The results show good performance in discrete emotion classification, with accuracies typically ranging from 60% to 80% on six-emotion labels and from 60% to 75% on a more challenging 12-category task. The predictions of anger and surprise had the lowest accuracy in all datasets. For continuous rating prediction, models showed moderate to strong alignment with humans (r > 0.75) but also exhibited consistent biases, notably weaker performance on arousal, and a tendency to overestimate response strength. Rater-conditioned prompting resulted in only small, inconsistent changes in predictions. Overall, VLMs capture broad affective trends but lack the nuance found in validated psychological ratings, highlighting their potential and current limitations for affective computing and mental health-related applications.", "date": "2026-01-27", "categories": []}
{"id": "2602.00360", "title": "Leveraging Textual-Cues for Enhancing Multimodal Sentiment Analysis by Object Recognition", "abstract": "Multimodal sentiment analysis, which includes both image and text data, presents several challenges due to the dissimilarities in the modalities of text and image, the ambiguity of sentiment, and the complexities of contextual meaning. In this work, we experiment with finding the sentiments of image and text data, individually and in combination, on two datasets. Part of the approach introduces the novel `Textual-Cues for Enhancing Multimodal Sentiment Analysis' (TEMSA) based on object recognition methods to address the difficulties in multimodal sentiment analysis. Specifically, we extract the names of all objects detected in an image and combine them with associated text; we call this combination of text and image data TEMS. Our results demonstrate that only TEMS improves the results when considering all the object names for the overall sentiment of multimodal data compared to individual analysis. This research contributes to advancing multimodal sentiment analysis and offers insights into the efficacy of TEMSA in combining image and text data for multimodal sentiment analysis.", "date": "2026-01-30", "categories": []}
{"id": "2602.00811", "title": "MissMAC-Bench: Building Solid Benchmark for Missing Modality Issue in Robust Multimodal Affective Computing", "abstract": "As a knowledge discovery task over heterogeneous data sources, current Multimodal Affective Computing (MAC) heavily rely on the completeness of multiple modalities to accurately understand human's affective state. However, in real-world scenarios, the availability of modality data is often dynamic and uncertain, leading to substantial performance fluctuations due to the distribution shifts and semantic deficiencies of the incomplete multimodal inputs. Known as the missing modality issue, this challenge poses a critical barrier to the robustness and practical deployment of MAC models. To systematically quantify this issue, we introduce MissMAC-Bench, a comprehensive benchmark designed to establish fair and unified evaluation standards from the perspective of cross-modal synergy. Two guiding principles are proposed, including no missing prior during training, and one single model capable of handling both complete and incomplete modality scenarios, thereby ensuring better generalization. Moreover, to bridge the gap between academic research and real-world applications, our benchmark integrates evaluation protocols with both fixed and random missing patterns at the dataset and instance levels. Extensive experiments conducted on 3 widely-used language models across 4 datasets validate the effectiveness of diverse MAC approaches in tackling the missing modality issue. Our benchmark provides a solid foundation for advancing robust multimodal affective computing and promotes the development of multimedia data mining.", "date": "2026-01-31", "categories": []}
{"id": "2602.00914", "title": "A Baseline Multimodal Approach to Emotion Recognition in Conversations", "abstract": "We present a lightweight multimodal baseline for emotion recognition in conversations using the SemEval-2024 Task 3 dataset built from the sitcom Friends. The goal of this report is not to propose a novel state-of-the-art method, but to document an accessible reference implementation that combines (i) a transformer-based text classifier and (ii) a self-supervised speech representation model, with a simple late-fusion ensemble. We report the baseline setup and empirical results obtained under a limited training protocol, highlighting when multimodal fusion improves over unimodal models. This preprint is provided for transparency and to support future, more rigorous comparisons.", "date": "2026-01-31", "categories": []}
{"id": "2509.04889", "title": "SpiderNets: Vision Models Predict Human Fear From Aversive Images", "abstract": "Phobias are common and impairing, and exposure therapy, which involves confronting patients with fear-provoking visual stimuli, is the most effective treatment. Scalable computerized exposure therapy requires automated prediction of fear directly from image content to adapt stimulus selection and treatment intensity. Whether such predictions can be made reliably and generalize across individuals and stimuli, however, remains unknown. Here we show that pretrained convolutional and transformer vision models, adapted via transfer learning, accurately predict group-level perceived fear for spider-related images, even when evaluated on new people and new images, achieving a mean absolute error (MAE) below 10 units on the 0-100 fear scale. Visual explanation analyses indicate that predictions are driven by spider-specific regions in the images. Learning-curve analyses show that transformer models are data efficient and approach performance saturation with the available data (~300 images). Prediction errors increase for very low and very high fear levels and within specific categories of images. These results establish transparent, data-driven fear estimation from images, laying the groundwork for adaptive digital mental health tools.", "date": "2025-09-05", "categories": []}
{"id": "2502.04379", "title": "Can Large Language Models Capture Video Game Engagement?", "abstract": "Can out-of-the-box pretrained Large Language Models (LLMs) detect human affect successfully when observing a video? To address this question, for the first time, we evaluate comprehensively the capacity of popular LLMs for successfully predicting continuous affect annotations of videos when prompted by a sequence of text and video frames in a multimodal fashion. In this paper, we test LLMs' ability to correctly label changes of in-game engagement in 80 minutes of annotated videogame footage from 20 first-person shooter games of the GameVibe corpus. We run over 4,800 experiments to investigate the impact of LLM architecture, model size, input modality, prompting strategy, and ground truth processing method on engagement prediction. Our findings suggest that while LLMs rightfully claim human-like performance across multiple domains and able to outperform traditional machine learning baselines, they generally fall behind continuous experience annotations provided by humans. We examine some of the underlying causes for a fluctuating performance across games, highlight the cases where LLMs exceed expectations, and draw a roadmap for the further exploration of automated emotion labelling via LLMs.", "date": "2025-02-05", "categories": []}
{"id": "2601.21130", "title": "What You Feel Is Not What They See: On Predicting Self-Reported Emotion from Third-Party Observer Labels", "abstract": "Self-reported emotion labels capture internal experience, while third-party labels reflect external perception. These perspectives often diverge, limiting the applicability of third-party-trained models to self-report contexts. This gap is critical in mental health, where accurate self-report modeling is essential for guiding intervention. We present the first cross-corpus evaluation of third-party-trained models on self-reports. We find activation unpredictable (CCC approximately 0) and valence moderately predictable (CCC approximately 0.3). Crucially, when content is personally significant to the speaker, models achieve high performance for valence (CCC approximately 0.6-0.8). Our findings point to personal significance as a key pathway for aligning external perception with internal experience and underscore the challenge of self-report activation modeling.", "date": "2026-01-29", "categories": []}
{"id": "2505.11051", "title": "CAMEO: Collection of Multilingual Emotional Speech Corpora", "abstract": "This paper presents CAMEO -- a curated collection of multilingual emotional speech datasets designed to facilitate research in emotion recognition and other speech-related tasks. The main objectives were to ensure easy access to the data, to allow reproducibility of the results, and to provide a standardized benchmark for evaluating speech emotion recognition (SER) systems across different emotional states and languages. The paper describes the dataset selection criteria, the curation and normalization process, and provides performance results for several models. The collection, along with metadata, and a leaderboard, is publicly available via the Hugging Face platform.", "date": "2025-05-16", "categories": []}
{"id": "2601.18414", "title": "Comparative Evaluation of Machine Learning Algorithms for Affective State Recognition from Children's Drawings", "abstract": "Autism spectrum disorder (ASD) represents a neurodevelopmental condition characterized by difficulties in expressing emotions and communication, particularly during early childhood. Understanding the affective state of children at an early age remains challenging, as conventional assessment methods are often intrusive, subjective, or difficult to apply consistently. This paper builds upon previous work on affective state recognition from children's drawings by presenting a comparative evaluation of machine learning models for emotion classification. Three deep learning architectures -- MobileNet, EfficientNet, and VGG16 -- are evaluated within a unified experimental framework to analyze classification performance, robustness, and computational efficiency. The models are trained using transfer learning on a dataset of children's drawings annotated with emotional labels provided by psychological experts. The results highlight important trade-offs between lightweight and deeper architectures when applied to drawing-based affective computing tasks, particularly in mobile and real-time application contexts.", "date": "2026-01-26", "categories": []}
{"id": "2601.18908", "title": "Enhancing Speech Emotion Recognition using Dynamic Spectral Features and Kalman Smoothing", "abstract": "Speech Emotion Recognition systems often use static features like Mel-Frequency Cepstral Coefficients (MFCCs), Zero Crossing Rate (ZCR), and Root Mean Square Energy (RMSE). Because of this, they can misclassify emotions when there is acoustic noise in vocal signals. To address this, we added dynamic features using Dynamic Spectral features (Deltas and Delta-Deltas) along with the Kalman Smoothing algorithm. This approach reduces noise and improves emotion classification. Since emotion changes over time, the Kalman Smoothing filter also helped make the classifier outputs more stable. Tests on the RAVDESS dataset showed that this method achieved a state-of-the-art accuracy of 87\\% and reduced misclassification between emotions with similar acoustic features", "date": "2026-01-26", "categories": []}
{"id": "2506.08717", "title": "Multi-Teacher Language-Aware Knowledge Distillation for Multilingual Speech Emotion Recognition", "abstract": "Speech Emotion Recognition (SER) is crucial for improving human-computer interaction. Despite strides in monolingual SER, extending them to build a multilingual system remains challenging. Our goal is to train a single model capable of multilingual SER by distilling knowledge from multiple teacher models. To address this, we introduce a novel language-aware multi-teacher knowledge distillation method to advance SER in English, Finnish, and French. It leverages Wav2Vec2.0 as the foundation of monolingual teacher models and then distills their knowledge into a single multilingual student model. The student model demonstrates state-of-the-art performance, with a weighted recall of 72.9 on the English dataset and an unweighted recall of 63.4 on the Finnish dataset, surpassing fine-tuning and knowledge distillation baselines. Our method excels in improving recall for sad and neutral emotions, although it still faces challenges in recognizing anger and happiness.", "date": "2025-06-10", "categories": []}
{"id": "2601.18228", "title": "Facial Emotion Recognition on FER-2013 using an EfficientNetB2-Based Approach", "abstract": "Detection of human emotions based on facial images in real-world scenarios is a difficult task due to low image quality, variations in lighting, pose changes, background distractions, small inter-class variations, noisy crowd-sourced labels, and severe class imbalance, as observed in the FER-2013 dataset of 48x48 grayscale images. Although recent approaches using large CNNs such as VGG and ResNet achieve reasonable accuracy, they are computationally expensive and memory-intensive, limiting their practicality for real-time applications. We address these challenges using a lightweight and efficient facial emotion recognition pipeline based on EfficientNetB2, trained using a two-stage warm-up and fine-tuning strategy. The model is enhanced with AdamW optimization, decoupled weight decay, label smoothing (epsilon = 0.06) to reduce annotation noise, and clipped class weights to mitigate class imbalance, along with dropout, mixed-precision training, and extensive real-time data augmentation. The model is trained using a stratified 87.5%/12.5% train-validation split while keeping the official test set intact, achieving a test accuracy of 68.78% with nearly ten times fewer parameters than VGG16-based baselines. Experimental results, including per-class metrics and learning dynamics, demonstrate stable training and strong generalization, making the proposed approach suitable for real-time and edge-based applications.", "date": "2026-01-26", "categories": []}
{"id": "2601.16811", "title": "Incorporating Eye-Tracking Signals Into Multimodal Deep Visual Models For Predicting User Aesthetic Experience In Residential Interiors", "abstract": "Understanding how people perceive and evaluate interior spaces is essential for designing environments that promote well-being. However, predicting aesthetic experiences remains difficult due to the subjective nature of perception and the complexity of visual responses. This study introduces a dual-branch CNN-LSTM framework that fuses visual features with eye-tracking signals to predict aesthetic evaluations of residential interiors. We collected a dataset of 224 interior design videos paired with synchronized gaze data from 28 participants who rated 15 aesthetic dimensions. The proposed model attains 72.2% accuracy on objective dimensions (e.g., light) and 66.8% on subjective dimensions (e.g., relaxation), outperforming state-of-the-art video baselines and showing clear gains on subjective evaluation tasks. Notably, models trained with eye-tracking retain comparable performance when deployed with visual input alone. Ablation experiments further reveal that pupil responses contribute most to objective assessments, while the combination of gaze and visual cues enhances subjective evaluations. These findings highlight the value of incorporating eye-tracking as privileged information during training, enabling more practical tools for aesthetic assessment in interior design.", "date": "2026-01-23", "categories": []}
{"id": "2507.03251", "title": "Toward Efficient Speech Emotion Recognition via Spectral Learning and Attention", "abstract": "Speech Emotion Recognition (SER) traditionally relies on auditory data analysis for emotion classification. Several studies have adopted different methods for SER. However, existing SER methods often struggle to capture subtle emotional variations and generalize across diverse datasets. In this article, we use Mel-Frequency Cepstral Coefficients (MFCCs) as spectral features to bridge the gap between computational emotion processing and human auditory perception. To further improve robustness and feature diversity, we propose a novel 1D-CNN-based SER framework that integrates data augmentation techniques. MFCC features extracted from the augmented data are processed using a 1D Convolutional Neural Network (CNN) architecture enhanced with channel and spatial attention mechanisms. These attention modules allow the model to highlight key emotional patterns, enhancing its ability to capture subtle variations in speech signals. The proposed method delivers cutting-edge performance, achieving the accuracy of 97.49% for SAVEE, 99.23% for RAVDESS, 89.31% for CREMA-D, 99.82% for TESS, 99.53% for EMO-DB, and 96.39% for EMOVO. Experimental results show new benchmarks in SER, demonstrating the effectiveness of our approach in recognizing emotional expressions with high precision. Our evaluation demonstrates that the integration of advanced Deep Learning (DL) methods substantially enhances generalization across diverse datasets, underscoring their potential to advance SER for real-world deployment in assistive technologies and human-computer interaction.", "date": "2025-07-04", "categories": []}
{"id": "2601.15615", "title": "Region-aware Spatiotemporal Modeling with Collaborative Domain Generalization for Cross-Subject EEG Emotion Recognition", "abstract": "Cross-subject EEG-based emotion recognition (EER) remains challenging due to strong inter-subject variability, which induces substantial distribution shifts in EEG signals, as well as the high complexity of emotion-related neural representations in both spatial organization and temporal evolution. Existing approaches typically improve spatial modeling, temporal modeling, or generalization strategies in isolation, which limits their ability to align representations across subjects while capturing multi-scale dynamics and suppressing subject-specific bias within a unified framework. To address these gaps, we propose a Region-aware Spatiotemporal Modeling framework with Collaborative Domain Generalization (RSM-CoDG) for cross-subject EEG emotion recognition. RSM-CoDG incorporates neuroscience priors derived from functional brain region partitioning to construct region-level spatial representations, thereby improving cross-subject comparability. It also employs multi-scale temporal modeling to characterize the dynamic evolution of emotion-evoked neural activity. In addition, the framework employs a collaborative domain generalization strategy, incorporating multidimensional constraints to reduce subject-specific bias in a fully unseen target subject setting, which enhances the generalization to unknown individuals. Extensive experimental results on SEED series datasets demonstrate that RSM-CoDG consistently outperforms existing competing methods, providing an effective approach for improving robustness. The source code is available at https://github.com/RyanLi-X/RSM-CoDG.", "date": "2026-01-22", "categories": []}
{"id": "2601.12534", "title": "Encoding Emotion Through Self-Supervised Eye Movement Reconstruction", "abstract": "The relationship between emotional expression and eye movement is well-documented, with literature establishing gaze patterns are reliable indicators of emotion. However, most studies utilize specialized, high-resolution eye-tracking equipment, limiting the potential reach of findings. We investigate how eye movement can be used to predict multimodal markers of emotional expression from naturalistic, low-resolution videos. We utilize a collection of video interviews from the USC Shoah Foundation's Visual History Archive with Holocaust survivors as they recount their experiences in the Auschwitz concentration camp. Inspired by pretraining methods on language models, we develop a novel gaze detection model that uses self-supervised eye movement reconstruction that can effectively leverage unlabeled video. We use this model's encoder embeddings to fine-tune models on two downstream tasks related to emotional expression. The first is aligning eye movement with directional emotion estimates from speech. The second task is using eye gaze as a predictor of three momentary manifestations of emotional behaviors: laughing, crying/sobbing, and sighing. We find our new model is predictive of emotion outcomes and observe a positive correlation between pretraining performance and emotion processing performance for both experiments. We conclude self-supervised eye movement reconstruction is an effective method for encoding the affective signal they carry.", "date": "2026-01-18", "categories": []}
{"id": "2601.14259", "title": "A Cloud-Based Cross-Modal Transformer for Emotion Recognition and Adaptive Human-Computer Interaction", "abstract": "Emotion recognition is a fundamental component of next-generation human-computer interaction (HCI), enabling machines to perceive, understand, and respond to users' affective states. However, existing systems often rely on single-modality analysis such as facial expressions, speech tone, or textual sentiment, resulting in limited robustness and poor generalization in real-world environments. To address these challenges, this study proposes a Cloud-Based Cross-Modal Transformer (CMT) framework for multimodal emotion recognition and adaptive human-computer interaction. The proposed model integrates visual, auditory, and textual signals using pretrained encoders (Vision Transformer, Wav2Vec2, and BERT) and employs a cross-modal attention mechanism to capture complex interdependencies among heterogeneous features. By leveraging cloud computing infrastructure with distributed training on Kubernetes and TensorFlow Serving, the system enables scalable, low-latency emotion recognition for large-scale user interactions. Experiments conducted on benchmark datasets including IEMOCAP, MELD, and AffectNet demonstrate that the CMT achieves state-of-the-art performance, improving the F1-score by 3.0 percent and reducing cross-entropy loss by 12.9 percent compared to strong multimodal baselines. Additionally, cloud deployment evaluations show an average response latency of 128 ms, representing a 35 percent reduction compared with conventional transformer-based fusion systems. These results confirm that the proposed framework enables efficient, real-time emotion recognition and adaptive feedback in applications such as intelligent customer service, virtual tutoring systems, and affective computing interfaces, marking an important step toward cloud-native affective computing and emotionally intelligent interactive systems.", "date": "2025-11-21", "categories": []}
{"id": "2601.14274", "title": "Divide and Refine: Enhancing Multimodal Representation and Explainability for Emotion Recognition in Conversation", "abstract": "Multimodal emotion recognition in conversation (MERC) requires representations that effectively integrate signals from multiple modalities. These signals include modality-specific cues, information shared across modalities, and interactions that emerge only when modalities are combined. In information-theoretic terms, these correspond to \\emph{unique}, \\emph{redundant}, and \\emph{synergistic} contributions. An ideal representation should leverage all three, yet achieving such balance remains challenging. Recent advances in contrastive learning and augmentation-based methods have made progress, but they often overlook the role of data preparation in preserving these components. In particular, applying augmentations directly to raw inputs or fused embeddings can blur the boundaries between modality-unique and cross-modal signals. To address this challenge, we propose a two-phase framework \\emph{\\textbf{D}ivide and \\textbf{R}efine} (\\textbf{DnR}). In the \\textbf{Divide} phase, each modality is explicitly decomposed into uniqueness, pairwise redundancy, and synergy. In the \\textbf{Refine} phase, tailored objectives enhance the informativeness of these components while maintaining their distinct roles. The refined representations are plug-and-play compatible with diverse multimodal pipelines. Extensive experiments on IEMOCAP and MELD demonstrate consistent improvements across multiple MERC backbones. These results highlight the effectiveness of explicitly dividing, refining, and recombining multimodal representations as a principled strategy for advancing emotion recognition. Our implementation is available at https://github.com/mattam301/DnR-WACV2026", "date": "2026-01-10", "categories": []}
{"id": "2601.14620", "title": "Scaling Ambiguity: Augmenting Human Annotation in Speech Emotion Recognition with Audio-Language Models", "abstract": "Speech Emotion Recognition models typically use single categorical labels, overlooking the inherent ambiguity of human emotions. Ambiguous Emotion Recognition addresses this by representing emotions as probability distributions, but progress is limited by unreliable ground-truth distributions inferred from sparse human annotations. This paper explores whether Large Audio-Language Models (ALMs) can mitigate the annotation bottleneck by generating high-quality synthetic annotations. We introduce a framework leveraging ALMs to create Synthetic Perceptual Proxies, augmenting human annotations to improve ground-truth distribution reliability. We validate these proxies through statistical analysis of their alignment with human distributions and evaluate their impact by fine-tuning ALMs with the augmented emotion distributions. Furthermore, to address class imbalance and enable unbiased evaluation, we propose DiME-Aug, a Distribution-aware Multimodal Emotion Augmentation strategy. Experiments on IEMOCAP and MSP-Podcast show that synthetic annotations enhance emotion distribution, especially in low-ambiguity regions where annotation agreement is high. However, benefits diminish for highly ambiguous emotions with greater human disagreement. This work provides the first evidence that ALMs could address annotation scarcity in ambiguous emotion recognition, but highlights the need for more advanced prompting or generation strategies to handle highly ambiguous cases.", "date": "2026-01-21", "categories": []}
{"id": "2601.14651", "title": "READ-Net: Clarifying Emotional Ambiguity via Adaptive Feature Recalibration for Audio-Visual Depression Detection", "abstract": "Depression is a severe global mental health issue that impairs daily functioning and overall quality of life. Although recent audio-visual approaches have improved automatic depression detection, methods that ignore emotional cues often fail to capture subtle depressive signals hidden within emotional expressions. Conversely, those incorporating emotions frequently confuse transient emotional expressions with stable depressive symptoms in feature representations, a phenomenon termed \\emph{Emotional Ambiguity}, thereby leading to detection errors. To address this critical issue, we propose READ-Net, the first audio-visual depression detection framework explicitly designed to resolve Emotional Ambiguity through Adaptive Feature Recalibration (AFR). The core insight of AFR is to dynamically adjust the weights of emotional features to enhance depression-related signals. Rather than merely overlooking or naively combining emotional information, READ-Net innovatively identifies and preserves depressive-relevant cues within emotional features, while adaptively filtering out irrelevant emotional noise. This recalibration strategy significantly clarifies feature representations, and effectively mitigates the persistent challenge of emotional interference. Additionally, READ-Net can be easily integrated into existing frameworks for improved performance. Extensive evaluations on three publicly available datasets show that READ-Net outperforms state-of-the-art methods, with average gains of 4.55\\% in accuracy and 1.26\\% in F1-score, demonstrating its robustness to emotional disturbances and improving audio-visual depression detection.", "date": "2026-01-21", "categories": []}
{"id": "2407.18552", "title": "Multimodal Emotion Recognition using Audio-Video Transformer Fusion with Cross Attention", "abstract": "Multimodal emotion recognition (MER) aims to infer human affect by jointly modeling audio and visual cues; however, existing approaches often struggle with temporal misalignment, weakly discriminative feature representations, and suboptimal fusion of heterogeneous modalities. To address these challenges, we propose AVT-CA, an Audio-Video Transformer architecture with cross attention for robust emotion recognition. The proposed model introduces a hierarchical video feature representation that combines channel attention, spatial attention, and local feature extraction to emphasize emotionally salient regions while suppressing irrelevant information. These refined visual features are integrated with audio representations through an intermediate transformer-based fusion mechanism that captures interlinked temporal dependencies across modalities. Furthermore, a cross-attention module selectively reinforces mutually consistent audio-visual cues, enabling effective feature selection and noise-aware fusion. Extensive experiments on three benchmark datasets, CMU-MOSEI, RAVDESS, and CREMA-D, demonstrate that AVT-CA consistently outperforms state-of-the-art baselines, achieving significant improvements in both accuracy and F1-score. Our source code is publicly available at https://github.com/shravan-18/AVTCA.", "date": "2024-07-26", "categories": []}
{"id": "2507.14867", "title": "Hybrid-supervised Hypergraph-enhanced Transformer for Micro-gesture Based Emotion Recognition", "abstract": "Micro-gestures are unconsciously performed body gestures that can convey the emotion states of humans and start to attract more research attention in the fields of human behavior understanding and affective computing as an emerging topic. However, the modeling of human emotion based on micro-gestures has not been explored sufficiently. In this work, we propose to recognize the emotion states based on the micro-gestures by reconstructing the behavior patterns with a hypergraph-enhanced Transformer in a hybrid-supervised framework. In the framework, hypergraph Transformer based encoder and decoder are separately designed by stacking the hypergraph-enhanced self-attention and multiscale temporal convolution modules. Especially, to better capture the subtle motion of micro-gestures, we construct a decoder with additional upsampling operations for a reconstruction task in a self-supervised learning manner. We further propose a hypergraph-enhanced self-attention module where the hyperedges between skeleton joints are gradually updated to present the relationships of body joints for modeling the subtle local motion. Lastly, for exploiting the relationship between the emotion states and local motion of micro-gestures, an emotion recognition head from the output of encoder is designed with a shallow architecture and learned in a supervised way. The end-to-end framework is jointly trained in a one-stage way by comprehensively utilizing self-reconstruction and supervision information. The proposed method is evaluated on two publicly available datasets, namely iMiGUE and SMG, and achieves the best performance under multiple metrics, which is superior to the existing methods.", "date": "2025-07-20", "categories": []}
{"id": "2508.07902", "title": "Tailored Emotional LLM-Supporter: Enhancing Cultural Sensitivity", "abstract": "Large language models (LLMs) show promise in offering emotional support and generating empathetic responses for individuals in distress, but their ability to deliver culturally sensitive support remains underexplored due to a lack of resources. In this work, we introduce CultureCare, the first dataset designed for this task, spanning four cultures and including 1729 distress messages, 1523 cultural signals, and 1041 support strategies with fine-grained emotional and cultural annotations. Leveraging CultureCare, we (i) develop and test four adaptation strategies for guiding three state-of-the-art LLMs toward culturally sensitive responses; (ii) conduct comprehensive evaluations using LLM-as-a-Judge, in-culture human annotators, and clinical psychologists; (iii) show that adapted LLMs outperform anonymous online peer responses, and that simple cultural role-play is insufficient for cultural sensitivity; and (iv) explore the application of LLMs in clinical training, where experts highlight their potential in fostering cultural competence in novice therapists.", "date": "2025-08-11", "categories": []}
{"id": "2509.21849", "title": "Following the TRACE: A Structured Path to Empathetic Response Generation with Multi-Agent Models", "abstract": "Empathetic response generation is a crucial task for creating more human-like and supportive conversational agents. However, existing methods face a core trade-off between the analytical depth of specialized models and the generative fluency of Large Language Models (LLMs). To address this, we propose TRACE, Task-decomposed Reasoning for Affective Communication and Empathy, a novel framework that models empathy as a structured cognitive process by decomposing the task into a pipeline for analysis and synthesis. By building a comprehensive understanding before generation, TRACE unites deep analysis with expressive generation. Experimental results show that our framework significantly outperforms strong baselines in both automatic and LLM-based evaluations, confirming that our structured decomposition is a promising paradigm for creating more capable and interpretable empathetic agents. Our code is available at https://anonymous.4open.science/r/TRACE-18EF/README.md.", "date": "2025-09-26", "categories": []}
{"id": "2601.12326", "title": "EmoKGEdit: Training-free Affective Injection via Visual Cue Transformation", "abstract": "Existing image emotion editing methods struggle to disentangle emotional cues from latent content representations, often yielding weak emotional expression and distorted visual structures. To bridge this gap, we propose EmoKGEdit, a novel training-free framework for precise and structure-preserving image emotion editing. Specifically, we construct a Multimodal Sentiment Association Knowledge Graph (MSA-KG) to disentangle the intricate relationships among objects, scenes, attributes, visual clues and emotion. MSA-KG explicitly encode the causal chain among object-attribute-emotion, and as external knowledge to support chain of thought reasoning, guiding the multimodal large model to infer plausible emotion-related visual cues and generate coherent instructions. In addition, based on MSA-KG, we design a disentangled structure-emotion editing module that explicitly separates emotional attributes from layout features within the latent space, which ensures that the target emotion is effectively injected while strictly maintaining visual spatial coherence. Extensive experiments demonstrate that EmoKGEdit achieves excellent performance in both emotion fidelity and content preservation, and outperforms the state-of-the-art methods.", "date": "2026-01-18", "categories": []}
{"id": "2601.12341", "title": "Time-Continuous Modeling for Temporal Affective Pattern Recognition in LLMs", "abstract": "This paper introduces a dataset and conceptual framework for LLMs to mimic real world emotional dynamics through time and in-context learning leveraging physics-informed neural network, opening a possibility for interpretable dialogue modeling.", "date": "2026-01-18", "categories": []}
{"id": "2601.12392", "title": "Psych\\=eChat: An Empathic Framework Focused on Emotion Shift Tracking and Safety Risk Analysis in Psychological Counseling", "abstract": "Large language models (LLMs) have demonstrated notable advancements in psychological counseling. However, existing models generally do not explicitly model seekers' emotion shifts across counseling sessions, a core focus in classical psychological schools. Moreover, how to align counselor models' responses with these emotion shifts while proactively mitigating safety risks remains underexplored. To bridge these gaps, we propose Psych\\=eChat, which explicitly integrates emotion shift tracking and safety risk analysis for psychological counseling. Specifically, we employ interactive role-playing to synthesize counselor--seeker dialogues, incorporating two modules: Emotion Management Module, to capture seekers' current emotions and emotion shifts; and Risk Control Module, to anticipate seekers' subsequent reactions and identify potential risks. Furthermore, we introduce two modeling paradigms. The Agent Mode structures emotion management, risk control, and counselor responses into a collaborative multi-agent pipeline. The LLM Mode integrates these stages into a unified chain-of-thought for end-to-end inference, balancing efficiency and performance. Extensive experiments, including interactive scoring, dialogue-level evaluation, and human assessment, demonstrate that Psych\\=eChat outperforms existing methods for emotional insight and safety control.", "date": "2026-01-18", "categories": []}
{"id": "2601.12402", "title": "Weaknesses of Facial Emotion Recognition Systems", "abstract": "Emotion detection from faces is one of the machine learning problems needed for human-computer interaction. The variety of methods used is enormous, which motivated an in-depth review of articles and scientific studies. Three of the most interesting and best solutions are selected, followed by the selection of three datasets that stood out for the diversity and number of images in them. The selected neural networks are trained, and then a series of experiments are performed to compare their performance, including testing on different datasets than a model was trained on. This reveals weaknesses in existing solutions, including differences between datasets, unequal levels of difficulty in recognizing certain emotions and the challenges in differentiating between closely related emotions.", "date": "2026-01-18", "categories": []}
{"id": "2601.13024", "title": "Tears or Cheers? Benchmarking LLMs via Culturally Elicited Distinct Affective Responses", "abstract": "Culture serves as a fundamental determinant of human affective processing and profoundly shapes how individuals perceive and interpret emotional stimuli. Despite this intrinsic link extant evaluations regarding cultural alignment within Large Language Models primarily prioritize declarative knowledge such as geographical facts or established societal customs. These benchmarks remain insufficient to capture the subjective interpretative variance inherent to diverse sociocultural lenses. To address this limitation, we introduce CEDAR, a multimodal benchmark constructed entirely from scenarios capturing Culturally \\underline{\\textsc{E}}licited \\underline{\\textsc{D}}istinct \\underline{\\textsc{A}}ffective \\underline{\\textsc{R}}esponses. To construct CEDAR, we implement a novel pipeline that leverages LLM-generated provisional labels to isolate instances yielding cross-cultural emotional distinctions, and subsequently derives reliable ground-truth annotations through rigorous human evaluation. The resulting benchmark comprises 10,962 instances across seven languages and 14 fine-grained emotion categories, with each language including 400 multimodal and 1,166 text-only samples. Comprehensive evaluations of 17 representative multilingual models reveal a dissociation between language consistency and cultural alignment, demonstrating that culturally grounded affective understanding remains a significant challenge for current models.", "date": "2026-01-19", "categories": []}
{"id": "2601.13225", "title": "Not all Blends are Equal: The BLEMORE Dataset of Blended Emotion Expressions with Relative Salience Annotations", "abstract": "Humans often experience not just a single basic emotion at a time, but rather a blend of several emotions with varying salience. Despite the importance of such blended emotions, most video-based emotion recognition approaches are designed to recognize single emotions only. The few approaches that have attempted to recognize blended emotions typically cannot assess the relative salience of the emotions within a blend. This limitation largely stems from the lack of datasets containing a substantial number of blended emotion samples annotated with relative salience. To address this shortcoming, we introduce BLEMORE, a novel dataset for multimodal (video, audio) blended emotion recognition that includes information on the relative salience of each emotion within a blend. BLEMORE comprises over 3,000 clips from 58 actors, performing 6 basic emotions and 10 distinct blends, where each blend has 3 different salience configurations (50/50, 70/30, and 30/70). Using this dataset, we conduct extensive evaluations of state-of-the-art video classification approaches on two blended emotion prediction tasks: (1) predicting the presence of emotions in a given sample, and (2) predicting the relative salience of emotions in a blend. Our results show that unimodal classifiers achieve up to 29% presence accuracy and 13% salience accuracy on the validation set, while multimodal methods yield clear improvements, with ImageBind + WavLM reaching 35% presence accuracy and HiCMAE 18% salience accuracy. On the held-out test set, the best models achieve 33% presence accuracy (VideoMAEv2 + HuBERT) and 18% salience accuracy (HiCMAE). In sum, the BLEMORE dataset provides a valuable resource to advancing research on emotion recognition systems that account for the complexity and significance of blended emotion expressions.", "date": "2026-01-19", "categories": []}
{"id": "2601.13589", "title": "Motion-to-Response Content Generation via Multi-Agent AI System with Real-Time Safety Verification", "abstract": "This paper proposes a multi-agent artificial intelligence system that generates response-oriented media content in real time based on audio-derived emotional signals. Unlike conventional speech emotion recognition studies that focus primarily on classification accuracy, our approach emphasizes the transformation of inferred emotional states into safe, age-appropriate, and controllable response content through a structured pipeline of specialized AI agents. The proposed system comprises four cooperative agents: (1) an Emotion Recognition Agent with CNN-based acoustic feature extraction, (2) a Response Policy Decision Agent for mapping emotions to response modes, (3) a Content Parameter Generation Agent for producing media control parameters, and (4) a Safety Verification Agent enforcing age-appropriateness and stimulation constraints. We introduce an explicit safety verification loop that filters generated content before output, ensuring compliance with predefined rules. Experimental results on public datasets demonstrate that the system achieves 73.2% emotion recognition accuracy, 89.4% response mode consistency, and 100% safety compliance while maintaining sub-100ms inference latency suitable for on-device deployment. The modular architecture enables interpretability and extensibility, making it applicable to child-adjacent media, therapeutic applications, and emotionally responsive smart devices.", "date": "2026-01-20", "categories": []}
{"id": "2601.13659", "title": "Temporal-Spatial Decouple before Act: Disentangled Representation Learning for Multimodal Sentiment Analysis", "abstract": "Multimodal Sentiment Analysis integrates Linguistic, Visual, and Acoustic. Mainstream approaches based on modality-invariant and modality-specific factorization or on complex fusion still rely on spatiotemporal mixed modeling. This ignores spatiotemporal heterogeneity, leading to spatiotemporal information asymmetry and thus limited performance. Hence, we propose TSDA, Temporal-Spatial Decouple before Act, which explicitly decouples each modality into temporal dynamics and spatial structural context before any interaction. For every modality, a temporal encoder and a spatial encoder project signals into separate temporal and spatial body. Factor-Consistent Cross-Modal Alignment then aligns temporal features only with their temporal counterparts across modalities, and spatial features only with their spatial counterparts. Factor specific supervision and decorrelation regularization reduce cross factor leakage while preserving complementarity. A Gated Recouple module subsequently recouples the aligned streams for task. Extensive experiments show that TSDA outperforms baselines. Ablation analysis studies confirm the necessity and interpretability of the design.", "date": "2026-01-20", "categories": []}
{"id": "2508.09145", "title": "MoLAN: A Unified Modality-Aware Noise Dynamic Editing Framework for Multimodal Sentiment Analysis", "abstract": "Multimodal Sentiment Analysis aims to integrate information from various modalities, such as audio, visual, and text, to make complementary predictions. However, it often struggles with irrelevant or misleading visual and auditory information. Most existing approaches typically treat the entire modality information (e.g., a whole image, audio segment, or text paragraph) as an independent unit for feature enhancement or denoising. They often suppress the redundant and noise information at the risk of losing critical information. To address this challenge, we propose MoLAN, a unified ModaLity-aware noise dynAmic editiNg framework. Specifically, MoLAN performs modality-aware blocking by dividing the features of each modality into multiple blocks. Each block is then dynamically assigned a distinct denoising strength based on its noise level and semantic relevance, enabling fine-grained noise suppression while preserving essential multimodal information. Notably, MoLAN is a unified and flexible framework that can be seamlessly integrated into a wide range of multimodal models. Building upon this framework, we further introduce MoLAN+, a new multimodal sentiment analysis approach. Experiments across five models and four datasets demonstrate the broad effectiveness of the MoLAN framework. Extensive evaluations show that MoLAN+ achieves the state-of-the-art performance. The code is publicly available at https://github.com/betterfly123/MoLAN-Framework.", "date": "2025-07-31", "categories": []}
{"id": "2511.06404", "title": "InfoAffect: Affective Annotations of Infographics in Information Spread", "abstract": "Infographics are widely used in social media to convey complex information, yet how they influence users' affects remains underexplored due to the scarcity of relevant datasets. To address this gap, we introduce a 3.5k-sample affect-annotated InfoAffect dataset, which combines textual content with real-world infographics. We first collected the raw data from six fields and aligned it via preprocessing, the accompanied-text-priority method, and three strategies to guarantee quality and compliance. After that, we constructed an Affect Table to constrain annotation. We used five state-of-the-art multimodal large language models (MLLMs) to analyze both modalities, and their outputs were fused with the Reciprocal Rank Fusion (RRF) algorithm to yield robust affects and confidences. We conducted a user study with two experiments to validate usability and assess the InfoAffect dataset using the Composite Affect Consistency Index (CACI), achieving an overall score of 0.608, which indicates high accuracy. The InfoAffect dataset is available in a public repository at https://github.com/bulichuchu/InfoAffect-dataset.", "date": "2025-11-09", "categories": []}
{"id": "2601.10342", "title": "C-GRASP: Clinically-Grounded Reasoning for Affective Signal Processing", "abstract": "Heart rate variability (HRV) is a pivotal noninvasive marker for autonomic monitoring; however, applying Large Language Models (LLMs) to HRV interpretation is hindered by physiological hallucinations. These include respiratory sinus arrhythmia (RSA) contamination, short-data instability in nonlinear metrics, and the neglect of individualized baselines in favor of population norms. We propose C-GRASP (Clinically-Grounded Reasoning for Affective Signal Processing), a guardrailed RAG-enhanced pipeline that decomposes HRV interpretation into eight traceable reasoning steps. Central to C-GRASP is a Z-score Priority Hierarchy that enforces the weighting of individualized baseline shifts over normative statistics. The system effectively mitigates spectral hallucinations through automated RSA-aware guardrails, preventing contamination of frequency-domain indices. Evaluated on 414 trials from the DREAMER dataset, C-GRASP integrated with high-scale reasoning models (e.g., MedGemma3-thinking) achieved superior performance in 4-class emotion classification (37.3% accuracy) and a Clinical Reasoning Consistency (CRC) score of 69.6%. Ablation studies confirm that the individualized Delta Z-score module serves as the critical logical anchor, preventing the \"population bias\" common in native LLMs. Ultimately, C-GRASP transitions affective computing from black-box classification to transparent, evidence-based clinical decision support, paving the way for safer AI integration in biomedical engineering.", "date": "2026-01-15", "categories": []}
{"id": "2601.10513", "title": "AEQ-Bench: Measuring Empathy of Omni-Modal Large Models", "abstract": "While the automatic evaluation of omni-modal large models (OLMs) is essential, assessing empathy remains a significant challenge due to its inherent affectivity. To investigate this challenge, we introduce AEQ-Bench (Audio Empathy Quotient Benchmark), a novel benchmark to systematically assess two core empathetic capabilities of OLMs: (i) generating empathetic responses by comprehending affective cues from multi-modal inputs (audio + text), and (ii) judging the empathy of audio responses without relying on text transcription. Compared to existing benchmarks, AEQ-Bench incorporates two novel settings that vary in context specificity and speech tone. Comprehensive assessment across linguistic and paralinguistic metrics reveals that (1) OLMs trained with audio output capabilities generally outperformed models with text-only outputs, and (2) while OLMs align with human judgments for coarse-grained quality assessment, they remain unreliable for evaluating fine-grained paralinguistic expressiveness.", "date": "2026-01-15", "categories": []}
{"id": "2304.00377", "title": "Beyond One-Size-Fits-All: A Survey of Personalized Affective Computing in Human-Agent Interaction", "abstract": "In personalized machine learning, the aim of personalization is to train a model that caters to a specific individual or group of individuals by optimizing one or more performance metrics and adhering to specific constraints. In this paper, we discuss the need for personalization in affective computing and present the first survey of existing approaches for personalization in affective computing. Our review spans training techniques and objectives towards the personalization of affective computing models across various interaction modes and contexts. We develop a taxonomy that clusters existing approaches into Data-level and Model-level approaches. Across the Data-Level and Model-Level broad categories, we group existing approaches into seven sub-categories: (1) User-Specific Models, (2) Group-Specific Models, (3) Weighting-Based Approaches, (4) Feature Augmentation, (5) Generative-Based Models which fall into the Data-Level approaches, (6) Fine-Tuning Approaches, and (7) Multitask Learning Approaches falling under the model-level approaches. We provide a problem formulation for personalized affective computing, and to each of the identified sub-categories. Additionally, we provide a statistical analysis of the surveyed literature, analyzing the prevalence of different affective computing tasks, interaction modes (i.e. Human-Computer Interaction (HCI), Human-Human interaction (HHI), Human-Robot Interaction (HRI)), interaction contexts (e.g. educative, social, gaming, etc.), and the level of personalization among the surveyed works. Based on our analysis, we provide a road-map for researchers interested in exploring this direction.", "date": "2023-04-01", "categories": []}
{"id": "2507.07046", "title": "A Novel Hybrid Deep Learning Technique for Speech Emotion Detection using Feature Engineering", "abstract": "Nowadays, speech emotion recognition (SER) plays a vital role in the field of human-computer interaction (HCI) and the evolution of artificial intelligence (AI). Our proposed DCRF-BiLSTM model is used to recognize seven emotions: neutral, happy, sad, angry, fear, disgust, and surprise, which are trained on five datasets: RAVDESS (R), TESS (T), SAVEE (S), EmoDB (E), and Crema-D (C). The model achieves high accuracy on individual datasets, including 97.83% on RAVDESS, 97.02% on SAVEE, 95.10% for CREMA-D, and a perfect 100% on both TESS and EMO-DB. For the combined (R+T+S) datasets, it achieves 98.82% accuracy, outperforming previously reported results. To our knowledge, no existing study has evaluated a single SER model across all five benchmark datasets (i.e., R+T+S+C+E) simultaneously. In our work, we introduce this comprehensive combination and achieve a remarkable overall accuracy of 93.76%. These results confirm the robustness and generalizability of our DCRF-BiLSTM framework across diverse datasets.", "date": "2025-07-09", "categories": []}
{"id": "2601.09606", "title": "GRCF: Two-Stage Groupwise Ranking and Calibration Framework for Multimodal Sentiment Analysis", "abstract": "Most Multimodal Sentiment Analysis research has focused on point-wise regression. While straightforward, this approach is sensitive to label noise and neglects whether one sample is more positive than another, resulting in unstable predictions and poor correlation alignment. Pairwise ordinal learning frameworks emerged to address this gap, capturing relative order by learning from comparisons. Yet, they introduce two new trade-offs: First, they assign uniform importance to all comparisons, failing to adaptively focus on hard-to-rank samples. Second, they employ static ranking margins, which fail to reflect the varying semantic distances between sentiment groups. To address this, we propose a Two-Stage Group-wise Ranking and Calibration Framework (GRCF) that adapts the philosophy of Group Relative Policy Optimization (GRPO). Our framework resolves these trade-offs by simultaneously preserving relative ordinal structure, ensuring absolute score calibration, and adaptively focusing on difficult samples. Specifically, Stage 1 introduces a GRPO-inspired Advantage-Weighted Dynamic Margin Ranking Loss to build a fine-grained ordinal structure. Stage 2 then employs an MAE-driven objective to align prediction magnitudes. To validate its generalizability, we extend GRCF to classification tasks, including multimodal humor detection and sarcasm detection. GRCF achieves state-of-the-art performance on core regression benchmarks, while also showing strong generalizability in classification tasks.", "date": "2026-01-14", "categories": []}
{"id": "2601.06575", "title": "Are Emotions Arranged in a Circle? Geometric Analysis of Emotion Representations via Hyperspherical Contrastive Learning", "abstract": "Psychological research has long utilized circumplex models to structure emotions, placing similar emotions adjacently and opposing ones diagonally. Although frequently used to interpret deep learning representations, these models are rarely directly incorporated into the representation learning of language models, leaving their geometric validity unexplored. This paper proposes a method to induce circular emotion representations within language model embeddings via contrastive learning on a hypersphere. We show that while this circular alignment offers superior interpretability and robustness against dimensionality reduction, it underperforms compared to conventional designs in high-dimensional settings and fine-grained classification. Our findings elucidate the trade-offs involved in applying psychological circumplex models to deep learning architectures.", "date": "2026-01-10", "categories": []}
{"id": "2601.07858", "title": "Affect and Effect: Limitations of regularisation-based continual learning in EEG-based emotion classification", "abstract": "Generalisation to unseen subjects in EEG-based emotion classification remains a challenge due to high inter-and intra-subject variability. Continual learning (CL) poses a promising solution by learning from a sequence of tasks while mitigating catastrophic forgetting. Regularisation-based CL approaches, such as Elastic Weight Consolidation (EWC), Synaptic Intelligence (SI), and Memory Aware Synapses (MAS), are commonly used as baselines in EEG-based CL studies, yet their suitability for this problem remains underexplored. This study theoretically and empirically finds that regularisation-based CL methods show limited performance for EEG-based emotion classification on the DREAMER and SEED datasets. We identify a fundamental misalignment in the stability-plasticity trade-off, where regularisation-based methods prioritise mitigating catastrophic forgetting (backward transfer) over adapting to new subjects (forward transfer). We investigate this limitation under subject-incremental sequences and observe that: (1) the heuristics for estimating parameter importance become less reliable under noisy data and covariate shift, (2) gradients on parameters deemed important by these heuristics often interfere with gradient updates required for new subjects, moving optimisation away from the minimum, (3) importance values accumulated across tasks over-constrain the model, and (4) performance is sensitive to subject order. Forward transfer showed no statistically significant improvement over sequential fine-tuning (p > 0.05 across approaches and datasets). The high variability of EEG signals means past subjects provide limited value to future subjects. Regularisation-based continual learning approaches are therefore limited for robust generalisation to unseen subjects in EEG-based emotion classification.", "date": "2026-01-09", "categories": []}
{"id": "2601.07877", "title": "E^2-LLM: Bridging Neural Signals and Interpretable Affective Analysis", "abstract": "Emotion recognition from electroencephalography (EEG) signals remains challenging due to high inter-subject variability, limited labeled data, and the lack of interpretable reasoning in existing approaches. While recent multimodal large language models (MLLMs) have advanced emotion analysis, they have not been adapted to handle the unique spatiotemporal characteristics of neural signals. We present E^2-LLM (EEG-to-Emotion Large Language Model), the first MLLM framework for interpretable emotion analysis from EEG. E^2-LLM integrates a pretrained EEG encoder with Qwen-based LLMs through learnable projection layers, employing a multi-stage training pipeline that encompasses emotion-discriminative pretraining, cross-modal alignment, and instruction tuning with chain-of-thought reasoning. We design a comprehensive evaluation protocol covering basic emotion prediction, multi-task reasoning, and zero-shot scenario understanding. Experiments on the dataset across seven emotion categories demonstrate that E^2-LLM achieves excellent performance on emotion classification, with larger variants showing enhanced reliability and superior zero-shot generalization to complex reasoning scenarios. Our work establishes a new paradigm combining physiological signals with LLM reasoning capabilities, showing that model scaling improves both recognition accuracy and interpretable emotional understanding in affective computing.", "date": "2026-01-11", "categories": []}
{"id": "2601.08049", "title": "Integrating Attendance Tracking and Emotion Detection for Enhanced Student Engagement in Smart Classrooms", "abstract": "The increasing adoption of smart classroom technologies in higher education has mainly focused on automating attendance, with limited attention given to students' emotional and cognitive engagement during lectures. This limits instructors' ability to identify disengagement and adapt teaching strategies in real time. This paper presents SCASED (Smart Classroom Attendance System with Emotion Detection), an IoT-based system that integrates automated attendance tracking with facial emotion recognition to support classroom engagement monitoring. The system uses a Raspberry Pi camera and OpenCV for face detection, and a finetuned MobileNetV2 model to classify four learning-related emotional states: engagement, boredom, confusion, and frustration. A session-based mechanism is implemented to manage attendance and emotion monitoring by recording attendance once per session and performing continuous emotion analysis thereafter. Attendance and emotion data are visualized through a cloud-based dashboard to provide instructors with insights into classroom dynamics. Experimental evaluation using the DAiSEE dataset achieved an emotion classification accuracy of 89.5%. The results show that integrating attendance data with emotion analytics can provide instructors with additional insight into classroom dynamics and support more responsive teaching practices.", "date": "2026-01-12", "categories": []}
{"id": "2305.07216", "title": "Versatile audio-visual learning for emotion recognition", "abstract": "Most current audio-visual emotion recognition models lack the flexibility needed for deployment in practical applications. We envision a multimodal system that works even when only one modality is available and can be implemented interchangeably for either predicting emotional attributes or recognizing categorical emotions. Achieving such flexibility in a multimodal emotion recognition system is difficult due to the inherent challenges in accurately interpreting and integrating varied data sources. It is also a challenge to robustly handle missing or partial information while allowing direct switch between regression or classification tasks. This study proposes a versatile audio-visual learning (VAVL) framework for handling unimodal and multimodal systems for emotion regression or emotion classification tasks. We implement an audio-visual framework that can be trained even when audio and visual paired data is not available for part of the training set (i.e., audio only or only video is present). We achieve this effective representation learning with audio-visual shared layers, residual connections over shared layers, and a unimodal reconstruction task. Our experimental results reveal that our architecture significantly outperforms strong baselines on the CREMA-D, MSP-IMPROV, and CMU-MOSEI corpora. Notably, VAVL attains a new state-of-the-art performance in the emotional attribute prediction task on the MSP-IMPROV corpus.", "date": "2023-05-12", "categories": []}
{"id": "2311.07093", "title": "A Comprehensive Study on the Effectiveness of ASR Representations for Noise-Robust Speech Emotion Recognition", "abstract": "This paper proposes an efficient attempt to noisy speech emotion recognition (NSER). Conventional NSER approaches have proven effective in mitigating the impact of artificial noise sources, such as white Gaussian noise, but are limited to non-stationary noises in real-world environments due to their complexity and uncertainty. To overcome this limitation, we introduce a new method for NSER by adopting the automatic speech recognition (ASR) model as a noise-robust feature extractor to eliminate non-vocal information in noisy speech. We first obtain intermediate layer information from the ASR model as a feature representation for emotional speech and then apply this representation for the downstream NSER task. Our experimental results show that 1) the proposed method achieves better NSER performance compared with the conventional noise reduction method, 2) outperforms self-supervised learning approaches, and 3) even outperforms text-based approaches using ASR transcription or the ground truth transcription of noisy speech.", "date": "2023-11-13", "categories": []}
{"id": "2507.22393", "title": "Gems: Group Emotion Profiling Through Multimodal Situational Understanding", "abstract": "Understanding individual, group and event level emotions along with contextual information is crucial for analyzing a multi-person social situation. To achieve this, we frame emotion comprehension as the task of predicting fine-grained individual emotion to coarse grained group and event level emotion. We introduce GEMS that leverages a multimodal swin-transformer and S3Attention based architecture, which processes an input scene, group members, and context information to generate joint predictions. Existing multi-person emotion related benchmarks mainly focus on atomic interactions primarily based on emotion perception over time and group level. To this end, we extend and propose VGAF-GEMS to provide more fine grained and holistic analysis on top of existing group level annotation of VGAF dataset. GEMS aims to predict basic discrete and continuous emotions (including valence and arousal) as well as individual, group and event level perceived emotions. Our benchmarking effort links individual, group and situational emotional responses holistically. The quantitative and qualitative comparisons with adapted state-of-the-art models demonstrate the effectiveness of GEMS framework on VGAF-GEMS benchmarking. We believe that it will pave the way of further research. The code and data is available at: https://github.com/katariaak579/GEMS", "date": "2025-07-30", "categories": []}
{"id": "2601.07565", "title": "A Unified Framework for Emotion Recognition and Sentiment Analysis via Expert-Guided Multimodal Fusion with Large Language Models", "abstract": "Multimodal emotion understanding requires effective integration of text, audio, and visual modalities for both discrete emotion recognition and continuous sentiment analysis. We present EGMF, a unified framework combining expert-guided multimodal fusion with large language models. Our approach features three specialized expert networks--a fine-grained local expert for subtle emotional nuances, a semantic correlation expert for cross-modal relationships, and a global context expert for long-range dependencies--adaptively integrated through hierarchical dynamic gating for context-aware feature selection. Enhanced multimodal representations are integrated with LLMs via pseudo token injection and prompt-based conditioning, enabling a single generative framework to handle both classification and regression through natural language generation. We employ LoRA fine-tuning for computational efficiency. Experiments on bilingual benchmarks (MELD, CHERMA, MOSEI, SIMS-V2) demonstrate consistent improvements over state-of-the-art methods, with superior cross-lingual robustness revealing universal patterns in multimodal emotional expressions across English and Chinese. We will release the source code publicly.", "date": "2026-01-12", "categories": []}
{"id": "2601.04297", "title": "ArtCognition: A Multimodal AI Framework for Affective State Sensing from Visual and Kinematic Drawing Cues", "abstract": "The objective assessment of human affective and psychological states presents a significant challenge, particularly through non-verbal channels. This paper introduces digital drawing as a rich and underexplored modality for affective sensing. We present a novel multimodal framework, named ArtCognition, for the automated analysis of the House-Tree-Person (HTP) test, a widely used psychological instrument. ArtCognition uniquely fuses two distinct data streams: static visual features from the final artwork, captured by computer vision models, and dynamic behavioral kinematic cues derived from the drawing process itself, such as stroke speed, pauses, and smoothness. To bridge the gap between low-level features and high-level psychological interpretation, we employ a Retrieval-Augmented Generation (RAG) architecture. This grounds the analysis in established psychological knowledge, enhancing explainability and reducing the potential for model hallucination. Our results demonstrate that the fusion of visual and behavioral kinematic cues provides a more nuanced assessment than either modality alone. We show significant correlations between the extracted multimodal features and standardized psychological metrics, validating the framework's potential as a scalable tool to support clinicians. This work contributes a new methodology for non-intrusive affective state assessment and opens new avenues for technology-assisted mental healthcare.", "date": "2026-01-07", "categories": []}
{"id": "2511.16952", "title": "Point-Supervised Facial Expression Spotting with Gaussian-Based Instance-Adaptive Intensity Modeling", "abstract": "Automatic facial expression spotting, which aims to identify facial expression instances in untrimmed videos, is crucial for facial expression analysis. Existing methods primarily focus on fully-supervised learning and rely on costly, time-consuming temporal boundary annotations. In this paper, we investigate point-supervised facial expression spotting (P-FES), where only a single timestamp annotation per instance is required for training. We propose a unique two-branch framework for P-FES. First, to mitigate the limitation of hard pseudo-labeling, which often confuses neutral and expression frames with various intensities, we propose a Gaussian-based instance-adaptive intensity modeling (GIM) module to model instance-level expression intensity distribution for soft pseudo-labeling. By detecting the pseudo-apex frame around each point label, estimating the duration, and constructing an instance-level Gaussian distribution, GIM assigns soft pseudo-labels to expression frames for more reliable intensity supervision. The GIM module is incorporated into our framework to optimize the class-agnostic expression intensity branch. Second, we design a class-aware apex classification branch that distinguishes macro- and micro-expressions solely based on their pseudo-apex frames. During inference, the two branches work independently: the class-agnostic expression intensity branch generates expression proposals, while the class-aware apex-classification branch is responsible for macro- and micro-expression classification. Furthermore, we introduce an intensity-aware contrastive loss to enhance discriminative feature learning and suppress neutral noise by contrasting neutral frames with expression frames with various intensities. Extensive experiments on the SAMM-LV, CAS(ME)$^2$, and CAS(ME)$^3$ datasets demonstrate the effectiveness of our proposed framework.", "date": "2025-11-21", "categories": []}
{"id": "2601.02414", "title": "MIAR: Modality Interaction and Alignment Representation Fuison for Multimodal Emotion", "abstract": "Multimodal Emotion Recognition (MER) aims to perceive human emotions through three modes: language, vision, and audio. Previous methods primarily focused on modal fusion without adequately addressing significant distributional differences among modalities or considering their varying contributions to the task. They also lacked robust generalization capabilities across diverse textual model features, thus limiting performance in multimodal scenarios. Therefore, we propose a novel approach called Modality Interaction and Alignment Representation (MIAR). This network integrates contextual features across different modalities using a feature interaction to generate feature tokens to represent global representations of this modality extracting information from other modalities. These four tokens represent global representations of how each modality extracts information from others. MIAR aligns different modalities using contrastive learning and normalization strategies. We conduct experiments on two benchmarks: CMU-MOSI and CMU-MOSEI datasets, experimental results demonstrate the MIAR outperforms state-of-the-art MER methods.", "date": "2026-01-03", "categories": []}
{"id": "2601.02415", "title": "Multimodal Sentiment Analysis based on Multi-channel and Symmetric Mutual Promotion Feature Fusion", "abstract": "Multimodal sentiment analysis is a key technology in the fields of human-computer interaction and affective computing. Accurately recognizing human emotional states is crucial for facilitating smooth communication between humans and machines. Despite some progress in multimodal sentiment analysis research, numerous challenges remain. The first challenge is the limited and insufficiently rich features extracted from single modality data. Secondly, most studies focus only on the consistency of inter-modal feature information, neglecting the differences between features, resulting in inadequate feature information fusion. In this paper, we first extract multi-channel features to obtain more comprehensive feature information. We employ dual-channel features in both the visual and auditory modalities to enhance intra-modal feature representation. Secondly, we propose a symmetric mutual promotion (SMP) inter-modal feature fusion method. This method combines symmetric cross-modal attention mechanisms and self-attention mechanisms, where the cross-modal attention mechanism captures useful information from other modalities, and the self-attention mechanism models contextual information. This approach promotes the exchange of useful information between modalities, thereby strengthening inter-modal interactions. Furthermore, we integrate intra-modal features and inter-modal fused features, fully leveraging the complementarity of inter-modal feature information while considering feature information differences. Experiments conducted on two benchmark datasets demonstrate the effectiveness and superiority of our proposed method.", "date": "2026-01-03", "categories": []}
{"id": "2506.09827", "title": "EmoNet-Voice: A Fine-Grained, Expert-Verified Benchmark for Speech Emotion Detection", "abstract": "Speech emotion recognition (SER) systems are constrained by existing datasets that typically cover only 6-10 basic emotions, lack scale and diversity, and face ethical challenges when collecting sensitive emotional states. We introduce EMONET-VOICE, a comprehensive resource addressing these limitations through two components: (1) EmoNet-Voice Big, a 5,000-hour multilingual pre-training dataset spanning 40 fine-grained emotion categories across 11 voices and 4 languages, and (2) EmoNet-Voice Bench, a rigorously validated benchmark of 4,7k samples with unanimous expert consensus on emotion presence and intensity levels. Using state-of-the-art synthetic voice generation, our privacy-preserving approach enables ethical inclusion of sensitive emotions (e.g., pain, shame) while maintaining controlled experimental conditions. Each sample underwent validation by three psychology experts. We demonstrate that our Empathic Insight models trained on our synthetic data achieve strong real-world dataset generalization, as tested on EmoDB and RAVDESS. Furthermore, our comprehensive evaluation reveals that while high-arousal emotions (e.g., anger: 95% accuracy) are readily detected, the benchmark successfully exposes the difficulty of distinguishing perceptually similar emotions (e.g., sadness vs. distress: 63% discrimination), providing quantifiable metrics for advancing nuanced emotion AI. EMONET-VOICE establishes a new paradigm for large-scale, ethically-sourced, fine-grained SER research.", "date": "2025-06-11", "categories": []}
{"id": "2510.23506", "title": "Emotion-Coherent Reasoning for Multimodal LLMs via Emotional Rationale Verifier", "abstract": "The recent advancement of Multimodal Large Language Models (MLLMs) is transforming human-computer interaction (HCI) from surface-level exchanges into more nuanced and emotionally intelligent communication. To realize this shift, emotion understanding becomes essential allowing systems to capture subtle cues underlying user intent. Furthermore, providing faithful explanations for predicted emotions is crucial to ensure interpretability and build user trust. However, current MLLM-based methods often generate emotion explanations that diverge from the target labels and sometimes even contradict their own predicted emotions. This inconsistency poses a critical risk for misunderstanding and erodes reliability in interactive settings. To address this, we propose a novel approach: the Emotional Rationale Verifier (ERV) and an Explanation Reward. Our method guides the model to produce reasoning that is explicitly consistent with the target emotion during multimodal emotion recognition without modifying the model architecture or requiring additional paired video-description annotations. Our method significantly improves faithful explanation-prediction consistency and explanation emotion accuracy on the MAFW and DFEW datasets. Through extensive experiments and human evaluations, we show that our approach not only enhances alignment between explanation and prediction but also empowers MLLMs to deliver emotionally coherent, trustworthy interactions, marking a key step toward truly human-like HCI systems.", "date": "2025-10-27", "categories": []}
{"id": "2601.00812", "title": "Free Energy-Based Modeling of Emotional Dynamics in Video Advertisements", "abstract": "Emotional responses during advertising video viewing are recognized as essential for understanding media effects because they have influenced attention, memory, and purchase intention. To establish a methodological basis for explainable emotion estimation without relying on external information such as physiological signals or subjective ratings, we have quantified \"pleasantness,\" \"surprise,\" and \"habituation\" solely from scene-level expression features of advertising videos, drawing on the free energy(FE) principle, which has provided a unified account of perception, learning, and behavior. In this framework, Kullback-Leibler divergence (KLD) has captured prediction error, Bayesian surprise (BS) has captured belief updates, and uncertainty (UN) has reflected prior ambiguity, and together they have formed the core components of FE. Using 1,059 15 s food video advertisements, the experiments have shown that KLD has reflected \"pleasantness\" associated with brand presentation, BS has captured \"surprise\" arising from informational complexity, and UN has reflected \"surprise\" driven by uncertainty in element types and spatial arrangements, as well as by the variability and quantity of presented elements. This study also identified three characteristic emotional patterns, namely uncertain stimulus, sustained high emotion, and momentary peak and decay, demonstrating the usefulness of the proposed method. Robustness across nine hyperparameter settings and generalization tests with six types of Japanese advertising videos (three genres and two durations) confirmed that these tendencies remained stable. This work can be extended by integrating a wider range of expression elements and validating the approach through subjective ratings, ultimately guiding the development of technologies that can support the creation of more engaging advertising videos.", "date": "2025-12-22", "categories": []}
{"id": "2601.01530", "title": "EmoHarbor: Evaluating Personalized Emotional Support by Simulating the User's Internal World", "abstract": "Current evaluation paradigms for emotional support conversations tend to reward generic empathetic responses, yet they fail to assess whether the support is genuinely personalized to users' unique psychological profiles and contextual needs. We introduce EmoHarbor, an automated evaluation framework that adopts a User-as-a-Judge paradigm by simulating the user's inner world. EmoHarbor employs a Chain-of-Agent architecture that decomposes users' internal processes into three specialized roles, enabling agents to interact with supporters and complete assessments in a manner similar to human users. We instantiate this benchmark using 100 real-world user profiles that cover a diverse range of personality traits and situations, and define 10 evaluation dimensions of personalized support quality. Comprehensive evaluation of 20 advanced LLMs on EmoHarbor reveals a critical insight: while these models excel at generating empathetic responses, they consistently fail to tailor support to individual user contexts. This finding reframes the central challenge, shifting research focus from merely enhancing generic empathy to developing truly user-aware emotional support. EmoHarbor provides a reproducible and scalable framework to guide the development and evaluation of more nuanced and user-aware emotional support systems.", "date": "2026-01-04", "categories": []}
{"id": "2508.03609", "title": "evTransFER: A Transfer Learning Framework for Event-based Facial Expression Recognition", "abstract": "Event-based cameras are bio-inspired sensors that asynchronously capture pixel intensity changes with microsecond latency, high temporal resolution, and high dynamic range, providing information on the spatiotemporal dynamics of a scene. We propose evTransFER, a transfer learning-based framework for facial expression recognition using event-based cameras. The main contribution is a feature extractor designed to encode facial spatiotemporal dynamics, built by training an adversarial generative method on facial reconstruction and transferring the encoder weights to the facial expression recognition system. We demonstrate that the proposed transfer learning method improves facial expression recognition compared to training a network from scratch. We propose an architecture that incorporates an LSTM to capture longer-term facial expression dynamics and introduces a new event-based representation called TIE. We evaluated the framework using both the synthetic event-based facial expression database e-CK+ and the real neuromorphic dataset NEFER. On e-CK+, evTransFER achieved a recognition rate of 93.6\\%, surpassing state-of-the-art methods. For NEFER, which comprises event sequence with real sensor noise and sparse activity, the proposed transfer learning strategy achieved an accuracy of up to 76.7\\%. In both datasets, the outcomes surpassed current methodologies and exceeded results when compared with models trained from scratch.", "date": "2025-08-05", "categories": []}
{"id": "2512.23435", "title": "Distilled HuBERT for Mobile Speech Emotion Recognition: A Cross-Corpus Validation Study", "abstract": "Speech Emotion Recognition (SER) has significant potential for mobile applications, yet deployment remains constrained by the computational demands of state-of-the-art transformer architectures. This paper presents a mobile-efficient SER system based on DistilHuBERT, a distilled and 8-bit quantized transformer that achieves approximately 92% parameter reduction compared to full-scale Wav2Vec 2.0 models while maintaining competitive accuracy. We conduct a rigorous 5-fold Leave-One-Session-Out (LOSO) cross-validation on the IEMOCAP dataset to ensure speaker independence, augmented with cross-corpus training on CREMA-D to enhance generalization. Cross-corpus training with CREMA-D yields a 1.2% improvement in Weighted Accuracy, a 1.4% gain in Macro F1-score, and a 32% reduction in cross-fold variance, with the Neutral class showing the most substantial benefit at 5.4% F1-score improvement. Our approach achieves an Unweighted Accuracy of 61.4% with a quantized model footprint of only 23 MB, representing approximately 91% of the Unweighted Accuracy of a full-scale baseline. Cross-corpus evaluation on RAVDESS reveals that the theatrical nature of acted emotions causes predictions to cluster by arousal level rather than by specific emotion categories - happiness predictions systematically bleed into anger predictions, and sadness predictions bleed into neutral predictions, due to acoustic saturation when actors prioritize clarity over subtlety. Despite this theatricality effect reducing overall RAVDESS accuracy to 46.64%, the model maintains robust arousal detection with 99% recall for anger, 55% recall for neutral, and 27% recall for sadness. These findings demonstrate a Pareto-optimal tradeoff between model size and accuracy, enabling practical affect recognition on resource-constrained mobile devices.", "date": "2025-12-29", "categories": []}
{"id": "2512.24231", "title": "MotivNet: Evolving Meta-Sapiens into an Emotionally Intelligent Foundation Model", "abstract": "In this paper, we introduce MotivNet, a generalizable facial emotion recognition model for robust real-world application. Current state-of-the-art FER models tend to have weak generalization when tested on diverse data, leading to deteriorated performance in the real world and hindering FER as a research domain. Though researchers have proposed complex architectures to address this generalization issue, they require training cross-domain to obtain generalizable results, which is inherently contradictory for real-world application. Our model, MotivNet, achieves competitive performance across datasets without cross-domain training by using Meta-Sapiens as a backbone. Sapiens is a human vision foundational model with state-of-the-art generalization in the real world through large-scale pretraining of a Masked Autoencoder. We propose MotivNet as an additional downstream task for Sapiens and define three criteria to evaluate MotivNet's viability as a Sapiens task: benchmark performance, model similarity, and data similarity. Throughout this paper, we describe the components of MotivNet, our training approach, and our results showing MotivNet is generalizable across domains. We demonstrate that MotivNet can be benchmarked against existing SOTA models and meets the listed criteria, validating MotivNet as a Sapiens downstream task, and making FER more incentivizing for in-the-wild application. The code is available at https://github.com/OSUPCVLab/EmotionFromFaceImages.", "date": "2025-12-30", "categories": []}
{"id": "2502.18186", "title": "Steering Language Model to Stable Speech Emotion Recognition via Contextual Perception and Chain of Thought", "abstract": "Large-scale audio language models (ALMs), such as Qwen2-Audio, are capable of comprehending diverse audio signal, performing audio analysis and generating textual responses. However, in speech emotion recognition (SER), ALMs often suffer from hallucinations, resulting in misclassifications or irrelevant outputs. To address these challenges, we propose C$^2$SER, a novel ALM designed to enhance the stability and accuracy of SER through Contextual perception and Chain of Thought (CoT). C$^2$SER integrates the Whisper encoder for semantic perception and Emotion2Vec-S for acoustic perception, where Emotion2Vec-S extends Emotion2Vec with semi-supervised learning to enhance emotional discrimination. Additionally, C$^2$SER employs a CoT approach, processing SER in a step-by-step manner while leveraging speech content and speaking styles to improve recognition. To further enhance stability, C$^2$SER introduces self-distillation from explicit CoT to implicit CoT, mitigating error accumulation and boosting recognition accuracy. Extensive experiments show that C$^2$SER outperforms existing popular ALMs, such as Qwen2-Audio and SECap, delivering more stable and precise emotion recognition. We release the training code, checkpoints, and test sets to facilitate further research.", "date": "2025-02-25", "categories": []}
{"id": "2512.11239", "title": "Cross-modal Prompting for Balanced Incomplete Multi-modal Emotion Recognition", "abstract": "Incomplete multi-modal emotion recognition (IMER) aims at understanding human intentions and sentiments by comprehensively exploring the partially observed multi-source data. Although the multi-modal data is expected to provide more abundant information, the performance gap and modality under-optimization problem hinder effective multi-modal learning in practice, and are exacerbated in the confrontation of the missing data. To address this issue, we devise a novel Cross-modal Prompting (ComP) method, which emphasizes coherent information by enhancing modality-specific features and improves the overall recognition accuracy by boosting each modality's performance. Specifically, a progressive prompt generation module with a dynamic gradient modulator is proposed to produce concise and consistent modality semantic cues. Meanwhile, cross-modal knowledge propagation selectively amplifies the consistent information in modality features with the delivered prompts to enhance the discrimination of the modality-specific output. Additionally, a coordinator is designed to dynamically re-weight the modality outputs as a complement to the balance strategy to improve the model's efficacy. Extensive experiments on 4 datasets with 7 SOTA methods under different missing rates validate the effectiveness of our proposed method.", "date": "2025-12-12", "categories": []}
{"id": "2512.22333", "title": "Emotion classification using EEG headset signals and Random Forest", "abstract": "Emotions are one of the important components of the human being, thus they are a valuable part of daily activities such as interaction with people, decision making and learning. For this reason, it is important to detect, recognize and understand emotions using computational systems to improve communication between people and machines, which would facilitate the ability of computers to understand the communication between humans. This study proposes the creation of a model that allows the classification of people's emotions based on their EEG signals, for which the brain-computer interface EMOTIV EPOC was used. This allowed the collection of electroencephalographic information from 50 people, all of whom were shown audiovisual resources that helped to provoke the desired mood. The information obtained was stored in a database for the generation of the model and the corresponding classification analysis. Random Forest model was created for emotion prediction (happiness, sadness and relaxation), based on the signals of any person. The results obtained were 97.21% accurate for happiness, 76% for relaxation and 76% for sadness. Finally, the model was used to generate a real-time emotion prediction algorithm; it captures the person's EEG signals, executes the generated algorithm and displays the result on the screen with the help of images representative of each emotion.", "date": "2025-12-26", "categories": []}
{"id": "2512.22741", "title": "Text-Routed Sparse Mixture-of-Experts Model with Explanation and Temporal Alignment for Multi-Modal Sentiment Analysis", "abstract": "Human-interaction-involved applications underscore the need for Multi-modal Sentiment Analysis (MSA). Although many approaches have been proposed to address the subtle emotions in different modalities, the power of explanations and temporal alignments is still underexplored. Thus, this paper proposes the Text-routed sparse mixture-of-Experts model with eXplanation and Temporal alignment for MSA (TEXT). TEXT first augments explanations for MSA via Multi-modal Large Language Models (MLLM), and then novelly aligns the epresentations of audio and video through a temporality-oriented neural network block. TEXT aligns different modalities with explanations and facilitates a new text-routed sparse mixture-of-experts with gate fusion. Our temporal alignment block merges the benefits of Mamba and temporal cross-attention. As a result, TEXT achieves the best performance cross four datasets among all tested models, including three recently proposed approaches and three MLLMs. TEXT wins on at least four metrics out of all six metrics. For example, TEXT decreases the mean absolute error to 0.353 on the CH-SIMS dataset, which signifies a 13.5% decrement compared with recently proposed approaches.", "date": "2025-12-28", "categories": []}
{"id": "2512.22823", "title": "NepEMO: A Multi-Label Emotion and Sentiment Analysis on Nepali Reddit with Linguistic Insights and Temporal Trends", "abstract": "Social media (SM) platforms (e.g. Facebook, Twitter, and Reddit) are increasingly leveraged to share opinions and emotions, specifically during challenging events, such as natural disasters, pandemics, and political elections, and joyful occasions like festivals and celebrations. Among the SM platforms, Reddit provides a unique space for its users to anonymously express their experiences and thoughts on sensitive issues such as health and daily life. In this work, we present a novel dataset, called NepEMO, for multi-label emotion (MLE) and sentiment classification (SC) on the Nepali subreddit post. We curate and build a manually annotated dataset of 4,462 posts (January 2019- June 2025) written in English, Romanised Nepali and Devanagari script for five emotions (fear, anger, sadness, joy, and depression) and three sentiment classes (positive, negative, and neutral). We perform a detailed analysis of posts to capture linguistic insights, including emotion trends, co-occurrence of emotions, sentiment-specific n-grams, and topic modelling using Latent Dirichlet Allocation and TF-IDF keyword extraction. Finally, we compare various traditional machine learning (ML), deep learning (DL), and transformer models for MLE and SC tasks. The result shows that transformer models consistently outperform the ML and DL models for both tasks.", "date": "2025-12-28", "categories": []}
{"id": "2512.23163", "title": "Why We Need a New Framework for Emotional Intelligence in AI", "abstract": "In this paper, we develop the position that current frameworks for evaluating emotional intelligence (EI) in artificial intelligence (AI) systems need refinement because they do not adequately or comprehensively measure the various aspects of EI relevant in AI. Human EI often involves a phenomenological component and a sense of understanding that artificially intelligent systems lack; therefore, some aspects of EI are irrelevant in evaluating AI systems. However, EI also includes an ability to sense an emotional state, explain it, respond appropriately, and adapt to new contexts (e.g., multicultural), and artificially intelligent systems can do such things to greater or lesser degrees. Several benchmark frameworks specialize in evaluating the capacity of different AI models to perform some tasks related to EI, but these often lack a solid foundation regarding the nature of emotion and what it is to be emotionally intelligent. In this project, we begin by reviewing different theories about emotion and general EI, evaluating the extent to which each is applicable to artificial systems. We then critically evaluate the available benchmark frameworks, identifying where each falls short in light of the account of EI developed in the first section. Lastly, we outline some options for improving evaluation strategies to avoid these shortcomings in EI evaluation in AI systems.", "date": "2025-12-29", "categories": []}
{"id": "2512.23291", "title": "Multi-Track Multimodal Learning on iMiGUE: Micro-Gesture and Emotion Recognition", "abstract": "Micro-gesture recognition and behavior-based emotion prediction are both highly challenging tasks that require modeling subtle, fine-grained human behaviors, primarily leveraging video and skeletal pose data. In this work, we present two multimodal frameworks designed to tackle both problems on the iMiGUE dataset. For micro-gesture classification, we explore the complementary strengths of RGB and 3D pose-based representations to capture nuanced spatio-temporal patterns. To comprehensively represent gestures, video, and skeletal embeddings are extracted using MViTv2-S and 2s-AGCN, respectively. Then, they are integrated through a Cross-Modal Token Fusion module to combine spatial and pose information. For emotion recognition, our framework extends to behavior-based emotion prediction, a binary classification task identifying emotional states based on visual cues. We leverage facial and contextual embeddings extracted using SwinFace and MViTv2-S models and fuse them through an InterFusion module designed to capture emotional expressions and body gestures. Experiments conducted on the iMiGUE dataset, within the scope of the MiGA 2025 Challenge, demonstrate the robust performance and accuracy of our method in the behavior-based emotion prediction task, where our approach secured 2nd place.", "date": "2025-12-29", "categories": []}
{"id": "2304.09914", "title": "The Face of Populism: Examining Differences in Facial Emotional Expressions of Political Leaders Using Machine Learning", "abstract": "Populist rhetoric employed on online media is characterized as deeply impassioned and often imbued with strong emotions. The aim of this paper is to empirically investigate the differences in affective nonverbal communication of political leaders. We use a deep-learning approach to process a sample of 220 YouTube videos of political leaders from 15 different countries, analyze their facial expressions of emotion and then examine differences in average emotion scores representing the relative presence of 6 emotional states (anger, disgust, fear, happiness, sadness, and surprise) and a neutral expression for each frame of the YouTube video. Based on a sample of manually coded images, we find that this deep-learning approach has 53-60\\% agreement with human labels. We observe statistically significant differences in the average score of negative emotions between groups of leaders with varying degrees of populist rhetoric.", "date": "2023-04-19", "categories": []}
{"id": "2505.18699", "title": "Affective Image Editing: Shaping Emotional Factors via Text Descriptions", "abstract": "In daily life, images as common affective stimuli have widespread applications. Despite significant progress in text-driven image editing, there is limited work focusing on understanding users' emotional requests. In this paper, we introduce AIEdiT for Affective Image Editing using Text descriptions, which evokes specific emotions by adaptively shaping multiple emotional factors across the entire images. To represent universal emotional priors, we build the continuous emotional spectrum and extract nuanced emotional requests. To manipulate emotional factors, we design the emotional mapper to translate visually-abstract emotional requests to visually-concrete semantic representations. To ensure that editing results evoke specific emotions, we introduce an MLLM to supervise the model training. During inference, we strategically distort visual elements and subsequently shape corresponding emotional factors to edit images according to users' instructions. Additionally, we introduce a large-scale dataset that includes the emotion-aligned text and image pair set for training and evaluation. Extensive experiments demonstrate that AIEdiT achieves superior performance, effectively reflecting users' emotional requests.", "date": "2025-05-24", "categories": []}
{"id": "2512.17376", "title": "Towards Deeper Emotional Reflection: Crafting Affective Image Filters with Generative Priors", "abstract": "Social media platforms enable users to express emotions by posting text with accompanying images. In this paper, we propose the Affective Image Filter (AIF) task, which aims to reflect visually-abstract emotions from text into visually-concrete images, thereby creating emotionally compelling results. We first introduce the AIF dataset and the formulation of the AIF models. Then, we present AIF-B as an initial attempt based on a multi-modal transformer architecture. After that, we propose AIF-D as an extension of AIF-B towards deeper emotional reflection, effectively leveraging generative priors from pre-trained large-scale diffusion models. Quantitative and qualitative experiments demonstrate that AIF models achieve superior performance for both content consistency and emotional fidelity compared to state-of-the-art methods. Extensive user study experiments demonstrate that AIF models are significantly more effective at evoking specific emotions. Based on the presented results, we comprehensively discuss the value and potential of AIF models.", "date": "2025-12-19", "categories": []}
{"id": "2512.20548", "title": "Advancing Multimodal Teacher Sentiment Analysis:The Large-Scale T-MED Dataset & The Effective AAM-TSA Model", "abstract": "Teachers' emotional states are critical in educational scenarios, profoundly impacting teaching efficacy, student engagement, and learning achievements. However, existing studies often fail to accurately capture teachers' emotions due to the performative nature and overlook the critical impact of instructional information on emotional expression. In this paper, we systematically investigate teacher sentiment analysis by building both the dataset and the model accordingly. We construct the first large-scale teacher multimodal sentiment analysis dataset, T-MED. To ensure labeling accuracy and efficiency, we employ a human-machine collaborative labeling process. The T-MED dataset includes 14,938 instances of teacher emotional data from 250 real classrooms across 11 subjects ranging from K-12 to higher education, integrating multimodal text, audio, video, and instructional information. Furthermore, we propose a novel asymmetric attention-based multimodal teacher sentiment analysis model, AAM-TSA. AAM-TSA introduces an asymmetric attention mechanism and hierarchical gating unit to enable differentiated cross-modal feature fusion and precise emotional classification. Experimental results demonstrate that AAM-TSA significantly outperforms existing state-of-the-art methods in terms of accuracy and interpretability on the T-MED dataset.", "date": "2025-12-23", "categories": []}
{"id": "2508.09337", "title": "Decoding Neural Emotion Patterns through Large Language Model Embeddings", "abstract": "Understanding how emotional expression in language relates to brain function is a challenge in computational neuroscience and affective computing. Traditional neuroimaging is costly and lab-bound, but abundant digital text offers new avenues for emotion-brain mapping. Prior work has largely examined neuroimaging-based emotion localization or computational text analysis separately, with little integration. We propose a computational framework that maps textual emotional content to anatomically defined brain regions without requiring neuroimaging. Using OpenAI's text-embedding-ada-002, we generate high-dimensional semantic representations, apply dimensionality reduction and clustering to identify emotional groups, and map them to 18 brain regions linked to emotional processing. Three experiments were conducted: i) analyzing conversational data from healthy vs. depressed subjects (DIAC-WOZ dataset) to compare mapping patterns, ii) applying the method to the GoEmotions dataset and iii) comparing human-written text with large language model (LLM) responses to assess differences in inferred brain activation. Emotional intensity was scored via lexical analysis. Results showed neuroanatomically plausible mappings with high spatial specificity. Depressed subjects exhibited greater limbic engagement tied to negative affect. Discrete emotions were successfully differentiated. LLM-generated text matched humans in basic emotion distribution but lacked nuanced activation in empathy and self-referential regions (medial prefrontal and posterior cingulate cortex). This cost-effective, scalable approach enables large-scale analysis of naturalistic language, distinguishes between clinical populations, and offers a brain-based benchmark for evaluating AI emotional expression.", "date": "2025-08-12", "categories": []}
{"id": "2511.21728", "title": "Affective Multimodal Agents with Proactive Knowledge Grounding for Emotionally Aligned Marketing Dialogue", "abstract": "Recent advances in large language models (LLMs) have enabled fluent dialogue systems, but most remain reactive and struggle in emotionally rich, goal-oriented settings such as marketing conversations. To address this limitation, we propose AffectMind, a multimodal affective dialogue agent that performs proactive reasoning and dynamic knowledge grounding to sustain emotionally aligned and persuasive interactions. AffectMind combines three components: a Proactive Knowledge Grounding Network (PKGN) that continuously updates factual and affective context from text, vision, and prosody; an Emotion--Intent Alignment Model (EIAM) that jointly models user emotion and purchase intent to adapt persuasion strategies; and a Reinforced Discourse Loop (RDL) that optimizes emotional coherence and engagement via reinforcement signals from user responses. Experiments on two newly curated marketing dialogue datasets, MM-ConvMarket and AffectPromo, show that AffectMind outperforms strong LLM-based baselines in emotional consistency (+26\\%), persuasive success rate (+19\\%), and long-term user engagement (+23\\%), highlighting emotion-grounded proactivity as a key capability for commercial multimodal agents.", "date": "2025-11-21", "categories": []}
{"id": "2512.19479", "title": "Emotion-Director: Bridging Affective Shortcut in Emotion-Oriented Image Generation", "abstract": "Image generation based on diffusion models has demonstrated impressive capability, motivating exploration into diverse and specialized applications. Owing to the importance of emotion in advertising, emotion-oriented image generation has attracted increasing attention. However, current emotion-oriented methods suffer from an affective shortcut, where emotions are approximated to semantics. As evidenced by two decades of research, emotion is not equivalent to semantics. To this end, we propose Emotion-Director, a cross-modal collaboration framework consisting of two modules. First, we propose a cross-Modal Collaborative diffusion model, abbreviated as MC-Diffusion. MC-Diffusion integrates visual prompts with textual prompts for guidance, enabling the generation of emotion-oriented images beyond semantics. Further, we improve the DPO optimization by a negative visual prompt, enhancing the model's sensitivity to different emotions under the same semantics. Second, we propose MC-Agent, a cross-Modal Collaborative Agent system that rewrites textual prompts to express the intended emotions. To avoid template-like rewrites, MC-Agent employs multi-agents to simulate human subjectivity toward emotions, and adopts a chain-of-concept workflow that improves the visual expressiveness of the rewritten prompts. Extensive qualitative and quantitative experiments demonstrate the superiority of Emotion-Director in emotion-oriented image generation.", "date": "2025-12-22", "categories": []}
{"id": "2512.15528", "title": "EmoCaliber: Advancing Reliable Visual Emotion Comprehension via Confidence Verbalization and Calibration", "abstract": "Visual Emotion Comprehension (VEC) aims to infer sentiment polarities or emotion categories from affective cues embedded in images. In recent years, Multimodal Large Language Models (MLLMs) have established a popular paradigm in VEC, leveraging their generalizability to unify VEC tasks defined under diverse emotion taxonomies. While this paradigm achieves notable success, it typically formulates VEC as a deterministic task, requiring the model to output a single, definitive emotion label for each image. Such a formulation insufficiently accounts for the inherent subjectivity of emotion perception, overlooking alternative interpretations that may be equally plausible to different viewers. To address this limitation, we propose equipping MLLMs with capabilities to verbalize their confidence in emotion predictions. This additional signal provides users with an estimate of both the plausibility of alternative interpretations and the MLLMs' self-assessed competence, thereby enhancing reliability in practice. Building on this insight, we introduce a three-stage training framework that progressively endows with structured reasoning, teaches to verbalize confidence, and calibrates confidence expression, culminating in EmoCaliber, a confidence-aware MLLM for VEC. Through fair and comprehensive evaluations on the unified benchmark VECBench, EmoCaliber demonstrates overall superiority against existing methods in both emotion prediction and confidence estimation. These results validate the effectiveness of our approach and mark a feasible step toward more reliable VEC systems. Project page: https://github.com/wdqqdw/EmoCaliber.", "date": "2025-12-17", "categories": []}
{"id": "2512.17630", "title": "Confidence-Credibility Aware Weighted Ensembles of Small LLMs Outperform Large LLMs in Emotion Detection", "abstract": "This paper introduces a confidence-weighted, credibility-aware ensemble framework for text-based emotion detection, inspired by Condorcet's Jury Theorem (CJT). Unlike conventional ensembles that often rely on homogeneous architectures, our approach combines architecturally diverse small transformer-based large language models (sLLMs) - BERT, RoBERTa, DistilBERT, DeBERTa, and ELECTRA, each fully fine-tuned for emotion classification. To preserve error diversity, we minimize parameter convergence while taking advantage of the unique biases of each model. A dual-weighted voting mechanism integrates both global credibility (validation F1 score) and local confidence (instance-level probability) to dynamically weight model contributions. Experiments on the DAIR-AI dataset demonstrate that our credibility-confidence ensemble achieves a macro F1 score of 93.5 percent, surpassing state-of-the-art benchmarks and significantly outperforming large-scale LLMs, including Falcon, Mistral, Qwen, and Phi, even after task-specific Low-Rank Adaptation (LoRA). With only 595M parameters in total, our small LLMs ensemble proves more parameter-efficient and robust than models up to 7B parameters, establishing that carefully designed ensembles of small, fine-tuned models can outperform much larger LLMs in specialized natural language processing (NLP) tasks such as emotion detection.", "date": "2025-12-19", "categories": []}
{"id": "2501.12424", "title": "Multi-Modality Collaborative Learning for Sentiment Analysis", "abstract": "Multimodal sentiment analysis (MSA) identifies individuals' sentiment states in videos by integrating visual, audio, and text modalities. Despite progress in existing methods, the inherent modality heterogeneity limits the effective capture of interactive sentiment features across modalities. In this paper, by introducing a Multi-Modality Collaborative Learning (MMCL) framework, we facilitate cross-modal interactions and capture enhanced and complementary features from modality-common and modality-specific representations, respectively. Specifically, we design a parameter-free decoupling module and separate uni-modality into modality-common and modality-specific components through semantics assessment of cross-modal elements. For modality-specific representations, inspired by the act-reward mechanism in reinforcement learning, we design policy models to adaptively mine complementary sentiment features under the guidance of a joint reward. For modality-common representations, intra-modal attention is employed to highlight crucial components, playing enhanced roles among modalities. Experimental results, including superiority evaluations on four databases, effectiveness verification of each module, and assessment of complementary features, demonstrate that MMCL successfully learns collaborative features across modalities and significantly improves performance. The code can be available at https://github.com/smwanghhh/MMCL.", "date": "2025-01-21", "categories": []}
{"id": "2505.24342", "title": "VAEER: Visual Attention-Inspired Emotion Elicitation Reasoning", "abstract": "Images shared online strongly influence emotions and public well-being. Understanding the emotions an image elicits is therefore vital for fostering healthier and more sustainable digital communities, especially during public crises. We study Visual Emotion Elicitation (VEE), predicting the set of emotions that an image evokes in viewers. We introduce VAEER, an interpretable multi-label VEE framework that combines attention-inspired cue extraction with knowledge-grounded reasoning. VAEER isolates salient visual foci and contextual signals, aligns them with structured affective knowledge, and performs per-emotion inference to yield transparent, emotion-specific rationales. Across three heterogeneous benchmarks, including social imagery and disaster-related photos, VAEER achieves state-of-the-art results with up to 19% per-emotion improvements and a 12.3% average gain over strong CNN and VLM baselines. Our findings highlight interpretable multi-label emotion elicitation as a scalable foundation for responsible visual media analysis and emotionally sustainable online ecosystems.", "date": "2025-05-30", "categories": []}
{"id": "2512.16485", "title": "Smile on the Face, Sadness in the Eyes: Bridging the Emotion Gap with a Multimodal Dataset of Eye and Facial Behaviors", "abstract": "Emotion Recognition (ER) is the process of analyzing and identifying human emotions from sensing data. Currently, the field heavily relies on facial expression recognition (FER) because visual channel conveys rich emotional cues. However, facial expressions are often used as social tools rather than manifestations of genuine inner emotions. To understand and bridge this gap between FER and ER, we introduce eye behaviors as an important emotional cue and construct an Eye-behavior-aided Multimodal Emotion Recognition (EMER) dataset. To collect data with genuine emotions, spontaneous emotion induction paradigm is exploited with stimulus material, during which non-invasive eye behavior data, like eye movement sequences and eye fixation maps, is captured together with facial expression videos. To better illustrate the gap between ER and FER, multi-view emotion labels for mutimodal ER and FER are separately annotated. Furthermore, based on the new dataset, we design a simple yet effective Eye-behavior-aided MER Transformer (EMERT) that enhances ER by bridging the emotion gap. EMERT leverages modality-adversarial feature decoupling and a multitask Transformer to model eye behaviors as a strong complement to facial expressions. In the experiment, we introduce seven multimodal benchmark protocols for a variety of comprehensive evaluations of the EMER dataset. The results show that the EMERT outperforms other state-of-the-art multimodal methods by a great margin, revealing the importance of modeling eye behaviors for robust ER. To sum up, we provide a comprehensive analysis of the importance of eye behaviors in ER, advancing the study on addressing the gap between FER and ER for more robust ER performance. Our EMER dataset and the trained EMERT models will be publicly available at https://github.com/kejun1/EMER.", "date": "2025-12-18", "categories": []}
{"id": "2512.13998", "title": "Memo2496: Expert-Annotated Dataset and Dual-View Adaptive Framework for Music Emotion Recognition", "abstract": "Music Emotion Recogniser (MER) research faces challenges due to limited high-quality annotated datasets and difficulties in addressing cross-track feature drift. This work presents two primary contributions to address these issues. Memo2496, a large-scale dataset, offers 2496 instrumental music tracks with continuous valence arousal labels, annotated by 30 certified music specialists. Annotation quality is ensured through calibration with extreme emotion exemplars and a consistency threshold of 0.25, measured by Euclidean distance in the valence arousal space. Furthermore, the Dual-view Adaptive Music Emotion Recogniser (DAMER) is introduced. DAMER integrates three synergistic modules: Dual Stream Attention Fusion (DSAF) facilitates token-level bidirectional interaction between Mel spectrograms and cochleagrams via cross attention mechanisms; Progressive Confidence Labelling (PCL) generates reliable pseudo labels employing curriculum-based temperature scheduling and consistency quantification using Jensen Shannon divergence; and Style Anchored Memory Learning (SAML) maintains a contrastive memory queue to mitigate cross-track feature drift. Extensive experiments on the Memo2496, 1000songs, and PMEmo datasets demonstrate DAMER's state-of-the-art performance, improving arousal dimension accuracy by 3.43%, 2.25%, and 0.17%, respectively. Ablation studies and visualisation analyses validate each module's contribution. Both the dataset and source code are publicly available.", "date": "2025-12-16", "categories": []}
{"id": "2512.15376", "title": "Emotion Recognition in Signers", "abstract": "Recognition of signers' emotions suffers from one theoretical challenge and one practical challenge, namely, the overlap between grammatical and affective facial expressions and the scarcity of data for model training. This paper addresses these two challenges in a cross-lingual setting using our eJSL dataset, a new benchmark dataset for emotion recognition in Japanese Sign Language signers, and BOBSL, a large British Sign Language dataset with subtitles. In eJSL, two signers expressed 78 distinct utterances with each of seven different emotional states, resulting in 1,092 video clips. We empirically demonstrate that 1) textual emotion recognition in spoken language mitigates data scarcity in sign language, 2) temporal segment selection has a significant impact, and 3) incorporating hand motion enhances emotion recognition in signers. Finally we establish a stronger baseline than spoken language LLMs.", "date": "2025-12-17", "categories": []}
{"id": "2507.06080", "title": "CAST-Phys: Contactless Affective States Through Physiological signals Database", "abstract": "In recent years, affective computing and its applications have become a fast-growing research topic. Despite significant advancements, the lack of affective multi-modal datasets remains a major bottleneck in developing accurate emotion recognition systems. Furthermore, the use of contact-based devices during emotion elicitation often unintentionally influences the emotional experience, reducing or altering the genuine spontaneous emotional response. This limitation highlights the need for methods capable of extracting affective cues from multiple modalities without physical contact, such as remote physiological emotion recognition. To address this, we present the Contactless Affective States Through Physiological Signals Database (CAST-Phys), a novel high-quality dataset explicitly designed for multi-modal remote physiological emotion recognition using facial and physiological cues. The dataset includes diverse physiological signals, such as photoplethysmography (PPG), electrodermal activity (EDA), and respiration rate (RR), alongside high-resolution uncompressed facial video recordings, enabling the potential for remote signal recovery. Our analysis highlights the crucial role of physiological signals in realistic scenarios where facial expressions alone may not provide sufficient emotional information. Furthermore, we demonstrate the potential of remote multi-modal emotion recognition by evaluating the impact of individual and fused modalities, showcasing its effectiveness in advancing contactless emotion recognition technologies.", "date": "2025-07-08", "categories": []}
{"id": "2512.11905", "title": "Smartphone monitoring of smiling as a behavioral proxy of well-being in everyday life", "abstract": "Subjective well-being is a cornerstone of individual and societal health, yet its scientific measurement has traditionally relied on self-report methods prone to recall bias and high participant burden. This has left a gap in our understanding of well-being as it is expressed in everyday life. We hypothesized that candid smiles captured during natural smartphone interactions could serve as a scalable, objective behavioral correlate of positive affect. To test this, we analyzed 405,448 video clips passively recorded from 233 consented participants over one week. Using a deep learning model to quantify smile intensity, we identified distinct diurnal and daily patterns. Daily patterns of smile intensity across the week showed strong correlation with national survey data on happiness (r=0.92), and diurnal rhythms documented close correspondence with established results from the day reconstruction method (r=0.80). Higher daily mean smile intensity was significantly associated with more physical activity (Beta coefficient = 0.043, 95% CI [0.001, 0.085]) and greater light exposure (Beta coefficient = 0.038, [0.013, 0.063]), whereas no significant effects were found for smartphone use. These findings suggest that passive smartphone sensing could serve as a powerful, ecologically valid methodology for studying the dynamics of affective behavior and open the door to understanding this behavior at a population scale.", "date": "2025-12-10", "categories": []}
{"id": "2512.12208", "title": "A Hybrid Deep Learning Framework for Emotion Recognition in Children with Autism During NAO Robot-Mediated Interaction", "abstract": "Understanding emotional responses in children with Autism Spectrum Disorder (ASD) during social interaction remains a critical challenge in both developmental psychology and human-robot interaction. This study presents a novel deep learning pipeline for emotion recognition in autistic children in response to a name-calling event by a humanoid robot (NAO), under controlled experimental settings. The dataset comprises of around 50,000 facial frames extracted from video recordings of 15 children with ASD. A hybrid model combining a fine-tuned ResNet-50-based Convolutional Neural Network (CNN) and a three-layer Graph Convolutional Network (GCN) trained on both visual and geometric features extracted from MediaPipe FaceMesh landmarks. Emotions were probabilistically labeled using a weighted ensemble of two models: DeepFace's and FER, each contributing to soft-label generation across seven emotion classes. Final classification leveraged a fused embedding optimized via Kullback-Leibler divergence. The proposed method demonstrates robust performance in modeling subtle affective responses and offers significant promise for affective profiling of ASD children in clinical and therapeutic human-robot interaction contexts, as the pipeline effectively captures micro emotional cues in neurodivergent children, addressing a major gap in autism-specific HRI research. This work represents the first such large-scale, real-world dataset and pipeline from India on autism-focused emotion analysis using social robotics, contributing an essential foundation for future personalized assistive technologies.", "date": "2025-12-13", "categories": []}
{"id": "2512.12277", "title": "Feature Aggregation for Efficient Continual Learning of Complex Facial Expressions", "abstract": "As artificial intelligence (AI) systems become increasingly embedded in our daily life, the ability to recognize and adapt to human emotions is essential for effective human-computer interaction. Facial expression recognition (FER) provides a primary channel for inferring affective states, but the dynamic and culturally nuanced nature of emotions requires models that can learn continuously without forgetting prior knowledge. In this work, we propose a hybrid framework for FER in a continual learning setting that mitigates catastrophic forgetting. Our approach integrates two complementary modalities: deep convolutional features and facial Action Units (AUs) derived from the Facial Action Coding System (FACS). The combined representation is modelled through Bayesian Gaussian Mixture Models (BGMMs), which provide a lightweight, probabilistic solution that avoids retraining while offering strong discriminative power. Using the Compound Facial Expression of Emotion (CFEE) dataset, we show that our model can first learn basic expressions and then progressively recognize compound expressions. Experiments demonstrate improved accuracy, stronger knowledge retention, and reduced forgetting. This framework contributes to the development of emotionally intelligent AI systems with applications in education, healthcare, and adaptive user interfaces.", "date": "2025-12-13", "categories": []}
{"id": "2512.13363", "title": "Detecting Emotion Drift in Mental Health Text Using Pre-Trained Transformers", "abstract": "This study investigates emotion drift: the change in emotional state across a single text, within mental health-related messages. While sentiment analysis typically classifies an entire message as positive, negative, or neutral, the nuanced shift of emotions over the course of a message is often overlooked. This study detects sentence-level emotions and measures emotion drift scores using pre-trained transformer models such as DistilBERT and RoBERTa. The results provide insights into patterns of emotional escalation or relief in mental health conversations. This methodology can be applied to better understand emotional dynamics in content.", "date": "2025-12-15", "categories": []}
{"id": "2511.06782", "title": "HEDN: A Hard-Easy Dual Network with Source Reliability Assessment for Cross-Subject EEG Emotion Recognition", "abstract": "Cross-subject electroencephalography (EEG) emotion recognition remains a major challenge in brain-computer interfaces (BCIs) due to substantial inter-subject variability. Multi-Source Domain Adaptation (MSDA) offers a potential solution, but existing MSDA frameworks typically assume equal source quality, leading to negative transfer from low-reliability domains and prohibitive computational overhead due to multi-branch model designs. To address these limitations, we propose the Hard-Easy Dual Network (HEDN), a lightweight reliability-aware MSDA framework. HEDN introduces a novel Source Reliability Assessment (SRA) mechanism that dynamically evaluates the structural integrity of each source domain during training. Based on this assessment, sources are routed to two specialized branches: an Easy Network that exploits high-quality sources to construct fine-grained, structure-aware prototypes for reliable pseudo-label generation, and a Hard Network that utilizes adversarial training to refine and align low-quality sources. Furthermore, a cross-network consistency loss aligns predictions between branches to preserve semantic coherence. Extensive experiments conducted on SEED, SEED-IV, and DEAP datasets demonstrate that HEDN achieves state-of-the-art performance across both cross-subject and cross-dataset evaluation protocols while reducing adaptation complexity.", "date": "2025-11-10", "categories": []}
{"id": "2512.10963", "title": "Emotion-Driven Personalized Recommendation for AI-Generated Content Using Multi-Modal Sentiment and Intent Analysis", "abstract": "With the rapid growth of AI-generated content (AIGC) across domains such as music, video, and literature, the demand for emotionally aware recommendation systems has become increasingly important. Traditional recommender systems primarily rely on user behavioral data such as clicks, views, or ratings, while neglecting users' real-time emotional and intentional states during content interaction. To address this limitation, this study proposes a Multi-Modal Emotion and Intent Recognition Model (MMEI) based on a BERT-based Cross-Modal Transformer with Attention-Based Fusion, integrated into a cloud-native personalized AIGC recommendation framework. The proposed system jointly processes visual (facial expression), auditory (speech tone), and textual (comments or utterances) modalities through pretrained encoders ViT, Wav2Vec2, and BERT, followed by an attention-based fusion module to learn emotion-intent representations. These embeddings are then used to drive personalized content recommendations through a contextual matching layer. Experiments conducted on benchmark emotion datasets (AIGC-INT, MELD, and CMU-MOSEI) and an AIGC interaction dataset demonstrate that the proposed MMEI model achieves a 4.3% improvement in F1-score and a 12.3% reduction in cross-entropy loss compared to the best fusion-based transformer baseline. Furthermore, user-level online evaluations reveal that emotion-driven recommendations increase engagement time by 15.2% and enhance satisfaction scores by 11.8%, confirming the model's effectiveness in aligning AI-generated content with users' affective and intentional states. This work highlights the potential of cross-modal emotional intelligence for next-generation AIGC ecosystems, enabling adaptive, empathetic, and context-aware recommendation experiences.", "date": "2025-11-25", "categories": []}
{"id": "2512.10975", "title": "Agent-Based Modular Learning for Multimodal Emotion Recognition in Human-Agent Systems", "abstract": "Effective human-agent interaction (HAI) relies on accurate and adaptive perception of human emotional states. While multimodal deep learning models - leveraging facial expressions, speech, and textual cues - offer high accuracy in emotion recognition, their training and maintenance are often computationally intensive and inflexible to modality changes. In this work, we propose a novel multi-agent framework for training multimodal emotion recognition systems, where each modality encoder and the fusion classifier operate as autonomous agents coordinated by a central supervisor. This architecture enables modular integration of new modalities (e.g., audio features via emotion2vec), seamless replacement of outdated components, and reduced computational overhead during training. We demonstrate the feasibility of our approach through a proof-of-concept implementation supporting vision, audio, and text modalities, with the classifier serving as a shared decision-making agent. Our framework not only improves training efficiency but also contributes to the design of more flexible, scalable, and maintainable perception modules for embodied and virtual agents in HAI scenarios.", "date": "2025-12-02", "categories": []}
{"id": "2512.10441", "title": "Decoding Student Minds: Leveraging Conversational Agents for Psychological and Learning Analysis", "abstract": "This paper presents a psychologically-aware conversational agent designed to enhance both learning performance and emotional well-being in educational settings. The system combines Large Language Models (LLMs), a knowledge graph-enhanced BERT (KG-BERT), and a bidirectional Long Short-Term Memory (LSTM) with attention to classify students' cognitive and affective states in real time. Unlike prior chatbots limited to either tutoring or affective support, our approach leverages multimodal data-including textual semantics, prosodic speech features, and temporal behavioral trends-to infer engagement, stress, and conceptual understanding. A pilot study with university students demonstrated improved motivation, reduced stress, and moderate academic gains compared to baseline methods. These results underline the promise of integrating semantic reasoning, multimodal fusion, and temporal modeling to support adaptive, student-centered educational interventions.", "date": "2025-12-11", "categories": []}
{"id": "2512.09592", "title": "CS3D: An Efficient Facial Expression Recognition via Event Vision", "abstract": "Responsive and accurate facial expression recognition is crucial to human-robot interaction for daily service robots. Nowadays, event cameras are becoming more widely adopted as they surpass RGB cameras in capturing facial expression changes due to their high temporal resolution, low latency, computational efficiency, and robustness in low-light conditions. Despite these advantages, event-based approaches still encounter practical challenges, particularly in adopting mainstream deep learning models. Traditional deep learning methods for facial expression analysis are energy-intensive, making them difficult to deploy on edge computing devices and thereby increasing costs, especially for high-frequency, dynamic, event vision-based approaches. To address this challenging issue, we proposed the CS3D framework by decomposing the Convolutional 3D method to reduce the computational complexity and energy consumption. Additionally, by utilizing soft spiking neurons and a spatial-temporal attention mechanism, the ability to retain information is enhanced, thus improving the accuracy of facial expression detection. Experimental results indicate that our proposed CS3D method attains higher accuracy on multiple datasets compared to architectures such as the RNN, Transformer, and C3D, while the energy consumption of the CS3D method is just 21.97\\% of the original C3D required on the same device.", "date": "2025-12-10", "categories": []}
{"id": "2512.06447", "title": "Towards Stable Cross-Domain Depression Recognition under Missing Modalities", "abstract": "Depression poses serious public health risks, including suicide, underscoring the urgency of timely and scalable screening. Multimodal automatic depression detection (ADD) offers a promising solution; however, widely studied audio- and video-based ADD methods lack a unified, generalizable framework for diverse depression recognition scenarios and show limited stability to missing modalities, which are common in real-world data. In this work, we propose a unified framework for Stable Cross-Domain Depression Recognition based on Multimodal Large Language Model (SCD-MLLM). The framework supports the integration and processing of heterogeneous depression-related data collected from varied sources while maintaining stability in the presence of incomplete modality inputs. Specifically, SCD-MLLM introduces two key components: (i) Multi-Source Data Input Adapter (MDIA), which employs masking mechanism and task-specific prompts to transform heterogeneous depression-related inputs into uniform token sequences, addressing inconsistency across diverse data sources; (ii) Modality-Aware Adaptive Fusion Module (MAFM), which adaptively integrates audio and visual features via a shared projection mechanism, enhancing resilience under missing modality conditions. e conduct comprehensive experiments under multi-dataset joint training settings on five publicly available and heterogeneous depression datasets from diverse scenarios: CMDC, AVEC2014, DAIC-WOZ, DVlog, and EATD. Across both complete and partial modality settings, SCD-MLLM outperforms state-of-the-art (SOTA) models as well as leading commercial LLMs (Gemini and GPT), demonstrating superior cross-domain generalization, enhanced ability to capture multimodal cues of depression, and strong stability to missing modality cases in real-world applications.", "date": "2025-12-06", "categories": []}
{"id": "2512.07430", "title": "MIDG: Mixture of Invariant Experts with knowledge injection for Domain Generalization in Multimodal Sentiment Analysis", "abstract": "Existing methods in domain generalization for Multimodal Sentiment Analysis (MSA) often overlook inter-modal synergies during invariant features extraction, which prevents the accurate capture of the rich semantic information within multimodal data. Additionally, while knowledge injection techniques have been explored in MSA, they often suffer from fragmented cross-modal knowledge, overlooking specific representations that exist beyond the confines of unimodal. To address these limitations, we propose a novel MSA framework designed for domain generalization. Firstly, the framework incorporates a Mixture of Invariant Experts model to extract domain-invariant features, thereby enhancing the model's capacity to learn synergistic relationships between modalities. Secondly, we design a Cross-Modal Adapter to augment the semantic richness of multimodal representations through cross-modal knowledge injection. Extensive domain experiments conducted on three datasets demonstrate that the proposed MIDG achieves superior performance.", "date": "2025-12-08", "categories": []}
{"id": "2512.05515", "title": "DashFusion: Dual-stream Alignment with Hierarchical Bottleneck Fusion for Multimodal Sentiment Analysis", "abstract": "Multimodal sentiment analysis (MSA) integrates various modalities, such as text, image, and audio, to provide a more comprehensive understanding of sentiment. However, effective MSA is challenged by alignment and fusion issues. Alignment requires synchronizing both temporal and semantic information across modalities, while fusion involves integrating these aligned features into a unified representation. Existing methods often address alignment or fusion in isolation, leading to limitations in performance and efficiency. To tackle these issues, we propose a novel framework called Dual-stream Alignment with Hierarchical Bottleneck Fusion (DashFusion). Firstly, dual-stream alignment module synchronizes multimodal features through temporal and semantic alignment. Temporal alignment employs cross-modal attention to establish frame-level correspondences among multimodal sequences. Semantic alignment ensures consistency across the feature space through contrastive learning. Secondly, supervised contrastive learning leverages label information to refine the modality features. Finally, hierarchical bottleneck fusion progressively integrates multimodal information through compressed bottleneck tokens, which achieves a balance between performance and computational efficiency. We evaluate DashFusion on three datasets: CMU-MOSI, CMU-MOSEI, and CH-SIMS. Experimental results demonstrate that DashFusion achieves state-of-the-art performance across various metrics, and ablation studies confirm the effectiveness of our alignment and fusion techniques. The codes for our experiments are available at https://github.com/ultramarineX/DashFusion.", "date": "2025-12-05", "categories": []}
{"id": "2512.05669", "title": "Deep Learning-Based Real-Time Sequential Facial Expression Analysis Using Geometric Features", "abstract": "Facial expression recognition is a crucial component in enhancing human-computer interaction and developing emotion-aware systems. Real-time detection and interpretation of facial expressions have become increasingly important for various applications, from user experience personalization to intelligent surveillance systems. This study presents a novel approach to real-time sequential facial expression recognition using deep learning and geometric features. The proposed method utilizes MediaPipe FaceMesh for rapid and accurate facial landmark detection. Geometric features, including Euclidean distances and angles, are extracted from these landmarks. Temporal dynamics are incorporated by analyzing feature differences between consecutive frames, enabling the detection of onset, apex, and offset phases of expressions. For classification, a ConvLSTM1D network followed by multilayer perceptron blocks is employed. The method's performance was evaluated on multiple publicly available datasets, including CK+, Oulu-CASIA (VIS and NIR), and MMI. Accuracies of 93%, 79%, 77%, and 68% were achieved respectively. Experiments with composite datasets were also conducted to assess the model's generalization capabilities. The approach demonstrated real-time applicability, processing approximately 165 frames per second on consumer-grade hardware. This research contributes to the field of facial expression analysis by providing a fast, accurate, and adaptable solution. The findings highlight the potential for further advancements in emotion-aware technologies and personalized user experiences, paving the way for more sophisticated human-computer interaction systems. To facilitate further research in this field, the complete source code for this study has been made publicly available on GitHub: https://github.com/miralab-ai/facial-expression-analysis.", "date": "2025-12-05", "categories": []}
{"id": "2512.04728", "title": "Measuring the Unspoken: A Disentanglement Model and Benchmark for Psychological Analysis in the Wild", "abstract": "Generative psychological analysis of in-the-wild conversations faces two fundamental challenges: (1) existing Vision-Language Models (VLMs) fail to resolve Articulatory-Affective Ambiguity, where visual patterns of speech mimic emotional expressions; and (2) progress is stifled by a lack of verifiable evaluation metrics capable of assessing visual grounding and reasoning depth. We propose a complete ecosystem to address these twin challenges. First, we introduce Multilevel Insight Network for Disentanglement(MIND), a novel hierarchical visual encoder that introduces a Status Judgment module to algorithmically suppress ambiguous lip features based on their temporal feature variance, achieving explicit visual disentanglement. Second, we construct ConvoInsight-DB, a new large-scale dataset with expert annotations for micro-expressions and deep psychological inference. Third, Third, we designed the Mental Reasoning Insight Rating Metric (PRISM), an automated dimensional framework that uses expert-guided LLM to measure the multidimensional performance of large mental vision models. On our PRISM benchmark, MIND significantly outperforms all baselines, achieving a +86.95% gain in micro-expression detection over prior SOTA. Ablation studies confirm that our Status Judgment disentanglement module is the most critical component for this performance leap. Our code has been opened.", "date": "2025-12-04", "categories": []}
{"id": "2209.15402", "title": "Rethinking the Learning Paradigm for Facial Expression Recognition", "abstract": "Due to the subjective crowdsourcing annotations and the inherent inter-class similarity of facial expressions, the real-world Facial Expression Recognition (FER) datasets usually exhibit ambiguous annotation. To simplify the learning paradigm, most previous methods convert ambiguous annotation results into precise one-hot annotations and train FER models in an end-to-end supervised manner. In this paper, we rethink the existing training paradigm and propose that it is better to use weakly supervised strategies to train FER models with original ambiguous annotation.", "date": "2022-09-30", "categories": []}
{"id": "2511.18751", "title": "Robust Multimodal Sentiment Analysis of Image-Text Pairs by Distribution-Based Feature Recovery and Fusion", "abstract": "As posts on social media increase rapidly, analyzing the sentiments embedded in image-text pairs has become a popular research topic in recent years. Although existing works achieve impressive accomplishments in simultaneously harnessing image and text information, they lack the considerations of possible low-quality and missing modalities. In real-world applications, these issues might frequently occur, leading to urgent needs for models capable of predicting sentiment robustly. Therefore, we propose a Distribution-based feature Recovery and Fusion (DRF) method for robust multimodal sentiment analysis of image-text pairs. Specifically, we maintain a feature queue for each modality to approximate their feature distributions, through which we can simultaneously handle low-quality and missing modalities in a unified framework. For low-quality modalities, we reduce their contributions to the fusion by quantitatively estimating modality qualities based on the distributions. For missing modalities, we build inter-modal mapping relationships supervised by samples and distributions, thereby recovering the missing modalities from available ones. In experiments, two disruption strategies that corrupt and discard some modalities in samples are adopted to mimic the low-quality and missing modalities in various real-world scenarios. Through comprehensive experiments on three publicly available image-text datasets, we demonstrate the universal improvements of DRF compared to SOTA methods under both two strategies, validating its effectiveness in robust multimodal sentiment analysis.", "date": "2025-11-24", "categories": []}
{"id": "2512.01282", "title": "Kardia-R1: Unleashing LLMs to Reason toward Understanding and Empathy for Emotional Support via Rubric-as-Judge Reinforcement Learning", "abstract": "As web platforms evolve towards greater personalization and emotional complexity, conversational agents must transcend superficial empathy to demonstrate identity-aware emotional reasoning. However, existing systems face two limitations: (1) reliance on situation-centric datasets lacking persistent user identity, which hampers the capture of personalized affective nuances; and (2) dependence on opaque, coarse reward signals that hinder development of verifiable empathetic reasoning. To address these gaps, we introduce KardiaBench, a large-scale user-grounded benchmark comprising 178,080 QA pairs across 22,080 multi-turn conversations anchored to 671 real-world profiles. The dataset is constructed via a model-in-the-loop pipeline with iterative rubric-guided refinement to ensure psychological plausibility and persona consistency. This progressive empathy pipeline that integrates user comprehension, contextual reasoning, and emotion perception into conversations, followed by iterative critique and rubric-based refinement to ensure psychological plausibility, emotional fidelity, and persona consistency. Building on this, we propose Kardia-R1, a framework that trains models for interpretable, stepwise empathetic cognition. Kardia-R1 leverages Rubric-as-Judge Empathetic Reinforcement Learning (Rubric-ERL), a GRPO-based method that uses explainable, human-aligned rubric rewards to tightly couple user understanding, emotional inference, and supportive response generation. Extensive experiments across four LLM backbones demonstrate that Kardia-R1 consistently outperforms othet methods in emotion accuracy, empathy, relevance, persona consistency, and safety. Our dataset and model will be released at https://github.com/JhCircle/Kardia-R1.", "date": "2025-12-01", "categories": []}
{"id": "2512.02558", "title": "Empathy Level Prediction in Multi-Modal Scenario with Supervisory Documentation Assistance", "abstract": "Prevalent empathy prediction techniques primarily concentrate on a singular modality, typically textual, thus neglecting multi-modal processing capabilities. They also overlook the utilization of certain privileged information, which may encompass additional empathetic content. In response, we introduce an advanced multi-modal empathy prediction method integrating video, audio, and text information. The method comprises the Multi-Modal Empathy Prediction and Supervisory Documentation Assisted Training. We use pre-trained networks in the empathy prediction network to extract features from various modalities, followed by a cross-modal fusion. This process yields a multi-modal feature representation, which is employed to predict empathy labels. To enhance the extraction of text features, we incorporate supervisory documents as privileged information during the assisted training phase. Specifically, we apply the Latent Dirichlet Allocation model to identify potential topic distributions to constrain text features. These supervisory documents, created by supervisors, focus on the counseling topics and the counselor's display of empathy. Notably, this privileged information is only available during training and is not accessible during the prediction phase. Experimental results on the multi-modal and dialogue empathy datasets demonstrate that our approach is superior to the existing methods.", "date": "2025-12-02", "categories": []}
{"id": "2512.02699", "title": "Learning What to Attend First: Modality-Importance-Guided Reasoning for Reliable Multimodal Emotion Understanding", "abstract": "In this paper, we present Modality-Importance-Guided Reasoning (MIGR), a framework designed to improve the reliability of reasoning-based multimodal emotion understanding in multimodal large language models. Although existing methods have advanced emotion understanding, they often suffer from reasoning drift: models gradually rely on their own generated text instead of multimodal evidence, and their explanations are overly shaped by visually initiated reasoning paths. To address these issues, we introduce Modality Importance (MI), a simple yet effective mechanism for identifying the emotion-dominant modality. Using MI, MIGR reorganizes reasoning sequences so that explanations begin from the modality most critical to the target emotion, preventing early reasoning from being misled by less informative cues. Our two-stage framework-comprising modality-aligned supervised fine-tuning and modality-aware reward optimization-encourages models to generate emotionally grounded, causally relevant, and coherence-preserving explanations. Experimental results on the DFEW benchmark show that MIGR substantially improves reasoning reliability, decreasing instances of correct predictions accompanied by emotionally inconsistent explanations from 18.10% to 7.37%. These results confirm the benefit of initiating reasoning from the emotion-dominant modality.", "date": "2025-12-02", "categories": []}
{"id": "2407.16406", "title": "Hi-EF: Benchmarking Emotion Forecasting in Human-interaction", "abstract": "Affective Forecasting is an psychology task that involves predicting an individual's future emotional responses, often hampered by reliance on external factors leading to inaccuracies, and typically remains at a qualitative analysis stage. To address these challenges, we narrows the scope of Affective Forecasting by introducing the concept of Human-interaction-based Emotion Forecasting (EF). This task is set within the context of a two-party interaction, positing that an individual's emotions are significantly influenced by their interaction partner's emotional expressions and informational cues. This dynamic provides a structured perspective for exploring the patterns of emotional change, thereby enhancing the feasibility of emotion forecasting.", "date": "2024-07-23", "categories": []}
{"id": "2511.12077", "title": "Learning to Hear by Seeing: It's Time for Vision Language Models to Understand Artistic Emotion from Sight and Sound", "abstract": "Emotion understanding is critical for making Large Language Models (LLMs) more general, reliable, and aligned with humans. Art conveys emotion through the joint design of visual and auditory elements, yet most prior work is human-centered or single-modality, overlooking the emotion intentionally expressed by the artwork. Meanwhile, current Audio-Visual Language Models (AVLMs) typically require large-scale audio pretraining to endow Visual Language Models (VLMs) with hearing, which limits scalability. We present Vision Anchored Audio-Visual Emotion LLM (VAEmotionLLM), a two-stage framework that teaches a VLM to hear by seeing with limited audio pretraining and to understand emotion across modalities. In Stage 1, Vision-Guided Audio Alignment (VG-Align) distills the frozen visual pathway into a new audio pathway by aligning next-token distributions of the shared LLM on synchronized audio-video clips, enabling hearing without a large audio dataset. In Stage 2, a lightweight Cross-Modal Emotion Adapter (EmoAdapter), composed of the Emotion Enhancer and the Emotion Supervisor, injects emotion-sensitive residuals and applies emotion supervision to enhance cross-modal emotion understanding. We also construct ArtEmoBenchmark, an art-centric emotion benchmark that evaluates content and emotion understanding under audio-only, visual-only, and audio-visual inputs. VAEmotionLLM achieves state-of-the-art results on ArtEmoBenchmark, outperforming audio-only, visual-only, and audio-visual baselines. Ablations show that the proposed components are complementary.", "date": "2025-11-15", "categories": []}
{"id": "2512.01145", "title": "Weakly Supervised Continuous Micro-Expression Intensity Estimation Using Temporal Deep Neural Network", "abstract": "Micro-facial expressions are brief and involuntary facial movements that reflect genuine emotional states. While most prior work focuses on classifying discrete micro-expression categories, far fewer studies address the continuous evolution of intensity over time. Progress in this direction is limited by the lack of frame-level intensity labels, which makes fully supervised regression impractical.\n  We propose a unified framework for continuous micro-expression intensity estimation using only weak temporal labels (onset, apex, offset). A simple triangular prior converts sparse temporal landmarks into dense pseudo-intensity trajectories, and a lightweight temporal regression model that combines a ResNet18 encoder with a bidirectional GRU predicts frame-wise intensity directly from image sequences. The method requires no frame-level annotation effort and is applied consistently across datasets through a single preprocessing and temporal alignment pipeline.\n  Experiments on SAMM and CASME II show strong temporal agreement with the pseudo-intensity trajectories. On SAMM, the model reaches a Spearman correlation of 0.9014 and a Kendall correlation of 0.7999, outperforming a frame-wise baseline. On CASME II, it achieves up to 0.9116 and 0.8168, respectively, when trained without the apex-ranking term. Ablation studies confirm that temporal modeling and structured pseudo labels are central to capturing the rise-apex-fall dynamics of micro-facial movements.\n  To our knowledge, this is the first unified approach for continuous micro-expression intensity estimation using only sparse temporal annotations.", "date": "2025-11-30", "categories": []}
{"id": "2512.01410", "title": "DyFuLM: An Advanced Multimodal Framework for Sentiment Analysis", "abstract": "Understanding sentiment in complex textual expressions remains a fundamental challenge in affective computing. To address this, we propose a Dynamic Fusion Learning Model (DyFuLM), a multimodal framework designed to capture both hierarchical semantic representations and fine-grained emotional nuances. DyFuLM introduces two key moodules: a Hierarchical Dynamic Fusion module that adaptively integrates multi-level features, and a Gated Feature Aggregation module that regulates cross-layer information ffow to achieve balanced representation learning. Comprehensive experiments on multi-task sentiment datasets demonstrate that DyFuLM achieves 82.64% coarse-grained and 68.48% fine-grained accuracy, yielding the lowest regression errors (MAE = 0.0674, MSE = 0.0082) and the highest R^2 coefficient of determination (R^2= 0.6903). Furthermore, the ablation study validates the effectiveness of each module in DyFuLM. When all modules are removed, the accuracy drops by 0.91% for coarse-grained and 0.68% for fine-grained tasks. Keeping only the gated fusion module causes decreases of 0.75% and 0.55%, while removing the dynamic loss mechanism results in drops of 0.78% and 0.26% for coarse-grained and fine-grained sentiment classification, respectively. These results demonstrate that each module contributes significantly to feature interaction and task balance. Overall, the experimental findings further validate that DyFuLM enhances sentiment representation and overall performance through effective hierarchical feature fusion.", "date": "2025-12-01", "categories": []}
{"id": "2512.01442", "title": "PSA-MF: Personality-Sentiment Aligned Multi-Level Fusion for Multimodal Sentiment Analysis", "abstract": "Multimodal sentiment analysis (MSA) is a research field that recognizes human sentiments by combining textual, visual, and audio modalities. The main challenge lies in integrating sentiment-related information from different modalities, which typically arises during the unimodal feature extraction phase and the multimodal feature fusion phase. Existing methods extract only shallow information from unimodal features during the extraction phase, neglecting sentimental differences across different personalities. During the fusion phase, they directly merge the feature information from each modality without considering differences at the feature level. This ultimately affects the model's recognition performance. To address this problem, we propose a personality-sentiment aligned multi-level fusion framework. We introduce personality traits during the feature extraction phase and propose a novel personality-sentiment alignment method to obtain personalized sentiment embeddings from the textual modality for the first time. In the fusion phase, we introduce a novel multi-level fusion method. This method gradually integrates sentimental information from textual, visual, and audio modalities through multimodal pre-fusion and a multi-level enhanced fusion strategy. Our method has been evaluated through multiple experiments on two commonly used datasets, achieving state-of-the-art results.", "date": "2025-12-01", "categories": []}
{"id": "2506.19079", "title": "Reading Smiles: Proxy Bias in Foundation Models for Facial Emotion Recognition", "abstract": "Foundation Models (FMs) are rapidly transforming Affective Computing (AC), with Vision Language Models (VLMs) now capable of recognising emotions in zero shot settings. This paper probes a critical but underexplored question: what visual cues do these models rely on to infer affect, and are these cues psychologically grounded or superficially learnt? We benchmark varying scale VLMs on a teeth annotated subset of AffectNet dataset and find consistent performance shifts depending on the presence of visible teeth. Through structured introspection of, the best-performing model, i.e., GPT-4o, we show that facial attributes like eyebrow position drive much of its affective reasoning, revealing a high degree of internal consistency in its valence-arousal predictions. These patterns highlight the emergent nature of FMs behaviour, but also reveal risks: shortcut learning, bias, and fairness issues especially in sensitive domains like mental health and education.", "date": "2025-06-23", "categories": []}
{"id": "2511.21909", "title": "A Customer Journey in the Land of Oz: Leveraging the Wizard of Oz Technique to Model Emotions in Customer Service Interactions", "abstract": "Emotion-aware customer service needs in-domain conversational data, rich annotations, and predictive capabilities, but existing resources for emotion recognition are often out-of-domain, narrowly labeled, and focused on post-hoc detection. To address this, we conducted a controlled Wizard of Oz (WOZ) experiment to elicit interactions with targeted affective trajectories. The resulting corpus, EmoWOZ-CS, contains 2,148 bilingual (Dutch-English) written dialogues from 179 participants across commercial aviation, e-commerce, online travel agencies, and telecommunication scenarios. Our contributions are threefold: (1) Evaluate WOZ-based operator-steered valence trajectories as a design for emotion research; (2) Quantify human annotation performance and variation, including divergences between self-reports and third-party judgments; (3) Benchmark detection and forward-looking emotion inference in real-time support. Findings show neutral dominates participant messages; desire and gratitude are the most frequent non-neutral emotions. Agreement is moderate for multilabel emotions and valence, lower for arousal and dominance; self-reports diverge notably from third-party labels, aligning most for neutral, gratitude, and anger. Objective strategies often elicit neutrality or gratitude, while suboptimal strategies increase anger, annoyance, disappointment, desire, and confusion. Some affective strategies (cheerfulness, gratitude) foster positive reciprocity, whereas others (apology, empathy) can also leave desire, anger, or annoyance. Temporal analysis confirms successful conversation-level steering toward prescribed trajectories, most distinctly for negative targets; positive and neutral targets yield similar final valence distributions. Benchmarks highlight the difficulty of forward-looking emotion inference from prior turns, underscoring the complexity of proactive emotion-aware support.", "date": "2025-11-26", "categories": []}
{"id": "2511.22188", "title": "ARPGNet: Appearance- and Relation-aware Parallel Graph Attention Fusion Network for Facial Expression Recognition", "abstract": "The key to facial expression recognition is to learn discriminative spatial-temporal representations that embed facial expression dynamics. Previous studies predominantly rely on pre-trained Convolutional Neural Networks (CNNs) to learn facial appearance representations, overlooking the relationships between facial regions. To address this issue, this paper presents an Appearance- and Relation-aware Parallel Graph attention fusion Network (ARPGNet) to learn mutually enhanced spatial-temporal representations of appearance and relation information. Specifically, we construct a facial region relation graph and leverage the graph attention mechanism to model the relationships between facial regions. The resulting relational representation sequences, along with CNN-based appearance representation sequences, are then fed into a parallel graph attention fusion module for mutual interaction and enhancement. This module simultaneously explores the complementarity between different representation sequences and the temporal dynamics within each sequence. Experimental results on three facial expression recognition datasets demonstrate that the proposed ARPGNet outperforms or is comparable to state-of-the-art methods.", "date": "2025-11-27", "categories": []}
{"id": "2511.23115", "title": "Analyzing Image Beyond Visual Aspect: Image Emotion Classification via Multiple-Affective Captioning", "abstract": "Image emotion classification (IEC) is a longstanding research field that has received increasing attention with the rapid progress of deep learning. Although recent advances have leveraged the knowledge encoded in pre-trained visual models, their effectiveness is constrained by the \"affective gap\" , limits the applicability of pre-training knowledge for IEC tasks. It has been demonstrated in psychology that language exhibits high variability, encompasses diverse and abundant information, and can effectively eliminate the \"affective gap\". Inspired by this, we propose a novel Affective Captioning for Image Emotion Classification (ACIEC) to classify image emotion based on pure texts, which effectively capture the affective information in the image. In our method, a hierarchical multi-level contrastive loss is designed for detecting emotional concepts from images, while an emotional attribute chain-of-thought reasoning is proposed to generate affective sentences. Then, a pre-trained language model is leveraged to synthesize emotional concepts and affective sentences to conduct IEC. Additionally, a contrastive loss based on semantic similarity sampling is designed to solve the problem of large intra-class differences and small inter-class differences in affective datasets. Moreover, we also take the images with embedded texts into consideration, which were ignored by previous studies. Extensive experiments illustrate that our method can effectively bridge the affective gap and achieve superior results on multiple benchmarks.", "date": "2025-11-28", "categories": []}
{"id": "2511.19982", "title": "EmoFeedback$^2$: Reinforcement of Continuous Emotional Image Generation via LVLM-based Reward and Textual Feedback", "abstract": "Continuous emotional image generation (C-EICG) is emerging rapidly due to its ability to produce images aligned with both user descriptions and continuous emotional values. However, existing approaches lack emotional feedback from generated images, limiting the control of emotional continuity. Additionally, their simple alignment between emotions and naively generated texts fails to adaptively adjust emotional prompts according to image content, leading to insufficient emotional fidelity. To address these concerns, we propose a novel generation-understanding-feedback reinforcement paradigm (EmoFeedback$^2$) for C-EICG, which exploits the reasoning capability of the fine-tuned large vision-language model (LVLM) to provide reward and textual feedback for generating high-quality images with continuous emotions. Specifically, we introduce an emotion-aware reward feedback strategy, where the LVLM evaluates the emotional values of generated images and computes the reward against target emotions, guiding the reinforcement fine-tuning of the generative model and enhancing the emotional continuity of images. Furthermore, we design a self-promotion textual feedback framework, in which the LVLM iteratively analyzes the emotional content of generated images and adaptively produces refinement suggestions for the next-round prompt, improving the emotional fidelity with fine-grained content. Extensive experimental results demonstrate that our approach effectively generates high-quality images with the desired emotions, outperforming existing state-of-the-art methods in our custom dataset. The code and dataset will be released soon.", "date": "2025-11-25", "categories": []}
{"id": "2511.21051", "title": "MUSE: Manipulating Unified Framework for Synthesizing Emotions in Images via Test-Time Optimization", "abstract": "Images evoke emotions that profoundly influence perception, often prioritized over content. Current Image Emotional Synthesis (IES) approaches artificially separate generation and editing tasks, creating inefficiencies and limiting applications where these tasks naturally intertwine, such as therapeutic interventions or storytelling. In this work, we introduce MUSE, the first unified framework capable of both emotional generation and editing. By adopting a strategy conceptually aligned with Test-Time Scaling (TTS) that widely used in both LLM and diffusion model communities, it avoids the requirement for additional updating diffusion model and specialized emotional synthesis datasets. More specifically, MUSE addresses three key questions in emotional synthesis: (1) HOW to stably guide synthesis by leveraging an off-the-shelf emotion classifier with gradient-based optimization of emotional tokens; (2) WHEN to introduce emotional guidance by identifying the optimal timing using semantic similarity as a supervisory signal; and (3) WHICH emotion to guide synthesis through a multi-emotion loss that reduces interference from inherent and similar emotions. Experimental results show that MUSE performs favorably against all methods for both generation and editing, improving emotional accuracy and semantic diversity while maintaining an optimal balance between desired content, adherence to text prompts, and realistic emotional expression. It establishes a new paradigm for emotion synthesis.", "date": "2025-11-26", "categories": []}
{"id": "2511.17585", "title": "PaSE: Prototype-aligned Calibration and Shapley-based Equilibrium for Multimodal Sentiment Analysis", "abstract": "Multimodal Sentiment Analysis (MSA) seeks to understand human emotions by integrating textual, acoustic, and visual signals. Although multimodal fusion is designed to leverage cross-modal complementarity, real-world scenarios often exhibit modality competition: dominant modalities tend to overshadow weaker ones, leading to suboptimal performance. In this paper, we propose PaSE, a novel Prototype-aligned Calibration and Shapley-optimized Equilibrium framework, which enhances collaboration while explicitly mitigating modality competition. PaSE first applies Prototype-guided Calibration Learning (PCL) to refine unimodal representations and align them through an Entropic Optimal Transport mechanism that ensures semantic consistency. To further stabilize optimization, we introduce a Dual-Phase Optimization strategy. A prototype-gated fusion module is first used to extract shared representations, followed by Shapley-based Gradient Modulation (SGM), which adaptively adjusts gradients according to the contribution of each modality. Extensive experiments on IEMOCAP, MOSI, and MOSEI confirm that PaSE achieves the superior performance and effectively alleviates modality competition.", "date": "2025-11-16", "categories": []}
{"id": "2511.19816", "title": "Breaking Bad: Norms for Valence, Arousal, and Dominance for over 10k English Multiword Expressions", "abstract": "Factor analysis studies have shown that the primary dimensions of word meaning are Valence (V), Arousal (A), and Dominance (D). Existing lexicons such as the NRC VAD Lexicon, published in 2018, include VAD association ratings for words. Here, we present a complement to it, which has human ratings of valence, arousal, and dominance for 10k English Multiword Expressions (MWEs) and their constituent words. We also increase the coverage of unigrams, especially words that have become more common since 2018. In all, the new NRC VAD Lexicon v2 now has entries for 10k MWEs and 25k words, in addition to the entries in v1. We show that the associations are highly reliable. We use the lexicon to examine emotional characteristics of MWEs, including: 1. The degree to which MWEs (idioms, noun compounds, and verb particle constructions) exhibit strong emotionality; 2. The degree of emotional compositionality in MWEs. The lexicon enables a wide variety of research in NLP, Psychology, Public Health, Digital Humanities, and Social Sciences. The NRC VAD Lexicon v2 is freely available through the project webpage: http://saifmohammad.com/WebPages/nrc-vad.html", "date": "2025-11-25", "categories": []}
{"id": "2511.20106", "title": "EM2LDL: A Multilingual Speech Corpus for Mixed Emotion Recognition through Label Distribution Learning", "abstract": "This study introduces EM2LDL, a novel multilingual speech corpus designed to advance mixed emotion recognition through label distribution learning. Addressing the limitations of predominantly monolingual and single-label emotion corpora \\textcolor{black}{that restrict linguistic diversity, are unable to model mixed emotions, and lack ecological validity}, EM2LDL comprises expressive utterances in English, Mandarin, and Cantonese, capturing the intra-utterance code-switching prevalent in multilingual regions like Hong Kong and Macao. The corpus integrates spontaneous emotional expressions from online platforms, annotated with fine-grained emotion distributions across 32 categories. Experimental baselines using self-supervised learning models demonstrate robust performance in speaker-independent gender-, age-, and personality-based evaluations, with HuBERT-large-EN achieving optimal results. By incorporating linguistic diversity and ecological validity, EM2LDL enables the exploration of complex emotional dynamics in multilingual settings. This work provides a versatile testbed for developing adaptive, empathetic systems for applications in affective computing, including mental health monitoring and cross-cultural communication. The dataset, annotations, and baseline codes are publicly available at https://github.com/xingfengli/EM2LDL.", "date": "2025-11-25", "categories": []}
{"id": "2411.06160", "title": "Expansion Quantization Network: An Efficient Micro-emotion Annotation and Detection Framework", "abstract": "Text emotion detection constitutes a crucial foundation for advancing artificial intelligence from basic comprehension to the exploration of emotional reasoning. Most existing emotion detection datasets rely on manual annotations, which are associated with high costs, substantial subjectivity, and severe label imbalances. This is particularly evident in the inadequate annotation of micro-emotions and the absence of emotional intensity representation, which fail to capture the rich emotions embedded in sentences and adversely affect the quality of downstream task completion. By proposing an all-labels and training-set label regression method, we map label values to energy intensity levels, thereby fully leveraging the learning capabilities of machine models and the interdependencies among labels to uncover multiple emotions within samples. This led to the establishment of the Emotion Quantization Network (EQN) framework for micro-emotion detection and annotation. Using five commonly employed sentiment datasets, we conducted comparative experiments with various models, validating the broad applicability of our framework within NLP machine learning models. Based on the EQN framework, emotion detection and annotation are conducted on the GoEmotions dataset. A comprehensive comparison with the results from Google literature demonstrates that the EQN framework possesses a high capability for automatic detection and annotation of micro-emotions. The EQN framework is the first to achieve automatic micro-emotion annotation with energy-level scores, providing strong support for further emotion detection analysis and the quantitative research of emotion computing.", "date": "2024-11-09", "categories": []}
{"id": "2508.03520", "title": "UPLME: Uncertainty-Aware Probabilistic Language Modelling for Robust Empathy Regression", "abstract": "Noisy self-reported empathy scores challenge supervised learning for empathy regression. While many algorithms have been proposed for learning with noisy labels in textual classification problems, the regression counterpart is relatively under-explored. We propose UPLME, an uncertainty-aware probabilistic language modelling framework to capture label noise in empathy regression tasks. One of the novelties in UPLME is a probabilistic language model that predicts both empathy scores and heteroscedastic uncertainty, and is trained using Bayesian concepts with variational model ensembling. We further introduce two novel loss components: one penalises degenerate Uncertainty Quantification (UQ), and another enforces similarity between the input pairs on which empathy is being predicted. UPLME achieves state-of-the-art performance (Pearson Correlation Coefficient: $0.558\\rightarrow0.580$ and $0.629\\rightarrow0.634$) in terms of the performance reported in the literature on two public benchmarks with label noise. Through synthetic label noise injection, we demonstrate that UPLME is effective in distinguishing between noisy and clean samples based on the predicted uncertainty. UPLME further outperform (Calibration error: $0.571\\rightarrow0.376$) a recent variational model ensembling-based UQ method designed for regression problems. Code is publicly available at https://github.com/hasan-rakibul/UPLME.", "date": "2025-08-05", "categories": []}
{"id": "2511.19063", "title": "Logic of Montage", "abstract": "In expressing emotions, as an expression form separate from natural language, we propose an alternative form that complements natural language, acting as a proxy or window for emotional states. First, we set up an expression form \"Effect of Contradictory Structure.\" \"Effect of Contradictory Structure\" is not static but dynamic. Effect in \"Effect of Contradictory Structure\" is unpleasant or pleasant, and the orientation to avoid that unpleasantness is considered pseudo-expression of will. Second, \"Effect of Contradictory Structure\" can be overlapped with each other. This overlapping operation is called \"montage.\" A broader \"Structure\" that includes related \"Effect of Contradictory Structure\" and \"Effect of Structure\" are set up. Montage produces \"Effect of Structure\". In montage, it is necessary to set something like \"strength,\" so we adopted Deleuze and Deleuze/Guattari's word \"intensity\" and set it as an element of our model. We set up a general theoretical framework - Word Import Between Systems (Models) and justified the import of \"intensity\" through Austin's use of the word \"force.\" \"Effect of Structure\" process is demonstrated using the example of proceeding to the next level of education.", "date": "2025-11-24", "categories": []}
{"id": "2511.15675", "title": "MF-GCN: A Multi-Frequency Graph Convolutional Network for Tri-Modal Depression Detection Using Eye-Tracking, Facial, and Acoustic Features", "abstract": "Depression is a prevalent global mental health disorder, characterised by persistent low mood and anhedonia. However, it remains underdiagnosed because current diagnostic methods depend heavily on subjective clinical assessments. To enable objective detection, we introduce a gold standard dataset of 103 clinically assessed participants collected through a tripartite data approach which uniquely integrated eye tracking data with audio and video to give a comprehensive representation of depressive symptoms. Eye tracking data quantifies the attentional bias towards negative stimuli that is frequently observed in depressed groups. Audio and video data capture the affective flattening and psychomotor retardation characteristic of depression. Statistical validation confirmed their significant discriminative power in distinguishing depressed from non depressed groups. We address a critical limitation of existing graph-based models that focus on low-frequency information and propose a Multi-Frequency Graph Convolutional Network (MF-GCN). This framework consists of a novel Multi-Frequency Filter Bank Module (MFFBM), which can leverage both low and high frequency signals. Extensive evaluation against traditional machine learning algorithms and deep learning frameworks demonstrates that MF-GCN consistently outperforms baselines. In binary classification, the model achieved a sensitivity of 0.96 and F2 score of 0.94. For the 3 class classification task, the proposed method achieved a sensitivity of 0.79 and specificity of 0.87 and siginificantly suprassed other models. To validate generalizability, the model was also evaluated on the Chinese Multimodal Depression Corpus (CMDC) dataset and achieved a sensitivity of 0.95 and F2 score of 0.96. These results confirm that our trimodal, multi frequency framework effectively captures cross modal interaction for accurate depression detection.", "date": "2025-11-19", "categories": []}
{"id": "2511.17103", "title": "Bridging Visual Affective Gap: Borrowing Textual Knowledge by Learning from Noisy Image-Text Pairs", "abstract": "Visual emotion recognition (VER) is a longstanding field that has garnered increasing attention with the advancement of deep neural networks. Although recent studies have achieved notable improvements by leveraging the knowledge embedded within pre-trained visual models, the lack of direct association between factual-level features and emotional categories, called the \"affective gap\", limits the applicability of pre-training knowledge for VER tasks. On the contrary, the explicit emotional expression and high information density in textual modality eliminate the \"affective gap\". Therefore, we propose borrowing the knowledge from the pre-trained textual model to enhance the emotional perception of pre-trained visual models. We focus on the factual and emotional connections between images and texts in noisy social media data, and propose Partitioned Adaptive Contrastive Learning (PACL) to leverage these connections. Specifically, we manage to separate different types of samples and devise distinct contrastive learning strategies for each type. By dynamically constructing negative and positive pairs, we fully exploit the potential of noisy samples. Through comprehensive experiments, we demonstrate that bridging the \"affective gap\" significantly improves the performance of various pre-trained visual models in downstream emotion-related tasks. Our code is released on https://github.com/wdqqdw/PACL.", "date": "2025-11-21", "categories": []}
{"id": "2501.01094", "title": "MMVA: Multimodal Matching Based on Valence and Arousal across Images, Music, and Musical Captions", "abstract": "We introduce Multimodal Matching based on Valence and Arousal (MMVA), a tri-modal encoder framework designed to capture emotional content across images, music, and musical captions. To support this framework, we expand the Image-Music-Emotion-Matching-Net (IMEMNet) dataset, creating IMEMNet-C which includes 24,756 images and 25,944 music clips with corresponding musical captions. We employ multimodal matching scores based on the continuous valence (emotional positivity) and arousal (emotional intensity) values. This continuous matching score allows for random sampling of image-music pairs during training by computing similarity scores from the valence-arousal values across different modalities. Consequently, the proposed approach achieves state-of-the-art performance in valence-arousal prediction tasks. Furthermore, the framework demonstrates its efficacy in various zeroshot tasks, highlighting the potential of valence and arousal predictions in downstream applications.", "date": "2025-01-02", "categories": []}
{"id": "2507.15205", "title": "Long-Short Distance Graph Neural Networks and Improved Curriculum Learning for Emotion Recognition in Conversation", "abstract": "Emotion Recognition in Conversation (ERC) is a practical and challenging task. This paper proposes a novel multimodal approach, the Long-Short Distance Graph Neural Network (LSDGNN). Based on the Directed Acyclic Graph (DAG), it constructs a long-distance graph neural network and a short-distance graph neural network to obtain multimodal features of distant and nearby utterances, respectively. To ensure that long- and short-distance features are as distinct as possible in representation while enabling mutual influence between the two modules, we employ a Differential Regularizer and incorporate a BiAffine Module to facilitate feature interaction. In addition, we propose an Improved Curriculum Learning (ICL) to address the challenge of data imbalance. By computing the similarity between different emotions to emphasize the shifts in similar emotions, we design a \"weighted emotional shift\" metric and develop a difficulty measurer, enabling a training process that prioritizes learning easy samples before harder ones. Experimental results on the IEMOCAP and MELD datasets demonstrate that our model outperforms existing benchmarks.", "date": "2025-07-21", "categories": []}
{"id": "2511.15902", "title": "EEG Emotion Recognition Through Deep Learning", "abstract": "An advanced emotion classification model was developed using a CNN-Transformer architecture for emotion recognition from EEG brain wave signals, effectively distinguishing among three emotional states, positive, neutral and negative. The model achieved a testing accuracy of 91%, outperforming traditional models such as SVM, DNN, and Logistic Regression. Training was conducted on a custom dataset created by merging data from SEED, SEED-FRA, and SEED-GER repositories, comprising 1,455 samples with EEG recordings labeled according to emotional states. The combined dataset represents one of the largest and most culturally diverse collections available. Additionally, the model allows for the reduction of the requirements of the EEG apparatus, by leveraging only 5 electrodes of the 62. This reduction demonstrates the feasibility of deploying a more affordable consumer-grade EEG headset, thereby enabling accessible, at-home use, while also requiring less computational power. This advancement sets the groundwork for future exploration into mood changes induced by media content consumption, an area that remains underresearched. Integration into medical, wellness, and home-health platforms could enable continuous, passive emotional monitoring, particularly beneficial in clinical or caregiving settings where traditional behavioral cues, such as facial expressions or vocal tone, are diminished, restricted, or difficult to interpret, thus potentially transforming mental health diagnostics and interventions...", "date": "2025-11-19", "categories": []}
{"id": "2511.14969", "title": "Quality-Controlled Multimodal Emotion Recognition in Conversations with Identity-Based Transfer Learning and MAMBA Fusion", "abstract": "This paper addresses data quality issues in multimodal emotion recognition in conversation (MERC) through systematic quality control and multi-stage transfer learning. We implement a quality control pipeline for MELD and IEMOCAP datasets that validates speaker identity, audio-text alignment, and face detection. We leverage transfer learning from speaker and face recognition, assuming that identity-discriminative embeddings capture not only stable acoustic and Facial traits but also person-specific patterns of emotional expression. We employ RecoMadeEasy(R) engines for extracting 512-dimensional speaker and face embeddings, fine-tune MPNet-v2 for emotion-aware text representations, and adapt these features through emotion-specific MLPs trained on unimodal datasets. MAMBA-based trimodal fusion achieves 64.8% accuracy on MELD and 74.3% on IEMOCAP. These results show that combining identity-based audio and visual embeddings with emotion-tuned text representations on a quality-controlled subset of data yields consistent competitive performance for multimodal emotion recognition in conversation and provides a basis for further improvement on challenging, low-frequency emotion classes.", "date": "2025-11-18", "categories": []}
{"id": "2511.15085", "title": "TiCAL:Typicality-Based Consistency-Aware Learning for Multimodal Emotion Recognition", "abstract": "Multimodal Emotion Recognition (MER) aims to accurately identify human emotional states by integrating heterogeneous modalities such as visual, auditory, and textual data. Existing approaches predominantly rely on unified emotion labels to supervise model training, often overlooking a critical challenge: inter-modal emotion conflicts, wherein different modalities within the same sample may express divergent emotional tendencies. In this work, we address this overlooked issue by proposing a novel framework, Typicality-based Consistent-aware Multimodal Emotion Recognition (TiCAL), inspired by the stage-wise nature of human emotion perception. TiCAL dynamically assesses the consistency of each training sample by leveraging pseudo unimodal emotion labels alongside a typicality estimation. To further enhance emotion representation, we embed features in a hyperbolic space, enabling the capture of fine-grained distinctions among emotional categories. By incorporating consistency estimates into the learning process, our method improves model performance, particularly on samples exhibiting high modality inconsistency. Extensive experiments on benchmark datasets, e.g, CMU-MOSEI and MER2023, validate the effectiveness of TiCAL in mitigating inter-modal emotional conflicts and enhancing overall recognition accuracy, e.g., with about 2.6% improvements over the state-of-the-art DMD.", "date": "2025-11-19", "categories": []}
{"id": "2511.15138", "title": "Cross-Modal Consistency-Guided Active Learning for Affective BCI Systems", "abstract": "Deep learning models perform best with abundant, high-quality labels, yet such conditions are rarely achievable in EEG-based emotion recognition. Electroencephalogram (EEG) signals are easily corrupted by artifacts and individual variability, while emotional labels often stem from subjective and inconsistent reports-making robust affective decoding particularly difficult. We propose an uncertainty-aware active learning framework that enhances robustness to label noise by jointly leveraging model uncertainty and cross-modal consistency. Instead of relying solely on EEG-based uncertainty estimates, the method evaluates cross-modal alignment to determine whether uncertainty originates from cognitive ambiguity or sensor noise. A representation alignment module embeds EEG and face features into a shared latent space, enforcing semantic coherence between modalities. Residual discrepancies are treated as noise-induced inconsistencies, and these samples are selectively queried for oracle feedback during active learning. This feedback-driven process guides the network toward reliable, informative samples and reduces the impact of noisy labels. Experiments on the ASCERTAIN dataset examine the efficiency and robustness of ours, highlighting its potential as a data-efficient and noise-tolerant approach for EEG-based affective decoding in brain-computer interface systems.", "date": "2025-11-19", "categories": []}
{"id": "2502.07843", "title": "Emotional EEG Classification using Upscaled Connectivity Matrices", "abstract": "In recent studies of emotional EEG classification, connectivity matrices have been successfully employed as input to convolutional neural networks (CNNs), which can effectively consider inter-regional interaction patterns in EEG. However, we find that such an approach has a limitation that important patterns in connectivity matrices may be lost during the convolutional operations in CNNs. To resolve this issue, we propose and validate an idea to upscale the connectivity matrices to strengthen the local patterns. Experimental results demonstrate that this simple idea can significantly enhance the classification performance.", "date": "2025-02-11", "categories": []}
{"id": "2510.15471", "title": "A Novel Combined Optical Flow Approach for Comprehensive Micro-Expression Recognition", "abstract": "Facial micro-expressions are brief, involuntary facial movements that reveal hidden emotions. Most Micro-Expression Recognition (MER) methods that rely on optical flow typically focus on the onset-to-apex phase, neglecting the apex-to-offset phase, which holds key temporal dynamics. This study introduces a Combined Optical Flow (COF), integrating both phases to enhance feature representation. COF provides a more comprehensive motion analysis, improving MER performance. Experimental results on CASMEII and SAMM datasets show that COF outperforms single optical flow-based methods, demonstrating its effectiveness in capturing micro-expression dynamics.", "date": "2025-10-17", "categories": []}
{"id": "2511.11691", "title": "Beyond saliency: enhancing explanation of speech emotion recognition with expert-referenced acoustic cues", "abstract": "Explainable AI (XAI) for Speech Emotion Recognition (SER) is critical for building transparent, trustworthy models. Current saliency-based methods, adapted from vision, highlight spectrogram regions but fail to show whether these regions correspond to meaningful acoustic markers of emotion, limiting faithfulness and interpretability. We propose a framework that overcomes these limitations by quantifying the magnitudes of cues within salient regions. This clarifies \"what\" is highlighted and connects it to \"why\" it matters, linking saliency to expert-referenced acoustic cues of speech emotions. Experiments on benchmark SER datasets show that our approach improves explanation quality by explicitly linking salient regions to theory-driven speech emotions expert-referenced acoustics. Compared to standard saliency methods, it provides more understandable and plausible explanations of SER models, offering a foundational step towards trustworthy speech-based affective computing.", "date": "2025-11-12", "categories": []}
{"id": "2511.13530", "title": "Towards Affect-Adaptive Human-Robot Interaction: A Protocol for Multimodal Dataset Collection on Social Anxiety", "abstract": "Social anxiety is a prevalent condition that affects interpersonal interactions and social functioning. Recent advances in artificial intelligence and social robotics offer new opportunities to examine social anxiety in the human-robot interaction context. Accurate detection of affective states and behaviours associated with social anxiety requires multimodal datasets, where each signal modality provides complementary insights into its manifestations. However, such datasets remain scarce, limiting progress in both research and applications. To address this, this paper presents a protocol for multimodal dataset collection designed to reflect social anxiety in a human-robot interaction context. The dataset will consist of synchronised audio, video, and physiological recordings acquired from at least 70 participants, grouped according to their level of social anxiety, as they engage in approximately 10-minute interactive Wizard-of-Oz role-play scenarios with the Furhat social robot under controlled experimental conditions. In addition to multimodal data, the dataset will be enriched with contextual data providing deeper insight into individual variability in social anxiety responses. This work can contribute to research on affect-adaptive human-robot interaction by providing support for robust multimodal detection of social anxiety.", "date": "2025-11-17", "categories": []}
{"id": "2511.05863", "title": "EMOD: A Unified EEG Emotion Representation Framework Leveraging V-A Guided Contrastive Learning", "abstract": "Emotion recognition from EEG signals is essential for affective computing and has been widely explored using deep learning. While recent deep learning approaches have achieved strong performance on single EEG emotion datasets, their generalization across datasets remains limited due to the heterogeneity in annotation schemes and data formats. Existing models typically require dataset-specific architectures tailored to input structure and lack semantic alignment across diverse emotion labels. To address these challenges, we propose EMOD: A Unified EEG Emotion Representation Framework Leveraging Valence-Arousal (V-A) Guided Contrastive Learning. EMOD learns transferable and emotion-aware representations from heterogeneous datasets by bridging both semantic and structural gaps. Specifically, we project discrete and continuous emotion labels into a unified V-A space and formulate a soft-weighted supervised contrastive loss that encourages emotionally similar samples to cluster in the latent space. To accommodate variable EEG formats, EMOD employs a flexible backbone comprising a Triple-Domain Encoder followed by a Spatial-Temporal Transformer, enabling robust extraction and integration of temporal, spectral, and spatial features. We pretrain EMOD on 8 public EEG datasets and evaluate its performance on three benchmark datasets. Experimental results show that EMOD achieves the state-of-the-art performance, demonstrating strong adaptability and generalization across diverse EEG-based emotion recognition scenarios.", "date": "2025-11-08", "categories": []}
{"id": "2511.10846", "title": "Reinforcing Stereotypes of Anger: Emotion AI on African American Vernacular English", "abstract": "Automated emotion detection is widely used in applications ranging from well-being monitoring to high-stakes domains like mental health and hiring. However, models often rely on annotations that reflect dominant cultural norms, limiting model ability to recognize emotional expression in dialects often excluded from training data distributions, such as African American Vernacular English (AAVE). This study examines emotion recognition model performance on AAVE compared to General American English (GAE). We analyze 2.7 million tweets geo-tagged within Los Angeles. Texts are scored for strength of AAVE using computational approximations of dialect features. Annotations of emotion presence and intensity are collected on a dataset of 875 tweets with both high and low AAVE densities. To assess model accuracy on a task as subjective as emotion perception, we calculate community-informed \"silver\" labels where AAVE-dense tweets are labeled by African American, AAVE-fluent (ingroup) annotators. On our labeled sample, GPT and BERT-based models exhibit false positive prediction rates of anger on AAVE more than double than on GAE. SpanEmo, a popular text-based emotion model, increases false positive rates of anger from 25 percent on GAE to 60 percent on AAVE. Additionally, a series of linear regressions reveals that models and non-ingroup annotations are significantly more correlated with profanity-based AAVE features than ingroup annotations. Linking Census tract demographics, we observe that neighborhoods with higher proportions of African American residents are associated with higher predictions of anger (Pearson's correlation r = 0.27) and lower joy (r = -0.10). These results find an emergent safety issue of emotion AI reinforcing racial stereotypes through biased emotion classification. We emphasize the need for culturally and dialect-informed affective computing systems.", "date": "2025-11-13", "categories": []}
{"id": "2511.10892", "title": "MCN-CL: Multimodal Cross-Attention Network and Contrastive Learning for Multimodal Emotion Recognition", "abstract": "Multimodal emotion recognition plays a key role in many domains, including mental health monitoring, educational interaction, and human-computer interaction. However, existing methods often face three major challenges: unbalanced category distribution, the complexity of dynamic facial action unit time modeling, and the difficulty of feature fusion due to modal heterogeneity. With the explosive growth of multimodal data in social media scenarios, the need for building an efficient cross-modal fusion framework for emotion recognition is becoming increasingly urgent. To this end, this paper proposes Multimodal Cross-Attention Network and Contrastive Learning (MCN-CL) for multimodal emotion recognition. It uses a triple query mechanism and hard negative mining strategy to remove feature redundancy while preserving important emotional cues, effectively addressing the issues of modal heterogeneity and category imbalance. Experiment results on the IEMOCAP and MELD datasets show that our proposed method outperforms state-of-the-art approaches, with Weighted F1 scores improving by 3.42% and 5.73%, respectively.", "date": "2025-11-14", "categories": []}
{"id": "2511.10940", "title": "Facial Expression Recognition with YOLOv11 and YOLOv12: A Comparative Study", "abstract": "Facial Expression Recognition remains a challenging task, especially in unconstrained, real-world environments. This study investigates the performance of two lightweight models, YOLOv11n and YOLOv12n, which are the nano variants of the latest official YOLO series, within a unified detection and classification framework for FER. Two benchmark classification datasets, FER2013 and KDEF, are converted into object detection format and model performance is evaluated using mAP 0.5, precision, recall, and confusion matrices. Results show that YOLOv12n achieves the highest overall performance on the clean KDEF dataset with a mAP 0.5 of 95.6, and also outperforms YOLOv11n on the FER2013 dataset in terms of mAP 63.8, reflecting stronger sensitivity to varied expressions. In contrast, YOLOv11n demonstrates higher precision 65.2 on FER2013, indicating fewer false positives and better reliability in noisy, real-world conditions. On FER2013, both models show more confusion between visually similar expressions, while clearer class separation is observed on the cleaner KDEF dataset. These findings underscore the trade-off between sensitivity and precision, illustrating how lightweight YOLO models can effectively balance performance and efficiency. The results demonstrate adaptability across both controlled and real-world conditions, establishing these models as strong candidates for real-time, resource-constrained emotion-aware AI applications.", "date": "2025-11-14", "categories": []}
{"id": "2511.11002", "title": "EmoVid: A Multimodal Emotion Video Dataset for Emotion-Centric Video Understanding and Generation", "abstract": "Emotion plays a pivotal role in video-based expression, but existing video generation systems predominantly focus on low-level visual metrics while neglecting affective dimensions. Although emotion analysis has made progress in the visual domain, the video community lacks dedicated resources to bridge emotion understanding with generative tasks, particularly for stylized and non-realistic contexts. To address this gap, we introduce EmoVid, the first multimodal, emotion-annotated video dataset specifically designed for creative media, which includes cartoon animations, movie clips, and animated stickers. Each video is annotated with emotion labels, visual attributes (brightness, colorfulness, hue), and text captions. Through systematic analysis, we uncover spatial and temporal patterns linking visual features to emotional perceptions across diverse video forms. Building on these insights, we develop an emotion-conditioned video generation technique by fine-tuning the Wan2.1 model. The results show a significant improvement in both quantitative metrics and the visual quality of generated videos for text-to-video and image-to-video tasks. EmoVid establishes a new benchmark for affective video computing. Our work not only offers valuable insights into visual emotion analysis in artistically styled videos, but also provides practical methods for enhancing emotional expression in video generation.", "date": "2025-11-14", "categories": []}
{"id": "2511.11126", "title": "Enhancing Meme Emotion Understanding with Multi-Level Modality Enhancement and Dual-Stage Modal Fusion", "abstract": "With the rapid rise of social media and Internet culture, memes have become a popular medium for expressing emotional tendencies. This has sparked growing interest in Meme Emotion Understanding (MEU), which aims to classify the emotional intent behind memes by leveraging their multimodal contents. While existing efforts have achieved promising results, two major challenges remain: (1) a lack of fine-grained multimodal fusion strategies, and (2) insufficient mining of memes' implicit meanings and background knowledge. To address these challenges, we propose MemoDetector, a novel framework for advancing MEU. First, we introduce a four-step textual enhancement module that utilizes the rich knowledge and reasoning capabilities of Multimodal Large Language Models (MLLMs) to progressively infer and extract implicit and contextual insights from memes. These enhanced texts significantly enrich the original meme contents and provide valuable guidance for downstream classification. Next, we design a dual-stage modal fusion strategy: the first stage performs shallow fusion on raw meme image and text, while the second stage deeply integrates the enhanced visual and textual features. This hierarchical fusion enables the model to better capture nuanced cross-modal emotional cues. Experiments on two datasets, MET-MEME and MOOD, demonstrate that our method consistently outperforms state-of-the-art baselines. Specifically, MemoDetector improves F1 scores by 4.3\\% on MET-MEME and 3.4\\% on MOOD. Further ablation studies and in-depth analyses validate the effectiveness and robustness of our approach, highlighting its strong potential for advancing MEU. Our code is available at https://github.com/singing-cat/MemoDetector.", "date": "2025-11-14", "categories": []}
{"id": "2312.05735", "title": "A Comprehensive Survey on Multi-modal Conversational Emotion Recognition with Deep Learning", "abstract": "Multi-modal conversation emotion recognition (MCER) aims to recognize and track the speaker's emotional state using text, speech, and visual information in the conversation scene. Analyzing and studying MCER issues is significant to affective computing, intelligent recommendations, and human-computer interaction fields. Unlike the traditional single-utterance multi-modal emotion recognition or single-modal conversation emotion recognition, MCER is a more challenging problem that needs to deal with more complex emotional interaction relationships. The critical issue is learning consistency and complementary semantics for multi-modal feature fusion based on emotional interaction relationships. To solve this problem, people have conducted extensive research on MCER based on deep learning technology, but there is still a lack of systematic review of the modeling methods. Therefore, a timely and comprehensive overview of MCER's recent advances in deep learning is of great significance to academia and industry. In this survey, we provide a comprehensive overview of MCER modeling methods and roughly divide MCER methods into four categories, i.e., context-free modeling, sequential context modeling, speaker-differentiated modeling, and speaker-relationship modeling. In addition, we further discuss MCER's publicly available popular datasets, multi-modal feature extraction methods, application areas, existing challenges, and future development directions. We hope that our review can help MCER researchers understand the current research status in emotion recognition, provide some inspiration, and develop more efficient models.", "date": "2023-12-10", "categories": []}
{"id": "2510.04688", "title": "A Study on the Data Distribution Gap in Music Emotion Recognition", "abstract": "Music Emotion Recognition (MER) is a task deeply connected to human perception, relying heavily on subjective annotations collected from contributors. Prior studies tend to focus on specific musical styles rather than incorporating a diverse range of genres, such as rock and classical, within a single framework. In this paper, we address the task of recognizing emotion from audio content by investigating five datasets with dimensional emotion annotations -- EmoMusic, DEAM, PMEmo, WTC, and WCMED -- which span various musical styles. We demonstrate the problem of out-of-distribution generalization in a systematic experiment. By closely looking at multiple data and feature sets, we provide insight into genre-emotion relationships in existing data and examine potential genre dominance and dataset biases in certain feature representations. Based on these experiments, we arrive at a simple yet effective framework that combines embeddings extracted from the Jukebox model with chroma features and demonstrate how, alongside a combination of several diverse training sets, this permits us to train models with substantially improved cross-dataset generalization capabilities.", "date": "2025-10-06", "categories": []}
{"id": "2511.04914", "title": "MERaLiON-SER: Robust Speech Emotion Recognition Model for English and SEA Languages", "abstract": "We present MERaLiON-SER, a robust speech emotion recognition model designed for English and Southeast Asian languages. The model is trained using a hybrid objective combining weighted categorical cross-entropy and Concordance Correlation Coefficient (CCC) losses for joint discrete and dimensional emotion modelling. This dual approach enables the model to capture both the distinct categories of emotion (like happy or angry) and the fine-grained, such as arousal (intensity), valence (positivity/negativity), and dominance (sense of control), leading to a more comprehensive and robust representation of human affect. Extensive evaluations across multilingual Singaporean languages (English, Chinese, Malay, and Tamil ) and other public benchmarks show that MERaLiON-SER consistently surpasses both open-source speech encoders and large Audio-LLMs. These results underscore the importance of specialised speech-only models for accurate paralinguistic understanding and cross-lingual generalisation. Furthermore, the proposed framework provides a foundation for integrating emotion-aware perception into future agentic audio systems, enabling more empathetic and contextually adaptive multimodal reasoning.", "date": "2025-11-07", "categories": []}
{"id": "2511.09003", "title": "Detecting Emotional Dynamic Trajectories: An Evaluation Framework for Emotional Support in Language Models", "abstract": "Emotional support is a core capability in human-AI interaction, with applications including psychological counseling, role play, and companionship. However, existing evaluations of large language models (LLMs) often rely on short, static dialogues and fail to capture the dynamic and long-term nature of emotional support. To overcome this limitation, we shift from snapshot-based evaluation to trajectory-based assessment, adopting a user-centered perspective that evaluates models based on their ability to improve and stabilize user emotional states over time. Our framework constructs a large-scale benchmark consisting of 328 emotional contexts and 1,152 disturbance events, simulating realistic emotional shifts under evolving dialogue scenarios. To encourage psychologically grounded responses, we constrain model outputs using validated emotion regulation strategies such as situation selection and cognitive reappraisal. User emotional trajectories are modeled as a first-order Markov process, and we apply causally-adjusted emotion estimation to obtain unbiased emotional state tracking. Based on this framework, we introduce three trajectory-level metrics: Baseline Emotional Level (BEL), Emotional Trajectory Volatility (ETV), and Emotional Centroid Position (ECP). These metrics collectively capture user emotional dynamics over time and support comprehensive evaluation of long-term emotional support performance of LLMs. Extensive evaluations across a diverse set of LLMs reveal significant disparities in emotional support capabilities and provide actionable insights for model development.", "date": "2025-11-12", "categories": []}
{"id": "2505.13979", "title": "Mixed Signals: Understanding Model Disagreement in Multimodal Empathy Detection", "abstract": "Multimodal models play a key role in empathy detection, but their performance can suffer when modalities provide conflicting cues. To understand these failures, we examine cases where unimodal and multimodal predictions diverge. Using fine-tuned models for text, audio, and video, along with a gated fusion model, we find that such disagreements often reflect underlying ambiguity, as evidenced by annotator uncertainty. Our analysis shows that dominant signals in one modality can mislead fusion when unsupported by others. We also observe that humans, like models, do not consistently benefit from multimodal input. These insights position disagreement as a useful diagnostic signal for identifying challenging examples and improving empathy system robustness.", "date": "2025-05-20", "categories": []}
{"id": "2511.07955", "title": "Speech Emotion Recognition with Phonation Excitation Information and Articulatory Kinematics", "abstract": "Speech emotion recognition (SER) has advanced significantly for the sake of deep-learning methods, while textual information further enhances its performance. However, few studies have focused on the physiological information during speech production, which also encompasses speaker traits, including emotional states. To bridge this gap, we conducted a series of experiments to investigate the potential of the phonation excitation information and articulatory kinematics for SER. Due to the scarcity of training data for this purpose, we introduce a portrayed emotional dataset, STEM-E2VA, which includes audio and physiological data such as electroglottography (EGG) and electromagnetic articulography (EMA). EGG and EMA provide information of phonation excitation and articulatory kinematics, respectively. Additionally, we performed emotion recognition using estimated physiological data derived through inversion methods from speech, instead of collected EGG and EMA, to explore the feasibility of applying such physiological information in real-world SER. Experimental results confirm the effectiveness of incorporating physiological information about speech production for SER and demonstrate its potential for practical use in real-world scenarios.", "date": "2025-11-11", "categories": []}
{"id": "2511.08444", "title": "One Model for All: Universal Pre-training for EEG based Emotion Recognition across Heterogeneous Datasets and Paradigms", "abstract": "EEG-based emotion recognition is hampered by profound dataset heterogeneity (channel/subject variability), hindering generalizable models. Existing approaches struggle to transfer knowledge effectively. We propose 'One Model for All', a universal pre-training framework for EEG analysis across disparate datasets. Our paradigm decouples learning into two stages: (1) Univariate pre-training via self-supervised contrastive learning on individual channels, enabled by a Unified Channel Schema (UCS) that leverages the channel union (e.g., SEED-62ch, DEAP-32ch); (2) Multivariate fine-tuning with a novel 'ART' (Adaptive Resampling Transformer) and 'GAT' (Graph Attention Network) architecture to capture complex spatio-temporal dependencies. Experiments show universal pre-training is an essential stabilizer, preventing collapse on SEED (vs. scratch) and yielding substantial gains on DEAP (+7.65%) and DREAMER (+3.55%). Our framework achieves new SOTA performance on all within-subject benchmarks: SEED (99.27%), DEAP (93.69%), and DREAMER (93.93%). We also show SOTA cross-dataset transfer, achieving 94.08% (intersection) and 93.05% (UCS) on the unseen DREAMER dataset, with the former surpassing the within-domain pre-training benchmark. Ablation studies validate our architecture: the GAT module is critical, yielding a +22.19% gain over GCN on the high-noise DEAP dataset, and its removal causes a catastrophic -16.44% performance drop. This work paves the way for more universal, scalable, and effective pre-trained models for diverse EEG analysis tasks.", "date": "2025-11-11", "categories": []}
{"id": "2511.06231", "title": "Synheart Emotion: Privacy-Preserving On-Device Emotion Recognition from Biosignals", "abstract": "Human-computer interaction increasingly demands systems that recognize not only explicit user inputs but also implicit emotional states. While substantial progress has been made in affective computing, most emotion recognition systems rely on cloud-based inference, introducing privacy vulnerabilities and latency constraints unsuitable for real-time applications. This work presents a comprehensive evaluation of machine learning architectures for on-device emotion recognition from wrist-based photoplethysmography (PPG), systematically comparing different models spanning classical ensemble methods, deep neural networks, and transformers on the WESAD stress detection dataset. Results demonstrate that classical ensemble methods substantially outperform deep learning on small physiological datasets, with ExtraTrees achieving F1 = 0.826 on combined features and F1 = 0.623 on wrist-only features, compared to transformers achieving only F1 = 0.509-0.577. We deploy the wrist-only ExtraTrees model optimized via ONNX conversion, achieving a 4.08 MB footprint, 0.05 ms inference latency, and 152x speedup over the original implementation. Furthermore, ONNX optimization yields a 30.5% average storage reduction and 40.1x inference speedup, highlighting the feasibility of privacy-preserving on-device emotion recognition for real-world wearables.", "date": "2025-11-09", "categories": []}
{"id": "2511.04995", "title": "Enhancing Public Speaking Skills in Engineering Students Through AI", "abstract": "This research-to-practice full paper was inspired by the persistent challenge in effective communication among engineering students. Public speaking is a necessary skill for future engineers as they have to communicate technical knowledge with diverse stakeholders. While universities offer courses or workshops, they are unable to offer sustained and personalized training to students. Providing comprehensive feedback on both verbal and non-verbal aspects of public speaking is time-intensive, making consistent and individualized assessment impractical. This study integrates research on verbal and non-verbal cues in public speaking to develop an AI-driven assessment model for engineering students. Our approach combines speech analysis, computer vision, and sentiment detection into a multi-modal AI system that provides assessment and feedback. The model evaluates (1) verbal communication (pitch, loudness, pacing, intonation), (2) non-verbal communication (facial expressions, gestures, posture), and (3) expressive coherence, a novel integration ensuring alignment between speech and body language. Unlike previous systems that assess these aspects separately, our model fuses multiple modalities to deliver personalized, scalable feedback. Preliminary testing demonstrated that our AI-generated feedback was moderately aligned with expert evaluations. Among the state-of-the-art AI models evaluated, all of which were Large Language Models (LLMs), including Gemini and OpenAI models, Gemini Pro emerged as the best-performing, showing the strongest agreement with human annotators. By eliminating reliance on human evaluators, this AI-driven public speaking trainer enables repeated practice, helping students naturally align their speech with body language and emotion, crucial for impactful and professional communication.", "date": "2025-11-07", "categories": []}
{"id": "2410.21882", "title": "Building Altruistic and Moral AI Agent with Brain-inspired Emotional Empathy Mechanisms", "abstract": "As AI closely interacts with human society, it is crucial to ensure that its behavior is safe, altruistic, and aligned with human ethical and moral values. However, existing research on embedding ethical considerations into AI remains insufficient, and previous external constraints based on principles and rules are inadequate to provide AI with long-term stability and generalization capabilities. Emotional empathy intrinsically motivates altruistic behaviors aimed at alleviating others' negative emotions through emotional sharing and contagion mechanisms. Motivated by this, we draw inspiration from the neural mechanism of human emotional empathy-driven altruistic decision making, and simulate the shared self-other perception-mirroring-empathy neural circuits, to construct a brain-inspired emotional empathy-driven altruistic decision-making model. Here, empathy directly impacts dopamine release to form intrinsic altruistic motivation. The proposed model exhibits consistent altruistic behaviors across three experimental settings: emotional contagion-integrated two-agent altruistic rescue, multi-agent gaming, and robotic emotional empathy interaction scenarios. In-depth analyses validate the positive correlation between empathy levels and altruistic preferences (consistent with psychological behavioral experiment findings), while also demonstrating how interaction partners' empathy levels influence the agent's behavioral patterns. We further test the proposed model's performance and stability in moral dilemmas involving conflicts between self-interest and others' well-being, partially observable environments, and adversarial defense scenarios. This work provides preliminary exploration of human-like empathy-driven altruistic moral decision making, contributing potential perspectives for developing ethically-aligned AI.", "date": "2024-10-29", "categories": []}
{"id": "2511.01892", "title": "Retrieval-Augmented Multimodal Depression Detection", "abstract": "Multimodal deep learning has shown promise in depression detection by integrating text, audio, and video signals. Recent work leverages sentiment analysis to enhance emotional understanding, yet suffers from high computational cost, domain mismatch, and static knowledge limitations. To address these issues, we propose a novel Retrieval-Augmented Generation (RAG) framework. Given a depression-related text, our method retrieves semantically relevant emotional content from a sentiment dataset and uses a Large Language Model (LLM) to generate an Emotion Prompt as an auxiliary modality. This prompt enriches emotional representation and improves interpretability. Experiments on the AVEC 2019 dataset show our approach achieves state-of-the-art performance with CCC of 0.593 and MAE of 3.95, surpassing previous transfer learning and multi-task learning baselines.", "date": "2025-10-29", "categories": []}
{"id": "2511.00344", "title": "Federated Dialogue-Semantic Diffusion for Emotion Recognition under Incomplete Modalities", "abstract": "Multimodal Emotion Recognition in Conversations (MERC) enhances emotional understanding through the fusion of multimodal signals. However, unpredictable modality absence in real-world scenarios significantly degrades the performance of existing methods. Conventional missing-modality recovery approaches, which depend on training with complete multimodal data, often suffer from semantic distortion under extreme data distributions, such as fixed-modality absence. To address this, we propose the Federated Dialogue-guided and Semantic-Consistent Diffusion (FedDISC) framework, pioneering the integration of federated learning into missing-modality recovery. By federated aggregation of modality-specific diffusion models trained on clients and broadcasting them to clients missing corresponding modalities, FedDISC overcomes single-client reliance on modality completeness. Additionally, the DISC-Diffusion module ensures consistency in context, speaker identity, and semantics between recovered and available modalities, using a Dialogue Graph Network to capture conversational dependencies and a Semantic Conditioning Network to enforce semantic alignment. We further introduce a novel Alternating Frozen Aggregation strategy, which cyclically freezes recovery and classifier modules to facilitate collaborative optimization. Extensive experiments on the IEMOCAP, CMUMOSI, and CMUMOSEI datasets demonstrate that FedDISC achieves superior emotion classification performance across diverse missing modality patterns, outperforming existing approaches.", "date": "2025-11-01", "categories": []}
{"id": "2511.00402", "title": "Emotion Detection in Speech Using Lightweight and Transformer-Based Models: A Comparative and Ablation Study", "abstract": "Emotion recognition from speech plays a vital role in the development of empathetic human-computer interaction systems. This paper presents a comparative analysis of lightweight transformer-based models, DistilHuBERT and PaSST, by classifying six core emotions from the CREMA-D dataset. We benchmark their performance against a traditional CNN-LSTM baseline model using MFCC features. DistilHuBERT demonstrates superior accuracy (70.64%) and F1 score (70.36%) while maintaining an exceptionally small model size (0.02 MB), outperforming both PaSST and the baseline. Furthermore, we conducted an ablation study on three variants of the PaSST, Linear, MLP, and Attentive Pooling heads, to understand the effect of classification head architecture on model performance. Our results indicate that PaSST with an MLP head yields the best performance among its variants but still falls short of DistilHuBERT. Among the emotion classes, angry is consistently the most accurately detected, while disgust remains the most challenging. These findings suggest that lightweight transformers like DistilHuBERT offer a compelling solution for real-time speech emotion recognition on edge devices. The code is available at: https://github.com/luckymaduabuchi/Emotion-detection-.", "date": "2025-11-01", "categories": []}
{"id": "2511.01444", "title": "Robust Multimodal Sentiment Analysis via Double Information Bottleneck", "abstract": "Multimodal sentiment analysis has received significant attention across diverse research domains. Despite advancements in algorithm design, existing approaches suffer from two critical limitations: insufficient learning of noise-contaminated unimodal data, leading to corrupted cross-modal interactions, and inadequate fusion of multimodal representations, resulting in discarding discriminative unimodal information while retaining multimodal redundant information. To address these challenges, this paper proposes a Double Information Bottleneck (DIB) strategy to obtain a powerful, unified compact multimodal representation. Implemented within the framework of low-rank Renyi's entropy functional, DIB offers enhanced robustness against diverse noise sources and computational tractability for high-dimensional data, as compared to the conventional Shannon entropy-based methods. The DIB comprises two key modules: 1) learning a sufficient and compressed representation of individual unimodal data by maximizing the task-relevant information and discarding the superfluous information, and 2) ensuring the discriminative ability of multimodal representation through a novel attention bottleneck fusion mechanism. Consequently, DIB yields a multimodal representation that effectively filters out noisy information from unimodal data while capturing inter-modal complementarity. Extensive experiments on CMU-MOSI, CMU-MOSEI, CH-SIMS, and MVSA-Single validate the effectiveness of our method. The model achieves 47.4% accuracy under the Acc-7 metric on CMU-MOSI and 81.63% F1-score on CH-SIMS, outperforming the second-best baseline by 1.19%. Under noise, it shows only 0.36% and 0.29% performance degradation on CMU-MOSI and CMU-MOSEI respectively.", "date": "2025-11-03", "categories": []}
{"id": "2503.16532", "title": "Modelling Emotions in Face-to-Face Setting: The Interplay of Eye-Tracking, Personality, and Temporal Dynamics", "abstract": "Accurate emotion recognition is pivotal for nuanced and engaging human-computer interactions, yet remains difficult to achieve, especially in dynamic, conversation-like settings. In this study, we showcase how integrating eye-tracking data, temporal dynamics, and personality traits can substantially enhance the detection of both perceived and felt emotions. Seventy-three participants viewed short, speech-containing videos from the CREMA-D dataset, while being recorded for eye-tracking signals (pupil size, fixation patterns), Big Five personality assessments, and self-reported emotional states. Our neural network models combined these diverse inputs including stimulus emotion labels for contextual cues and yielded marked performance gains compared to the state-of-the-art. Specifically, perceived valence predictions reached a macro F1-score of 0.76, and models incorporating personality traits and stimulus information demonstrated significant improvements in felt emotion accuracy. These results highlight the benefit of unifying physiological, individual and contextual factors to address the subjectivity and complexity of emotional expression. Beyond validating the role of user-specific data in capturing subtle internal states, our findings inform the design of future affective computing and human-agent systems, paving the way for more adaptive and cross-individual emotional intelligence in real-world interactions.", "date": "2025-03-18", "categories": []}
{"id": "2510.26823", "title": "Cross-Corpus Validation of Speech Emotion Recognition in Urdu using Domain-Knowledge Acoustic Features", "abstract": "Speech Emotion Recognition (SER) is a key affective computing technology that enables emotionally intelligent artificial intelligence. While SER is challenging in general, it is particularly difficult for low-resource languages such as Urdu. This study investigates Urdu SER in a cross-corpus setting, an area that has remained largely unexplored. We employ a cross-corpus evaluation framework across three different Urdu emotional speech datasets to test model generalization. Two standard domain-knowledge based acoustic feature sets, eGeMAPS and ComParE, are used to represent speech signals as feature vectors which are then passed to Logistic Regression and Multilayer Perceptron classifiers. Classification performance is assessed using unweighted average recall (UAR) whilst considering class-label imbalance. Results show that Self-corpus validation often overestimates performance, with UAR exceeding cross-corpus evaluation by up to 13%, underscoring that cross-corpus evaluation offers a more realistic measure of model robustness. Overall, this work emphasizes the importance of cross-corpus validation for Urdu SER and its implications contribute to advancing affective computing research for underrepresented language communities.", "date": "2025-10-28", "categories": []}
{"id": "2409.06154", "title": "Static for Dynamic: Towards a Deeper Understanding of Dynamic Facial Expressions Using Static Expression Data", "abstract": "Dynamic facial expression recognition (DFER) infers emotions from the temporal evolution of expressions, unlike static facial expression recognition (SFER), which relies solely on a single snapshot. This temporal analysis provides richer information and promises greater recognition capability. However, current DFER methods often exhibit unsatisfied performance largely due to fewer training samples compared to SFER. Given the inherent correlation between static and dynamic expressions, we hypothesize that leveraging the abundant SFER data can enhance DFER. To this end, we propose Static-for-Dynamic (S4D), a unified dual-modal learning framework that integrates SFER data as a complementary resource for DFER. Specifically, S4D employs dual-modal self-supervised pre-training on facial images and videos using a shared Vision Transformer (ViT) encoder-decoder architecture, yielding improved spatiotemporal representations. The pre-trained encoder is then fine-tuned on static and dynamic expression datasets in a multi-task learning setup to facilitate emotional information interaction. Unfortunately, vanilla multi-task learning in our study results in negative transfer. To address this, we propose an innovative Mixture of Adapter Experts (MoAE) module that facilitates task-specific knowledge acquisition while effectively extracting shared knowledge from both static and dynamic expression data. Extensive experiments demonstrate that S4D achieves a deeper understanding of DFER, setting new state-of-the-art performance on FERV39K, MAFW, and DFEW benchmarks, with weighted average recall (WAR) of 53.65\\%, 58.44\\%, and 76.68\\%, respectively. Additionally, a systematic correlation analysis between SFER and DFER tasks is presented, which further elucidates the potential benefits of leveraging SFER.", "date": "2024-09-10", "categories": []}
{"id": "2510.25054", "title": "Evaluating Emotion Recognition in Spoken Language Models on Emotionally Incongruent Speech", "abstract": "Advancements in spoken language processing have driven the development of spoken language models (SLMs), designed to achieve universal audio understanding by jointly learning text and audio representations for a wide range of tasks. Although promising results have been achieved, there is growing discussion regarding these models' generalization capabilities and the extent to which they truly integrate audio and text modalities in their internal representations. In this work, we evaluate four SLMs on the task of speech emotion recognition using a dataset of emotionally incongruent speech samples, a condition under which the semantic content of the spoken utterance conveys one emotion while speech expressiveness conveys another. Our results indicate that SLMs rely predominantly on textual semantics rather than speech emotion to perform the task, indicating that text-related representations largely dominate over acoustic representations. We release both the code and the Emotionally Incongruent Synthetic Speech dataset (EMIS) to the community.", "date": "2025-10-29", "categories": []}
{"id": "2510.26304", "title": "Exploring the correlation between the type of music and the emotions evoked: A study using subjective questionnaires and EEG", "abstract": "The subject of this work is to check how different types of music affect human emotions. While listening to music, a subjective survey and brain activity measurements were carried out using an EEG helmet. The aim is to demonstrate the impact of different music genres on emotions. The research involved a diverse group of participants of different gender and musical preferences. This had the effect of capturing a wide range of emotional responses to music. After the experiment, a relationship analysis of the respondents' questionnaires with EEG signals was performed. The analysis revealed connections between emotions and observed brain activity.", "date": "2025-10-30", "categories": []}
{"id": "2510.24720", "title": "Modelling the Interplay of Eye-Tracking Temporal Dynamics and Personality for Emotion Detection in Face-to-Face Settings", "abstract": "Accurate recognition of human emotions is critical for adaptive human-computer interaction, yet remains challenging in dynamic, conversation-like settings. This work presents a personality-aware multimodal framework that integrates eye-tracking sequences, Big Five personality traits, and contextual stimulus cues to predict both perceived and felt emotions. Seventy-three participants viewed speech-containing clips from the CREMA-D dataset while providing eye-tracking signals, personality assessments, and emotion ratings. Our neural models captured temporal gaze dynamics and fused them with trait and stimulus information, yielding consistent gains over SVM and literature baselines. Results show that (i) stimulus cues strongly enhance perceived-emotion predictions (macro F1 up to 0.77), while (ii) personality traits provide the largest improvements for felt emotion recognition (macro F1 up to 0.58). These findings highlight the benefit of combining physiological, trait-level, and contextual information to address the inherent subjectivity of emotion. By distinguishing between perceived and felt responses, our approach advances multimodal affective computing and points toward more personalized and ecologically valid emotion-aware systems.", "date": "2025-09-19", "categories": []}
{"id": "2510.24827", "title": "MCIHN: A Hybrid Network Model Based on Multi-path Cross-modal Interaction for Multimodal Emotion Recognition", "abstract": "Multimodal emotion recognition is crucial for future human-computer interaction. However, accurate emotion recognition still faces significant challenges due to differences between different modalities and the difficulty of characterizing unimodal emotional information. To solve these problems, a hybrid network model based on multipath cross-modal interaction (MCIHN) is proposed. First, adversarial autoencoders (AAE) are constructed separately for each modality. The AAE learns discriminative emotion features and reconstructs the features through a decoder to obtain more discriminative information about the emotion classes. Then, the latent codes from the AAE of different modalities are fed into a predefined Cross-modal Gate Mechanism model (CGMM) to reduce the discrepancy between modalities, establish the emotional relationship between interacting modalities, and generate the interaction features between different modalities. Multimodal fusion using the Feature Fusion module (FFM) for better emotion recognition. Experiments were conducted on publicly available SIMS and MOSI datasets, demonstrating that MCIHN achieves superior performance.", "date": "2025-10-28", "categories": []}
{"id": "2510.03490", "title": "SEER: The Span-based Emotion Evidence Retrieval Benchmark", "abstract": "We introduce the SEER (Span-based Emotion Evidence Retrieval) Benchmark to test Large Language Models' (LLMs) ability to identify the specific spans of text that express emotion. Unlike traditional emotion recognition tasks that assign a single label to an entire sentence, SEER targets the underexplored task of emotion evidence detection: pinpointing which exact phrases convey emotion. This span-level approach is crucial for applications like empathetic dialogue and clinical support, which need to know how emotion is expressed, not just what the emotion is. SEER includes two tasks: identifying emotion evidence within a single sentence, and identifying evidence across a short passage of five consecutive sentences. It contains new annotations for both emotion and emotion evidence on 1200 real-world sentences. We evaluate 14 open-source LLMs and find that, while some models approach average human performance on single-sentence inputs, their accuracy degrades in longer passages. Our error analysis reveals key failure modes, including overreliance on emotion keywords and false positives in neutral text.", "date": "2025-10-03", "categories": []}
{"id": "2510.23617", "title": "An Enhanced Dual Transformer Contrastive Network for Multimodal Sentiment Analysis", "abstract": "Multimodal Sentiment Analysis (MSA) seeks to understand human emotions by jointly analyzing data from multiple modalities typically text and images offering a richer and more accurate interpretation than unimodal approaches. In this paper, we first propose BERT-ViT-EF, a novel model that combines powerful Transformer-based encoders BERT for textual input and ViT for visual input through an early fusion strategy. This approach facilitates deeper cross-modal interactions and more effective joint representation learning. To further enhance the model's capability, we propose an extension called the Dual Transformer Contrastive Network (DTCN), which builds upon BERT-ViT-EF. DTCN incorporates an additional Transformer encoder layer after BERT to refine textual context (before fusion) and employs contrastive learning to align text and image representations, fostering robust multimodal feature learning. Empirical results on two widely used MSA benchmarks MVSA-Single and TumEmo demonstrate the effectiveness of our approach. DTCN achieves best accuracy (78.4%) and F1-score (78.3%) on TumEmo, and delivers competitive performance on MVSA-Single, with 76.6% accuracy and 75.9% F1-score. These improvements highlight the benefits of early fusion and deeper contextual modeling in Transformer-based multimodal sentiment analysis.", "date": "2025-10-20", "categories": []}
{"id": "2504.19549", "title": "DEEMO: De-identity Multimodal Emotion Recognition and Reasoning", "abstract": "Emotion understanding is a critical yet challenging task. Most existing approaches rely heavily on identity-sensitive information, such as facial expressions and speech, which raises concerns about personal privacy. To address this, we introduce the De-identity Multimodal Emotion Recognition and Reasoning (DEEMO), a novel task designed to enable emotion understanding using de-identified video and audio inputs. The DEEMO dataset consists of two subsets: DEEMO-NFBL, which includes rich annotations of Non-Facial Body Language (NFBL), and DEEMO-MER, an instruction dataset for Multimodal Emotion Recognition and Reasoning using identity-free cues. This design supports emotion understanding without compromising identity privacy. In addition, we propose DEEMO-LLaMA, a Multimodal Large Language Model (MLLM) that integrates de-identified audio, video, and textual information to enhance both emotion recognition and reasoning. Extensive experiments show that DEEMO-LLaMA achieves state-of-the-art performance on both tasks, outperforming existing MLLMs by a significant margin, achieving 74.49% accuracy and 74.45% F1-score in de-identity emotion recognition, and 6.20 clue overlap and 7.66 label overlap in de-identity emotion reasoning. Our work contributes to ethical AI by advancing privacy-preserving emotion understanding and promoting responsible affective computing.", "date": "2025-04-28", "categories": []}
{"id": "2508.03543", "title": "EmoSteer-TTS: Fine-Grained and Training-Free Emotion-Controllable Text-to-Speech via Activation Steering", "abstract": "Text-to-speech (TTS) has shown great progress in recent years. However, most existing TTS systems offer only coarse and rigid emotion control, typically via discrete emotion labels or a carefully crafted and detailed emotional text prompt, making fine-grained emotion manipulation either inaccessible or unstable. These models also require extensive, high-quality datasets for training. To address these limitations, we propose EmoSteer-TTS, a novel training-free approach, to achieve fine-grained speech emotion control (conversion, interpolation, erasure) by activation steering. We first empirically observe that modifying a subset of the internal activations within a flow matching-based TTS model can effectively alter the emotional tone of synthesized speech. Building on this insight, we then develop a training-free and efficient algorithm, including activation extraction, emotional token searching, and inference-time steering, which can be seamlessly integrated into a wide range of pretrained models (e.g., F5-TTS, CosyVoice2, and E2-TTS). In addition, to derive effective steering vectors, we construct a curated emotional speech dataset with diverse speakers. Extensive experiments demonstrate that EmoSteer-TTS enables fine-grained, interpretable, and continuous control over speech emotion, outperforming the state-of-the-art (SOTA). To the best of our knowledge, this is the first method that achieves training-free and continuous fine-grained emotion control in TTS. Demo samples are available at https://emosteer-tts-demo.pages.dev/.", "date": "2025-08-05", "categories": []}
{"id": "2510.13493", "title": "ExpressNet-MoE: A Hybrid Deep Neural Network for Emotion Recognition", "abstract": "In many domains, including online education, healthcare, security, and human-computer interaction, facial emotion recognition (FER) is essential. Real-world FER is still difficult despite its significance because of some factors such as variable head positions, occlusions, illumination shifts, and demographic diversity. Engagement detection, which is essential for applications like virtual learning and customer services, is frequently challenging due to FER limitations by many current models. In this article, we propose ExpressNet-MoE, a novel hybrid deep learning model that blends both Convolution Neural Networks (CNNs) and Mixture of Experts (MoE) framework, to overcome the difficulties. Our model dynamically chooses the most pertinent expert networks, thus it aids in the generalization and providing flexibility to model across a wide variety of datasets. Our model improves on the accuracy of emotion recognition by utilizing multi-scale feature extraction to collect both global and local facial features. ExpressNet-MoE includes numerous CNN-based feature extractors, a MoE module for adaptive feature selection, and finally a residual network backbone for deep feature learning. To demonstrate efficacy of our proposed model we evaluated on several datasets, and compared with current state-of-the-art methods. Our model achieves accuracies of 74.77% on AffectNet (v7), 72.55% on AffectNet (v8), 84.29% on RAF-DB, and 64.66% on FER-2013. The results show how adaptive our model is and how it may be used to develop end-to-end emotion recognition systems in practical settings. Reproducible codes and results are made publicly accessible at https://github.com/DeeptimaanB/ExpressNet-MoE.", "date": "2025-10-15", "categories": []}
{"id": "2510.18016", "title": "ViBED-Net: Video Based Engagement Detection Network Using Face-Aware and Scene-Aware Spatiotemporal Cues", "abstract": "Engagement detection in online learning environments is vital for improving student outcomes and personalizing instruction. We present ViBED-Net (Video-Based Engagement Detection Network), a novel deep learning framework designed to assess student engagement from video data using a dual-stream architecture. ViBED-Net captures both facial expressions and full-scene context by processing facial crops and entire video frames through EfficientNetV2 for spatial feature extraction. These features are then analyzed over time using two temporal modeling strategies: Long Short-Term Memory (LSTM) networks and Transformer encoders. Our model is evaluated on the DAiSEE dataset, a large-scale benchmark for affective state recognition in e-learning. To enhance performance on underrepresented engagement classes, we apply targeted data augmentation techniques. Among the tested variants, ViBED-Net with LSTM achieves 73.43\\% accuracy, outperforming existing state-of-the-art approaches. ViBED-Net demonstrates that combining face-aware and scene-aware spatiotemporal cues significantly improves engagement detection accuracy. Its modular design allows flexibility for application across education, user experience research, and content personalization. This work advances video-based affective computing by offering a scalable, high-performing solution for real-world engagement analysis. The source code for this project is available on https://github.com/prateek-gothwal/ViBED-Net .", "date": "2025-10-20", "categories": []}
{"id": "2510.22197", "title": "Multi-dataset Joint Pre-training of Emotional EEG Enables Generalizable Affective Computing", "abstract": "Task-specific pre-training is essential when task representations diverge from generic pre-training features. Existing task-general pre-training EEG models struggle with complex tasks like emotion recognition due to mismatches between task-specific features and broad pre-training approaches. This work aims to develop a task-specific multi-dataset joint pre-training framework for cross-dataset emotion recognition, tackling problems of large inter-dataset distribution shifts, inconsistent emotion category definitions, and substantial inter-subject variability. We introduce a cross-dataset covariance alignment loss to align second-order statistical properties across datasets, enabling robust generalization without the need for extensive labels or per-subject calibration. To capture the long-term dependency and complex dynamics of EEG, we propose a hybrid encoder combining a Mamba-like linear attention channel encoder and a spatiotemporal dynamics model. Our method outperforms state-of-the-art large-scale EEG models by an average of 4.57% in AUROC for few-shot emotion recognition and 11.92% in accuracy for zero-shot generalization to a new dataset. Performance scales with the increase of datasets used in pre-training. Multi-dataset joint pre-training achieves a performance gain of 8.55% over single-dataset training. This work provides a scalable framework for task-specific pre-training and highlights its benefit in generalizable affective computing. Our code is available at https://github.com/ncclab-sustech/mdJPT_nips2025.", "date": "2025-10-25", "categories": []}
{"id": "2510.20239", "title": "Tri-Modal Severity Fused Diagnosis across Depression and Post-traumatic Stress Disorders", "abstract": "Depression and post traumatic stress disorder (PTSD) often co-occur with connected symptoms, complicating automated assessment, which is often binary and disorder specific. Clinically useful diagnosis needs severity aware cross disorder estimates and decision support explanations. Our unified tri modal affective severity framework synchronizes and fuses interview text with sentence level transformer embeddings, audio with log Mel statistics with deltas, and facial signals with action units, gaze, head and pose descriptors to output graded severities for diagnosing both depression (PHQ-8; 5 classes) and PTSD (3 classes). Standardized features are fused via a calibrated late fusion classifier, yielding per disorder probabilities and feature-level attributions. This severity aware tri-modal affective fusion approach is demoed on multi disorder concurrent depression and PTSD assessment. Stratified cross validation on DAIC derived corpora outperforms unimodal/ablation baselines. The fused model matches the strongest unimodal baseline on accuracy and weighted F1, while improving decision curve utility and robustness under noisy or missing modalities. For PTSD specifically, fusion reduces regression error and improves class concordance. Errors cluster between adjacent severities; extreme classes are identified reliably. Ablations show text contributes most to depression severity, audio and facial cues are critical for PTSD, whereas attributions align with linguistic and behavioral markers. Our approach offers reproducible evaluation and clinician in the loop support for affective clinical decision making.", "date": "2025-10-23", "categories": []}
{"id": "2510.20256", "title": "Calibrating Multimodal Consensus for Emotion Recognition", "abstract": "In recent years, Multimodal Emotion Recognition (MER) has made substantial progress. Nevertheless, most existing approaches neglect the semantic inconsistencies that may arise across modalities, such as conflicting emotional cues between text and visual inputs. Besides, current methods are often dominated by the text modality due to its strong representational capacity, which can compromise recognition accuracy. To address these challenges, we propose a model termed Calibrated Multimodal Consensus (CMC). CMC introduces a Pseudo Label Generation Module (PLGM) to produce pseudo unimodal labels, enabling unimodal pretraining in a self-supervised fashion. It then employs a Parameter-free Fusion Module (PFM) and a Multimodal Consensus Router (MCR) for multimodal finetuning, thereby mitigating text dominance and guiding the fusion process toward a more reliable consensus. Experimental results demonstrate that CMC achieves performance on par with or superior to state-of-the-art methods across four datasets, CH-SIMS, CH-SIMS v2, CMU-MOSI, and CMU-MOSEI, and exhibits notable advantages in scenarios with semantic inconsistencies on CH-SIMS and CH-SIMS v2. The implementation of this work is publicly accessible at https://github.com/gw-zhong/CMC.", "date": "2025-10-23", "categories": []}
{"id": "2510.19668", "title": "Unraveling Emotions with Pre-Trained Models", "abstract": "Transformer models have significantly advanced the field of emotion recognition. However, there are still open challenges when exploring open-ended queries for Large Language Models (LLMs). Although current models offer good results, automatic emotion analysis in open texts presents significant challenges, such as contextual ambiguity, linguistic variability, and difficulty interpreting complex emotional expressions. These limitations make the direct application of generalist models difficult. Accordingly, this work compares the effectiveness of fine-tuning and prompt engineering in emotion detection in three distinct scenarios: (i) performance of fine-tuned pre-trained models and general-purpose LLMs using simple prompts; (ii) effectiveness of different emotion prompt designs with LLMs; and (iii) impact of emotion grouping techniques on these models. Experimental tests attain metrics above 70% with a fine-tuned pre-trained model for emotion recognition. Moreover, the findings highlight that LLMs require structured prompt engineering and emotion grouping to enhance their performance. These advancements improve sentiment analysis, human-computer interaction, and understanding of user behavior across various domains.", "date": "2025-10-22", "categories": []}
{"id": "2510.18014", "title": "ManzaiSet: A Multimodal Dataset of Viewer Responses to Japanese Manzai Comedy", "abstract": "We present ManzaiSet, the first large scale multimodal dataset of viewer responses to Japanese manzai comedy, capturing facial videos and audio from 241 participants watching up to 10 professional performances in randomized order (94.6 percent watched >= 8; analyses focus on n=228). This addresses the Western centric bias in affective computing. Three key findings emerge: (1) k means clustering identified three distinct viewer types: High and Stable Appreciators (72.8 percent, n=166), Low and Variable Decliners (13.2 percent, n=30), and Variable Improvers (14.0 percent, n=32), with heterogeneity of variance (Brown Forsythe p < 0.001); (2) individual level analysis revealed a positive viewing order effect (mean slope = 0.488, t(227) = 5.42, p < 0.001, permutation p < 0.001), contradicting fatigue hypotheses; (3) automated humor classification (77 instances, 131 labels) plus viewer level response modeling found no type wise differences after FDR correction. The dataset enables culturally aware emotion AI development and personalized entertainment systems tailored to non Western contexts.", "date": "2025-10-20", "categories": []}
{"id": "2510.18036", "title": "Transformer Redesign for Late Fusion of Audio-Text Features on Ultra-Low-Power Edge Hardware", "abstract": "Deploying emotion recognition systems in real-world environments where devices must be small, low-power, and private remains a significant challenge. This is especially relevant for applications such as tension monitoring, conflict de-escalation, and responsive wearables, where cloud-based solutions are impractical. Multimodal emotion recognition has advanced through deep learning, but most systems remain unsuitable for deployment on ultra-constrained edge devices. Prior work typically relies on powerful hardware, lacks real-time performance, or uses unimodal input. This paper addresses that gap by presenting a hardware-aware emotion recognition system that combines acoustic and linguistic features using a late-fusion architecture optimised for Edge TPU. The design integrates a quantised transformer-based acoustic model with frozen keyword embeddings from a DSResNet-SE network, enabling real-time inference within a 1.8MB memory budget and 21-23ms latency. The pipeline ensures spectrogram alignment between training and deployment using MicroFrontend and MLTK. Evaluation on re-recorded, segmented IEMOCAP samples captured through the Coral Dev Board Micro microphone shows a 6.3% macro F1 improvement over unimodal baselines. This work demonstrates that accurate, real-time multimodal emotion inference is achievable on microcontroller-class edge platforms through task-specific fusion and hardware-guided model design.", "date": "2025-10-20", "categories": []}
{"id": "2510.16086", "title": "FSRF: Factorization-guided Semantic Recovery for Incomplete Multimodal Sentiment Analysis", "abstract": "In recent years, Multimodal Sentiment Analysis (MSA) has become a research hotspot that aims to utilize multimodal data for human sentiment understanding. Previous MSA studies have mainly focused on performing interaction and fusion on complete multimodal data, ignoring the problem of missing modalities in real-world applications due to occlusion, personal privacy constraints, and device malfunctions, resulting in low generalizability.\n  To this end, we propose a Factorization-guided Semantic Recovery Framework (FSRF) to mitigate the modality missing problem in the MSA task.\n  Specifically, we propose a de-redundant homo-heterogeneous factorization module that factorizes modality into modality-homogeneous, modality-heterogeneous, and noisy representations and design elaborate constraint paradigms for representation learning.\n  Furthermore, we design a distribution-aligned self-distillation module that fully recovers the missing semantics by utilizing bidirectional knowledge transfer.\n  Comprehensive experiments on two datasets indicate that FSRF has a significant performance advantage over previous methods with uncertain missing modalities.", "date": "2025-10-17", "categories": []}
{"id": "2510.17252", "title": "How News Feels: Understanding Affective Bias in Multilingual Headlines for Human-Centered Media Design", "abstract": "News media often shape the public mood not only by what they report but by how they frame it. The same event can appear calm in one outlet and alarming in another, reflecting subtle emotional bias in reporting. Negative or emotionally charged headlines tend to attract more attention and spread faster, which in turn encourages outlets to frame stories in ways that provoke stronger reactions. This research explores that tendency through large-scale emotion analysis of Bengali news. Using zero-shot inference with Gemma-3 4B, we analyzed 300000 Bengali news headlines and their content to identify the dominant emotion and overall tone of each. The findings reveal a clear dominance of negative emotions, particularly anger, fear, and disappointment, and significant variation in how similar stories are emotionally portrayed across outlets. Based on these insights, we propose design ideas for a human-centered news aggregator that visualizes emotional cues and helps readers recognize hidden affective framing in daily news.", "date": "2025-10-20", "categories": []}
{"id": "2510.17475", "title": "DAMSDAN: Distribution-Aware Multi-Source Domain Adaptation Network for Cross-Domain EEG-based Emotion Recognition", "abstract": "Significant inter-individual variability limits the generalization of EEG-based emotion recognition under cross-domain settings. We address two core challenges in multi-source adaptation: (1) dynamically modeling distributional heterogeneity across sources and quantifying their relevance to a target to reduce negative transfer; and (2) achieving fine-grained semantic consistency to strengthen class discrimination. We propose a distribution-aware multi-source domain adaptation network (DAMSDAN). DAMSDAN integrates prototype-based constraints with adversarial learning to drive the encoder toward discriminative, domain-invariant emotion representations. A domain-aware source weighting strategy based on maximum mean discrepancy (MMD) dynamically estimates inter-domain shifts and reweights source contributions. In addition, a prototype-guided conditional alignment module with dual pseudo-label interaction enhances pseudo-label reliability and enables category-level, fine-grained alignment, mitigating noise propagation and semantic drift. Experiments on SEED and SEED-IV show average accuracies of 94.86\\% and 79.78\\% for cross-subject, and 95.12\\% and 83.15\\% for cross-session protocols. On the large-scale FACED dataset, DAMSDAN achieves 82.88\\% (cross-subject). Extensive ablations and interpretability analyses corroborate the effectiveness of the proposed framework for cross-domain EEG-based emotion recognition.", "date": "2025-10-20", "categories": []}
{"id": "2505.19693", "title": "EmoSphere-SER: Enhancing Speech Emotion Recognition Through Spherical Representation with Auxiliary Classification", "abstract": "Speech emotion recognition predicts a speaker's emotional state from speech signals using discrete labels or continuous dimensions such as arousal, valence, and dominance (VAD). We propose EmoSphere-SER, a joint model that integrates spherical VAD region classification to guide VAD regression for improved emotion prediction. In our framework, VAD values are transformed into spherical coordinates that are divided into multiple spherical regions, and an auxiliary classification task predicts which spherical region each point belongs to, guiding the regression process. Additionally, we incorporate a dynamic weighting scheme and a style pooling layer with multi-head self-attention to capture spectral and temporal dynamics, further boosting performance. This combined training strategy reinforces structured learning and improves prediction consistency. Experimental results show that our approach exceeds baseline methods, confirming the validity of the proposed framework.", "date": "2025-05-26", "categories": []}
{"id": "2510.13534", "title": "High Semantic Features for the Continual Learning of Complex Emotions: a Lightweight Solution", "abstract": "Incremental learning is a complex process due to potential catastrophic forgetting of old tasks when learning new ones. This is mainly due to transient features that do not fit from task to task. In this paper, we focus on complex emotion recognition. First, we learn basic emotions and then, incrementally, like humans, complex emotions. We show that Action Units, describing facial muscle movements, are non-transient, highly semantical features that outperform those extracted by both shallow and deep convolutional neural networks. Thanks to this ability, our approach achieves interesting results when learning incrementally complex, compound emotions with an accuracy of 0.75 on the CFEE dataset and can be favorably compared to state-of-the-art results. Moreover, it results in a lightweight model with a small memory footprint.", "date": "2025-10-15", "categories": []}
{"id": "2510.13862", "title": "Ensembling Large Language Models to Characterize Affective Dynamics in Student-AI Tutor Dialogues", "abstract": "While recent studies have examined the leaning impact of large language model (LLM) in educational contexts, the affective dynamics of LLM-mediated tutoring remain insufficiently understood. This work introduces the first ensemble-LLM framework for large-scale affect sensing in tutoring dialogues, advancing the conversation on responsible pathways for integrating generative AI into education by attending to learners' evolving affective states. To achieve this, we analyzed two semesters' worth of 16,986 conversational turns exchanged between PyTutor, an LLM-powered AI tutor, and 261 undergraduate learners across three U.S. institutions. To investigate learners' emotional experiences, we generate zero-shot affect annotations from three frontier LLMs (Gemini, GPT-4o, Claude), including scalar ratings of valence, arousal, and learning-helpfulness, along with free-text emotion labels. These estimates are fused through rank-weighted intra-model pooling and plurality consensus across models to produce robust emotion profiles. Our analysis shows that during interaction with the AI tutor, students typically report mildly positive affect and moderate arousal. Yet learning is not uniformly smooth: confusion and curiosity are frequent companions to problem solving, and frustration, while less common, still surfaces in ways that can derail progress. Emotional states are short-lived--positive moments last slightly longer than neutral or negative ones, but they are fragile and easily disrupted. Encouragingly, negative emotions often resolve quickly, sometimes rebounding directly into positive states. Neutral moments frequently act as turning points, more often steering students upward than downward, suggesting opportunities for tutors to intervene at precisely these junctures.", "date": "2025-10-13", "categories": []}
{"id": "2510.13906", "title": "Switchboard-Affect: Emotion Perception Labels from Conversational Speech", "abstract": "Understanding the nuances of speech emotion dataset curation and labeling is essential for assessing speech emotion recognition (SER) model potential in real-world applications. Most training and evaluation datasets contain acted or pseudo-acted speech (e.g., podcast speech) in which emotion expressions may be exaggerated or otherwise intentionally modified. Furthermore, datasets labeled based on crowd perception often lack transparency regarding the guidelines given to annotators. These factors make it difficult to understand model performance and pinpoint necessary areas for improvement. To address this gap, we identified the Switchboard corpus as a promising source of naturalistic conversational speech, and we trained a crowd to label the dataset for categorical emotions (anger, contempt, disgust, fear, sadness, surprise, happiness, tenderness, calmness, and neutral) and dimensional attributes (activation, valence, and dominance). We refer to this label set as Switchboard-Affect (SWB-Affect). In this work, we present our approach in detail, including the definitions provided to annotators and an analysis of the lexical and paralinguistic cues that may have played a role in their perception. In addition, we evaluate state-of-the-art SER models, and we find variable performance across the emotion categories with especially poor generalization for anger. These findings underscore the importance of evaluation with datasets that capture natural affective variations in speech. We release the labels for SWB-Affect to enable further analysis in this domain.", "date": "2025-10-14", "categories": []}
{"id": "2504.13211", "title": "MIRROR: Multimodal Cognitive Reframing Therapy for Rolling with Resistance", "abstract": "Recent studies have explored the use of large language models (LLMs) in psychotherapy; however, text-based cognitive behavioral therapy (CBT) models often struggle with client resistance, which can weaken therapeutic alliance. To address this, we propose a multimodal approach that incorporates nonverbal cues, which allows the AI therapist to better align its responses with the client's negative emotional state. Specifically, we introduce a new synthetic dataset, Mirror (Multimodal Interactive Rolling with Resistance), which is a novel synthetic dataset that pairs each client's statements with corresponding facial images. Using this dataset, we train baseline vision language models (VLMs) so that they can analyze facial cues, infer emotions, and generate empathetic responses to effectively manage client resistance. These models are then evaluated in terms of both their counseling skills as a therapist, and the strength of therapeutic alliance in the presence of client resistance. Our results demonstrate that Mirror significantly enhances the AI therapist's ability to handle resistance, which outperforms existing text-based CBT approaches. Human expert evaluations further confirm the effectiveness of our approach in managing client resistance and fostering therapeutic alliance.", "date": "2025-04-16", "categories": []}
{"id": "2509.19330", "title": "LibEMER: A novel benchmark and algorithms library for EEG-based Multimodal Emotion Recognition", "abstract": "EEG-based multimodal emotion recognition(EMER) has gained significant attention and witnessed notable advancements, the inherent complexity of human neural systems has motivated substantial efforts toward multimodal approaches. However, this field currently suffers from three critical limitations: (i) the absence of open-source implementations. (ii) the lack of standardized and transparent benchmarks for fair performance analysis. (iii) in-depth discussion regarding main challenges and promising research directions is a notable scarcity. To address these challenges, we introduce LibEMER, a unified evaluation framework that provides fully reproducible PyTorch implementations of curated deep learning methods alongside standardized protocols for data preprocessing, model realization, and experimental setups. This framework enables unbiased performance assessment on three widely-used public datasets across two learning tasks. The open-source library is publicly accessible at: https://anonymous.4open.science/r/2025ULUIUBUEUMUEUR485384", "date": "2025-09-14", "categories": []}
{"id": "2510.12819", "title": "Beyond Discrete Categories: Multi-Task Valence-Arousal Modeling for Pet Vocalization Analysis", "abstract": "Traditional pet emotion recognition from vocalizations, based on discrete classification, struggles with ambiguity and capturing intensity variations. We propose a continuous Valence-Arousal (VA) model that represents emotions in a two-dimensional space. Our method uses an automatic VA label generation algorithm, enabling large-scale annotation of 42,553 pet vocalization samples. A multi-task learning framework jointly trains VA regression with auxiliary tasks (emotion, body size, gender) to enhance prediction by improving feature learning. Our Audio Transformer model achieves a validation Valence Pearson correlation of r = 0.9024 and an Arousal r = 0.7155, effectively resolving confusion between discrete categories like \"territorial\" and \"happy.\" This work introduces the first continuous VA framework for pet vocalization analysis, offering a more expressive representation for human-pet interaction, veterinary diagnostics, and behavioral training. The approach shows strong potential for deployment in consumer products like AI pet emotion translators.", "date": "2025-10-09", "categories": []}
{"id": "2510.13557", "title": "Modeling Cultural Bias in Facial Expression Recognition with Adaptive Agents", "abstract": "Facial expression recognition (FER) must remain robust under both cultural variation and perceptually degraded visual conditions, yet most existing evaluations assume homogeneous data and high-quality imagery. We introduce an agent-based, streaming benchmark that reveals how cross-cultural composition and progressive blurring interact to shape face recognition robustness. Each agent operates in a frozen CLIP feature space with a lightweight residual adapter trained online at sigma=0 and fixed during testing. Agents move and interact on a 5x5 lattice, while the environment provides inputs with sigma-scheduled Gaussian blur. We examine monocultural populations (Western-only, Asian-only) and mixed environments with balanced (5/5) and imbalanced (8/2, 2/8) compositions, as well as different spatial contact structures. Results show clear asymmetric degradation curves between cultural groups: JAFFE (Asian) populations maintain higher performance at low blur but exhibit sharper drops at intermediate stages, whereas KDEF (Western) populations degrade more uniformly. Mixed populations exhibit intermediate patterns, with balanced mixtures mitigating early degradation, but imbalanced settings amplify majority-group weaknesses under high blur. These findings quantify how cultural composition and interaction structure influence the robustness of FER as perceptual conditions deteriorate.", "date": "2025-10-15", "categories": []}
{"id": "2407.13753", "title": "Exploring Facial Biomarkers for Depression through Temporal Analysis of Action Units", "abstract": "Depression is characterized by persistent sadness and loss of interest, significantly impairing daily functioning and now a widespread mental disorder. Traditional diagnostic methods rely on subjective assessments, necessitating objective approaches for accurate diagnosis. Our study investigates the use of facial action units (AUs) and emotions as biomarkers for depression. We analyzed facial expressions from video data of participants classified with or without depression. Our methodology involved detailed feature extraction, mean intensity comparisons of key AUs, and the application of time series classification models. Furthermore, we employed Principal Component Analysis (PCA) and various clustering algorithms to explore the variability in emotional expression patterns. Results indicate significant differences in the intensities of AUs associated with sadness and happiness between the groups, highlighting the potential of facial analysis in depression assessment.", "date": "2024-07-18", "categories": []}
{"id": "2506.19887", "title": "MATER: Multi-level Acoustic and Textual Emotion Representation for Interpretable Speech Emotion Recognition", "abstract": "This paper presents our contributions to the Speech Emotion Recognition in Naturalistic Conditions (SERNC) Challenge, where we address categorical emotion recognition and emotional attribute prediction. To handle the complexities of natural speech, including intra- and inter-subject variability, we propose Multi-level Acoustic-Textual Emotion Representation (MATER), a novel hierarchical framework that integrates acoustic and textual features at the word, utterance, and embedding levels. By fusing low-level lexical and acoustic cues with high-level contextualized representations, MATER effectively captures both fine-grained prosodic variations and semantic nuances. Additionally, we introduce an uncertainty-aware ensemble strategy to mitigate annotator inconsistencies, improving robustness in ambiguous emotional expressions. MATER ranks fourth in both tasks with a Macro-F1 of 41.01% and an average CCC of 0.5928, securing second place in valence prediction with an impressive CCC of 0.6941.", "date": "2025-06-24", "categories": []}
{"id": "2505.05043", "title": "xTrace: A Facial Expressive Behaviour Analysis Tool for Continuous Affect Recognition", "abstract": "Recognising expressive behaviours in face videos is a long-standing challenge in Affective Computing. Despite significant advancements in recent years, it still remains a challenge to build a robust and reliable system for naturalistic and in-the-wild facial expressive behaviour analysis in real time. This paper addresses two key challenges in building such a system: (1). The paucity of large-scale labelled facial affect video datasets with extensive coverage of the 2D emotion space, and (2). The difficulty of extracting facial video features that are discriminative, interpretable, robust, and computationally efficient. Toward addressing these challenges, this work introduces xTrace, a robust tool for facial expressive behaviour analysis and predicting continuous values of dimensional emotions, namely valence and arousal, from in-the-wild face videos. To address challenge (1), the proposed affect recognition model is trained on the largest facial affect video data set, containing $\\sim$450k videos that cover most emotion zones in the dimensional emotion space, making xTrace highly versatile in analysing a wide spectrum of naturalistic expressive behaviours. To address challenge (2), xTrace uses facial affect descriptors that are not only explainable, but can also achieve a high degree of accuracy and robustness with low computational complexity. The key components of xTrace are benchmarked against three existing tools: MediaPipe, OpenFace, and Augsburg Affect Toolbox. On an in-the-wild benchmarking set composed of $\\sim$50k videos, xTrace achieves 0.86 mean Concordance Correlation Coefficient (CCC) and on the SEWA test set it achieves 0.75 mean CCC, outperforming existing SOTA by $\\sim$7.1\\%.", "date": "2025-05-08", "categories": []}
{"id": "2508.01181", "title": "Benchmarking and Bridging Emotion Conflicts for Multimodal Emotion Reasoning", "abstract": "Despite their strong performance in multimodal emotion reasoning, existing Multimodal Large Language Models (MLLMs) often overlook the scenarios involving emotion conflicts, where emotional cues from different modalities are inconsistent. To fill this gap, we first introduce CA-MER, a new benchmark designed to examine MLLMs under realistic emotion conflicts. It consists of three subsets: video-aligned, audio-aligned, and consistent, where only one or all modalities reflect the true emotion. However, evaluations on our CA-MER reveal that current state-of-the-art emotion MLLMs systematically over-rely on audio signal during emotion conflicts, neglecting critical cues from visual modality. To mitigate this bias, we propose MoSEAR, a parameter-efficient framework that promotes balanced modality integration. MoSEAR consists of two modules: (1)MoSE, modality-specific experts with a regularized gating mechanism that reduces modality bias in the fine-tuning heads; and (2)AR, an attention reallocation mechanism that rebalances modality contributions in frozen backbones during inference. Our framework offers two key advantages: it mitigates emotion conflicts and improves performance on consistent samples-without incurring a trade-off between audio and visual modalities. Experiments on multiple benchmarks-including MER2023, EMER, DFEW, and our CA-MER-demonstrate that MoSEAR achieves state-of-the-art performance, particularly under modality conflict conditions.", "date": "2025-08-02", "categories": []}
{"id": "2406.02642", "title": "E-ICL: Enhancing Fine-Grained Emotion Recognition through the Lens of Prototype Theory", "abstract": "In-context learning (ICL) achieves remarkable performance in various domains such as knowledge acquisition, commonsense reasoning, and semantic understanding. However, its performance significantly deteriorates for emotion detection tasks, especially fine-grained emotion recognition. The underlying reasons for this remain unclear. In this paper, we identify the reasons behind ICL's poor performance from the perspective of prototype theory and propose a method to address this issue. Specifically, we conduct extensive pilot experiments and find that ICL conforms to the prototype theory on fine-grained emotion recognition. Based on this theory, we uncover the following deficiencies in ICL: (1) It relies on prototypes (example-label pairs) that are semantically similar but emotionally inaccurate to predict emotions. (2) It is prone to interference from irrelevant categories, affecting the accuracy and robustness of the predictions. To address these issues, we propose an Emotion Context Learning method (E-ICL) on fine-grained emotion recognition. E-ICL relies on more emotionally accurate prototypes to predict categories by referring to emotionally similar examples with dynamic labels. Simultaneously, E-ICL employs an exclusionary emotion prediction strategy to avoid interference from irrelevant categories, thereby increasing its accuracy and robustness. Note that the entire process is accomplished with the assistance of a plug-and-play emotion auxiliary model, without additional training. Experiments on the fine-grained emotion datasets EDOS, Empathetic-Dialogues, EmpatheticIntent, and GoEmotions show that E-ICL achieves superior emotion prediction performance. Furthermore, even when the emotion auxiliary model used is lower than 10% of the LLMs, E-ICL can still boost the performance of LLMs by over 4% on multiple datasets.", "date": "2024-06-04", "categories": []}
{"id": "2510.08606", "title": "Centering Emotion Hotspots: Multimodal Local-Global Fusion and Cross-Modal Alignment for Emotion Recognition in Conversations", "abstract": "Emotion Recognition in Conversations (ERC) is hard because discriminative evidence is sparse, localized, and often asynchronous across modalities. We center ERC on emotion hotspots and present a unified model that detects per-utterance hotspots in text, audio, and video, fuses them with global features via Hotspot-Gated Fusion, and aligns modalities using a routed Mixture-of-Aligners; a cross-modal graph encodes conversational structure. This design focuses modeling on salient spans, mitigates misalignment, and preserves context. Experiments on standard ERC benchmarks show consistent gains over strong baselines, with ablations confirming the contributions of HGF and MoA. Our results point to a hotspot-centric view that can inform future multimodal learning, offering a new perspective on modality fusion in ERC.", "date": "2025-10-07", "categories": []}
{"id": "2510.08802", "title": "Edu-EmotionNet: Cross-Modality Attention Alignment with Temporal Feedback Loops", "abstract": "Understanding learner emotions in online education is critical for improving engagement and personalized instruction. While prior work in emotion recognition has explored multimodal fusion and temporal modeling, existing methods often rely on static fusion strategies and assume that modality inputs are consistently reliable, which is rarely the case in real-world learning environments. We introduce Edu-EmotionNet, a novel framework that jointly models temporal emotion evolution and modality reliability for robust affect recognition. Our model incorporates three key components: a Cross-Modality Attention Alignment (CMAA) module for dynamic cross-modal context sharing, a Modality Importance Estimator (MIE) that assigns confidence-based weights to each modality at every time step, and a Temporal Feedback Loop (TFL) that leverages previous predictions to enforce temporal consistency. Evaluated on educational subsets of IEMOCAP and MOSEI, re-annotated for confusion, curiosity, boredom, and frustration, Edu-EmotionNet achieves state-of-the-art performance and demonstrates strong robustness to missing or noisy modalities. Visualizations confirm its ability to capture emotional transitions and adaptively prioritize reliable signals, making it well suited for deployment in real-time learning systems", "date": "2025-10-09", "categories": []}
{"id": "2510.09072", "title": "Emotion-Disentangled Embedding Alignment for Noise-Robust and Cross-Corpus Speech Emotion Recognition", "abstract": "Effectiveness of speech emotion recognition in real-world scenarios is often hindered by noisy environments and variability across datasets. This paper introduces a two-step approach to enhance the robustness and generalization of speech emotion recognition models through improved representation learning. First, our model employs EDRL (Emotion-Disentangled Representation Learning) to extract class-specific discriminative features while preserving shared similarities across emotion categories. Next, MEA (Multiblock Embedding Alignment) refines these representations by projecting them into a joint discriminative latent subspace that maximizes covariance with the original speech input. The learned EDRL-MEA embeddings are subsequently used to train an emotion classifier using clean samples from publicly available datasets, and are evaluated on unseen noisy and cross-corpus speech samples. Improved performance under these challenging conditions demonstrates the effectiveness of the proposed method.", "date": "2025-10-10", "categories": []}
{"id": "2510.07320", "title": "Deep Learning Based Approach to Enhanced Recognition of Emotions and Behavioral Patterns of Autistic Children", "abstract": "Autism Spectrum Disorder significantly influences the communication abilities, learning processes, behavior, and social interactions of individuals. Although early intervention and customized educational strategies are critical to improving outcomes, there is a pivotal gap in understanding and addressing nuanced behavioral patterns and emotional identification in autistic children prior to skill development. This extended research delves into the foundational step of recognizing and mapping these patterns as a prerequisite to improving learning and soft skills. Using a longitudinal approach to monitor emotions and behaviors, this study aims to establish a baseline understanding of the unique needs and challenges faced by autistic students, particularly in the Information Technology domain, where opportunities are markedly limited. Through a detailed analysis of behavioral trends over time, we propose a targeted framework for developing applications and technical aids designed to meet these identified needs. Our research underscores the importance of a sequential and evidence-based intervention approach that prioritizes a deep understanding of each child's behavioral and emotional landscape as the basis for effective skill development. By shifting the focus toward early identification of behavioral patterns, we aim to foster a more inclusive and supportive learning environment that can significantly improve the educational and developmental trajectory of children with ASD.", "date": "2025-08-29", "categories": []}
{"id": "2510.06600", "title": "Fine-Grained Emotion Recognition via In-Context Learning", "abstract": "Fine-grained emotion recognition aims to identify the emotional type in queries through reasoning and decision-making processes, playing a crucial role in various systems. Recent methods use In-Context Learning (ICL), enhancing the representation of queries in the reasoning process through semantically similar examples, while further improving emotion recognition by explaining the reasoning mechanisms. However, these methods enhance the reasoning process but overlook the decision-making process. This paper investigates decision-making in fine-grained emotion recognition through prototype theory. We show that ICL relies on similarity matching between query representations and emotional prototypes within the model, where emotion-accurate representations are critical. However, semantically similar examples often introduce emotional discrepancies, hindering accurate representations and causing errors. To address this, we propose Emotion In-Context Learning (EICL), which introduces emotionally similar examples and uses a dynamic soft-label strategy to improve query representations in the emotion reasoning process. A two-stage exclusion strategy is then employed to assess similarity from multiple angles, further optimizing the decision-making process. Extensive experiments show that EICL significantly outperforms ICL on multiple datasets.", "date": "2025-10-08", "categories": []}
{"id": "2510.05826", "title": "Leveraging Vision Transformers for Enhanced Classification of Emotions using ECG Signals", "abstract": "Biomedical signals provide insights into various conditions affecting the human body. Beyond diagnostic capabilities, these signals offer a deeper understanding of how specific organs respond to an individual's emotions and feelings. For instance, ECG data can reveal changes in heart rate variability linked to emotional arousal, stress levels, and autonomic nervous system activity. This data offers a window into the physiological basis of our emotional states. Recent advancements in the field diverge from conventional approaches by leveraging the power of advanced transformer architectures, which surpass traditional machine learning and deep learning methods. We begin by assessing the effectiveness of the Vision Transformer (ViT), a forefront model in image classification, for identifying emotions in imaged ECGs. Following this, we present and evaluate an improved version of ViT, integrating both CNN and SE blocks, aiming to bolster performance on imaged ECGs associated with emotion detection. Our method unfolds in two critical phases: first, we apply advanced preprocessing techniques for signal purification and converting signals into interpretable images using continuous wavelet transform and power spectral density analysis; second, we unveil a performance-boosted vision transformer architecture, cleverly enhanced with convolutional neural network components, to adeptly tackle the challenges of emotion recognition. Our methodology's robustness and innovation were thoroughly tested using ECG data from the YAAD and DREAMER datasets, leading to remarkable outcomes. For the YAAD dataset, our approach outperformed existing state-of-the-art methods in classifying seven unique emotional states, as well as in valence and arousal classification. Similarly, in the DREAMER dataset, our method excelled in distinguishing between valence, arousal and dominance, surpassing current leading techniques.", "date": "2025-10-07", "categories": []}
{"id": "2510.06072", "title": "EmoHRNet: High-Resolution Neural Network Based Speech Emotion Recognition", "abstract": "Speech emotion recognition (SER) is pivotal for enhancing human-machine interactions. This paper introduces \"EmoHRNet\", a novel adaptation of High-Resolution Networks (HRNet) tailored for SER. The HRNet structure is designed to maintain high-resolution representations from the initial to the final layers. By transforming audio samples into spectrograms, EmoHRNet leverages the HRNet architecture to extract high-level features. EmoHRNet's unique architecture maintains high-resolution representations throughout, capturing both granular and overarching emotional cues from speech signals. The model outperforms leading models, achieving accuracies of 92.45% on RAVDESS, 80.06% on IEMOCAP, and 92.77% on EMOVO. Thus, we show that EmoHRNet sets a new benchmark in the SER domain.", "date": "2025-10-07", "categories": []}
{"id": "2306.08933", "title": "Towards Interpretability in Audio and Visual Affective Machine Learning: A Review", "abstract": "Machine learning is frequently used in affective computing, but presents challenges due the opacity of state-of-the-art machine learning methods. Because of the impact affective machine learning systems may have on an individual's life, it is important that models be made transparent to detect and mitigate biased decision making. In this regard, affective machine learning could benefit from the recent advancements in explainable artificial intelligence (XAI) research. We perform a structured literature review to examine the use of interpretability in the context of affective machine learning. We focus on studies using audio, visual, or audiovisual data for model training and identified 29 research articles. Our findings show an emergence of the use of interpretability methods in the last five years. However, their use is currently limited regarding the range of methods used, the depth of evaluations, and the consideration of use-cases. We outline the main gaps in the research and provide recommendations for researchers that aim to implement interpretable methods for affective machine learning.", "date": "2023-06-15", "categories": []}
{"id": "2505.23631", "title": "Human Empathy as Encoder: AI-Assisted Depression Assessment in Special Education", "abstract": "Assessing student depression in sensitive environments like special education is challenging. Standardized questionnaires may not fully reflect students' true situations. Furthermore, automated methods often falter with rich student narratives, lacking the crucial, individualized insights stemming from teachers' empathetic connections with students. Existing methods often fail to address this ambiguity or effectively integrate educator understanding. To address these limitations by fostering a synergistic human-AI collaboration, this paper introduces Human Empathy as Encoder (HEAE), a novel, human-centered AI framework for transparent and socially responsible depression severity assessment. Our approach uniquely integrates student narrative text with a teacher-derived, 9-dimensional \"Empathy Vector\" (EV), its dimensions guided by the PHQ-9 framework,to explicitly translate tacit empathetic insight into a structured AI input enhancing rather than replacing human judgment. Rigorous experiments optimized the multimodal fusion, text representation, and classification architecture, achieving 82.74% accuracy for 7-level severity classification. This work demonstrates a path toward more responsible and ethical affective computing by structurally embedding human empathy", "date": "2025-05-29", "categories": []}
{"id": "2510.03337", "title": "Error correction in multiclass image classification of facial emotion on unbalanced samples", "abstract": "This paper considers the problem of error correction in multi-class classification of face images on unbalanced samples. The study is based on the analysis of a data frame containing images labeled by seven different emotional states of people of different ages. Particular attention is paid to the problem of class imbalance, in which some emotions significantly prevail over others. To solve the classification problem, a neural network model based on LSTM with an attention mechanism focusing on key areas of the face that are informative for emotion recognition is used. As part of the experiments, the model is trained on all possible configurations of subsets of six classes with subsequent error correction for the seventh class, excluded at the training stage. The results show that correction is possible for all classes, although the degree of success varies: some classes are better restored, others are worse. In addition, on the test sample, when correcting some classes, an increase in key quality metrics for small classes was recorded, which indicates the promise of the proposed approach in solving applied problems related to the search for rare events, for example, in anti-fraud systems. Thus, the proposed method can be effectively applied in facial expression analysis systems and in tasks requiring stable classification under skewed class distribution.", "date": "2025-10-02", "categories": []}
{"id": "2510.03060", "title": "Semantic Differentiation in Speech Emotion Recognition: Insights from Descriptive and Expressive Speech Roles", "abstract": "Speech Emotion Recognition (SER) is essential for improving human-computer interaction, yet its accuracy remains constrained by the complexity of emotional nuances in speech. In this study, we distinguish between descriptive semantics, which represents the contextual content of speech, and expressive semantics, which reflects the speaker's emotional state. After watching emotionally charged movie segments, we recorded audio clips of participants describing their experiences, along with the intended emotion tags for each clip, participants' self-rated emotional responses, and their valence/arousal scores. Through experiments, we show that descriptive semantics align with intended emotions, while expressive semantics correlate with evoked emotions. Our findings inform SER applications in human-AI interaction and pave the way for more context-aware AI systems.", "date": "2025-10-03", "categories": []}
{"id": "2510.03066", "title": "InsideOut: An EfficientNetV2-S Based Deep Learning Framework for Robust Multi-Class Facial Emotion Recognition", "abstract": "Facial Emotion Recognition (FER) is a key task in affective computing, enabling applications in human-computer interaction, e-learning, healthcare, and safety systems. Despite advances in deep learning, FER remains challenging due to occlusions, illumination and pose variations, subtle intra-class differences, and dataset imbalance that hinders recognition of minority emotions. We present InsideOut, a reproducible FER framework built on EfficientNetV2-S with transfer learning, strong data augmentation, and imbalance-aware optimization. The approach standardizes FER2013 images, applies stratified splitting and augmentation, and fine-tunes a lightweight classification head with class-weighted loss to address skewed distributions. InsideOut achieves 62.8% accuracy with a macro averaged F1 of 0.590 on FER2013, showing competitive results compared to conventional CNN baselines. The novelty lies in demonstrating that efficient architectures, combined with tailored imbalance handling, can provide practical, transparent, and reproducible FER solutions.", "date": "2025-10-03", "categories": []}
{"id": "2411.17674", "title": "Push the Limit of Multi-modal Emotion Recognition by Prompting LLMs with Receptive-Field-Aware Attention Weighting", "abstract": "Understanding the emotions in a dialogue usually requires external knowledge to accurately understand the contents. As the LLMs become more and more powerful, we do not want to settle on the limited ability of the pre-trained language model. However, the LLMs either can only process text modality or are too expensive to process the multimedia information. We aim to utilize both the power of LLMs and the supplementary features from the multimedia modalities. In this paper, we present a framework, Lantern, that can improve the performance of a certain vanilla model by prompting large language models with receptive-field-aware attention weighting. This framework trained a multi-task vanilla model to produce probabilities of emotion classes and dimension scores. These predictions are fed into the LLMs as references to adjust the predicted probabilities of each emotion class with its external knowledge and contextual understanding. We slice the dialogue into different receptive fields, and each sample is included in exactly t receptive fields. Finally, the predictions of LLMs are merged with a receptive-field-aware attention-driven weighting module. In the experiments, vanilla models CORECT and SDT are deployed in Lantern with GPT-4 or Llama-3.1-405B. The experiments in IEMOCAP with 4-way and 6-way settings demonstrated that the Lantern can significantly improve the performance of current vanilla models by up to 1.23% and 1.80%.", "date": "2024-11-26", "categories": []}
{"id": "2510.01677", "title": "Beyond Simple Fusion: Adaptive Gated Fusion for Robust Multimodal Sentiment Analysis", "abstract": "Multimodal sentiment analysis (MSA) leverages information fusion from diverse modalities (e.g., text, audio, visual) to enhance sentiment prediction. However, simple fusion techniques often fail to account for variations in modality quality, such as those that are noisy, missing, or semantically conflicting. This oversight leads to suboptimal performance, especially in discerning subtle emotional nuances. To mitigate this limitation, we introduce a simple yet efficient \\textbf{A}daptive \\textbf{G}ated \\textbf{F}usion \\textbf{N}etwork that adaptively adjusts feature weights via a dual gate fusion mechanism based on information entropy and modality importance. This mechanism mitigates the influence of noisy modalities and prioritizes informative cues following unimodal encoding and cross-modal interaction. Experiments on CMU-MOSI and CMU-MOSEI show that AGFN significantly outperforms strong baselines in accuracy, effectively discerning subtle emotions with robust performance. Visualization analysis of feature representations demonstrates that AGFN enhances generalization by learning from a broader feature distribution, achieved by reducing the correlation between feature location and prediction error, thereby decreasing reliance on specific locations and creating more robust multimodal feature representations.", "date": "2025-10-02", "categories": []}
{"id": "2505.15772", "title": "MIKU-PAL: An Automated and Standardized Multi-Modal Method for Speech Paralinguistic and Affect Labeling", "abstract": "Acquiring large-scale emotional speech data with strong consistency remains a challenge for speech synthesis. This paper presents MIKU-PAL, a fully automated multimodal pipeline for extracting high-consistency emotional speech from unlabeled video data. Leveraging face detection and tracking algorithms, we developed an automatic emotion analysis system using a multimodal large language model (MLLM). Our results demonstrate that MIKU-PAL can achieve human-level accuracy (68.5% on MELD) and superior consistency (0.93 Fleiss kappa score) while being much cheaper and faster than human annotation. With the high-quality, flexible, and consistent annotation from MIKU-PAL, we can annotate fine-grained speech emotion categories of up to 26 types, validated by human annotators with 83% rationality ratings. Based on our proposed system, we further released a fine-grained emotional speech dataset MIKU-EmoBench(131.2 hours) as a new benchmark for emotional text-to-speech and visual voice cloning.", "date": "2025-05-21", "categories": []}
{"id": "2506.06820", "title": "Beyond Classification: Towards Speech Emotion Reasoning with Multitask AudioLLMs", "abstract": "Audio Large Language Models (AudioLLMs) have achieved strong results in semantic tasks like speech recognition and translation, but remain limited in modeling paralinguistic cues such as emotion. Existing approaches often treat emotion understanding as a classification problem, offering little insight into the underlying rationale behind predictions. In this work, we explore emotion reasoning, a strategy that leverages the generative capabilities of AudioLLMs to enhance emotion recognition by producing semantically aligned, evidence-grounded explanations. To support this in multitask AudioLLMs, we introduce a unified framework combining reasoning-augmented data supervision, dual-encoder architecture, and task-alternating training. This approach enables AudioLLMs to effectively learn different tasks while incorporating emotional reasoning. Experiments on IEMOCAP and MELD show that our approach not only improves emotion prediction accuracy but also enhances the coherence and evidential grounding of the generated responses. Experiments on two out-of-domain datasets demonstrate the generalization capabilities of the resulting model.", "date": "2025-06-07", "categories": []}
{"id": "2509.22729", "title": "Multi-Modal Sentiment Analysis with Dynamic Attention Fusion", "abstract": "Traditional sentiment analysis has long been a unimodal task, relying solely on text. This approach overlooks non-verbal cues such as vocal tone and prosody that are essential for capturing true emotional intent. We introduce Dynamic Attention Fusion (DAF), a lightweight framework that combines frozen text embeddings from a pretrained language model with acoustic features from a speech encoder, using an adaptive attention mechanism to weight each modality per utterance. Without any finetuning of the underlying encoders, our proposed DAF model consistently outperforms both static fusion and unimodal baselines on a large multimodal benchmark. We report notable gains in F1-score and reductions in prediction error and perform a variety of ablation studies that support our hypothesis that the dynamic weighting strategy is crucial for modeling emotionally complex inputs. By effectively integrating verbal and non-verbal information, our approach offers a more robust foundation for sentiment prediction and carries broader impact for affective computing applications -- from emotion recognition and mental health assessment to more natural human computer interaction.", "date": "2025-09-25", "categories": []}
{"id": "2509.24214", "title": "Scalable Audio-Visual Masked Autoencoders for Efficient Affective Video Facial Analysis", "abstract": "Affective video facial analysis (AVFA) has emerged as a key research field for building emotion-aware intelligent systems, yet this field continues to suffer from limited data availability. In recent years, the self-supervised learning (SSL) technique of Masked Autoencoders (MAE) has gained momentum, with growing adaptations in its audio-visual contexts. While scaling has proven essential for breakthroughs in general multi-modal learning domains, its specific impact on AVFA remains largely unexplored. Another core challenge in this field is capturing both intra- and inter-modal correlations through scalable audio-visual representations. To tackle these issues, we propose AVF-MAE++, a family of audio-visual MAE models designed to efficiently investigate the scaling properties in AVFA while enhancing cross-modal correlation modeling. Our framework introduces a novel dual masking strategy across audio and visual modalities and strengthens modality encoders with a more holistic design to better support scalable pre-training. Additionally, we present the Iterative Audio-Visual Correlation Learning Module, which improves correlation learning within the SSL paradigm, bridging the limitations of previous methods. To support smooth adaptation and reduce overfitting risks, we further introduce a progressive semantic injection strategy, organizing the model training into three structured stages. Extensive experiments conducted on 17 datasets, covering three major AVFA tasks, demonstrate that AVF-MAE++ achieves consistent state-of-the-art performance across multiple benchmarks. Comprehensive ablation studies further highlight the importance of each proposed component and provide deeper insights into the design choices driving these improvements. Our code and models have been publicly released at Github.", "date": "2025-09-29", "categories": []}
{"id": "2509.24298", "title": "Bridging the behavior-neural gap: A multimodal AI reveals the brain's geometry of emotion more accurately than human self-reports", "abstract": "The ability to represent emotion plays a significant role in human cognition and social interaction, yet the high-dimensional geometry of this affective space and its neural underpinnings remain debated. A key challenge, the `behavior-neural gap,' is the limited ability of human self-reports to predict brain activity. Here we test the hypothesis that this gap arises from the constraints of traditional rating scales and that large-scale similarity judgments can more faithfully capture the brain's affective geometry. Using AI models as `cognitive agents,' we collected millions of triplet odd-one-out judgments from a multimodal large language model (MLLM) and a language-only model (LLM) in response to 2,180 emotionally evocative videos. We found that the emergent 30-dimensional embeddings from these models are highly interpretable and organize emotion primarily along categorical lines, yet in a blended fashion that incorporates dimensional properties. Most remarkably, the MLLM's representation predicted neural activity in human emotion-processing networks with the highest accuracy, outperforming not only the LLM but also, counterintuitively, representations derived directly from human behavioral ratings. This result supports our primary hypothesis and suggests that sensory grounding--learning from rich visual data--is critical for developing a truly neurally-aligned conceptual framework for emotion. Our findings provide compelling evidence that MLLMs can autonomously develop rich, neurally-aligned affective representations, offering a powerful paradigm to bridge the gap between subjective experience and its neural substrates. Project page: https://reedonepeck.github.io/ai-emotion.github.io/.", "date": "2025-09-29", "categories": []}
{"id": "2509.24322", "title": "Multimodal Large Language Models Meet Multimodal Emotion Recognition and Reasoning: A Survey", "abstract": "In recent years, large language models (LLMs) have driven major advances in language understanding, marking a significant step toward artificial general intelligence (AGI). With increasing demands for higher-level semantics and cross-modal fusion, multimodal large language models (MLLMs) have emerged, integrating diverse information sources (e.g., text, vision, and audio) to enhance modeling and reasoning in complex scenarios. In AI for Science, multimodal emotion recognition and reasoning has become a rapidly growing frontier. While LLMs and MLLMs have achieved notable progress in this area, the field still lacks a systematic review that consolidates recent developments. To address this gap, this paper provides a comprehensive survey of LLMs and MLLMs for emotion recognition and reasoning, covering model architectures, datasets, and performance benchmarks. We further highlight key challenges and outline future research directions, aiming to offer researchers both an authoritative reference and practical insights for advancing this domain. To the best of our knowledge, this paper is the first attempt to comprehensively survey the intersection of MLLMs with multimodal emotion recognition and reasoning. The summary of existing methods mentioned is in our Github: \\href{https://github.com/yuntaoshou/Awesome-Emotion-Reasoning}{https://github.com/yuntaoshou/Awesome-Emotion-Reasoning}.", "date": "2025-09-29", "categories": []}
{"id": "2505.23297", "title": "EmoBench-UA: A Benchmark Dataset for Emotion Detection in Ukrainian", "abstract": "While Ukrainian NLP has seen progress in many texts processing tasks, emotion classification remains an underexplored area with no publicly available benchmark to date. In this work, we introduce EmoBench-UA, the first annotated dataset for emotion detection in Ukrainian texts. Our annotation schema is adapted from the previous English-centric works on emotion detection (Mohammad et al., 2018; Mohammad, 2022) guidelines. The dataset was created through crowdsourcing using the Toloka.ai platform ensuring high-quality of the annotation process. Then, we evaluate a range of approaches on the collected dataset, starting from linguistic-based baselines, synthetic data translated from English, to large language models (LLMs). Our findings highlight the challenges of emotion classification in non-mainstream languages like Ukrainian and emphasize the need for further development of Ukrainian-specific models and training resources.", "date": "2025-05-29", "categories": []}
{"id": "2509.21381", "title": "Toward a Realistic Encoding Model of Auditory Affective Understanding in the Brain", "abstract": "In affective neuroscience and emotion-aware AI, understanding how complex auditory stimuli drive emotion arousal dynamics remains unresolved. This study introduces a computational framework to model the brain's encoding of naturalistic auditory inputs into dynamic behavioral/neural responses across three datasets (SEED, LIRIS, self-collected BAVE). Guided by neurobiological principles of parallel auditory hierarchy, we decompose audio into multilevel auditory features (through classical algorithms and wav2vec 2.0/Hubert) from the original and isolated human voice/background soundtrack elements, mapping them to emotion-related responses via cross-dataset analyses. Our analysis reveals that high-level semantic representations (derived from the final layer of wav2vec 2.0/Hubert) exert a dominant role in emotion encoding, outperforming low-level acoustic features with significantly stronger mappings to behavioral annotations and dynamic neural synchrony across most brain regions ($p < 0.05$). Notably, middle layers of wav2vec 2.0/hubert (balancing acoustic-semantic information) surpass the final layers in emotion induction across datasets. Moreover, human voices and soundtracks show dataset-dependent emotion-evoking biases aligned with stimulus energy distribution (e.g., LIRIS favors soundtracks due to higher background energy), with neural analyses indicating voices dominate prefrontal/temporal activity while soundtracks excel in limbic regions. By integrating affective computing and neuroscience, this work uncovers hierarchical mechanisms of auditory-emotion encoding, providing a foundation for adaptive emotion-aware systems and cross-disciplinary explorations of audio-affective interactions.", "date": "2025-09-23", "categories": []}
{"id": "2509.21747", "title": "Incorporating Scene Context and Semantic Labels for Enhanced Group-level Emotion Recognition", "abstract": "Group-level emotion recognition (GER) aims to identify holistic emotions within a scene involving multiple individuals. Current existed methods underestimate the importance of visual scene contextual information in modeling individual relationships. Furthermore, they overlook the crucial role of semantic information from emotional labels for complete understanding of emotions. To address this limitation, we propose a novel framework that incorporates visual scene context and label-guided semantic information to improve GER performance. It involves the visual context encoding module that leverages multi-scale scene information to diversely encode individual relationships. Complementarily, the emotion semantic encoding module utilizes group-level emotion labels to prompt a large language model to generate nuanced emotion lexicons. These lexicons, in conjunction with the emotion labels, are then subsequently refined into comprehensive semantic representations through the utilization of a structured emotion tree. Finally, similarity-aware interaction is proposed to align and integrate visual and semantic information, thereby generating enhanced group-level emotion representations and subsequently improving the performance of GER. Experiments on three widely adopted GER datasets demonstrate that our proposed method achieves competitive performance compared to state-of-the-art methods.", "date": "2025-09-26", "categories": []}
{"id": "2509.22168", "title": "Teaching AI to Feel: A Collaborative, Full-Body Exploration of Emotive Communication", "abstract": "Commonaiverse is an interactive installation exploring human emotions through full-body motion tracking and real-time AI feedback. Participants engage in three phases: Teaching, Exploration and the Cosmos Phase, collaboratively expressing and interpreting emotions with the system. The installation integrates MoveNet for precise motion tracking and a multi-recommender AI system to analyze emotional states dynamically, responding with adaptive audiovisual outputs. By shifting from top-down emotion classification to participant-driven, culturally diverse definitions, we highlight new pathways for inclusive, ethical affective computing. We discuss how this collaborative, out-of-the-box approach pushes multimedia research beyond single-user facial analysis toward a more embodied, co-created paradigm of emotional AI. Furthermore, we reflect on how this reimagined framework fosters user agency, reduces bias, and opens avenues for advanced interactive applications.", "date": "2025-09-26", "categories": []}
{"id": "2505.10599", "title": "UDDETTS: Unifying Discrete and Dimensional Emotions for Controllable Emotional Text-to-Speech", "abstract": "Recent large language models (LLMs) have made great progress in the field of text-to-speech (TTS), but they still face major challenges in synthesizing fine-grained emotional speech in an interpretable manner. Traditional methods rely on discrete emotion labels to control emotion categories and intensities, which cannot capture the complexity and continuity of human emotional perception and expression. The lack of large-scale emotional speech datasets with balanced emotion distributions and fine-grained emotional annotations often causes overfitting in synthesis models and impedes effective emotion control. To address these issues, we propose UDDETTS, a universal LLM framework unifying discrete and dimensional emotions for controllable emotional TTS. This model introduces the interpretable Arousal-Dominance-Valence (ADV) space for dimensional emotion description and supports emotion control driven by either discrete emotion labels or nonlinearly quantified ADV values. Furthermore, a semi-supervised training strategy is designed to comprehensively utilize diverse speech datasets with different types of emotional annotations to train the UDDETTS. Experiments show that UDDETTS achieves linear emotion control along three interpretable dimensions, and exhibits superior end-to-end emotional speech synthesis capabilities. Code and demos are available at: https://anonymous.4open.science/w/UDDETTS.", "date": "2025-05-15", "categories": []}
{"id": "2506.01047", "title": "CHEER-Ekman: Fine-grained Embodied Emotion Classification", "abstract": "Emotions manifest through physical experiences and bodily reactions, yet identifying such embodied emotions in text remains understudied. We present an embodied emotion classification dataset, CHEER-Ekman, extending the existing binary embodied emotion dataset with Ekman's six basic emotion categories. Using automatic best-worst scaling with large language models, we achieve performance superior to supervised approaches on our new dataset. Our investigation reveals that simplified prompting instructions and chain-of-thought reasoning significantly improve emotion recognition accuracy, enabling smaller models to achieve competitive performance with larger ones. Our dataset is publicly available at: https://github.com/menamerai/cheer-ekman.", "date": "2025-06-01", "categories": []}
{"id": "2508.19993", "title": "MathBuddy: A Multimodal System for Affective Math Tutoring", "abstract": "The rapid adoption of LLM-based conversational systems is already transforming the landscape of educational technology. However, the current state-of-the-art learning models do not take into account the student's affective states. Multiple studies in educational psychology support the claim that positive or negative emotional states can impact a student's learning capabilities. To bridge this gap, we present MathBuddy, an emotionally aware LLM-powered Math Tutor, which dynamically models the student's emotions and maps them to relevant pedagogical strategies, making the tutor-student conversation a more empathetic one. The student's emotions are captured from the conversational text as well as from their facial expressions. The student's emotions are aggregated from both modalities to confidently prompt our LLM Tutor for an emotionally-aware response. We have evaluated our model using automatic evaluation metrics across eight pedagogical dimensions and user studies. We report a massive 23 point performance gain using the win rate and a 3 point gain at an overall level using DAMR scores which strongly supports our hypothesis of improving LLM-based tutor's pedagogical abilities by modeling students' emotions. Our dataset and code are available at: https://github.com/ITU-NLP/MathBuddy .", "date": "2025-08-27", "categories": []}
{"id": "2509.20153", "title": "Affective Computing and Emotional Data: Challenges and Implications in Privacy Regulations, The AI Act, and Ethics in Large Language Models", "abstract": "This paper examines the integration of emotional intelligence into artificial intelligence systems, with a focus on affective computing and the growing capabilities of Large Language Models (LLMs), such as ChatGPT and Claude, to recognize and respond to human emotions. Drawing on interdisciplinary research that combines computer science, psychology, and neuroscience, the study analyzes foundational neural architectures - CNNs for processing facial expressions and RNNs for sequential data, such as speech and text - that enable emotion recognition. It examines the transformation of human emotional experiences into structured emotional data, addressing the distinction between explicit emotional data collected with informed consent in research settings and implicit data gathered passively through everyday digital interactions. That raises critical concerns about lawful processing, AI transparency, and individual autonomy over emotional expressions in digital environments. The paper explores implications across various domains, including healthcare, education, and customer service, while addressing challenges of cultural variations in emotional expression and potential biases in emotion recognition systems across different demographic groups. From a regulatory perspective, the paper examines emotional data in the context of the GDPR and the EU AI Act frameworks, highlighting how emotional data may be considered sensitive personal data that requires robust safeguards, including purpose limitation, data minimization, and meaningful consent mechanisms.", "date": "2025-09-24", "categories": []}
{"id": "2509.20373", "title": "Speaker Style-Aware Phoneme Anchoring for Improved Cross-Lingual Speech Emotion Recognition", "abstract": "Cross-lingual speech emotion recognition (SER) remains a challenging task due to differences in phonetic variability and speaker-specific expressive styles across languages. Effectively capturing emotion under such diverse conditions requires a framework that can align the externalization of emotions across different speakers and languages. To address this problem, we propose a speaker-style aware phoneme anchoring framework that aligns emotional expression at the phonetic and speaker levels. Our method builds emotion-specific speaker communities via graph-based clustering to capture shared speaker traits. Using these groups, we apply dual-space anchoring in speaker and phonetic spaces to enable better emotion transfer across languages. Evaluations on the MSP-Podcast (English) and BIIC-Podcast (Taiwanese Mandarin) corpora demonstrate improved generalization over competitive baselines and provide valuable insights into the commonalities in cross-lingual emotion representation.", "date": "2025-09-19", "categories": []}
{"id": "2509.19595", "title": "Anatomy of a Feeling: Narrating Embodied Emotions via Large Vision-Language Models", "abstract": "The embodiment of emotional reactions from body parts contains rich information about our affective experiences. We propose a framework that utilizes state-of-the-art large vision-language models (LVLMs) to generate Embodied LVLM Emotion Narratives (ELENA). These are well-defined, multi-layered text outputs, primarily comprising descriptions that focus on the salient body parts involved in emotional reactions. We also employ attention maps and observe that contemporary models exhibit a persistent bias towards the facial region. Despite this limitation, we observe that our employed framework can effectively recognize embodied emotions in face-masked images, outperforming baselines without any fine-tuning. ELENA opens a new trajectory for embodied emotion analysis across the modality of vision and enriches modeling in an affect-aware setting.", "date": "2025-09-23", "categories": []}
{"id": "2509.19965", "title": "SynchroRaMa : Lip-Synchronized and Emotion-Aware Talking Face Generation via Multi-Modal Emotion Embedding", "abstract": "Audio-driven talking face generation has received growing interest, particularly for applications requiring expressive and natural human-avatar interaction. However, most existing emotion-aware methods rely on a single modality (either audio or image) for emotion embedding, limiting their ability to capture nuanced affective cues. Additionally, most methods condition on a single reference image, restricting the model's ability to represent dynamic changes in actions or attributes across time. To address these issues, we introduce SynchroRaMa, a novel framework that integrates a multi-modal emotion embedding by combining emotional signals from text (via sentiment analysis) and audio (via speech-based emotion recognition and audio-derived valence-arousal features), enabling the generation of talking face videos with richer and more authentic emotional expressiveness and fidelity. To ensure natural head motion and accurate lip synchronization, SynchroRaMa includes an audio-to-motion (A2M) module that generates motion frames aligned with the input audio. Finally, SynchroRaMa incorporates scene descriptions generated by Large Language Model (LLM) as additional textual input, enabling it to capture dynamic actions and high-level semantic attributes. Conditioning the model on both visual and textual cues enhances temporal consistency and visual realism. Quantitative and qualitative experiments on benchmark datasets demonstrate that SynchroRaMa outperforms the state-of-the-art, achieving improvements in image quality, expression preservation, and motion realism. A user study further confirms that SynchroRaMa achieves higher subjective ratings than competing methods in overall naturalness, motion diversity, and video smoothness. Our project page is available at <https://novicemm.github.io/synchrorama>.", "date": "2025-09-24", "categories": []}
{"id": "2509.11101", "title": "Seeing is Not Understanding: A Benchmark on Perception-Cognition Disparities in Large Language Models", "abstract": "With the rapid advancement of Multimodal Large Language Models (MLLMs), they have demonstrated exceptional capabilities across a variety of vision-language tasks. However, current evaluation benchmarks predominantly focus on objective visual question answering or captioning, inadequately assessing the models' ability to understand complex and subjective human emotions. To bridge this gap, we introduce EmoBench-Reddit, a novel, hierarchical benchmark for multimodal emotion understanding. The dataset comprises 350 meticulously curated samples from the social media platform Reddit, each containing an image, associated user-provided text, and an emotion category (sad, humor, sarcasm, happy) confirmed by user flairs. We designed a hierarchical task framework that progresses from basic perception to advanced cognition, with each data point featuring six multiple-choice questions and one open-ended question of increasing difficulty. Perception tasks evaluate the model's ability to identify basic visual elements (e.g., colors, objects), while cognition tasks require scene reasoning, intent understanding, and deep empathy integrating textual context. We ensured annotation quality through a combination of AI assistance (Claude 4) and manual verification.We conducted a comprehensive evaluation of nine leading MLLMs, including GPT-5, Gemini-2.5-pro, and GPT-4o, on EmoBench-Reddit.", "date": "2025-09-14", "categories": []}
{"id": "2509.18175", "title": "ERFC: Happy Customers with Emotion Recognition and Forecasting in Conversation in Call Centers", "abstract": "Emotion Recognition in Conversation has been seen to be widely applicable in call center analytics, opinion mining, finance, retail, healthcare, and other industries. In a call center scenario, the role of the call center agent is not just confined to receiving calls but to also provide good customer experience by pacifying the frustration or anger of the customers. This can be achieved by maintaining neutral and positive emotion from the agent. As in any conversation, the emotion of one speaker is usually dependent on the emotion of other speaker. Hence the positive emotion of an agent, accompanied with the right resolution will help in enhancing customer experience. This can change an unhappy customer to a happy one. Imparting the right resolution at right time becomes easier if the agent has the insight of the emotion of future utterances. To predict the emotions of the future utterances we propose a novel architecture, Emotion Recognition and Forecasting in Conversation. Our proposed ERFC architecture considers multi modalities, different attributes of emotion, context and the interdependencies of the utterances of the speakers in the conversation. Our intensive experiments on the IEMOCAP dataset have shown the feasibility of the proposed ERFC. This approach can provide a tremendous business value for the applications like call center, where the happiness of customer is utmost important.", "date": "2025-09-17", "categories": []}
{"id": "2509.18550", "title": "HadaSmileNet: Hadamard fusion of handcrafted and deep-learning features for enhancing facial emotion recognition of genuine smiles", "abstract": "The distinction between genuine and posed emotions represents a fundamental pattern recognition challenge with significant implications for data mining applications in social sciences, healthcare, and human-computer interaction. While recent multi-task learning frameworks have shown promise in combining deep learning architectures with handcrafted D-Marker features for smile facial emotion recognition, these approaches exhibit computational inefficiencies due to auxiliary task supervision and complex loss balancing requirements. This paper introduces HadaSmileNet, a novel feature fusion framework that directly integrates transformer-based representations with physiologically grounded D-Markers through parameter-free multiplicative interactions. Through systematic evaluation of 15 fusion strategies, we demonstrate that Hadamard multiplicative fusion achieves optimal performance by enabling direct feature interactions while maintaining computational efficiency. The proposed approach establishes new state-of-the-art results for deep learning methods across four benchmark datasets: UvA-NEMO (88.7 percent, +0.8), MMI (99.7 percent), SPOS (98.5 percent, +0.7), and BBC (100 percent, +5.0). Comprehensive computational analysis reveals 26 percent parameter reduction and simplified training compared to multi-task alternatives, while feature visualization demonstrates enhanced discriminative power through direct domain knowledge integration. The framework's efficiency and effectiveness make it particularly suitable for practical deployment in multimedia data mining applications that require real-time affective computing capabilities.", "date": "2025-09-23", "categories": []}
{"id": "2505.14660", "title": "EmoGist: Efficient In-Context Learning for Visual Emotion Understanding", "abstract": "In this paper, we introduce EmoGist, a training-free, in-context learning method for performing visual emotion classification with LVLMs. The key intuition of our approach is that context-dependent definition of emotion labels could allow more accurate predictions of emotions, as the ways in which emotions manifest within images are highly context dependent and nuanced. EmoGist pre-generates multiple descriptions of emotion labels, by analyzing the clusters of example images belonging to each label. At test time, we retrieve a version of description based on the cosine similarity of test image to cluster centroids, and feed it together with the test image to a fast LVLM for classification. Through our experiments, we show that EmoGist allows up to 12 points improvement in micro F1 scores with the multi-label Memotion dataset, and up to 8 points in macro F1 in the multi-class FI dataset.", "date": "2025-05-20", "categories": []}
{"id": "2509.16623", "title": "CGTGait: Collaborative Graph and Transformer for Gait Emotion Recognition", "abstract": "Skeleton-based gait emotion recognition has received significant attention due to its wide-ranging applications. However, existing methods primarily focus on extracting spatial and local temporal motion information, failing to capture long-range temporal representations. In this paper, we propose \\textbf{CGTGait}, a novel framework that collaboratively integrates graph convolution and transformers to extract discriminative spatiotemporal features for gait emotion recognition. Specifically, CGTGait consists of multiple CGT blocks, where each block employs graph convolution to capture frame-level spatial topology and the transformer to model global temporal dependencies. Additionally, we introduce a Bidirectional Cross-Stream Fusion (BCSF) module to effectively aggregate posture and motion spatiotemporal features, facilitating the exchange of complementary information between the two streams. We evaluate our method on two widely used datasets, Emotion-Gait and ELMD, demonstrating that our CGTGait achieves state-of-the-art or at least competitive performance while reducing computational complexity by approximately \\textbf{82.2\\%} (only requiring 0.34G FLOPs) during testing. Code is available at \\small{https://github.com/githubzjj1/CGTGait.}", "date": "2025-09-20", "categories": []}
{"id": "2509.17711", "title": "DA-Mamba: Dialogue-aware selective state-space model for multimodal engagement estimation", "abstract": "Human engagement estimation in conversational scenarios is essential for applications such as adaptive tutoring, remote healthcare assessment, and socially aware human--computer interaction. Engagement is a dynamic, multimodal signal conveyed by facial expressions, speech, gestures, and behavioral cues over time. In this work we introduce DA-Mamba, a dialogue-aware multimodal architecture that replaces attention-heavy dialogue encoders with Mamba-based selective state-space processing to achieve linear time and memory complexity while retaining expressive cross-modal reasoning. We design a Mamba dialogue-aware selective state-space model composed of three core modules: a Dialogue-Aware Encoder, and two Mamba-based fusion mechanisms: Modality-Group Fusion and Partner-Group Fusion, these modules achieve expressive dialogue understanding. Extensive experiments on three standard benchmarks (NoXi, NoXi-Add, and MPIIGI) show that DA-Mamba surpasses prior state-of-the-art (SOTA) methods in concordance correlation coefficient (CCC), while reducing training time and peak memory; these gains enable processing much longer sequences and facilitate real-time deployment in resource-constrained, multi-party conversational settings. The source code will be available at: https://github.com/kksssssss-ssda/MMEA.", "date": "2025-09-22", "categories": []}
{"id": "2501.16870", "title": "Experimenting with Affective Computing Models in Video Interviews with Spanish-speaking Older Adults", "abstract": "Understanding emotional signals in older adults is crucial for designing virtual assistants that support their well-being. However, existing affective computing models often face significant limitations: (1) limited availability of datasets representing older adults, especially in non-English-speaking populations, and (2) poor generalization of models trained on younger or homogeneous demographics. To address these gaps, this study evaluates state-of-the-art affective computing models -- including facial expression recognition, text sentiment analysis, and smile detection -- using videos of older adults interacting with either a person or a virtual avatar. As part of this effort, we introduce a novel dataset featuring Spanish-speaking older adults engaged in human-to-human video interviews. Through three comprehensive analyses, we investigate (1) the alignment between human-annotated labels and automatic model outputs, (2) the relationships between model outputs across different modalities, and (3) individual variations in emotional signals. Using both the Wizard of Oz (WoZ) dataset and our newly collected dataset, we uncover limited agreement between human annotations and model predictions, weak consistency across modalities, and significant variability among individuals. These findings highlight the shortcomings of generalized emotion perception models and emphasize the need of incorporating personal variability and cultural nuances into future systems.", "date": "2025-01-28", "categories": []}
{"id": "2509.15986", "title": "EmoHeal: An End-to-End System for Personalized Therapeutic Music Retrieval from Fine-grained Emotions", "abstract": "Existing digital mental wellness tools often overlook the nuanced emotional states underlying everyday challenges. For example, pre-sleep anxiety affects more than 1.5 billion people worldwide, yet current approaches remain largely static and \"one-size-fits-all\", failing to adapt to individual needs. In this work, we present EmoHeal, an end-to-end system that delivers personalized, three-stage supportive narratives. EmoHeal detects 27 fine-grained emotions from user text with a fine-tuned XLM-RoBERTa model, mapping them to musical parameters via a knowledge graph grounded in music therapy principles (GEMS, iso-principle). EmoHeal retrieves audiovisual content using the CLAMP3 model to guide users from their current state toward a calmer one (\"match-guide-target\"). A within-subjects study (N=40) demonstrated significant supportive effects, with participants reporting substantial mood improvement (M=4.12, p<0.001) and high perceived emotion recognition accuracy (M=4.05, p<0.001). A strong correlation between perceived accuracy and therapeutic outcome (r=0.72, p<0.001) validates our fine-grained approach. These findings establish the viability of theory-driven, emotion-aware digital wellness tools and provides a scalable AI blueprint for operationalizing music therapy principles.", "date": "2025-09-19", "categories": []}
{"id": "2504.03643", "title": "Potential Indicator for Continuous Emotion Arousal by Dynamic Neural Synchrony", "abstract": "The need for automatic and high-quality emotion annotation is paramount in applications such as continuous emotion recognition and video highlight detection, yet achieving this through manual human annotations is challenging. Inspired by inter-subject correlation (ISC) utilized in neuroscience, this study introduces a novel Electroencephalography (EEG) based ISC methodology that leverages a single-electrode and feature-based dynamic approach. Our contributions are three folds. Firstly, we reidentify two potent emotion features suitable for classifying emotions-first-order difference (FD) an differential entropy (DE). Secondly, through the use of overall correlation analysis, we demonstrate the heterogeneous synchronized performance of electrodes. This performance aligns with neural emotion patterns established in prior studies, thus validating the effectiveness of our approach. Thirdly, by employing a sliding window correlation technique, we showcase the significant consistency of dynamic ISCs across various features or key electrodes in each analyzed film clip. Our findings indicate the method's reliability in capturing consistent, dynamic shared neural synchrony among individuals, triggered by evocative film stimuli. This underscores the potential of our approach to serve as an indicator of continuous human emotion arousal. The implications of this research are significant for advancements in affective computing and the broader neuroscience field, suggesting a streamlined and effective tool for emotion analysis in real-world applications.", "date": "2025-01-24", "categories": []}
{"id": "2509.14611", "title": "Leveraging IndoBERT and DistilBERT for Indonesian Emotion Classification in E-Commerce Reviews", "abstract": "Understanding emotions in the Indonesian language is essential for improving customer experiences in e-commerce. This study focuses on enhancing the accuracy of emotion classification in Indonesian by leveraging advanced language models, IndoBERT and DistilBERT. A key component of our approach was data processing, specifically data augmentation, which included techniques such as back-translation and synonym replacement. These methods played a significant role in boosting the model's performance. After hyperparameter tuning, IndoBERT achieved an accuracy of 80\\%, demonstrating the impact of careful data processing. While combining multiple IndoBERT models led to a slight improvement, it did not significantly enhance performance. Our findings indicate that IndoBERT was the most effective model for emotion classification in Indonesian, with data augmentation proving to be a vital factor in achieving high accuracy. Future research should focus on exploring alternative architectures and strategies to improve generalization for Indonesian NLP tasks.", "date": "2025-09-18", "categories": []}
{"id": "2508.18655", "title": "Empathy Omni: Enabling Empathetic Speech Response Generation through Large Language Models", "abstract": "With the development of speech large language models (speech LLMs), users can now interact directly with assistants via speech. However, most existing models only convert response content into speech without fully capturing the rich emotional cues in user queries, where the same sentence may convey different meanings depending on the expression. Emotional understanding is thus essential for improving human-machine interaction. Most empathetic speech LLMs rely on massive datasets, demanding high computational cost. A key challenge is to build models that generate empathetic responses with limited data and without large-scale training. To this end, we propose Emotion Omni, a model that understands emotional content in user speech and generates empathetic responses. We further developed a data pipeline to construct a 200k emotional dialogue dataset supporting empathetic speech assistants. Experiments show that Emotion Omni achieves comparable instruction-following ability without large-scale pretraining, while surpassing existing models in speech quality (UTMOS:4.41) and empathy (Emotion GPT Score: 3.97). These results confirm its improvements in both speech fidelity and emotional expressiveness. Demos are available at https://w311411.github.io/omni_demo/.", "date": "2025-08-26", "categories": []}
{"id": "2509.13725", "title": "WatchAnxiety: A Transfer Learning Approach for State Anxiety Prediction from Smartwatch Data", "abstract": "Social anxiety is a common mental health condition linked to significant challenges in academic, social, and occupational functioning. A core feature is elevated momentary (state) anxiety in social situations, yet little prior work has measured or predicted fluctuations in this anxiety throughout the day. Capturing these intra-day dynamics is critical for designing real-time, personalized interventions such as Just-In-Time Adaptive Interventions (JITAIs). To address this gap, we conducted a study with socially anxious college students (N=91; 72 after exclusions) using our custom smartwatch-based system over an average of 9.03 days (SD = 2.95). Participants received seven ecological momentary assessments (EMAs) per day to report state anxiety. We developed a base model on over 10,000 days of external heart rate data, transferred its representations to our dataset, and fine-tuned it to generate probabilistic predictions. These were combined with trait-level measures in a meta-learner. Our pipeline achieved 60.4% balanced accuracy in state anxiety detection in our dataset. To evaluate generalizability, we applied the training approach to a separate hold-out set from the TILES-18 dataset-the same dataset used for pretraining. On 10,095 once-daily EMAs, our method achieved 59.1% balanced accuracy, outperforming prior work by at least 7%.", "date": "2025-09-17", "categories": []}
{"id": "2509.10847", "title": "A funny companion: Distinct neural responses to perceived AI- versus human-generated humor", "abstract": "As AI companions become capable of human-like communication, including telling jokes, understanding how people cognitively and emotionally respond to AI humor becomes increasingly important. This study used electroencephalography (EEG) to compare how people process humor from AI versus human sources. Behavioral analysis revealed that participants rated AI and human humor as comparably funny. However, neurophysiological data showed that AI humor elicited a smaller N400 effect, suggesting reduced cognitive effort during the processing of incongruity. This was accompanied by a larger Late Positive Potential (LPP), indicating a greater degree of surprise and emotional response. This enhanced LPP likely stems from the violation of low initial expectations regarding AI's comedic capabilities. Furthermore, a key temporal dynamic emerged: human humor showed habituation effects, marked by an increasing N400 and a decreasing LPP over time. In contrast, AI humor demonstrated increasing processing efficiency and emotional reward, with a decreasing N400 and an increasing LPP. This trajectory reveals how the brain can dynamically update its predictive model of AI capabilities. This process of cumulative reinforcement challenges \"algorithm aversion\" in humor, as it demonstrates how cognitive adaptation to AI's language patterns can lead to an intensified emotional reward. Additionally, participants' social attitudes toward AI modulated these neural responses, with higher perceived AI trustworthiness correlating with enhanced emotional engagement. These findings indicate that the brain responds to AI humor with surprisingly positive and intense reactions, highlighting humor's potential for fostering genuine engagement in human-AI social interaction.", "date": "2025-09-13", "categories": []}
{"id": "2509.12295", "title": "More Similar than Dissimilar: Modeling Annotators for Cross-Corpus Speech Emotion Recognition", "abstract": "Speech emotion recognition systems often predict a consensus value generated from the ratings of multiple annotators. However, these models have limited ability to predict the annotation of any one person. Alternatively, models can learn to predict the annotations of all annotators. Adapting such models to new annotators is difficult as new annotators must individually provide sufficient labeled training data. We propose to leverage inter-annotator similarity by using a model pre-trained on a large annotator population to identify a similar, previously seen annotator. Given a new, previously unseen, annotator and limited enrollment data, we can make predictions for a similar annotator, enabling off-the-shelf annotation of unseen data in target datasets, providing a mechanism for extremely low-cost personalization. We demonstrate our approach significantly outperforms other off-the-shelf approaches, paving the way for lightweight emotion adaptation, practical for real-world deployment.", "date": "2025-09-15", "categories": []}
{"id": "2408.11956", "title": "The Whole Is Bigger Than the Sum of Its Parts: Modeling Individual Annotators to Capture Emotional Variability", "abstract": "Emotion expression and perception are nuanced, complex, and highly subjective processes. When multiple annotators label emotional data, the resulting labels contain high variability. Most speech emotion recognition tasks address this by averaging annotator labels as ground truth. However, this process omits the nuance of emotion and inter-annotator variability, which are important signals to capture. Previous work has attempted to learn distributions to capture emotion variability, but these methods also lose information about the individual annotators. We address these limitations by learning to predict individual annotators and by introducing a novel method to create distributions from continuous model outputs that permit the learning of emotion distributions during model training. We show that this combined approach can result in emotion distributions that are more accurate than those seen in prior work, in both within- and cross-corpus settings.", "date": "2024-08-21", "categories": []}
{"id": "2507.13380", "title": "Persona-Based Synthetic Data Generation Using Multi-Stage Conditioning with Large Language Models for Emotion Recognition", "abstract": "In the field of emotion recognition, the development of high-performance models remains a challenge due to the scarcity of high-quality, diverse emotional datasets. Emotional expressions are inherently subjective, shaped by individual personality traits, socio-cultural backgrounds, and contextual factors, making large-scale, generalizable data collection both ethically and practically difficult. To address this issue, we introduce PersonaGen, a novel framework for generating emotionally rich text using a Large Language Model (LLM) through multi-stage persona-based conditioning. PersonaGen constructs layered virtual personas by combining demographic attributes, socio-cultural backgrounds, and detailed situational contexts, which are then used to guide emotion expression generation. We conduct comprehensive evaluations of the generated synthetic data, assessing semantic diversity through clustering and distributional metrics, human-likeness via LLM-based quality scoring, realism through comparison with real-world emotion corpora, and practical utility in downstream emotion classification tasks. Experimental results show that PersonaGen significantly outperforms baseline methods in generating diverse, coherent, and discriminative emotion expressions, demonstrating its potential as a robust alternative for augmenting or replacing real-world emotional datasets.", "date": "2025-07-15", "categories": []}
{"id": "2502.19860", "title": "MIND: Towards Immersive Psychological Healing with Multi-agent Inner Dialogue", "abstract": "Mental health issues are worsening in today's competitive society, such as depression and anxiety. Traditional healings like counseling and chatbots fail to engage effectively, they often provide generic responses lacking emotional depth. Although large language models (LLMs) have the potential to create more human-like interactions, they still struggle to capture subtle emotions. This requires LLMs to be equipped with human-like adaptability and warmth. To fill this gap, we propose the MIND (Multi-agent INner Dialogue), a novel paradigm that provides more immersive psychological healing environments. Considering the strong generative and role-playing ability of LLM agents, we predefine an interactive healing framework and assign LLM agents different roles within the framework to engage in interactive inner dialogues with users, thereby providing an immersive healing experience. We conduct extensive human experiments in various real-world healing dimensions, and find that MIND provides a more user-friendly experience than traditional paradigms. This demonstrates that MIND effectively leverages the significant potential of LLMs in psychological healing.", "date": "2025-02-27", "categories": []}
{"id": "2509.06015", "title": "Micro-Expression Recognition via Fine-Grained Dynamic Perception", "abstract": "Facial micro-expression recognition (MER) is a challenging task, due to the transience, subtlety, and dynamics of micro-expressions (MEs). Most existing methods resort to hand-crafted features or deep networks, in which the former often additionally requires key frames, and the latter suffers from small-scale and low-diversity training data. In this paper, we develop a novel fine-grained dynamic perception (FDP) framework for MER. We propose to rank frame-level features of a sequence of raw frames in chronological order, in which the rank process encodes the dynamic information of both ME appearances and motions. Specifically, a novel local-global feature-aware transformer is proposed for frame representation learning. A rank scorer is further adopted to calculate rank scores of each frame-level feature. Afterwards, the rank features from rank scorer are pooled in temporal dimension to capture dynamic representation. Finally, the dynamic representation is shared by a MER module and a dynamic image construction module, in which the former predicts the ME category, and the latter uses an encoder-decoder structure to construct the dynamic image. The design of dynamic image construction task is beneficial for capturing facial subtle actions associated with MEs and alleviating the data scarcity issue. Extensive experiments show that our method (i) significantly outperforms the state-of-the-art MER methods, and (ii) works well for dynamic image construction. Particularly, our FDP improves by 4.05%, 2.50%, 7.71%, and 2.11% over the previous best results in terms of F1-score on the CASME II, SAMM, CAS(ME)^2, and CAS(ME)^3 datasets, respectively. The code is available at https://github.com/CYF-cuber/FDP.", "date": "2025-09-07", "categories": []}
{"id": "2509.09593", "title": "Fluent but Unfeeling: The Emotional Blind Spots of Language Models", "abstract": "The versatility of Large Language Models (LLMs) in natural language understanding has made them increasingly popular in mental health research. While many studies explore LLMs' capabilities in emotion recognition, a critical gap remains in evaluating whether LLMs align with human emotions at a fine-grained level. Existing research typically focuses on classifying emotions into predefined, limited categories, overlooking more nuanced expressions. To address this gap, we introduce EXPRESS, a benchmark dataset curated from Reddit communities featuring 251 fine-grained, self-disclosed emotion labels. Our comprehensive evaluation framework examines predicted emotion terms and decomposes them into eight basic emotions using established emotion theories, enabling a fine-grained comparison. Systematic testing of prevalent LLMs under various prompt settings reveals that accurately predicting emotions that align with human self-disclosed emotions remains challenging. Qualitative analysis further shows that while certain LLMs generate emotion terms consistent with established emotion theories and definitions, they sometimes fail to capture contextual cues as effectively as human self-disclosures. These findings highlight the limitations of LLMs in fine-grained emotion alignment and offer insights for future research aimed at enhancing their contextual understanding.", "date": "2025-09-11", "categories": []}
{"id": "2505.20511", "title": "Multimodal Emotion Recognition in Conversations: A Survey of Methods, Trends, Challenges and Prospects", "abstract": "While text-based emotion recognition methods have achieved notable success, real-world dialogue systems often demand a more nuanced emotional understanding than any single modality can offer. Multimodal Emotion Recognition in Conversations (MERC) has thus emerged as a crucial direction for enhancing the naturalness and emotional understanding of human-computer interaction. Its goal is to accurately recognize emotions by integrating information from various modalities such as text, speech, and visual signals.\n  This survey offers a systematic overview of MERC, including its motivations, core tasks, representative methods, and evaluation strategies. We further examine recent trends, highlight key challenges, and outline future directions. As interest in emotionally intelligent systems grows, this survey provides timely guidance for advancing MERC research.", "date": "2025-05-26", "categories": []}
{"id": "2408.04638", "title": "Affective Computing in the Era of Large Language Models: A Survey from the NLP Perspective", "abstract": "Affective Computing (AC) integrates computer science, psychology, and cognitive science to enable machines to recognize, interpret, and simulate human emotions across domains such as social media, finance, healthcare, and education. AC commonly centers on two task families: Affective Understanding (AU) and Affective Generation (AG). While fine-tuned pre-trained language models (PLMs) have achieved solid AU performance, they often generalize poorly across tasks and remain limited for AG, especially in producing diverse, emotionally appropriate responses. The advent of Large Language Models (LLMs) (e.g., ChatGPT and LLaMA) has catalyzed a paradigm shift by offering in-context learning, broader world knowledge, and stronger sequence generation. This survey presents an NLP-oriented overview of AC in the LLM era. We (i) consolidate traditional AC tasks and preliminary LLM-based studies; (ii) review adaptation techniques that improve AU/AG, including Instruction Tuning (full and parameter-efficient methods such as LoRA, P-/Prompt-Tuning), Prompt Engineering (zero/few-shot, chain-of-thought, agent-based prompting), and Reinforcement Learning. For the latter, we summarize RL from human preferences (RLHF), verifiable/programmatic rewards (RLVR), and AI feedback (RLAIF), which provide preference- or rule-grounded optimization signals that can help steer AU/AG toward empathy, safety, and planning, achieving finer-grained or multi-objective control. To assess progress, we compile benchmarks and evaluation practices for both AU and AG. We also discuss open challenges-from ethics, data quality, and safety to robust evaluation and resource efficiency-and outline research directions. We hope this survey clarifies the landscape and offers practical guidance for building affect-aware, reliable, and responsible LLM systems.", "date": "2024-07-30", "categories": []}
{"id": "2509.02100", "title": "E-THER: A Multimodal Dataset for Empathic AI -- Towards Emotional Mismatch Awareness", "abstract": "A prevalent shortfall among current empathic AI systems is their inability to recognize when verbal expressions may not fully reflect underlying emotional states. This is because the existing datasets, used for the training of these systems, focus on surface-level emotion recognition without addressing the complex verbal-visual incongruence (mismatch) patterns useful for empathic understanding. In this paper, we present E-THER, the first Person-Centered Therapy-grounded multimodal dataset with multidimensional annotations for verbal-visual incongruence detection, enabling training of AI systems that develop genuine rather than performative empathic capabilities. The annotations included in the dataset are drawn from humanistic approach, i.e., identifying verbal-visual emotional misalignment in client-counsellor interactions - forming a framework for training and evaluating AI on empathy tasks. Additional engagement scores provide behavioral annotations for research applications. Notable gains in empathic and therapeutic conversational qualities are observed in state-of-the-art vision-language models (VLMs), such as IDEFICS and VideoLLAVA, using evaluation metrics grounded in empathic and therapeutic principles. Empirical findings indicate that our incongruence-trained models outperform general-purpose models in critical traits, such as sustaining therapeutic engagement, minimizing artificial or exaggerated linguistic patterns, and maintaining fidelity to PCT theoretical framework.", "date": "2025-09-02", "categories": []}
{"id": "2509.05298", "title": "Livia: An Emotion-Aware AR Companion Powered by Modular AI Agents and Progressive Memory Compression", "abstract": "Loneliness and social isolation pose significant emotional and health challenges, prompting the development of technology-based solutions for companionship and emotional support. This paper introduces Livia, an emotion-aware augmented reality (AR) companion app designed to provide personalized emotional support by combining modular artificial intelligence (AI) agents, multimodal affective computing, progressive memory compression, and AR driven embodied interaction. Livia employs a modular AI architecture with specialized agents responsible for emotion analysis, dialogue generation, memory management, and behavioral orchestration, ensuring robust and adaptive interactions. Two novel algorithms-Temporal Binary Compression (TBC) and Dynamic Importance Memory Filter (DIMF)-effectively manage and prioritize long-term memory, significantly reducing storage requirements while retaining critical context. Our multimodal emotion detection approach achieves high accuracy, enhancing proactive and empathetic engagement. User evaluations demonstrated increased emotional bonds, improved satisfaction, and statistically significant reductions in loneliness. Users particularly valued Livia's adaptive personality evolution and realistic AR embodiment. Future research directions include expanding gesture and tactile interactions, supporting multi-user experiences, and exploring customized hardware implementations.", "date": "2025-08-12", "categories": []}
{"id": "2509.05330", "title": "MVRS: The Multimodal Virtual Reality Stimuli-based Emotion Recognition Dataset", "abstract": "Automatic emotion recognition has become increasingly important with the rise of AI, especially in fields like healthcare, education, and automotive systems. However, there is a lack of multimodal datasets, particularly involving body motion and physiological signals, which limits progress in the field. To address this, the MVRS dataset is introduced, featuring synchronized recordings from 13 participants aged 12 to 60 exposed to VR based emotional stimuli (relaxation, fear, stress, sadness, joy). Data were collected using eye tracking (via webcam in a VR headset), body motion (Kinect v2), and EMG and GSR signals (Arduino UNO), all timestamp aligned. Participants followed a unified protocol with consent and questionnaires. Features from each modality were extracted, fused using early and late fusion techniques, and evaluated with classifiers to confirm the datasets quality and emotion separability, making MVRS a valuable contribution to multimodal affective computing.", "date": "2025-08-31", "categories": []}
{"id": "2509.05617", "title": "From Joy to Fear: A Benchmark of Emotion Estimation in Pop Song Lyrics", "abstract": "The emotional content of song lyrics plays a pivotal role in shaping listener experiences and influencing musical preferences. This paper investigates the task of multi-label emotional attribution of song lyrics by predicting six emotional intensity scores corresponding to six fundamental emotions. A manually labeled dataset is constructed using a mean opinion score (MOS) approach, which aggregates annotations from multiple human raters to ensure reliable ground-truth labels. Leveraging this dataset, we conduct a comprehensive evaluation of several publicly available large language models (LLMs) under zero-shot scenarios. Additionally, we fine-tune a BERT-based model specifically for predicting multi-label emotion scores. Experimental results reveal the relative strengths and limitations of zero-shot and fine-tuned models in capturing the nuanced emotional content of lyrics. Our findings highlight the potential of LLMs for emotion recognition in creative texts, providing insights into model selection strategies for emotion-based music information retrieval applications. The labeled dataset is available at https://github.com/LLM-HITCS25S/LyricsEmotionAttribution.", "date": "2025-09-06", "categories": []}
{"id": "2509.05634", "title": "On the Contribution of Lexical Features to Speech Emotion Recognition", "abstract": "Although paralinguistic cues are often considered the primary drivers of speech emotion recognition (SER), we investigate the role of lexical content extracted from speech and show that it can achieve competitive and in some cases higher performance compared to acoustic models. On the MELD dataset, our lexical-based approach obtains a weighted F1-score (WF1) of 51.5%, compared to 49.3% for an acoustic-only pipeline with a larger parameter count. Furthermore, we analyze different self-supervised (SSL) speech and text representations, conduct a layer-wise study of transformer-based encoders, and evaluate the effect of audio denoising.", "date": "2025-09-06", "categories": []}
{"id": "2509.04480", "title": "Discrete Prompt Tuning via Recursive Utilization of Black-box Multimodal Large Language Model for Personalized Visual Emotion Recognition", "abstract": "Visual Emotion Recognition (VER) is an important research topic due to its wide range of applications, including opinion mining and advertisement design. Extending this capability to recognize emotions at the individual level further broadens its potential applications. Recently, Multimodal Large Language Models (MLLMs) have attracted increasing attention and demonstrated performance comparable to that of conventional VER methods. However, MLLMs are trained on large and diverse datasets containing general opinions, which causes them to favor majority viewpoints and familiar patterns. This tendency limits their performance in a personalized VER, which is crucial for practical and real-world applications, and indicates a key area for improvement. To address this limitation, the proposed method employs discrete prompt tuning inspired by the process of humans' prompt engineering to adapt the VER task to each individual. Our method selects the best natural language representation from the generated prompts and uses it to update the prompt for the realization of accurate personalized VER.", "date": "2025-08-30", "categories": []}
{"id": "2506.09556", "title": "MEDUSA: A Multimodal Deep Fusion Multi-Stage Training Framework for Speech Emotion Recognition in Naturalistic Conditions", "abstract": "SER is a challenging task due to the subjective nature of human emotions and their uneven representation under naturalistic conditions. We propose MEDUSA, a multimodal framework with a four-stage training pipeline, which effectively handles class imbalance and emotion ambiguity. The first two stages train an ensemble of classifiers that utilize DeepSER, a novel extension of a deep cross-modal transformer fusion mechanism from pretrained self-supervised acoustic and linguistic representations. Manifold MixUp is employed for further regularization. The last two stages optimize a trainable meta-classifier that combines the ensemble predictions. Our training approach incorporates human annotation scores as soft targets, coupled with balanced data sampling and multitask learning. MEDUSA ranked 1st in Task 1: Categorical Emotion Recognition in the Interspeech 2025: Speech Emotion Recognition in Naturalistic Conditions Challenge.", "date": "2025-06-11", "categories": []}
{"id": "2506.18201", "title": "Deciphering Emotions in Children Storybooks: A Comparative Analysis of Multimodal LLMs in Educational Applications", "abstract": "Emotion recognition capabilities in multimodal AI systems are crucial for developing culturally responsive educational technologies, yet remain underexplored for Arabic language contexts where culturally appropriate learning tools are critically needed. This study evaluates the emotion recognition performance of two advanced multimodal large language models, GPT-4o and Gemini 1.5 Pro, when processing Arabic children's storybook illustrations. We assessed both models across three prompting strategies (zero-shot, few-shot, and chain-of-thought) using 75 images from seven Arabic storybooks, comparing model predictions with human annotations based on Plutchik's emotional framework. GPT-4o consistently outperformed Gemini across all conditions, achieving the highest macro F1-score of 59% with chain-of-thought prompting compared to Gemini's best performance of 43%. Error analysis revealed systematic misclassification patterns, with valence inversions accounting for 60.7% of errors, while both models struggled with culturally nuanced emotions and ambiguous narrative contexts. These findings highlight fundamental limitations in current models' cultural understanding and emphasize the need for culturally sensitive training approaches to develop effective emotion-aware educational technologies for Arabic-speaking learners.", "date": "2025-06-22", "categories": []}
{"id": "2509.00813", "title": "AImoclips: A Benchmark for Evaluating Emotion Conveyance in Text-to-Music Generation", "abstract": "Recent advances in text-to-music (TTM) generation have enabled controllable and expressive music creation using natural language prompts. However, the emotional fidelity of TTM systems remains largely underexplored compared to human preference or text alignment. In this study, we introduce AImoclips, a benchmark for evaluating how well TTM systems convey intended emotions to human listeners, covering both open-source and commercial models. We selected 12 emotion intents spanning four quadrants of the valence-arousal space, and used six state-of-the-art TTM systems to generate over 1,000 music clips. A total of 111 participants rated the perceived valence and arousal of each clip on a 9-point Likert scale. Our results show that commercial systems tend to produce music perceived as more pleasant than intended, while open-source systems tend to perform the opposite. Emotions are more accurately conveyed under high-arousal conditions across all models. Additionally, all systems exhibit a bias toward emotional neutrality, highlighting a key limitation in affective controllability. This benchmark offers valuable insights into model-specific emotion rendering characteristics and supports future development of emotionally aligned TTM systems.", "date": "2025-08-31", "categories": []}
{"id": "2509.03212", "title": "AIVA: An AI-based Virtual Companion for Emotion-aware Interaction", "abstract": "Recent advances in Large Language Models (LLMs) have significantly improved natural language understanding and generation, enhancing Human-Computer Interaction (HCI). However, LLMs are limited to unimodal text processing and lack the ability to interpret emotional cues from non-verbal signals, hindering more immersive and empathetic interactions. This work explores integrating multimodal sentiment perception into LLMs to create emotion-aware agents. We propose \\ours, an AI-based virtual companion that captures multimodal sentiment cues, enabling emotionally aligned and animated HCI. \\ours introduces a Multimodal Sentiment Perception Network (MSPN) using a cross-modal fusion transformer and supervised contrastive learning to provide emotional cues. Additionally, we develop an emotion-aware prompt engineering strategy for generating empathetic responses and integrate a Text-to-Speech (TTS) system and animated avatar module for expressive interactions. \\ours provides a framework for emotion-aware agents with applications in companion robotics, social care, mental health, and human-centered AI.", "date": "2025-09-03", "categories": []}
{"id": "2502.10435", "title": "RAMer: Reconstruction-based Adversarial Model for Multi-party Multi-modal Multi-label Emotion Recognition", "abstract": "Conventional Multi-modal multi-label emotion recognition (MMER) assumes complete access to visual, textual, and acoustic modalities. However, real-world multi-party settings often violate this assumption, as non-speakers frequently lack acoustic and textual inputs, leading to a significant degradation in model performance. Existing approaches also tend to unify heterogeneous modalities into a single representation, overlooking each modality's unique characteristics. To address these challenges, we propose RAMer (Reconstruction-based Adversarial Model for Emotion Recognition), which refines multi-modal representations by not only exploring modality commonality and specificity but crucially by leveraging reconstructed features, enhanced by contrastive learning, to overcome data incompleteness and enrich feature quality. RAMer also introduces a personality auxiliary task to complement missing modalities using modality-level attention, improving emotion reasoning. To further strengthen the model's ability to capture label and modality interdependency, we propose a stack shuffle strategy to enrich correlations between labels and modality-specific features. Experiments on three benchmarks, i.e., MEmoR, CMU-MOSEI, and $M^3ED$, demonstrate that RAMer achieves state-of-the-art performance in dyadic and multi-party MMER scenarios.", "date": "2025-02-09", "categories": []}
{"id": "2509.00025", "title": "DeepEmoNet: Building Machine Learning Models for Automatic Emotion Recognition in Human Speeches", "abstract": "Speech emotion recognition (SER) has been a challenging problem in spoken language processing research, because it is unclear how human emotions are connected to various components of sounds such as pitch, loudness, and energy. This paper aims to tackle this problem using machine learning. Particularly, we built several machine learning models using SVMs, LTSMs, and CNNs to classify emotions in human speeches. In addition, by leveraging transfer learning and data augmentation, we efficiently trained our models to attain decent performances on a relatively small dataset. Our best model was a ResNet34 network, which achieved an accuracy of $66.7\\%$ and an F1 score of $0.631$.", "date": "2025-08-20", "categories": []}
{"id": "2509.00077", "title": "Amplifying Emotional Signals: Data-Efficient Deep Learning for Robust Speech Emotion Recognition", "abstract": "Speech Emotion Recognition (SER) presents a significant yet persistent challenge in human-computer interaction. While deep learning has advanced spoken language processing, achieving high performance on limited datasets remains a critical hurdle. This paper confronts this issue by developing and evaluating a suite of machine learning models, including Support Vector Machines (SVMs), Long Short-Term Memory networks (LSTMs), and Convolutional Neural Networks (CNNs), for automated emotion classification in human speech. We demonstrate that by strategically employing transfer learning and innovative data augmentation techniques, our models can achieve impressive performance despite the constraints of a relatively small dataset. Our most effective model, a ResNet34 architecture, establishes a new performance benchmark on the combined RAVDESS and SAVEE datasets, attaining an accuracy of 66.7% and an F1 score of 0.631. These results underscore the substantial benefits of leveraging pre-trained models and data augmentation to overcome data scarcity, thereby paving the way for more robust and generalizable SER systems.", "date": "2025-08-26", "categories": []}
{"id": "2509.01260", "title": "Annotation and modeling of emotions in a textual corpus: an evaluative approach", "abstract": "Emotion is a crucial phenomenon in the functioning of human beings in society. However, it remains a widely open subject, particularly in its textual manifestations. This paper examines an industrial corpus manually annotated following an evaluative approach to emotion. This theoretical framework, which is currently underutilized, offers a different perspective that complements traditional approaches. Noting that the annotations we collected exhibit significant disagreement, we hypothesized that they nonetheless follow stable statistical trends. Using language models trained on these annotations, we demonstrate that it is possible to model the labeling process and that variability is driven by underlying linguistic features. Conversely, our results indicate that language models seem capable of distinguishing emotional situations based on evaluative criteria.", "date": "2025-09-01", "categories": []}
{"id": "2509.01401", "title": "ArabEmoNet: A Lightweight Hybrid 2D CNN-BiLSTM Model with Attention for Robust Arabic Speech Emotion Recognition", "abstract": "Speech emotion recognition is vital for human-computer interaction, particularly for low-resource languages like Arabic, which face challenges due to limited data and research. We introduce ArabEmoNet, a lightweight architecture designed to overcome these limitations and deliver state-of-the-art performance. Unlike previous systems relying on discrete MFCC features and 1D convolutions, which miss nuanced spectro-temporal patterns, ArabEmoNet uses Mel spectrograms processed through 2D convolutions, preserving critical emotional cues often lost in traditional methods.\n  While recent models favor large-scale architectures with millions of parameters, ArabEmoNet achieves superior results with just 1 million parameters, 90 times smaller than HuBERT base and 74 times smaller than Whisper. This efficiency makes it ideal for resource-constrained environments. ArabEmoNet advances Arabic speech emotion recognition, offering exceptional performance and accessibility for real-world applications.", "date": "2025-09-01", "categories": []}
{"id": "2506.00863", "title": "L3Cube-MahaEmotions: A Marathi Emotion Recognition Dataset with Synthetic Annotations using CoTR prompting and Large Language Models", "abstract": "Emotion recognition in low-resource languages like Marathi remains challenging due to limited annotated data. We present L3Cube-MahaEmotions, a high-quality Marathi emotion recognition dataset with 11 fine-grained emotion labels. The training data is synthetically annotated using large language models (LLMs), while the validation and test sets are manually labeled to serve as a reliable gold-standard benchmark. Building on the MahaSent dataset, we apply the Chain-of-Translation (CoTR) prompting technique, where Marathi sentences are translated into English and emotion labeled via a single prompt. GPT-4 and Llama3-405B were evaluated, with GPT-4 selected for training data annotation due to superior label quality. We evaluate model performance using standard metrics and explore label aggregation strategies (e.g., Union, Intersection). While GPT-4 predictions outperform fine-tuned BERT models, BERT-based models trained on synthetic labels fail to surpass GPT-4. This highlights both the importance of high-quality human-labeled data and the inherent complexity of emotion recognition. An important finding of this work is that generic LLMs like GPT-4 and Llama3-405B generalize better than fine-tuned BERT for complex low-resource emotion recognition tasks. The dataset and model are shared publicly at https://github.com/l3cube-pune/MarathiNLP", "date": "2025-06-01", "categories": []}
{"id": "2508.20796", "title": "Speech Emotion Recognition via Entropy-Aware Score Selection", "abstract": "In this paper, we propose a multimodal framework for speech emotion recognition that leverages entropy-aware score selection to combine speech and textual predictions. The proposed method integrates a primary pipeline that consists of an acoustic model based on wav2vec2.0 and a secondary pipeline that consists of a sentiment analysis model using RoBERTa-XLM, with transcriptions generated via Whisper-large-v3. We propose a late score fusion approach based on entropy and varentropy thresholds to overcome the confidence constraints of primary pipeline predictions. A sentiment mapping strategy translates three sentiment categories into four target emotion classes, enabling coherent integration of multimodal predictions. The results on the IEMOCAP and MSP-IMPROV datasets show that the proposed method offers a practical and reliable enhancement over traditional single-modality systems.", "date": "2025-08-28", "categories": []}
{"id": "2508.20805", "title": "Exploring Machine Learning and Language Models for Multimodal Depression Detection", "abstract": "This paper presents our approach to the first Multimodal Personality-Aware Depression Detection Challenge, focusing on multimodal depression detection using machine learning and deep learning models. We explore and compare the performance of XGBoost, transformer-based architectures, and large language models (LLMs) on audio, video, and text features. Our results highlight the strengths and limitations of each type of model in capturing depression-related signals across modalities, offering insights into effective multimodal representation strategies for mental health prediction.", "date": "2025-08-28", "categories": []}
{"id": "2508.19193", "title": "Emotions as Ambiguity-aware Ordinal Representations", "abstract": "Emotions are inherently ambiguous and dynamic phenomena, yet existing continuous emotion recognition approaches either ignore their ambiguity or treat ambiguity as an independent and static variable over time. Motivated by this gap in the literature, in this paper we introduce ambiguity-aware ordinal emotion representations, a novel framework that captures both the ambiguity present in emotion annotation and the inherent temporal dynamics of emotional traces. Specifically, we propose approaches that model emotion ambiguity through its rate of change. We evaluate our framework on two affective corpora -- RECOLA and GameVibe -- testing our proposed approaches on both bounded (arousal, valence) and unbounded (engagement) continuous traces. Our results demonstrate that ordinal representations outperform conventional ambiguity-aware models on unbounded labels, achieving the highest Concordance Correlation Coefficient (CCC) and Signed Differential Agreement (SDA) scores, highlighting their effectiveness in modeling the traces' dynamics. For bounded traces, ordinal representations excel in SDA, revealing their superior ability to capture relative changes of annotated emotion traces.", "date": "2025-08-26", "categories": []}
{"id": "2508.19533", "title": "Emotion Transfer with Enhanced Prototype for Unseen Emotion Recognition in Conversation", "abstract": "Current Emotion Recognition in Conversation (ERC) research follows a closed-domain assumption. However, there is no clear consensus on emotion classification in psychology, which presents a challenge for models when it comes to recognizing previously unseen emotions in real-world applications. To bridge this gap, we introduce the Unseen Emotion Recognition in Conversation (UERC) task for the first time and propose ProEmoTrans, a solid prototype-based emotion transfer framework. This prototype-based approach shows promise but still faces key challenges: First, implicit expressions complicate emotion definition, which we address by proposing an LLM-enhanced description approach. Second, utterance encoding in long conversations is difficult, which we tackle with a proposed parameter-free mechanism for efficient encoding and overfitting prevention. Finally, the Markovian flow nature of emotions is hard to transfer, which we address with an improved Attention Viterbi Decoding (AVD) method to transfer seen emotion transitions to unseen emotions. Extensive experiments on three datasets show that our method serves as a strong baseline for preliminary exploration in this new area.", "date": "2025-08-27", "categories": []}
{"id": "2505.22863", "title": "Large Language Models for Depression Recognition in Spoken Language Integrating Psychological Knowledge", "abstract": "Depression is a growing concern gaining attention in both public discourse and AI research. While deep neural networks (DNNs) have been used for recognition, they still lack real-world effectiveness. Large language models (LLMs) show strong potential but require domain-specific fine-tuning and struggle with non-textual cues. Since depression is often expressed through vocal tone and behaviour rather than explicit text, relying on language alone is insufficient. Diagnostic accuracy also suffers without incorporating psychological expertise. To address these limitations, we present, to the best of our knowledge, the first application of LLMs to multimodal depression detection using the DAIC-WOZ dataset. We extract the audio features using the pre-trained model Wav2Vec, and mapped it to text-based LLMs for further processing. We also propose a novel strategy for incorporating psychological knowledge into LLMs to enhance diagnostic performance, specifically using a question and answer set to grant authorised knowledge to LLMs. Our approach yields a notable improvement in both Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) compared to a base score proposed by the related original paper. The codes are available at https://github.com/myxp-lyp/Depression-detection.git", "date": "2025-05-28", "categories": []}
{"id": "2508.17623", "title": "EMO-Reasoning: Benchmarking Emotional Reasoning Capabilities in Spoken Dialogue Systems", "abstract": "Speech emotions play a crucial role in human-computer interaction, shaping engagement and context-aware communication. Despite recent advances in spoken dialogue systems, a holistic system for evaluating emotional reasoning is still lacking. To address this, we introduce EMO-Reasoning, a benchmark for assessing emotional coherence in dialogue systems. It leverages a curated dataset generated via text-to-speech to simulate diverse emotional states, overcoming the scarcity of emotional speech data. We further propose the Cross-turn Emotion Reasoning Score to assess the emotion transitions in multi-turn dialogues. Evaluating seven dialogue systems through continuous, categorical, and perceptual metrics, we show that our framework effectively detects emotional inconsistencies, providing insights for improving current dialogue systems. By releasing a systematic evaluation benchmark, we aim to advance emotion-aware spoken dialogue modeling toward more natural and adaptive interactions.", "date": "2025-08-25", "categories": []}
{"id": "2508.18322", "title": "Structures Meet Semantics: Multimodal Fusion via Graph Contrastive Learning", "abstract": "Multimodal sentiment analysis (MSA) aims to infer emotional states by effectively integrating textual, acoustic, and visual modalities. Despite notable progress, existing multimodal fusion methods often neglect modality-specific structural dependencies and semantic misalignment, limiting their quality, interpretability, and robustness. To address these challenges, we propose a novel framework called the Structural-Semantic Unifier (SSU), which systematically integrates modality-specific structural information and cross-modal semantic grounding for enhanced multimodal representations. Specifically, SSU dynamically constructs modality-specific graphs by leveraging linguistic syntax for text and a lightweight, text-guided attention mechanism for acoustic and visual modalities, thus capturing detailed intra-modal relationships and semantic interactions. We further introduce a semantic anchor, derived from global textual semantics, that serves as a cross-modal alignment hub, effectively harmonizing heterogeneous semantic spaces across modalities. Additionally, we develop a multiview contrastive learning objective that promotes discriminability, semantic consistency, and structural coherence across intra- and inter-modal views. Extensive evaluations on two widely used benchmark datasets, CMU-MOSI and CMU-MOSEI, demonstrate that SSU consistently achieves state-of-the-art performance while significantly reducing computational overhead compared to prior methods. Comprehensive qualitative analyses further validate SSU's interpretability and its ability to capture nuanced emotional patterns through semantically grounded interactions.", "date": "2025-08-24", "categories": []}
{"id": "2508.18740", "title": "M3HG: Multimodal, Multi-scale, and Multi-type Node Heterogeneous Graph for Emotion Cause Triplet Extraction in Conversations", "abstract": "Emotion Cause Triplet Extraction in Multimodal Conversations (MECTEC) has recently gained significant attention in social media analysis, aiming to extract emotion utterances, cause utterances, and emotion categories simultaneously. However, the scarcity of related datasets, with only one published dataset featuring highly uniform dialogue scenarios, hinders model development in this field. To address this, we introduce MECAD, the first multimodal, multi-scenario MECTEC dataset, comprising 989 conversations from 56 TV series spanning a wide range of dialogue contexts. In addition, existing MECTEC methods fail to explicitly model emotional and causal contexts and neglect the fusion of semantic information at different levels, leading to performance degradation. In this paper, we propose M3HG, a novel model that explicitly captures emotional and causal contexts and effectively fuses contextual information at both inter- and intra-utterance levels via a multimodal heterogeneous graph. Extensive experiments demonstrate the effectiveness of M3HG compared with existing state-of-the-art methods. The codes and dataset are available at https://github.com/redifinition/M3HG.", "date": "2025-08-26", "categories": []}
{"id": "2508.18782", "title": "Long-Term Variability in Physiological-Arousal Relationships for Robust Emotion Estimation", "abstract": "Estimating emotional states from physiological signals is a central topic in affective computing and psychophysiology. While many emotion estimation systems implicitly assume a stable relationship between physiological features and subjective affect, this assumption has rarely been tested over long timeframes. This study investigates whether such relationships remain consistent across several months within individuals. We developed a custom measurement system and constructed a longitudinal dataset by collecting physiological signals -- including blood volume pulse, electrodermal activity (EDA), skin temperature, and acceleration--along with self-reported emotional states from 24 participants over two three-month periods. Data were collected in naturalistic working environments, allowing analysis of the relationship between physiological features and subjective arousal in everyday contexts. We examined how physiological-arousal relationships evolve over time by using Explainable Boosting Machines (EBMs) to ensure model interpretability. A model trained on 1st-period data showed a 5\\% decrease in accuracy when tested on 2nd-period data, indicating long-term variability in physiological-arousal associations. EBM-based comparisons further revealed that while heart rate remained a relatively stable predictor, minimum EDA exhibited substantial individual-level fluctuations between periods. While the number of participants is limited, these findings highlight the need to account for temporal variability in physiological-arousal relationships and suggest that emotion estimation models should be periodically updated -- e.g., every five months -- based on observed shift trends to maintain robust performance over time.", "date": "2025-08-26", "categories": []}
{"id": "2508.16859", "title": "Beyond Emotion Recognition: A Multi-Turn Multimodal Emotion Understanding and Reasoning Benchmark", "abstract": "Multimodal large language models (MLLMs) have been widely applied across various fields due to their powerful perceptual and reasoning capabilities. In the realm of psychology, these models hold promise for a deeper understanding of human emotions and behaviors. However, recent research primarily focuses on enhancing their emotion recognition abilities, leaving the substantial potential in emotion reasoning, which is crucial for improving the naturalness and effectiveness of human-machine interactions. Therefore, in this paper, we introduce a multi-turn multimodal emotion understanding and reasoning (MTMEUR) benchmark, which encompasses 1,451 video data from real-life scenarios, along with 5,101 progressive questions. These questions cover various aspects, including emotion recognition, potential causes of emotions, future action prediction, etc. Besides, we propose a multi-agent framework, where each agent specializes in a specific aspect, such as background context, character dynamics, and event details, to improve the system's reasoning capabilities. Furthermore, we conduct experiments with existing MLLMs and our agent-based method on the proposed benchmark, revealing that most models face significant challenges with this task.", "date": "2025-08-23", "categories": []}
{"id": "2508.18108", "title": "SentiMM: A Multimodal Multi-Agent Framework for Sentiment Analysis in Social Media", "abstract": "With the increasing prevalence of multimodal content on social media, sentiment analysis faces significant challenges in effectively processing heterogeneous data and recognizing multi-label emotions. Existing methods often lack effective cross-modal fusion and external knowledge integration. We propose SentiMM, a novel multi-agent framework designed to systematically address these challenges. SentiMM processes text and visual inputs through specialized agents, fuses multimodal features, enriches context via knowledge retrieval, and aggregates results for final sentiment classification. We also introduce SentiMMD, a large-scale multimodal dataset with seven fine-grained sentiment categories. Extensive experiments demonstrate that SentiMM achieves superior performance compared to state-of-the-art baselines, validating the effectiveness of our structured approach.", "date": "2025-08-25", "categories": []}
{"id": "2505.01068", "title": "Multimodal Transformers are Hierarchical Modal-wise Heterogeneous Graphs", "abstract": "Multimodal Sentiment Analysis (MSA) is a rapidly developing field that integrates multimodal information to recognize sentiments, and existing models have made significant progress in this area. The central challenge in MSA is multimodal fusion, which is predominantly addressed by Multimodal Transformers (MulTs). Although act as the paradigm, MulTs suffer from efficiency concerns. In this work, from the perspective of efficiency optimization, we propose and prove that MulTs are hierarchical modal-wise heterogeneous graphs (HMHGs), and we introduce the graph-structured representation pattern of MulTs. Based on this pattern, we propose an Interlaced Mask (IM) mechanism to design the Graph-Structured and Interlaced-Masked Multimodal Transformer (GsiT). It is formally equivalent to MulTs which achieves an efficient weight-sharing mechanism without information disorder through IM, enabling All-Modal-In-One fusion with only 1/3 of the parameters of pure MulTs. A Triton kernel called Decomposition is implemented to ensure avoiding additional computational overhead. Moreover, it achieves significantly higher performance than traditional MulTs. To further validate the effectiveness of GsiT itself and the HMHG concept, we integrate them into multiple state-of-the-art models and demonstrate notable performance improvements and parameter reduction on widely used MSA datasets.", "date": "2025-05-02", "categories": []}
{"id": "2508.15852", "title": "PGF-Net: A Progressive Gated-Fusion Framework for Efficient Multimodal Sentiment Analysis", "abstract": "We introduce PGF-Net (Progressive Gated-Fusion Network), a novel deep learning framework designed for efficient and interpretable multimodal sentiment analysis. Our framework incorporates three primary innovations. Firstly, we propose a Progressive Intra-Layer Fusion paradigm, where a Cross-Attention mechanism empowers the textual representation to dynamically query and integrate non-linguistic features from audio and visual streams within the deep layers of a Transformer encoder. This enables a deeper, context-dependent fusion process. Secondly, the model incorporates an Adaptive Gated Arbitration mechanism, which acts as a dynamic controller to balance the original linguistic information against the newly fused multimodal context, ensuring stable and meaningful integration while preventing noise from overwhelming the signal. Lastly, a hybrid Parameter-Efficient Fine-Tuning (PEFT) strategy is employed, synergistically combining global adaptation via LoRA with local refinement through Post-Fusion Adapters. This significantly reduces trainable parameters, making the model lightweight and suitable for resource-limited scenarios. These innovations are integrated into a hierarchical encoder architecture, enabling PGF-Net to perform deep, dynamic, and interpretable multimodal sentiment analysis while maintaining exceptional parameter efficiency. Experimental results on MOSI dataset demonstrate that our proposed PGF-Net achieves state-of-the-art performance, with a Mean Absolute Error (MAE) of 0.691 and an F1-Score of 86.9%. Notably, our model achieves these results with only 3.09M trainable parameters, showcasing a superior balance between performance and computational efficiency.", "date": "2025-08-20", "categories": []}
{"id": "2508.14920", "title": "Human Feedback Driven Dynamic Speech Emotion Recognition", "abstract": "This work proposes to explore a new area of dynamic speech emotion recognition. Unlike traditional methods, we assume that each audio track is associated with a sequence of emotions active at different moments in time. The study particularly focuses on the animation of emotional 3D avatars. We propose a multi-stage method that includes the training of a classical speech emotion recognition model, synthetic generation of emotional sequences, and further model improvement based on human feedback. Additionally, we introduce a novel approach to modeling emotional mixtures based on the Dirichlet distribution. The models are evaluated based on ground-truth emotions extracted from a dataset of 3D facial animations. We compare our models against the sliding window approach. Our experimental results show the effectiveness of Dirichlet-based approach in modeling emotional mixtures. Incorporating human feedback further improves the model quality while providing a simplified annotation procedure.", "date": "2025-08-18", "categories": []}
{"id": "2409.16126", "title": "VisioPhysioENet: Visual Physiological Engagement Detection Network", "abstract": "This paper presents VisioPhysioENet, a novel multimodal system that leverages visual and physiological signals to detect learner engagement. It employs a two-level approach for extracting both visual and physiological features. For visual feature extraction, Dlib is used to detect facial landmarks, while OpenCV provides additional estimations. The face recognition library, built on Dlib, is used to identify the facial region of interest specifically for physiological signal extraction. Physiological signals are then extracted using the plane-orthogonal-toskin method to assess cardiovascular activity. These features are integrated using advanced machine learning classifiers, enhancing the detection of various levels of engagement. We thoroughly tested VisioPhysioENet on the DAiSEE dataset. It achieved an accuracy of 63.09%. This shows it can better identify different levels of engagement compared to many existing methods. It performed 8.6% better than the only other model that uses both physiological and visual features.", "date": "2024-09-24", "categories": []}
{"id": "2508.14130", "title": "EmoSLLM: Parameter-Efficient Adaptation of LLMs for Speech Emotion Recognition", "abstract": "Emotion recognition from speech is a challenging task that requires capturing both linguistic and paralinguistic cues, with critical applications in human-computer interaction and mental health monitoring. Recent works have highlighted the ability of Large Language Models (LLMs) to perform tasks outside of the sole natural language area. In particular, recent approaches have investigated coupling LLMs with other data modalities by using pre-trained backbones and different fusion mechanisms. This work proposes a novel approach that fine-tunes an LLM with audio and text representations for emotion prediction. Our method first extracts audio features using an audio feature extractor, which are then mapped into the LLM's representation space via a learnable interfacing module. The LLM takes as input (1) the transformed audio features, (2) additional features in the form of natural language (e.g., the transcript), and (3) a textual prompt describing the emotion prediction task. To efficiently adapt the LLM to this multimodal task, we employ Low-Rank Adaptation (LoRA), enabling parameter-efficient fine-tuning. Experimental results on standard emotion recognition benchmarks demonstrate that our model outperforms all but one existing Speech-Text LLMs in the literature, while requiring less than half the parameters of competing approaches. This highlights our approach's effectiveness in integrating multi-modal inputs for speech-based emotion understanding while maintaining significant computational efficiency.", "date": "2025-08-19", "categories": []}
{"id": "2508.14214", "title": "Large Language Models are Highly Aligned with Human Ratings of Emotional Stimuli", "abstract": "Emotions exert an immense influence over human behavior and cognition in both commonplace and high-stress tasks. Discussions of whether or how to integrate large language models (LLMs) into everyday life (e.g., acting as proxies for, or interacting with, human agents), should be informed by an understanding of how these tools evaluate emotionally loaded stimuli or situations. A model's alignment with human behavior in these cases can inform the effectiveness of LLMs for certain roles or interactions. To help build this understanding, we elicited ratings from multiple popular LLMs for datasets of words and images that were previously rated for their emotional content by humans. We found that when performing the same rating tasks, GPT-4o responded very similarly to human participants across modalities, stimuli and most rating scales (r = 0.9 or higher in many cases). However, arousal ratings were less well aligned between human and LLM raters, while happiness ratings were most highly aligned. Overall LLMs aligned better within a five-category (happiness, anger, sadness, fear, disgust) emotion framework than within a two-dimensional (arousal and valence) organization. Finally, LLM ratings were substantially more homogenous than human ratings. Together these results begin to describe how LLM agents interpret emotional stimuli and highlight similarities and differences among biological and artificial intelligence in key behavioral domains.", "date": "2025-08-19", "categories": []}
{"id": "2504.14658", "title": "EmoSEM: Segment and Explain Emotion Stimuli in Visual Art", "abstract": "This paper focuses on a key challenge in visual emotion understanding: given an art image, the model pinpoints pixel regions that trigger a specific human emotion, and generates linguistic explanations for it. Despite advances in general segmentation, pixel-level emotion understanding still faces a dual challenge: first, the subjectivity of emotion limits general segmentation models like SAM to adapt to emotion-oriented segmentation tasks; and second, the abstract nature of art expression makes it hard for captioning models to balance pixel-level semantics and emotion reasoning. To solve the above problems, this paper proposes the Emotion stimuli Segmentation and Explanation Model (EmoSEM) model to endow the segmentation framework with emotion comprehension capability. First, to enable the model to perform segmentation under the guidance of emotional intent well, we introduce an emotional prompt with a learnable mask token as the conditional input for segmentation decoding. Then, we design an emotion projector to establish the association between emotion and visual features. Next, more importantly, to address emotion-visual stimuli alignment, we develop a lightweight prefix adapter, a module that fuses the learned emotional mask with the corresponding emotion into a unified representation compatible with the language model. Finally, we input the joint visual, mask, and emotional tokens into the language model and output the emotional explanations. It ensures that the generated interpretations remain semantically and emotionally coherent with the visual stimuli. Our method realizes end-to-end modeling from low-level pixel features to high-level emotion interpretation, delivering the first interpretable fine-grained framework for visual emotion analysis. Extensive experiments validate the effectiveness of our model. Code will be made publicly available.", "date": "2025-04-20", "categories": []}
{"id": "2506.13971", "title": "Multimodal Fusion with Semi-Supervised Learning Minimizes Annotation Quantity for Modeling Videoconference Conversation Experience", "abstract": "Group conversations over videoconferencing are a complex social behavior. However, the subjective moments of negative experience, where the conversation loses fluidity or enjoyment remain understudied. These moments are infrequent in naturalistic data, and thus training a supervised learning (SL) model requires costly manual data annotation. We applied semi-supervised learning (SSL) to leverage targeted labeled and unlabeled clips for training multimodal (audio, facial, text) deep features to predict non-fluid or unenjoyable moments in holdout videoconference sessions. The modality-fused co-training SSL achieved an ROC-AUC of 0.9 and an F1 score of 0.6, outperforming SL models by up to 4% with the same amount of labeled data. Remarkably, the best SSL model with just 8% labeled data matched 96% of the SL model's full-data performance. This shows an annotation-efficient framework for modeling videoconference experience.", "date": "2025-06-01", "categories": []}
{"id": "2506.22807", "title": "FreqDGT: Frequency-Adaptive Dynamic Graph Networks with Transformer for Cross-subject EEG Emotion Recognition", "abstract": "Electroencephalography (EEG) serves as a reliable and objective signal for emotion recognition in affective brain-computer interfaces, offering unique advantages through its high temporal resolution and ability to capture authentic emotional states that cannot be consciously controlled. However, cross-subject generalization remains a fundamental challenge due to individual variability, cognitive traits, and emotional responses. We propose FreqDGT, a frequency-adaptive dynamic graph transformer that systematically addresses these limitations through an integrated framework. FreqDGT introduces frequency-adaptive processing (FAP) to dynamically weight emotion-relevant frequency bands based on neuroscientific evidence, employs adaptive dynamic graph learning (ADGL) to learn input-specific brain connectivity patterns, and implements multi-scale temporal disentanglement network (MTDN) that combines hierarchical temporal transformers with adversarial feature disentanglement to capture both temporal dynamics and ensure cross-subject robustness. Comprehensive experiments demonstrate that FreqDGT significantly improves cross-subject emotion recognition accuracy, confirming the effectiveness of integrating frequency-adaptive, spatial-dynamic, and temporal-hierarchical modeling while ensuring robustness to individual differences. The code is available at https://github.com/NZWANG/FreqDGT.", "date": "2025-06-28", "categories": []}
{"id": "2507.14698", "title": "Spatial-Temporal Transformer with Curriculum Learning for EEG-Based Emotion Recognition", "abstract": "EEG-based emotion recognition plays an important role in developing adaptive brain-computer communication systems, yet faces two fundamental challenges in practical implementations: (1) effective integration of non-stationary spatial-temporal neural patterns, (2) robust adaptation to dynamic emotional intensity variations in real-world scenarios. This paper proposes SST-CL, a novel framework integrating spatial-temporal transformers with curriculum learning. Our method introduces two core components: a spatial encoder that models inter-channel relationships and a temporal encoder that captures multi-scale dependencies through windowed attention mechanisms, enabling simultaneous extraction of spatial correlations and temporal dynamics from EEG signals. Complementing this architecture, an intensity-aware curriculum learning strategy progressively guides training from high-intensity to low-intensity emotional states through dynamic sample scheduling based on a dual difficulty assessment. Comprehensive experiments on three benchmark datasets demonstrate state-of-the-art performance across various emotional intensity levels, with ablation studies confirming the necessity of both architectural components and the curriculum learning mechanism.", "date": "2025-07-19", "categories": []}
{"id": "2508.13524", "title": "Evaluating Open-Source Vision Language Models for Facial Emotion Recognition against Traditional Deep Learning Models", "abstract": "Facial Emotion Recognition (FER) is crucial for applications such as human-computer interaction and mental health diagnostics. This study presents the first empirical comparison of open-source Vision-Language Models (VLMs), including Phi-3.5 Vision and CLIP, against traditional deep learning models VGG19, ResNet-50, and EfficientNet-B0 on the challenging FER-2013 dataset, which contains 35,887 low-resolution grayscale images across seven emotion classes. To address the mismatch between VLM training assumptions and the noisy nature of FER data, we introduce a novel pipeline that integrates GFPGAN-based image restoration with FER evaluation. Results show that traditional models, particularly EfficientNet-B0 (86.44%) and ResNet-50 (85.72%), significantly outperform VLMs like CLIP (64.07%) and Phi-3.5 Vision (51.66%), highlighting the limitations of VLMs in low-quality visual tasks. In addition to performance evaluation using precision, recall, F1-score, and accuracy, we provide a detailed computational cost analysis covering preprocessing, training, inference, and evaluation phases, offering practical insights for deployment. This work underscores the need for adapting VLMs to noisy environments and provides a reproducible benchmark for future research in emotion recognition.", "date": "2025-08-19", "categories": []}
{"id": "2507.15714", "title": "Chinchunmei at SemEval-2025 Task 11: Boosting the Large Language Model's Capability of Emotion Perception using Contrastive Learning", "abstract": "The SemEval-2025 Task 11, Bridging the Gap in Text-Based Emotion Detection, introduces an emotion recognition challenge spanning over 28 languages. This competition encourages researchers to explore more advanced approaches to address the challenges posed by the diversity of emotional expressions and background variations. It features two tracks: multi-label classification (Track A) and emotion intensity prediction (Track B), covering six emotion categories: anger, fear, joy, sadness, surprise, and disgust. In our work, we systematically explore the benefits of two contrastive learning approaches: sample-based (Contrastive Reasoning Calibration) and generation-based (DPO, SimPO) contrastive learning. The sample-based contrastive approach trains the model by comparing two samples to generate more reliable predictions. The generation-based contrastive approach trains the model to differentiate between correct and incorrect generations, refining its prediction. All models are fine-tuned from LLaMa3-Instruct-8B. Our system achieves 9th place in Track A and 6th place in Track B for English, while ranking among the top-tier performing systems for other languages.", "date": "2025-07-21", "categories": []}
{"id": "2508.11889", "title": "In-Context Examples Matter: Improving Emotion Recognition in Conversation with Instruction Tuning", "abstract": "Emotion recognition in conversation (ERC) aims to identify the emotion of each utterance in a conversation, playing a vital role in empathetic artificial intelligence. With the growing of large language models (LLMs), instruction tuning has emerged as a critical paradigm for ERC. Existing studies mainly focus on multi-stage instruction tuning, which first endows LLMs with speaker characteristics, and then conducts context-aware instruction tuning to comprehend emotional states. However, these methods inherently constrains the capacity to jointly capture the dynamic interaction between speaker characteristics and conversational context, resulting in weak alignment among speaker identity, contextual cues, and emotion states within a unified framework. In this paper, we propose InitERC, a simple yet effective one-stage in-context instruction tuning framework for ERC. InitERC adapts LLMs to learn speaker-context-emotion alignment from context examples via in-context instruction tuning. Specifically, InitERC comprises four components, i.e., demonstration pool construction, in-context example selection, prompt template design, and in-context instruction tuning. To explore the impact of in-context examples, we conduct a comprehensive study on three key factors: retrieval strategy, example ordering, and the number of examples. Extensive experiments on three widely used datasets demonstrate that our proposed InitERC achieves substantial improvements over the state-of-the-art baselines.", "date": "2025-08-16", "categories": []}
{"id": "2508.12669", "title": "Leveraging Large Language Models for Predictive Analysis of Human Misery", "abstract": "This study investigates the use of Large Language Models (LLMs) for predicting human-perceived misery scores from natural language descriptions of real-world scenarios. The task is framed as a regression problem, where the model assigns a scalar value from 0 to 100 to each input statement. We evaluate multiple prompting strategies, including zero-shot, fixed-context few-shot, and retrieval-based prompting using BERT sentence embeddings. Few-shot approaches consistently outperform zero-shot baselines, underscoring the value of contextual examples in affective prediction. To move beyond static evaluation, we introduce the \"Misery Game Show\", a novel gamified framework inspired by a television format. It tests LLMs through structured rounds involving ordinal comparison, binary classification, scalar estimation, and feedback-driven reasoning. This setup enables us to assess not only predictive accuracy but also the model's ability to adapt based on corrective feedback. The gamified evaluation highlights the broader potential of LLMs in dynamic emotional reasoning tasks beyond standard regression. Code and data link: https://github.com/abhi1nandy2/Misery_Data_Exps_GitHub", "date": "2025-08-18", "categories": []}
{"id": "2508.12742", "title": "On the Importance of Behavioral Nuances: Amplifying Non-Obvious Motor Noise Under True Empirical Considerations May Lead to Briefer Assays and Faster Classification Processes", "abstract": "There is a tradeoff between attaining statistical power with large, difficult to gather data sets, and producing highly scalable assays that register brief data samples. Often, as grand-averaging techniques a priori assume normally-distributed parameters and linear, stationary processes in biorhythmic, time series data, important information is lost, averaged out as gross data. We developed an affective computing platform that enables taking brief data samples while maintaining personalized statistical power. This is achieved by combining a new data type derived from the micropeaks present in time series data registered from brief (5-second-long) face videos with recent advances in AI-driven face-grid estimation methods. By adopting geometric and nonlinear dynamical systems approaches to analyze the kinematics, especially the speed data, the new methods capture all facial micropeaks. These include as well the nuances of different affective micro expressions. We offer new ways to differentiate dynamical and geometric patterns present in autistic individuals from those found more commonly in neurotypical development.", "date": "2025-08-18", "categories": []}
{"id": "2508.12854", "title": "E3RG: Building Explicit Emotion-driven Empathetic Response Generation System with Multimodal Large Language Model", "abstract": "Multimodal Empathetic Response Generation (MERG) is crucial for building emotionally intelligent human-computer interactions. Although large language models (LLMs) have improved text-based ERG, challenges remain in handling multimodal emotional content and maintaining identity consistency. Thus, we propose E3RG, an Explicit Emotion-driven Empathetic Response Generation System based on multimodal LLMs which decomposes MERG task into three parts: multimodal empathy understanding, empathy memory retrieval, and multimodal response generation. By integrating advanced expressive speech and video generative models, E3RG delivers natural, emotionally rich, and identity-consistent responses without extra training. Experiments validate the superiority of our system on both zero-shot and few-shot settings, securing Top-1 position in the Avatar-based Multimodal Empathy Challenge on ACM MM 25. Our code is available at https://github.com/RH-Lin/E3RG.", "date": "2025-08-18", "categories": []}
{"id": "2508.06800", "title": "Hardness-Aware Dynamic Curriculum Learning for Robust Multimodal Emotion Recognition with Missing Modalities", "abstract": "Missing modalities have recently emerged as a critical research direction in multimodal emotion recognition (MER). Conventional approaches typically address this issue through missing modality reconstruction. However, these methods fail to account for variations in reconstruction difficulty across different samples, consequently limiting the model's ability to handle hard samples effectively. To overcome this limitation, we propose a novel Hardness-Aware Dynamic Curriculum Learning framework, termed HARDY-MER. Our framework operates in two key stages: first, it estimates the hardness level of each sample, and second, it strategically emphasizes hard samples during training to enhance model performance on these challenging instances. Specifically, we first introduce a Multi-view Hardness Evaluation mechanism that quantifies reconstruction difficulty by considering both Direct Hardness (modality reconstruction errors) and Indirect Hardness (cross-modal mutual information). Meanwhile, we introduce a Retrieval-based Dynamic Curriculum Learning strategy that dynamically adjusts the training curriculum by retrieving samples with similar semantic information and balancing the learning focus between easy and hard instances. Extensive experiments on benchmark datasets demonstrate that HARDY-MER consistently outperforms existing methods in missing-modality scenarios. Our code will be made publicly available at https://github.com/HARDY-MER/HARDY-MER.", "date": "2025-08-09", "categories": []}
{"id": "2508.10561", "title": "Reproducible Physiological Features in Affective Computing: A Preliminary Analysis on Arousal Modeling", "abstract": "In Affective Computing, a key challenge lies in reliably linking subjective emotional experiences with objective physiological markers. This preliminary study addresses the issue of reproducibility by identifying physiological features from cardiovascular and electrodermal signals that are associated with continuous self-reports of arousal levels. Using the Continuously Annotated Signal of Emotion dataset, we analyzed 164 features extracted from cardiac and electrodermal signals of 30 participants exposed to short emotion-evoking videos. Feature selection was performed using the Terminating-Random Experiments (T-Rex) method, which performs variable selection systematically controlling a user-defined target False Discovery Rate. Remarkably, among all candidate features, only two electrodermal-derived features exhibited reproducible and statistically significant associations with arousal, achieving a 100\\% confirmation rate. These results highlight the necessity of rigorous reproducibility assessments in physiological features selection, an aspect often overlooked in Affective Computing. Our approach is particularly promising for applications in safety-critical environments requiring trustworthy and reliable white box models, such as mental disorder recognition and human-robot interaction systems.", "date": "2025-08-14", "categories": []}
{"id": "2508.10828", "title": "A Multimodal Neural Network for Recognizing Subjective Self-Disclosure Towards Social Robots", "abstract": "Subjective self-disclosure is an important feature of human social interaction. While much has been done in the social and behavioural literature to characterise the features and consequences of subjective self-disclosure, little work has been done thus far to develop computational systems that are able to accurately model it. Even less work has been done that attempts to model specifically how human interactants self-disclose with robotic partners. It is becoming more pressing as we require social robots to work in conjunction with and establish relationships with humans in various social settings. In this paper, our aim is to develop a custom multimodal attention network based on models from the emotion recognition literature, training this model on a large self-collected self-disclosure video corpus, and constructing a new loss function, the scale preserving cross entropy loss, that improves upon both classification and regression versions of this problem. Our results show that the best performing model, trained with our novel loss function, achieves an F1 score of 0.83, an improvement of 0.48 from the best baseline model. This result makes significant headway in the aim of allowing social robots to pick up on an interaction partner's self-disclosures, an ability that will be essential in social robots with social cognition.", "date": "2025-08-14", "categories": []}
{"id": "2412.05296", "title": "Revisiting Your Memory: Reconstruction of Affect-Contextualized Memory via EEG-guided Audiovisual Generation", "abstract": "In this paper, we introduce RevisitAffectiveMemory, a novel task designed to reconstruct autobiographical memories through audio-visual generation guided by affect extracted from electroencephalogram (EEG) signals. To support this pioneering task, we present the EEG-AffectiveMemory dataset, which encompasses textual descriptions, visuals, music, and EEG recordings collected during memory recall from nine participants. Furthermore, we propose RYM (Revisit Your Memory), a three-stage framework for generating synchronized audio-visual contents while maintaining dynamic personal memory affect trajectories. Experimental results demonstrate our method successfully decodes individual affect dynamics trajectories from neural signals during memory recall (F1=0.9). Also, our approach faithfully reconstructs affect-contextualized audio-visual memory across all subjects, both qualitatively and quantitatively, with participants reporting strong affective concordance between their recalled memories and the generated content. Especially, contents generated from subject-reported affect dynamics showed higher correlation with participants' reported affect dynamics trajectories (r=0.265, p<.05) and received stronger user preference (preference=56%) compared to those generated from randomly reordered affect dynamics. Our approaches advance affect decoding research and its practical applications in personalized media creation via neural-based affect comprehension. Codes and the dataset are available at https://github.com/ioahKwon/Revisiting-Your-Memory.", "date": "2024-11-24", "categories": []}
{"id": "2505.06685", "title": "Emotion-Qwen: A Unified Framework for Emotion and Vision Understanding", "abstract": "Accurate emotion understanding in videos necessitates effectively recognizing and interpreting emotional states by integrating visual, textual, auditory, and contextual cues. Although recent Large Multimodal Models (LMMs) have exhibited significant progress in general vision-language (VL) tasks, their performance often deteriorates in emotion-specific scenarios, exhibiting catastrophic forgetting when fine-tuned on emotion-centric tasks. To overcome these limitations, we propose Emotion-Qwen, a unified multimodal framework designed to simultaneously enable robust emotion understanding and preserve general VL reasoning capabilities. Emotion-Qwen introduces a novel Hybrid Compressor based on a Mixture-of-Experts (MoE) architecture, dynamically routing inputs to optimally balance emotion-specific processing and general multimodal reasoning. We further propose a carefully structured three-stage pre-training pipeline, leveraging extensive general and emotion-focused datasets to strengthen multimodal representation robustness and model adaptability. Additionally, we develop the Video Emotion Reasoning (VER) dataset, a large-scale bilingual resource containing over 40K video clips annotated with detailed context-aware emotional descriptions, significantly facilitating research on fine-grained emotional reasoning. Extensive experiments confirm that Emotion-Qwen achieves state-of-the-art performance across multiple emotion recognition and reasoning benchmarks, while maintaining highly competitive results in general VL tasks.", "date": "2025-05-10", "categories": []}
{"id": "2508.06564", "title": "Grounding Emotion Recognition with Visual Prototypes: VEGA -- Revisiting CLIP in MERC", "abstract": "Multimodal Emotion Recognition in Conversations remains a challenging task due to the complex interplay of textual, acoustic and visual signals. While recent models have improved performance via advanced fusion strategies, they often lack psychologically meaningful priors to guide multimodal alignment. In this paper, we revisit the use of CLIP and propose a novel Visual Emotion Guided Anchoring (VEGA) mechanism that introduces class-level visual semantics into the fusion and classification process. Distinct from prior work that primarily utilizes CLIP's textual encoder, our approach leverages its image encoder to construct emotion-specific visual anchors based on facial exemplars. These anchors guide unimodal and multimodal features toward a perceptually grounded and psychologically aligned representation space, drawing inspiration from cognitive theories (prototypical emotion categories and multisensory integration). A stochastic anchor sampling strategy further enhances robustness by balancing semantic stability and intra-class diversity. Integrated into a dual-branch architecture with self-distillation, our VEGA-augmented model achieves sota performance on IEMOCAP and MELD. Code is available at: https://github.com/dkollias/VEGA.", "date": "2025-08-06", "categories": []}
{"id": "2508.09188", "title": "Synthetic Data Generation for Emotional Depth Faces: Optimizing Conditional DCGANs via Genetic Algorithms in the Latent Space and Stabilizing Training with Knowledge Distillation", "abstract": "Affective computing faces a major challenge: the lack of high-quality, diverse depth facial datasets for recognizing subtle emotional expressions. We propose a framework for synthetic depth face generation using an optimized GAN with Knowledge Distillation (EMA teacher models) to stabilize training, improve quality, and prevent mode collapse. We also apply Genetic Algorithms to evolve GAN latent vectors based on image statistics, boosting diversity and visual quality for target emotions. The approach outperforms GAN, VAE, GMM, and KDE in both diversity and quality. For classification, we extract and concatenate LBP, HOG, Sobel edge, and intensity histogram features, achieving 94% and 96% accuracy with XGBoost. Evaluation using FID, IS, SSIM, and PSNR shows consistent improvement over state-of-the-art methods.", "date": "2025-08-07", "categories": []}
{"id": "2408.10328", "title": "Decoding Human Emotions: Analyzing Multi-Channel EEG Data using LSTM Networks", "abstract": "Emotion recognition from electroencephalogram (EEG) signals is a thriving field, particularly in neuroscience and Human-Computer Interaction (HCI). This study aims to understand and improve the predictive accuracy of emotional state classification through metrics such as valence, arousal, dominance, and likeness by applying a Long Short-Term Memory (LSTM) network to analyze EEG signals. Using a popular dataset of multi-channel EEG recordings known as DEAP, we look towards leveraging LSTM networks' properties to handle temporal dependencies within EEG signal data. This allows for a more comprehensive understanding and classification of emotional parameter states. We obtain accuracies of 89.89%, 90.33%, 90.70%, and 90.54% for arousal, valence, dominance, and likeness, respectively, demonstrating significant improvements in emotion recognition model capabilities. This paper elucidates the methodology and architectural specifics of our LSTM model and provides a benchmark analysis with existing papers.", "date": "2024-08-19", "categories": []}
{"id": "2412.02935", "title": "Dynamic Graph Neural ODE Network for Multi-modal Emotion Recognition in Conversation", "abstract": "Multimodal emotion recognition in conversation (MERC) refers to identifying and classifying human emotional states by combining data from multiple different modalities (e.g., audio, images, text, video, etc.). Most existing multimodal emotion recognition methods use GCN to improve performance, but existing GCN methods are prone to overfitting and cannot capture the temporal dependency of the speaker's emotions. To address the above problems, we propose a Dynamic Graph Neural Ordinary Differential Equation Network (DGODE) for MERC, which combines the dynamic changes of emotions to capture the temporal dependency of speakers' emotions, and effectively alleviates the overfitting problem of GCNs. Technically, the key idea of DGODE is to utilize an adaptive mixhop mechanism to improve the generalization ability of GCNs and use the graph ODE evolution network to characterize the continuous dynamics of node representations over time and capture temporal dependencies. Extensive experiments on two publicly available multimodal emotion recognition datasets demonstrate that the proposed DGODE model has superior performance compared to various baselines. Furthermore, the proposed DGODE can also alleviate the over-smoothing problem, thereby enabling the construction of a deep GCN network.", "date": "2024-12-04", "categories": []}
{"id": "2412.03300", "title": "Touch and Tell: Multimodal Decoding of Human Emotions and Social Gestures for Robots", "abstract": "Human emotions are complex and can be conveyed through nuanced touch gestures. Previous research has primarily focused on how humans recognize emotions through touch or on identifying key features of emotional expression for robots. However, there is a gap in understanding how reliably these emotions and gestures can be communicated to robots via touch and interpreted using data driven methods. This study investigates the consistency and distinguishability of emotional and gestural expressions through touch and sound. To this end, we integrated a custom piezoresistive pressure sensor as well as a microphone on a social robot. Twenty-eight participants first conveyed ten different emotions to the robot using spontaneous touch gestures, then they performed six predefined social touch gestures. Our findings reveal statistically significant consistency in both emotion and gesture expression among participants. However, some emotions exhibited low intraclass correlation values, and certain emotions with similar levels of arousal or valence did not show significant differences in their conveyance. To investigate emotion and social gesture decoding within affective human-robot tactile interaction, we developed single-modality models and multimodal models integrating tactile and auditory features. A support vector machine (SVM) model trained on multimodal features achieved the highest accuracy for classifying ten emotions, reaching 40 %.For gesture classification, a Convolutional Neural Network- Long Short-Term Memory Network (CNN-LSTM) achieved 90.74 % accuracy. Our results demonstrate that even though the unimodal models have the potential to decode emotions and touch gestures, the multimodal integration of touch and sound significantly outperforms unimodal approaches, enhancing the decoding of both emotions and gestures.", "date": "2024-12-04", "categories": []}
{"id": "2502.12454", "title": "Zero-shot Emotion Annotation in Facial Images Using Large Multimodal Models: Benchmarking and Prospects for Multi-Class, Multi-Frame Approaches", "abstract": "This study investigates the feasibility and performance of using large multimodal models (LMMs) to automatically annotate human emotions in everyday scenarios. We conducted experiments on the DailyLife subset of the publicly available FERV39k dataset, employing the GPT-4o-mini model for rapid, zero-shot labeling of key frames extracted from video segments. Under a seven-class emotion taxonomy (\"Angry,\" \"Disgust,\" \"Fear,\" \"Happy,\" \"Neutral,\" \"Sad,\" \"Surprise\"), the LMM achieved an average precision of approximately 50%. In contrast, when limited to ternary emotion classification (negative/neutral/positive), the average precision increased to approximately 64%. Additionally, we explored a strategy that integrates multiple frames within 1-2 second video clips to enhance labeling performance and reduce costs. The results indicate that this approach can slightly improve annotation accuracy. Overall, our preliminary findings highlight the potential application of zero-shot LMMs in human facial emotion annotation tasks, offering new avenues for reducing labeling costs and broadening the applicability of LMMs in complex multimodal environments.", "date": "2025-02-18", "categories": []}
{"id": "2508.08650", "title": "UWB at WASSA-2024 Shared Task 2: Cross-lingual Emotion Detection", "abstract": "This paper presents our system built for the WASSA-2024 Cross-lingual Emotion Detection Shared Task. The task consists of two subtasks: first, to assess an emotion label from six possible classes for a given tweet in one of five languages, and second, to predict words triggering the detected emotions in binary and numerical formats. Our proposed approach revolves around fine-tuning quantized large language models, specifically Orca~2, with low-rank adapters (LoRA) and multilingual Transformer-based models, such as XLM-R and mT5. We enhance performance through machine translation for both subtasks and trigger word switching for the second subtask. The system achieves excellent performance, ranking 1st in numerical trigger words detection, 3rd in binary trigger words detection, and 7th in emotion detection.", "date": "2025-08-12", "categories": []}
{"id": "2508.08830", "title": "Silicon Minds versus Human Hearts: The Wisdom of Crowds Beats the Wisdom of AI in Emotion Recognition", "abstract": "The ability to discern subtle emotional cues is fundamental to human social intelligence. As artificial intelligence (AI) becomes increasingly common, AI's ability to recognize and respond to human emotions is crucial for effective human-AI interactions. In particular, whether such systems can match or surpass human experts remains to be seen. However, the emotional intelligence of AI, particularly multimodal large language models (MLLMs), remains largely unexplored. This study evaluates the emotion recognition abilities of MLLMs using the Reading the Mind in the Eyes Test (RMET) and its multiracial counterpart (MRMET), and compares their performance against human participants. Results show that, on average, MLLMs outperform humans in accurately identifying emotions across both tests. This trend persists even when comparing performance across low, medium, and expert-level performing groups. Yet when we aggregate independent human decisions to simulate collective intelligence, human groups significantly surpass the performance of aggregated MLLM predictions, highlighting the wisdom of the crowd. Moreover, a collaborative approach (augmented intelligence) that combines human and MLLM predictions achieves greater accuracy than either humans or MLLMs alone. These results suggest that while MLLMs exhibit strong emotion recognition at the individual level, the collective intelligence of humans and the synergistic potential of human-AI collaboration offer the most promising path toward effective emotional AI. We discuss the implications of these findings for the development of emotionally intelligent AI systems and future research directions.", "date": "2025-08-12", "categories": []}
{"id": "2311.00721", "title": "Empathy Detection from Text, Audiovisual, Audio or Physiological Signals: A Systematic Review of Task Formulations and Machine Learning Methods", "abstract": "Empathy indicates an individual's ability to understand others. Over the past few years, empathy has drawn attention from various disciplines, including but not limited to Affective Computing, Cognitive Science, and Psychology. Detecting empathy has potential applications in society, healthcare and education. Despite being a broad and overlapping topic, the avenue of empathy detection leveraging Machine Learning remains underexplored from a systematic literature review perspective. We collected 849 papers from 10 well-known academic databases, systematically screened them and analysed the final 82 papers. Our analyses reveal several prominent task formulations - including empathy on localised utterances or overall expressions, unidirectional or parallel empathy, and emotional contagion - in monadic, dyadic and group interactions. Empathy detection methods are summarised based on four input modalities - text, audiovisual, audio and physiological signals - thereby presenting modality-specific network architecture design protocols. We discuss challenges, research gaps and potential applications in the Affective Computing-based empathy domain, which can facilitate new avenues of exploration. We further enlist the public availability of datasets and codes. This paper, therefore, provides a structured overview of recent advancements and remaining challenges towards developing a robust empathy detection system that could meaningfully contribute to enhancing human well-being.", "date": "2023-10-30", "categories": []}
{"id": "2508.06890", "title": "Maestro-EVC: Controllable Emotional Voice Conversion Guided by References and Explicit Prosody", "abstract": "Emotional voice conversion (EVC) aims to modify the emotional style of speech while preserving its linguistic content. In practical EVC, controllability, the ability to independently control speaker identity and emotional style using distinct references, is crucial. However, existing methods often struggle to fully disentangle these attributes and lack the ability to model fine-grained emotional expressions such as temporal dynamics. We propose Maestro-EVC, a controllable EVC framework that enables independent control of content, speaker identity, and emotion by effectively disentangling each attribute from separate references. We further introduce a temporal emotion representation and an explicit prosody modeling with prosody augmentation to robustly capture and transfer the temporal dynamics of the target emotion, even under prosody-mismatched conditions. Experimental results confirm that Maestro-EVC achieves high-quality, controllable, and emotionally expressive speech synthesis.", "date": "2025-08-09", "categories": []}
{"id": "2508.06902", "title": "eMotions: A Large-Scale Dataset and Audio-Visual Fusion Network for Emotion Analysis in Short-form Videos", "abstract": "Short-form videos (SVs) have become a vital part of our online routine for acquiring and sharing information. Their multimodal complexity poses new challenges for video analysis, highlighting the need for video emotion analysis (VEA) within the community. Given the limited availability of SVs emotion data, we introduce eMotions, a large-scale dataset consisting of 27,996 videos with full-scale annotations. To ensure quality and reduce subjective bias, we emphasize better personnel allocation and propose a multi-stage annotation procedure. Additionally, we provide the category-balanced and test-oriented variants through targeted sampling to meet diverse needs. While there have been significant studies on videos with clear emotional cues (e.g., facial expressions), analyzing emotions in SVs remains a challenging task. The challenge arises from the broader content diversity, which introduces more distinct semantic gaps and complicates the representations learning of emotion-related features. Furthermore, the prevalence of audio-visual co-expressions in SVs leads to the local biases and collective information gaps caused by the inconsistencies in emotional expressions. To tackle this, we propose AV-CANet, an end-to-end audio-visual fusion network that leverages video transformer to capture semantically relevant representations. We further introduce the Local-Global Fusion Module designed to progressively capture the correlations of audio-visual features. Besides, EP-CE Loss is constructed to globally steer optimizations with tripolar penalties. Extensive experiments across three eMotions-related datasets and four public VEA datasets demonstrate the effectiveness of our proposed AV-CANet, while providing broad insights for future research. Moreover, we conduct ablation studies to examine the critical components of our method. Dataset and code will be made available at Github.", "date": "2025-08-09", "categories": []}
{"id": "2508.07625", "title": "A Trustworthy Method for Multimodal Emotion Recognition", "abstract": "Existing emotion recognition methods mainly focus on enhancing performance by employing complex deep models, typically resulting in significantly higher model complexity. Although effective, it is also crucial to ensure the reliability of the final decision, especially for noisy, corrupted and out-of-distribution data. To this end, we propose a novel emotion recognition method called trusted emotion recognition (TER), which utilizes uncertainty estimation to calculate the confidence value of predictions. TER combines the results from multiple modalities based on their confidence values to output the trusted predictions. We also provide a new evaluation criterion to assess the reliability of predictions. Specifically, we incorporate trusted precision and trusted recall to determine the trusted threshold and formulate the trusted Acc. and trusted F1 score to evaluate the model's trusted performance. The proposed framework combines the confidence module that accordingly endows the model with reliability and robustness against possible noise or corruption. The extensive experimental results validate the effectiveness of our proposed model. The TER achieves state-of-the-art performance on the Music-video, achieving 82.40% Acc. In terms of trusted performance, TER outperforms other methods on the IEMOCAP and Music-video, achieving trusted F1 scores of 0.7511 and 0.9035, respectively.", "date": "2025-08-11", "categories": []}
{"id": "2508.08093", "title": "MDD-Net: Multimodal Depression Detection through Mutual Transformer", "abstract": "Depression is a major mental health condition that severely impacts the emotional and physical well-being of individuals. The simple nature of data collection from social media platforms has attracted significant interest in properly utilizing this information for mental health research. A Multimodal Depression Detection Network (MDD-Net), utilizing acoustic and visual data obtained from social media networks, is proposed in this work where mutual transformers are exploited to efficiently extract and fuse multimodal features for efficient depression detection. The MDD-Net consists of four core modules: an acoustic feature extraction module for retrieving relevant acoustic attributes, a visual feature extraction module for extracting significant high-level patterns, a mutual transformer for computing the correlations among the generated features and fusing these features from multiple modalities, and a detection layer for detecting depression using the fused feature representations. The extensive experiments are performed using the multimodal D-Vlog dataset, and the findings reveal that the developed multimodal depression detection network surpasses the state-of-the-art by up to 17.37% for F1-Score, demonstrating the greater performance of the proposed system. The source code is accessible at https://github.com/rezwanh001/Multimodal-Depression-Detection.", "date": "2025-08-11", "categories": []}
{"id": "2501.08393", "title": "Empathetic Conversational Agents: Utilizing Neural and Physiological Signals for Enhanced Empathetic Interactions", "abstract": "Conversational agents (CAs) are revolutionizing human-computer interaction by evolving from text-based chatbots to empathetic digital humans (DHs) capable of rich emotional expressions. This paper explores the integration of neural and physiological signals into the perception module of CAs to enhance empathetic interactions. By leveraging these cues, the study aims to detect emotions in real-time and generate empathetic responses and expressions. We conducted a user study where participants engaged in conversations with a DH about emotional topics. The DH responded and displayed expressions by mirroring detected emotions in real-time using neural and physiological cues. The results indicate that participants experienced stronger emotions and greater engagement during interactions with the Empathetic DH, demonstrating the effectiveness of incorporating neural and physiological signals for real-time emotion recognition. However, several challenges were identified, including recognition accuracy, emotional transition speeds, individual personality effects, and limitations in voice tone modulation. Addressing these challenges is crucial for further refining Empathetic DHs and fostering meaningful connections between humans and artificial entities. Overall, this research advances human-agent interaction and highlights the potential of real-time neural and physiological emotion recognition in creating empathetic DHs.", "date": "2025-01-14", "categories": []}
{"id": "2508.05933", "title": "REFS: Robust EEG feature selection with missing multi-dimensional annotation for emotion recognition", "abstract": "The affective brain-computer interface is a crucial technology for affective interaction and emotional intelligence, emerging as a significant area of research in the human-computer interaction. Compared to single-type features, multi-type EEG features provide a multi-level representation for analyzing multi-dimensional emotions. However, the high dimensionality of multi-type EEG features, combined with the relatively small number of high-quality EEG samples, poses challenges such as classifier overfitting and suboptimal real-time performance in multi-dimensional emotion recognition. Moreover, practical applications of affective brain-computer interface frequently encounters partial absence of multi-dimensional emotional labels due to the open nature of the acquisition environment, and ambiguity and variability in individual emotion perception. To address these challenges, this study proposes a novel EEG feature selection method for missing multi-dimensional emotion recognition. The method leverages adaptive orthogonal non-negative matrix factorization to reconstruct the multi-dimensional emotional label space through second-order and higher-order correlations, which could reduce the negative impact of missing values and outliers on label reconstruction. Simultaneously, it employs least squares regression with graph-based manifold learning regularization and global feature redundancy minimization regularization to enable EEG feature subset selection despite missing information, ultimately achieving robust EEG-based multi-dimensional emotion recognition. Simulation experiments on three widely used multi-dimensional emotional datasets, DREAMER, DEAP and HDED, reveal that the proposed method outperforms thirteen advanced feature selection methods in terms of robustness for EEG emotional feature selection.", "date": "2025-08-08", "categories": []}
{"id": "2508.05991", "title": "ECMF: Enhanced Cross-Modal Fusion for Multimodal Emotion Recognition in MER-SEMI Challenge", "abstract": "Emotion recognition plays a vital role in enhancing human-computer interaction. In this study, we tackle the MER-SEMI challenge of the MER2025 competition by proposing a novel multimodal emotion recognition framework. To address the issue of data scarcity, we leverage large-scale pre-trained models to extract informative features from visual, audio, and textual modalities. Specifically, for the visual modality, we design a dual-branch visual encoder that captures both global frame-level features and localized facial representations. For the textual modality, we introduce a context-enriched method that employs large language models to enrich emotional cues within the input text. To effectively integrate these multimodal features, we propose a fusion strategy comprising two key components, i.e., self-attention mechanisms for dynamic modality weighting, and residual connections to preserve original representations. Beyond architectural design, we further refine noisy labels in the training set by a multi-source labeling strategy. Our approach achieves a substantial performance improvement over the official baseline on the MER2025-SEMI dataset, attaining a weighted F-score of 87.49% compared to 78.63%, thereby validating the effectiveness of the proposed framework.", "date": "2025-08-08", "categories": []}
{"id": "2508.06036", "title": "More Is Better: A MoE-Based Emotion Recognition Framework with Human Preference Alignment", "abstract": "In this paper, we present our solution for the semi-supervised learning track (MER-SEMI) in MER2025. We propose a comprehensive framework, grounded in the principle that \"more is better,\" to construct a robust Mixture of Experts (MoE) emotion recognition system. Our approach integrates a diverse range of input modalities as independent experts, including novel signals such as knowledge from large Vision-Language Models (VLMs) and temporal Action Unit (AU) information. To effectively utilize unlabeled data, we introduce a consensus-based pseudo-labeling strategy, generating high-quality labels from the agreement between a baseline model and Gemini, which are then used in a two-stage training paradigm. Finally, we employ a multi-expert voting ensemble combined with a rule-based re-ranking process to correct prediction bias and better align the outputs with human preferences. Evaluated on the MER2025-SEMI challenge dataset, our method achieves an F1-score of 0.8772 on the test set, ranking 2nd in the track. Our code is available at https://github.com/zhuyjan/MER2025-MRAC25.", "date": "2025-08-08", "categories": []}
{"id": "2508.06321", "title": "EmoAugNet: A Signal-Augmented Hybrid CNN-LSTM Framework for Speech Emotion Recognition", "abstract": "Recognizing emotional signals in speech has a significant impact on enhancing the effectiveness of human-computer interaction (HCI). This study introduces EmoAugNet, a hybrid deep learning framework, that incorporates Long Short-Term Memory (LSTM) layers with one-dimensional Convolutional Neural Networks (1D-CNN) to enable reliable Speech Emotion Recognition (SER). The quality and variety of the features that are taken from speech signals have a significant impact on how well SER systems perform. A comprehensive speech data augmentation strategy was used to combine both traditional methods, such as noise addition, pitch shifting, and time stretching, with a novel combination-based augmentation pipeline to enhance generalization and reduce overfitting. Each audio sample was transformed into a high-dimensional feature vector using root mean square energy (RMSE), Mel-frequency Cepstral Coefficient (MFCC), and zero-crossing rate (ZCR). Our model with ReLU activation has a weighted accuracy of 95.78\\% and unweighted accuracy of 92.52\\% on the IEMOCAP dataset and, with ELU activation, has a weighted accuracy of 96.75\\% and unweighted accuracy of 91.28\\%. On the RAVDESS dataset, we get a weighted accuracy of 94.53\\% and 94.98\\% unweighted accuracy for ReLU activation and 93.72\\% weighted accuracy and 94.64\\% unweighted accuracy for ELU activation. These results highlight EmoAugNet's effectiveness in improving the robustness and performance of SER systems through integated data augmentation and hybrid modeling.", "date": "2025-08-06", "categories": []}
{"id": "2508.04723", "title": "Wearable Music2Emotion : Assessing Emotions Induced by AI-Generated Music through Portable EEG-fNIRS Fusion", "abstract": "Emotions critically influence mental health, driving interest in music-based affective computing via neurophysiological signals with Brain-computer Interface techniques. While prior studies leverage music's accessibility for emotion induction, three key limitations persist: \\textbf{(1) Stimulus Constraints}: Music stimuli are confined to small corpora due to copyright and curation costs, with selection biases from heuristic emotion-music mappings that ignore individual affective profiles. \\textbf{(2) Modality Specificity}: Overreliance on unimodal neural data (e.g., EEG) ignores complementary insights from cross-modal signal fusion.\\textbf{ (3) Portability Limitation}: Cumbersome setups (e.g., 64+ channel gel-based EEG caps) hinder real-world applicability due to procedural complexity and portability barriers. To address these limitations, we propose MEEtBrain, a portable and multimodal framework for emotion analysis (valence/arousal), integrating AI-generated music stimuli with synchronized EEG-fNIRS acquisition via a wireless headband. By MEEtBrain, the music stimuli can be automatically generated by AI on a large scale, eliminating subjective selection biases while ensuring music diversity. We use our developed portable device that is designed in a lightweight headband-style and uses dry electrodes, to simultaneously collect EEG and fNIRS recordings. A 14-hour dataset from 20 participants was collected in the first recruitment to validate the framework's efficacy, with AI-generated music eliciting target emotions (valence/arousal). We are actively expanding our multimodal dataset (44 participants in the latest dataset) and make it publicly available to promote further research and practical applications. \\textbf{The dataset is available at https://zju-bmi-lab.github.io/ZBra.", "date": "2025-08-05", "categories": []}
{"id": "2508.05228", "title": "CWEFS: Brain volume conduction effects inspired channel-wise EEG feature selection for multi-dimensional emotion recognition", "abstract": "Due to the intracranial volume conduction effects, high-dimensional multi-channel electroencephalography (EEG) features often contain substantial redundant and irrelevant information. This issue not only hinders the extraction of discriminative emotional representations but also compromises the real-time performance. Feature selection has been established as an effective approach to address the challenges while enhancing the transparency and interpretability of emotion recognition models. However, existing EEG feature selection research overlooks the influence of latent EEG feature structures on emotional label correlations and assumes uniform importance across various channels, directly limiting the precise construction of EEG feature selection models for multi-dimensional affective computing. To address these limitations, a novel channel-wise EEG feature selection (CWEFS) method is proposed for multi-dimensional emotion recognition. Specifically, inspired by brain volume conduction effects, CWEFS integrates EEG emotional feature selection into a shared latent structure model designed to construct a consensus latent space across diverse EEG channels. To preserve the local geometric structure, this consensus space is further integrated with the latent semantic analysis of multi-dimensional emotional labels. Additionally, CWEFS incorporates adaptive channel-weight learning to automatically determine the significance of different EEG channels in the emotional feature selection task. The effectiveness of CWEFS was validated using three popular EEG datasets with multi-dimensional emotional labels. Comprehensive experimental results, compared against nineteen feature selection methods, demonstrate that the EEG feature subsets chosen by CWEFS achieve optimal emotion recognition performance across six evaluation metrics.", "date": "2025-08-07", "categories": []}
{"id": "2508.05474", "title": "Can Large Language Models Generate Effective Datasets for Emotion Recognition in Conversations?", "abstract": "Emotion recognition in conversations (ERC) focuses on identifying emotion shifts within interactions, representing a significant step toward advancing machine intelligence. However, ERC data remains scarce, and existing datasets face numerous challenges due to their highly biased sources and the inherent subjectivity of soft labels. Even though Large Language Models (LLMs) have demonstrated their quality in many affective tasks, they are typically expensive to train, and their application to ERC tasks--particularly in data generation--remains limited. To address these challenges, we employ a small, resource-efficient, and general-purpose LLM to synthesize ERC datasets with diverse properties, supplementing the three most widely used ERC benchmarks. We generate six novel datasets, with two tailored to enhance each benchmark. We evaluate the utility of these datasets to (1) supplement existing datasets for ERC classification, and (2) analyze the effects of label imbalance in ERC. Our experimental results indicate that ERC classifier models trained on the generated datasets exhibit strong robustness and consistently achieve statistically significant performance improvements on existing ERC benchmarks.", "date": "2025-08-07", "categories": []}
{"id": "2505.22548", "title": "Emotion-o1: Adaptive Long Reasoning for Emotion Understanding in LLMs", "abstract": "Long chain-of-thought (CoT) reasoning has shown great promise in enhancing the emotion understanding performance of large language models (LLMs). However, current fixed-length CoT methods struggle to balance reasoning depth and efficiency. Simple tasks (e.g., sentiment classification) are over-reasoned, while complex tasks (e.g., sarcasm understanding) lack depth. To fill this gap, we present Emotion-o1, an adaptive CoT framework that dynamically adjusts reasoning length based on emotion-task complexity. Emotion-o1 is trained by distilling adaptive CoT patterns from a reasoning-oriented LLM, followed by supervised fine-tuning and reinforcement learning with a four-part reward targeting accuracy, brevity, structure, and redundancy. Experimental results on four emotion tasks highlight: (1) Emotion-o1 demonstrates significant improvements over its backbone, with F1 score increases of 10%(Sentiment), 5%(Emotion), 18%(Humor), and 27%(Sarcasm). (2) In sentiment and sarcasm tasks, our 8B model demonstrates superior performance against advanced LLMs, outperforming Grok-3 by 1.1% and Claude-3.7 by 2%. (3) The framework maintains accuracy while reducing reasoning length by 83% compared to OpenAI-o1, demonstrating effective precision-efficiency optimization. Emotion-o1 effectively balances reasoning depth and efficiency for emotion understanding in LLMs.", "date": "2025-05-28", "categories": []}
{"id": "2508.03722", "title": "Multimodal Video Emotion Recognition with Reliable Reasoning Priors", "abstract": "This study investigates the integration of trustworthy prior reasoning knowledge from MLLMs into multimodal emotion recognition. We employ Gemini to generate fine-grained, modality-separable reasoning traces, which are injected as priors during the fusion stage to enrich cross-modal interactions. To mitigate the pronounced class-imbalance in multimodal emotion recognition, we introduce Balanced Dual-Contrastive Learning, a loss formulation that jointly balances inter-class and intra-class distributions. Applied to the MER2024 benchmark, our prior-enhanced framework yields substantial performance gains, demonstrating that the reliability of MLLM-derived reasoning can be synergistically combined with the domain adaptability of lightweight fusion networks for robust, scalable emotion recognition.", "date": "2025-07-29", "categories": []}
{"id": "2508.03729", "title": "Privileged Contrastive Pretraining for Multimodal Affect Modelling", "abstract": "Affective Computing (AC) has made significant progress with the advent of deep learning, yet a persistent challenge remains: the reliable transfer of affective models from controlled laboratory settings (in-vitro) to uncontrolled real-world environments (in-vivo). To address this challenge we introduce the Privileged Contrastive Pretraining (PriCon) framework according to which models are first pretrained via supervised contrastive learning (SCL) and then act as teacher models within a Learning Using Privileged Information (LUPI) framework. PriCon both leverages privileged information during training and enhances the robustness of derived affect models via SCL. Experiments conducted on two benchmark affective corpora, RECOLA and AGAIN, demonstrate that models trained using PriCon consistently outperform LUPI and end to end models. Remarkably, in many cases, PriCon models achieve performance comparable to models trained with access to all modalities during both training and testing. The findings underscore the potential of PriCon as a paradigm towards further bridging the gap between in-vitro and in-vivo affective modelling, offering a scalable and practical solution for real-world applications.", "date": "2025-07-30", "categories": []}
{"id": "2508.04481", "title": "Emotion Detection Using Conditional Generative Adversarial Networks (cGAN): A Deep Learning Approach", "abstract": "This paper presents a deep learning-based approach to emotion detection using Conditional Generative Adversarial Networks (cGANs). Unlike traditional unimodal techniques that rely on a single data type, we explore a multimodal framework integrating text, audio, and facial expressions. The proposed cGAN architecture is trained to generate synthetic emotion-rich data and improve classification accuracy across multiple modalities. Our experimental results demonstrate significant improvements in emotion recognition performance compared to baseline models. This work highlights the potential of cGANs in enhancing human-computer interaction systems by enabling more nuanced emotional understanding.", "date": "2025-08-06", "categories": []}
{"id": "2507.22676", "title": "Listening to the Unspoken: Exploring \"365\" Aspects of Multimodal Interview Performance Assessment", "abstract": "Interview performance assessment is essential for determining candidates' suitability for professional positions. To ensure holistic and fair evaluations, we propose a novel and comprehensive framework that explores ``365'' aspects of interview performance by integrating \\textit{three} modalities (video, audio, and text), \\textit{six} responses per candidate, and \\textit{five} key evaluation dimensions. The framework employs modality-specific feature extractors to encode heterogeneous data streams and subsequently fused via a Shared Compression Multilayer Perceptron. This module compresses multimodal embeddings into a unified latent space, facilitating efficient feature interaction. To enhance prediction robustness, we incorporate a two-level ensemble learning strategy: (1) independent regression heads predict scores for each response, and (2) predictions are aggregated across responses using a mean-pooling mechanism to produce final scores for the five target dimensions. By listening to the unspoken, our approach captures both explicit and implicit cues from multimodal data, enabling comprehensive and unbiased assessments. Achieving a multi-dimensional average MSE of 0.1824, our framework secured first place in the AVI Challenge 2025, demonstrating its effectiveness and robustness in advancing automated and multimodal interview performance assessment. The full implementation is available at https://github.com/MSA-LMC/365Aspects.", "date": "2025-07-30", "categories": []}
{"id": "2508.02680", "title": "AnnoSense: A Framework for Physiological Emotion Data Collection in Everyday Settings for AI", "abstract": "Emotional and mental well-being are vital components of quality of life, and with the rise of smart devices like smartphones, wearables, and artificial intelligence (AI), new opportunities for monitoring emotions in everyday settings have emerged. However, for AI algorithms to be effective, they require high-quality data and accurate annotations. As the focus shifts towards collecting emotion data in real-world environments to capture more authentic emotional experiences, the process of gathering emotion annotations has become increasingly complex. This work explores the challenges of everyday emotion data collection from the perspectives of key stakeholders. We collected 75 survey responses, performed 32 interviews with the public, and 3 focus group discussions (FGDs) with 12 mental health professionals. The insights gained from a total of 119 stakeholders informed the development of our framework, AnnoSense, designed to support everyday emotion data collection for AI. This framework was then evaluated by 25 emotion AI experts for its clarity, usefulness, and adaptability. Lastly, we discuss the potential next steps and implications of AnnoSense for future research in emotion AI, highlighting its potential to enhance the collection and analysis of emotion data in real-world contexts.", "date": "2025-07-17", "categories": []}
{"id": "2508.03535", "title": "CoEmoGen: Towards Semantically-Coherent and Scalable Emotional Image Content Generation", "abstract": "Emotional Image Content Generation (EICG) aims to generate semantically clear and emotionally faithful images based on given emotion categories, with broad application prospects. While recent text-to-image diffusion models excel at generating concrete concepts, they struggle with the complexity of abstract emotions. There have also emerged methods specifically designed for EICG, but they excessively rely on word-level attribute labels for guidance, which suffer from semantic incoherence, ambiguity, and limited scalability. To address these challenges, we propose CoEmoGen, a novel pipeline notable for its semantic coherence and high scalability. Specifically, leveraging multimodal large language models (MLLMs), we construct high-quality captions focused on emotion-triggering content for context-rich semantic guidance. Furthermore, inspired by psychological insights, we design a Hierarchical Low-Rank Adaptation (HiLoRA) module to cohesively model both polarity-shared low-level features and emotion-specific high-level semantics. Extensive experiments demonstrate CoEmoGen's superiority in emotional faithfulness and semantic coherence from quantitative, qualitative, and user study perspectives. To intuitively showcase scalability, we curate EmoArt, a large-scale dataset of emotionally evocative artistic images, providing endless inspiration for emotion-driven artistic creation. The dataset and code are available at https://github.com/yuankaishen2001/CoEmoGen.", "date": "2025-08-05", "categories": []}
{"id": "2505.02331", "title": "VAEmo: Efficient Representation Learning for Visual-Audio Emotion with Knowledge Injection", "abstract": "Audiovisual emotion recognition (AVER) aims to infer human emotions from nonverbal visual-audio (VA) cues, offering modality-complementary and language-agnostic advantages. However, AVER remains challenging due to the inherent ambiguity of emotional expressions, cross-modal expressive disparities, and the scarcity of reliably annotated data. Recent self-supervised AVER approaches have introduced strong multimodal representations, yet they predominantly rely on modality-specific encoders and coarse content-level alignment, limiting fine-grained emotional semantic modeling. To address these issues, we propose VAEmo, an efficient two-stage framework for emotion-centric joint VA representation learning with external knowledge injection. In Stage~1, a unified and lightweight representation network is pre-trained on large-scale speaker-centric VA corpora via masked reconstruction and contrastive objectives, mitigating the modality gap and learning expressive, complementary representations without emotion labels. In Stage~2, multimodal large language models automatically generate detailed affective descriptions according to our well-designed chain-of-thought prompting for only a small subset of VA samples; these rich textual semantics are then injected by aligning their corresponding embeddings with VA representations through dual-path contrastive learning, further bridging the emotion gap. Extensive experiments on multiple downstream AVER benchmarks show that VAEmo achieves state-of-the-art performance with a compact design, highlighting the benefit of unified cross-modal encoding and emotion-aware semantic guidance for efficient, generalizable VA emotion representations.", "date": "2025-05-05", "categories": []}
{"id": "2508.01161", "title": "CSIRO-LT at SemEval-2025 Task 11: Adapting LLMs for Emotion Recognition for Multiple Languages", "abstract": "Detecting emotions across different languages is challenging due to the varied and culturally nuanced ways of emotional expressions. The \\textit{Semeval 2025 Task 11: Bridging the Gap in Text-Based emotion} shared task was organised to investigate emotion recognition across different languages. The goal of the task is to implement an emotion recogniser that can identify the basic emotional states that general third-party observers would attribute to an author based on their written text snippet, along with the intensity of those emotions. We report our investigation of various task-adaptation strategies for LLMs in emotion recognition. We show that the most effective method for this task is to fine-tune a pre-trained multilingual LLM with LoRA setting separately for each language.", "date": "2025-08-02", "categories": []}
{"id": "2508.01644", "title": "DRKF: Decoupled Representations with Knowledge Fusion for Multimodal Emotion Recognition", "abstract": "Multimodal emotion recognition (MER) aims to identify emotional states by integrating and analyzing information from multiple modalities. However, inherent modality heterogeneity and inconsistencies in emotional cues remain key challenges that hinder performance. To address these issues, we propose a Decoupled Representations with Knowledge Fusion (DRKF) method for MER. DRKF consists of two main modules: an Optimized Representation Learning (ORL) Module and a Knowledge Fusion (KF) Module. ORL employs a contrastive mutual information estimation method with progressive modality augmentation to decouple task-relevant shared representations and modality-specific features while mitigating modality heterogeneity. KF includes a lightweight self-attention-based Fusion Encoder (FE) that identifies the dominant modality and integrates emotional information from other modalities to enhance the fused representation. To handle potential errors from incorrect dominant modality selection under emotionally inconsistent conditions, we introduce an Emotion Discrimination Submodule (ED), which enforces the fused representation to retain discriminative cues of emotional inconsistency. This ensures that even if the FE selects an inappropriate dominant modality, the Emotion Classification Submodule (EC) can still make accurate predictions by leveraging preserved inconsistency information. Experiments show that DRKF achieves state-of-the-art (SOTA) performance on IEMOCAP, MELD, and M3ED. The source code is publicly available at https://github.com/PANPANKK/DRKF.", "date": "2025-08-03", "categories": []}
{"id": "2508.02417", "title": "The Role of Review Process Failures in Affective State Estimation: An Empirical Investigation of DEAP Dataset", "abstract": "The reliability of affective state estimation using EEG data is in question, given the variability in reported performance and the lack of standardized evaluation protocols. To investigate this, we reviewed 101 studies, focusing on the widely used DEAP dataset for emotion recognition. Our analysis revealed widespread methodological issues that include data leakage from improper segmentation, biased feature selection, flawed hyperparameter optimization, neglect of class imbalance, and insufficient methodological reporting. Notably, we found that nearly 87% of the reviewed papers contained one or more of these errors. Moreover, through experimental analysis, we observed that such methodological flaws can inflate the classification accuracy by up to 46%. These findings reveal fundamental gaps in standardized evaluation practices and highlight critical deficiencies in the peer review process for machine learning applications in neuroscience, emphasizing the urgent need for stricter methodological standards and evaluation protocols.", "date": "2025-08-04", "categories": []}
{"id": "2508.02429", "title": "Multimodal Large Language Models for End-to-End Affective Computing: Benchmarking and Boosting with Generative Knowledge Prompting", "abstract": "Multimodal Affective Computing (MAC) aims to recognize and interpret human emotions by integrating information from diverse modalities such as text, video, and audio. Recent advancements in Multimodal Large Language Models (MLLMs) have significantly reshaped the landscape of MAC by offering a unified framework for processing and aligning cross-modal information. However, practical challenges remain, including performance variability across complex MAC tasks and insufficient understanding of how architectural designs and data characteristics impact affective analysis. To address these gaps, we conduct a systematic benchmark evaluation of state-of-the-art open-source MLLMs capable of concurrently processing audio, visual, and textual modalities across multiple established MAC datasets. Our evaluation not only compares the performance of these MLLMs but also provides actionable insights into model optimization by analyzing the influence of model architectures and dataset properties. Furthermore, we propose a novel hybrid strategy that combines generative knowledge prompting with supervised fine-tuning to enhance MLLMs' affective computing capabilities. Experimental results demonstrate that this integrated approach significantly improves performance across various MAC tasks, offering a promising avenue for future research and development in this field. Our code is released on https://github.com/LuoMSen/MLLM-MAC.", "date": "2025-08-04", "categories": []}
{"id": "2409.09545", "title": "Multi-Microphone and Multi-Modal Emotion Recognition in Reverberant Environment", "abstract": "This paper presents a Multi-modal Emotion Recognition (MER) system designed to enhance emotion recognition accuracy in challenging acoustic conditions. Our approach combines a modified and extended Hierarchical Token-semantic Audio Transformer (HTS-AT) for multi-channel audio processing with an R(2+1)D Convolutional Neural Networks (CNN) model for video analysis. We evaluate our proposed method on a reverberated version of the Ryerson audio-visual database of emotional speech and song (RAVDESS) dataset using synthetic and real-world Room Impulse Responsess (RIRs). Our results demonstrate that integrating audio and video modalities yields superior performance compared to uni-modal approaches, especially in challenging acoustic conditions. Moreover, we show that the multimodal (audiovisual) approach that utilizes multiple microphones outperforms its single-microphone counterpart.", "date": "2024-09-14", "categories": []}
{"id": "2507.21395", "title": "Sync-TVA: A Graph-Attention Framework for Multimodal Emotion Recognition with Cross-Modal Fusion", "abstract": "Multimodal emotion recognition (MER) is crucial for enabling emotionally intelligent systems that perceive and respond to human emotions. However, existing methods suffer from limited cross-modal interaction and imbalanced contributions across modalities. To address these issues, we propose Sync-TVA, an end-to-end graph-attention framework featuring modality-specific dynamic enhancement and structured cross-modal fusion. Our design incorporates a dynamic enhancement module for each modality and constructs heterogeneous cross-modal graphs to model semantic relations across text, audio, and visual features. A cross-attention fusion mechanism further aligns multimodal cues for robust emotion inference. Experiments on MELD and IEMOCAP demonstrate consistent improvements over state-of-the-art models in both accuracy and weighted F1 score, especially under class-imbalanced conditions.", "date": "2025-07-29", "categories": []}
{"id": "2507.21778", "title": "AU-LLM: Micro-Expression Action Unit Detection via Enhanced LLM-Based Feature Fusion", "abstract": "The detection of micro-expression Action Units (AUs) is a formidable challenge in affective computing, pivotal for decoding subtle, involuntary human emotions. While Large Language Models (LLMs) demonstrate profound reasoning abilities, their application to the fine-grained, low-intensity domain of micro-expression AU detection remains unexplored. This paper pioneers this direction by introducing \\textbf{AU-LLM}, a novel framework that for the first time uses LLM to detect AUs in micro-expression datasets with subtle intensities and the scarcity of data. We specifically address the critical vision-language semantic gap, the \\textbf{Enhanced Fusion Projector (EFP)}. The EFP employs a Multi-Layer Perceptron (MLP) to intelligently fuse mid-level (local texture) and high-level (global semantics) visual features from a specialized 3D-CNN backbone into a single, information-dense token. This compact representation effectively empowers the LLM to perform nuanced reasoning over subtle facial muscle movements.Through extensive evaluations on the benchmark CASME II and SAMM datasets, including stringent Leave-One-Subject-Out (LOSO) and cross-domain protocols, AU-LLM establishes a new state-of-the-art, validating the significant potential and robustness of LLM-based reasoning for micro-expression analysis. The codes are available at https://github.com/ZS-liu-JLU/AU-LLMs.", "date": "2025-07-29", "categories": []}
{"id": "2210.16604", "title": "A Critical Reflection and Forward Perspective on Empathy and Natural Language Processing", "abstract": "We review the state of research on empathy in natural language processing and identify the following issues: (1) empathy definitions are absent or abstract, which (2) leads to low construct validity and reproducibility. Moreover, (3) emotional empathy is overemphasized, skewing our focus to a narrow subset of simplified tasks. We believe these issues hinder research progress and argue that current directions will benefit from a clear conceptualization that includes operationalizing cognitive empathy components. Our main objectives are to provide insight and guidance on empathy conceptualization for NLP research objectives and to encourage researchers to pursue the overlooked opportunities in this area, highly relevant, e.g., for clinical and educational sectors.", "date": "2022-10-29", "categories": []}
{"id": "2412.18988", "title": "MTCAE-DFER: Multi-Task Cascaded Autoencoder for Dynamic Facial Expression Recognition", "abstract": "This paper expands the cascaded network branch of the autoencoder-based multi-task learning (MTL) framework for dynamic facial expression recognition, namely Multi-Task Cascaded Autoencoder for Dynamic Facial Expression Recognition (MTCAE-DFER). MTCAE-DFER builds a plug-and-play cascaded decoder module, which is based on the Vision Transformer (ViT) architecture and employs the decoder concept of Transformer to reconstruct the multi-head attention module. The decoder output from the previous task serves as the query (Q), representing local dynamic features, while the Video Masked Autoencoder (VideoMAE) shared encoder output acts as both the key (K) and value (V), representing global dynamic features. This setup facilitates interaction between global and local dynamic features across related tasks. Additionally, this proposal aims to alleviate overfitting of complex large model. We utilize autoencoder-based multi-task cascaded learning approach to explore the impact of dynamic face detection and dynamic face landmark on dynamic facial expression recognition, which enhances the model's generalization ability. After we conduct extensive ablation experiments and comparison with state-of-the-art (SOTA) methods on various public datasets for dynamic facial expression recognition, the robustness of the MTCAE-DFER model and the effectiveness of global-local dynamic feature interaction among related tasks have been proven.", "date": "2024-12-25", "categories": []}
{"id": "2507.06821", "title": "HeLo: Heterogeneous Multi-Modal Fusion with Label Correlation for Emotion Distribution Learning", "abstract": "Multi-modal emotion recognition has garnered increasing attention as it plays a significant role in human-computer interaction (HCI) in recent years. Since different discrete emotions may exist at the same time, compared with single-class emotion recognition, emotion distribution learning (EDL) that identifies a mixture of basic emotions has gradually emerged as a trend. However, existing EDL methods face challenges in mining the heterogeneity among multiple modalities. Besides, rich semantic correlations across arbitrary basic emotions are not fully exploited. In this paper, we propose a multi-modal emotion distribution learning framework, named HeLo, aimed at fully exploring the heterogeneity and complementary information in multi-modal emotional data and label correlation within mixed basic emotions. Specifically, we first adopt cross-attention to effectively fuse the physiological data. Then, an optimal transport (OT)-based heterogeneity mining module is devised to mine the interaction and heterogeneity between the physiological and behavioral representations. To facilitate label correlation learning, we introduce a learnable label embedding optimized by correlation matrix alignment. Finally, the learnable label embeddings and label correlation matrices are integrated with the multi-modal representations through a novel label correlation-driven cross-attention mechanism for accurate emotion distribution learning. Experimental results on two publicly available datasets demonstrate the superiority of our proposed method in emotion distribution learning.", "date": "2025-07-09", "categories": []}
{"id": "2507.20557", "title": "FED-PsyAU: Privacy-Preserving Micro-Expression Recognition via Psychological AU Coordination and Dynamic Facial Motion Modeling", "abstract": "Micro-expressions (MEs) are brief, low-intensity, often localized facial expressions. They could reveal genuine emotions individuals may attempt to conceal, valuable in contexts like criminal interrogation and psychological counseling. However, ME recognition (MER) faces challenges, such as small sample sizes and subtle features, which hinder efficient modeling. Additionally, real-world applications encounter ME data privacy issues, leaving the task of enhancing recognition across settings under privacy constraints largely unexplored. To address these issues, we propose a FED-PsyAU research framework. We begin with a psychological study on the coordination of upper and lower facial action units (AUs) to provide structured prior knowledge of facial muscle dynamics. We then develop a DPK-GAT network that combines these psychological priors with statistical AU patterns, enabling hierarchical learning of facial motion features from regional to global levels, effectively enhancing MER performance. Additionally, our federated learning framework advances MER capabilities across multiple clients without data sharing, preserving privacy and alleviating the limited-sample issue for each client. Extensive experiments on commonly-used ME databases demonstrate the effectiveness of our approach.", "date": "2025-07-28", "categories": []}
{"id": "2507.20737", "title": "Multi-Masked Querying Network for Robust Emotion Recognition from Incomplete Multi-Modal Physiological Signals", "abstract": "Emotion recognition from physiological data is crucial for mental health assessment, yet it faces two significant challenges: incomplete multi-modal signals and interference from body movements and artifacts. This paper presents a novel Multi-Masked Querying Network (MMQ-Net) to address these issues by integrating multiple querying mechanisms into a unified framework. Specifically, it uses modality queries to reconstruct missing data from incomplete signals, category queries to focus on emotional state features, and interference queries to separate relevant information from noise. Extensive experiment results demonstrate the superior emotion recognition performance of MMQ-Net compared to existing approaches, particularly under high levels of data incompleteness.", "date": "2025-07-28", "categories": []}
{"id": "2507.21015", "title": "Learning Transferable Facial Emotion Representations from Large-Scale Semantically Rich Captions", "abstract": "Current facial emotion recognition systems are predominately trained to predict a fixed set of predefined categories or abstract dimensional values. This constrained form of supervision hinders generalization and applicability, as it reduces the rich and nuanced spectrum of emotions into oversimplified labels or scales. In contrast, natural language provides a more flexible, expressive, and interpretable way to represent emotions, offering a much broader source of supervision. Yet, leveraging semantically rich natural language captions as supervisory signals for facial emotion representation learning remains relatively underexplored, primarily due to two key challenges: 1) the lack of large-scale caption datasets with rich emotional semantics, and 2) the absence of effective frameworks tailored to harness such rich supervision. To this end, we introduce EmoCap100K, a large-scale facial emotion caption dataset comprising over 100,000 samples, featuring rich and structured semantic descriptions that capture both global affective states and fine-grained local facial behaviors. Building upon this dataset, we further propose EmoCapCLIP, which incorporates a joint global-local contrastive learning framework enhanced by a cross-modal guided positive mining module. This design facilitates the comprehensive exploitation of multi-level caption information while accommodating semantic similarities between closely related expressions. Extensive evaluations on over 20 benchmarks covering five tasks demonstrate the superior performance of our method, highlighting the promise of learning facial emotion representations from large-scale semantically rich captions. The code and data will be available at https://github.com/sunlicai/EmoCapCLIP.", "date": "2025-07-28", "categories": []}
{"id": "2506.06740", "title": "AI PsyRoom: Artificial Intelligence Platform for Segmented Yearning and Reactive Outcome Optimization Method", "abstract": "Psychological counseling faces huge challenges due to the growing demand for mental health services and the shortage of trained professionals. Large language models (LLMs) have shown potential to assist psychological counseling, especially in empathy and emotional support. However, existing models lack a deep understanding of emotions and are unable to generate personalized treatment plans based on fine-grained emotions. To address these shortcomings, we present AI PsyRoom, a multi-agent simulation framework designed to enhance psychological counseling by generating empathetic and emotionally nuanced conversations. By leveraging fine-grained emotion classification and a multi-agent framework, we construct a multi-agent PsyRoom A for dialogue reconstruction, generating a high-quality dialogue dataset EmoPsy, which contains 35 sub-emotions, 423 specific emotion scenarios, and 12,350 dialogues. We also propose PsyRoom B for generating personalized treatment plans. Quantitative evaluations demonstrate that AI PsyRoom significantly outperforms state-of-the-art methods, achieving 18% improvement in problem orientation, 23% in expression, 24% in Empathy, and 16% in interactive communication quality. The datasets and models are publicly available, providing a foundation for advancing AI-assisted psychological counseling research.", "date": "2025-06-07", "categories": []}
{"id": "2507.18929", "title": "MGHFT: Multi-Granularity Hierarchical Fusion Transformer for Cross-Modal Sticker Emotion Recognition", "abstract": "Although pre-trained visual models with text have demonstrated strong capabilities in visual feature extraction, sticker emotion understanding remains challenging due to its reliance on multi-view information, such as background knowledge and stylistic cues. To address this, we propose a novel multi-granularity hierarchical fusion transformer (MGHFT), with a multi-view sticker interpreter based on Multimodal Large Language Models. Specifically, inspired by the human ability to interpret sticker emotions from multiple views, we first use Multimodal Large Language Models to interpret stickers by providing rich textual context via multi-view descriptions. Then, we design a hierarchical fusion strategy to fuse the textual context into visual understanding, which builds upon a pyramid visual transformer to extract both global and local sticker features at multiple stages. Through contrastive learning and attention mechanisms, textual features are injected at different stages of the visual backbone, enhancing the fusion of global- and local-granularity visual semantics with textual guidance. Finally, we introduce a text-guided fusion attention mechanism to effectively integrate the overall multimodal features, enhancing semantic understanding. Extensive experiments on 2 public sticker emotion datasets demonstrate that MGHFT significantly outperforms existing sticker emotion recognition approaches, achieving higher accuracy and more fine-grained emotion recognition. Compared to the best pre-trained visual models, our MGHFT also obtains an obvious improvement, 5.4% on F1 and 4.0% on accuracy. The code is released at https://github.com/cccccj-03/MGHFT_ACMMM2025.", "date": "2025-07-25", "categories": []}
{"id": "2507.15401", "title": "Rethinking Occlusion in FER: A Semantic-Aware Perspective and Go Beyond", "abstract": "Facial expression recognition (FER) is a challenging task due to pervasive occlusion and dataset biases. Especially when facial information is partially occluded, existing FER models struggle to extract effective facial features, leading to inaccurate classifications. In response, we present ORSANet, which introduces the following three key contributions: First, we introduce auxiliary multi-modal semantic guidance to disambiguate facial occlusion and learn high-level semantic knowledge, which is two-fold: 1) we introduce semantic segmentation maps as dense semantics prior to generate semantics-enhanced facial representations; 2) we introduce facial landmarks as sparse geometric prior to mitigate intrinsic noises in FER, such as identity and gender biases. Second, to facilitate the effective incorporation of these two multi-modal priors, we customize a Multi-scale Cross-interaction Module (MCM) to adaptively fuse the landmark feature and semantics-enhanced representations within different scales. Third, we design a Dynamic Adversarial Repulsion Enhancement Loss (DARELoss) that dynamically adjusts the margins of ambiguous classes, further enhancing the model's ability to distinguish similar expressions. We further construct the first occlusion-oriented FER dataset to facilitate specialized robustness analysis on various real-world occlusion conditions, dubbed Occlu-FER. Extensive experiments on both public benchmarks and Occlu-FER demonstrate that our proposed ORSANet achieves SOTA recognition performance. Code is publicly available at https://github.com/Wenyuzhy/ORSANet-master.", "date": "2025-07-21", "categories": []}
{"id": "2507.17959", "title": "OPEN: A Benchmark Dataset and Baseline for Older Adult Patient Engagement Recognition in Virtual Rehabilitation Learning Environments", "abstract": "Engagement in virtual learning is essential for participant satisfaction, performance, and adherence, particularly in online education and virtual rehabilitation, where interactive communication plays a key role. Yet, accurately measuring engagement in virtual group settings remains a challenge. There is increasing interest in using artificial intelligence (AI) for large-scale, real-world, automated engagement recognition. While engagement has been widely studied in younger academic populations, research and datasets focused on older adults in virtual and telehealth learning settings remain limited. Existing methods often neglect contextual relevance and the longitudinal nature of engagement across sessions. This paper introduces OPEN (Older adult Patient ENgagement), a novel dataset supporting AI-driven engagement recognition. It was collected from eleven older adults participating in weekly virtual group learning sessions over six weeks as part of cardiac rehabilitation, producing over 35 hours of data, making it the largest dataset of its kind. To protect privacy, raw video is withheld; instead, the released data include facial, hand, and body joint landmarks, along with affective and behavioral features extracted from video. Annotations include binary engagement states, affective and behavioral labels, and context-type indicators, such as whether the instructor addressed the group or an individual. The dataset offers versions with 5-, 10-, 30-second, and variable-length samples. To demonstrate utility, multiple machine learning and deep learning models were trained, achieving engagement recognition accuracy of up to 81 percent. OPEN provides a scalable foundation for personalized engagement modeling in aging populations and contributes to broader engagement recognition research.", "date": "2025-07-23", "categories": []}
{"id": "2507.18026", "title": "Emotion Recognition from Skeleton Data: A Comprehensive Survey", "abstract": "Emotion recognition through body movements has emerged as a compelling and privacy-preserving alternative to traditional methods that rely on facial expressions or physiological signals. Recent advancements in 3D skeleton acquisition technologies and pose estimation algorithms have significantly enhanced the feasibility of emotion recognition based on full-body motion. This survey provides a comprehensive and systematic review of skeleton-based emotion recognition techniques. First, we introduce psychological models of emotion and examine the relationship between bodily movements and emotional expression. Next, we summarize publicly available datasets, highlighting the differences in data acquisition methods and emotion labeling strategies. We then categorize existing methods into posture-based and gait-based approaches, analyzing them from both data-driven and technical perspectives. In particular, we propose a unified taxonomy that encompasses four primary technical paradigms: Traditional approaches, Feat2Net, FeatFusionNet, and End2EndNet. Representative works within each category are reviewed and compared, with benchmarking results across commonly used datasets. Finally, we explore the extended applications of emotion recognition in mental health assessment, such as detecting depression and autism, and discuss the open challenges and future research directions in this rapidly evolving field.", "date": "2025-07-24", "categories": []}
{"id": "2405.08427", "title": "Impact of Stickers on Multimodal Sentiment and Intent in Social Media: A New Task, Dataset and Baseline", "abstract": "Stickers are increasingly used in social media to express sentiment and intent. Despite their significant impact on sentiment analysis and intent recognition, little research has been conducted in this area. To address this gap, we propose a new task: \\textbf{M}ultimodal chat \\textbf{S}entiment \\textbf{A}nalysis and \\textbf{I}ntent \\textbf{R}ecognition involving \\textbf{S}tickers (MSAIRS). Additionally, we introduce a novel multimodal dataset containing Chinese chat records and stickers excerpted from several mainstream social media platforms. Our dataset includes paired data with the same text but different stickers, the same sticker but different contexts, and various stickers consisting of the same images with different texts, allowing us to better understand the impact of stickers on chat sentiment and intent. We also propose an effective multimodal joint model, MMSAIR, featuring differential vector construction and cascaded attention mechanisms for enhanced multimodal fusion. Our experiments demonstrate the necessity and effectiveness of jointly modeling sentiment and intent, as they mutually reinforce each other's recognition accuracy. MMSAIR significantly outperforms traditional models and advanced MLLMs, demonstrating the challenge and uniqueness of sticker interpretation in social media. Our dataset and code are available on https://github.com/FakerBoom/MSAIRS-Dataset.", "date": "2024-05-14", "categories": []}
{"id": "2507.17450", "title": "Persistent Patterns in Eye Movements: A Topological Approach to Emotion Recognition", "abstract": "We present a topological pipeline for automated multiclass emotion recognition from eye-tracking data. Delay embeddings of gaze trajectories are analyzed using persistent homology. From the resulting persistence diagrams, we extract shape-based features such as mean persistence, maximum persistence, and entropy. A random forest classifier trained on these features achieves up to $75.6\\%$ accuracy on four emotion classes, which are the quadrants the Circumplex Model of Affect. The results demonstrate that persistence diagram geometry effectively encodes discriminative gaze dynamics, suggesting a promising topological approach for affective computing and human behavior analysis.", "date": "2025-07-23", "categories": []}
{"id": "2403.08255", "title": "Make Me Happier: Evoking Emotions Through Image Diffusion Models", "abstract": "Despite the rapid progress in image generation, emotional image editing remains under-explored. The semantics, context, and structure of an image can evoke emotional responses, making emotional image editing techniques valuable for various real-world applications, including treatment of psychological disorders, commercialization of products, and artistic design. First, we present a novel challenge of emotion-evoked image generation, aiming to synthesize images that evoke target emotions while retaining the semantics and structures of the original scenes. To address this challenge, we propose a diffusion model capable of effectively understanding and editing source images to convey desired emotions and sentiments. Moreover, due to the lack of emotion editing datasets, we provide a unique dataset consisting of 340,000 pairs of images and their emotion annotations. Furthermore, we conduct human psychophysics experiments and introduce a new evaluation metric to systematically benchmark all the methods. Experimental results demonstrate that our method surpasses all competitive baselines. Our diffusion model is capable of identifying emotional cues from original images, editing images that elicit desired emotions, and meanwhile, preserving the semantic structure of the original images. All code, model, and dataset are available at GitHub.", "date": "2024-03-13", "categories": []}
{"id": "2410.09767", "title": "LibEER: A Comprehensive Benchmark and Algorithm Library for EEG-based Emotion Recognition", "abstract": "EEG-based emotion recognition (EER) has gained significant attention due to its potential for understanding and analyzing human emotions. While recent advancements in deep learning techniques have substantially improved EER, the field lacks a convincing benchmark and comprehensive open-source libraries. This absence complicates fair comparisons between models and creates reproducibility challenges for practitioners, which collectively hinder progress. To address these issues, we introduce LibEER, a comprehensive benchmark and algorithm library designed to facilitate fair comparisons in EER. LibEER carefully selects popular and powerful baselines, harmonizes key implementation details across methods, and provides a standardized codebase in PyTorch. By offering a consistent evaluation framework with standardized experimental settings, LibEER enables unbiased assessments of seventeen representative deep learning models for EER across the six most widely used datasets. Additionally, we conduct a thorough, reproducible comparison of model performance and efficiency, providing valuable insights to guide researchers in the selection and design of EER models. Moreover, we make observations and in-depth analysis on the experiment results and identify current challenges in this community. We hope that our work will not only lower entry barriers for newcomers to EEG-based emotion recognition but also contribute to the standardization of research in this domain, fostering steady development. The library and source code are publicly available at https://github.com/XJTU-EEG/LibEER.", "date": "2024-10-13", "categories": []}
{"id": "2507.14173", "title": "Enhancing Generalization in PPG-Based Emotion Measurement with a CNN-TCN-LSTM Model", "abstract": "Human computer interaction has become integral to modern life, driven by advancements in machine learning technologies. Affective computing, in particular, has focused on systems that recognize, interpret, and respond to human emotions, often using wearable devices, which provide continuous data streams of physiological signals. Among various physiological signals, the photoplethysmogram (PPG) has gained prominence due to its ease of acquisition from widely available devices. However, the generalization of PPG-based emotion recognition models across individuals remains an unresolved challenge. This paper introduces a novel hybrid architecture that combines Convolutional Neural Networks (CNNs), Long Short-Term Memory networks (LSTMs), and Temporal Convolutional Networks (TCNs) to address this issue. The proposed model integrates the strengths of these architectures to improve robustness and generalization. Raw PPG signals are fed into the CNN for feature extraction. These features are processed separately by LSTM and TCN. The outputs from these components are concatenated to generate a final feature representation, which serves as the input for classifying valence and arousal, the primary dimensions of emotion. Experiments using the Photoplethysmogram Dataset for Emotional Analysis (PPGE) demonstrate that the proposed hybrid model achieves better model generalization than standalone CNN and LSTM architectures. Our results show that the proposed solution outperforms the state-of-the-art CNN architecture, as well as a CNN-LSTM model, in emotion recognition tasks with PPG signals. Using metrics such as Area Under the Curve (AUC) and F1 Score, we highlight the model's effectiveness in handling subject variability.", "date": "2025-07-10", "categories": []}
{"id": "2507.15470", "title": "FedMultiEmo: Real-Time Emotion Recognition via Multimodal Federated Learning", "abstract": "In-vehicle emotion recognition underpins adaptive driver-assistance systems and, ultimately, occupant safety. However, practical deployment is hindered by (i) modality fragility - poor lighting and occlusions degrade vision-based methods; (ii) physiological variability - heart-rate and skin-conductance patterns differ across individuals; and (iii) privacy risk - centralized training requires transmission of sensitive data. To address these challenges, we present FedMultiEmo, a privacy-preserving framework that fuses two complementary modalities at the decision level: visual features extracted by a Convolutional Neural Network from facial images, and physiological cues (heart rate, electrodermal activity, and skin temperature) classified by a Random Forest. FedMultiEmo builds on three key elements: (1) a multimodal federated learning pipeline with majority-vote fusion, (2) an end-to-end edge-to-cloud prototype on Raspberry Pi clients and a Flower server, and (3) a personalized Federated Averaging scheme that weights client updates by local data volume. Evaluated on FER2013 and a custom physiological dataset, the federated Convolutional Neural Network attains 77% accuracy, the Random Forest 74%, and their fusion 87%, matching a centralized baseline while keeping all raw data local. The developed system converges in 18 rounds, with an average round time of 120 seconds and a per-client memory footprint below 200 MB. These results indicate that FedMultiEmo offers a practical approach to real-time, privacy-aware emotion recognition in automotive settings.", "date": "2025-07-21", "categories": []}
{"id": "2507.15878", "title": "Salience Adjustment for Context-Based Emotion Recognition", "abstract": "Emotion recognition in dynamic social contexts requires an understanding of the complex interaction between facial expressions and situational cues. This paper presents a salience-adjusted framework for context-aware emotion recognition with Bayesian Cue Integration (BCI) and Visual-Language Models (VLMs) to dynamically weight facial and contextual information based on the expressivity of facial cues. We evaluate this approach using human annotations and automatic emotion recognition systems in prisoner's dilemma scenarios, which are designed to evoke emotional reactions. Our findings demonstrate that incorporating salience adjustment enhances emotion recognition performance, offering promising directions for future research to extend this framework to broader social contexts and multimodal applications.", "date": "2025-07-17", "categories": []}
{"id": "2507.15914", "title": "MSGM: A Multi-Scale Spatiotemporal Graph Mamba for EEG Emotion Recognition", "abstract": "EEG-based emotion recognition struggles with capturing multi-scale spatiotemporal dynamics and ensuring computational efficiency for real-time applications. Existing methods often oversimplify temporal granularity and spatial hierarchies, limiting accuracy. To overcome these challenges, we propose the Multi-Scale Spatiotemporal Graph Mamba (MSGM), a novel framework integrating multi-window temporal segmentation, bimodal spatial graph modeling, and efficient fusion via the Mamba architecture. By segmenting EEG signals across diverse temporal scales and constructing global-local graphs with neuroanatomical priors, MSGM effectively captures fine-grained emotional fluctuations and hierarchical brain connectivity. A multi-depth Graph Convolutional Network (GCN) and token embedding fusion module, paired with Mamba's state-space modeling, enable dynamic spatiotemporal interaction at linear complexity. Notably, with just one MSST-Mamba layer, MSGM surpasses leading methods in the field on the SEED, THU-EP, and FACED datasets, outperforming baselines in subject-independent emotion classification while achieving robust accuracy and millisecond-level inference on the NVIDIA Jetson Xavier NX.", "date": "2025-07-21", "categories": []}
{"id": "2507.14549", "title": "Synthesizing Images on Perceptual Boundaries of ANNs for Uncovering Human Perceptual Variability on Facial Expressions", "abstract": "A fundamental challenge in affective cognitive science is to develop models that accurately capture the relationship between external emotional stimuli and human internal experiences. While ANNs have demonstrated remarkable accuracy in facial expression recognition, their ability to model inter-individual differences in human perception remains underexplored. This study investigates the phenomenon of high perceptual variability-where individuals exhibit significant differences in emotion categorization even when viewing the same stimulus. Inspired by the similarity between ANNs and human perception, we hypothesize that facial expression samples that are ambiguous for ANN classifiers also elicit divergent perceptual judgments among human observers. To examine this hypothesis, we introduce a novel perceptual boundary sampling method to generate facial expression stimuli that lie along ANN decision boundaries. These ambiguous samples form the basis of the varEmotion dataset, constructed through large-scale human behavioral experiments. Our analysis reveals that these ANN-confusing stimuli also provoke heightened perceptual uncertainty in human participants, highlighting shared computational principles in emotion perception. Finally, by fine-tuning ANN representations using behavioral data, we achieve alignment between ANN predictions and both group-level and individual-level human perceptual patterns. Our findings establish a systematic link between ANN decision boundaries and human perceptual variability, offering new insights into personalized modeling of emotional interpretation.", "date": "2025-07-19", "categories": []}
{"id": "2507.14593", "title": "Coordinate Heart System: A Geometric Framework for Emotion Representation", "abstract": "This paper presents the Coordinate Heart System (CHS), a geometric framework for emotion representation in artificial intelligence applications. We position eight core emotions as coordinates on a unit circle, enabling mathematical computation of complex emotional states through coordinate mixing and vector operations. Our initial five-emotion model revealed significant coverage gaps in the emotion space, leading to the development of an eight-emotion system that provides complete geometric coverage with mathematical guarantees. The framework converts natural language input to emotion coordinates and supports real-time emotion interpolation through computational algorithms. The system introduces a re-calibrated stability parameter S in [0,1], which dynamically integrates emotional load, conflict resolution, and contextual drain factors. This stability model leverages advanced Large Language Model interpretation of textual cues and incorporates hybrid temporal tracking mechanisms to provide nuanced assessment of psychological well-being states. Our key contributions include: (i) mathematical proof demonstrating why five emotions are insufficient for complete geometric coverage, (ii) an eight-coordinate system that eliminates representational blind spots, (iii) novel algorithms for emotion mixing, conflict resolution, and distance calculation in emotion space, and (iv) a comprehensive computational framework for AI emotion recognition with enhanced multi-dimensional stability modeling. Experimental validation through case studies demonstrates the system's capability to handle emotionally conflicted states, contextual distress factors, and complex psychological scenarios that traditional categorical emotion models cannot adequately represent. This work establishes a new mathematical foundation for emotion modeling in artificial intelligence systems.", "date": "2025-07-19", "categories": []}
{"id": "2507.14887", "title": "MEKiT: Multi-source Heterogeneous Knowledge Injection Method via Instruction Tuning for Emotion-Cause Pair Extraction", "abstract": "Although large language models (LLMs) excel in text comprehension and generation, their performance on the Emotion-Cause Pair Extraction (ECPE) task, which requires reasoning ability, is often underperform smaller language model. The main reason is the lack of auxiliary knowledge, which limits LLMs' ability to effectively perceive emotions and reason causes. To address this issue, we propose a novel \\textbf{M}ulti-source h\\textbf{E}terogeneous \\textbf{K}nowledge \\textbf{i}njection me\\textbf{T}hod, MEKiT, which integrates heterogeneous internal emotional knowledge and external causal knowledge. Specifically, for these two distinct aspects and structures of knowledge, we apply the approaches of incorporating instruction templates and mixing data for instruction-tuning, which respectively facilitate LLMs in more comprehensively identifying emotion and accurately reasoning causes. Experimental results demonstrate that MEKiT provides a more effective and adaptable solution for the ECPE task, exhibiting an absolute performance advantage over compared baselines and dramatically improving the performance of LLMs on the ECPE task.", "date": "2025-07-20", "categories": []}
{"id": "2407.07046", "title": "CorMulT: A Semi-supervised Modality Correlation-aware Multimodal Transformer for Sentiment Analysis", "abstract": "Multimodal sentiment analysis is an active research area that combines multiple data modalities, e.g., text, image and audio, to analyze human emotions and benefits a variety of applications. Existing multimodal sentiment analysis methods can be classified as modality interaction-based methods, modality transformation-based methods and modality similarity-based methods. However, most of these methods highly rely on the strong correlations between modalities, and cannot fully uncover and utilize the correlations between modalities to enhance sentiment analysis. Therefore, these methods usually achieve bad performance for identifying the sentiment of multimodal data with weak correlations. To address this issue, we proposed a two-stage semi-supervised model termed Correlation-aware Multimodal Transformer (CorMulT) which consists pre-training stage and prediction stage. At the pre-training stage, a modality correlation contrastive learning module is designed to efficiently learn modality correlation coefficients between different modalities. At the prediction stage, the learned correlation coefficients are fused with modality representations to make the sentiment prediction. According to the experiments on the popular multimodal dataset CMU-MOSEI, CorMulT obviously surpasses state-of-the-art multimodal sentiment analysis methods.", "date": "2024-07-09", "categories": []}
{"id": "2409.18877", "title": "UniEmoX: Cross-modal Semantic-Guided Large-Scale Pretraining for Universal Scene Emotion Perception", "abstract": "Visual emotion analysis holds significant research value in both computer vision and psychology. However, existing methods for visual emotion analysis suffer from limited generalizability due to the ambiguity of emotion perception and the diversity of data scenarios. To tackle this issue, we introduce UniEmoX, a cross-modal semantic-guided large-scale pretraining framework. Inspired by psychological research emphasizing the inseparability of the emotional exploration process from the interaction between individuals and their environment, UniEmoX integrates scene-centric and person-centric low-level image spatial structural information, aiming to derive more nuanced and discriminative emotional representations. By exploiting the similarity between paired and unpaired image-text samples, UniEmoX distills rich semantic knowledge from the CLIP model to enhance emotional embedding representations more effectively. To the best of our knowledge, this is the first large-scale pretraining framework that integrates psychological theories with contemporary contrastive learning and masked image modeling techniques for emotion analysis across diverse scenarios. Additionally, we develop a visual emotional dataset titled Emo8. Emo8 samples cover a range of domains, including cartoon, natural, realistic, science fiction and advertising cover styles, covering nearly all common emotional scenes. Comprehensive experiments conducted on six benchmark datasets across two downstream tasks validate the effectiveness of UniEmoX. The source code is available at https://github.com/chincharles/u-emo.", "date": "2024-09-27", "categories": []}
{"id": "2507.14024", "title": "Moodifier: MLLM-Enhanced Emotion-Driven Image Editing", "abstract": "Bridging emotions and visual content for emotion-driven image editing holds great potential in creative industries, yet precise manipulation remains challenging due to the abstract nature of emotions and their varied manifestations across different contexts. We tackle this challenge with an integrated approach consisting of three complementary components. First, we introduce MoodArchive, an 8M+ image dataset with detailed hierarchical emotional annotations generated by LLaVA and partially validated by human evaluators. Second, we develop MoodifyCLIP, a vision-language model fine-tuned on MoodArchive to translate abstract emotions into specific visual attributes. Third, we propose Moodifier, a training-free editing model leveraging MoodifyCLIP and multimodal large language models (MLLMs) to enable precise emotional transformations while preserving content integrity. Our system works across diverse domains such as character expressions, fashion design, jewelry, and home d\\'ecor, enabling creators to quickly visualize emotional variations while preserving identity and structure. Extensive experimental evaluations show that Moodifier outperforms existing methods in both emotional accuracy and content preservation, providing contextually appropriate edits. By linking abstract emotions to concrete visual changes, our solution unlocks new possibilities for emotional content creation in real-world applications. We will release the MoodArchive dataset, MoodifyCLIP model, and make the Moodifier code and demo publicly available upon acceptance.", "date": "2025-07-18", "categories": []}
{"id": "2507.14084", "title": "The Emotion-Memory Link: Do Memorability Annotations Matter for Intelligent Systems?", "abstract": "Humans have a selective memory, remembering relevant episodes and forgetting the less relevant information. Possessing awareness of event memorability for a user could help intelligent systems in more accurate user modelling, especially for such applications as meeting support systems, memory augmentation, and meeting summarisation. Emotion recognition has been widely studied, since emotions are thought to signal moments of high personal relevance to users. The emotional experience of situations and their memorability have traditionally been considered to be closely tied to one another: moments that are experienced as highly emotional are considered to also be highly memorable. This relationship suggests that emotional annotations could serve as proxies for memorability. However, existing emotion recognition systems rely heavily on third-party annotations, which may not accurately represent the first-person experience of emotional relevance and memorability. This is why, in this study, we empirically examine the relationship between perceived group emotions (Pleasure-Arousal) and group memorability in the context of conversational interactions. Our investigation involves continuous time-based annotations of both emotions and memorability in dynamic, unstructured group settings, approximating conditions of real-world conversational AI applications such as online meeting support systems. Our results show that the observed relationship between affect and memorability annotations cannot be reliably distinguished from what might be expected under random chance. We discuss the implications of this surprising finding for the development and applications of Affective Computing technology. In addition, we contextualise our findings in broader discourses in the Affective Computing and point out important targets for future research efforts.", "date": "2025-07-18", "categories": []}
{"id": "2507.12889", "title": "Camera-based implicit mind reading by capturing higher-order semantic dynamics of human gaze within environmental context", "abstract": "Emotion recognition,as a step toward mind reading,seeks to infer internal states from external cues.Most existing methods rely on explicit signals-such as facial expressions,speech,or gestures-that reflect only bodily responses and overlook the influence of environmental context.These cues are often voluntary,easy to mask,and insufficient for capturing deeper,implicit emotions. Physiological signal-based approaches offer more direct access to internal states but require complex sensors that compromise natural behavior and limit scalability.Gaze-based methods typically rely on static fixation analysis and fail to capture the rich,dynamic interactions between gaze and the environment,and thus cannot uncover the deep connection between emotion and implicit behavior.To address these limitations,we propose a novel camera-based,user-unaware emotion recognition approach that integrates gaze fixation patterns with environmental semantics and temporal dynamics.Leveraging standard HD cameras,our method unobtrusively captures users'eye appearance and head movements in natural settings-without the need for specialized hardware or active user participation.From these visual cues,the system estimates gaze trajectories over time and space, providing the basis for modeling the spatial, semantic,and temporal dimensions of gaze behavior. This allows us to capture the dynamic interplay between visual attention and the surrounding environment,revealing that emotions are not merely physiological responses but complex outcomes of human-environment interactions.The proposed approach enables user-unaware,real-time,and continuous emotion recognition,offering high generalizability and low deployment cost.", "date": "2025-07-17", "categories": []}
{"id": "2404.07078", "title": "VLLMs Provide Better Context for Emotion Understanding Through Common Sense Reasoning", "abstract": "Recognising emotions in context involves identifying an individual's apparent emotions while considering contextual cues from the surrounding scene. Previous approaches to this task have typically designed explicit scene-encoding architectures or incorporated external scene-related information, such as captions. However, these methods often utilise limited contextual information or rely on intricate training pipelines to decouple noise from relevant information. In this work, we leverage the capabilities of Vision-and-Large-Language Models (VLLMs) to enhance in-context emotion classification in a more straightforward manner. Our proposed method follows a simple yet effective two-stage approach. First, we prompt VLLMs to generate natural language descriptions of the subject's apparent emotion in relation to the visual context. Second, the descriptions, along with the visual input, are used to train a transformer-based architecture that fuses text and visual features before the final classification task. This method not only simplifies the training process but also significantly improves performance. Experimental results demonstrate that the textual descriptions effectively guide the model to constrain the noisy visual input, allowing our fused architecture to outperform individual modalities. Our approach achieves state-of-the-art performance across three datasets, BoLD, EMOTIC, and CAER-S, without bells and whistles. The code will be made publicly available on github: https://github.com/NickyFot/EmoCommonSense.git", "date": "2024-04-10", "categories": []}
{"id": "2410.00166", "title": "EEG Emotion Copilot: Optimizing Lightweight LLMs for Emotional EEG Interpretation with Assisted Medical Record Generation", "abstract": "In the fields of affective computing (AC) and brain-machine interface (BMI), the analysis of physiological and behavioral signals to discern individual emotional states has emerged as a critical research frontier. While deep learning-based approaches have made notable strides in EEG emotion recognition, particularly in feature extraction and pattern recognition, significant challenges persist in achieving end-to-end emotion computation, including real-time processing, individual adaptation, and seamless user interaction. This paper presents the EEG Emotion Copilot, a system optimizing a lightweight large language model (LLM) with 0.5B parameters operating in a local setting, which first recognizes emotional states directly from EEG signals, subsequently generates personalized diagnostic and treatment suggestions, and finally supports the automation of assisted electronic medical records. Specifically, we demonstrate the critical techniques in the novel data structure of prompt, model pruning and fine-tuning training, and deployment strategies aiming at improving real-time performance and computational efficiency. Extensive experiments show that our optimized lightweight LLM-based copilot achieves an enhanced intuitive interface for participant interaction, superior accuracy of emotion recognition and assisted electronic medical records generation, in comparison to such models with similar scale parameters or large-scale parameters such as 1.5B, 1.8B, 3B and 7B. In summary, through these efforts, the proposed copilot is expected to advance the application of AC in the medical domain, offering innovative solution to mental health monitoring. The codes will be released at https://github.com/NZWANG/EEG_Emotion_Copilot.", "date": "2024-09-30", "categories": []}
{"id": "2505.06110", "title": "Multimodal Sentiment Analysis on CMU-MOSEI Dataset using Transformer-based Models", "abstract": "This project performs multimodal sentiment analysis using the CMU-MOSEI dataset, using transformer-based models with early fusion to integrate text, audio, and visual modalities. We employ BERT-based encoders for each modality, extracting embeddings that are concatenated before classification. The model achieves strong performance, with 97.87% 7-class accuracy and a 0.9682 F1-score on the test set, demonstrating the effectiveness of early fusion in capturing cross-modal interactions. The training utilized Adam optimization (lr=1e-4), dropout (0.3), and early stopping to ensure generalization and robustness. Results highlight the superiority of transformer architectures in modeling multimodal sentiment, with a low MAE (0.1060) indicating precise sentiment intensity prediction. Future work may compare fusion strategies or enhance interpretability. This approach utilizes multimodal learning by effectively combining linguistic, acoustic, and visual cues for sentiment analysis.", "date": "2025-05-09", "categories": []}
{"id": "2507.10861", "title": "Visually grounded emotion regulation via diffusion models and user-driven reappraisal", "abstract": "Cognitive reappraisal is a key strategy in emotion regulation, involving reinterpretation of emotionally charged stimuli to alter affective responses. Despite its central role in clinical and cognitive science, real-world reappraisal interventions remain cognitively demanding, abstract, and primarily verbal. This reliance on higher-order cognitive and linguistic processes is often impaired in individuals with trauma or depression, limiting the effectiveness of standard approaches. Here, we propose a novel, visually based augmentation of cognitive reappraisal by integrating large-scale text-to-image diffusion models into the emotional regulation process. Specifically, we introduce a system in which users reinterpret emotionally negative images via spoken reappraisals, which are transformed into supportive, emotionally congruent visualizations using stable diffusion models with a fine-tuned IP-adapter. This generative transformation visually instantiates users' reappraisals while maintaining structural similarity to the original stimuli, externalizing and reinforcing regulatory intent. To test this approach, we conducted a within-subject experiment (N = 20) using a modified cognitive emotion regulation (CER) task. Participants reappraised or described aversive images from the International Affective Picture System (IAPS), with or without AI-generated visual feedback. Results show that AI-assisted reappraisal significantly reduced negative affect compared to both non-AI and control conditions. Further analyses reveal that sentiment alignment between participant reappraisals and generated images correlates with affective relief, suggesting that multimodal coherence enhances regulatory efficacy. These findings demonstrate that generative visual input can support cogitive reappraisal and open new directions at the intersection of generative AI, affective computing, and therapeutic technology.", "date": "2025-07-14", "categories": []}
{"id": "2507.10895", "title": "Commuting Distance Regularization for Timescale-Dependent Label Inconsistency in EEG Emotion Recognition", "abstract": "In this work, we address the often-overlooked issue of Timescale Dependent Label Inconsistency (TsDLI) in training neural network models for EEG-based human emotion recognition. To mitigate TsDLI and enhance model generalization and explainability, we propose two novel regularization strategies: Local Variation Loss (LVL) and Local-Global Consistency Loss (LGCL). Both methods incorporate classical mathematical principles--specifically, functions of bounded variation and commute-time distances--within a graph theoretic framework. Complementing our regularizers, we introduce a suite of new evaluation metrics that better capture the alignment between temporally local predictions and their associated global emotion labels. We validate our approach through comprehensive experiments on two widely used EEG emotion datasets, DREAMER and DEAP, across a range of neural architectures including LSTM and transformer-based models. Performance is assessed using five distinct metrics encompassing both quantitative accuracy and qualitative consistency. Results consistently show that our proposed methods outperform state-of-the-art baselines, delivering superior aggregate performance and offering a principled trade-off between interpretability and predictive power under label inconsistency. Notably, LVL achieves the best aggregate rank across all benchmarked backbones and metrics, while LGCL frequently ranks the second, highlighting the effectiveness of our framework.", "date": "2025-07-15", "categories": []}
{"id": "2507.09248", "title": "AGCD-Net: Attention Guided Context Debiasing Network for Emotion Recognition", "abstract": "Context-aware emotion recognition (CAER) enhances affective computing in real-world scenarios, but traditional methods often suffer from context bias-spurious correlation between background context and emotion labels (e.g. associating ``garden'' with ``happy''). In this paper, we propose \\textbf{AGCD-Net}, an Attention Guided Context Debiasing model that introduces \\textit{Hybrid ConvNeXt}, a novel convolutional encoder that extends the ConvNeXt backbone by integrating Spatial Transformer Network and Squeeze-and-Excitation layers for enhanced feature recalibration. At the core of AGCD-Net is the Attention Guided - Causal Intervention Module (AG-CIM), which applies causal theory, perturbs context features, isolates spurious correlations, and performs an attention-driven correction guided by face features to mitigate context bias. Experimental results on the CAER-S dataset demonstrate the effectiveness of AGCD-Net, achieving state-of-the-art performance and highlighting the importance of causal debiasing for robust emotion recognition in complex settings.", "date": "2025-07-12", "categories": []}
{"id": "2507.08167", "title": "Emotion Detection in Older Adults Using Physiological Signals from Wearable Sensors", "abstract": "Emotion detection in older adults is crucial for understanding their cognitive and emotional well-being, especially in hospital and assisted living environments. In this work, we investigate an edge-based, non-obtrusive approach to emotion identification that uses only physiological signals obtained via wearable sensors. Our dataset includes data from 40 older individuals. Emotional states were obtained using physiological signals from the Empatica E4 and Shimmer3 GSR+ wristband and facial expressions were recorded using camera-based emotion recognition with the iMotion's Facial Expression Analysis (FEA) module. The dataset also contains twelve emotion categories in terms of relative intensities. We aim to study how well emotion recognition can be accomplished using simply physiological sensor data, without the requirement for cameras or intrusive facial analysis. By leveraging classical machine learning models, we predict the intensity of emotional responses based on physiological signals. We achieved the highest 0.782 r2 score with the lowest 0.0006 MSE on the regression task. This method has significant implications for individuals with Alzheimer's Disease and Related Dementia (ADRD), as well as veterans coping with Post-Traumatic Stress Disorder (PTSD) or other cognitive impairments. Our results across multiple classical regression models validate the feasibility of this method, paving the way for privacy-preserving and efficient emotion recognition systems in real-world settings.", "date": "2025-07-10", "categories": []}
{"id": "2507.08175", "title": "Emotion Recognition in Older Adults with Quantum Machine Learning and Wearable Sensors", "abstract": "We investigate the feasibility of inferring emotional states exclusively from physiological signals, thereby presenting a privacy-preserving alternative to conventional facial recognition techniques. We conduct a performance comparison of classical machine learning algorithms and hybrid quantum machine learning (QML) methods with a quantum kernel-based model. Our results indicate that the quantum-enhanced SVM surpasses classical counterparts in classification performance across all emotion categories, even when trained on limited datasets. The F1 scores over all classes are over 80% with around a maximum of 36% improvement in the recall values. The integration of wearable sensor data with quantum machine learning not only enhances accuracy and robustness but also facilitates unobtrusive emotion recognition. This methodology holds promise for populations with impaired communication abilities, such as individuals with Alzheimer's Disease and Related Dementias (ADRD) and veterans with Post-Traumatic Stress Disorder (PTSD). The findings establish an early foundation for passive emotional monitoring in clinical and assisted living conditions.", "date": "2025-07-10", "categories": []}
{"id": "2507.07678", "title": "Action Unit Enhance Dynamic Facial Expression Recognition", "abstract": "Dynamic Facial Expression Recognition(DFER) is a rapidly evolving field of research that focuses on the recognition of time-series facial expressions. While previous research on DFER has concentrated on feature learning from a deep learning perspective, we put forward an AU-enhanced Dynamic Facial Expression Recognition architecture, namely AU-DFER, that incorporates AU-expression knowledge to enhance the effectiveness of deep learning modeling. In particular, the contribution of the Action Units(AUs) to different expressions is quantified, and a weight matrix is designed to incorporate a priori knowledge. Subsequently, the knowledge is integrated with the learning outcomes of a conventional deep learning network through the introduction of AU loss. The design is incorporated into the existing optimal model for dynamic expression recognition for the purpose of validation. Experiments are conducted on three recent mainstream open-source approaches to DFER on the principal datasets in this field. The results demonstrate that the proposed architecture outperforms the state-of-the-art(SOTA) methods without the need for additional arithmetic and generally produces improved results. Furthermore, we investigate the potential of AU loss function redesign to address data label imbalance issues in established dynamic expression datasets. To the best of our knowledge, this is the first attempt to integrate quantified AU-expression knowledge into various DFER models. We also devise strategies to tackle label imbalance, or minor class problems. Our findings suggest that employing a diverse strategy of loss function design can enhance the effectiveness of DFER. This underscores the criticality of addressing data imbalance challenges in mainstream datasets within this domain. The source code is available at https://github.com/Cross-Innovation-Lab/AU-DFER.", "date": "2025-07-10", "categories": []}
{"id": "2507.05624", "title": "ADMC: Attention-based Diffusion Model for Missing Modalities Feature Completion", "abstract": "Multimodal emotion and intent recognition is essential for automated human-computer interaction, It aims to analyze users' speech, text, and visual information to predict their emotions or intent. One of the significant challenges is that missing modalities due to sensor malfunctions or incomplete data. Traditional methods that attempt to reconstruct missing information often suffer from over-coupling and imprecise generation processes, leading to suboptimal outcomes. To address these issues, we introduce an Attention-based Diffusion model for Missing Modalities feature Completion (ADMC). Our framework independently trains feature extraction networks for each modality, preserving their unique characteristics and avoiding over-coupling. The Attention-based Diffusion Network (ADN) generates missing modality features that closely align with authentic multimodal distribution, enhancing performance across all missing-modality scenarios. Moreover, ADN's cross-modal generation offers improved recognition even in full-modality contexts. Our approach achieves state-of-the-art results on the IEMOCAP and MIntRec benchmarks, demonstrating its effectiveness in both missing and complete modality scenarios.", "date": "2025-07-08", "categories": []}
{"id": "2507.05918", "title": "Few-shot text-based emotion detection", "abstract": "This paper describes the approach of the Unibuc - NLP team in tackling the SemEval 2025 Workshop, Task 11: Bridging the Gap in Text-Based Emotion Detection. We mainly focused on experiments using large language models (Gemini, Qwen, DeepSeek) with either few-shot prompting or fine-tuning. With our final system, for the multi-label emotion detection track (track A), we got an F1-macro of $0.7546$ (26/96 teams) for the English subset, $0.1727$ (35/36 teams) for the Portuguese (Mozambican) subset and $0.325$ (\\textbf{1}/31 teams) for the Emakhuwa subset.", "date": "2025-07-08", "categories": []}
{"id": "2412.08049", "title": "EmoVerse: Exploring Multimodal Large Language Models for Sentiment and Emotion Understanding", "abstract": "Sentiment and emotion understanding are essential to applications such as human-computer interaction and depression detection. While Multimodal Large Language Models (MLLMs) demonstrate robust general capabilities, they face considerable challenges in the field of affective computing, particularly in detecting subtle facial expressions and handling complex emotion-related tasks, such as emotion reason inference and understanding emotions in long-context scenarios. Furthermore, there is a lack of a unified MLLM that can effectively handle both sentiment and emotion-related tasks. To address these challenges, we explore multi-task training strategies for MLLMs in affective computing and introduce Emotion Universe (EmoVerse), an MLLM designed to handle a broad spectrum of sentiment and emotion-related tasks. In addition, EmoVerse is capable of deeply analyzing the underlying causes of emotional states. We also introduce the Affective Multitask (AMT) Dataset, which supports multimodal sentiment analysis, multimodal emotion recognition, facial expression recognition, emotion reason inference, and emotion cause-pair extraction tasks. Extensive experiments demonstrate that EmoVerse outperforms existing methods, achieving state-of-the-art results in sentiment and emotion-related tasks. The code is available at https://github.com/liaolea/EmoVerse.", "date": "2024-12-11", "categories": []}
{"id": "2503.03335", "title": "iNews: A Multimodal Dataset for Modeling Personalized Affective Responses to News", "abstract": "Understanding how individuals perceive and react to information is fundamental for advancing social and behavioral sciences and developing human-centered AI systems. Current approaches often lack the granular data needed to model these personalized responses, relying instead on aggregated labels that obscure the rich variability driven by individual differences. We introduce iNews, a novel large-scale dataset specifically designed to facilitate the modeling of personalized affective responses to news content. Our dataset comprises annotations from 291 demographically diverse UK participants across 2,899 multimodal Facebook news posts from major UK outlets, with an average of 5.18 annotators per sample. For each post, annotators provide multifaceted labels including valence, arousal, dominance, discrete emotions, content relevance judgments, sharing likelihood, and modality importance ratings. Crucially, we collect comprehensive annotator persona information covering demographics, personality, media trust, and consumption patterns, which explain 15.2% of annotation variance - substantially higher than existing NLP datasets. Incorporating this information yields a 7% accuracy gain in zero-shot prediction and remains beneficial even with 32-shot in-context learning. iNews opens new possibilities for research in LLM personalization, subjectivity, affective computing, and human behavior simulation.", "date": "2025-03-05", "categories": []}
{"id": "2504.12151", "title": "Towards Explainable Fusion and Balanced Learning in Multimodal Sentiment Analysis", "abstract": "Multimodal Sentiment Analysis (MSA) faces two critical challenges: the lack of interpretability in the decision logic of multimodal fusion and modality imbalance caused by disparities in inter-modal information density. To address these issues, we propose KAN-MCP, a novel framework that integrates the interpretability of Kolmogorov-Arnold Networks (KAN) with the robustness of the Multimodal Clean Pareto (MCPareto) framework. First, KAN leverages its univariate function decomposition to achieve transparent analysis of cross-modal interactions. This structural design allows direct inspection of feature transformations without relying on external interpretation tools, thereby ensuring both high expressiveness and interpretability. Second, the proposed MCPareto enhances robustness by addressing modality imbalance and noise interference. Specifically, we introduce the Dimensionality Reduction and Denoising Modal Information Bottleneck (DRD-MIB) method, which jointly denoises and reduces feature dimensionality. This approach provides KAN with discriminative low-dimensional inputs to reduce the modeling complexity of KAN while preserving critical sentiment-related information. Furthermore, MCPareto dynamically balances gradient contributions across modalities using the purified features output by DRD-MIB, ensuring lossless transmission of auxiliary signals and effectively alleviating modality imbalance. This synergy of interpretability and robustness not only achieves superior performance on benchmark datasets such as CMU-MOSI, CMU-MOSEI, and CH-SIMS v2 but also offers an intuitive visualization interface through KAN's interpretable architecture. Our code is released on https://github.com/LuoMSen/KAN-MCP.", "date": "2025-04-16", "categories": []}
{"id": "2507.02205", "title": "Team RAS in 9th ABAW Competition: Multimodal Compound Expression Recognition Approach", "abstract": "Compound Expression Recognition (CER), a subfield of affective computing, aims to detect complex emotional states formed by combinations of basic emotions. In this work, we present a novel zero-shot multimodal approach for CER that combines six heterogeneous modalities into a single pipeline: static and dynamic facial expressions, scene and label matching, scene context, audio, and text. Unlike previous approaches relying on task-specific training data, our approach uses zero-shot components, including Contrastive Language-Image Pretraining (CLIP)-based label matching and Qwen-VL for semantic scene understanding. We further introduce a Multi-Head Probability Fusion (MHPF) module that dynamically weights modality-specific predictions, followed by a Compound Expressions (CE) transformation module that uses Pair-Wise Probability Aggregation (PPA) and Pair-Wise Feature Similarity Aggregation (PFSA) methods to produce interpretable compound emotion outputs. Evaluated under multi-corpus training, the proposed approach shows F1 scores of 46.95% on AffWild2, 49.02% on Acted Facial Expressions in The Wild (AFEW), and 34.85% on C-EXPR-DB via zero-shot testing, which is comparable to the results of supervised approaches trained on target data. This demonstrates the effectiveness of the proposed approach for capturing CE without domain adaptation. The source code is publicly available.", "date": "2025-07-02", "categories": []}
{"id": "2507.04708", "title": "Why We Feel What We Feel: Joint Detection of Emotions and Their Opinion Triggers in E-commerce", "abstract": "Customer reviews on e-commerce platforms capture critical affective signals that drive purchasing decisions. However, no existing research has explored the joint task of emotion detection and explanatory span identification in e-commerce reviews - a crucial gap in understanding what triggers customer emotional responses. To bridge this gap, we propose a novel joint task unifying Emotion detection and Opinion Trigger extraction (EOT), which explicitly models the relationship between causal text spans (opinion triggers) and affective dimensions (emotion categories) grounded in Plutchik's theory of 8 primary emotions. In the absence of labeled data, we introduce EOT-X, a human-annotated collection of 2,400 reviews with fine-grained emotions and opinion triggers. We evaluate 23 Large Language Models (LLMs) and present EOT-DETECT, a structured prompting framework with systematic reasoning and self-reflection. Our framework surpasses zero-shot and chain-of-thought techniques, across e-commerce domains.", "date": "2025-07-07", "categories": []}
{"id": "2507.02537", "title": "Are You Listening to Me? Fine-Tuning Chatbots for Empathetic Dialogue", "abstract": "Conversational agents have made significant progress since ELIZA, expanding their role across various domains, including healthcare, education, and customer service. As these agents become increasingly integrated into daily human interactions, the need for emotional intelligence, particularly empathetic listening, becomes increasingly essential. In this study, we explore how Large Language Models (LLMs) respond when tasked with generating emotionally rich interactions. Starting from a small dataset manually crafted by an expert to reflect empathic behavior, we extended the conversations using two LLMs: ChatGPT and Gemini. We analyzed the emotional progression of the dialogues using both sentiment analysis (via VADER) and expert assessments. While the generated conversations often mirrored the intended emotional structure, human evaluation revealed important differences in the perceived empathy and coherence of the responses. These findings suggest that emotion modeling in dialogues requires not only structural alignment in the expressed emotions but also qualitative depth, highlighting the importance of combining automated and humancentered methods in the development of emotionally competent agents.", "date": "2025-07-03", "categories": []}
{"id": "2507.01081", "title": "AI-guided digital intervention with physiological monitoring reduces intrusive memories after experimental trauma", "abstract": "Trauma prevalence is vast globally. Evidence-based digital treatments can help, but most require human guidance. Human guides provide tailored instructions and responsiveness to internal cognitive states, but limit scalability. Can generative AI and neurotechnology provide a scalable alternative? Here we test ANTIDOTE, combining AI guidance and pupillometry to automatically deliver and monitor an evidence-based digital treatment, specifically the Imagery Competing Task Intervention (ICTI), to reduce intrusive memories after psychological trauma. One hundred healthy volunteers were exposed to videos of traumatic events and randomly assigned to an intervention or active control condition. As predicted, intervention participants reported significantly fewer intrusive memories over the following week. Post-hoc assessment against clinical rubrics confirmed the AI guide delivered the intervention successfully. Additionally, pupil size tracked intervention engagement and predicted symptom reduction, providing a candidate biomarker of intervention effectiveness. These findings open a path toward rigorous AI-guided digital interventions that can scale to trauma prevalence.", "date": "2025-07-01", "categories": []}
{"id": "2507.00055", "title": "Leveraging Unlabeled Audio-Visual Data in Speech Emotion Recognition using Knowledge Distillation", "abstract": "Voice interfaces integral to the human-computer interaction systems can benefit from speech emotion recognition (SER) to customize responses based on user emotions. Since humans convey emotions through multi-modal audio-visual cues, developing SER systems using both the modalities is beneficial. However, collecting a vast amount of labeled data for their development is expensive. This paper proposes a knowledge distillation framework called LightweightSER (LiSER) that leverages unlabeled audio-visual data for SER, using large teacher models built on advanced speech and face representation models. LiSER transfers knowledge regarding speech emotions and facial expressions from the teacher models to lightweight student models. Experiments conducted on two benchmark datasets, RAVDESS and CREMA-D, demonstrate that LiSER can reduce the dependence on extensive labeled datasets for SER tasks.", "date": "2025-06-26", "categories": []}
{"id": "2507.00161", "title": "Designing an Adaptive Storytelling Platform to Promote Civic Education in Politically Polarized Learning Environments", "abstract": "Political polarization undermines democratic civic education by exacerbating identity-based resistance to opposing viewpoints. Emerging AI technologies offer new opportunities to advance interventions that reduce polarization and promote political open-mindedness. We examined novel design strategies that leverage adaptive and emotionally-responsive civic narratives that may sustain students' emotional engagement in stories, and in turn, promote perspective-taking toward members of political out-groups. Drawing on theories from political psychology and narratology, we investigate how affective computing techniques can support three storytelling mechanisms: transportation into a story world, identification with characters, and interaction with the storyteller. Using a design-based research (DBR) approach, we iteratively developed and refined an AI-mediated Digital Civic Storytelling (AI-DCS) platform. Our prototype integrates facial emotion recognition and attention tracking to assess users' affective and attentional states in real time. Narrative content is organized around pre-structured story outlines, with beat-by-beat language adaptation implemented via GPT-4, personalizing linguistic tone to sustain students' emotional engagement in stories that center political perspectives different from their own. Our work offers a foundation for AI-supported, emotionally-sensitive strategies that address affective polarization while preserving learner autonomy. We conclude with implications for civic education interventions, algorithmic literacy, and HCI challenges associated with AI dialogue management and affect-adaptive learning environments.", "date": "2025-06-30", "categories": []}
{"id": "2507.00586", "title": "Context-Aware Academic Emotion Dataset and Benchmark", "abstract": "Academic emotion analysis plays a crucial role in evaluating students' engagement and cognitive states during the learning process. This paper addresses the challenge of automatically recognizing academic emotions through facial expressions in real-world learning environments. While significant progress has been made in facial expression recognition for basic emotions, academic emotion recognition remains underexplored, largely due to the scarcity of publicly available datasets. To bridge this gap, we introduce RAER, a novel dataset comprising approximately 2,700 video clips collected from around 140 students in diverse, natural learning contexts such as classrooms, libraries, laboratories, and dormitories, covering both classroom sessions and individual study. Each clip was annotated independently by approximately ten annotators using two distinct sets of academic emotion labels with varying granularity, enhancing annotation consistency and reliability. To our knowledge, RAER is the first dataset capturing diverse natural learning scenarios. Observing that annotators naturally consider context cues-such as whether a student is looking at a phone or reading a book-alongside facial expressions, we propose CLIP-CAER (CLIP-based Context-aware Academic Emotion Recognition). Our method utilizes learnable text prompts within the vision-language model CLIP to effectively integrate facial expression and context cues from videos. Experimental results demonstrate that CLIP-CAER substantially outperforms state-of-the-art video-based facial expression recognition methods, which are primarily designed for basic emotions, emphasizing the crucial role of context in accurately recognizing academic emotions. Project page: https://zgsfer.github.io/CAER", "date": "2025-07-01", "categories": []}
{"id": "2407.21536", "title": "Tracing Intricate Cues in Dialogue: Joint Graph Structure and Sentiment Dynamics for Multimodal Emotion Recognition", "abstract": "Multimodal emotion recognition in conversation (MERC) has garnered substantial research attention recently. Existing MERC methods face several challenges: (1) they fail to fully harness direct inter-modal cues, possibly leading to less-than-thorough cross-modal modeling; (2) they concurrently extract information from the same and different modalities at each network layer, potentially triggering conflicts from the fusion of multi-source data; (3) they lack the agility required to detect dynamic sentimental changes, perhaps resulting in inaccurate classification of utterances with abrupt sentiment shifts. To address these issues, a novel approach named GraphSmile is proposed for tracking intricate emotional cues in multimodal dialogues. GraphSmile comprises two key components, i.e., GSF and SDP modules. GSF ingeniously leverages graph structures to alternately assimilate inter-modal and intra-modal emotional dependencies layer by layer, adequately capturing cross-modal cues while effectively circumventing fusion conflicts. SDP is an auxiliary task to explicitly delineate the sentiment dynamics between utterances, promoting the model's ability to distinguish sentimental discrepancies. GraphSmile is effortlessly applied to multimodal sentiment analysis in conversation (MSAC), thus enabling simultaneous execution of MERC and MSAC tasks. Empirical results on multiple benchmarks demonstrate that GraphSmile can handle complex emotional and sentimental patterns, significantly outperforming baseline models.", "date": "2024-07-31", "categories": []}
{"id": "2506.18915", "title": "Automatic Depression Assessment using Machine Learning: A Comprehensive Survey", "abstract": "Depression is a common mental illness across current human society. Traditional depression assessment relying on inventories and interviews with psychologists frequently suffer from subjective diagnosis results, slow and expensive diagnosis process as well as lack of human resources. Since there is a solid evidence that depression is reflected by various human internal brain activities and external expressive behaviours, early traditional machine learning (ML) and advanced deep learning (DL) models have been widely explored for human behaviour-based automatic depression assessment (ADA) since 2012. However, recent ADA surveys typically only focus on a limited number of human behaviour modalities. Despite being used as a theoretical basis for developing ADA approaches, existing ADA surveys lack a comprehensive review and summary of multi-modal depression-related human behaviours. To bridge this gap, this paper specifically summarises depression-related human behaviours across a range of modalities (e.g. the human brain, verbal language and non-verbal audio/facial/body behaviours). We focus on conducting an up-to-date and comprehensive survey of ML-based ADA approaches for learning depression cues from these behaviours as well as discussing and comparing their distinctive features and limitations. In addition, we also review existing ADA competitions and datasets, identify and discuss the main challenges and opportunities to provide further research directions for future ADA researchers.", "date": "2025-06-09", "categories": []}
{"id": "2501.12050", "title": "Representation Learning with Parameterised Quantum Circuits for Advancing Speech Emotion Recognition", "abstract": "Quantum machine learning (QML) offers a promising avenue for advancing representation learning in complex signal domains. In this study, we investigate the use of parameterised quantum circuits (PQCs) for speech emotion recognition (SER) a challenging task due to the subtle temporal variations and overlapping affective states in vocal signals. We propose a hybrid quantum classical architecture that integrates PQCs into a conventional convolutional neural network (CNN), leveraging quantum properties such as superposition and entanglement to enrich emotional feature representations. Experimental evaluations on three benchmark datasets IEMOCAP, RECOLA, and MSP-IMPROV demonstrate that our hybrid model achieves improved classification performance relative to a purely classical CNN baseline, with over 50% reduction in trainable parameters. This work provides early evidence of the potential for QML to enhance emotion recognition and lays the foundation for future quantum-enabled affective computing systems.", "date": "2025-01-21", "categories": []}
{"id": "2506.20204", "title": "Affective Priming Score: A Data-Driven Method to Detect Priming in Sequential Datasets", "abstract": "Affective priming exemplifies the challenge of ambiguity in affective computing. While the community has largely addressed this issue from a label-based perspective, identifying data points in the sequence affected by the priming effect, the impact of priming on data itself, particularly in physiological signals, remains underexplored. Data affected by priming can lead to misclassifications when used in learning models. This study proposes the Affective Priming Score (APS), a data-driven method to detect data points influenced by the priming effect. The APS assigns a score to each data point, quantifying the extent to which it is affected by priming. To validate this method, we apply it to the SEED and SEED-VII datasets, which contain sufficient transitions between emotional events to exhibit priming effects. We train models with the same configuration using both the original data and priming-free sequences. The misclassification rate is significantly reduced when using priming-free sequences compared to the original data. This work contributes to the broader challenge of ambiguity by identifying and mitigating priming effects at the data level, enhancing model robustness, and offering valuable insights for the design and collection of affective computing datasets.", "date": "2025-06-25", "categories": []}
{"id": "2404.15615", "title": "M3D: Manifold-based Domain Adaptation with Dynamic Distribution for Non-Deep Transfer Learning in Cross-subject and Cross-session EEG-based Emotion Recognition", "abstract": "Emotion decoding using Electroencephalography (EEG)-based affective brain-computer interfaces (aBCIs) plays a crucial role in affective computing but is limited by challenges such as EEG's non-stationarity, individual variability, and the high cost of large labeled datasets. While deep learning methods are effective, they require extensive computational resources and large data volumes, limiting their practical application. To overcome these issues, we propose Manifold-based Domain Adaptation with Dynamic Distribution (M3D), a lightweight, non-deep transfer learning framework. M3D consists of four key modules: manifold feature transformation, dynamic distribution alignment, classifier learning, and ensemble learning. The data is mapped to an optimal Grassmann manifold space, enabling dynamic alignment of source and target domains. This alignment is designed to prioritize both marginal and conditional distributions, improving adaptation efficiency across diverse datasets. In classifier learning, the principle of structural risk minimization is applied to build robust classification models. Additionally, dynamic distribution alignment iteratively refines the classifier. The ensemble learning module aggregates classifiers from different optimization stages to leverage diversity and enhance prediction accuracy. M3D is evaluated on two EEG emotion recognition datasets using two validation protocols (cross-subject single-session and cross-subject cross-session) and a clinical EEG dataset for Major Depressive Disorder (MDD). Experimental results show that M3D outperforms traditional non-deep learning methods with a 4.47% average improvement and achieves deep learning-level performance with reduced data and computational requirements, demonstrating its potential for real-world aBCI applications.", "date": "2024-04-24", "categories": []}
{"id": "2406.08809", "title": "Are We There Yet? A Brief Survey of Music Emotion Prediction Datasets, Models and Outstanding Challenges", "abstract": "Deep learning models for music have advanced drastically in recent years, but how good are machine learning models at capturing emotion, and what challenges are researchers facing? In this paper, we provide a comprehensive overview of the available music-emotion datasets and discuss evaluation standards as well as competitions in the field. We also offer a brief overview of various types of music emotion prediction models that have been built over the years, providing insights into the diverse approaches within the field. Through this examination, we highlight the challenges that persist in accurately capturing emotion in music, including issues related to dataset quality, annotation consistency, and model generalization. Additionally, we explore the impact of different modalities, such as audio, MIDI, and physiological signals, on the effectiveness of emotion prediction models. Through this examination, we identify persistent challenges in music emotion recognition (MER), including issues related to dataset quality, the ambiguity in emotion labels, and the difficulties of cross-dataset generalization. We argue that future advancements in MER require standardized benchmarks, larger and more diverse datasets, and improved model interpretability. Recognizing the dynamic nature of this field, we have complemented our findings with an accompanying GitHub repository. This repository contains a comprehensive list of music emotion datasets and recent predictive models.", "date": "2024-06-13", "categories": []}
{"id": "2506.19280", "title": "Emotion Detection on User Front-Facing App Interfaces for Enhanced Schedule Optimization: A Machine Learning Approach", "abstract": "Human-Computer Interaction (HCI) has evolved significantly to incorporate emotion recognition capabilities, creating unprecedented opportunities for adaptive and personalized user experiences. This paper explores the integration of emotion detection into calendar applications, enabling user interfaces to dynamically respond to users' emotional states and stress levels, thereby enhancing both productivity and engagement. We present and evaluate two complementary approaches to emotion detection: a biometric-based method utilizing heart rate (HR) data extracted from electrocardiogram (ECG) signals processed through Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) neural networks to predict the emotional dimensions of Valence, Arousal, and Dominance; and a behavioral method analyzing computer activity through multiple machine learning models to classify emotions based on fine-grained user interactions such as mouse movements, clicks, and keystroke patterns. Our comparative analysis, from real-world datasets, reveals that while both approaches demonstrate effectiveness, the computer activity-based method delivers superior consistency and accuracy, particularly for mouse-related interactions, which achieved approximately 90\\% accuracy. Furthermore, GRU networks outperformed LSTM models in the biometric approach, with Valence prediction reaching 84.38\\% accuracy.", "date": "2025-06-24", "categories": []}
{"id": "2503.11290", "title": "EmoAgent: A Multi-Agent Framework for Diverse Affective Image Manipulation", "abstract": "Affective Image Manipulation (AIM) aims to alter visual elements within an image to evoke specific emotional responses from viewers. However, existing AIM approaches rely on rigid \\emph{one-to-one} mappings between emotions and visual cues, making them ill-suited for the inherently subjective and diverse ways in which humans perceive and express emotion.To address this, we introduce a novel task setting termed \\emph{Diverse AIM (D-AIM)}, aiming to generate multiple visually distinct yet emotionally consistent image edits from a single source image and target emotion. We propose \\emph{EmoAgent}, the first multi-agent framework tailored specifically for D-AIM. EmoAgent explicitly decomposes the manipulation process into three specialized phases executed by collaborative agents: a Planning Agent that generates diverse emotional editing strategies, an Editing Agent that precisely executes these strategies, and a Critic Agent that iteratively refines the results to ensure emotional accuracy. This collaborative design empowers EmoAgent to model \\emph{one-to-many} emotion-to-visual mappings, enabling semantically diverse and emotionally faithful edits.Extensive quantitative and qualitative evaluations demonstrate that EmoAgent substantially outperforms state-of-the-art approaches in both emotional fidelity and semantic diversity, effectively generating multiple distinct visual edits that convey the same target emotion.", "date": "2025-03-14", "categories": []}
{"id": "2506.16388", "title": "HausaNLP at SemEval-2025 Task 11: Hausa Text Emotion Detection", "abstract": "This paper presents our approach to multi-label emotion detection in Hausa, a low-resource African language, for SemEval Track A. We fine-tuned AfriBERTa, a transformer-based model pre-trained on African languages, to classify Hausa text into six emotions: anger, disgust, fear, joy, sadness, and surprise. Our methodology involved data preprocessing, tokenization, and model fine-tuning using the Hugging Face Trainer API. The system achieved a validation accuracy of 74.00%, with an F1-score of 73.50%, demonstrating the effectiveness of transformer-based models for emotion detection in low-resource languages.", "date": "2025-06-19", "categories": []}
{"id": "2506.18716", "title": "Multi-modal Anchor Gated Transformer with Knowledge Distillation for Emotion Recognition in Conversation", "abstract": "Emotion Recognition in Conversation (ERC) aims to detect the emotions of individual utterances within a conversation. Generating efficient and modality-specific representations for each utterance remains a significant challenge. Previous studies have proposed various models to integrate features extracted using different modality-specific encoders. However, they neglect the varying contributions of modalities to this task and introduce high complexity by aligning modalities at the frame level. To address these challenges, we propose the Multi-modal Anchor Gated Transformer with Knowledge Distillation (MAGTKD) for the ERC task. Specifically, prompt learning is employed to enhance textual modality representations, while knowledge distillation is utilized to strengthen representations of weaker modalities. Furthermore, we introduce a multi-modal anchor gated transformer to effectively integrate utterance-level representations across modalities. Extensive experiments on the IEMOCAP and MELD datasets demonstrate the effectiveness of knowledge distillation in enhancing modality representations and achieve state-of-the-art performance in emotion recognition. Our code is available at: https://github.com/JieLi-dd/MAGTKD.", "date": "2025-06-23", "categories": []}
{"id": "2506.16448", "title": "Consumer-friendly EEG-based Emotion Recognition System: A Multi-scale Convolutional Neural Network Approach", "abstract": "EEG is a non-invasive, safe, and low-risk method to record electrophysiological signals inside the brain. Especially with recent technology developments like dry electrodes, consumer-grade EEG devices, and rapid advances in machine learning, EEG is commonly used as a resource for automatic emotion recognition. With the aim to develop a deep learning model that can perform EEG-based emotion recognition in a real-life context, we propose a novel approach to utilize multi-scale convolutional neural networks to accomplish such tasks. By implementing feature extraction kernels with many ratio coefficients as well as a new type of kernel that learns key information from four separate areas of the brain, our model consistently outperforms the state-of-the-art TSception model in predicting valence, arousal, and dominance scores across many performance evaluation metrics.", "date": "2025-06-19", "categories": []}
{"id": "2506.17191", "title": "Facial Landmark Visualization and Emotion Recognition Through Neural Networks", "abstract": "Emotion recognition from facial images is a crucial task in human-computer interaction, enabling machines to learn human emotions through facial expressions. Previous studies have shown that facial images can be used to train deep learning models; however, most of these studies do not include a through dataset analysis. Visualizing facial landmarks can be challenging when extracting meaningful dataset insights; to address this issue, we propose facial landmark box plots, a visualization technique designed to identify outliers in facial datasets. Additionally, we compare two sets of facial landmark features: (i) the landmarks' absolute positions and (ii) their displacements from a neutral expression to the peak of an emotional expression. Our results indicate that a neural network achieves better performance than a random forest classifier.", "date": "2025-06-20", "categories": []}
{"id": "2405.12661", "title": "EmoEdit: Evoking Emotions through Image Manipulation", "abstract": "Affective Image Manipulation (AIM) seeks to modify user-provided images to evoke specific emotional responses. This task is inherently complex due to its twofold objective: significantly evoking the intended emotion, while preserving the original image composition. Existing AIM methods primarily adjust color and style, often failing to elicit precise and profound emotional shifts. Drawing on psychological insights, we introduce EmoEdit, which extends AIM by incorporating content modifications to enhance emotional impact. Specifically, we first construct EmoEditSet, a large-scale AIM dataset comprising 40,120 paired data through emotion attribution and data construction. To make existing generative models emotion-aware, we design the Emotion adapter and train it using EmoEditSet. We further propose an instruction loss to capture the semantic variations in data pairs. Our method is evaluated both qualitatively and quantitatively, demonstrating superior performance compared to existing state-of-the-art techniques. Additionally, we showcase the portability of our Emotion adapter to other diffusion-based models, enhancing their emotion knowledge with diverse semantics.", "date": "2024-05-21", "categories": []}
{"id": "2407.06060", "title": "MERGE -- A Bimodal Audio-Lyrics Dataset for Static Music Emotion Recognition", "abstract": "The Music Emotion Recognition (MER) field has seen steady developments in recent years, with contributions from feature engineering, machine learning, and deep learning. The landscape has also shifted from audio-centric systems to bimodal ensembles that combine audio and lyrics. However, a lack of public, sizable and quality-controlled bimodal databases has hampered the development and improvement of bimodal audio-lyrics systems. This article proposes three new audio, lyrics, and bimodal MER research datasets, collectively referred to as MERGE, which were created using a semi-automatic approach. To comprehensively assess the proposed datasets and establish a baseline for benchmarking, we conducted several experiments for each modality, using feature engineering, machine learning, and deep learning methodologies. Additionally, we propose and validate fixed train-validation-test splits. The obtained results confirm the viability of the proposed datasets, achieving the best overall result of 81.74\\% F1-score for bimodal classification.", "date": "2024-07-08", "categories": []}
{"id": "2505.19802", "title": "GraphAU-Pain: Graph-based Action Unit Representation for Pain Intensity Estimation", "abstract": "Understanding pain-related facial behaviors is essential for digital healthcare in terms of effective monitoring, assisted diagnostics, and treatment planning, particularly for patients unable to communicate verbally. Existing data-driven methods of detecting pain from facial expressions are limited due to interpretability and severity quantification. To this end, we propose GraphAU-Pain, leveraging a graph-based framework to model facial Action Units (AUs) and their interrelationships for pain intensity estimation. AUs are represented as graph nodes, with co-occurrence relationships as edges, enabling a more expressive depiction of pain-related facial behaviors. By utilizing a relational graph neural network, our framework offers improved interpretability and significant performance gains. Experiments conducted on the publicly available UNBC dataset demonstrate the effectiveness of the GraphAU-Pain, achieving an F1-score of 66.21% and accuracy of 87.61% in pain intensity estimation.", "date": "2025-05-26", "categories": []}
{"id": "2506.13978", "title": "AI shares emotion with humans across languages and cultures", "abstract": "Effective and safe human-machine collaboration requires the regulated and meaningful exchange of emotions between humans and artificial intelligence (AI). Current AI systems based on large language models (LLMs) can provide feedback that makes people feel heard. Yet it remains unclear whether LLMs represent emotion in language as humans do, or whether and how the emotional tone of their output can be controlled. We assess human-AI emotional alignment across linguistic-cultural groups and model-families, using interpretable LLM features translated from concept-sets for over twenty nuanced emotion categories (including six basic emotions). Our analyses reveal that LLM-derived emotion spaces are structurally congruent with human perception, underpinned by the fundamental affective dimensions of valence and arousal. Furthermore, these emotion-related features also accurately predict large-scale behavioural data on word ratings along these two core dimensions, reflecting both universal and language-specific patterns. Finally, by leveraging steering vectors derived solely from human-centric emotion concepts, we show that model expressions can be stably and naturally modulated across distinct emotion categories, which provides causal evidence that human emotion concepts can be used to systematically induce LLMs to produce corresponding affective states when conveying content. These findings suggest AI not only shares emotional representations with humans but its affective outputs can be precisely guided using psychologically grounded emotion concepts.", "date": "2025-06-11", "categories": []}
{"id": "2405.20072", "title": "Faces of the Mind: Unveiling Mental Health States Through Facial Expressions in 11,427 Adolescents", "abstract": "Mood disorders such as depression and anxiety often manifest through facial expressions, but existing machine learning algorithms designed to assess these disorders have been hindered by small datasets and limited real-world applicability. To address this gap, we analyzed facial videos of 11,427 participants - a dataset two orders of magnitude larger than those used in previous studies - including standardized facial expression videos and psychological assessments of depression, anxiety, and stress. However, scaling up the dataset introduces significant challenges due to increased symptom heterogeneity, making it difficult for models to learn accurate representations. To address this, we introduced the Symptom Discrepancy Index (SDI), a novel metric for quantifying dataset heterogeneity caused by variability in individual symptoms among samples with identical total scores. By removing the 10% most heterogeneous cases as identified by the SDI, we raised the F1 scores of all models from approximately 50% to 80%. Notably, comparable performance gains were observed within both the retained and excluded subsets. These findings demonstrate symptom heterogeneity, not model capacity, as the principal bottleneck in large scale automated assessment and provide a general solution that is readily applicable to other psychometric data sets.", "date": "2024-05-30", "categories": []}
{"id": "2506.12325", "title": "GSDNet: Revisiting Incomplete Multimodal-Diffusion from Graph Spectrum Perspective for Conversation Emotion Recognition", "abstract": "Multimodal emotion recognition in conversations (MERC) aims to infer the speaker's emotional state by analyzing utterance information from multiple sources (i.e., video, audio, and text). Compared with unimodality, a more robust utterance representation can be obtained by fusing complementary semantic information from different modalities. However, the modality missing problem severely limits the performance of MERC in practical scenarios. Recent work has achieved impressive performance on modality completion using graph neural networks and diffusion models, respectively. This inspires us to combine these two dimensions through the graph diffusion model to obtain more powerful modal recovery capabilities. Unfortunately, existing graph diffusion models may destroy the connectivity and local structure of the graph by directly adding Gaussian noise to the adjacency matrix, resulting in the generated graph data being unable to retain the semantic and topological information of the original graph. To this end, we propose a novel Graph Spectral Diffusion Network (GSDNet), which maps Gaussian noise to the graph spectral space of missing modalities and recovers the missing data according to its original distribution. Compared with previous graph diffusion methods, GSDNet only affects the eigenvalues of the adjacency matrix instead of destroying the adjacency matrix directly, which can maintain the global topological information and important spectral features during the diffusion process. Extensive experiments have demonstrated that GSDNet achieves state-of-the-art emotion recognition performance in various modality loss scenarios.", "date": "2025-06-14", "categories": []}
{"id": "2506.12437", "title": "Feeling Machines: Ethics, Culture, and the Rise of Emotional AI", "abstract": "This paper explores the growing presence of emotionally responsive artificial intelligence through a critical and interdisciplinary lens. Bringing together the voices of early-career researchers from multiple fields, it explores how AI systems that simulate or interpret human emotions are reshaping our interactions in areas such as education, healthcare, mental health, caregiving, and digital life. The analysis is structured around four central themes: the ethical implications of emotional AI, the cultural dynamics of human-machine interaction, the risks and opportunities for vulnerable populations, and the emerging regulatory, design, and technical considerations. The authors highlight the potential of affective AI to support mental well-being, enhance learning, and reduce loneliness, as well as the risks of emotional manipulation, over-reliance, misrepresentation, and cultural bias. Key challenges include simulating empathy without genuine understanding, encoding dominant sociocultural norms into AI systems, and insufficient safeguards for individuals in sensitive or high-risk contexts. Special attention is given to children, elderly users, and individuals with mental health challenges, who may interact with AI in emotionally significant ways. However, there remains a lack of cognitive or legal protections which are necessary to navigate such engagements safely. The report concludes with ten recommendations, including the need for transparency, certification frameworks, region-specific fine-tuning, human oversight, and longitudinal research. A curated supplementary section provides practical tools, models, and datasets to support further work in this domain.", "date": "2025-06-14", "categories": []}
{"id": "2506.12486", "title": "DinoCompanion: An Attachment-Theory Informed Multimodal Robot for Emotionally Responsive Child-AI Interaction", "abstract": "Children's emotional development fundamentally relies on secure attachment relationships, yet current AI companions lack the theoretical foundation to provide developmentally appropriate emotional support. We introduce DinoCompanion, the first attachment-theory-grounded multimodal robot for emotionally responsive child-AI interaction. We address three critical challenges in child-AI systems: the absence of developmentally-informed AI architectures, the need to balance engagement with safety, and the lack of standardized evaluation frameworks for attachment-based capabilities. Our contributions include: (i) a multimodal dataset of 128 caregiver-child dyads containing 125,382 annotated clips with paired preference-risk labels, (ii) CARPO (Child-Aware Risk-calibrated Preference Optimization), a novel training objective that maximizes engagement while applying epistemic-uncertainty-weighted risk penalties, and (iii) AttachSecure-Bench, a comprehensive evaluation benchmark covering ten attachment-centric competencies with strong expert consensus (\\k{appa}=0.81). DinoCompanion achieves state-of-the-art performance (57.15%), outperforming GPT-4o (50.29%) and Claude-3.7-Sonnet (53.43%), with exceptional secure base behaviors (72.99%, approaching human expert levels of 78.4%) and superior attachment risk detection (69.73%). Ablations validate the critical importance of multimodal fusion, uncertainty-aware risk modeling, and hierarchical memory for coherent, emotionally attuned interactions.", "date": "2025-06-14", "categories": []}
{"id": "2506.12848", "title": "Towards Fine-Grained Emotion Understanding via Skeleton-Based Micro-Gesture Recognition", "abstract": "We present our solution to the MiGA Challenge at IJCAI 2025, which aims to recognize micro-gestures (MGs) from skeleton sequences for the purpose of hidden emotion understanding. MGs are characterized by their subtlety, short duration, and low motion amplitude, making them particularly challenging to model and classify. We adopt PoseC3D as the baseline framework and introduce three key enhancements: (1) a topology-aware skeleton representation specifically designed for the iMiGUE dataset to better capture fine-grained motion patterns; (2) an improved temporal processing strategy that facilitates smoother and more temporally consistent motion modeling; and (3) the incorporation of semantic label embeddings as auxiliary supervision to improve the model generalization. Our method achieves a Top-1 accuracy of 67.01\\% on the iMiGUE test set. As a result of these contributions, our approach ranks third on the official MiGA Challenge leaderboard. The source code is available at \\href{https://github.com/EGO-False-Sleep/Miga25_track1}{https://github.com/EGO-False-Sleep/Miga25\\_track1}.", "date": "2025-06-15", "categories": []}
{"id": "2506.10010", "title": "Multimodal Emotion Coupling via Speech-to-Facial and Bodily Gestures in Dyadic Interaction", "abstract": "Human emotional expression emerges through coordinated vocal, facial, and gestural signals. While speech face alignment is well established, the broader dynamics linking emotionally expressive speech to regional facial and hand motion remains critical for gaining a deeper insight into how emotional and behavior cues are communicated in real interactions. Further modulating the coordination is the structure of conversational exchange like sequential turn taking, which creates stable temporal windows for multimodal synchrony, and simultaneous speech, often indicative of high arousal moments, disrupts this alignment and impacts emotional clarity. Understanding these dynamics enhances realtime emotion detection by improving the accuracy of timing and synchrony across modalities in both human interactions and AI systems. This study examines multimodal emotion coupling using region specific motion capture from dyadic interactions in the IEMOCAP corpus. Speech features included low level prosody, MFCCs, and model derived arousal, valence, and categorical emotions (Happy, Sad, Angry, Neutral), aligned with 3D facial and hand marker displacements. Expressive activeness was quantified through framewise displacement magnitudes, and speech to gesture prediction mapped speech features to facial and hand movements. Nonoverlapping speech consistently elicited greater activeness particularly in the lower face and mouth. Sadness showed increased expressivity during nonoverlap, while anger suppressed gestures during overlaps. Predictive mapping revealed highest accuracy for prosody and MFCCs in articulatory regions while arousal and valence had lower and more context sensitive correlations. Notably, hand speech synchrony was enhanced under low arousal and overlapping speech, but not for valence.", "date": "2025-05-08", "categories": []}
{"id": "2506.10154", "title": "Analyzing Emotions in Bangla Social Media Comments Using Machine Learning and LIME", "abstract": "Research on understanding emotions in written language continues to expand, especially for understudied languages with distinctive regional expressions and cultural features, such as Bangla. This study examines emotion analysis using 22,698 social media comments from the EmoNoBa dataset. For language analysis, we employ machine learning models: Linear SVM, KNN, and Random Forest with n-gram data from a TF-IDF vectorizer. We additionally investigated how PCA affects the reduction of dimensionality. Moreover, we utilized a BiLSTM model and AdaBoost to improve decision trees. To make our machine learning models easier to understand, we used LIME to explain the predictions of the AdaBoost classifier, which uses decision trees. With the goal of advancing sentiment analysis in languages with limited resources, our work examines various techniques to find efficient techniques for emotion identification in Bangla.", "date": "2025-06-11", "categories": []}
{"id": "2506.10334", "title": "Using Vision Language Models to Detect Students' Academic Emotion through Facial Expressions", "abstract": "Students' academic emotions significantly influence their social behavior and learning performance. Traditional approaches to automatically and accurately analyze these emotions have predominantly relied on supervised machine learning algorithms. However, these models often struggle to generalize across different contexts, necessitating repeated cycles of data collection, annotation, and training. The emergence of Vision-Language Models (VLMs) offers a promising alternative, enabling generalization across visual recognition tasks through zero-shot prompting without requiring fine-tuning. This study investigates the potential of VLMs to analyze students' academic emotions via facial expressions in an online learning environment. We employed two VLMs, Llama-3.2-11B-Vision-Instruct and Qwen2.5-VL-7B-Instruct, to analyze 5,000 images depicting confused, distracted, happy, neutral, and tired expressions using zero-shot prompting. Preliminary results indicate that both models demonstrate moderate performance in academic facial expression recognition, with Qwen2.5-VL-7B-Instruct outperforming Llama-3.2-11B-Vision-Instruct. Notably, both models excel in identifying students' happy emotions but fail to detect distracted behavior. Additionally, Qwen2.5-VL-7B-Instruct exhibits relatively high performance in recognizing students' confused expressions, highlighting its potential for practical applications in identifying content that causes student confusion.", "date": "2025-06-12", "categories": []}
{"id": "2506.10452", "title": "Towards Robust Multimodal Emotion Recognition under Missing Modalities and Distribution Shifts", "abstract": "Recent advancements in Multimodal Emotion Recognition (MER) face challenges in addressing both modality missing and Out-Of-Distribution (OOD) data simultaneously. Existing methods often rely on specific models or introduce excessive parameters, which limits their practicality. To address these issues, we propose a novel robust MER framework, Causal Inference Distiller (CIDer), and introduce a new task, Random Modality Feature Missing (RMFM), to generalize the definition of modality missing. CIDer integrates two key components: a Model-Specific Self-Distillation (MSSD) module and a Model-Agnostic Causal Inference (MACI) module. MSSD enhances robustness under the RMFM task through a weight-sharing self-distillation approach applied across low-level features, attention maps, and high-level representations. Additionally, a Word-level Self-aligned Attention Module (WSAM) reduces computational complexity, while a Multimodal Composite Transformer (MCT) facilitates efficient multimodal fusion. To tackle OOD challenges, MACI employs a tailored causal graph to mitigate label and language biases using a Multimodal Causal Module (MCM) and fine-grained counterfactual texts. Notably, MACI can independently enhance OOD generalization with minimal additional parameters. Furthermore, we also introduce the new repartitioned MER OOD datasets. Experimental results demonstrate that CIDer achieves robust performance in both RMFM and OOD scenarios, with fewer parameters and faster training compared to state-of-the-art methods. The implementation of this work is publicly accessible at https://github.com/gw-zhong/CIDer.", "date": "2025-06-12", "categories": []}
{"id": "2506.10930", "title": "Developing a High-performance Framework for Speech Emotion Recognition in Naturalistic Conditions Challenge for Emotional Attribute Prediction", "abstract": "Speech emotion recognition (SER) in naturalistic conditions presents a significant challenge for the speech processing community. Challenges include disagreement in labeling among annotators and imbalanced data distributions. This paper presents a reproducible framework that achieves superior (top 1) performance in the Emotion Recognition in Naturalistic Conditions Challenge (IS25-SER Challenge) - Task 2, evaluated on the MSP-Podcast dataset. Our system is designed to tackle the aforementioned challenges through multimodal learning, multi-task learning, and imbalanced data handling. Specifically, our best system is trained by adding text embeddings, predicting gender, and including ``Other'' (O) and ``No Agreement'' (X) samples in the training set. Our system's results secured both first and second places in the IS25-SER Challenge, and the top performance was achieved by a simple two-system ensemble.", "date": "2025-06-12", "categories": []}
{"id": "2408.08979", "title": "Electroencephalogram Emotion Recognition via AUC Maximization", "abstract": "Imbalanced datasets pose significant challenges in areas including neuroscience, cognitive science, and medical diagnostics, where accurately detecting minority classes is essential for robust model performance. This study addresses the issue of class imbalance, using the `Liking' label in the DEAP dataset as an example. Such imbalances are often overlooked by prior research, which typically focuses on the more balanced arousal and valence labels and predominantly uses accuracy metrics to measure model performance. To tackle this issue, we adopt numerical optimization techniques aimed at maximizing the area under the curve (AUC), thus enhancing the detection of underrepresented classes. Our approach, which begins with a linear classifier, is compared against traditional linear classifiers, including logistic regression and support vector machines (SVM). Our method significantly outperforms these models, increasing recall from 41.6\\% to 79.7\\% and improving the F1-score from 0.506 to 0.632. These results highlight the efficacy of AUC maximization via numerical optimization in managing imbalanced datasets, providing an effective solution for enhancing predictive accuracy in detecting minority but crucial classes in out-of-sample datasets.", "date": "2024-08-16", "categories": []}
{"id": "2506.07086", "title": "Representation Decomposition for Learning Similarity and Contrastness Across Modalities for Affective Computing", "abstract": "Multi-modal affective computing aims to automatically recognize and interpret human attitudes from diverse data sources such as images and text, thereby enhancing human-computer interaction and emotion understanding. Existing approaches typically rely on unimodal analysis or straightforward fusion of cross-modal information that fail to capture complex and conflicting evidence presented across different modalities. In this paper, we propose a novel LLM-based approach for affective computing that explicitly deconstructs visual and textual representations into shared (modality-invariant) and modality-specific components. Specifically, our approach firstly encodes and aligns input modalities using pre-trained multi-modal encoders, then employs a representation decomposition framework to separate common emotional content from unique cues, and finally integrates these decomposed signals via an attention mechanism to form a dynamic soft prompt for a multi-modal LLM. Extensive experiments on three representative tasks for affective computing, namely, multi-modal aspect-based sentiment analysis, multi-modal emotion analysis, and hateful meme detection, demonstrate the effectiveness of our approach, which consistently outperforms strong baselines and state-of-the-art models.", "date": "2025-06-08", "categories": []}
{"id": "2506.04409", "title": "Empaths at SemEval-2025 Task 11: Retrieval-Augmented Approach to Perceived Emotions Prediction", "abstract": "This paper describes EmoRAG, a system designed to detect perceived emotions in text for SemEval-2025 Task 11, Subtask A: Multi-label Emotion Detection. We focus on predicting the perceived emotions of the speaker from a given text snippet, labeling it with emotions such as joy, sadness, fear, anger, surprise, and disgust. Our approach does not require additional model training and only uses an ensemble of models to predict emotions. EmoRAG achieves results comparable to the best performing systems, while being more efficient, scalable, and easier to implement.", "date": "2025-06-04", "categories": []}
{"id": "2506.04652", "title": "EMO-Debias: Benchmarking Gender Debiasing Techniques in Multi-Label Speech Emotion Recognition", "abstract": "Speech emotion recognition (SER) systems often exhibit gender bias. However, the effectiveness and robustness of existing debiasing methods in such multi-label scenarios remain underexplored. To address this gap, we present EMO-Debias, a large-scale comparison of 13 debiasing methods applied to multi-label SER. Our study encompasses techniques from pre-processing, regularization, adversarial learning, biased learners, and distributionally robust optimization. Experiments conducted on acted and naturalistic emotion datasets, using WavLM and XLSR representations, evaluate each method under conditions of gender imbalance. Our analysis quantifies the trade-offs between fairness and accuracy, identifying which approaches consistently reduce gender performance gaps without compromising overall model performance. The findings provide actionable insights for selecting effective debiasing strategies and highlight the impact of dataset distributions.", "date": "2025-06-05", "categories": []}
{"id": "2506.05095", "title": "FG 2025 TrustFAA: the First Workshop on Towards Trustworthy Facial Affect Analysis: Advancing Insights of Fairness, Explainability, and Safety (TrustFAA)", "abstract": "With the increasing prevalence and deployment of Emotion AI-powered facial affect analysis (FAA) tools, concerns about the trustworthiness of these systems have become more prominent. This first workshop on \"Towards Trustworthy Facial Affect Analysis: Advancing Insights of Fairness, Explainability, and Safety (TrustFAA)\" aims to bring together researchers who are investigating different challenges in relation to trustworthiness-such as interpretability, uncertainty, biases, and privacy-across various facial affect analysis tasks, including macro/ micro-expression recognition, facial action unit detection, other corresponding applications such as pain and depression detection, as well as human-robot interaction and collaboration. In alignment with FG2025's emphasis on ethics, as demonstrated by the inclusion of an Ethical Impact Statement requirement for this year's submissions, this workshop supports FG2025's efforts by encouraging research, discussion and dialogue on trustworthy FAA.", "date": "2025-06-05", "categories": []}
{"id": "2411.02430", "title": "Generative Emotion Cause Explanation in Multimodal Conversations", "abstract": "Multimodal conversation, a crucial form of human communication, carries rich emotional content, making the exploration of the causes of emotions within it a research endeavor of significant importance. However, existing research on the causes of emotions typically employs an utterance selection method within a single textual modality to locate causal utterances. This approach remains limited to coarse-grained assessments, lacks nuanced explanations of emotional causation, and demonstrates inadequate capability in identifying multimodal emotional triggers. Therefore, we introduce a task-\\textbf{Multimodal Emotion Cause Explanation in Conversation (MECEC)}. This task aims to generate a summary based on the multimodal context of conversations, clearly and intuitively describing the reasons that trigger a given emotion. To adapt to this task, we develop a new dataset (ECEM) based on the MELD dataset. ECEM combines video clips with detailed explanations of character emotions, helping to explore the causal factors behind emotional expression in multimodal conversations. A novel approach, FAME-Net, is further proposed, that harnesses the power of Large Language Models (LLMs) to analyze visual data and accurately interpret the emotions conveyed through facial expressions in videos. By exploiting the contagion effect of facial emotions, FAME-Net effectively captures the emotional causes of individuals engaged in conversations. Our experimental results on the newly constructed dataset show that FAME-Net outperforms several excellent baselines. Code and dataset are available at https://github.com/3222345200/FAME-Net.", "date": "2024-11-01", "categories": []}
{"id": "2412.17616", "title": "Facial Expression Analysis and Its Potentials in IoT Systems: A Contemporary Survey", "abstract": "Facial expressions convey human emotions and can be categorized into macro-expressions (MaEs) and micro-expressions (MiEs) based on duration and intensity. While MaEs are voluntary and easily recognized, MiEs are involuntary, rapid, and can reveal concealed emotions. The integration of facial expression analysis with Internet-of-Thing (IoT) systems has significant potential across diverse scenarios. IoT-enhanced MaE analysis enables real-time monitoring of patient emotions, facilitating improved mental health care in smart healthcare. Similarly, IoT-based MiE detection enhances surveillance accuracy and threat detection in smart security. Our work aims to provide a comprehensive overview of research progress in facial expression analysis and explores its potential integration with IoT systems. We discuss the distinctions between our work and existing surveys, elaborate on advancements in MaE and MiE analysis techniques across various learning paradigms, and examine their potential applications in IoT. We highlight challenges and future directions for the convergence of facial expression-based technologies and IoT systems, aiming to foster innovation in this domain. By presenting recent developments and practical applications, our work offers a systematic understanding of the ways of facial expression analysis to enhance IoT systems in healthcare, security, and beyond.", "date": "2024-12-23", "categories": []}
{"id": "2506.03993", "title": "Words of Warmth: Trust and Sociability Norms for over 26k English Words", "abstract": "Social psychologists have shown that Warmth (W) and Competence (C) are the primary dimensions along which we assess other people and groups. These dimensions impact various aspects of our lives from social competence and emotion regulation to success in the work place and how we view the world. More recent work has started to explore how these dimensions develop, why they have developed, and what they constitute. Of particular note, is the finding that warmth has two distinct components: Trust (T) and Sociability (S). In this work, we introduce Words of Warmth, the first large-scale repository of manually derived word--warmth (as well as word--trust and word--sociability) associations for over 26k English words. We show that the associations are highly reliable. We use the lexicons to study the rate at which children acquire WCTS words with age. Finally, we show that the lexicon enables a wide variety of bias and stereotype research through case studies on various target entities. Words of Warmth is freely available at: http://saifmohammad.com/warmth.html", "date": "2025-06-04", "categories": []}
{"id": "2409.00304", "title": "StimuVAR: Spatiotemporal Stimuli-aware Video Affective Reasoning with Multimodal Large Language Models", "abstract": "Predicting and reasoning how a video would make a human feel is crucial for developing socially intelligent systems. Although Multimodal Large Language Models (MLLMs) have shown impressive video understanding capabilities, they tend to focus more on the semantic content of videos, often overlooking emotional stimuli. Hence, most existing MLLMs fall short in estimating viewers' emotional reactions and providing plausible explanations. To address this issue, we propose StimuVAR, a spatiotemporal Stimuli-aware framework for Video Affective Reasoning (VAR) with MLLMs. StimuVAR incorporates a two-level stimuli-aware mechanism: frame-level awareness and token-level awareness. Frame-level awareness involves sampling video frames with events that are most likely to evoke viewers' emotions. Token-level awareness performs tube selection in the token space to make the MLLM concentrate on emotion-triggered spatiotemporal regions. Furthermore, we create VAR instruction data to perform affective training, steering MLLMs' reasoning strengths towards emotional focus and thereby enhancing their affective reasoning ability. To thoroughly assess the effectiveness of VAR, we provide a comprehensive evaluation protocol with extensive metrics. StimuVAR is the first MLLM-based method for viewer-centered VAR. Experiments demonstrate its superiority in understanding viewers' emotional responses to videos and providing coherent and insightful explanations. Our code is available at https://github.com/EthanG97/StimuVAR", "date": "2024-08-31", "categories": []}
{"id": "2412.13541", "title": "Spatio-Temporal Fuzzy-oriented Multi-Modal Meta-Learning for Fine-grained Emotion Recognition", "abstract": "Fine-grained emotion recognition (FER) plays a vital role in various fields, such as disease diagnosis, personalized recommendations, and multimedia mining. However, existing FER methods face three key challenges in real-world applications: (i) they rely on large amounts of continuously annotated data to ensure accuracy since emotions are complex and ambiguous in reality, which is costly and time-consuming; (ii) they cannot capture the temporal heterogeneity caused by changing emotion patterns, because they usually assume that the temporal correlation within sampling periods is the same; (iii) they do not consider the spatial heterogeneity of different FER scenarios, that is, the distribution of emotion information in different data may have bias or interference. To address these challenges, we propose a Spatio-Temporal Fuzzy-oriented Multi-modal Meta-learning framework (ST-F2M). Specifically, ST-F2M first divides the multi-modal videos into multiple views, and each view corresponds to one modality of one emotion. Multiple randomly selected views for the same emotion form a meta-training task. Next, ST-F2M uses an integrated module with spatial and temporal convolutions to encode the data of each task, reflecting the spatial and temporal heterogeneity. Then it adds fuzzy semantic information to each task based on generalized fuzzy rules, which helps handle the complexity and ambiguity of emotions. Finally, ST-F2M learns emotion-related general meta-knowledge through meta-recurrent neural networks to achieve fast and robust fine-grained emotion recognition. Extensive experiments show that ST-F2M outperforms various state-of-the-art methods in terms of accuracy and model efficiency. In addition, we construct ablation studies and further analysis to explore why ST-F2M performs well.", "date": "2024-12-18", "categories": []}
{"id": "2502.10973", "title": "Akan Cinematic Emotions (ACE): A Multimodal Multi-party Dataset for Emotion Recognition in Movie Dialogues", "abstract": "In this paper, we introduce the Akan Conversation Emotion (ACE) dataset, the first multimodal emotion dialogue dataset for an African language, addressing the significant lack of resources for low-resource languages in emotion recognition research. ACE, developed for the Akan language, contains 385 emotion-labeled dialogues and 6,162 utterances across audio, visual, and textual modalities, along with word-level prosodic prominence annotations. The presence of prosodic labels in this dataset also makes it the first prosodically annotated African language dataset. We demonstrate the quality and utility of ACE through experiments using state-of-the-art emotion recognition methods, establishing solid baselines for future research. We hope ACE inspires further work on inclusive, linguistically and culturally diverse NLP resources.", "date": "2025-02-16", "categories": []}
{"id": "2506.02088", "title": "Enhancing Speech Emotion Recognition with Graph-Based Multimodal Fusion and Prosodic Features for the Speech Emotion Recognition in Naturalistic Conditions Challenge at Interspeech 2025", "abstract": "Training SER models in natural, spontaneous speech is especially challenging due to the subtle expression of emotions and the unpredictable nature of real-world audio. In this paper, we present a robust system for the INTERSPEECH 2025 Speech Emotion Recognition in Naturalistic Conditions Challenge, focusing on categorical emotion recognition. Our method combines state-of-the-art audio models with text features enriched by prosodic and spectral cues. In particular, we investigate the effectiveness of Fundamental Frequency (F0) quantization and the use of a pretrained audio tagging model. We also employ an ensemble model to improve robustness. On the official test set, our system achieved a Macro F1-score of 39.79% (42.20% on validation). Our results underscore the potential of these methods, and analysis of fusion techniques confirmed the effectiveness of Graph Attention Networks. Our source code is publicly available.", "date": "2025-06-02", "categories": []}
{"id": "2506.02239", "title": "Investigating the Impact of Word Informativeness on Speech Emotion Recognition", "abstract": "In emotion recognition from speech, a key challenge lies in identifying speech signal segments that carry the most relevant acoustic variations for discerning specific emotions. Traditional approaches compute functionals for features such as energy and F0 over entire sentences or longer speech portions, potentially missing essential fine-grained variation in the long-form statistics. This research investigates the use of word informativeness, derived from a pre-trained language model, to identify semantically important segments. Acoustic features are then computed exclusively for these identified segments, enhancing emotion recognition accuracy. The methodology utilizes standard acoustic prosodic features, their functionals, and self-supervised representations. Results indicate a notable improvement in recognition performance when features are computed on segments selected based on word informativeness, underscoring the effectiveness of this approach.", "date": "2025-06-02", "categories": []}
{"id": "2409.10289", "title": "ReflectDiffu:Reflect between Emotion-intent Contagion and Mimicry for Empathetic Response Generation via a RL-Diffusion Framework", "abstract": "Empathetic response generation necessitates the integration of emotional and intentional dynamics to foster meaningful interactions. Existing research either neglects the intricate interplay between emotion and intent, leading to suboptimal controllability of empathy, or resorts to large language models (LLMs), which incur significant computational overhead. In this paper, we introduce ReflectDiffu, a lightweight and comprehensive framework for empathetic response generation. This framework incorporates emotion contagion to augment emotional expressiveness and employs an emotion-reasoning mask to pinpoint critical emotional elements. Additionally, it integrates intent mimicry within reinforcement learning for refinement during diffusion. By harnessing an intent twice reflect mechanism of Exploring-Sampling-Correcting, ReflectDiffu adeptly translates emotional decision-making into precise intent actions, thereby addressing empathetic response misalignments stemming from emotional misrecognition. Through reflection, the framework maps emotional states to intents, markedly enhancing both response empathy and flexibility. Comprehensive experiments reveal that ReflectDiffu outperforms existing models regarding relevance, controllability, and informativeness, achieving state-of-the-art results in both automatic and human evaluations.", "date": "2024-09-16", "categories": []}
{"id": "2506.00865", "title": "GIA-MIC: Multimodal Emotion Recognition with Gated Interactive Attention and Modality-Invariant Learning Constraints", "abstract": "Multimodal emotion recognition (MER) extracts emotions from multimodal data, including visual, speech, and text inputs, playing a key role in human-computer interaction. Attention-based fusion methods dominate MER research, achieving strong classification performance. However, two key challenges remain: effectively extracting modality-specific features and capturing cross-modal similarities despite distribution differences caused by modality heterogeneity. To address these, we propose a gated interactive attention mechanism to adaptively extract modality-specific features while enhancing emotional information through pairwise interactions. Additionally, we introduce a modality-invariant generator to learn modality-invariant representations and constrain domain shifts by aligning cross-modal similarities. Experiments on IEMOCAP demonstrate that our method outperforms state-of-the-art MER approaches, achieving WA 80.7% and UA 81.3%.", "date": "2025-06-01", "categories": []}
{"id": "2506.00903", "title": "Leveraging CLIP Encoder for Multimodal Emotion Recognition", "abstract": "Multimodal emotion recognition (MER) aims to identify human emotions by combining data from various modalities such as language, audio, and vision. Despite the recent advances of MER approaches, the limitations in obtaining extensive datasets impede the improvement of performance. To mitigate this issue, we leverage a Contrastive Language-Image Pre-training (CLIP)-based architecture and its semantic knowledge from massive datasets that aims to enhance the discriminative multimodal representation. We propose a label encoder-guided MER framework based on CLIP (MER-CLIP) to learn emotion-related representations across modalities. Our approach introduces a label encoder that treats labels as text embeddings to incorporate their semantic information, leading to the learning of more representative emotional features. To further exploit label semantics, we devise a cross-modal decoder that aligns each modality to a shared embedding space by sequentially fusing modality features based on emotion-related input from the label encoder. Finally, the label encoder-guided prediction enables generalization across diverse labels by embedding their semantic information as well as word labels. Experimental results show that our method outperforms the state-of-the-art MER methods on the benchmark datasets, CMU-MOSI and CMU-MOSEI.", "date": "2025-06-01", "categories": []}
{"id": "2506.01490", "title": "Confidence-Aware Self-Distillation for Multimodal Sentiment Analysis with Incomplete Modalities", "abstract": "Multimodal sentiment analysis (MSA) aims to understand human sentiment through multimodal data. In real-world scenarios, practical factors often lead to uncertain modality missingness. Existing methods for handling modality missingness are based on data reconstruction or common subspace projections. However, these methods neglect the confidence in multimodal combinations and impose constraints on intra-class representation, hindering the capture of modality-specific information and resulting in suboptimal performance. To address these challenges, we propose a Confidence-Aware Self-Distillation (CASD) strategy that effectively incorporates multimodal probabilistic embeddings via a mixture of Student's $t$-distributions, enhancing its robustness by incorporating confidence and accommodating heavy-tailed properties. This strategy estimates joint distributions with uncertainty scores and reduces uncertainty in the student network by consistency distillation. Furthermore, we introduce a reparameterization representation module that facilitates CASD in robust multimodal learning by sampling embeddings from the joint distribution for the prediction module to calculate the task loss. As a result, the directional constraint from the loss minimization is alleviated by the sampled representation. Experimental results on three benchmark datasets demonstrate that our method achieves state-of-the-art performance.", "date": "2025-06-02", "categories": []}
{"id": "2506.01698", "title": "When LLMs Team Up: The Emergence of Collaborative Affective Computing", "abstract": "Affective Computing (AC) is essential in bridging the gap between human emotional experiences and machine understanding. Traditionally, AC tasks in natural language processing (NLP) have been approached through pipeline architectures, which often suffer from structure rigidity that leads to inefficiencies and limited adaptability. The advent of Large Language Models (LLMs) has revolutionized this field by offering a unified approach to affective understanding and generation tasks, enhancing the potential for dynamic, real-time interactions. However, LLMs face cognitive limitations in affective reasoning, such as misinterpreting cultural nuances or contextual emotions, and hallucination problems in decision-making. To address these challenges, recent research advocates for LLM-based collaboration systems that emphasize interactions among specialized models and LLMs, mimicking human-like affective intelligence through the synergy of emotional and rational thinking that aligns with Dual Process Theory in psychology. This survey aims to provide a comprehensive overview of LLM-based collaboration systems in AC, exploring from structured collaborations to autonomous collaborations. Specifically, it includes: (1) A systematic review of existing methods, focusing on collaboration strategies, mechanisms, key functions, and applications; (2) Experimental comparisons of collaboration strategies across representative tasks in affective understanding and generation; (3) An analysis highlighting the potential of these systems to enhance robustness and adaptability in complex affective reasoning; (4) A discussion of key challenges and future research directions to further advance the field. This work is the first to systematically explore collaborative intelligence with LLMs in AC, paving the way for more powerful applications that approach human-like social intelligence.", "date": "2025-06-02", "categories": []}
{"id": "2505.13455", "title": "Spatiotemporal Emotional Synchrony in Dyadic Interactions: The Role of Speech Conditions in Facial and Vocal Affective Alignment", "abstract": "Understanding how humans express and synchronize emotions across multiple communication channels particularly facial expressions and speech has significant implications for emotion recognition systems and human computer interaction. Motivated by the notion that non-overlapping speech promotes clearer emotional coordination, while overlapping speech disrupts synchrony, this study examines how these conversational dynamics shape the spatial and temporal alignment of arousal and valence across facial and vocal modalities. Using dyadic interactions from the IEMOCAP dataset, we extracted continuous emotion estimates via EmoNet (facial video) and a Wav2Vec2-based model (speech audio). Segments were categorized based on speech overlap, and emotional alignment was assessed using Pearson correlation, lag adjusted analysis, and Dynamic Time Warping (DTW). Across analyses, non overlapping speech was associated with more stable and predictable emotional synchrony than overlapping speech. While zero-lag correlations were low and not statistically different, non overlapping speech showed reduced variability, especially for arousal. Lag adjusted correlations and best-lag distributions revealed clearer, more consistent temporal alignment in these segments. In contrast, overlapping speech exhibited higher variability and flatter lag profiles, though DTW indicated unexpectedly tighter alignment suggesting distinct coordination strategies. Notably, directionality patterns showed that facial expressions more often preceded speech during turn-taking, while speech led during simultaneous vocalizations. These findings underscore the importance of conversational structure in regulating emotional communication and provide new insight into the spatial and temporal dynamics of multimodal affective alignment in real world interaction.", "date": "2025-04-29", "categories": []}
{"id": "2505.24423", "title": "MMAFFBen: A Multilingual and Multimodal Affective Analysis Benchmark for Evaluating LLMs and VLMs", "abstract": "Large language models and vision-language models (which we jointly call LMs) have transformed NLP and CV, demonstrating remarkable potential across various fields. However, their capabilities in affective analysis (i.e. sentiment analysis and emotion detection) remain underexplored. This gap is largely due to the absence of comprehensive evaluation benchmarks, and the inherent complexity of affective analysis tasks. In this paper, we introduce MMAFFBen, the first extensive open-source benchmark for multilingual multimodal affective analysis. MMAFFBen encompasses text, image, and video modalities across 35 languages, covering four key affective analysis tasks: sentiment polarity, sentiment intensity, emotion classification, and emotion intensity. Moreover, we construct the MMAFFIn dataset for fine-tuning LMs on affective analysis tasks, and further develop MMAFFLM-3b and MMAFFLM-7b based on it. We evaluate various representative LMs, including GPT-4o-mini, providing a systematic comparison of their affective understanding capabilities. This project is available at https://github.com/lzw108/MMAFFBen.", "date": "2025-05-30", "categories": []}
{"id": "2505.24427", "title": "Donate or Create? Comparing Data Collection Strategies for Emotion-labeled Multimodal Social Media Posts", "abstract": "Accurate modeling of subjective phenomena such as emotion expression requires data annotated with authors' intentions. Commonly such data is collected by asking study participants to donate and label genuine content produced in the real world, or create content fitting particular labels during the study. Asking participants to create content is often simpler to implement and presents fewer risks to participant privacy than data donation. However, it is unclear if and how study-created content may differ from genuine content, and how differences may impact models. We collect study-created and genuine multimodal social media posts labeled for emotion and compare them on several dimensions, including model performance. We find that compared to genuine posts, study-created posts are longer, rely more on their text and less on their images for emotion expression, and focus more on emotion-prototypical events. The samples of participants willing to donate versus create posts are demographically different. Study-created data is valuable to train models that generalize well to genuine data, but realistic effectiveness estimates require genuine data.", "date": "2025-05-30", "categories": []}
{"id": "2505.24493", "title": "MELT: Towards Automated Multimodal Emotion Data Annotation by Leveraging LLM Embedded Knowledge", "abstract": "Although speech emotion recognition (SER) has advanced significantly with deep learning, annotation remains a major hurdle. Human annotation is not only costly but also subject to inconsistencies annotators often have different preferences and may lack the necessary contextual knowledge, which can lead to varied and inaccurate labels. Meanwhile, Large Language Models (LLMs) have emerged as a scalable alternative for annotating text data. However, the potential of LLMs to perform emotional speech data annotation without human supervision has yet to be thoroughly investigated. To address these problems, we apply GPT-4o to annotate a multimodal dataset collected from the sitcom Friends, using only textual cues as inputs. By crafting structured text prompts, our methodology capitalizes on the knowledge GPT-4o has accumulated during its training, showcasing that it can generate accurate and contextually relevant annotations without direct access to multimodal inputs. Therefore, we propose MELT, a multimodal emotion dataset fully annotated by GPT-4o. We demonstrate the effectiveness of MELT by fine-tuning four self-supervised learning (SSL) backbones and assessing speech emotion recognition performance across emotion datasets. Additionally, our subjective experiments\\' results demonstrate a consistence performance improvement on SER.", "date": "2025-05-30", "categories": []}
{"id": "2502.11926", "title": "BRIGHTER: BRIdging the Gap in Human-Annotated Textual Emotion Recognition Datasets for 28 Languages", "abstract": "People worldwide use language in subtle and complex ways to express emotions. Although emotion recognition--an umbrella term for several NLP tasks--impacts various applications within NLP and beyond, most work in this area has focused on high-resource languages. This has led to significant disparities in research efforts and proposed solutions, particularly for under-resourced languages, which often lack high-quality annotated datasets. In this paper, we present BRIGHTER--a collection of multi-labeled, emotion-annotated datasets in 28 different languages and across several domains. BRIGHTER primarily covers low-resource languages from Africa, Asia, Eastern Europe, and Latin America, with instances labeled by fluent speakers. We highlight the challenges related to the data collection and annotation processes, and then report experimental results for monolingual and crosslingual multi-label emotion identification, as well as emotion intensity recognition. We analyse the variability in performance across languages and text domains, both with and without the use of LLMs, and show that the BRIGHTER datasets represent a meaningful step towards addressing the gap in text-based emotion recognition.", "date": "2025-02-17", "categories": []}
{"id": "2505.23427", "title": "On the Validity of Head Motion Patterns as Generalisable Depression Biomarkers", "abstract": "Depression is a debilitating mood disorder negatively impacting millions worldwide. While researchers have explored multiple verbal and non-verbal behavioural cues for automated depression assessment, head motion has received little attention thus far. Further, the common practice of validating machine learning models via a single dataset can limit model generalisability. This work examines the effectiveness and generalisability of models utilising elementary head motion units, termed kinemes, for depression severity estimation. Specifically, we consider three depression datasets from different western cultures (German: AVEC2013, Australian: Blackdog and American: Pitt datasets) with varied contextual and recording settings to investigate the generalisability of the derived kineme patterns via two methods: (i) k-fold cross-validation over individual/multiple datasets, and (ii) model reuse on other datasets. Evaluating classification and regression performance with classical machine learning methods, our results show that: (1) head motion patterns are efficient biomarkers for estimating depression severity, achieving highly competitive performance for both classification and regression tasks on a variety of datasets, including achieving the second best Mean Absolute Error (MAE) on the AVEC2013 dataset, and (2) kineme-based features are more generalisable than (a) raw head motion descriptors for binary severity classification, and (b) other visual behavioural cues for severity estimation (regression).", "date": "2025-05-29", "categories": []}
{"id": "2505.23732", "title": "EmotionRankCLAP: Bridging Natural Language Speaking Styles and Ordinal Speech Emotion via Rank-N-Contrast", "abstract": "Current emotion-based contrastive language-audio pretraining (CLAP) methods typically learn by na\\\"ively aligning audio samples with corresponding text prompts. Consequently, this approach fails to capture the ordinal nature of emotions, hindering inter-emotion understanding and often resulting in a wide modality gap between the audio and text embeddings due to insufficient alignment. To handle these drawbacks, we introduce EmotionRankCLAP, a supervised contrastive learning approach that uses dimensional attributes of emotional speech and natural language prompts to jointly capture fine-grained emotion variations and improve cross-modal alignment. Our approach utilizes a Rank-N-Contrast objective to learn ordered relationships by contrasting samples based on their rankings in the valence-arousal space. EmotionRankCLAP outperforms existing emotion-CLAP methods in modeling emotion ordinality across modalities, measured via a cross-modal retrieval task.", "date": "2025-05-29", "categories": []}
{"id": "2503.10688", "title": "CULEMO: Cultural Lenses on Emotion -- Benchmarking LLMs for Cross-Cultural Emotion Understanding", "abstract": "NLP research has increasingly focused on subjective tasks such as emotion analysis. However, existing emotion benchmarks suffer from two major shortcomings: (1) they largely rely on keyword-based emotion recognition, overlooking crucial cultural dimensions required for deeper emotion understanding, and (2) many are created by translating English-annotated data into other languages, leading to potentially unreliable evaluation. To address these issues, we introduce Cultural Lenses on Emotion (CuLEmo), the first benchmark designed to evaluate culture-aware emotion prediction across six languages: Amharic, Arabic, English, German, Hindi, and Spanish. CuLEmo comprises 400 crafted questions per language, each requiring nuanced cultural reasoning and understanding. We use this benchmark to evaluate several state-of-the-art LLMs on culture-aware emotion prediction and sentiment analysis tasks. Our findings reveal that (1) emotion conceptualizations vary significantly across languages and cultures, (2) LLMs performance likewise varies by language and cultural context, and (3) prompting in English with explicit country context often outperforms in-language prompts for culture-aware emotion and sentiment understanding. The dataset and evaluation code are publicly available.", "date": "2025-03-12", "categories": []}
{"id": "2505.20033", "title": "EmoNet-Face: An Expert-Annotated Benchmark for Synthetic Emotion Recognition", "abstract": "Effective human-AI interaction relies on AI's ability to accurately perceive and interpret human emotions. Current benchmarks for vision and vision-language models are severely limited, offering a narrow emotional spectrum that overlooks nuanced states (e.g., bitterness, intoxication) and fails to distinguish subtle differences between related feelings (e.g., shame vs. embarrassment). Existing datasets also often use uncontrolled imagery with occluded faces and lack demographic diversity, risking significant bias. To address these critical gaps, we introduce EmoNet Face, a comprehensive benchmark suite. EmoNet Face features: (1) A novel 40-category emotion taxonomy, meticulously derived from foundational research to capture finer details of human emotional experiences. (2) Three large-scale, AI-generated datasets (EmoNet HQ, Binary, and Big) with explicit, full-face expressions and controlled demographic balance across ethnicity, age, and gender. (3) Rigorous, multi-expert annotations for training and high-fidelity evaluation. (4) We built EmpathicInsight-Face, a model achieving human-expert-level performance on our benchmark. The publicly released EmoNet Face suite - taxonomy, datasets, and model - provides a robust foundation for developing and evaluating AI systems with a deeper understanding of human emotions.", "date": "2025-05-26", "categories": []}
{"id": "2505.20571", "title": "Emotion Classification In-Context in Spanish", "abstract": "Classifying customer feedback into distinct emotion categories is essential for understanding sentiment and improving customer experience. In this paper, we classify customer feedback in Spanish into three emotion categories--positive, neutral, and negative--using advanced NLP and ML techniques. Traditional methods translate feedback from widely spoken languages to less common ones, resulting in a loss of semantic integrity and contextual nuances inherent to the original language. To address this limitation, we propose a hybrid approach that combines TF-IDF with BERT embeddings, effectively transforming Spanish text into rich numerical representations that preserve the semantic depth of the original language by using a Custom Stacking Ensemble (CSE) approach. To evaluate emotion classification, we utilize a range of models, including Logistic Regression, KNN, Bagging classifier with LGBM, and AdaBoost. The CSE model combines these classifiers as base models and uses a one-vs-all Logistic Regression as the meta-model. Our experimental results demonstrate that CSE significantly outperforms the individual and BERT model, achieving a test accuracy of 93.3% on the native Spanish dataset--higher than the accuracy obtained from the translated version. These findings underscore the challenges of emotion classification in Spanish and highlight the advantages of combining vectorization techniques like TF-IDF with BERT for improved accuracy. Our results provide valuable insights for businesses seeking to leverage emotion classification to enhance customer feedback analysis and service improvements.", "date": "2025-05-26", "categories": []}
{"id": "2505.21196", "title": "Learning Annotation Consensus for Continuous Emotion Recognition", "abstract": "In affective computing, datasets often contain multiple annotations from different annotators, which may lack full agreement. Typically, these annotations are merged into a single gold standard label, potentially losing valuable inter-rater variability. We propose a multi-annotator training approach for continuous emotion recognition (CER) that seeks a consensus across all annotators rather than relying on a single reference label. Our method employs a consensus network to aggregate annotations into a unified representation, guiding the main arousal-valence predictor to better reflect collective inputs. Tested on the RECOLA and COGNIMUSE datasets, our approach outperforms traditional methods that unify annotations into a single label. This underscores the benefits of fully leveraging multi-annotator data in emotion recognition and highlights its applicability across various fields where annotations are abundant yet inconsistent.", "date": "2025-05-27", "categories": []}
{"id": "2208.11450", "title": "VISTANet: VIsual Spoken Textual Additive Net for Interpretable Multimodal Emotion Recognition", "abstract": "This paper proposes a multimodal emotion recognition system, VIsual Spoken Textual Additive Net (VISTANet), to classify emotions reflected by input containing image, speech, and text into discrete classes. A new interpretability technique, K-Average Additive exPlanation (KAAP), has been developed that identifies important visual, spoken, and textual features leading to predicting a particular emotion class. The VISTANet fuses information from image, speech, and text modalities using a hybrid of intermediate and late fusion. It automatically adjusts the weights of their intermediate outputs while computing the weighted average. The KAAP technique computes the contribution of each modality and corresponding features toward predicting a particular emotion class. To mitigate the insufficiency of multimodal emotion datasets labelled with discrete emotion classes, we have constructed the IIT-R MMEmoRec dataset consisting of images, corresponding speech and text, and emotion labels ('angry,' 'happy,' 'hate,' and 'sad'). The VISTANet has resulted in an overall emotion recognition accuracy of 80.11% on the IIT-R MMEmoRec dataset using visual, spoken, and textual modalities, outperforming single or dual-modality configurations. The code and data can be accessed at https://github.com/MIntelligence-Group/MMEmoRec.", "date": "2022-08-24", "categories": []}
{"id": "2505.18169", "title": "Interpretable Multi-Task PINN for Emotion Recognition and EDA Prediction", "abstract": "Understanding and predicting human emotional and physiological states using wearable sensors has important applications in stress monitoring, mental health assessment, and affective computing. This study presents a novel Multi-Task Physics-Informed Neural Network (PINN) that performs Electrodermal Activity (EDA) prediction and emotion classification simultaneously, using the publicly available WESAD dataset. The model integrates psychological self-report features (PANAS and SAM) with a physics-inspired differential equation representing EDA dynamics, enforcing biophysically grounded constraints through a custom loss function. This loss combines EDA regression, emotion classification, and a physics residual term for improved interpretability.\n  The architecture supports dual outputs for both tasks and is trained under a unified multi-task framework. Evaluated using 5-fold cross-validation, the model achieves an average EDA RMSE of 0.0362, Pearson correlation of 0.9919, and F1-score of 94.08 percent. These results outperform classical models such as SVR and XGBoost, as well as ablated variants like emotion-only and EDA-only models.\n  In addition, the learned physical parameters including decay rate (alpha_0), emotional sensitivity (beta), and time scaling (gamma) are interpretable and stable across folds, aligning with known principles of human physiology. This work is the first to introduce a multi-task PINN framework for wearable emotion recognition, offering improved performance, generalizability, and model transparency. The proposed system provides a foundation for future interpretable and multimodal applications in healthcare and human-computer interaction.", "date": "2025-05-14", "categories": []}
{"id": "2505.18175", "title": "Evaluation in EEG Emotion Recognition: State-of-the-Art Review and Unified Framework", "abstract": "Electroencephalography-based Emotion Recognition (EEG-ER) has become a growing research area in recent years. Analyzing 216 papers published between 2018 and 2023, we uncover that the field lacks a unified evaluation protocol, which is essential to fairly define the state of the art, compare new approaches and to track the field's progress. We report the main inconsistencies between the used evaluation protocols, which are related to ground truth definition, evaluation metric selection, data splitting types (e.g., subject-dependent or subject-independent) and the use of different datasets. Capitalizing on this state-of-the-art research, we propose a unified evaluation protocol, EEGain (https://github.com/EmotionLab/EEGain), which enables an easy and efficient evaluation of new methods and datasets. EEGain is a novel open source software framework, offering the capability to compare - and thus define - state-of-the-art results. EEGain includes standardized methods for data pre-processing, data splitting, evaluation metrics, and the ability to load the six most relevant datasets (i.e., AMIGOS, DEAP, DREAMER, MAHNOB-HCI, SEED, SEED-IV) in EEG-ER with only a single line of code. In addition, we have assessed and validated EEGain using these six datasets on the four most common publicly available methods (EEGNet, DeepConvNet, ShallowConvNet, TSception). This is a significant step to make research on EEG-ER more reproducible and comparable, thereby accelerating the overall progress of the field.", "date": "2025-05-14", "categories": []}
{"id": "2505.18217", "title": "ABHINAYA -- A System for Speech Emotion Recognition In Naturalistic Conditions Challenge", "abstract": "Speech emotion recognition (SER) in naturalistic settings remains a challenge due to the intrinsic variability, diverse recording conditions, and class imbalance. As participants in the Interspeech Naturalistic SER Challenge which focused on these complexities, we present Abhinaya, a system integrating speech-based, text-based, and speech-text models. Our approach fine-tunes self-supervised and speech large language models (SLLM) for speech representations, leverages large language models (LLM) for textual context, and employs speech-text modeling with an SLLM to capture nuanced emotional cues. To combat class imbalance, we apply tailored loss functions and generate categorical decisions through majority voting. Despite one model not being fully trained, the Abhinaya system ranked 4th among 166 submissions. Upon completion of training, it achieved state-of-the-art performance among published results, demonstrating the effectiveness of our approach for SER in real-world conditions.", "date": "2025-05-23", "categories": []}
{"id": "2505.19178", "title": "Saliency-guided Emotion Modeling: Predicting Viewer Reactions from Video Stimuli", "abstract": "Understanding the emotional impact of videos is crucial for applications in content creation, advertising, and Human-Computer Interaction (HCI). Traditional affective computing methods rely on self-reported emotions, facial expression analysis, and biosensing data, yet they often overlook the role of visual saliency -- the naturally attention-grabbing regions within a video. In this study, we utilize deep learning to introduce a novel saliency-based approach to emotion prediction by extracting two key features: saliency area and number of salient regions. Using the HD2S saliency model and OpenFace facial action unit analysis, we examine the relationship between video saliency and viewer emotions. Our findings reveal three key insights: (1) Videos with multiple salient regions tend to elicit high-valence, low-arousal emotions, (2) Videos with a single dominant salient region are more likely to induce low-valence, high-arousal responses, and (3) Self-reported emotions often misalign with facial expression-based emotion detection, suggesting limitations in subjective reporting. By leveraging saliency-driven insights, this work provides a computationally efficient and interpretable alternative for emotion modeling, with implications for content creation, personalized media experiences, and affective computing research.", "date": "2025-05-25", "categories": []}
{"id": "2505.19694", "title": "Knowledge-Aligned Counterfactual-Enhancement Diffusion Perception for Unsupervised Cross-Domain Visual Emotion Recognition", "abstract": "Visual Emotion Recognition (VER) is a critical yet challenging task aimed at inferring emotional states of individuals based on visual cues. However, existing works focus on single domains, e.g., realistic images or stickers, limiting VER models' cross-domain generalizability. To fill this gap, we introduce an Unsupervised Cross-Domain Visual Emotion Recognition (UCDVER) task, which aims to generalize visual emotion recognition from the source domain (e.g., realistic images) to the low-resource target domain (e.g., stickers) in an unsupervised manner. Compared to the conventional unsupervised domain adaptation problems, UCDVER presents two key challenges: a significant emotional expression variability and an affective distribution shift. To mitigate these issues, we propose the Knowledge-aligned Counterfactual-enhancement Diffusion Perception (KCDP) framework. Specifically, KCDP leverages a VLM to align emotional representations in a shared knowledge space and guides diffusion models for improved visual affective perception. Furthermore, a Counterfactual-Enhanced Language-image Emotional Alignment (CLIEA) method generates high-quality pseudo-labels for the target domain. Extensive experiments demonstrate that our model surpasses SOTA models in both perceptibility and generalization, e.g., gaining 12% improvements over the SOTA VER model TGCA-PVT. The project page is at https://yinwen2019.github.io/ucdver.", "date": "2025-05-26", "categories": []}
{"id": "2505.12746", "title": "Correspondence of high-dimensional emotion structures elicited by video clips between humans and Multimodal LLMs", "abstract": "Recent studies have revealed that human emotions exhibit a high-dimensional, complex structure. A full capturing of this complexity requires new approaches, as conventional models that disregard high dimensionality risk overlooking key nuances of human emotions. Here, we examined the extent to which the latest generation of rapidly evolving Multimodal Large Language Models (MLLMs) capture these high-dimensional, intricate emotion structures, including capabilities and limitations. Specifically, we compared self-reported emotion ratings from participants watching videos with model-generated estimates (e.g., Gemini or GPT). We evaluated performance not only at the individual video level but also from emotion structures that account for inter-video relationships. At the level of simple correlation between emotion structures, our results demonstrated strong similarity between human and model-inferred emotion structures. To further explore whether the similarity between humans and models is at the signle item level or the coarse-categorical level, we applied Gromov Wasserstein Optimal Transport. We found that although performance was not necessarily high at the strict, single-item level, performance across video categories that elicit similar emotions was substantial, indicating that the model could infer human emotional experiences at the category level. Our results suggest that current state-of-the-art MLLMs broadly capture the complex high-dimensional emotion structures at the category level, as well as their apparent limitations in accurately capturing entire structures at the single-item level.", "date": "2025-05-19", "categories": []}
{"id": "2505.17090", "title": "EmoSign: A Multimodal Dataset for Understanding Emotions in American Sign Language", "abstract": "Unlike spoken languages where the use of prosodic features to convey emotion is well studied, indicators of emotion in sign language remain poorly understood, creating communication barriers in critical settings. Sign languages present unique challenges as facial expressions and hand movements simultaneously serve both grammatical and emotional functions. To address this gap, we introduce EmoSign, the first sign video dataset containing sentiment and emotion labels for 200 American Sign Language (ASL) videos. We also collect open-ended descriptions of emotion cues. Annotations were done by 3 Deaf ASL signers with professional interpretation experience. Alongside the annotations, we include baseline models for sentiment and emotion classification. This dataset not only addresses a critical gap in existing sign language research but also establishes a new benchmark for understanding model capabilities in multimodal emotion recognition for sign languages. The dataset is made available at https://huggingface.co/datasets/catfang/emosign.", "date": "2025-05-20", "categories": []}
{"id": "2505.17223", "title": "REACT 2025: the Third Multiple Appropriate Facial Reaction Generation Challenge", "abstract": "In dyadic interactions, a broad spectrum of human facial reactions might be appropriate for responding to each human speaker behaviour. Following the successful organisation of the REACT 2023 and REACT 2024 challenges, we are proposing the REACT 2025 challenge encouraging the development and benchmarking of Machine Learning (ML) models that can be used to generate multiple appropriate, diverse, realistic and synchronised human-style facial reactions expressed by human listeners in response to an input stimulus (i.e., audio-visual behaviours expressed by their corresponding speakers). As a key of the challenge, we provide challenge participants with the first natural and large-scale multi-modal MAFRG dataset (called MARS) recording 137 human-human dyadic interactions containing a total of 2856 interaction sessions covering five different topics. In addition, this paper also presents the challenge guidelines and the performance of our baselines on the two proposed sub-challenges: Offline MAFRG and Online MAFRG, respectively. The challenge baseline code is publicly available at https://github.com/reactmultimodalchallenge/baseline_react2025", "date": "2025-05-22", "categories": []}
{"id": "2505.17501", "title": "RoHyDR: Robust Hybrid Diffusion Recovery for Incomplete Multimodal Emotion Recognition", "abstract": "Multimodal emotion recognition analyzes emotions by combining data from multiple sources. However, real-world noise or sensor failures often cause missing or corrupted data, creating the Incomplete Multimodal Emotion Recognition (IMER) challenge. In this paper, we propose Robust Hybrid Diffusion Recovery (RoHyDR), a novel framework that performs missing-modality recovery at unimodal, multimodal, feature, and semantic levels. For unimodal representation recovery of missing modalities, RoHyDR exploits a diffusion-based generator to generate distribution-consistent and semantically aligned representations from Gaussian noise, using available modalities as conditioning. For multimodal fusion recovery, we introduce adversarial learning to produce a realistic fused multimodal representation and recover missing semantic content. We further propose a multi-stage optimization strategy that enhances training stability and efficiency. In contrast to previous work, the hybrid diffusion and adversarial learning-based recovery mechanism in RoHyDR allows recovery of missing information in both unimodal representation and multimodal fusion, at both feature and semantic levels, effectively mitigating performance degradation caused by suboptimal optimization. Comprehensive experiments conducted on two widely used multimodal emotion recognition benchmarks demonstrate that our proposed method outperforms state-of-the-art IMER methods, achieving robust recognition performance under various missing-modality scenarios. Our code will be made publicly available upon acceptance.", "date": "2025-05-23", "categories": []}
{"id": "2505.17833", "title": "Investigating Affect Mining Techniques for Annotation Sample Selection in the Creation of Finnish Affective Speech Corpus", "abstract": "Study of affect in speech requires suitable data, as emotional expression and perception vary across languages. Until now, no corpus has existed for natural expression of affect in spontaneous Finnish, existing data being acted or from a very specific communicative setting. This paper presents the first such corpus, created by annotating 12,000 utterances for emotional arousal and valence, sampled from three large-scale Finnish speech corpora. To ensure diverse affective expression, sample selection was conducted with an affect mining approach combining acoustic, cross-linguistic speech emotion, and text sentiment features. We compare this method to random sampling in terms of annotation diversity, and conduct post-hoc analyses to identify sampling choices that would have maximized the diversity. As an outcome, the work introduces a spontaneous Finnish affective speech corpus and informs sampling strategies for affective speech corpus creation in other languages or domains.", "date": "2025-05-23", "categories": []}
{"id": "2505.18040", "title": "Contrastive Distillation of Emotion Knowledge from LLMs for Zero-Shot Emotion Recognition", "abstract": "The ability to handle various emotion labels without dedicated training is crucial for building adaptable Emotion Recognition (ER) systems. Conventional ER models rely on training using fixed label sets and struggle to generalize beyond them. On the other hand, Large Language Models (LLMs) have shown strong zero-shot ER performance across diverse label spaces, but their scale limits their use on edge devices. In this work, we propose a contrastive distillation framework that transfers rich emotional knowledge from LLMs into a compact model without the use of human annotations. We use GPT-4 to generate descriptive emotion annotations, offering rich supervision beyond fixed label sets. By aligning text samples with emotion descriptors in a shared embedding space, our method enables zero-shot prediction on different emotion classes, granularity, and label schema. The distilled model is effective across multiple datasets and label spaces, outperforming strong baselines of similar size and approaching GPT-4's zero-shot performance, while being over 10,000 times smaller.", "date": "2025-05-23", "categories": []}
{"id": "2505.12007", "title": "Multi-modal Collaborative Optimization and Expansion Network for Event-assisted Single-eye Expression Recognition", "abstract": "In this paper, we proposed a Multi-modal Collaborative Optimization and Expansion Network (MCO-E Net), to use event modalities to resist challenges such as low light, high exposure, and high dynamic range in single-eye expression recognition tasks. The MCO-E Net introduces two innovative designs: Multi-modal Collaborative Optimization Mamba (MCO-Mamba) and Heterogeneous Collaborative and Expansion Mixture-of-Experts (HCE-MoE). MCO-Mamba, building upon Mamba, leverages dual-modal information to jointly optimize the model, facilitating collaborative interaction and fusion of modal semantics. This approach encourages the model to balance the learning of both modalities and harness their respective strengths. HCE-MoE, on the other hand, employs a dynamic routing mechanism to distribute structurally varied experts (deep, attention, and focal), fostering collaborative learning of complementary semantics. This heterogeneous architecture systematically integrates diverse feature extraction paradigms to comprehensively capture expression semantics. Extensive experiments demonstrate that our proposed network achieves competitive performance in the task of single-eye expression recognition, especially under poor lighting conditions.", "date": "2025-05-17", "categories": []}
{"id": "2505.16220", "title": "Meta-PerSER: Few-Shot Listener Personalized Speech Emotion Recognition via Meta-learning", "abstract": "This paper introduces Meta-PerSER, a novel meta-learning framework that personalizes Speech Emotion Recognition (SER) by adapting to each listener's unique way of interpreting emotion. Conventional SER systems rely on aggregated annotations, which often overlook individual subtleties and lead to inconsistent predictions. In contrast, Meta-PerSER leverages a Model-Agnostic Meta-Learning (MAML) approach enhanced with Combined-Set Meta-Training, Derivative Annealing, and per-layer per-step learning rates, enabling rapid adaptation with only a few labeled examples. By integrating robust representations from pre-trained self-supervised models, our framework first captures general emotional cues and then fine-tunes itself to personal annotation styles. Experiments on the IEMOCAP corpus demonstrate that Meta-PerSER significantly outperforms baseline methods in both seen and unseen data scenarios, highlighting its promise for personalized emotion recognition.", "date": "2025-05-22", "categories": []}
{"id": "2505.16455", "title": "Psychology-driven LLM Agents for Explainable Panic Prediction on Social Media during Sudden Disaster Events", "abstract": "During sudden disaster events, accurately predicting public panic sentiment on social media is crucial for proactive governance and crisis management. Current efforts on this problem face three main challenges: lack of finely annotated data hinders emotion prediction studies, unmodeled risk perception causes prediction inaccuracies, and insufficient interpretability of panic formation mechanisms. We address these issues by proposing a Psychology-driven generative Agent framework (PsychoAgent) for explainable panic prediction based on emotion arousal theory. Specifically, we first construct a fine-grained open panic emotion dataset (namely COPE) via human-large language models (LLMs) collaboration to mitigate semantic bias. Then, we develop a framework integrating cross-domain heterogeneous data grounded in psychological mechanisms to model risk perception and cognitive differences in emotion generation. To enhance interpretability, we design an LLM-based role-playing agent that simulates individual psychological chains through dedicatedly designed prompts. Experimental results on our annotated dataset show that PsychoAgent improves panic emotion prediction performance by 12.6% to 21.7% compared to baseline models. Furthermore, the explainability and generalization of our approach is validated. Crucially, this represents a paradigm shift from opaque \"data-driven fitting\" to transparent \"role-based simulation with mechanistic interpretation\" for panic emotion prediction during emergencies. Our implementation is publicly available at: https://anonymous.4open.science/r/PsychoAgent-19DD.", "date": "2025-05-22", "categories": []}
{"id": "2505.16460", "title": "University of Indonesia at SemEval-2025 Task 11: Evaluating State-of-the-Art Encoders for Multi-Label Emotion Detection", "abstract": "This paper presents our approach for SemEval 2025 Task 11 Track A, focusing on multilabel emotion classification across 28 languages. We explore two main strategies: fully fine-tuning transformer models and classifier-only training, evaluating different settings such as fine-tuning strategies, model architectures, loss functions, encoders, and classifiers. Our findings suggest that training a classifier on top of prompt-based encoders such as mE5 and BGE yields significantly better results than fully fine-tuning XLMR and mBERT. Our best-performing model on the final leaderboard is an ensemble combining multiple BGE models, where CatBoost serves as the classifier, with different configurations. This ensemble achieves an average F1-macro score of 56.58 across all languages.", "date": "2025-05-22", "categories": []}
{"id": "2505.15348", "title": "The Super Emotion Dataset", "abstract": "Despite the wide-scale usage and development of emotion classification datasets in NLP, the field lacks a standardized, large-scale resource that follows a psychologically grounded taxonomy. Existing datasets either use inconsistent emotion categories, suffer from limited sample size, or focus on specific domains. The Super Emotion Dataset addresses this gap by harmonizing diverse text sources into a unified framework based on Shaver's empirically validated emotion taxonomy, enabling more consistent cross-domain emotion recognition research.", "date": "2025-05-21", "categories": []}
{"id": "2505.13483", "title": "EmoMeta: A Multimodal Dataset for Fine-grained Emotion Classification in Chinese Metaphors", "abstract": "Metaphors play a pivotal role in expressing emotions, making them crucial for emotional intelligence. The advent of multimodal data and widespread communication has led to a proliferation of multimodal metaphors, amplifying the complexity of emotion classification compared to single-mode scenarios. However, the scarcity of research on constructing multimodal metaphorical fine-grained emotion datasets hampers progress in this domain. Moreover, existing studies predominantly focus on English, overlooking potential variations in emotional nuances across languages. To address these gaps, we introduce a multimodal dataset in Chinese comprising 5,000 text-image pairs of metaphorical advertisements. Each entry is meticulously annotated for metaphor occurrence, domain relations and fine-grained emotion classification encompassing joy, love, trust, fear, sadness, disgust, anger, surprise, anticipation, and neutral. Our dataset is publicly accessible (https://github.com/DUTIR-YSQ/EmoMeta), facilitating further advancements in this burgeoning field.", "date": "2025-05-12", "categories": []}
{"id": "2505.13805", "title": "ClapFM-EVC: High-Fidelity and Flexible Emotional Voice Conversion with Dual Control from Natural Language and Speech", "abstract": "Despite great advances, achieving high-fidelity emotional voice conversion (EVC) with flexible and interpretable control remains challenging. This paper introduces ClapFM-EVC, a novel EVC framework capable of generating high-quality converted speech driven by natural language prompts or reference speech with adjustable emotion intensity. We first propose EVC-CLAP, an emotional contrastive language-audio pre-training model, guided by natural language prompts and categorical labels, to extract and align fine-grained emotional elements across speech and text modalities. Then, a FuEncoder with an adaptive intensity gate is presented to seamless fuse emotional features with Phonetic PosteriorGrams from a pre-trained ASR model. To further improve emotion expressiveness and speech naturalness, we propose a flow matching model conditioned on these captured features to reconstruct Mel-spectrogram of source speech. Subjective and objective evaluations validate the effectiveness of ClapFM-EVC.", "date": "2025-05-20", "categories": []}
{"id": "2505.14143", "title": "Multimodal Mixture of Low-Rank Experts for Sentiment Analysis and Emotion Recognition", "abstract": "Multi-task learning (MTL) enables the efficient transfer of extra knowledge acquired from other tasks. The high correlation between multimodal sentiment analysis (MSA) and multimodal emotion recognition (MER) supports their joint training. However, existing methods primarily employ hard parameter sharing, ignoring parameter conflicts caused by complex task correlations. In this paper, we present a novel MTL method for MSA and MER, termed Multimodal Mixture of Low-Rank Experts (MMoLRE). MMoLRE utilizes shared and task-specific experts to distinctly model common and unique task characteristics, thereby avoiding parameter conflicts. Additionally, inspired by low-rank structures in the Mixture of Experts (MoE) framework, we design low-rank expert networks to reduce parameter and computational overhead as the number of experts increases. Extensive experiments on the CMU-MOSI and CMU-MOSEI benchmarks demonstrate that MMoLRE achieves state-of-the-art performance on the MSA task and competitive results on the MER task.", "date": "2025-05-20", "categories": []}
{"id": "2505.10575", "title": "Robust Emotion Recognition via Bi-Level Self-Supervised Continual Learning", "abstract": "Emotion recognition through physiological signals such as electroencephalogram (EEG) has become an essential aspect of affective computing and provides an objective way to capture human emotions. However, physiological data characterized by cross-subject variability and noisy labels hinder the performance of emotion recognition models. Existing domain adaptation and continual learning methods struggle to address these issues, especially under realistic conditions where data is continuously streamed and unlabeled. To overcome these limitations, we propose a novel bi-level self-supervised continual learning framework, SSOCL, based on a dynamic memory buffer. This bi-level architecture iteratively refines the dynamic buffer and pseudo-label assignments to effectively retain representative samples, enabling generalization from continuous, unlabeled physiological data streams for emotion recognition. The assigned pseudo-labels are subsequently leveraged for accurate emotion prediction. Key components of the framework, including a fast adaptation module and a cluster-mapping module, enable robust learning and effective handling of evolving data streams. Experimental validation on two mainstream EEG tasks demonstrates the framework's ability to adapt to continuous data streams while maintaining strong generalization across subjects, outperforming existing approaches.", "date": "2025-05-13", "categories": []}
{"id": "2505.12160", "title": "Emotion Recognition for Low-Resource Turkish: Fine-Tuning BERTurk on TREMO and Testing on Xenophobic Political Discourse", "abstract": "Social media platforms like X (formerly Twitter) play a crucial role in shaping public discourse and societal norms. This study examines the term Sessiz Istila (Silent Invasion) on Turkish social media, highlighting the rise of anti-refugee sentiment amidst the Syrian refugee influx. Using BERTurk and the TREMO dataset, we developed an advanced Emotion Recognition Model (ERM) tailored for Turkish, achieving 92.62% accuracy in categorizing emotions such as happiness, fear, anger, sadness, disgust, and surprise. By applying this model to large-scale X data, the study uncovers emotional nuances in Turkish discourse, contributing to computational social science by advancing sentiment analysis in underrepresented languages and enhancing our understanding of global digital discourse and the unique linguistic challenges of Turkish. The findings underscore the transformative potential of localized NLP tools, with our ERM model offering practical applications for real-time sentiment analysis in Turkish-language contexts. By addressing critical areas, including marketing, public relations, and crisis management, these models facilitate improved decision-making through timely and accurate sentiment tracking. This highlights the significance of advancing research that accounts for regional and linguistic nuances.", "date": "2025-05-17", "categories": []}
{"id": "2505.13244", "title": "JNLP at SemEval-2025 Task 11: Cross-Lingual Multi-Label Emotion Detection Using Generative Models", "abstract": "With the rapid advancement of global digitalization, users from different countries increasingly rely on social media for information exchange. In this context, multilingual multi-label emotion detection has emerged as a critical research area. This study addresses SemEval-2025 Task 11: Bridging the Gap in Text-Based Emotion Detection. Our paper focuses on two sub-tracks of this task: (1) Track A: Multi-label emotion detection, and (2) Track B: Emotion intensity. To tackle multilingual challenges, we leverage pre-trained multilingual models and focus on two architectures: (1) a fine-tuned BERT-based classification model and (2) an instruction-tuned generative LLM. Additionally, we propose two methods for handling multi-label classification: the base method, which maps an input directly to all its corresponding emotion labels, and the pairwise method, which models the relationship between the input text and each emotion category individually. Experimental results demonstrate the strong generalization ability of our approach in multilingual emotion recognition. In Track A, our method achieved Top 4 performance across 10 languages, ranking 1st in Hindi. In Track B, our approach also secured Top 5 performance in 7 languages, highlighting its simplicity and effectiveness\\footnote{Our code is available at https://github.com/yingjie7/mlingual_multilabel_emo_detection.", "date": "2025-05-19", "categories": []}
{"id": "2505.09967", "title": "TKFNet: Learning Texture Key Factor Driven Feature for Facial Expression Recognition", "abstract": "Facial expression recognition (FER) in the wild remains a challenging task due to the subtle and localized nature of expression-related features, as well as the complex variations in facial appearance. In this paper, we introduce a novel framework that explicitly focuses on Texture Key Driver Factors (TKDF), localized texture regions that exhibit strong discriminative power across emotional categories. By carefully observing facial image patterns, we identify that certain texture cues, such as micro-changes in skin around the brows, eyes, and mouth, serve as primary indicators of emotional dynamics. To effectively capture and leverage these cues, we propose a FER architecture comprising a Texture-Aware Feature Extractor (TAFE) and Dual Contextual Information Filtering (DCIF). TAFE employs a ResNet-based backbone enhanced with multi-branch attention to extract fine-grained texture representations, while DCIF refines these features by filtering context through adaptive pooling and attention mechanisms. Experimental results on RAF-DB and KDEF datasets demonstrate that our method achieves state-of-the-art performance, verifying the effectiveness and robustness of incorporating TKDFs into FER pipelines.", "date": "2025-05-15", "categories": []}
{"id": "2401.12133", "title": "VRMN-bD: A Multi-modal Natural Behavior Dataset of Immersive Human Fear Responses in VR Stand-up Interactive Games", "abstract": "Understanding and recognizing emotions are important and challenging issues in the metaverse era. Understanding, identifying, and predicting fear, which is one of the fundamental human emotions, in virtual reality (VR) environments plays an essential role in immersive game development, scene development, and next-generation virtual human-computer interaction applications. In this article, we used VR horror games as a medium to analyze fear emotions by collecting multi-modal data (posture, audio, and physiological signals) from 23 players. We used an LSTM-based model to predict fear with accuracies of 65.31% and 90.47% under 6-level classification (no fear and five different levels of fear) and 2-level classification (no fear and fear), respectively. We constructed a multi-modal natural behavior dataset of immersive human fear responses (VRMN-bD) and compared it with existing relevant advanced datasets. The results show that our dataset has fewer limitations in terms of collection method, data scale and audience scope. We are unique and advanced in targeting multi-modal datasets of fear and behavior in VR stand-up interactive environments. Moreover, we discussed the implications of this work for communities and applications. The dataset and pre-trained model are available at https://github.com/KindOPSTAR/VRMN-bD.", "date": "2024-01-22", "categories": []}
{"id": "2505.06536", "title": "TACFN: Transformer-based Adaptive Cross-modal Fusion Network for Multimodal Emotion Recognition", "abstract": "The fusion technique is the key to the multimodal emotion recognition task. Recently, cross-modal attention-based fusion methods have demonstrated high performance and strong robustness. However, cross-modal attention suffers from redundant features and does not capture complementary features well. We find that it is not necessary to use the entire information of one modality to reinforce the other during cross-modal interaction, and the features that can reinforce a modality may contain only a part of it. To this end, we design an innovative Transformer-based Adaptive Cross-modal Fusion Network (TACFN). Specifically, for the redundant features, we make one modality perform intra-modal feature selection through a self-attention mechanism, so that the selected features can adaptively and efficiently interact with another modality. To better capture the complementary information between the modalities, we obtain the fused weight vector by splicing and use the weight vector to achieve feature reinforcement of the modalities. We apply TCAFN to the RAVDESS and IEMOCAP datasets. For fair comparison, we use the same unimodal representations to validate the effectiveness of the proposed fusion method. The experimental results show that TACFN brings a significant performance improvement compared to other methods and reaches the state-of-the-art. All code and models could be accessed from https://github.com/shuzihuaiyu/TACFN.", "date": "2025-05-10", "categories": []}
{"id": "2505.07020", "title": "R-CAGE: A Structural Model for Emotion Output Design in Human-AI Interaction", "abstract": "This paper presents R-CAGE (Rhythmic Control Architecture for Guarding Ego), a theoretical framework for restructuring emotional output in long-term human-AI interaction. While prior affective computing approaches emphasized expressiveness, immersion, and responsiveness, they often neglected the cognitive and structural consequences of repeated emotional engagement. R-CAGE instead conceptualizes emotional output not as reactive expression but as ethical design structure requiring architectural intervention. The model is grounded in experiential observations of subtle affective symptoms such as localized head tension, interpretive fixation, and emotional lag arising from prolonged interaction with affective AI systems. These indicate a mismatch between system-driven emotion and user interpretation that cannot be fully explained by biometric data or observable behavior. R-CAGE adopts a user-centered stance prioritizing psychological recovery, interpretive autonomy, and identity continuity. The framework consists of four control blocks: (1) Control of Rhythmic Expression regulates output pacing to reduce fatigue; (2) Architecture of Sensory Structuring adjusts intensity and timing of affective stimuli; (3) Guarding of Cognitive Framing reduces semantic pressure to allow flexible interpretation; (4) Ego-Aligned Response Design supports self-reference recovery during interpretive lag. By structurally regulating emotional rhythm, sensory intensity, and interpretive affordances, R-CAGE frames emotion not as performative output but as sustainable design unit. The goal is to protect users from oversaturation and cognitive overload while sustaining long-term interpretive agency in AI-mediated environments.", "date": "2025-05-11", "categories": []}
{"id": "2305.03568", "title": "A vector quantized masked autoencoder for audiovisual speech emotion recognition", "abstract": "An important challenge in emotion recognition is to develop methods that can leverage unlabeled training data. In this paper, we propose the VQ-MAE-AV model, a self-supervised multimodal model that leverages masked autoencoders to learn representations of audiovisual speech without labels. The model includes vector quantized variational autoencoders that compress raw audio and visual speech data into discrete tokens. The audiovisual speech tokens are used to train a multimodal masked autoencoder that consists of an encoder-decoder architecture with attention mechanisms. The model is designed to extract both local (i.e., at the frame level) and global (i.e., at the sequence level) representations of audiovisual speech. During self-supervised pre-training, the VQ-MAE-AV model is trained on a large-scale unlabeled dataset of audiovisual speech, for the task of reconstructing randomly masked audiovisual speech tokens and with a contrastive learning strategy. During this pre-training, the encoder learns to extract a representation of audiovisual speech that can be subsequently leveraged for emotion recognition. During the supervised fine-tuning stage, a small classification model is trained on top of the VQ-MAE-AV encoder for an emotion recognition task. The proposed approach achieves state-of-the-art emotion recognition results across several datasets in both controlled and in-the-wild conditions.", "date": "2023-05-05", "categories": []}
{"id": "2404.04545", "title": "TCAN: Text-oriented Cross Attention Network for Multimodal Sentiment\n  Analysis", "abstract": "  Multimodal Sentiment Analysis (MSA) endeavors to understand human sentiment\nby leveraging language, visual, and acoustic modalities. Despite the remarkable\nperformance exhibited by previous MSA approaches, the presence of inherent\nmultimodal heterogeneities poses a challenge, with the contribution of\ndifferent modalities varying considerably. Past research predominantly focused\non improving representation learning techniques and feature fusion strategies.\nHowever, many of these efforts overlooked the variation in semantic richness\namong different modalities, treating each modality uniformly. This approach may\nlead to underestimating the significance of strong modalities while\noveremphasizing the importance of weak ones. Motivated by these insights, we\nintroduce a Text-oriented Cross-Attention Network (TCAN), emphasizing the\npredominant role of the text modality in MSA. Specifically, for each multimodal\nsample, by taking unaligned sequences of the three modalities as inputs, we\ninitially allocate the extracted unimodal features into a visual-text and an\nacoustic-text pair. Subsequently, we implement self-attention on the text\nmodality and apply text-queried cross-attention to the visual and acoustic\nmodalities. To mitigate the influence of noise signals and redundant features,\nwe incorporate a gated control mechanism into the framework. Additionally, we\nintroduce unimodal joint learning to gain a deeper understanding of homogeneous\nemotional tendencies across diverse modalities through backpropagation.\nExperimental results demonstrate that TCAN consistently outperforms\nstate-of-the-art MSA methods on two datasets (CMU-MOSI and CMU-MOSEI).\n", "date": "2024-04-06", "categories": []}
{"id": "2505.04642", "title": "Rethinking Multimodal Sentiment Analysis: A High-Accuracy, Simplified\n  Fusion Architecture", "abstract": "  Multimodal sentiment analysis, a pivotal task in affective computing, seeks\nto understand human emotions by integrating cues from language, audio, and\nvisual signals. While many recent approaches leverage complex attention\nmechanisms and hierarchical architectures, we propose a lightweight, yet\neffective fusion-based deep learning model tailored for utterance-level emotion\nclassification. Using the benchmark IEMOCAP dataset, which includes aligned\ntext, audio-derived numeric features, and visual descriptors, we design a\nmodality-specific encoder using fully connected layers followed by dropout\nregularization. The modality-specific representations are then fused using\nsimple concatenation and passed through a dense fusion layer to capture\ncross-modal interactions. This streamlined architecture avoids computational\noverhead while preserving performance, achieving a classification accuracy of\n92% across six emotion categories. Our approach demonstrates that with careful\nfeature engineering and modular design, simpler fusion strategies can\noutperform or match more complex models, particularly in resource-constrained\nenvironments.\n", "date": "2025-05-05", "categories": []}
{"id": "2503.22712", "title": "Coverage-Guaranteed Speech Emotion Recognition via Calibrated\n  Uncertainty-Adaptive Prediction Sets", "abstract": "  Road rage, often triggered by emotional suppression and sudden outbursts,\nsignificantly threatens road safety by causing collisions and aggressive\nbehavior. Speech emotion recognition technologies can mitigate this risk by\nidentifying negative emotions early and issuing timely alerts. However, current\nSER methods, such as those based on hidden markov models and Long short-term\nmemory networks, primarily handle one-dimensional signals, frequently\nexperience overfitting, and lack calibration, limiting their safety-critical\neffectiveness. We propose a novel risk-controlled prediction framework\nproviding statistically rigorous guarantees on prediction accuracy. This\napproach employs a calibration set to define a binary loss function indicating\nwhether the true label is included in the prediction set. Using a data-driven\nthreshold $\\beta$, we optimize a joint loss function to maintain an expected\ntest loss bounded by a user-specified risk level $\\alpha$. Evaluations across\nsix baseline models and two benchmark datasets demonstrate our framework\nconsistently achieves a minimum coverage of $1 - \\alpha$, effectively\ncontrolling marginal error rates despite varying calibration-test split ratios\n(e.g., 0.1). The robustness and generalizability of the framework are further\nvalidated through an extension to small-batch online calibration under a local\nexchangeability assumption. We construct a non-negative test martingale to\nmaintain prediction validity even in dynamic and non-exchangeable environments.\nCross-dataset tests confirm our method's ability to uphold reliable statistical\nguarantees in realistic, evolving data scenarios.\n", "date": "2025-03-24", "categories": []}
{"id": "2310.04306", "title": "Towards A Robust Group-level Emotion Recognition via Uncertainty-Aware\n  Learning", "abstract": "  Group-level emotion recognition (GER) is an inseparable part of human\nbehavior analysis, aiming to recognize an overall emotion in a multi-person\nscene. However, the existing methods are devoted to combing diverse emotion\ncues while ignoring the inherent uncertainties under unconstrained\nenvironments, such as congestion and occlusion occurring within a group.\nAdditionally, since only group-level labels are available, inconsistent emotion\npredictions among individuals in one group can confuse the network. In this\npaper, we propose an uncertainty-aware learning (UAL) method to extract more\nrobust representations for GER. By explicitly modeling the uncertainty of each\nindividual, we utilize stochastic embedding drawn from a Gaussian distribution\ninstead of deterministic point embedding. This representation captures the\nprobabilities of different emotions and generates diverse predictions through\nthis stochasticity during the inference stage. Furthermore,\nuncertainty-sensitive scores are adaptively assigned as the fusion weights of\nindividuals' face within each group. Moreover, we develop an image enhancement\nmodule to enhance the model's robustness against severe noise. The overall\nthree-branch model, encompassing face, object, and scene component, is guided\nby a proportional-weighted fusion strategy and integrates the proposed\nuncertainty-aware method to produce the final group-level output. Experimental\nresults demonstrate the effectiveness and generalization ability of our method\nacross three widely used databases.\n", "date": "2023-10-06", "categories": []}
{"id": "2503.10707", "title": "CALLM: Understanding Cancer Survivors' Emotions and Intervention\n  Opportunities via Mobile Diaries and Context-Aware Language Models", "abstract": "  Cancer survivors face unique emotional challenges that impact their quality\nof life. Mobile diary entries provide a promising method for tracking emotional\nstates, improving self-awareness, and promoting well-being outcome. This paper\naims to, through mobile diaries, understand cancer survivors' emotional states\nand key variables related to just-in-time intervention opportunities, including\nthe desire to regulate emotions and the availability to engage in\ninterventions. Although emotion analysis tools show potential for recognizing\nemotions from text, current methods lack the contextual understanding necessary\nto interpret brief mobile diary narratives. Our analysis of diary entries from\ncancer survivors (N=407) reveals systematic relationships between described\ncontexts and emotional states, with administrative and health-related contexts\nassociated with negative affect and regulation needs, while leisure activities\npromote positive emotions. We propose CALLM, a Context-Aware framework\nleveraging Large Language Models (LLMs) with Retrieval-Augmented Generation\n(RAG) to analyze these brief entries by integrating retrieved peer experiences\nand personal diary history. CALLM demonstrates strong performance with balanced\naccuracies reaching 72.96% for positive affect, 73.29% for negative affect,\n73.72% for emotion regulation desire, and 60.09% for intervention availability,\noutperforming language model baselines. Post-hoc analysis reveals that model\nconfidence strongly predicts accuracy, with longer diary entries generally\nenhancing performance, and brief personalization periods yielding meaningful\nimprovements. Our findings demonstrate how contextual information in mobile\ndiaries can be effectively leveraged to understand emotional experiences,\npredict key states, and identify optimal intervention moments for personalized\njust-in-time support.\n", "date": "2025-03-12", "categories": []}
{"id": "2504.17974", "title": "Optimism, Expectation, or Sarcasm? Multi-Class Hope Speech Detection in\n  Spanish and English", "abstract": "  Hope is a complex and underexplored emotional state that plays a significant\nrole in education, mental health, and social interaction. Unlike basic\nemotions, hope manifests in nuanced forms ranging from grounded optimism to\nexaggerated wishfulness or sarcasm, making it difficult for Natural Language\nProcessing systems to detect accurately. This study introduces PolyHope V2, a\nmultilingual, fine-grained hope speech dataset comprising over 30,000 annotated\ntweets in English and Spanish. This resource distinguishes between four hope\nsubtypes Generalized, Realistic, Unrealistic, and Sarcastic and enhances\nexisting datasets by explicitly labeling sarcastic instances. We benchmark\nmultiple pretrained transformer models and compare them with large language\nmodels (LLMs) such as GPT 4 and Llama 3 under zero-shot and few-shot regimes.\nOur findings show that fine-tuned transformers outperform prompt-based LLMs,\nespecially in distinguishing nuanced hope categories and sarcasm. Through\nqualitative analysis and confusion matrices, we highlight systematic challenges\nin separating closely related hope subtypes. The dataset and results provide a\nrobust foundation for future emotion recognition tasks that demand greater\nsemantic and contextual sensitivity across languages.\n", "date": "2025-04-24", "categories": []}
{"id": "2401.15695", "title": "HappyRouting: Learning Emotion-Aware Route Trajectories for Scalable\n  In-The-Wild Navigation", "abstract": "  Routes represent an integral part of triggering emotions in drivers.\nNavigation systems allow users to choose a navigation strategy, such as the\nfastest or shortest route. However, they do not consider the driver's emotional\nwell-being. We present HappyRouting, a novel navigation-based empathic car\ninterface guiding drivers through real-world traffic while evoking positive\nemotions. We propose design considerations, derive a technical architecture,\nand implement a routing optimization framework. Our contribution is a machine\nlearning-based generated emotion map layer, predicting emotions along routes\nbased on static and dynamic contextual data. We evaluated HappyRouting in a\nreal-world driving study (N=13), finding that happy routes increase\nsubjectively perceived valence by 11% (p=.007). Although happy routes take 1.25\ntimes longer on average, participants perceived the happy route as shorter,\npresenting an emotion-enhanced alternative to today's fastest routing\nmechanisms. We discuss how emotion-based routing can be integrated into\nnavigation apps, promoting emotional well-being for mobility use.\n", "date": "2024-01-28", "categories": []}
{"id": "2505.01542", "title": "Emotions in the Loop: A Survey of Affective Computing for Emotional\n  Support", "abstract": "  In a world where technology is increasingly embedded in our everyday\nexperiences, systems that sense and respond to human emotions are elevating\ndigital interaction. At the intersection of artificial intelligence and\nhuman-computer interaction, affective computing is emerging with innovative\nsolutions where machines are humanized by enabling them to process and respond\nto user emotions. This survey paper explores recent research contributions in\naffective computing applications in the area of emotion recognition, sentiment\nanalysis and personality assignment developed using approaches like large\nlanguage models (LLMs), multimodal techniques, and personalized AI systems. We\nanalyze the key contributions and innovative methodologies applied by the\nselected research papers by categorizing them into four domains: AI chatbot\napplications, multimodal input systems, mental health and therapy applications,\nand affective computing for safety applications. We then highlight the\ntechnological strengths as well as the research gaps and challenges related to\nthese studies. Furthermore, the paper examines the datasets used in each study,\nhighlighting how modality, scale, and diversity impact the development and\nperformance of affective models. Finally, the survey outlines ethical\nconsiderations and proposes future directions to develop applications that are\nmore safe, empathetic and practical.\n", "date": "2025-05-02", "categories": []}
{"id": "2503.12623", "title": "MAVEN: Multi-modal Attention for Valence-Arousal Emotion Network", "abstract": "  Dynamic emotion recognition in the wild remains challenging due to the\ntransient nature of emotional expressions and temporal misalignment of\nmulti-modal cues. Traditional approaches predict valence and arousal and often\noverlook the inherent correlation between these two dimensions. The proposed\nMulti-modal Attention for Valence-Arousal Emotion Network (MAVEN) integrates\nvisual, audio, and textual modalities through a bi-directional cross-modal\nattention mechanism. MAVEN uses modality-specific encoders to extract features\nfrom synchronized video frames, audio segments, and transcripts, predicting\nemotions in polar coordinates following Russell's circumplex model. The\nevaluation of the Aff-Wild2 dataset using MAVEN achieved a concordance\ncorrelation coefficient (CCC) of 0.3061, surpassing the ResNet-50 baseline\nmodel with a CCC of 0.22. The multistage architecture captures the subtle and\ntransient nature of emotional expressions in conversational videos and improves\nemotion recognition in real-world situations. The code is available at:\nhttps://github.com/Vrushank-Ahire/MAVEN_8th_ABAW\n", "date": "2025-03-16", "categories": []}
{"id": "2505.00013", "title": "Performance Evaluation of Emotion Classification in Japanese Using\n  RoBERTa and DeBERTa", "abstract": "  Background Practical applications such as social media monitoring and\ncustomer-feedback analysis require accurate emotion detection for Japanese\ntext, yet resource scarcity and class imbalance hinder model performance.\n  Objective This study aims to build a high-accuracy model for predicting the\npresence or absence of eight Plutchik emotions in Japanese sentences.\n  Methods Using the WRIME corpus, we transform reader-averaged intensity scores\ninto binary labels and fine-tune four pre-trained language models (BERT,\nRoBERTa, DeBERTa-v3-base, DeBERTa-v3-large). For context, we also assess two\nlarge language models (TinySwallow-1.5B-Instruct and ChatGPT-4o). Accuracy and\nF1-score serve as evaluation metrics.\n  Results DeBERTa-v3-large attains the best mean accuracy (0.860) and F1-score\n(0.662), outperforming all other models. It maintains robust F1 across both\nhigh-frequency emotions (e.g., Joy, Anticipation) and low-frequency emotions\n(e.g., Anger, Trust). The LLMs lag, with ChatGPT-4o and\nTinySwallow-1.5B-Instruct scoring 0.527 and 0.292 in mean F1, respectively.\n  Conclusion The fine-tuned DeBERTa-v3-large model currently offers the most\nreliable solution for binary emotion classification in Japanese. We release\nthis model as a pip-installable package (pip install\ndeberta-emotion-predictor). Future work should augment data for rare emotions,\nreduce model size, and explore prompt engineering to improve LLM performance.\n  This manuscript is under review for possible publication in New Generation\nComputing.\n", "date": "2025-04-22", "categories": []}
{"id": "2409.15545", "title": "Addressing Emotion Bias in Music Emotion Recognition and Generation with\n  Frechet Audio Distance", "abstract": "  The complex nature of musical emotion introduces inherent bias in both\nrecognition and generation, particularly when relying on a single audio\nencoder, emotion classifier, or evaluation metric. In this work, we conduct a\nstudy on Music Emotion Recognition (MER) and Emotional Music Generation (EMG),\nemploying diverse audio encoders alongside Frechet Audio Distance (FAD), a\nreference-free evaluation metric. Our study begins with a benchmark evaluation\nof MER, highlighting the limitations of using a single audio encoder and the\ndisparities observed across different measurements. We then propose assessing\nMER performance using FAD derived from multiple encoders to provide a more\nobjective measure of musical emotion. Furthermore, we introduce an enhanced EMG\napproach designed to improve both the variability and prominence of generated\nmusical emotion, thereby enhancing its realism. Additionally, we investigate\nthe differences in realism between the emotions conveyed in real and synthetic\nmusic, comparing our EMG model against two baseline models. Experimental\nresults underscore the issue of emotion bias in both MER and EMG and\ndemonstrate the potential of using FAD and diverse audio encoders to evaluate\nmusical emotion more objectively and effectively.\n", "date": "2024-09-23", "categories": []}
{"id": "2504.21154", "title": "Emotion Recognition in Contemporary Dance Performances Using Laban\n  Movement Analysis", "abstract": "  This paper presents a novel framework for emotion recognition in contemporary\ndance by improving existing Laban Movement Analysis (LMA) feature descriptors\nand introducing robust, novel descriptors that capture both quantitative and\nqualitative aspects of the movement. Our approach extracts expressive\ncharacteristics from 3D keypoints data of professional dancers performing\ncontemporary dance under various emotional states, and trains multiple\nclassifiers, including Random Forests and Support Vector Machines.\nAdditionally, we provide in-depth explanation of features and their impact on\nmodel predictions using explainable machine learning methods. Overall, our\nstudy improves emotion recognition in contemporary dance and offers promising\napplications in performance analysis, dance training, and human--computer\ninteraction, with a highest accuracy of 96.85\\%.\n", "date": "2025-04-29", "categories": []}
{"id": "2504.21184", "title": "AffectEval: A Modular and Customizable Framework for Affective Computing", "abstract": "  The field of affective computing focuses on recognizing, interpreting, and\nresponding to human emotions, and has broad applications across education,\nchild development, and human health and wellness. However, developing affective\ncomputing pipelines remains labor-intensive due to the lack of software\nframeworks that support multimodal, multi-domain emotion recognition\napplications. This often results in redundant effort when building pipelines\nfor different applications. While recent frameworks attempt to address these\nchallenges, they remain limited in reducing manual effort and ensuring\ncross-domain generalizability. We introduce AffectEval, a modular and\ncustomizable framework to facilitate the development of affective computing\npipelines while reducing the manual effort and duplicate work involved in\ndeveloping such pipelines. We validate AffectEval by replicating prior\naffective computing experiments, and we demonstrate that our framework reduces\nprogramming effort by up to 90%, as measured by the reduction in raw lines of\ncode.\n", "date": "2025-04-29", "categories": []}
{"id": "2504.20342", "title": "Narrative-Centered Emotional Reflection: Scaffolding Autonomous\n  Emotional Literacy with AI", "abstract": "  Reflexion is an AI-powered platform designed to enable structured emotional\nself-reflection at scale. By integrating real-time emotion detection, layered\nreflective prompting, and metaphorical storytelling generation, Reflexion\nempowers users to engage in autonomous emotional exploration beyond basic\nsentiment categorization. Grounded in theories of expressive writing, cognitive\nrestructuring, self-determination, and critical consciousness development, the\nsystem scaffolds a progressive journey from surface-level emotional recognition\ntoward value-aligned action planning. Initial pilot studies with diverse\nparticipants demonstrate positive outcomes in emotional articulation, cognitive\nreframing, and perceived psychological resilience. Reflexion represents a\npromising direction for scalable, theory-informed affective computing\ninterventions aimed at fostering emotional literacy and psychological growth\nacross educational, therapeutic, and public health contexts.\n", "date": "2025-04-29", "categories": []}
{"id": "2504.18673", "title": "Can Third-parties Read Our Emotions?", "abstract": "  Natural Language Processing tasks that aim to infer an author's private\nstates, e.g., emotions and opinions, from their written text, typically rely on\ndatasets annotated by third-party annotators. However, the assumption that\nthird-party annotators can accurately capture authors' private states remains\nlargely unexamined. In this study, we present human subjects experiments on\nemotion recognition tasks that directly compare third-party annotations with\nfirst-party (author-provided) emotion labels. Our findings reveal significant\nlimitations in third-party annotations-whether provided by human annotators or\nlarge language models (LLMs)-in faithfully representing authors' private\nstates. However, LLMs outperform human annotators nearly across the board. We\nfurther explore methods to improve third-party annotation quality. We find that\ndemographic similarity between first-party authors and third-party human\nannotators enhances annotation performance. While incorporating first-party\ndemographic information into prompts leads to a marginal but statistically\nsignificant improvement in LLMs' performance. We introduce a framework for\nevaluating the limitations of third-party annotations and call for refined\nannotation practices to accurately represent and model authors' private states.\n", "date": "2025-04-25", "categories": []}
{"id": "2503.07269", "title": "SemEval-2025 Task 11: Bridging the Gap in Text-Based Emotion Detection", "abstract": "  We present our shared task on text-based emotion detection, covering more\nthan 30 languages from seven distinct language families. These languages are\npredominantly low-resource and are spoken across various continents. The data\ninstances are multi-labeled with six emotional classes, with additional\ndatasets in 11 languages annotated for emotion intensity. Participants were\nasked to predict labels in three tracks: (a) multilabel emotion detection, (b)\nemotion intensity score detection, and (c) cross-lingual emotion detection.\n  The task attracted over 700 participants. We received final submissions from\nmore than 200 teams and 93 system description papers. We report baseline\nresults, along with findings on the best-performing systems, the most common\napproaches, and the most effective methods across different tracks and\nlanguages. The datasets for this task are publicly available. The dataset is\navailable at SemEval2025 Task 11 https://brighter-dataset.github.io\n", "date": "2025-03-10", "categories": []}
{"id": "2504.16378", "title": "Cyberoception: Finding a Painlessly-Measurable New Sense in the\n  Cyberworld Towards Emotion-Awareness in Computing", "abstract": "  In Affective computing, recognizing users' emotions accurately is the basis\nof affective human-computer interaction. Understanding users' interoception\ncontributes to a better understanding of individually different emotional\nabilities, which is essential for achieving inter-individually accurate emotion\nestimation. However, existing interoception measurement methods, such as the\nheart rate discrimination task, have several limitations, including their\ndependence on a well-controlled laboratory environment and precision apparatus,\nmaking monitoring users' interoception challenging. This study aims to\ndetermine other forms of data that can explain users' interoceptive or similar\nstates in their real-world lives and propose a novel hypothetical concept\n\"cyberoception,\" a new sense (1) which has properties similar to interoception\nin terms of the correlation with other emotion-related abilities, and (2) which\ncan be measured only by the sensors embedded inside commodity smartphone\ndevices in users' daily lives. Results from a 10-day-long in-lab/in-the-wild\nhybrid experiment reveal a specific cyberoception type \"Turn On\" (users'\nsubjective sensory perception about the frequency of turning-on behavior on\ntheir smartphones), significantly related to participants' emotional valence.\nWe anticipate that cyberoception to serve as a fundamental building block for\ndeveloping more \"emotion-aware\", user-friendly applications and services.\n", "date": "2025-04-23", "categories": []}
{"id": "2504.16573", "title": "PsyCounAssist: A Full-Cycle AI-Powered Psychological Counseling\n  Assistant System", "abstract": "  Psychological counseling is a highly personalized and dynamic process that\nrequires therapists to continuously monitor emotional changes, document session\ninsights, and maintain therapeutic continuity. In this paper, we introduce\nPsyCounAssist, a comprehensive AI-powered counseling assistant system\nspecifically designed to augment psychological counseling practices.\nPsyCounAssist integrates multimodal emotion recognition combining speech and\nphotoplethysmography (PPG) signals for accurate real-time affective analysis,\nautomated structured session reporting using large language models (LLMs), and\npersonalized AI-generated follow-up support. Deployed on Android-based tablet\ndevices, the system demonstrates practical applicability and flexibility in\nreal-world counseling scenarios. Experimental evaluation confirms the\nreliability of PPG-based emotional classification and highlights the system's\npotential for non-intrusive, privacy-aware emotional support. PsyCounAssist\nrepresents a novel approach to ethically and effectively integrating AI into\npsychological counseling workflows.\n", "date": "2025-04-23", "categories": []}
{"id": "2504.16856", "title": "Emo Pillars: Knowledge Distillation to Support Fine-Grained\n  Context-Aware and Context-Less Emotion Classification", "abstract": "  Most datasets for sentiment analysis lack context in which an opinion was\nexpressed, often crucial for emotion understanding, and are mainly limited by a\nfew emotion categories. Foundation large language models (LLMs) like GPT-4\nsuffer from over-predicting emotions and are too resource-intensive. We design\nan LLM-based data synthesis pipeline and leverage a large model, Mistral-7b,\nfor the generation of training examples for more accessible, lightweight\nBERT-type encoder models. We focus on enlarging the semantic diversity of\nexamples and propose grounding the generation into a corpus of narratives to\nproduce non-repetitive story-character-centered utterances with unique contexts\nover 28 emotion classes. By running 700K inferences in 450 GPU hours, we\ncontribute with the dataset of 100K contextual and also 300K context-less\nexamples to cover both scenarios. We use it for fine-tuning pre-trained\nencoders, which results in several Emo Pillars models. We show that Emo Pillars\nmodels are highly adaptive to new domains when tuned to specific tasks such as\nGoEmotions, ISEAR, IEMOCAP, and EmoContext, reaching the SOTA performance on\nthe first three. We also validate our dataset, conducting statistical analysis\nand human evaluation, and confirm the success of our measures in utterance\ndiversification (although less for the neutral class) and context\npersonalization, while pointing out the need for improved handling of\nout-of-taxonomy labels within the pipeline.\n", "date": "2025-04-23", "categories": []}
{"id": "2410.21897", "title": "Semi-Supervised Self-Learning Enhanced Music Emotion Recognition", "abstract": "  Music emotion recognition (MER) aims to identify the emotions conveyed in a\ngiven musical piece. However, currently, in the field of MER, the available\npublic datasets have limited sample sizes. Recently, segment-based methods for\nemotion-related tasks have been proposed, which train backbone networks on\nshorter segments instead of entire audio clips, thereby naturally augmenting\ntraining samples without requiring additional resources. Then, the predicted\nsegment-level results are aggregated to obtain the entire song prediction. The\nmost commonly used method is that the segment inherits the label of the clip\ncontaining it, but music emotion is not constant during the whole clip. Doing\nso will introduce label noise and make the training easy to overfit. To handle\nthe noisy label issue, we propose a semi-supervised self-learning (SSSL)\nmethod, which can differentiate between samples with correct and incorrect\nlabels in a self-learning manner, thus effectively utilizing the augmented\nsegment-level data. Experiments on three public emotional datasets demonstrate\nthat the proposed method can achieve better or comparable performance.\n", "date": "2024-10-29", "categories": []}
{"id": "2504.14706", "title": "AI with Emotions: Exploring Emotional Expressions in Large Language\n  Models", "abstract": "  The human-level performance of Large Language Models (LLMs) across various\ntasks has raised expectations for the potential of Artificial Intelligence (AI)\nto possess emotions someday. To explore the capability of current LLMs to\nexpress emotions in their outputs, we conducted an experiment using several\nLLMs (OpenAI GPT, Google Gemini, Meta Llama3, and Cohere Command R+) to\nrole-play as agents answering questions with specified emotional states. We\ndefined the emotional states using Russell's Circumplex model, a\nwell-established framework that characterizes emotions along the\nsleepy-activated (arousal) and pleasure-displeasure (valence) axes. We chose\nthis model for its simplicity, utilizing two continuous parameters, which\nallows for better controllability in applications involving continuous changes\nin emotional states. The responses generated were evaluated using a sentiment\nanalysis model, independent of the LLMs, trained on the GoEmotions dataset. The\nevaluation showed that the emotional states of the generated answers were\nconsistent with the specifications, demonstrating the LLMs' capability for\nemotional expression. This indicates the potential for LLM-based AI agents to\nsimulate emotions, opening up a wide range of applications for emotion-based\ninteractions, such as advisors or consultants who can provide advice or\nopinions with a personal touch.\n", "date": "2025-04-20", "categories": []}
{"id": "2504.11460", "title": "Semantic Matters: Multimodal Features for Affective Analysis", "abstract": "  In this study, we present our methodology for two tasks: the Emotional\nMimicry Intensity (EMI) Estimation Challenge and the Behavioural\nAmbivalence/Hesitancy (BAH) Recognition Challenge, both conducted as part of\nthe 8th Workshop and Competition on Affective & Behavior Analysis in-the-wild.\nWe utilize a Wav2Vec 2.0 model pre-trained on a large podcast dataset to\nextract various audio features, capturing both linguistic and paralinguistic\ninformation. Our approach incorporates a valence-arousal-dominance (VAD) module\nderived from Wav2Vec 2.0, a BERT text encoder, and a vision transformer (ViT)\nwith predictions subsequently processed through a long short-term memory (LSTM)\narchitecture or a convolution-like method for temporal modeling. We integrate\nthe textual and visual modality into our analysis, recognizing that semantic\ncontent provides valuable contextual cues and underscoring that the meaning of\nspeech often conveys more critical insights than its acoustic counterpart\nalone. Fusing in the vision modality helps in some cases to interpret the\ntextual modality more precisely. This combined approach results in significant\nperformance improvements, achieving in EMI $\\rho_{\\text{TEST}} = 0.706$ and in\nBAH $F1_{\\text{TEST}} = 0.702$, securing first place in the EMI challenge and\nsecond place in the BAH challenge.\n", "date": "2025-03-16", "categories": []}
{"id": "2504.07521", "title": "Why We Feel: Breaking Boundaries in Emotional Reasoning with Multimodal\n  Large Language Models", "abstract": "  Most existing emotion analysis emphasizes which emotion arises (e.g., happy,\nsad, angry) but neglects the deeper why. We propose Emotion Interpretation\n(EI), focusing on causal factors-whether explicit (e.g., observable objects,\ninterpersonal interactions) or implicit (e.g., cultural context, off-screen\nevents)-that drive emotional responses. Unlike traditional emotion recognition,\nEI tasks require reasoning about triggers instead of mere labeling. To\nfacilitate EI research, we present EIBench, a large-scale benchmark\nencompassing 1,615 basic EI samples and 50 complex EI samples featuring\nmultifaceted emotions. Each instance demands rationale-based explanations\nrather than straightforward categorization. We further propose a Coarse-to-Fine\nSelf-Ask (CFSA) annotation pipeline, which guides Vision-Language Models\n(VLLMs) through iterative question-answer rounds to yield high-quality labels\nat scale. Extensive evaluations on open-source and proprietary large language\nmodels under four experimental settings reveal consistent performance\ngaps-especially for more intricate scenarios-underscoring EI's potential to\nenrich empathetic, context-aware AI applications. Our benchmark and methods are\npublicly available at: https://github.com/Lum1104/EIBench, offering a\nfoundation for advanced multimodal causal analysis and next-generation\naffective computing.\n", "date": "2025-04-10", "categories": []}
{"id": "2504.12331", "title": "Span-level Emotion-Cause-Category Triplet Extraction with Instruction\n  Tuning LLMs and Data Augmentation", "abstract": "  Span-level emotion-cause-category triplet extraction represents a novel and\ncomplex challenge within emotion cause analysis. This task involves identifying\nemotion spans, cause spans, and their associated emotion categories within the\ntext to form structured triplets. While prior research has predominantly\nconcentrated on clause-level emotion-cause pair extraction and span-level\nemotion-cause detection, these methods often confront challenges originating\nfrom redundant information retrieval and difficulty in accurately determining\nemotion categories, particularly when emotions are expressed implicitly or\nambiguously. To overcome these challenges, this study explores a fine-grained\napproach to span-level emotion-cause-category triplet extraction and introduces\nan innovative framework that leverages instruction tuning and data augmentation\ntechniques based on large language models. The proposed method employs\ntask-specific triplet extraction instructions and utilizes low-rank adaptation\nto fine-tune large language models, eliminating the necessity for intricate\ntask-specific architectures. Furthermore, a prompt-based data augmentation\nstrategy is developed to address data scarcity by guiding large language models\nin generating high-quality synthetic training data. Extensive experimental\nevaluations demonstrate that the proposed approach significantly outperforms\nexisting baseline methods, achieving at least a 12.8% improvement in span-level\nemotion-cause-category triplet extraction metrics. The results demonstrate the\nmethod's effectiveness and robustness, offering a promising avenue for\nadvancing research in emotion cause analysis. The source code is available at\nhttps://github.com/zxgnlp/InstruDa-LLM.\n", "date": "2025-04-13", "categories": []}
{"id": "2302.13170", "title": "Partial Label Learning for Emotion Recognition from EEG", "abstract": "  Fully supervised learning has recently achieved promising performance in\nvarious electroencephalography (EEG) learning tasks by training on large\ndatasets with ground truth labels. However, labeling EEG data for affective\nexperiments is challenging, as it can be difficult for participants to\naccurately distinguish between similar emotions, resulting in ambiguous\nlabeling (reporting multiple emotions for one EEG instance). This notion could\ncause model performance degradation, as the ground truth is hidden within\nmultiple candidate labels. To address this issue, Partial Label Learning (PLL)\nhas been proposed to identify the ground truth from candidate labels during the\ntraining phase, and has shown good performance in the computer vision domain.\nHowever, PLL methods have not yet been adopted for EEG representation learning\nor implemented for emotion recognition tasks. In this paper, we adapt and\nre-implement six state-of-the-art PLL approaches for emotion recognition from\nEEG on two large emotion datasets (SEED-IV and SEED-V). These datasets contain\nfour and five categories of emotions, respectively. We evaluate the performance\nof all methods in classical, circumplex-based and real-world experiments. The\nresults show that PLL methods can achieve strong results in affective computing\nfrom EEG and achieve comparable performance to fully supervised learning. We\nalso investigate the effect of label disambiguation, a key step in many PLL\nmethods. The results show that in most cases, label disambiguation would\nbenefit the model when the candidate labels are generated based on their\nsimilarities to the ground truth rather than obeying a uniform distribution.\nThis finding suggests the potential of using label disambiguation-based PLL\nmethods for circumplex-based and real-world affective tasks.\n", "date": "2023-02-25", "categories": []}
{"id": "2502.19935", "title": "Lotus at SemEval-2025 Task 11: RoBERTa with Llama-3 Generated\n  Explanations for Multi-Label Emotion Classification", "abstract": "  This paper presents a novel approach for multi-label emotion detection, where\nLlama-3 is used to generate explanatory content that clarifies ambiguous\nemotional expressions, thereby enhancing RoBERTa's emotion classification\nperformance. By incorporating explanatory context, our method improves\nF1-scores, particularly for emotions like fear, joy, and sadness, and\noutperforms text-only models. The addition of explanatory content helps resolve\nambiguity, addresses challenges like overlapping emotional cues, and enhances\nmulti-label classification, marking a significant advancement in emotion\ndetection tasks.\n", "date": "2025-02-27", "categories": []}
{"id": "2308.09013", "title": "Deep-seeded Clustering for Emotion Recognition from Wearable\n  Physiological Sensors", "abstract": "  According to the circumplex model of affect, an emotional response could\ncharacterized by a level of pleasure (valence) and intensity (arousal). As it\nreflects on the autonomic nervous system (ANS) activity, modern wearable\nwristbands can record non-invasively and during our everyday lives peripheral\nend-points of this response. While emotion recognition from physiological\nsignals is usually achieved using supervised machine learning algorithms that\nrequire ground truth labels for training, collecting it is cumbersome and\nparticularly unfeasible in naturalistic settings, and extracting meaningful\ninsights from these signals requires domain knowledge and might be prone to\nbias. Here, we propose and test a deep-seeded clustering algorithm that\nautomatically extracts and classifies features from those physiological signals\nwith minimal supervision - combining an autoencoder (AE) for unsupervised\nfeature representation and c-means clustering for fine-grained classification.\nWe also show that the model obtains good performance results across three\ndifferent datasets frequently used in affective computing studies (accuracies\nof 80.7% on WESAD, 64.2% on Stress-Predict and 61.0% on CEAP360-VR).\n", "date": "2023-08-17", "categories": []}
{"id": "2504.11082", "title": "DeepMLF: Multimodal language model with learnable tokens for deep fusion\n  in sentiment analysis", "abstract": "  While multimodal fusion has been extensively studied in Multimodal Sentiment\nAnalysis (MSA), the role of fusion depth and multimodal capacity allocation\nremains underexplored. In this work, we position fusion depth, scalability, and\ndedicated multimodal capacity as primary factors for effective fusion. We\nintroduce DeepMLF, a novel multimodal language model (LM) with learnable tokens\ntailored toward deep fusion. DeepMLF leverages an audiovisual encoder and a\npretrained decoder LM augmented with multimodal information across its layers.\nWe append learnable tokens to the LM that: 1) capture modality interactions in\na controlled fashion and 2) preserve independent information flow for each\nmodality. These fusion tokens gather linguistic information via causal\nself-attention in LM Blocks and integrate with audiovisual information through\ncross-attention MM Blocks. Serving as dedicated multimodal capacity, this\ndesign enables progressive fusion across multiple layers, providing depth in\nthe fusion process. Our training recipe combines modality-specific losses and\nlanguage modelling loss, with the decoder LM tasked to predict ground truth\npolarity. Across three MSA benchmarks with varying dataset characteristics,\nDeepMLF achieves state-of-the-art performance. Our results confirm that deeper\nfusion leads to better performance, with optimal fusion depths (5-7) exceeding\nthose of existing approaches. Additionally, our analysis on the number of\nfusion tokens reveals that small token sets ($\\sim$20) achieve optimal\nperformance. We examine the importance of representation learning order (fusion\ncurriculum) through audiovisual encoder initialization experiments. Our\nablation studies demonstrate the superiority of the proposed fusion design and\ngating while providing a holistic examination of DeepMLF's scalability to LLMs,\nand the impact of each training objective and embedding regularization.\n", "date": "2025-04-15", "categories": []}
{"id": "2504.09326", "title": "Infused Suppression Of Magnification Artefacts For Micro-AU Detection", "abstract": "  Facial micro-expressions are spontaneous, brief and subtle facial motions\nthat unveil the underlying, suppressed emotions. Detecting Action Units (AUs)\nin micro-expressions is crucial because it yields a finer representation of\nfacial motions than categorical emotions, effectively resolving the ambiguity\namong different expressions. One of the difficulties in micro-expression\nanalysis is that facial motions are subtle and brief, thereby increasing the\ndifficulty in correlating facial motion features to AU occurrence. To bridge\nthe subtlety issue, flow-related features and motion magnification are a few\ncommon approaches as they can yield descriptive motion changes and increased\nmotion amplitude respectively. While motion magnification can amplify the\nmotion changes, it also accounts for illumination changes and projection errors\nduring the amplification process, thereby creating motion artefacts that\nconfuse the model to learn inauthentic magnified motion features. The problem\nis further aggravated in the context of a more complicated task where more AU\nclasses are analyzed in cross-database settings. To address this issue, we\npropose InfuseNet, a layer-wise unitary feature infusion framework that\nleverages motion context to constrain the Action Unit (AU) learning within an\ninformative facial movement region, thereby alleviating the influence of\nmagnification artefacts. On top of that, we propose leveraging magnified latent\nfeatures instead of reconstructing magnified samples to limit the distortion\nand artefacts caused by the projection inaccuracy in the motion reconstruction\nprocess. Via alleviating the magnification artefacts, InfuseNet has surpassed\nthe state-of-the-art results in the CD6ME protocol. Further quantitative\nstudies have also demonstrated the efficacy of motion artefacts alleviation.\n", "date": "2025-04-12", "categories": []}
{"id": "2504.10351", "title": "Multimodal Representation Learning Techniques for Comprehensive Facial\n  State Analysis", "abstract": "  Multimodal foundation models have significantly improved feature\nrepresentation by integrating information from multiple modalities, making them\nhighly suitable for a broader set of applications. However, the exploration of\nmultimodal facial representation for understanding perception has been limited.\nUnderstanding and analyzing facial states, such as Action Units (AUs) and\nemotions, require a comprehensive and robust framework that bridges visual and\nlinguistic modalities. In this paper, we present a comprehensive pipeline for\nmultimodal facial state analysis. First, we compile a new Multimodal Face\nDataset (MFA) by generating detailed multilevel language descriptions of face,\nincorporating Action Unit (AU) and emotion descriptions, by leveraging GPT-4o.\nSecond, we introduce a novel Multilevel Multimodal Face Foundation model (MF^2)\ntailored for Action Unit (AU) and emotion recognition. Our model incorporates\ncomprehensive visual feature modeling at both local and global levels of face\nimage, enhancing its ability to represent detailed facial appearances. This\ndesign aligns visual representations with structured AU and emotion\ndescriptions, ensuring effective cross-modal integration. Third, we develop a\nDecoupled Fine-Tuning Network (DFN) that efficiently adapts MF^2 across various\ntasks and datasets. This approach not only reduces computational overhead but\nalso broadens the applicability of the foundation model to diverse scenarios.\nExperimentation show superior performance for AU and emotion detection tasks.\n", "date": "2025-04-14", "categories": []}
{"id": "2502.03979", "title": "Towards Unified Music Emotion Recognition across Dimensional and\n  Categorical Models", "abstract": "  One of the most significant challenges in Music Emotion Recognition (MER)\ncomes from the fact that emotion labels can be heterogeneous across datasets\nwith regard to the emotion representation, including categorical (e.g., happy,\nsad) versus dimensional labels (e.g., valence-arousal). In this paper, we\npresent a unified multitask learning framework that combines these two types of\nlabels and is thus able to be trained on multiple datasets. This framework uses\nan effective input representation that combines musical features (i.e., key and\nchords) and MERT embeddings. Moreover, knowledge distillation is employed to\ntransfer the knowledge of teacher models trained on individual datasets to a\nstudent model, enhancing its ability to generalize across multiple tasks. To\nvalidate our proposed framework, we conducted extensive experiments on a\nvariety of datasets, including MTG-Jamendo, DEAM, PMEmo, and EmoMusic.\nAccording to our experimental results, the inclusion of musical features,\nmultitask learning, and knowledge distillation significantly enhances\nperformance. In particular, our model outperforms the state-of-the-art models,\nincluding the best-performing model from the MediaEval 2021 competition on the\nMTG-Jamendo dataset. Our work makes a significant contribution to MER by\nallowing the combination of categorical and dimensional emotion labels in one\nunified framework, thus enabling training across datasets.\n", "date": "2025-02-06", "categories": []}
{"id": "2504.08543", "title": "UoB-NLP at SemEval-2025 Task 11: Leveraging Adapters for Multilingual\n  and Cross-Lingual Emotion Detection", "abstract": "  Emotion detection in natural language processing is a challenging task due to\nthe complexity of human emotions and linguistic diversity. While significant\nprogress has been made in high-resource languages, emotion detection in\nlow-resource languages remains underexplored. In this work, we address\nmultilingual and cross-lingual emotion detection by leveraging adapter-based\nfine-tuning with multilingual pre-trained language models. Adapters introduce a\nsmall number of trainable parameters while keeping the pre-trained model\nweights fixed, offering a parameter-efficient approach to adaptation. We\nexperiment with different adapter tuning strategies, including task-only\nadapters, target-language-ready task adapters, and language-family-based\nadapters. Our results show that target-language-ready task adapters achieve the\nbest overall performance, particularly for low-resource African languages with\nour team ranking 7th for Tigrinya, and 8th for Kinyarwanda in Track A. In Track\nC, our system ranked 3rd for Amharic, and 4th for Oromo, Tigrinya, Kinyarwanda,\nHausa, and Igbo. Our approach outperforms large language models in 11 languages\nand matches their performance in four others, despite our models having\nsignificantly fewer parameters. Furthermore, we find that adapter-based models\nretain cross-linguistic transfer capabilities while requiring fewer\ncomputational resources compared to full fine-tuning for each language.\n", "date": "2025-04-11", "categories": []}
{"id": "2406.06401", "title": "INTERSPEECH 2009 Emotion Challenge Revisited: Benchmarking 15 Years of\n  Progress in Speech Emotion Recognition", "abstract": "  We revisit the INTERSPEECH 2009 Emotion Challenge -- the first ever speech\nemotion recognition (SER) challenge -- and evaluate a series of deep learning\nmodels that are representative of the major advances in SER research in the\ntime since then. We start by training each model using a fixed set of\nhyperparameters, and further fine-tune the best-performing models of that\ninitial setup with a grid search. Results are always reported on the official\ntest set with a separate validation set only used for early stopping. Most\nmodels score below or close to the official baseline, while they marginally\noutperform the original challenge winners after hyperparameter tuning. Our work\nillustrates that, despite recent progress, FAU-AIBO remains a very challenging\nbenchmark. An interesting corollary is that newer methods do not consistently\noutperform older ones, showing that progress towards `solving' SER is not\nnecessarily monotonic.\n", "date": "2024-06-10", "categories": []}
{"id": "2406.06665", "title": "Enrolment-based personalisation for improving individual-level fairness\n  in speech emotion recognition", "abstract": "  The expression of emotion is highly individualistic. However, contemporary\nspeech emotion recognition (SER) systems typically rely on population-level\nmodels that adopt a `one-size-fits-all' approach for predicting emotion.\nMoreover, standard evaluation practices measure performance also on the\npopulation level, thus failing to characterise how models work across different\nspeakers. In the present contribution, we present a new method for capitalising\non individual differences to adapt an SER model to each new speaker using a\nminimal set of enrolment utterances. In addition, we present novel evaluation\nschemes for measuring fairness across different speakers. Our findings show\nthat aggregated evaluation metrics may obfuscate fairness issues on the\nindividual-level, which are uncovered by our evaluation, and that our proposed\nmethod can improve performance both in aggregated and disaggregated terms.\n", "date": "2024-06-10", "categories": []}
{"id": "2504.07660", "title": "End-to-End Facial Expression Detection in Long Videos", "abstract": "  Facial expression detection involves two interrelated tasks: spotting, which\nidentifies the onset and offset of expressions, and recognition, which\nclassifies them into emotional categories. Most existing methods treat these\ntasks separately using a two-step training pipelines. A spotting model first\ndetects expression intervals. A recognition model then classifies the detected\nsegments. However, this sequential approach leads to error propagation,\ninefficient feature learning, and suboptimal performance due to the lack of\njoint optimization of the two tasks. We propose FEDN, an end-to-end Facial\nExpression Detection Network that jointly optimizes spotting and recognition.\nOur model introduces a novel attention-based feature extraction module,\nincorporating segment attention and sliding window attention to improve facial\nfeature learning. By unifying two tasks within a single network, we greatly\nreduce error propagation and enhance overall performance. Experiments on\nCASME}^2 and CASME^3 demonstrate state-of-the-art accuracy for both spotting\nand detection, underscoring the benefits of joint optimization for robust\nfacial expression detection in long videos.\n", "date": "2025-04-10", "categories": []}
{"id": "2412.12225", "title": "DLF: Disentangled-Language-Focused Multimodal Sentiment Analysis", "abstract": "  Multimodal Sentiment Analysis (MSA) leverages heterogeneous modalities, such\nas language, vision, and audio, to enhance the understanding of human\nsentiment. While existing models often focus on extracting shared information\nacross modalities or directly fusing heterogeneous modalities, such approaches\ncan introduce redundancy and conflicts due to equal treatment of all modalities\nand the mutual transfer of information between modality pairs. To address these\nissues, we propose a Disentangled-Language-Focused (DLF) multimodal\nrepresentation learning framework, which incorporates a feature disentanglement\nmodule to separate modality-shared and modality-specific information. To\nfurther reduce redundancy and enhance language-targeted features, four\ngeometric measures are introduced to refine the disentanglement process. A\nLanguage-Focused Attractor (LFA) is further developed to strengthen language\nrepresentation by leveraging complementary modality-specific information\nthrough a language-guided cross-attention mechanism. The framework also employs\nhierarchical predictions to improve overall accuracy. Extensive experiments on\ntwo popular MSA datasets, CMU-MOSI and CMU-MOSEI, demonstrate the significant\nperformance gains achieved by the proposed DLF framework. Comprehensive\nablation studies further validate the effectiveness of the feature\ndisentanglement module, language-focused attractor, and hierarchical\npredictions. Our code is available at https://github.com/pwang322/DLF.\n", "date": "2024-12-16", "categories": []}
{"id": "2504.06578", "title": "Attributes-aware Visual Emotion Representation Learning", "abstract": "  Visual emotion analysis or recognition has gained considerable attention due\nto the growing interest in understanding how images can convey rich semantics\nand evoke emotions in human perception. However, visual emotion analysis poses\ndistinctive challenges compared to traditional vision tasks, especially due to\nthe intricate relationship between general visual features and the different\naffective states they evoke, known as the affective gap. Researchers have used\ndeep representation learning methods to address this challenge of extracting\ngeneralized features from entire images. However, most existing methods\noverlook the importance of specific emotional attributes such as brightness,\ncolorfulness, scene understanding, and facial expressions. Through this paper,\nwe introduce A4Net, a deep representation network to bridge the affective gap\nby leveraging four key attributes: brightness (Attribute 1), colorfulness\n(Attribute 2), scene context (Attribute 3), and facial expressions (Attribute\n4). By fusing and jointly training all aspects of attribute recognition and\nvisual emotion analysis, A4Net aims to provide a better insight into emotional\ncontent in images. Experimental results show the effectiveness of A4Net,\nshowcasing competitive performance compared to state-of-the-art methods across\ndiverse visual emotion datasets. Furthermore, visualizations of activation maps\ngenerated by A4Net offer insights into its ability to generalize across\ndifferent visual emotion datasets.\n", "date": "2025-04-09", "categories": []}
{"id": "2504.03695", "title": "Are Anxiety Detection Models Generalizable? A Cross-Activity and\n  Cross-Population Study Using Wearables", "abstract": "  Anxiety-provoking activities, such as public speaking, can trigger heightened\nanxiety responses in individuals with anxiety disorders. Recent research\nsuggests that physiological signals, including electrocardiogram (ECG) and\nelectrodermal activity (EDA), collected via wearable devices, can be used to\ndetect anxiety in such contexts through machine learning models. However, the\ngeneralizability of these anxiety prediction models across different activities\nand diverse populations remains underexplored-an essential step for assessing\nmodel bias and fostering user trust in broader applications. To address this\ngap, we conducted a study with 111 participants who engaged in three\nanxiety-provoking activities. Utilizing both our collected dataset and two\nwell-known publicly available datasets, we evaluated the generalizability of\nanxiety detection models within participants (for both same-activity and\ncross-activity scenarios) and across participants (within-activity and\ncross-activity). In total, we trained and tested more than 3348 anxiety\ndetection models (using six classifiers, 31 feature sets, and 18 train-test\nconfigurations). Our results indicate that three key metrics-AUROC, recall for\nanxious states, and recall for non-anxious states-were slightly above the\nbaseline score of 0.5. The best AUROC scores ranged from 0.62 to 0.73, with\nrecall for the anxious class spanning 35.19% to 74.3%. Interestingly, model\nperformance (as measured by AUROC) remained relatively stable across different\nactivities and participant groups, though recall for the anxious class did\nexhibit some variation.\n", "date": "2025-03-24", "categories": []}
{"id": "2504.03707", "title": "Towards Practical Emotion Recognition: An Unsupervised Source-Free\n  Approach for EEG Domain Adaptation", "abstract": "  Emotion recognition is crucial for advancing mental health, healthcare, and\ntechnologies like brain-computer interfaces (BCIs). However, EEG-based emotion\nrecognition models face challenges in cross-domain applications due to the high\ncost of labeled data and variations in EEG signals from individual differences\nand recording conditions. Unsupervised domain adaptation methods typically\nrequire access to source domain data, which may not always be feasible in\nreal-world scenarios due to privacy and computational constraints. Source-free\nunsupervised domain adaptation (SF-UDA) has recently emerged as a solution,\nenabling target domain adaptation without source data, but its application in\nemotion recognition remains unexplored. We propose a novel SF-UDA approach for\nEEG-based emotion classification across domains, introducing a multi-stage\nframework that enhances model adaptability without requiring source data. Our\napproach incorporates Dual-Loss Adaptive Regularization (DLAR) to minimize\nprediction discrepancies on confident samples and align predictions with\nexpected pseudo-labels. Additionally, we introduce Localized Consistency\nLearning (LCL), which enforces local consistency by promoting similar\npredictions from reliable neighbors. These techniques together address domain\nshift and reduce the impact of noisy pseudo-labels, a key challenge in\ntraditional SF-UDA models. Experiments on two widely used datasets, DEAP and\nSEED, demonstrate the effectiveness of our method. Our approach significantly\noutperforms state-of-the-art methods, achieving 65.84% accuracy when trained on\nDEAP and tested on SEED, and 58.99% accuracy in the reverse scenario. It excels\nat detecting both positive and negative emotions, making it well-suited for\npractical emotion recognition applications.\n", "date": "2025-03-26", "categories": []}
{"id": "2504.05158", "title": "Leveraging Label Potential for Enhanced Multimodal Emotion Recognition", "abstract": "  Multimodal emotion recognition (MER) seeks to integrate various modalities to\npredict emotional states accurately. However, most current research focuses\nsolely on the fusion of audio and text features, overlooking the valuable\ninformation in emotion labels. This oversight could potentially hinder the\nperformance of existing methods, as emotion labels harbor rich, insightful\ninformation that could significantly aid MER. We introduce a novel model called\nLabel Signal-Guided Multimodal Emotion Recognition (LSGMER) to overcome this\nlimitation. This model aims to fully harness the power of emotion label\ninformation to boost the classification accuracy and stability of MER.\nSpecifically, LSGMER employs a Label Signal Enhancement module that optimizes\nthe representation of modality features by interacting with audio and text\nfeatures through label embeddings, enabling it to capture the nuances of\nemotions precisely. Furthermore, we propose a Joint Objective Optimization(JOO)\napproach to enhance classification accuracy by introducing the\nAttribution-Prediction Consistency Constraint (APC), which strengthens the\nalignment between fused features and emotion categories. Extensive experiments\nconducted on the IEMOCAP and MELD datasets have demonstrated the effectiveness\nof our proposed LSGMER model.\n", "date": "2025-04-07", "categories": []}
{"id": "2504.03010", "title": "Emotion Recognition Using Convolutional Neural Networks", "abstract": "  Emotion has an important role in daily life, as it helps people better\ncommunicate with and understand each other more efficiently. Facial expressions\ncan be classified into 7 categories: angry, disgust, fear, happy, neutral, sad\nand surprise. How to detect and recognize these seven emotions has become a\npopular topic in the past decade. In this paper, we develop an emotion\nrecognition system that can apply emotion recognition on both still images and\nreal-time videos by using deep learning.\n  We build our own emotion recognition classification and regression system\nfrom scratch, which includes dataset collection, data preprocessing , model\ntraining and testing. Given a certain image or a real-time video, our system is\nable to show the classification and regression results for all of the 7\nemotions. The proposed system is tested on 2 different datasets, and achieved\nan accuracy of over 80\\%. Moreover, the result obtained from real-time testing\nproves the feasibility of implementing convolutional neural networks in real\ntime to detect emotions accurately and efficiently.\n", "date": "2025-04-03", "categories": []}
{"id": "2501.13432", "title": "Emotion estimation from video footage with LSTM", "abstract": "  Emotion estimation in general is a field that has been studied for a long\ntime, and several approaches exist using machine learning. in this paper, we\npresent an LSTM model, that processes the blend-shapes produced by the library\nMediaPipe, for a face detected in a live stream of a camera, to estimate the\nmain emotion from the facial expressions, this model is trained on the FER2013\ndataset and delivers a result of 71% accuracy and 62% f1-score which meets the\naccuracy benchmark of the FER2013 dataset, with significantly reduced\ncomputation costs.\nhttps://github.com/Samir-atra/Emotion_estimation_from_video_footage_with_LSTM_ML_algorithm\n", "date": "2025-01-23", "categories": []}
{"id": "2407.12787", "title": "GameVibe: A Multimodal Affective Game Corpus", "abstract": "  As online video and streaming platforms continue to grow, affective computing\nresearch has undergone a shift towards more complex studies involving multiple\nmodalities. However, there is still a lack of readily available datasets with\nhigh-quality audiovisual stimuli. In this paper, we present GameVibe, a novel\naffect corpus which consists of multimodal audiovisual stimuli, including\nin-game behavioural observations and third-person affect traces for viewer\nengagement. The corpus consists of videos from a diverse set of publicly\navailable gameplay sessions across 30 games, with particular attention to\nensure high-quality stimuli with good audiovisual and gameplay diversity.\nFurthermore, we present an analysis on the reliability of the annotators in\nterms of inter-annotator agreement.\n", "date": "2024-06-17", "categories": []}
{"id": "2503.06405", "title": "Heterogeneous bimodal attention fusion for speech emotion recognition", "abstract": "  Multi-modal emotion recognition in conversations is a challenging problem due\nto the complex and complementary interactions between different modalities.\nAudio and textual cues are particularly important for understanding emotions\nfrom a human perspective. Most existing studies focus on exploring interactions\nbetween audio and text modalities at the same representation level. However, a\ncritical issue is often overlooked: the heterogeneous modality gap between\nlow-level audio representations and high-level text representations. To address\nthis problem, we propose a novel framework called Heterogeneous Bimodal\nAttention Fusion (HBAF) for multi-level multi-modal interaction in\nconversational emotion recognition. The proposed method comprises three key\nmodules: the uni-modal representation module, the multi-modal fusion module,\nand the inter-modal contrastive learning module. The uni-modal representation\nmodule incorporates contextual content into low-level audio representations to\nbridge the heterogeneous multi-modal gap, enabling more effective fusion. The\nmulti-modal fusion module uses dynamic bimodal attention and a dynamic gating\nmechanism to filter incorrect cross-modal relationships and fully exploit both\nintra-modal and inter-modal interactions. Finally, the inter-modal contrastive\nlearning module captures complex absolute and relative interactions between\naudio and text modalities. Experiments on the MELD and IEMOCAP datasets\ndemonstrate that the proposed HBAF method outperforms existing state-of-the-art\nbaselines.\n", "date": "2025-03-09", "categories": []}
{"id": "2503.22687", "title": "Qieemo: Speech Is All You Need in the Emotion Recognition in\n  Conversations", "abstract": "  Emotion recognition plays a pivotal role in intelligent human-machine\ninteraction systems. Multimodal approaches benefit from the fusion of diverse\nmodalities, thereby improving the recognition accuracy. However, the lack of\nhigh-quality multimodal data and the challenge of achieving optimal alignment\nbetween different modalities significantly limit the potential for improvement\nin multimodal approaches. In this paper, the proposed Qieemo framework\neffectively utilizes the pretrained automatic speech recognition (ASR) model\nbackbone which contains naturally frame aligned textual and emotional features,\nto achieve precise emotion classification solely based on the audio modality.\nFurthermore, we design the multimodal fusion (MMF) module and cross-modal\nattention (CMA) module in order to fuse the phonetic posteriorgram (PPG) and\nemotional features extracted by the ASR encoder for improving recognition\naccuracy. The experimental results on the IEMOCAP dataset demonstrate that\nQieemo outperforms the benchmark unimodal, multimodal, and self-supervised\nmodels with absolute improvements of 3.0%, 1.2%, and 1.9% respectively.\n", "date": "2025-03-05", "categories": []}
{"id": "2503.22711", "title": "Modeling speech emotion with label variance and analyzing performance\n  across speakers and unseen acoustic conditions", "abstract": "  Spontaneous speech emotion data usually contain perceptual grades where\ngraders assign emotion score after listening to the speech files. Such\nperceptual grades introduce uncertainty in labels due to grader opinion\nvariation. Grader variation is addressed by using consensus grades as\ngroundtruth, where the emotion with the highest vote is selected. Consensus\ngrades fail to consider ambiguous instances where a speech sample may contain\nmultiple emotions, as captured through grader opinion uncertainty. We\ndemonstrate that using the probability density function of the emotion grades\nas targets instead of the commonly used consensus grades, provide better\nperformance on benchmark evaluation sets compared to results reported in the\nliterature. We show that a saliency driven foundation model (FM) representation\nselection helps to train a state-of-the-art speech emotion model for both\ndimensional and categorical emotion recognition. Comparing representations\nobtained from different FMs, we observed that focusing on overall test-set\nperformance can be deceiving, as it fails to reveal the models generalization\ncapacity across speakers and gender. We demonstrate that performance evaluation\nacross multiple test-sets and performance analysis across gender and speakers\nare useful in assessing usefulness of emotion models. Finally, we demonstrate\nthat label uncertainty and data-skew pose a challenge to model evaluation,\nwhere instead of using the best hypothesis, it is useful to consider the 2- or\n3-best hypotheses.\n", "date": "2025-03-24", "categories": []}
{"id": "2503.22715", "title": "Hierarchical Adaptive Expert for Multimodal Sentiment Analysis", "abstract": "  Multimodal sentiment analysis has emerged as a critical tool for\nunderstanding human emotions across diverse communication channels. While\nexisting methods have made significant strides, they often struggle to\neffectively differentiate and integrate modality-shared and modality-specific\ninformation, limiting the performance of multimodal learning. To address this\nchallenge, we propose the Hierarchical Adaptive Expert for Multimodal Sentiment\nAnalysis (HAEMSA), a novel framework that synergistically combines evolutionary\noptimization, cross-modal knowledge transfer, and multi-task learning. HAEMSA\nemploys a hierarchical structure of adaptive experts to capture both global and\nlocal modality representations, enabling more nuanced sentiment analysis. Our\napproach leverages evolutionary algorithms to dynamically optimize network\narchitectures and modality combinations, adapting to both partial and full\nmodality scenarios. Extensive experiments demonstrate HAEMSA's superior\nperformance across multiple benchmark datasets. On CMU-MOSEI, HAEMSA achieves a\n2.6% increase in 7-class accuracy and a 0.059 decrease in MAE compared to the\nprevious best method. For CMU-MOSI, we observe a 6.3% improvement in 7-class\naccuracy and a 0.058 reduction in MAE. On IEMOCAP, HAEMSA outperforms the\nstate-of-the-art by 2.84% in weighted-F1 score for emotion recognition. These\nresults underscore HAEMSA's effectiveness in capturing complex multimodal\ninteractions and generalizing across different emotional contexts.\n", "date": "2025-03-25", "categories": []}
{"id": "2503.23257", "title": "FIESTA: Fisher Information-based Efficient Selective Test-time\n  Adaptation", "abstract": "  Robust facial expression recognition in unconstrained, \"in-the-wild\"\nenvironments remains challenging due to significant domain shifts between\ntraining and testing distributions. Test-time adaptation (TTA) offers a\npromising solution by adapting pre-trained models during inference without\nrequiring labeled test data. However, existing TTA approaches typically rely on\nmanually selecting which parameters to update, potentially leading to\nsuboptimal adaptation and high computational costs. This paper introduces a\nnovel Fisher-driven selective adaptation framework that dynamically identifies\nand updates only the most critical model parameters based on their importance\nas quantified by Fisher information. By integrating this principled parameter\nselection approach with temporal consistency constraints, our method enables\nefficient and effective adaptation specifically tailored for video-based facial\nexpression recognition. Experiments on the challenging AffWild2 benchmark\ndemonstrate that our approach significantly outperforms existing TTA methods,\nachieving a 7.7% improvement in F1 score over the base model while adapting\nonly 22,000 parameters-more than 20 times fewer than comparable methods. Our\nablation studies further reveal that parameter importance can be effectively\nestimated from minimal data, with sampling just 1-3 frames sufficient for\nsubstantial performance gains. The proposed approach not only enhances\nrecognition accuracy but also dramatically reduces computational overhead,\nmaking test-time adaptation more practical for real-world affective computing\napplications.\n", "date": "2025-03-29", "categories": []}
{"id": "2503.23547", "title": "NRC VAD Lexicon v2: Norms for Valence, Arousal, and Dominance for over\n  55k English Terms", "abstract": "  Factor analysis studies have shown that the primary dimensions of word\nmeaning are Valence (V), Arousal (A), and Dominance (D) (also referred to in\nsocial cognition research as Competence (C)). These dimensions impact various\naspects of our lives from social competence and emotion regulation to success\nin the work place and how we view the world. We present here the NRC VAD\nLexicon v2, which has human ratings of valence, arousal, and dominance for more\nthan 55,000 English words and phrases. Notably, it adds entries for $\\sim$25k\nadditional words to v1.0. It also now includes for the first time entries for\ncommon multi-word phrases (~10k). We show that the associations are highly\nreliable. The lexicon enables a wide variety of research in psychology, NLP,\npublic health, digital humanities, and social sciences. The NRC VAD Lexicon v2\nis made freely available for research through our project webpage.\n", "date": "2025-03-30", "categories": []}
{"id": "2503.23721", "title": "Unimodal-driven Distillation in Multimodal Emotion Recognition with\n  Dynamic Fusion", "abstract": "  Multimodal Emotion Recognition in Conversations (MERC) identifies emotional\nstates across text, audio and video, which is essential for intelligent\ndialogue systems and opinion analysis. Existing methods emphasize heterogeneous\nmodal fusion directly for cross-modal integration, but often suffer from\ndisorientation in multimodal learning due to modal heterogeneity and lack of\ninstructive guidance. In this work, we propose SUMMER, a novel heterogeneous\nmultimodal integration framework leveraging Mixture of Experts with\nHierarchical Cross-modal Fusion and Interactive Knowledge Distillation. Key\ncomponents include a Sparse Dynamic Mixture of Experts (SDMoE) for capturing\ndynamic token-wise interactions, a Hierarchical Cross-Modal Fusion (HCMF) for\neffective fusion of heterogeneous modalities, and Interactive Knowledge\nDistillation (IKD), which uses a pre-trained unimodal teacher to guide\nmultimodal fusion in latent and logit spaces. Experiments on IEMOCAP and MELD\nshow SUMMER outperforms state-of-the-art methods, particularly in recognizing\nminority and semantically similar emotions.\n", "date": "2025-03-31", "categories": []}
{"id": "2503.23990", "title": "BeMERC: Behavior-Aware MLLM-based Framework for Multimodal Emotion\n  Recognition in Conversation", "abstract": "  Multimodal emotion recognition in conversation (MERC), the task of\nidentifying the emotion label for each utterance in a conversation, is vital\nfor developing empathetic machines. Current MLLM-based MERC studies focus\nmainly on capturing the speaker's textual or vocal characteristics, but ignore\nthe significance of video-derived behavior information. Different from text and\naudio inputs, learning videos with rich facial expression, body language and\nposture, provides emotion trigger signals to the models for more accurate\nemotion predictions. In this paper, we propose a novel behavior-aware\nMLLM-based framework (BeMERC) to incorporate speaker's behaviors, including\nsubtle facial micro-expression, body language and posture, into a vanilla\nMLLM-based MERC model, thereby facilitating the modeling of emotional dynamics\nduring a conversation. Furthermore, BeMERC adopts a two-stage instruction\ntuning strategy to extend the model to the conversations scenario for\nend-to-end training of a MERC predictor. Experiments demonstrate that BeMERC\nachieves superior performance than the state-of-the-art methods on two\nbenchmark datasets, and also provides a detailed discussion on the significance\nof video-derived behavior information in MERC.\n", "date": "2025-03-31", "categories": []}
{"id": "2503.24057", "title": "AMMSM: Adaptive Motion Magnification and Sparse Mamba for\n  Micro-Expression Recognition", "abstract": "  Micro-expressions are typically regarded as unconscious manifestations of a\nperson's genuine emotions. However, their short duration and subtle signals\npose significant challenges for downstream recognition. We propose a multi-task\nlearning framework named the Adaptive Motion Magnification and Sparse Mamba\n(AMMSM) to address this. This framework aims to enhance the accurate capture of\nmicro-expressions through self-supervised subtle motion magnification, while\nthe sparse spatial selection Mamba architecture combines sparse activation with\nthe advanced Visual Mamba model to model key motion regions and their valuable\nrepresentations more effectively. Additionally, we employ evolutionary search\nto optimize the magnification factor and the sparsity ratios of spatial\nselection, followed by fine-tuning to improve performance further. Extensive\nexperiments on two standard datasets demonstrate that the proposed AMMSM\nachieves state-of-the-art (SOTA) accuracy and robustness.\n", "date": "2025-03-31", "categories": []}
{"id": "2503.20919", "title": "GatedxLSTM: A Multimodal Affective Computing Approach for Emotion\n  Recognition in Conversations", "abstract": "  Affective Computing (AC) is essential for advancing Artificial General\nIntelligence (AGI), with emotion recognition serving as a key component.\nHowever, human emotions are inherently dynamic, influenced not only by an\nindividual's expressions but also by interactions with others, and\nsingle-modality approaches often fail to capture their full dynamics.\nMultimodal Emotion Recognition (MER) leverages multiple signals but\ntraditionally relies on utterance-level analysis, overlooking the dynamic\nnature of emotions in conversations. Emotion Recognition in Conversation (ERC)\naddresses this limitation, yet existing methods struggle to align multimodal\nfeatures and explain why emotions evolve within dialogues. To bridge this gap,\nwe propose GatedxLSTM, a novel speech-text multimodal ERC model that explicitly\nconsiders voice and transcripts of both the speaker and their conversational\npartner(s) to identify the most influential sentences driving emotional shifts.\nBy integrating Contrastive Language-Audio Pretraining (CLAP) for improved\ncross-modal alignment and employing a gating mechanism to emphasise emotionally\nimpactful utterances, GatedxLSTM enhances both interpretability and\nperformance. Additionally, the Dialogical Emotion Decoder (DED) refines emotion\npredictions by modelling contextual dependencies. Experiments on the IEMOCAP\ndataset demonstrate that GatedxLSTM achieves state-of-the-art (SOTA)\nperformance among open-source methods in four-class emotion classification.\nThese results validate its effectiveness for ERC applications and provide an\ninterpretability analysis from a psychological perspective.\n", "date": "2025-03-26", "categories": []}
{"id": "2503.21480", "title": "OmniVox: Zero-Shot Emotion Recognition with Omni-LLMs", "abstract": "  The use of omni-LLMs (large language models that accept any modality as\ninput), particularly for multimodal cognitive state tasks involving speech, is\nunderstudied. We present OmniVox, the first systematic evaluation of four\nomni-LLMs on the zero-shot emotion recognition task. We evaluate on two widely\nused multimodal emotion benchmarks: IEMOCAP and MELD, and find zero-shot\nomni-LLMs outperform or are competitive with fine-tuned audio models. Alongside\nour audio-only evaluation, we also evaluate omni-LLMs on text only and text and\naudio. We present acoustic prompting, an audio-specific prompting strategy for\nomni-LLMs which focuses on acoustic feature analysis, conversation context\nanalysis, and step-by-step reasoning. We compare our acoustic prompting to\nminimal prompting and full chain-of-thought prompting techniques. We perform a\ncontext window analysis on IEMOCAP and MELD, and find that using context helps,\nespecially on IEMOCAP. We conclude with an error analysis on the generated\nacoustic reasoning outputs from the omni-LLMs.\n", "date": "2025-03-27", "categories": []}
{"id": "2503.21806", "title": "Large Language Models Meet Contrastive Learning: Zero-Shot Emotion\n  Recognition Across Languages", "abstract": "  Multilingual speech emotion recognition aims to estimate a speaker's\nemotional state using a contactless method across different languages. However,\nvariability in voice characteristics and linguistic diversity poses significant\nchallenges for zero-shot speech emotion recognition, especially with\nmultilingual datasets. In this paper, we propose leveraging contrastive\nlearning to refine multilingual speech features and extend large language\nmodels for zero-shot multilingual speech emotion estimation. Specifically, we\nemploy a novel two-stage training framework to align speech signals with\nlinguistic features in the emotional space, capturing both emotion-aware and\nlanguage-agnostic speech representations. To advance research in this field, we\nintroduce a large-scale synthetic multilingual speech emotion dataset, M5SER.\nOur experiments demonstrate the effectiveness of the proposed method in both\nspeech emotion recognition and zero-shot multilingual speech emotion\nrecognition, including previously unseen datasets and languages.\n", "date": "2025-03-25", "categories": []}
{"id": "2503.21927", "title": "Hybrid Emotion Recognition: Enhancing Customer Interactions Through\n  Acoustic and Textual Analysis", "abstract": "  This research presents a hybrid emotion recognition system integrating\nadvanced Deep Learning, Natural Language Processing (NLP), and Large Language\nModels (LLMs) to analyze audio and textual data for enhancing customer\ninteractions in contact centers. By combining acoustic features with textual\nsentiment analysis, the system achieves nuanced emotion detection, addressing\nthe limitations of traditional approaches in understanding complex emotional\nstates. Leveraging LSTM and CNN models for audio analysis and DistilBERT for\ntextual evaluation, the methodology accommodates linguistic and cultural\nvariations while ensuring real-time processing. Rigorous testing on diverse\ndatasets demonstrates the system's robustness and accuracy, highlighting its\npotential to transform customer service by enabling personalized, empathetic\ninteractions and improving operational efficiency. This research establishes a\nfoundation for more intelligent and human-centric digital communication,\nredefining customer service standards.\n", "date": "2025-03-27", "categories": []}
{"id": "2503.21690", "title": "CMED: A Child Micro-Expression Dataset", "abstract": "  Micro-expressions are short bursts of emotion that are difficult to hide.\nTheir detection in children is an important cue to assist psychotherapists in\nconducting better therapy. However, existing research on the detection of\nmicro-expressions has focused on adults, whose expressions differ in their\ncharacteristics from those of children. The lack of research is a direct\nconsequence of the lack of a child-based micro-expressions dataset as it is\nmuch more challenging to capture children's facial expressions due to the lack\nof predictability and controllability. This study compiles a dataset of\nspontaneous child micro-expression videos, the first of its kind, to the best\nof the authors knowledge. The dataset is captured in the wild using video\nconferencing software. This dataset enables us to then explore key features and\ndifferences between adult and child micro-expressions. This study also\nestablishes a baseline for the automated spotting and recognition of\nmicro-expressions in children using three approaches comprising of hand-created\nand learning-based approaches.\n", "date": "2025-03-27", "categories": []}
{"id": "2503.20739", "title": "Emotion Detection and Music Recommendation System", "abstract": "  As artificial intelligence becomes more and more ingrained in daily life, we\npresent a novel system that uses deep learning for music recommendation and\nemotion-based detection. Through the use of facial recognition and the DeepFace\nframework, our method analyses human emotions in real-time and then plays music\nthat reflects the mood it has discovered. The system uses a webcam to take\npictures, analyses the most common facial expression, and then pulls a playlist\nfrom local storage that corresponds to the mood it has detected. An engaging\nand customised experience is ensured by allowing users to manually change the\nsong selection via a dropdown menu or navigation buttons. By continuously\nlooping over the playlist, the technology guarantees continuity. The objective\nof our system is to improve emotional well-being through music therapy by\noffering a responsive and automated music-selection experience.\n", "date": "2025-03-26", "categories": []}
{"id": "2403.08824", "title": "Computational Analysis of Stress, Depression and Engagement in Mental\n  Health: A Survey", "abstract": "  Analysis of stress, depression and engagement is less common and more complex\nthan that of frequently discussed emotions such as happiness, sadness, fear and\nanger. The importance of these psychological states has been increasingly\nrecognized due to their implications for mental health and well-being. Stress\nand depression are interrelated and together they impact engagement in daily\ntasks, highlighting the need to explore their interplay. This survey is the\nfirst to simultaneously explore computational methods for analyzing stress,\ndepression and engagement. We present a taxonomy and timeline of the\ncomputational approaches used to analyze them and we discuss the most commonly\nused datasets and input modalities, along with the categories and generic\npipeline of these approaches. Subsequently, we describe state-of-the-art\ncomputational approaches, including a performance summary on the most commonly\nused datasets. Following this, we explore the applications of stress,\ndepression and engagement analysis, along with the associated challenges,\nlimitations and future research directions.\n", "date": "2024-03-09", "categories": []}
{"id": "2503.10530", "title": "Lightweight Models for Emotional Analysis in Video", "abstract": "  In this study, we present an approach for efficient spatiotemporal feature\nextraction using MobileNetV4 and a multi-scale 3D MLP-Mixer-based temporal\naggregation module. MobileNetV4, with its Universal Inverted Bottleneck (UIB)\nblocks, serves as the backbone for extracting hierarchical feature\nrepresentations from input image sequences, ensuring both computational\nefficiency and rich semantic encoding. To capture temporal dependencies, we\nintroduce a three-level MLP-Mixer module, which processes spatial features at\nmultiple resolutions while maintaining structural integrity. Experimental\nresults on the ABAW 8th competition demonstrate the effectiveness of our\napproach, showing promising performance in affective behavior analysis. By\nintegrating an efficient vision backbone with a structured temporal modeling\nmechanism, the proposed framework achieves a balance between computational\nefficiency and predictive accuracy, making it well-suited for real-time\napplications in mobile and embedded computing environments.\n", "date": "2025-03-13", "categories": []}
{"id": "2503.10603", "title": "Technical Approach for the EMI Challenge in the 8th Affective Behavior\n  Analysis in-the-Wild Competition", "abstract": "  Emotional Mimicry Intensity (EMI) estimation plays a pivotal role in\nunderstanding human social behavior and advancing human-computer interaction.\nThe core challenges lie in dynamic correlation modeling and robust fusion of\nmultimodal temporal signals. To address the limitations of existing\nmethods--insufficient exploitation of cross-modal synergies, sensitivity to\nnoise, and constrained fine-grained alignment capabilities--this paper proposes\na dual-stage cross-modal alignment framework. Stage 1 develops vision-text and\naudio-text contrastive learning networks based on a CLIP architecture,\nachieving preliminary feature-space alignment through modality-decoupled\npre-training. Stage 2 introduces a temporal-aware dynamic fusion module\nintegrating Temporal Convolutional Networks (TCN) and gated bidirectional LSTM\nto capture macro-evolution patterns of facial expressions and local dynamics of\nacoustic features, respectively. A novel quality-guided fusion strategy further\nenables differentiable weight allocation for modality compensation under\nocclusion and noise. Experiments on the Hume-Vidmimic2 dataset demonstrate\nsuperior performance with an average Pearson correlation coefficient of 0.51\nacross six emotion dimensions on the validate set. Remarkably, our method\nachieved 0.68 on the test set, securing runner-up in the EMI Challenge Track of\nthe 8th ABAW (Affective Behavior Analysis in the Wild) Competition, offering a\nnovel pathway for fine-grained emotion analysis in open environments.\n", "date": "2025-03-13", "categories": []}
{"id": "2503.18964", "title": "Unifying EEG and Speech for Emotion Recognition: A Two-Step Joint\n  Learning Framework for Handling Missing EEG Data During Inference", "abstract": "  Computer interfaces are advancing towards using multi-modalities to enable\nbetter human-computer interactions. The use of automatic emotion recognition\n(AER) can make the interactions natural and meaningful thereby enhancing the\nuser experience. Though speech is the most direct and intuitive modality for\nAER, it is not reliable because it can be intentionally faked by humans. On the\nother hand, physiological modalities like EEG, are more reliable and impossible\nto fake. However, use of EEG is infeasible for realistic scenarios usage\nbecause of the need for specialized recording setup. In this paper, one of our\nprimary aims is to ride on the reliability of the EEG modality to facilitate\nrobust AER on the speech modality. Our approach uses both the modalities during\ntraining to reliably identify emotion at the time of inference, even in the\nabsence of the more reliable EEG modality. We propose, a two-step joint\nmulti-modal learning approach (JMML) that exploits both the intra- and inter-\nmodal characteristics to construct emotion embeddings that enrich the\nperformance of AER. In the first step, using JEC-SSL, intra-modal learning is\ndone independently on the individual modalities. This is followed by an\ninter-modal learning using the proposed extended variant of deep canonically\ncorrelated cross-modal autoencoder (E-DCC-CAE). The approach learns the joint\nproperties of both the modalities by mapping them into a common representation\nspace, such that the modalities are maximally correlated. These emotion\nembeddings, hold properties of both the modalities there by enhancing the\nperformance of ML classifier used for AER. Experimental results show the\nefficacy of the proposed approach. To best of our knowledge, this is the first\nattempt to combine speech and EEG with joint multi-modal learning approach for\nreliable AER.\n", "date": "2025-03-20", "categories": []}
{"id": "2503.19642", "title": "Exploring Cultural Nuances in Emotion Perception Across 15 African\n  Languages", "abstract": "  Understanding how emotions are expressed across languages is vital for\nbuilding culturally-aware and inclusive NLP systems. However, emotion\nexpression in African languages is understudied, limiting the development of\neffective emotion detection tools in these languages. In this work, we present\na cross-linguistic analysis of emotion expression in 15 African languages. We\nexamine four key dimensions of emotion representation: text length, sentiment\npolarity, emotion co-occurrence, and intensity variations. Our findings reveal\ndiverse language-specific patterns in emotional expression -- with Somali texts\ntypically longer, while others like IsiZulu and Algerian Arabic show more\nconcise emotional expression. We observe a higher prevalence of negative\nsentiment in several Nigerian languages compared to lower negativity in\nlanguages like IsiXhosa. Further, emotion co-occurrence analysis demonstrates\nstrong cross-linguistic associations between specific emotion pairs\n(anger-disgust, sadness-fear), suggesting universal psychological connections.\nIntensity distributions show multimodal patterns with significant variations\nbetween language families; Bantu languages display similar yet distinct\nprofiles, while Afroasiatic languages and Nigerian Pidgin demonstrate wider\nintensity ranges. These findings highlight the need for language-specific\napproaches to emotion detection while identifying opportunities for transfer\nlearning across related languages.\n", "date": "2025-03-25", "categories": []}
{"id": "2503.19677", "title": "Deep Learning for Speech Emotion Recognition: A CNN Approach Utilizing\n  Mel Spectrograms", "abstract": "  This paper explores the application of Convolutional Neural Networks CNNs for\nclassifying emotions in speech through Mel Spectrogram representations of audio\nfiles. Traditional methods such as Gaussian Mixture Models and Hidden Markov\nModels have proven insufficient for practical deployment, prompting a shift\ntowards deep learning techniques. By transforming audio data into a visual\nformat, the CNN model autonomously learns to identify intricate patterns,\nenhancing classification accuracy. The developed model is integrated into a\nuser-friendly graphical interface, facilitating realtime predictions and\npotential applications in educational environments. The study aims to advance\nthe understanding of deep learning in speech emotion recognition, assess the\nmodels feasibility, and contribute to the integration of technology in learning\ncontexts\n", "date": "2025-03-25", "categories": []}
{"id": "2401.06810", "title": "TONE: A 3-Tiered ONtology for Emotion analysis", "abstract": "  Emotions have played an important part in many sectors, including psychology,\nmedicine, mental health, computer science, and so on, and categorizing them has\nproven extremely useful in separating one emotion from another. Emotions can be\nclassified using the following two methods: (1) The supervised method's\nefficiency is strongly dependent on the size and domain of the data collected.\nA categorization established using relevant data from one domain may not work\nwell in another. (2) An unsupervised method that uses either domain expertise\nor a knowledge base of emotion types already exists. Though this second\napproach provides a suitable and generic categorization of emotions and is\ncost-effective, the literature doesn't possess a publicly available knowledge\nbase that can be directly applied to any emotion categorization-related task.\nThis pushes us to create a knowledge base that can be used for emotion\nclassification across domains, and ontology is often used for this purpose. In\nthis study, we provide TONE, an emotion-based ontology that effectively creates\nan emotional hierarchy based on Dr. Gerrod Parrot's group of emotions. In\naddition to ontology development, we introduce a semi-automated vocabulary\nconstruction process to generate a detailed collection of terms for emotions at\neach tier of the hierarchy. We also demonstrate automated methods for\nestablishing three sorts of dependencies in order to develop linkages between\ndifferent emotions. Our human and automatic evaluation results show the\nontology's quality. Furthermore, we describe three distinct use cases that\ndemonstrate the applicability of our ontology.\n", "date": "2024-01-11", "categories": []}
{"id": "2406.08353", "title": "Speech Emotion Recognition with ASR Transcripts: A Comprehensive Study\n  on Word Error Rate and Fusion Techniques", "abstract": "  Text data is commonly utilized as a primary input to enhance Speech Emotion\nRecognition (SER) performance and reliability. However, the reliance on\nhuman-transcribed text in most studies impedes the development of practical SER\nsystems, creating a gap between in-lab research and real-world scenarios where\nAutomatic Speech Recognition (ASR) serves as the text source. Hence, this study\nbenchmarks SER performance using ASR transcripts with varying Word Error Rates\n(WERs) from eleven models on three well-known corpora: IEMOCAP, CMU-MOSI, and\nMSP-Podcast. Our evaluation includes both text-only and bimodal SER with six\nfusion techniques, aiming for a comprehensive analysis that uncovers novel\nfindings and challenges faced by current SER research. Additionally, we propose\na unified ASR error-robust framework integrating ASR error correction and\nmodality-gated fusion, achieving lower WER and higher SER results compared to\nthe best-performing ASR transcript. These findings provide insights into SER\nwith ASR assistance, especially for real-world applications.\n", "date": "2024-06-12", "categories": []}
{"id": "2410.15029", "title": "Enhancing Multimodal Sentiment Analysis for Missing Modality through\n  Self-Distillation and Unified Modality Cross-Attention", "abstract": "  In multimodal sentiment analysis, collecting text data is often more\nchallenging than video or audio due to higher annotation costs and inconsistent\nautomatic speech recognition (ASR) quality. To address this challenge, our\nstudy has developed a robust model that effectively integrates multimodal\nsentiment information, even in the absence of text modality. Specifically, we\nhave developed a Double-Flow Self-Distillation Framework, including Unified\nModality Cross-Attention (UMCA) and Modality Imagination Autoencoder (MIA),\nwhich excels at processing both scenarios with complete modalities and those\nwith missing text modality. In detail, when the text modality is missing, our\nframework uses the LLM-based model to simulate the text representation from the\naudio modality, while the MIA module supplements information from the other two\nmodalities to make the simulated text representation similar to the real text\nrepresentation. To further align the simulated and real representations, and to\nenable the model to capture the continuous nature of sample orders in sentiment\nvalence regression tasks, we have also introduced the Rank-N Contrast (RNC)\nloss function. When testing on the CMU-MOSEI, our model achieved outstanding\nperformance on MAE and significantly outperformed other models when text\nmodality is missing. The code is available at:\nhttps://github.com/WarmCongee/SDUMC\n", "date": "2024-10-19", "categories": []}
{"id": "2503.05858", "title": "Bimodal Connection Attention Fusion for Speech Emotion Recognition", "abstract": "  Multi-modal emotion recognition is challenging due to the difficulty of\nextracting features that capture subtle emotional differences. Understanding\nmulti-modal interactions and connections is key to building effective bimodal\nspeech emotion recognition systems. In this work, we propose Bimodal Connection\nAttention Fusion (BCAF) method, which includes three main modules: the\ninteractive connection network, the bimodal attention network, and the\ncorrelative attention network. The interactive connection network uses an\nencoder-decoder architecture to model modality connections between audio and\ntext while leveraging modality-specific features. The bimodal attention network\nenhances semantic complementation and exploits intra- and inter-modal\ninteractions. The correlative attention network reduces cross-modal noise and\ncaptures correlations between audio and text. Experiments on the MELD and\nIEMOCAP datasets demonstrate that the proposed BCAF method outperforms existing\nstate-of-the-art baselines.\n", "date": "2025-03-08", "categories": []}
{"id": "2503.16883", "title": "Assessing the Reliability and Validity of GPT-4 in Annotating Emotion\n  Appraisal Ratings", "abstract": "  Appraisal theories suggest that emotions arise from subjective evaluations of\nevents, referred to as appraisals. The taxonomy of appraisals is quite diverse,\nand they are usually given ratings on a Likert scale to be annotated in an\nexperiencer-annotator or reader-annotator paradigm. This paper studies GPT-4 as\na reader-annotator of 21 specific appraisal ratings in different prompt\nsettings, aiming to evaluate and improve its performance compared to human\nannotators. We found that GPT-4 is an effective reader-annotator that performs\nclose to or even slightly better than human annotators, and its results can be\nsignificantly improved by using a majority voting of five completions. GPT-4\nalso effectively predicts appraisal ratings and emotion labels using a single\nprompt, but adding instruction complexity results in poorer performance. We\nalso found that longer event descriptions lead to more accurate annotations for\nboth model and human annotator ratings. This work contributes to the growing\nusage of LLMs in psychology and the strategies for improving GPT-4 performance\nin annotating appraisals.\n", "date": "2025-03-21", "categories": []}
{"id": "2503.17453", "title": "Feature-Based Dual Visual Feature Extraction Model for Compound\n  Multimodal Emotion Recognition", "abstract": "  This article presents our results for the eighth Affective Behavior Analysis\nin-the-wild (ABAW) competition.Multimodal emotion recognition (ER) has\nimportant applications in affective computing and human-computer interaction.\nHowever, in the real world, compound emotion recognition faces greater issues\nof uncertainty and modal conflicts. For the Compound Expression (CE)\nRecognition Challenge,this paper proposes a multimodal emotion recognition\nmethod that fuses the features of Vision Transformer (ViT) and Residual Network\n(ResNet). We conducted experiments on the C-EXPR-DB and MELD datasets. The\nresults show that in scenarios with complex visual and audio cues (such as\nC-EXPR-DB), the model that fuses the features of ViT and ResNet exhibits\nsuperior performance.Our code are avalible on\nhttps://github.com/MyGitHub-ax/8th_ABAW\n", "date": "2025-03-21", "categories": []}
{"id": "2503.17625", "title": "AI-Based Screening for Depression and Social Anxiety Through Eye\n  Tracking: An Exploratory Study", "abstract": "  Well-being is a dynamic construct that evolves over time and fluctuates\nwithin individuals, presenting challenges for accurate quantification. Reduced\nwell-being is often linked to depression or anxiety disorders, which are\ncharacterised by biases in visual attention towards specific stimuli, such as\nhuman faces. This paper introduces a novel approach to AI-assisted screening of\naffective disorders by analysing visual attention scan paths using\nconvolutional neural networks (CNNs). Data were collected from two studies\nexamining (1) attentional tendencies in individuals diagnosed with major\ndepression and (2) social anxiety. These data were processed using residual\nCNNs through images generated from eye-gaze patterns. Experimental results,\nobtained with ResNet architectures, demonstrated an average accuracy of 48% for\na three-class system and 62% for a two-class system. Based on these exploratory\nfindings, we propose that this method could be employed in rapid, ecological,\nand effective mental health screening systems to assess well-being through\neye-tracking.\n", "date": "2025-03-22", "categories": []}
{"id": "2503.12261", "title": "United we stand, Divided we fall: Handling Weak Complementary\n  Relationships for Audio-Visual Emotion Recognition in Valence-Arousal Space", "abstract": "  Audio and visual modalities are two predominant contact-free channels in\nvideos, which are often expected to carry a complementary relationship with\neach other. However, they may not always complement each other, resulting in\npoor audio-visual feature representations. In this paper, we introduce Gated\nRecursive Joint Cross Attention (GRJCA) using a gating mechanism that can\nadaptively choose the most relevant features to effectively capture the\nsynergic relationships across audio and visual modalities. Specifically, we\nimprove the performance of Recursive Joint Cross-Attention (RJCA) by\nintroducing a gating mechanism to control the flow of information between the\ninput features and the attended features of multiple iterations depending on\nthe strength of their complementary relationship. For instance, if the\nmodalities exhibit strong complementary relationships, the gating mechanism\nemphasizes cross-attended features, otherwise non-attended features. To further\nimprove the performance of the system, we also explored a hierarchical gating\napproach by introducing a gating mechanism at every iteration, followed by\nhigh-level gating across the gated outputs of each iteration. The proposed\napproach improves the performance of RJCA model by adding more flexibility to\ndeal with weak complementary relationships across audio and visual modalities.\nExtensive experiments are conducted on the challenging Affwild2 dataset to\ndemonstrate the robustness of the proposed approach. By effectively handling\nthe weak complementary relationships across the audio and visual modalities,\nthe proposed model achieves a Concordance Correlation Coefficient (CCC) of\n0.561 (0.623) and 0.620 (0.660) for valence and arousal respectively on the\ntest set (validation set).\n", "date": "2025-03-15", "categories": []}
{"id": "2503.16454", "title": "An Audio-Visual Fusion Emotion Generation Model Based on Neuroanatomical\n  Alignment", "abstract": "  In the field of affective computing, traditional methods for generating\nemotions predominantly rely on deep learning techniques and large-scale emotion\ndatasets. However, deep learning techniques are often complex and difficult to\ninterpret, and standardizing large-scale emotional datasets are difficult and\ncostly to establish. To tackle these challenges, we introduce a novel framework\nnamed Audio-Visual Fusion for Brain-like Emotion Learning(AVF-BEL). In contrast\nto conventional brain-inspired emotion learning methods, this approach improves\nthe audio-visual emotion fusion and generation model through the integration of\nmodular components, thereby enabling more lightweight and interpretable emotion\nlearning and generation processes. The framework simulates the integration of\nthe visual, auditory, and emotional pathways of the brain, optimizes the fusion\nof emotional features across visual and auditory modalities, and improves upon\nthe traditional Brain Emotional Learning (BEL) model. The experimental results\nindicate a significant improvement in the similarity of the audio-visual fusion\nemotion learning generation model compared to single-modality visual and\nauditory emotion learning and generation model. Ultimately, this aligns with\nthe fundamental phenomenon of heightened emotion generation facilitated by the\nintegrated impact of visual and auditory stimuli. This contribution not only\nenhances the interpretability and efficiency of affective intelligence but also\nprovides new insights and pathways for advancing affective computing\ntechnology. Our source code can be accessed here:\nhttps://github.com/OpenHUTB/emotion}{https://github.com/OpenHUTB/emotion.\n", "date": "2025-02-21", "categories": []}
{"id": "2503.15876", "title": "DeepPsy-Agent: A Stage-Aware and Deep-Thinking Emotional Support Agent\n  System", "abstract": "  This paper introduces DeepPsy-Agent, an innovative psychological support\nsystem that combines the three-stage helping theory in psychology with deep\nlearning techniques. The system consists of two core components: (1) a\nmulti-stage response-capable dialogue model (\\textit{deeppsy-chat}), which\nenhances reasoning capabilities through stage-awareness and deep-thinking\nanalysis to generate high-quality responses; and (2) a real-time stage\ntransition detection model that identifies contextual shifts to guide the\ndialogue towards more effective intervention stages. Based on 30,000 real\npsychological hotline conversations, we employ AI-simulated dialogues and\nexpert re-annotation strategies to construct a high-quality multi-turn dialogue\ndataset. Experimental results demonstrate that DeepPsy-Agent outperforms\ngeneral-purpose large language models (LLMs) in key metrics such as problem\nexposure completeness, cognitive restructuring success rate, and action\nadoption rate. Ablation studies further validate the effectiveness of\nstage-awareness and deep-thinking modules, showing that stage information\ncontributes 42.3\\% to performance, while the deep-thinking module increases\nroot-cause identification by 58.3\\% and reduces ineffective suggestions by\n72.1\\%. This system addresses critical challenges in AI-based psychological\nsupport through dynamic dialogue management and deep reasoning, advancing\nintelligent mental health services.\n", "date": "2025-03-20", "categories": []}
{"id": "2503.14513", "title": "Synthetic Data Generation of Body Motion Data by Neural Gas Network for\n  Emotion Recognition", "abstract": "  In the domain of emotion recognition using body motion, the primary challenge\nlies in the scarcity of diverse and generalizable datasets. Automatic emotion\nrecognition uses machine learning and artificial intelligence techniques to\nrecognize a person's emotional state from various data types, such as text,\nimages, sound, and body motion. Body motion poses unique challenges as many\nfactors, such as age, gender, ethnicity, personality, and illness, affect its\nappearance, leading to a lack of diverse and robust datasets specifically for\nemotion recognition. To address this, employing Synthetic Data Generation (SDG)\nmethods, such as Generative Adversarial Networks (GANs) and Variational Auto\nEncoders (VAEs), offers potential solutions, though these methods are often\ncomplex. This research introduces a novel application of the Neural Gas Network\n(NGN) algorithm for synthesizing body motion data and optimizing diversity and\ngeneration speed. By learning skeletal structure topology, the NGN fits the\nneurons or gas particles on body joints. Generated gas particles, which form\nthe skeletal structure later on, will be used to synthesize the new body\nposture. By attaching body postures over frames, the final synthetic body\nmotion appears. We compared our generated dataset against others generated by\nGANs, VAEs, and another benchmark algorithm, using benchmark metrics such as\nFr\\'echet Inception Distance (FID), Diversity, and a few more. Furthermore, we\ncontinued evaluation using classification metrics such as accuracy, precision,\nrecall, and a few others. Joint-related features or kinematic parameters were\nextracted, and the system assessed model performance against unseen data. Our\nfindings demonstrate that the NGN algorithm produces more realistic and\nemotionally distinct body motion data and does so with more synthesizing speed\nthan existing methods.\n", "date": "2025-03-11", "categories": []}
{"id": "2503.11115", "title": "Solution for 8th Competition on Affective & Behavior Analysis\n  in-the-wild", "abstract": "  In this report, we present our solution for the Action Unit (AU) Detection\nChallenge, in 8th Competition on Affective Behavior Analysis in-the-wild. In\norder to achieve robust and accurate classification of facial action unit in\nthe wild environment, we introduce an innovative method that leverages\naudio-visual multimodal data. Our method employs ConvNeXt as the image encoder\nand uses Whisper to extract Mel spectrogram features. For these features, we\nutilize a Transformer encoder-based feature fusion module to integrate the\naffective information embedded in audio and image features. This ensures the\nprovision of rich high-dimensional feature representations for the subsequent\nmultilayer perceptron (MLP) trained on the Aff-Wild2 dataset, enhancing the\naccuracy of AU detection.\n", "date": "2025-03-14", "categories": []}
{"id": "2407.16945", "title": "Affective Behaviour Analysis via Progressive Learning", "abstract": "  Affective Behavior Analysis aims to develop emotionally intelligent\ntechnology that can recognize and respond to human emotions. To advance this\nfield, the 7th Affective Behavior Analysis in-the-wild (ABAW) competition holds\nthe Multi-Task Learning Challenge based on the s-Aff-Wild2 database. The\nparticipants are required to develop a framework that achieves Valence-Arousal\nEstimation, Expression Recognition, and AU detection simultaneously. To achieve\nthis goal, we propose a progressive multi-task learning framework that fully\nleverages the distinct focuses of each task on facial emotion features.\nSpecifically, our method design can be summarized into three main aspects: 1)\nSeparate Training and Joint Training: We first train each task model separately\nand then perform joint training based on the pre-trained models, fully\nutilizing the feature focus aspects of each task to improve the overall\nframework performance. 2) Feature Fusion and Temporal Modeling:} We investigate\neffective strategies for fusing features extracted from each task-specific\nmodel and incorporate temporal feature modeling during the joint training\nphase, which further refines the performance of each task. 3) Joint Training\nStrategy Optimization: To identify the optimal joint training approach, we\nconduct a comprehensive strategy search, experimenting with various task\ncombinations and training methodologies to further elevate the overall\nperformance of each task. According to the official results, our team achieves\nfirst place in the MTL challenge with a total score of 1.5286 (i.e., AU F-score\n0.5580, Expression F-score 0.4286, CCC VA score 0.5420). Our code is publicly\navailable at https://github.com/YenanLiu/ABAW7th.\n", "date": "2024-07-24", "categories": []}
{"id": "2503.09929", "title": "Emotion Recognition with CLIP and Sequential Learning", "abstract": "  Human emotion recognition plays a crucial role in facilitating seamless\ninteractions between humans and computers. In this paper, we present our\ninnovative methodology for tackling the Valence-Arousal (VA) Estimation\nChallenge, the Expression Recognition Challenge, and the Action Unit (AU)\nDetection Challenge, all within the framework of the 8th Workshop and\nCompetition on Affective Behavior Analysis in-the-wild (ABAW).\n  Our approach introduces a novel framework aimed at enhancing continuous\nemotion recognition. This is achieved by fine-tuning the CLIP model with the\naff-wild2 dataset, which provides annotated expression labels. The result is a\nfine-tuned model that serves as an efficient visual feature extractor,\nsignificantly improving its robustness. To further boost the performance of\ncontinuous emotion recognition, we incorporate Temporal Convolutional Network\n(TCN) modules alongside Transformer Encoder modules into our system\narchitecture. The integration of these advanced components allows our model to\noutperform baseline performance, demonstrating its ability to recognize human\nemotions with greater accuracy and efficiency.\n", "date": "2025-03-13", "categories": []}
{"id": "2503.10104", "title": "Mamba-VA: A Mamba-based Approach for Continuous Emotion Recognition in\n  Valence-Arousal Space", "abstract": "  Continuous Emotion Recognition (CER) plays a crucial role in intelligent\nhuman-computer interaction, mental health monitoring, and autonomous driving.\nEmotion modeling based on the Valence-Arousal (VA) space enables a more nuanced\nrepresentation of emotional states. However, existing methods still face\nchallenges in handling long-term dependencies and capturing complex temporal\ndynamics. To address these issues, this paper proposes a novel emotion\nrecognition model, Mamba-VA, which leverages the Mamba architecture to\nefficiently model sequential emotional variations in video frames. First, the\nmodel employs a Masked Autoencoder (MAE) to extract deep visual features from\nvideo frames, enhancing the robustness of temporal information. Then, a\nTemporal Convolutional Network (TCN) is utilized for temporal modeling to\ncapture local temporal dependencies. Subsequently, Mamba is applied for\nlong-sequence modeling, enabling the learning of global emotional trends.\nFinally, a fully connected (FC) layer performs regression analysis to predict\ncontinuous valence and arousal values. Experimental results on the\nValence-Arousal (VA) Estimation task of the 8th competition on Affective\nBehavior Analysis in-the-wild (ABAW) demonstrate that the proposed model\nachieves valence and arousal scores of 0.5362 (0.5036) and 0.4310 (0.4119) on\nthe validation (test) set, respectively, outperforming the baseline. The source\ncode is available on GitHub:https://github.com/FreedomPuppy77/Charon.\n", "date": "2025-03-13", "categories": []}
{"id": "2503.10399", "title": "HSEmotion Team at ABAW-8 Competition: Audiovisual Ambivalence/Hesitancy,\n  Emotional Mimicry Intensity and Facial Expression Recognition", "abstract": "  This article presents our results for the eighth Affective Behavior Analysis\nin-the-Wild (ABAW) competition. We combine facial emotional descriptors\nextracted by pre-trained models, namely, our EmotiEffLib library, with acoustic\nfeatures and embeddings of texts recognized from speech. The frame-level\nfeatures are aggregated and fed into simple classifiers, e.g., multi-layered\nperceptron (feed-forward neural network with one hidden layer), to predict\nambivalence/hesitancy and facial expressions. In the latter case, we also use\nthe pre-trained facial expression recognition model to select high-score video\nframes and prevent their processing with a domain-specific video classifier.\nThe video-level prediction of emotional mimicry intensity is implemented by\nsimply aggregating frame-level features and training a multi-layered\nperceptron. Experimental results for three tasks from the ABAW challenge\ndemonstrate that our approach significantly increases validation metrics\ncompared to existing baselines.\n", "date": "2025-03-13", "categories": []}
{"id": "2503.10523", "title": "Interactive Multimodal Fusion with Temporal Modeling", "abstract": "  This paper presents our method for the estimation of valence-arousal (VA) in\nthe 8th Affective Behavior Analysis in-the-Wild (ABAW) competition. Our\napproach integrates visual and audio information through a multimodal\nframework. The visual branch uses a pre-trained ResNet model to extract spatial\nfeatures from facial images. The audio branches employ pre-trained VGG models\nto extract VGGish and LogMel features from speech signals. These features\nundergo temporal modeling using Temporal Convolutional Networks (TCNs). We then\napply cross-modal attention mechanisms, where visual features interact with\naudio features through query-key-value attention structures. Finally, the\nfeatures are concatenated and passed through a regression layer to predict\nvalence and arousal. Our method achieves competitive performance on the\nAff-Wild2 dataset, demonstrating effective multimodal fusion for VA estimation\nin-the-wild.\n", "date": "2025-03-13", "categories": []}
{"id": "2301.08995", "title": "REDAffectiveLM: Leveraging Affect Enriched Embedding and\n  Transformer-based Neural Language Model for Readers' Emotion Detection", "abstract": "  Technological advancements in web platforms allow people to express and share\nemotions towards textual write-ups written and shared by others. This brings\nabout different interesting domains for analysis; emotion expressed by the\nwriter and emotion elicited from the readers. In this paper, we propose a novel\napproach for Readers' Emotion Detection from short-text documents using a deep\nlearning model called REDAffectiveLM. Within state-of-the-art NLP tasks, it is\nwell understood that utilizing context-specific representations from\ntransformer-based pre-trained language models helps achieve improved\nperformance. Within this affective computing task, we explore how incorporating\naffective information can further enhance performance. Towards this, we\nleverage context-specific and affect enriched representations by using a\ntransformer-based pre-trained language model in tandem with affect enriched\nBi-LSTM+Attention. For empirical evaluation, we procure a new dataset REN-20k,\nbesides using RENh-4k and SemEval-2007. We evaluate the performance of our\nREDAffectiveLM rigorously across these datasets, against a vast set of\nstate-of-the-art baselines, where our model consistently outperforms baselines\nand obtains statistically significant results. Our results establish that\nutilizing affect enriched representation along with context-specific\nrepresentation within a neural architecture can considerably enhance readers'\nemotion detection. Since the impact of affect enrichment specifically in\nreaders' emotion detection isn't well explored, we conduct a detailed analysis\nover affect enriched Bi-LSTM+Attention using qualitative and quantitative model\nbehavior evaluation techniques. We observe that compared to conventional\nsemantic embedding, affect enriched embedding increases ability of the network\nto effectively identify and assign weightage to key terms responsible for\nreaders' emotion detection.\n", "date": "2023-01-21", "categories": []}
{"id": "2410.01306", "title": "Emotion-Aware Embedding Fusion in LLMs (Flan-T5, LLAMA 2, DeepSeek-R1,\n  and ChatGPT 4) for Intelligent Response Generation", "abstract": "  Empathetic and coherent responses are critical in auto-mated\nchatbot-facilitated psychotherapy. This study addresses the challenge of\nenhancing the emotional and contextual understanding of large language models\n(LLMs) in psychiatric applications. We introduce Emotion-Aware Embedding\nFusion, a novel framework integrating hierarchical fusion and attention\nmechanisms to prioritize semantic and emotional features in therapy\ntranscripts. Our approach combines multiple emotion lexicons, including NRC\nEmotion Lexicon, VADER, WordNet, and SentiWordNet, with state-of-the-art LLMs\nsuch as Flan-T5, LLAMA 2, DeepSeek-R1, and ChatGPT 4. Therapy session\ntranscripts, comprising over 2,000 samples are segmented into hierarchical\nlevels (word, sentence, and session) using neural networks, while hierarchical\nfusion combines these features with pooling techniques to refine emotional\nrepresentations. Atten-tion mechanisms, including multi-head self-attention and\ncross-attention, further prioritize emotional and contextual features, enabling\ntemporal modeling of emotion-al shifts across sessions. The processed\nembeddings, computed using BERT, GPT-3, and RoBERTa are stored in the Facebook\nAI similarity search vector database, which enables efficient similarity search\nand clustering across dense vector spaces. Upon user queries, relevant segments\nare retrieved and provided as context to LLMs, enhancing their ability to\ngenerate empathetic and con-textually relevant responses. The proposed\nframework is evaluated across multiple practical use cases to demonstrate\nreal-world applicability, including AI-driven therapy chatbots. The system can\nbe integrated into existing mental health platforms to generate personalized\nresponses based on retrieved therapy session data.\n", "date": "2024-10-02", "categories": []}
{"id": "2503.07943", "title": "Enhancing Sentiment Analysis through Multimodal Fusion: A BERT-DINOv2\n  Approach", "abstract": "  Multimodal sentiment analysis enhances conventional sentiment analysis, which\ntraditionally relies solely on text, by incorporating information from\ndifferent modalities such as images, text, and audio. This paper proposes a\nnovel multimodal sentiment analysis architecture that integrates text and image\ndata to provide a more comprehensive understanding of sentiments. For text\nfeature extraction, we utilize BERT, a natural language processing model. For\nimage feature extraction, we employ DINOv2, a vision-transformer-based model.\nThe textual and visual latent features are integrated using proposed fusion\ntechniques, namely the Basic Fusion Model, Self Attention Fusion Model, and\nDual Attention Fusion Model. Experiments on three datasets, Memotion 7k\ndataset, MVSA single dataset, and MVSA multi dataset, demonstrate the viability\nand practicality of the proposed multimodal architecture.\n", "date": "2025-03-11", "categories": []}
{"id": "2503.07969", "title": "7ABAW-Compound Expression Recognition via Curriculum Learning", "abstract": "  With the advent of deep learning, expression recognition has made significant\nadvancements. However, due to the limited availability of annotated compound\nexpression datasets and the subtle variations of compound expressions, Compound\nEmotion Recognition (CE) still holds considerable potential for exploration. To\nadvance this task, the 7th Affective Behavior Analysis in-the-wild (ABAW)\ncompetition introduces the Compound Expression Challenge based on C-EXPR-DB, a\nlimited dataset without labels. In this paper, we present a curriculum\nlearning-based framework that initially trains the model on single-expression\ntasks and subsequently incorporates multi-expression data. This design ensures\nthat our model first masters the fundamental features of basic expressions\nbefore being exposed to the complexities of compound emotions. Specifically,\nour designs can be summarized as follows: 1) Single-Expression Pre-training:\nThe model is first trained on datasets containing single expressions to learn\nthe foundational facial features associated with basic emotions. 2) Dynamic\nCompound Expression Generation: Given the scarcity of annotated compound\nexpression datasets, we employ CutMix and Mixup techniques on the original\nsingle-expression images to create hybrid images exhibiting characteristics of\nmultiple basic emotions. 3) Incremental Multi-Expression Integration: After\nperforming well on single-expression tasks, the model is progressively exposed\nto multi-expression data, allowing the model to adapt to the complexity and\nvariability of compound expressions. The official results indicate that our\nmethod achieves the \\textbf{best} performance in this competition track with an\nF-score of 0.6063. Our code is released at https://github.com/YenanLiu/ABAW7th.\n", "date": "2025-03-11", "categories": []}
{"id": "2310.15112", "title": "Learning Through AI-Clones: Enhancing Self-Perception and Presentation\n  Performance", "abstract": "  This study examines the impact of AI-generated digital clones with\nself-images on enhancing perceptions and skills in online presentations. A\nmixed-design experiment with 44 international students compared self-recording\nvideos (self-recording group) to AI-clone videos (AI-clone group) for online\nEnglish presentation practice. AI-clone videos were generated using voice\ncloning, face swapping, lip-syncing, and body-language simulation, refining the\nrepetition, filler words, and pronunciation of participants' original\npresentations. Through the lens of social comparison theory, the results showed\nthat AI clones functioned as positive \"role models\" for facilitating social\ncomparisons. When comparing the effects on self-perceptions, speech qualities,\nand self-kindness, the self-recording group showed an increase in pronunciation\nsatisfaction. However, the AI-clone group exhibited greater self-kindness,\nbroader observational coverage, and a meaningful transition from a corrective\nto an enhancive approach in self-critique. Moreover, machine-rated scores\nrevealed immediate performance gains only within the AI-clone group.\nConsidering individual differences, aligning interventions with participants'\nregulatory focus significantly enhanced their learning experience. These\nfindings highlight the theoretical, practical, and ethical implications of AI\nclones in supporting emotional and cognitive skill development.\n", "date": "2023-10-23", "categories": []}
{"id": "2501.03190", "title": "Multimodal Machine Learning Can Predict Videoconference Fluidity and\n  Enjoyment", "abstract": "  Videoconferencing is now a frequent mode of communication in both\nprofessional and informal settings, yet it often lacks the fluidity and\nenjoyment of in-person conversation. This study leverages multimodal machine\nlearning to predict moments of negative experience in videoconferencing. We\nsampled thousands of short clips from the RoomReader corpus, extracting audio\nembeddings, facial actions, and body motion features to train models for\nidentifying low conversational fluidity, low enjoyment, and classifying\nconversational events (backchanneling, interruption, or gap). Our best models\nachieved an ROC-AUC of up to 0.87 on hold-out videoconference sessions, with\ndomain-general audio features proving most critical. This work demonstrates\nthat multimodal audio-video signals can effectively predict high-level\nsubjective conversational outcomes. In addition, this is a contribution to\nresearch on videoconferencing user experience by showing that multimodal\nmachine learning can be used to identify rare moments of negative user\nexperience for further study or mitigation.\n", "date": "2025-01-06", "categories": []}
{"id": "2503.05379", "title": "R1-Omni: Explainable Omni-Multimodal Emotion Recognition with\n  Reinforcement Learning", "abstract": "  In this work, we present the first application of Reinforcement Learning with\nVerifiable Reward (RLVR) to an Omni-multimodal large language model in the\ncontext of emotion recognition, a task where both visual and audio modalities\nplay crucial roles. We leverage RLVR to optimize the Omni model, significantly\nenhancing its performance in three key aspects: reasoning capability, emotion\nrecognition accuracy, and generalization ability. The introduction of RLVR not\nonly improves the model's overall performance on in-distribution data but also\ndemonstrates superior robustness when evaluated on out-of-distribution\ndatasets. More importantly, the improved reasoning capability enables clear\nanalysis of the contributions of different modalities, particularly visual and\naudio information, in the emotion recognition process. This provides valuable\ninsights into the optimization of multimodal large language models.\n", "date": "2025-03-07", "categories": []}
{"id": "2503.06805", "title": "Multimodal Emotion Recognition and Sentiment Analysis in Multi-Party\n  Conversation Contexts", "abstract": "  Emotion recognition and sentiment analysis are pivotal tasks in speech and\nlanguage processing, particularly in real-world scenarios involving\nmulti-party, conversational data. This paper presents a multimodal approach to\ntackle these challenges on a well-known dataset. We propose a system that\nintegrates four key modalities/channels using pre-trained models: RoBERTa for\ntext, Wav2Vec2 for speech, a proposed FacialNet for facial expressions, and a\nCNN+Transformer architecture trained from scratch for video analysis. Feature\nembeddings from each modality are concatenated to form a multimodal vector,\nwhich is then used to predict emotion and sentiment labels. The multimodal\nsystem demonstrates superior performance compared to unimodal approaches,\nachieving an accuracy of 66.36% for emotion recognition and 72.15% for\nsentiment analysis.\n", "date": "2025-03-09", "categories": []}
{"id": "2503.04831", "title": "\"Only ChatGPT gets me\": An Empirical Analysis of GPT versus other Large\n  Language Models for Emotion Detection in Text", "abstract": "  This work investigates the capabilities of large language models (LLMs) in\ndetecting and understanding human emotions through text. Drawing upon emotion\nmodels from psychology, we adopt an interdisciplinary perspective that\nintegrates computational and affective sciences insights. The main goal is to\nassess how accurately they can identify emotions expressed in textual\ninteractions and compare different models on this specific task. This research\ncontributes to broader efforts to enhance human-computer interaction, making\nartificial intelligence technologies more responsive and sensitive to users'\nemotional nuances. By employing a methodology that involves comparisons with a\nstate-of-the-art model on the GoEmotions dataset, we aim to gauge LLMs'\neffectiveness as a system for emotional analysis, paving the way for potential\napplications in various fields that require a nuanced understanding of human\nlanguage.\n", "date": "2025-03-05", "categories": []}
{"id": "2503.03756", "title": "Efficient Finetuning for Dimensional Speech Emotion Recognition in the\n  Age of Transformers", "abstract": "  Accurate speech emotion recognition is essential for developing human-facing\nsystems. Recent advancements have included finetuning large, pretrained\ntransformer models like Wav2Vec 2.0. However, the finetuning process requires\nsubstantial computational resources, including high-memory GPUs and significant\nprocessing time. As the demand for accurate emotion recognition continues to\ngrow, efficient finetuning approaches are needed to reduce the computational\nburden. Our study focuses on dimensional emotion recognition, predicting\nattributes such as activation (calm to excited) and valence (negative to\npositive). We present various finetuning techniques, including full finetuning,\npartial finetuning of transformer layers, finetuning with mixed precision,\npartial finetuning with caching, and low-rank adaptation (LoRA) on the Wav2Vec\n2.0 base model. We find that partial finetuning with mixed precision achieves\nperformance comparable to full finetuning while increasing training speed by\n67%. Caching intermediate representations further boosts efficiency, yielding\nan 88% speedup and a 71% reduction in learnable parameters. We recommend\nfinetuning the final three transformer layers in mixed precision to balance\nperformance and training efficiency, and adding intermediate representation\ncaching for optimal speed with minimal performance trade-offs. These findings\nlower the barriers to finetuning speech emotion recognition systems, making\naccurate emotion recognition more accessible to a broader range of researchers\nand practitioners.\n", "date": "2025-02-17", "categories": []}
{"id": "2403.07260", "title": "LaERC-S: Improving LLM-based Emotion Recognition in Conversation with\n  Speaker Characteristics", "abstract": "  Emotion recognition in conversation (ERC), the task of discerning human\nemotions for each utterance within a conversation, has garnered significant\nattention in human-computer interaction systems. Previous ERC studies focus on\nspeaker-specific information that predominantly stems from relationships among\nutterances, which lacks sufficient information around conversations. Recent\nresearch in ERC has sought to exploit pre-trained large language models (LLMs)\nwith speaker modelling to comprehend emotional states. Although these methods\nhave achieved encouraging results, the extracted speaker-specific information\nstruggles to indicate emotional dynamics. In this paper, motivated by the fact\nthat speaker characteristics play a crucial role and LLMs have rich world\nknowledge, we present LaERC-S, a novel framework that stimulates LLMs to\nexplore speaker characteristics involving the mental state and behavior of\ninterlocutors, for accurate emotion predictions. To endow LLMs with this\nknowledge information, we adopt the two-stage learning to make the models\nreason speaker characteristics and track the emotion of the speaker in complex\nconversation scenarios. Extensive experiments on three benchmark datasets\ndemonstrate the superiority of LaERC-S, reaching the new state-of-the-art.\n", "date": "2024-03-12", "categories": []}
{"id": "2502.08813", "title": "Measuring Anxiety Levels with Head Motion Patterns in Severe Depression\n  Population", "abstract": "  Depression and anxiety are prevalent mental health disorders that frequently\ncooccur, with anxiety significantly influencing both the manifestation and\ntreatment of depression. An accurate assessment of anxiety levels in\nindividuals with depression is crucial to develop effective and personalized\ntreatment plans. This study proposes a new noninvasive method for quantifying\nanxiety severity by analyzing head movements -- specifically speed,\nacceleration, and angular displacement -- during video-recorded interviews with\npatients suffering from severe depression. Using data from a new CALYPSO\nDepression Dataset, we extracted head motion characteristics and applied\nregression analysis to predict clinically evaluated anxiety levels. Our results\ndemonstrate a high level of precision, achieving a mean absolute error (MAE) of\n0.35 in predicting the severity of psychological anxiety based on head movement\npatterns. This indicates that our approach can enhance the understanding of\nanxiety's role in depression and assist psychiatrists in refining treatment\nstrategies for individuals.\n", "date": "2025-02-12", "categories": []}
{"id": "2503.00049", "title": "Omni-SILA: Towards Omni-scene Driven Visual Sentiment Identifying,\n  Locating and Attributing in Videos", "abstract": "  Prior studies on Visual Sentiment Understanding (VSU) primarily rely on the\nexplicit scene information (e.g., facial expression) to judge visual\nsentiments, which largely ignore implicit scene information (e.g., human\naction, objection relation and visual background), while such information is\ncritical for precisely discovering visual sentiments. Motivated by this, this\npaper proposes a new Omni-scene driven visual Sentiment Identifying, Locating\nand Attributing in videos (Omni-SILA) task, aiming to interactively and\nprecisely identify, locate and attribute visual sentiments through both\nexplicit and implicit scene information. Furthermore, this paper believes that\nthis Omni-SILA task faces two key challenges: modeling scene and highlighting\nimplicit scene beyond explicit. To this end, this paper proposes an\nImplicit-enhanced Causal MoE (ICM) approach for addressing the Omni-SILA task.\nSpecifically, a Scene-Balanced MoE (SBM) and an Implicit-Enhanced Causal (IEC)\nblocks are tailored to model scene information and highlight the implicit scene\ninformation beyond explicit, respectively. Extensive experimental results on\nour constructed explicit and implicit Omni-SILA datasets demonstrate the great\nadvantage of the proposed ICM approach over advanced Video-LLMs.\n", "date": "2025-02-26", "categories": []}
{"id": "2503.00202", "title": "MIDAS: Mixing Ambiguous Data with Soft Labels for Dynamic Facial\n  Expression Recognition", "abstract": "  Dynamic facial expression recognition (DFER) is an important task in the\nfield of computer vision. To apply automatic DFER in practice, it is necessary\nto accurately recognize ambiguous facial expressions, which often appear in\ndata in the wild. In this paper, we propose MIDAS, a data augmentation method\nfor DFER, which augments ambiguous facial expression data with soft labels\nconsisting of probabilities for multiple emotion classes. In MIDAS, the\ntraining data are augmented by convexly combining pairs of video frames and\ntheir corresponding emotion class labels, which can also be regarded as an\nextension of mixup to soft-labeled video data. This simple extension is\nremarkably effective in DFER with ambiguous facial expression data. To evaluate\nMIDAS, we conducted experiments on the DFEW dataset. The results demonstrate\nthat the model trained on the data augmented by MIDAS outperforms the existing\nstate-of-the-art method trained on the original dataset.\n", "date": "2025-02-28", "categories": []}
{"id": "2502.20613", "title": "Continuous Adversarial Text Representation Learning for Affective\n  Recognition", "abstract": "  While pre-trained language models excel at semantic understanding, they often\nstruggle to capture nuanced affective information critical for affective\nrecognition tasks. To address these limitations, we propose a novel framework\nfor enhancing emotion-aware embeddings in transformer-based models. Our\napproach introduces a continuous valence-arousal labeling system to guide\ncontrastive learning, which captures subtle and multi-dimensional emotional\nnuances more effectively. Furthermore, we employ a dynamic token perturbation\nmechanism, using gradient-based saliency to focus on sentiment-relevant tokens,\nimproving model sensitivity to emotional cues. The experimental results\ndemonstrate that the proposed framework outperforms existing methods, achieving\nup to 15.5% improvement in the emotion classification benchmark, highlighting\nthe importance of employing continuous labels. This improvement demonstrates\nthat the proposed framework is effective in affective representation learning\nand enables precise and contextually relevant emotional understanding.\n", "date": "2025-02-28", "categories": []}
{"id": "2502.19856", "title": "Team A at SemEval-2025 Task 11: Breaking Language Barriers in Emotion\n  Detection with Multilingual Models", "abstract": "  This paper describes the system submitted by Team A to SemEval 2025 Task 11,\n``Bridging the Gap in Text-Based Emotion Detection.'' The task involved\nidentifying the perceived emotion of a speaker from text snippets, with each\ninstance annotated with one of six emotions: joy, sadness, fear, anger,\nsurprise, or disgust. A dataset provided by the task organizers served as the\nfoundation for training and evaluating our models. Among the various approaches\nexplored, the best performance was achieved using multilingual embeddings\ncombined with a fully connected layer. This paper details the system\narchitecture, discusses experimental results, and highlights the advantages of\nleveraging multilingual representations for robust emotion detection in text.\n", "date": "2025-02-27", "categories": []}
{"id": "2502.17172", "title": "Teleology-Driven Affective Computing: A Causal Framework for Sustained\n  Well-Being", "abstract": "  Affective computing has made significant strides in emotion recognition and\ngeneration, yet current approaches mainly focus on short-term pattern\nrecognition and lack a comprehensive framework to guide affective agents toward\nlong-term human well-being. To address this, we propose a teleology-driven\naffective computing framework that unifies major emotion theories (basic\nemotion, appraisal, and constructivist approaches) under the premise that\naffect is an adaptive, goal-directed process that facilitates survival and\ndevelopment. Our framework emphasizes aligning agent responses with both\npersonal/individual and group/collective well-being over extended timescales.\nWe advocate for creating a \"dataverse\" of personal affective events, capturing\nthe interplay between beliefs, goals, actions, and outcomes through real-world\nexperience sampling and immersive virtual reality. By leveraging causal\nmodeling, this \"dataverse\" enables AI systems to infer individuals' unique\naffective concerns and provide tailored interventions for sustained well-being.\nAdditionally, we introduce a meta-reinforcement learning paradigm to train\nagents in simulated environments, allowing them to adapt to evolving affective\nconcerns and balance hierarchical goals - from immediate emotional needs to\nlong-term self-actualization. This framework shifts the focus from statistical\ncorrelations to causal reasoning, enhancing agents' ability to predict and\nrespond proactively to emotional challenges, and offers a foundation for\ndeveloping personalized, ethically aligned affective systems that promote\nmeaningful human-AI interactions and societal well-being.\n", "date": "2025-02-24", "categories": []}
{"id": "2502.17445", "title": "Interpretable Dual-Filter Fuzzy Neural Networks for Affective\n  Brain-Computer Interfaces", "abstract": "  Fuzzy logic provides a robust framework for enhancing explainability,\nparticularly in domains requiring the interpretation of complex and ambiguous\nsignals, such as brain-computer interface (BCI) systems. Despite significant\nadvances in deep learning, interpreting human emotions remains a formidable\nchallenge. In this work, we present iFuzzyAffectDuo, a novel computational\nmodel that integrates a dual-filter fuzzy neural network architecture for\nimproved detection and interpretation of emotional states from neuroimaging\ndata. The model introduces a new membership function (MF) based on the Laplace\ndistribution, achieving superior accuracy and interpretability compared to\ntraditional approaches. By refining the extraction of neural signals associated\nwith specific emotions, iFuzzyAffectDuo offers a human-understandable framework\nthat unravels the underlying decision-making processes. We validate our\napproach across three neuroimaging datasets using functional Near-Infrared\nSpectroscopy (fNIRS) and Electroencephalography (EEG), demonstrating its\npotential to advance affective computing. These findings open new pathways for\nunderstanding the neural basis of emotions and their application in enhancing\nhuman-computer interaction.\n", "date": "2025-01-29", "categories": []}
{"id": "2502.16106", "title": "AnxietyFaceTrack: A Smartphone-Based Non-Intrusive Approach for\n  Detecting Social Anxiety Using Facial Features", "abstract": "  Social Anxiety Disorder (SAD) is a widespread mental health condition, yet\nits lack of objective markers hinders timely detection and intervention. While\nprevious research has focused on behavioral and non-verbal markers of SAD in\nstructured activities (e.g., speeches or interviews), these settings fail to\nreplicate real-world, unstructured social interactions fully. Identifying\nnon-verbal markers in naturalistic, unstaged environments is essential for\ndeveloping ubiquitous and non-intrusive monitoring solutions. To address this\ngap, we present AnxietyFaceTrack, a study leveraging facial video analysis to\ndetect anxiety in unstaged social settings. A cohort of 91 participants engaged\nin a social setting with unfamiliar individuals and their facial videos were\nrecorded using a low-cost smartphone camera. We examined facial features,\nincluding eye movements, head position, facial landmarks, and facial action\nunits, and used self-reported survey data to establish ground truth for\nmulticlass (anxious, neutral, non-anxious) and binary (e.g., anxious vs.\nneutral) classifications. Our results demonstrate that a Random Forest\nclassifier trained on the top 20% of features achieved the highest accuracy of\n91.0% for multiclass classification and an average accuracy of 92.33% across\nbinary classifications. Notably, head position and facial landmarks yielded the\nbest performance for individual facial regions, achieving 85.0% and 88.0%\naccuracy, respectively, in multiclass classification, and 89.66% and 91.0%\naccuracy, respectively, across binary classifications. This study introduces a\nnon-intrusive, cost-effective solution that can be seamlessly integrated into\neveryday smartphones for continuous anxiety monitoring, offering a promising\npathway for early detection and intervention.\n", "date": "2025-02-22", "categories": []}
{"id": "2502.15322", "title": "SentiFormer: Metadata Enhanced Transformer for Image Sentiment Analysis", "abstract": "  As more and more internet users post images online to express their daily\nemotions, image sentiment analysis has attracted increasing attention.\nRecently, researchers generally tend to design different neural networks to\nextract visual features from images for sentiment analysis. Despite the\nsignificant progress, metadata, the data (e.g., text descriptions and keyword\ntags) for describing the image, has not been sufficiently explored in this\ntask. In this paper, we propose a novel Metadata Enhanced Transformer for\nsentiment analysis (SentiFormer) to fuse multiple metadata and the\ncorresponding image into a unified framework. Specifically, we first obtain\nmultiple metadata of the image and unify the representations of diverse data.\nTo adaptively learn the appropriate weights for each metadata, we then design\nan adaptive relevance learning module to highlight more effective information\nwhile suppressing weaker ones. Moreover, we further develop a cross-modal\nfusion module to fuse the adaptively learned representations and make the final\nprediction. Extensive experiments on three publicly available datasets\ndemonstrate the superiority and rationality of our proposed method.\n", "date": "2025-02-21", "categories": []}
{"id": "2404.18398", "title": "UMETTS: A Unified Framework for Emotional Text-to-Speech Synthesis with\n  Multimodal Prompts", "abstract": "  Emotional Text-to-Speech (E-TTS) synthesis has garnered significant attention\nin recent years due to its potential to revolutionize human-computer\ninteraction. However, current E-TTS approaches often struggle to capture the\nintricacies of human emotions, primarily relying on oversimplified emotional\nlabels or single-modality input. In this paper, we introduce the Unified\nMultimodal Prompt-Induced Emotional Text-to-Speech System (UMETTS), a novel\nframework that leverages emotional cues from multiple modalities to generate\nhighly expressive and emotionally resonant speech. The core of UMETTS consists\nof two key components: the Emotion Prompt Alignment Module (EP-Align) and the\nEmotion Embedding-Induced TTS Module (EMI-TTS). (1) EP-Align employs\ncontrastive learning to align emotional features across text, audio, and visual\nmodalities, ensuring a coherent fusion of multimodal information. (2)\nSubsequently, EMI-TTS integrates the aligned emotional embeddings with\nstate-of-the-art TTS models to synthesize speech that accurately reflects the\nintended emotions. Extensive evaluations show that UMETTS achieves significant\nimprovements in emotion accuracy and speech naturalness, outperforming\ntraditional E-TTS methods on both objective and subjective metrics.\n", "date": "2024-04-29", "categories": []}
{"id": "2502.13954", "title": "Latent Distribution Decoupling: A Probabilistic Framework for\n  Uncertainty-Aware Multimodal Emotion Recognition", "abstract": "  Multimodal multi-label emotion recognition (MMER) aims to identify the\nconcurrent presence of multiple emotions in multimodal data. Existing studies\nprimarily focus on improving fusion strategies and modeling modality-to-label\ndependencies. However, they often overlook the impact of \\textbf{aleatoric\nuncertainty}, which is the inherent noise in the multimodal data and hinders\nthe effectiveness of modality fusion by introducing ambiguity into feature\nrepresentations. To address this issue and effectively model aleatoric\nuncertainty, this paper proposes Latent emotional Distribution Decomposition\nwith Uncertainty perception (LDDU) framework from a novel perspective of latent\nemotional space probabilistic modeling. Specifically, we introduce a\ncontrastive disentangled distribution mechanism within the emotion space to\nmodel the multimodal data, allowing for the extraction of semantic features and\nuncertainty. Furthermore, we design an uncertainty-aware fusion multimodal\nmethod that accounts for the dispersed distribution of uncertainty and\nintegrates distribution information. Experimental results show that LDDU\nachieves state-of-the-art performance on the CMU-MOSEI and M$^3$ED datasets,\nhighlighting the importance of uncertainty modeling in MMER. Code is available\nat https://github.com/201983290498/lddu\\_mmer.git.\n", "date": "2025-02-19", "categories": []}
{"id": "2408.07982", "title": "Toward a Dialogue System Using a Large Language Model to Recognize User\n  Emotions with a Camera", "abstract": "  The performance of ChatGPT\\copyright{} and other LLMs has improved\ntremendously, and in online environments, they are increasingly likely to be\nused in a wide variety of situations, such as ChatBot on web pages, call center\noperations using voice interaction, and dialogue functions using agents. In the\noffline environment, multimodal dialogue functions are also being realized,\nsuch as guidance by Artificial Intelligence agents (AI agents) using tablet\nterminals and dialogue systems in the form of LLMs mounted on robots. In this\nmultimodal dialogue, mutual emotion recognition between the AI and the user\nwill become important. So far, there have been methods for expressing emotions\non the part of the AI agent or for recognizing them using textual or voice\ninformation of the user's utterances, but methods for AI agents to recognize\nemotions from the user's facial expressions have not been studied. In this\nstudy, we examined whether or not LLM-based AI agents can interact with users\naccording to their emotional states by capturing the user in dialogue with a\ncamera, recognizing emotions from facial expressions, and adding such emotion\ninformation to prompts. The results confirmed that AI agents can have\nconversations according to the emotional state for emotional states with\nrelatively high scores, such as Happy and Angry.\n", "date": "2024-08-15", "categories": []}
{"id": "2502.12478", "title": "MSE-Adapter: A Lightweight Plugin Endowing LLMs with the Capability to\n  Perform Multimodal Sentiment Analysis and Emotion Recognition", "abstract": "  Current Multimodal Sentiment Analysis (MSA) and Emotion Recognition in\nConversations (ERC) methods based on pre-trained language models exhibit two\nprimary limitations:\n  1) Once trained for MSA and ERC tasks, these pre-trained language models lose\ntheir original generalized capabilities. 2) They demand considerable\ncomputational resources. As the size of pre-trained language models continues\nto grow, training larger multimodal sentiment analysis models using previous\napproaches could result in unnecessary computational cost. In response to this\nchallenge, we propose \\textbf{M}ultimodal \\textbf{S}entiment Analysis and\n\\textbf{E}motion Recognition \\textbf{Adapter} (MSE-Adapter), a lightweight and\nadaptable plugin. This plugin enables a large language model (LLM) to carry out\nMSA or ERC tasks with minimal computational overhead (only introduces\napproximately 2.6M to 2.8M trainable parameters upon the 6/7B models), while\npreserving the intrinsic capabilities of the LLM. In the MSE-Adapter, the\nText-Guide-Mixer (TGM) module is introduced to establish explicit connections\nbetween non-textual and textual modalities through the Hadamard product. This\nallows non-textual modalities to better align with textual modalities at the\nfeature level, promoting the generation of higher-quality pseudo tokens.\nExtensive experiments were conducted on four public English and Chinese\ndatasets using consumer-grade GPUs and open-source LLMs (Qwen-1.8B,\nChatGLM3-6B-base, and LLaMA2-7B) as the backbone. The results demonstrate the\neffectiveness of the proposed plugin. The code will be released on GitHub after\na blind review.\n", "date": "2025-02-18", "categories": []}
{"id": "2406.15490", "title": "Causal Discovery Inspired Unsupervised Domain Adaptation for\n  Emotion-Cause Pair Extraction", "abstract": "  This paper tackles the task of emotion-cause pair extraction in the\nunsupervised domain adaptation setting. The problem is challenging as the\ndistributions of the events causing emotions in target domains are dramatically\ndifferent than those in source domains, despite the distributions of emotional\nexpressions between domains are overlapped. Inspired by causal discovery, we\npropose a novel deep latent model in the variational autoencoder (VAE)\nframework, which not only captures the underlying latent structures of data but\nalso utilizes the easily transferable knowledge of emotions as the bridge to\nlink the distributions of events in different domains. To facilitate knowledge\ntransfer across domains, we also propose a novel variational posterior\nregularization technique to disentangle the latent representations of emotions\nfrom those of events in order to mitigate the damage caused by the spurious\ncorrelations related to the events in source domains. Through extensive\nexperiments, we demonstrate that our model outperforms the strongest baseline\nby approximately 11.05\\% on a Chinese benchmark and 2.45\\% on a English\nbenchmark in terms of weighted-average F1 score. We have released our source\ncode and the generated dataset publicly at:\nhttps://github.com/tk1363704/CAREL-VAE.\n", "date": "2024-06-18", "categories": []}
{"id": "2409.18339", "title": "AER-LLM: Ambiguity-aware Emotion Recognition Leveraging Large Language\n  Models", "abstract": "  Recent advancements in Large Language Models (LLMs) have demonstrated great\nsuccess in many Natural Language Processing (NLP) tasks. In addition to their\ncognitive intelligence, exploring their capabilities in emotional intelligence\nis also crucial, as it enables more natural and empathetic conversational AI.\nRecent studies have shown LLMs' capability in recognizing emotions, but they\noften focus on single emotion labels and overlook the complex and ambiguous\nnature of human emotions. This study is the first to address this gap by\nexploring the potential of LLMs in recognizing ambiguous emotions, leveraging\ntheir strong generalization capabilities and in-context learning. We design\nzero-shot and few-shot prompting and incorporate past dialogue as context\ninformation for ambiguous emotion recognition. Experiments conducted using\nthree datasets indicate significant potential for LLMs in recognizing ambiguous\nemotions, and highlight the substantial benefits of including context\ninformation. Furthermore, our findings indicate that LLMs demonstrate a high\ndegree of effectiveness in recognizing less ambiguous emotions and exhibit\npotential for identifying more ambiguous emotions, paralleling human perceptual\ncapabilities.\n", "date": "2024-09-26", "categories": []}
{"id": "2502.10813", "title": "Transformer-Driven Modeling of Variable Frequency Features for\n  Classifying Student Engagement in Online Learning", "abstract": "  The COVID-19 pandemic and the internet's availability have recently boosted\nonline learning. However, monitoring engagement in online learning is a\ndifficult task for teachers. In this context, timely automatic student\nengagement classification can help teachers in making adaptive adjustments to\nmeet students' needs. This paper proposes EngageFormer, a transformer based\narchitecture with sequence pooling using video modality for engagement\nclassification. The proposed architecture computes three views from the input\nvideo and processes them in parallel using transformer encoders; the global\nencoder then processes the representation from each encoder, and finally, multi\nlayer perceptron (MLP) predicts the engagement level. A learning centered\naffective state dataset is curated from existing open source databases. The\nproposed method achieved an accuracy of 63.9%, 56.73%, 99.16%, 65.67%, and\n74.89% on Dataset for Affective States in E-Environments (DAiSEE), Bahcesehir\nUniversity Multimodal Affective Database-1 (BAUM-1), Yawning Detection Dataset\n(YawDD), University of Texas at Arlington Real-Life Drowsiness Dataset\n(UTA-RLDD), and curated learning-centered affective state dataset respectively.\nThe achieved results on the BAUM-1, DAiSEE, and YawDD datasets demonstrate\nstate-of-the-art performance, indicating the superiority of the proposed model\nin accurately classifying affective states on these datasets. Additionally, the\nresults obtained on the UTA-RLDD dataset, which involves two-class\nclassification, serve as a baseline for future research. These results provide\na foundation for further investigations and serve as a point of reference for\nfuture works to compare and improve upon.\n", "date": "2025-02-15", "categories": []}
{"id": "2410.07062", "title": "TinyEmo: Scaling down Emotional Reasoning via Metric Projection", "abstract": "  This paper introduces TinyEmo, a family of small multi-modal language models\nfor emotional reasoning and classification. Our approach features: (1) a\nsynthetic emotional instruct dataset for both pre-training and fine-tuning\nstages, (2) a Metric Projector that delegates classification from the language\nmodel allowing for more efficient training and inference, (3) a multi-modal\nlarge language model (MM-LLM) for emotional reasoning, and (4) a semi-automated\nframework for bias detection. TinyEmo is able to perform emotion classification\nand emotional reasoning, all while using substantially fewer parameters than\ncomparable models. This efficiency allows us to freely incorporate more diverse\nemotional datasets, enabling strong performance on classification tasks, with\nour smallest model (700M parameters) outperforming larger state-of-the-art\nmodels based on general-purpose MM-LLMs with over 7B parameters. Additionally,\nthe Metric Projector allows for interpretability and indirect bias detection in\nlarge models without additional training, offering an approach to understand\nand improve AI systems. We release code, models, and dataset at\nhttps://github.com/ggcr/TinyEmo\n", "date": "2024-10-09", "categories": []}
{"id": "2502.09675", "title": "Multi-level Conflict-Aware Network for Multi-modal Sentiment Analysis", "abstract": "  Multimodal Sentiment Analysis (MSA) aims to recognize human emotions by\nexploiting textual, acoustic, and visual modalities, and thus how to make full\nuse of the interactions between different modalities is a central challenge of\nMSA. Interaction contains alignment and conflict aspects. Current works mainly\nemphasize alignment and the inherent differences between unimodal modalities,\nneglecting the fact that there are also potential conflicts between bimodal\ncombinations. Additionally, multi-task learning-based conflict modeling methods\noften rely on the unstable generated labels. To address these challenges, we\npropose a novel multi-level conflict-aware network (MCAN) for multimodal\nsentiment analysis, which progressively segregates alignment and conflict\nconstituents from unimodal and bimodal representations, and further exploits\nthe conflict constituents with the conflict modeling branch. In the conflict\nmodeling branch, we conduct discrepancy constraints at both the representation\nand predicted output levels, avoiding dependence on the generated labels.\nExperimental results on the CMU-MOSI and CMU-MOSEI datasets demonstrate the\neffectiveness of the proposed MCAN.\n", "date": "2025-02-13", "categories": []}
{"id": "2502.10145", "title": "Interpretable Concept-based Deep Learning Framework for Multimodal Human\n  Behavior Modeling", "abstract": "  In the contemporary era of intelligent connectivity, Affective Computing\n(AC), which enables systems to recognize, interpret, and respond to human\nbehavior states, has become an integrated part of many AI systems. As one of\nthe most critical components of responsible AI and trustworthiness in all\nhuman-centered systems, explainability has been a major concern in AC.\nParticularly, the recently released EU General Data Protection Regulation\nrequires any high-risk AI systems to be sufficiently interpretable, including\nbiometric-based systems and emotion recognition systems widely used in the\naffective computing field. Existing explainable methods often compromise\nbetween interpretability and performance. Most of them focus only on\nhighlighting key network parameters without offering meaningful,\ndomain-specific explanations to the stakeholders. Additionally, they also face\nchallenges in effectively co-learning and explaining insights from multimodal\ndata sources. To address these limitations, we propose a novel and\ngeneralizable framework, namely the Attention-Guided Concept Model (AGCM),\nwhich provides learnable conceptual explanations by identifying what concepts\nthat lead to the predictions and where they are observed. AGCM is extendable to\nany spatial and temporal signals through multimodal concept alignment and\nco-learning, empowering stakeholders with deeper insights into the model's\ndecision-making process. We validate the efficiency of AGCM on well-established\nFacial Expression Recognition benchmark datasets while also demonstrating its\ngeneralizability on more complex real-world human behavior understanding\napplications.\n", "date": "2025-02-14", "categories": []}
{"id": "2502.09294", "title": "Indeterminacy in Affective Computing: Considering Meaning and Context in\n  Data Collection Practices", "abstract": "  Automatic Affect Prediction (AAP) uses computational analysis of input data\nsuch as text, speech, images, and physiological signals to predict various\naffective phenomena (e.g., emotions or moods). These models are typically\nconstructed using supervised machine-learning algorithms, which rely heavily on\nlabeled training datasets. In this position paper, we posit that all AAP\ntraining data are derived from human Affective Interpretation Processes,\nresulting in a form of Affective Meaning. Research on human affect indicates a\nform of complexity that is fundamental to such meaning: it can possess what we\nrefer to here broadly as Qualities of Indeterminacy (QIs) - encompassing\nSubjectivity (meaning depends on who is interpreting), Uncertainty (lack of\nconfidence regarding meanings' correctness), Ambiguity (meaning contains\nmutually exclusive concepts) and Vagueness (meaning is situated at different\nlevels in a nested hierarchy). Failing to appropriately consider QIs leads to\nresults incapable of meaningful and reliable predictions. Based on this\npremise, we argue that a crucial step in adequately addressing indeterminacy in\nAAP is the development of data collection practices for modeling corpora that\ninvolve the systematic consideration of 1) a relevant set of QIs and 2) context\nfor the associated interpretation processes. To this end, we are 1) outlining a\nconceptual model of AIPs and the QIs associated with the meaning these produce\nand a conceptual structure of relevant context, supporting understanding of its\nrole. Finally, we use our framework for 2) discussing examples of\ncontext-sensitivity-related challenges for addressing QIs in data collection\nsetups. We believe our efforts can stimulate a structured discussion of both\nthe role of aspects of indeterminacy and context in research on AAP, informing\nthe development of better practices for data collection and analysis.\n", "date": "2025-02-13", "categories": []}
{"id": "2502.08573", "title": "A Novel Approach to for Multimodal Emotion Recognition : Multimodal\n  semantic information fusion", "abstract": "  With the advancement of artificial intelligence and computer vision\ntechnologies, multimodal emotion recognition has become a prominent research\ntopic. However, existing methods face challenges such as heterogeneous data\nfusion and the effective utilization of modality correlations. This paper\nproposes a novel multimodal emotion recognition approach, DeepMSI-MER, based on\nthe integration of contrastive learning and visual sequence compression. The\nproposed method enhances cross-modal feature fusion through contrastive\nlearning and reduces redundancy in the visual modality by leveraging visual\nsequence compression. Experimental results on two public datasets, IEMOCAP and\nMELD, demonstrate that DeepMSI-MER significantly improves the accuracy and\nrobustness of emotion recognition, validating the effectiveness of multimodal\nfeature fusion and the proposed approach.\n", "date": "2025-02-12", "categories": []}
{"id": "2502.06803", "title": "Emotion Recognition and Generation: A Comprehensive Review of Face,\n  Speech, and Text Modalities", "abstract": "  Emotion recognition and generation have emerged as crucial topics in\nArtificial Intelligence research, playing a significant role in enhancing\nhuman-computer interaction within healthcare, customer service, and other\nfields. Although several reviews have been conducted on emotion recognition and\ngeneration as separate entities, many of these works are either fragmented or\nlimited to specific methodologies, lacking a comprehensive overview of recent\ndevelopments and trends across different modalities. In this survey, we provide\na holistic review aimed at researchers beginning their exploration in emotion\nrecognition and generation. We introduce the fundamental principles underlying\nemotion recognition and generation across facial, vocal, and textual\nmodalities. This work categorises recent state-of-the-art research into\ndistinct technical approaches and explains the theoretical foundations and\nmotivations behind these methodologies, offering a clearer understanding of\ntheir application. Moreover, we discuss evaluation metrics, comparative\nanalyses, and current limitations, shedding light on the challenges faced by\nresearchers in the field. Finally, we propose future research directions to\naddress these challenges and encourage further exploration into developing\nrobust, effective, and ethically responsible emotion recognition and generation\nsystems.\n", "date": "2025-02-02", "categories": []}
{"id": "2502.06873", "title": "Multimodal Cognitive Reframing Therapy via Multi-hop Psychotherapeutic\n  Reasoning", "abstract": "  Previous research has revealed the potential of large language models (LLMs)\nto support cognitive reframing therapy; however, their focus was primarily on\ntext-based methods, often overlooking the importance of non-verbal evidence\ncrucial in real-life therapy. To alleviate this gap, we extend the textual\ncognitive reframing to multimodality, incorporating visual clues. Specifically,\nwe present a new dataset called Multi Modal-Cognitive Support Conversation\n(M2CoSC), which pairs each GPT-4-generated dialogue with an image that reflects\nthe virtual client's facial expressions. To better mirror real psychotherapy,\nwhere facial expressions lead to interpreting implicit emotional evidence, we\npropose a multi-hop psychotherapeutic reasoning approach that explicitly\nidentifies and incorporates subtle evidence. Our comprehensive experiments with\nboth LLMs and vision-language models (VLMs) demonstrate that the VLMs'\nperformance as psychotherapists is significantly improved with the M2CoSC\ndataset. Furthermore, the multi-hop psychotherapeutic reasoning method enables\nVLMs to provide more thoughtful and empathetic suggestions, outperforming\nstandard prompting methods.\n", "date": "2025-02-08", "categories": []}
{"id": "2502.06875", "title": "Beyond Vision: How Large Language Models Interpret Facial Expressions\n  from Valence-Arousal Values", "abstract": "  Large Language Models primarily operate through text-based inputs and\noutputs, yet human emotion is communicated through both verbal and non-verbal\ncues, including facial expressions. While Vision-Language Models analyze facial\nexpressions from images, they are resource-intensive and may depend more on\nlinguistic priors than visual understanding. To address this, this study\ninvestigates whether LLMs can infer affective meaning from dimensions of facial\nexpressions-Valence and Arousal values, structured numerical representations,\nrather than using raw visual input. VA values were extracted using Facechannel\nfrom images of facial expressions and provided to LLMs in two tasks: (1)\ncategorizing facial expressions into basic (on the IIMI dataset) and complex\nemotions (on the Emotic dataset) and (2) generating semantic descriptions of\nfacial expressions (on the Emotic dataset). Results from the categorization\ntask indicate that LLMs struggle to classify VA values into discrete emotion\ncategories, particularly for emotions beyond basic polarities (e.g., happiness,\nsadness). However, in the semantic description task, LLMs produced textual\ndescriptions that align closely with human-generated interpretations,\ndemonstrating a stronger capacity for free text affective inference of facial\nexpressions.\n", "date": "2025-02-08", "categories": []}
{"id": "2403.14318", "title": "A Lightweight Attention-based Deep Network via Multi-Scale Feature\n  Fusion for Multi-View Facial Expression Recognition", "abstract": "  Convolutional neural networks (CNNs) and their variations have shown\neffectiveness in facial expression recognition (FER). However, they face\nchallenges when dealing with high computational complexity and multi-view head\nposes in real-world scenarios. We introduce a lightweight attentional network\nincorporating multi-scale feature fusion (LANMSFF) to tackle these issues. For\nthe first challenge, we carefully design a lightweight network. We address the\nsecond challenge by presenting two novel components, namely mass attention\n(MassAtt) and point wise feature selection (PWFS) blocks. The MassAtt block\nsimultaneously generates channel and spatial attention maps to recalibrate\nfeature maps by emphasizing important features while suppressing irrelevant\nones. In addition, the PWFS block employs a feature selection mechanism that\ndiscards less meaningful features prior to the fusion process. This mechanism\ndistinguishes it from previous methods that directly fuse multi-scale features.\nOur proposed approach achieved results comparable to state-of-the-art methods\nin terms of parameter count and robustness to pose variation, with accuracy\nrates of 90.77% on KDEF, 70.44% on FER-2013, and 86.96% on FERPlus datasets.\nThe code for LANMSFF is available at https://github.com/AE-1129/LANMSFF.\n", "date": "2024-03-21", "categories": []}
{"id": "2412.07121", "title": "Bridging the Gap for Test-Time Multimodal Sentiment Analysis", "abstract": "  Multimodal sentiment analysis (MSA) is an emerging research topic that aims\nto understand and recognize human sentiment or emotions through multiple\nmodalities. However, in real-world dynamic scenarios, the distribution of\ntarget data is always changing and different from the source data used to train\nthe model, which leads to performance degradation. Common adaptation methods\nusually need source data, which could pose privacy issues or storage overheads.\nTherefore, test-time adaptation (TTA) methods are introduced to improve the\nperformance of the model at inference time. Existing TTA methods are always\nbased on probabilistic models and unimodal learning, and thus can not be\napplied to MSA which is often considered as a multimodal regression task. In\nthis paper, we propose two strategies: Contrastive Adaptation and Stable\nPseudo-label generation (CASP) for test-time adaptation for multimodal\nsentiment analysis. The two strategies deal with the distribution shifts for\nMSA by enforcing consistency and minimizing empirical risk, respectively.\nExtensive experiments show that CASP brings significant and consistent\nimprovements to the performance of the model across various distribution shift\nsettings and with different backbones, demonstrating its effectiveness and\nversatility. Our codes are available at https://github.com/zrguo/CASP.\n", "date": "2024-12-10", "categories": []}
{"id": "2405.13206", "title": "Identity-free Artificial Emotional Intelligence via Micro-Gesture\n  Understanding", "abstract": "  In this work, we focus on a special group of human body language -- the\nmicro-gesture (MG), which differs from the range of ordinary illustrative\ngestures in that they are not intentional behaviors performed to convey\ninformation to others, but rather unintentional behaviors driven by inner\nfeelings. This characteristic introduces two novel challenges regarding\nmicro-gestures that are worth rethinking. The first is whether strategies\ndesigned for other action recognition are entirely applicable to\nmicro-gestures. The second is whether micro-gestures, as supplementary data,\ncan provide additional insights for emotional understanding. In recognizing\nmicro-gestures, we explored various augmentation strategies that take into\naccount the subtle spatial and brief temporal characteristics of\nmicro-gestures, often accompanied by repetitiveness, to determine more suitable\naugmentation methods. Considering the significance of temporal domain\ninformation for micro-gestures, we introduce a simple and efficient\nplug-and-play spatiotemporal balancing fusion method. We not only studied our\nmethod on the considered micro-gesture dataset but also conducted experiments\non mainstream action datasets. The results show that our approach performs well\nin micro-gesture recognition and on other datasets, achieving state-of-the-art\nperformance compared to previous micro-gesture recognition methods. For\nemotional understanding based on micro-gestures, we construct complex emotional\nreasoning scenarios. Our evaluation, conducted with large language models,\nshows that micro-gestures play a significant and positive role in enhancing\ncomprehensive emotional understanding. The scenarios we developed can be\nextended to other micro-gesture-based tasks such as deception detection and\ninterviews. We confirm that our new insights contribute to advancing research\nin micro-gesture and emotional artificial intelligence.\n", "date": "2024-05-21", "categories": []}
{"id": "2502.02938", "title": "LLaVAC: Fine-tuning LLaVA as a Multimodal Sentiment Classifier", "abstract": "  We present LLaVAC, a method for constructing a classifier for multimodal\nsentiment analysis. This method leverages fine-tuning of the Large Language and\nVision Assistant (LLaVA) to predict sentiment labels across both image and text\nmodalities. Our approach involves designing a structured prompt that\nincorporates both unimodal and multimodal labels to fine-tune LLaVA, enabling\nit to perform sentiment classification effectively. Experiments on the\nMVSA-Single dataset demonstrate that LLaVAC outperforms existing methods in\nmultimodal sentiment analysis across three data processing procedures. The\nimplementation of LLaVAC is publicly available at\nhttps://github.com/tchayintr/llavac.\n", "date": "2025-02-05", "categories": []}
{"id": "2405.00574", "title": "EALD-MLLM: Emotion Analysis in Long-sequential and De-identity videos\n  with Multi-modal Large Language Model", "abstract": "  Emotion AI is the ability of computers to understand human emotional states.\nExisting works have achieved promising progress, but two limitations remain to\nbe solved: 1) Previous studies have been more focused on short sequential video\nemotion analysis while overlooking long sequential video. However, the emotions\nin short sequential videos only reflect instantaneous emotions, which may be\ndeliberately guided or hidden. In contrast, long sequential videos can reveal\nauthentic emotions; 2) Previous studies commonly utilize various signals such\nas facial, speech, and even sensitive biological signals (e.g.,\nelectrocardiogram). However, due to the increasing demand for privacy,\ndeveloping Emotion AI without relying on sensitive signals is becoming\nimportant. To address the aforementioned limitations, in this paper, we\nconstruct a dataset for Emotion Analysis in Long-sequential and De-identity\nvideos called EALD by collecting and processing the sequences of athletes'\npost-match interviews. In addition to providing annotations of the overall\nemotional state of each video, we also provide the Non-Facial Body Language\n(NFBL) annotations for each player. NFBL is an inner-driven emotional\nexpression and can serve as an identity-free clue to understanding the\nemotional state. Moreover, we provide a simple but effective baseline for\nfurther research. More precisely, we evaluate the Multimodal Large Language\nModels (MLLMs) with de-identification signals (e.g., visual, speech, and NFBLs)\nto perform emotion analysis. Our experimental results demonstrate that: 1)\nMLLMs can achieve comparable, even better performance than the supervised\nsingle-modal models, even in a zero-shot scenario; 2) NFBL is an important cue\nin long sequential emotion analysis. EALD will be available on the open-source\nplatform.\n", "date": "2024-05-01", "categories": []}
{"id": "2502.00310", "title": "SigWavNet: Learning Multiresolution Signal Wavelet Network for Speech\n  Emotion Recognition", "abstract": "  In the field of human-computer interaction and psychological assessment,\nspeech emotion recognition (SER) plays an important role in deciphering\nemotional states from speech signals. Despite advancements, challenges persist\ndue to system complexity, feature distinctiveness issues, and noise\ninterference. This paper introduces a new end-to-end (E2E) deep learning\nmulti-resolution framework for SER, addressing these limitations by extracting\nmeaningful representations directly from raw waveform speech signals. By\nleveraging the properties of the fast discrete wavelet transform (FDWT),\nincluding the cascade algorithm, conjugate quadrature filter, and coefficient\ndenoising, our approach introduces a learnable model for both wavelet bases and\ndenoising through deep learning techniques. The framework incorporates an\nactivation function for learnable asymmetric hard thresholding of wavelet\ncoefficients. Our approach exploits the capabilities of wavelets for effective\nlocalization in both time and frequency domains. We then combine\none-dimensional dilated convolutional neural networks (1D dilated CNN) with a\nspatial attention layer and bidirectional gated recurrent units (Bi-GRU) with a\ntemporal attention layer to efficiently capture the nuanced spatial and\ntemporal characteristics of emotional features. By handling variable-length\nspeech without segmentation and eliminating the need for pre or\npost-processing, the proposed model outperformed state-of-the-art methods on\nIEMOCAP and EMO-DB datasets. The source code of this paper is shared on the\nGithub repository:\nhttps://github.com/alaaNfissi/SigWavNet-Learning-Multiresolution-Signal-Wavelet-Network-for-Speech-Emotion-Recognition.\n", "date": "2025-02-01", "categories": []}
{"id": "2502.00547", "title": "Milmer: a Framework for Multiple Instance Learning based Multimodal\n  Emotion Recognition", "abstract": "  Emotions play a crucial role in human behavior and decision-making, making\nemotion recognition a key area of interest in human-computer interaction (HCI).\nThis study addresses the challenges of emotion recognition by integrating\nfacial expression analysis with electroencephalogram (EEG) signals, introducing\na novel multimodal framework-Milmer. The proposed framework employs a\ntransformer-based fusion approach to effectively integrate visual and\nphysiological modalities. It consists of an EEG preprocessing module, a facial\nfeature extraction and balancing module, and a cross-modal fusion module. To\nenhance visual feature extraction, we fine-tune a pre-trained Swin Transformer\non emotion-related datasets. Additionally, a cross-attention mechanism is\nintroduced to balance token representation across modalities, ensuring\neffective feature integration. A key innovation of this work is the adoption of\na multiple instance learning (MIL) approach, which extracts meaningful\ninformation from multiple facial expression images over time, capturing\ncritical temporal dynamics often overlooked in previous studies. Extensive\nexperiments conducted on the DEAP dataset demonstrate the superiority of the\nproposed framework, achieving a classification accuracy of 96.72% in the\nfour-class emotion recognition task. Ablation studies further validate the\ncontributions of each module, highlighting the significance of advanced feature\nextraction and fusion strategies in enhancing emotion recognition performance.\nOur code are available at https://github.com/liangyubuaa/Milmer.\n", "date": "2025-02-01", "categories": []}
{"id": "2502.00654", "title": "EmoTalkingGaussian: Continuous Emotion-conditioned Talking Head\n  Synthesis", "abstract": "  3D Gaussian splatting-based talking head synthesis has recently gained\nattention for its ability to render high-fidelity images with real-time\ninference speed. However, since it is typically trained on only a short video\nthat lacks the diversity in facial emotions, the resultant talking heads\nstruggle to represent a wide range of emotions. To address this issue, we\npropose a lip-aligned emotional face generator and leverage it to train our\nEmoTalkingGaussian model. It is able to manipulate facial emotions conditioned\non continuous emotion values (i.e., valence and arousal); while retaining\nsynchronization of lip movements with input audio. Additionally, to achieve the\naccurate lip synchronization for in-the-wild audio, we introduce a\nself-supervised learning method that leverages a text-to-speech network and a\nvisual-audio synchronization network. We experiment our EmoTalkingGaussian on\npublicly available videos and have obtained better results than\nstate-of-the-arts in terms of image quality (measured in PSNR, SSIM, LPIPS),\nemotion expression (measured in V-RMSE, A-RMSE, V-SA, A-SA, Emotion Accuracy),\nand lip synchronization (measured in LMD, Sync-E, Sync-C), respectively.\n", "date": "2025-02-02", "categories": []}
{"id": "2405.08992", "title": "Contextual Emotion Recognition using Large Vision Language Models", "abstract": "  \"How does the person in the bounding box feel?\" Achieving human-level\nrecognition of the apparent emotion of a person in real world situations\nremains an unsolved task in computer vision. Facial expressions are not enough:\nbody pose, contextual knowledge, and commonsense reasoning all contribute to\nhow humans perform this emotional theory of mind task. In this paper, we\nexamine two major approaches enabled by recent large vision language models: 1)\nimage captioning followed by a language-only LLM, and 2) vision language\nmodels, under zero-shot and fine-tuned setups. We evaluate the methods on the\nEmotions in Context (EMOTIC) dataset and demonstrate that a vision language\nmodel, fine-tuned even on a small dataset, can significantly outperform\ntraditional baselines. The results of this work aim to help robots and agents\nperform emotionally sensitive decision-making and interaction in the future.\n", "date": "2024-05-14", "categories": []}
{"id": "2501.18538", "title": "Mini-ResEmoteNet: Leveraging Knowledge Distillation for Human-Centered\n  Design", "abstract": "  Facial Emotion Recognition has emerged as increasingly pivotal in the domain\nof User Experience, notably within modern usability testing, as it facilitates\na deeper comprehension of user satisfaction and engagement. This study aims to\nextend the ResEmoteNet model by employing a knowledge distillation framework to\ndevelop Mini-ResEmoteNet models - lightweight student models - tailored for\nusability testing. Experiments were conducted on the FER2013 and RAF-DB\ndatasets to assess the efficacy of three student model architectures: Student\nModel A, Student Model B, and Student Model C. Their development involves\nreducing the number of feature channels in each layer of the teacher model by\napproximately 50%, 75%, and 87.5%. Demonstrating exceptional performance on the\nFER2013 dataset, Student Model A (E1) achieved a test accuracy of 76.33%,\nmarking a 0.21% absolute improvement over EmoNeXt. Moreover, the results\nexhibit absolute improvements in terms of inference speed and memory usage\nduring inference compared to the ResEmoteNet model. The findings indicate that\nthe proposed methods surpass other state-of-the-art approaches.\n", "date": "2025-01-30", "categories": []}
{"id": "2501.17261", "title": "NUS-Emo at SemEval-2024 Task 3: Instruction-Tuning LLM for Multimodal\n  Emotion-Cause Analysis in Conversations", "abstract": "  This paper describes the architecture of our system developed for Task 3 of\nSemEval-2024: Multimodal Emotion-Cause Analysis in Conversations. Our project\ntargets the challenges of subtask 2, dedicated to Multimodal Emotion-Cause Pair\nExtraction with Emotion Category (MECPE-Cat), and constructs a dual-component\nsystem tailored to the unique challenges of this task. We divide the task into\ntwo subtasks: emotion recognition in conversation (ERC) and emotion-cause pair\nextraction (ECPE). To address these subtasks, we capitalize on the abilities of\nLarge Language Models (LLMs), which have consistently demonstrated\nstate-of-the-art performance across various natural language processing tasks\nand domains. Most importantly, we design an approach of emotion-cause-aware\ninstruction-tuning for LLMs, to enhance the perception of the emotions with\ntheir corresponding causal rationales. Our method enables us to adeptly\nnavigate the complexities of MECPE-Cat, achieving a weighted average 34.71% F1\nscore of the task, and securing the 2nd rank on the leaderboard. The code and\nmetadata to reproduce our experiments are all made publicly available.\n", "date": "2024-08-22", "categories": []}
{"id": "2501.14981", "title": "The Muddy Waters of Modeling Empathy in Language: The Practical Impacts\n  of Theoretical Constructs", "abstract": "  Conceptual operationalizations of empathy in NLP are varied, with some having\nspecific behaviors and properties, while others are more abstract. How these\nvariations relate to one another and capture properties of empathy observable\nin text remains unclear. To provide insight into this, we analyze the transfer\nperformance of empathy models adapted to empathy tasks with different\ntheoretical groundings. We study (1) the dimensionality of empathy definitions,\n(2) the correspondence between the defined dimensions and measured/observed\nproperties, and (3) the conduciveness of the data to represent them, finding\nthey have a significant impact to performance compared to other transfer\nsetting features. Characterizing the theoretical grounding of empathy tasks as\ndirect, abstract, or adjacent further indicates that tasks that directly\npredict specified empathy components have higher transferability. Our work\nprovides empirical evidence for the need for precise and multidimensional\nempathy operationalizations.\n", "date": "2025-01-24", "categories": []}
{"id": "2501.15063", "title": "Cross-modal Context Fusion and Adaptive Graph Convolutional Network for\n  Multimodal Conversational Emotion Recognition", "abstract": "  Emotion recognition has a wide range of applications in human-computer\ninteraction, marketing, healthcare, and other fields. In recent years, the\ndevelopment of deep learning technology has provided new methods for emotion\nrecognition. Prior to this, many emotion recognition methods have been\nproposed, including multimodal emotion recognition methods, but these methods\nignore the mutual interference between different input modalities and pay\nlittle attention to the directional dialogue between speakers. Therefore, this\narticle proposes a new multimodal emotion recognition method, including a cross\nmodal context fusion module, an adaptive graph convolutional encoding module,\nand an emotion classification module. The cross modal context module includes a\ncross modal alignment module and a context fusion module, which are used to\nreduce the noise introduced by mutual interference between different input\nmodalities. The adaptive graph convolution module constructs a dialogue\nrelationship graph for extracting dependencies and self dependencies between\nspeakers. Our model has surpassed some state-of-the-art methods on publicly\navailable benchmark datasets and achieved high recognition accuracy.\n", "date": "2025-01-25", "categories": []}
{"id": "2501.15519", "title": "Fuzzy-aware Loss for Source-free Domain Adaptation in Visual Emotion\n  Recognition", "abstract": "  Source-free domain adaptation in visual emotion recognition (SFDA-VER) is a\nhighly challenging task that requires adapting VER models to the target domain\nwithout relying on source data, which is of great significance for data privacy\nprotection. However, due to the unignorable disparities between visual emotion\ndata and traditional image classification data, existing SFDA methods perform\npoorly on this task. In this paper, we investigate the SFDA-VER task from a\nfuzzy perspective and identify two key issues: fuzzy emotion labels and fuzzy\npseudo-labels. These issues arise from the inherent uncertainty of emotion\nannotations and the potential mispredictions in pseudo-labels. To address these\nissues, we propose a novel fuzzy-aware loss (FAL) to enable the VER model to\nbetter learn and adapt to new domains under fuzzy labels. Specifically, FAL\nmodifies the standard cross entropy loss and focuses on adjusting the losses of\nnon-predicted categories, which prevents a large number of uncertain or\nincorrect predictions from overwhelming the VER model during adaptation. In\naddition, we provide a theoretical analysis of FAL and prove its robustness in\nhandling the noise in generated pseudo-labels. Extensive experiments on 26\ndomain adaptation sub-tasks across three benchmark datasets demonstrate the\neffectiveness of our method.\n", "date": "2025-01-26", "categories": []}
{"id": "2501.10408", "title": "Leveraging Cross-Attention Transformer and Multi-Feature Fusion for\n  Cross-Linguistic Speech Emotion Recognition", "abstract": "  Speech Emotion Recognition (SER) plays a crucial role in enhancing\nhuman-computer interaction. Cross-Linguistic SER (CLSER) has been a challenging\nresearch problem due to significant variability in linguistic and acoustic\nfeatures of different languages. In this study, we propose a novel approach\nHuMP-CAT, which combines HuBERT, MFCC, and prosodic characteristics. These\nfeatures are fused using a cross-attention transformer (CAT) mechanism during\nfeature extraction. Transfer learning is applied to gain from a source\nemotional speech dataset to the target corpus for emotion recognition. We use\nIEMOCAP as the source dataset to train the source model and evaluate the\nproposed method on seven datasets in five languages (e.g., English, German,\nSpanish, Italian, and Chinese). We show that, by fine-tuning the source model\nwith a small portion of speech from the target datasets, HuMP-CAT achieves an\naverage accuracy of 78.75% across the seven datasets, with notable performance\nof 88.69% on EMODB (German language) and 79.48% on EMOVO (Italian language).\nOur extensive evaluation demonstrates that HuMP-CAT outperforms existing\nmethods across multiple target languages.\n", "date": "2025-01-06", "categories": []}
{"id": "2501.10666", "title": "Speech Emotion Detection Based on MFCC and CNN-LSTM Architecture", "abstract": "  Emotion detection techniques have been applied to multiple cases mainly from\nfacial image features and vocal audio features, of which the latter aspect is\ndisputed yet not only due to the complexity of speech audio processing but also\nthe difficulties of extracting appropriate features. Part of the SAVEE and\nRAVDESS datasets are selected and combined as the dataset, containing seven\nsorts of common emotions (i.e. happy, neutral, sad, anger, disgust, fear, and\nsurprise) and thousands of samples. Based on the Librosa package, this paper\nprocesses the initial audio input into waveplot and spectrum for analysis and\nconcentrates on multiple features including MFCC as targets for feature\nextraction. The hybrid CNN-LSTM architecture is adopted by virtue of its strong\ncapability to deal with sequential data and time series, which mainly consists\nof four convolutional layers and three long short-term memory layers. As a\nresult, the architecture achieved an accuracy of 61.07% comprehensively for the\ntest set, among which the detection of anger and neutral reaches a performance\nof 75.31% and 71.70% respectively. It can also be concluded that the\nclassification accuracy is dependent on the properties of emotion to some\nextent, with frequently-used and distinct-featured emotions having less\nprobability to be misclassified into other categories. Emotions like surprise\nwhose meaning depends on the specific context are more likely to confuse with\npositive or negative emotions, and negative emotions also have a possibility to\nget mixed with each other.\n", "date": "2025-01-18", "categories": []}
{"id": "2501.11166", "title": "AIMA at SemEval-2024 Task 10: History-Based Emotion Recognition in\n  Hindi-English Code-Mixed Conversations", "abstract": "  In this study, we introduce a solution to the SemEval 2024 Task 10 on subtask\n1, dedicated to Emotion Recognition in Conversation (ERC) in code-mixed\nHindi-English conversations. ERC in code-mixed conversations presents unique\nchallenges, as existing models are typically trained on monolingual datasets\nand may not perform well on code-mixed data. To address this, we propose a\nseries of models that incorporate both the previous and future context of the\ncurrent utterance, as well as the sequential information of the conversation.\nTo facilitate the processing of code-mixed data, we developed a\nHinglish-to-English translation pipeline to translate the code-mixed\nconversations into English. We designed four different base models, each\nutilizing powerful pre-trained encoders to extract features from the input but\nwith varying architectures. By ensembling all of these models, we developed a\nfinal model that outperforms all other baselines.\n", "date": "2025-01-19", "categories": []}
{"id": "2501.11170", "title": "AIMA at SemEval-2024 Task 3: Simple Yet Powerful Emotion Cause Pair\n  Analysis", "abstract": "  The SemEval-2024 Task 3 presents two subtasks focusing on emotion-cause pair\nextraction within conversational contexts. Subtask 1 revolves around the\nextraction of textual emotion-cause pairs, where causes are defined and\nannotated as textual spans within the conversation. Conversely, Subtask 2\nextends the analysis to encompass multimodal cues, including language, audio,\nand vision, acknowledging instances where causes may not be exclusively\nrepresented in the textual data. Our proposed model for emotion-cause analysis\nis meticulously structured into three core segments: (i) embedding extraction,\n(ii) cause-pair extraction & emotion classification, and (iii) cause extraction\nusing QA after finding pairs. Leveraging state-of-the-art techniques and\nfine-tuning on task-specific datasets, our model effectively unravels the\nintricate web of conversational dynamics and extracts subtle cues signifying\ncausality in emotional expressions. Our team, AIMA, demonstrated strong\nperformance in the SemEval-2024 Task 3 competition. We ranked as the 10th in\nsubtask 1 and the 6th in subtask 2 out of 23 teams.\n", "date": "2025-01-19", "categories": []}
{"id": "2501.11570", "title": "Uncertainty Estimation in the Real World: A Study on Music Emotion\n  Recognition", "abstract": "  Any data annotation for subjective tasks shows potential variations between\nindividuals. This is particularly true for annotations of emotional responses\nto musical stimuli. While older approaches to music emotion recognition systems\nfrequently addressed this uncertainty problem through probabilistic modeling,\nmodern systems based on neural networks tend to ignore the variability and\nfocus only on predicting central tendencies of human subjective responses. In\nthis work, we explore several methods for estimating not only the central\ntendencies of the subjective responses to a musical stimulus, but also for\nestimating the uncertainty associated with these responses. In particular, we\ninvestigate probabilistic loss functions and inference-time random sampling.\nExperimental results indicate that while the modeling of the central tendencies\nis achievable, modeling of the uncertainty in subjective responses proves\nsignificantly more challenging with currently available approaches even when\nempirical estimates of variations in the responses are available.\n", "date": "2025-01-20", "categories": []}
{"id": "2501.09502", "title": "Omni-Emotion: Extending Video MLLM with Detailed Face and Audio Modeling\n  for Multimodal Emotion Analysis", "abstract": "  Understanding emotions accurately is essential for fields like human-computer\ninteraction. Due to the complexity of emotions and their multi-modal nature\n(e.g., emotions are influenced by facial expressions and audio), researchers\nhave turned to using multi-modal models to understand human emotions rather\nthan single-modality. However, current video multi-modal large language models\n(MLLMs) encounter difficulties in effectively integrating audio and identifying\nsubtle facial micro-expressions. Furthermore, the lack of detailed emotion\nanalysis datasets also limits the development of multimodal emotion analysis.\nTo address these issues, we introduce a self-reviewed dataset and a\nhuman-reviewed dataset, comprising 24,137 coarse-grained samples and 3,500\nmanually annotated samples with detailed emotion annotations, respectively.\nThese datasets allow models to learn from diverse scenarios and better\ngeneralize to real-world applications. Moreover, in addition to the audio\nmodeling, we propose to explicitly integrate facial encoding models into the\nexisting advanced Video MLLM, enabling the MLLM to effectively unify audio and\nthe subtle facial cues for emotion understanding. By aligning these features\nwithin a unified space and employing instruction tuning in our proposed\ndatasets, our Omni-Emotion achieves state-of-the-art performance in both\nemotion recognition and reasoning tasks.\n", "date": "2025-01-16", "categories": []}
{"id": "2501.08696", "title": "Deep Learning-Based Feature Fusion for Emotion Analysis and Suicide Risk\n  Differentiation in Chinese Psychological Support Hotlines", "abstract": "  Mental health is a critical global public health issue, and psychological\nsupport hotlines play a pivotal role in providing mental health assistance and\nidentifying suicide risks at an early stage. However, the emotional expressions\nconveyed during these calls remain underexplored in current research. This\nstudy introduces a method that combines pitch acoustic features with deep\nlearning-based features to analyze and understand emotions expressed during\nhotline interactions. Using data from China's largest psychological support\nhotline, our method achieved an F1-score of 79.13% for negative binary emotion\nclassification.Additionally, the proposed approach was validated on an open\ndataset for multi-class emotion classification,where it demonstrated better\nperformance compared to the state-of-the-art methods. To explore its clinical\nrelevance, we applied the model to analysis the frequency of negative emotions\nand the rate of emotional change in the conversation, comparing 46 subjects\nwith suicidal behavior to those without. While the suicidal group exhibited\nmore frequent emotional changes than the non-suicidal group, the difference was\nnot statistically significant.Importantly, our findings suggest that emotional\nfluctuation intensity and frequency could serve as novel features for\npsychological assessment scales and suicide risk prediction.The proposed method\nprovides valuable insights into emotional dynamics and has the potential to\nadvance early intervention and improve suicide prevention strategies through\nintegration with clinical tools and assessments The source code is publicly\navailable at https://github.com/Sco-field/Speechemotionrecognition/tree/main.\n", "date": "2025-01-15", "categories": []}
{"id": "2501.07746", "title": "A Heterogeneous Multimodal Graph Learning Framework for Recognizing User\n  Emotions in Social Networks", "abstract": "  The rapid expansion of social media platforms has provided unprecedented\naccess to massive amounts of multimodal user-generated content. Comprehending\nuser emotions can provide valuable insights for improving communication and\nunderstanding of human behaviors. Despite significant advancements in Affective\nComputing, the diverse factors influencing user emotions in social networks\nremain relatively understudied. Moreover, there is a notable lack of deep\nlearning-based methods for predicting user emotions in social networks, which\ncould be addressed by leveraging the extensive multimodal data available. This\nwork presents a novel formulation of personalized emotion prediction in social\nnetworks based on heterogeneous graph learning. Building upon this formulation,\nwe design HMG-Emo, a Heterogeneous Multimodal Graph Learning Framework that\nutilizes deep learning-based features for user emotion recognition.\nAdditionally, we include a dynamic context fusion module in HMG-Emo that is\ncapable of adaptively integrating the different modalities in social media\ndata. Through extensive experiments, we demonstrate the effectiveness of\nHMG-Emo and verify the superiority of adopting a graph neural network-based\napproach, which outperforms existing baselines that use rich hand-crafted\nfeatures. To the best of our knowledge, HMG-Emo is the first multimodal and\ndeep-learning-based approach to predict personalized emotions within online\nsocial networks. Our work highlights the significance of exploiting advanced\ndeep learning techniques for less-explored problems in Affective Computing.\n", "date": "2025-01-13", "categories": []}
{"id": "2501.08085", "title": "Dynamic Multimodal Sentiment Analysis: Leveraging Cross-Modal Attention\n  for Enabled Classification", "abstract": "  This paper explores the development of a multimodal sentiment analysis model\nthat integrates text, audio, and visual data to enhance sentiment\nclassification. The goal is to improve emotion detection by capturing the\ncomplex interactions between these modalities, thereby enabling more accurate\nand nuanced sentiment interpretation. The study evaluates three feature fusion\nstrategies -- late stage fusion, early stage fusion, and multi-headed attention\n-- within a transformer-based architecture. Experiments were conducted using\nthe CMU-MOSEI dataset, which includes synchronized text, audio, and visual\ninputs labeled with sentiment scores. Results show that early stage fusion\nsignificantly outperforms late stage fusion, achieving an accuracy of 71.87\\%,\nwhile the multi-headed attention approach offers marginal improvement, reaching\n72.39\\%. The findings suggest that integrating modalities early in the process\nenhances sentiment classification, while attention mechanisms may have limited\nimpact within the current framework. Future work will focus on refining feature\nfusion techniques, incorporating temporal data, and exploring dynamic feature\nweighting to further improve model performance.\n", "date": "2025-01-14", "categories": []}
{"id": "2501.08182", "title": "CG-MER: A Card Game-based Multimodal dataset for Emotion Recognition", "abstract": "  The field of affective computing has seen significant advancements in\nexploring the relationship between emotions and emerging technologies. This\npaper presents a novel and valuable contribution to this field with the\nintroduction of a comprehensive French multimodal dataset designed specifically\nfor emotion recognition. The dataset encompasses three primary modalities:\nfacial expressions, speech, and gestures, providing a holistic perspective on\nemotions. Moreover, the dataset has the potential to incorporate additional\nmodalities, such as Natural Language Processing (NLP) to expand the scope of\nemotion recognition research. The dataset was curated through engaging\nparticipants in card game sessions, where they were prompted to express a range\nof emotions while responding to diverse questions. The study included 10\nsessions with 20 participants (9 females and 11 males). The dataset serves as a\nvaluable resource for furthering research in emotion recognition and provides\nan avenue for exploring the intricate connections between human emotions and\ndigital technologies.\n", "date": "2025-01-14", "categories": []}
{"id": "2501.08199", "title": "EmoNeXt: an Adapted ConvNeXt for Facial Emotion Recognition", "abstract": "  Facial expressions play a crucial role in human communication serving as a\npowerful and impactful means to express a wide range of emotions. With\nadvancements in artificial intelligence and computer vision, deep neural\nnetworks have emerged as effective tools for facial emotion recognition. In\nthis paper, we propose EmoNeXt, a novel deep learning framework for facial\nexpression recognition based on an adapted ConvNeXt architecture network. We\nintegrate a Spatial Transformer Network (STN) to focus on feature-rich regions\nof the face and Squeeze-and-Excitation blocks to capture channel-wise\ndependencies. Moreover, we introduce a self-attention regularization term,\nencouraging the model to generate compact feature vectors. We demonstrate the\nsuperiority of our model over existing state-of-the-art deep learning models on\nthe FER2013 dataset regarding emotion classification accuracy.\n", "date": "2025-01-14", "categories": []}
{"id": "2501.06215", "title": "Fitting Different Interactive Information: Joint Classification of\n  Emotion and Intention", "abstract": "  This paper is the first-place solution for ICASSP MEIJU@2025 Track I, which\nfocuses on low-resource multimodal emotion and intention recognition. How to\neffectively utilize a large amount of unlabeled data, while ensuring the mutual\npromotion of different difficulty levels tasks in the interaction stage, these\ntwo points become the key to the competition. In this paper, pseudo-label\nlabeling is carried out on the model trained with labeled data, and samples\nwith high confidence and their labels are selected to alleviate the problem of\nlow resources. At the same time, the characteristic of easy represented ability\nof intention recognition found in the experiment is used to make mutually\npromote with emotion recognition under different attention heads, and higher\nperformance of intention recognition is achieved through fusion. Finally, under\nthe refined processing data, we achieve the score of 0.5532 in the Test set,\nand win the championship of the track.\n", "date": "2025-01-05", "categories": []}
{"id": "2501.06475", "title": "Enhancing Multi-Modal Video Sentiment Classification Through\n  Semi-Supervised Clustering", "abstract": "  Understanding emotions in videos is a challenging task. However, videos\ncontain several modalities which make them a rich source of data for machine\nlearning and deep learning tasks. In this work, we aim to improve video\nsentiment classification by focusing on two key aspects: the video itself, the\naccompanying text, and the acoustic features. To address the limitations of\nrelying on large labeled datasets, we are developing a method that utilizes\nclustering-based semi-supervised pre-training to extract meaningful\nrepresentations from the data. This pre-training step identifies patterns in\nthe video and text data, allowing the model to learn underlying structures and\nrelationships without requiring extensive labeled information at the outset.\nOnce these patterns are established, we fine-tune the system in a supervised\nmanner to classify the sentiment expressed in videos. We believe that this\nmulti-modal approach, combining clustering with supervised fine-tuning, will\nlead to more accurate and insightful sentiment classification, especially in\ncases where labeled data is limited.\n", "date": "2025-01-11", "categories": []}
{"id": "2501.05461", "title": "Beyond Questionnaires: Video Analysis for Social Anxiety Detection", "abstract": "  Social Anxiety Disorder (SAD) significantly impacts individuals' daily lives\nand relationships. The conventional methods for SAD detection involve physical\nconsultations and self-reported questionnaires, but they have limitations such\nas time consumption and bias. This paper introduces video analysis as a\npromising method for early SAD detection. Specifically, we present a new\napproach for detecting SAD in individuals from various bodily features\nextracted from the video data. We conducted a study to collect video data of 92\nparticipants performing impromptu speech in a controlled environment. Using the\nvideo data, we studied the behavioral change in participants' head, body, eye\ngaze, and action units. By applying a range of machine learning and deep\nlearning algorithms, we achieved an accuracy rate of up to 74\\% in classifying\nparticipants as SAD or non-SAD. Video-based SAD detection offers a\nnon-intrusive and scalable approach that can be deployed in real-time,\npotentially enhancing early detection and intervention capabilities.\n", "date": "2024-12-26", "categories": []}
{"id": "2501.05474", "title": "Modality-Invariant Bidirectional Temporal Representation Distillation\n  Network for Missing Multimodal Sentiment Analysis", "abstract": "  Multimodal Sentiment Analysis (MSA) integrates diverse modalities(text,\naudio, and video) to comprehensively analyze and understand individuals'\nemotional states. However, the real-world prevalence of incomplete data poses\nsignificant challenges to MSA, mainly due to the randomness of modality\nmissing. Moreover, the heterogeneity issue in multimodal data has yet to be\neffectively addressed. To tackle these challenges, we introduce the\nModality-Invariant Bidirectional Temporal Representation Distillation Network\n(MITR-DNet) for Missing Multimodal Sentiment Analysis. MITR-DNet employs a\ndistillation approach, wherein a complete modality teacher model guides a\nmissing modality student model, ensuring robustness in the presence of modality\nmissing. Simultaneously, we developed the Modality-Invariant Bidirectional\nTemporal Representation Learning Module (MIB-TRL) to mitigate heterogeneity.\n", "date": "2025-01-07", "categories": []}
{"id": "2501.02539", "title": "AHMSA-Net: Adaptive Hierarchical Multi-Scale Attention Network for\n  Micro-Expression Recognition", "abstract": "  Micro-expression recognition (MER) presents a significant challenge due to\nthe transient and subtle nature of the motion changes involved. In recent\nyears, deep learning methods based on attention mechanisms have made some\nbreakthroughs in MER. However, these methods still suffer from the limitations\nof insufficient feature capture and poor dynamic adaptation when coping with\nthe instantaneous subtle movement changes of micro-expressions. Therefore, in\nthis paper, we design an Adaptive Hierarchical Multi-Scale Attention Network\n(AHMSA-Net) for MER. Specifically, we first utilize the onset and apex frames\nof the micro-expression sequence to extract three-dimensional (3D) optical flow\nmaps, including horizontal optical flow, vertical optical flow, and optical\nflow strain. Subsequently, the optical flow feature maps are inputted into\nAHMSA-Net, which consists of two parts: an adaptive hierarchical framework and\na multi-scale attention mechanism. Based on the adaptive downsampling\nhierarchical attention framework, AHMSA-Net captures the subtle changes of\nmicro-expressions from different granularities (fine and coarse) by dynamically\nadjusting the size of the optical flow feature map at each layer. Based on the\nmulti-scale attention mechanism, AHMSA-Net learns micro-expression action\ninformation by fusing features from different scales (channel and spatial).\nThese two modules work together to comprehensively improve the accuracy of MER.\nAdditionally, rigorous experiments demonstrate that the proposed method\nachieves competitive results on major micro-expression databases, with\nAHMSA-Net achieving recognition accuracy of up to 78.21% on composite databases\n(SMIC, SAMM, CASMEII) and 77.08% on the CASME^{}3 database.\n", "date": "2025-01-05", "categories": []}
{"id": "2501.03103", "title": "MVP: Multimodal Emotion Recognition based on Video and Physiological\n  Signals", "abstract": "  Human emotions entail a complex set of behavioral, physiological and\ncognitive changes. Current state-of-the-art models fuse the behavioral and\nphysiological components using classic machine learning, rather than recent\ndeep learning techniques. We propose to fill this gap, designing the Multimodal\nfor Video and Physio (MVP) architecture, streamlined to fuse video and\nphysiological signals. Differently then others approaches, MVP exploits the\nbenefits of attention to enable the use of long input sequences (1-2 minutes).\nWe have studied video and physiological backbones for inputting long sequences\nand evaluated our method with respect to the state-of-the-art. Our results show\nthat MVP outperforms former methods for emotion recognition based on facial\nvideos, EDA, and ECG/PPG.\n", "date": "2025-01-06", "categories": []}
{"id": "2412.17837", "title": "Evaluating the Capabilities of Large Language Models for Multi-label\n  Emotion Understanding", "abstract": "  Large Language Models (LLMs) show promising learning and reasoning abilities.\nCompared to other NLP tasks, multilingual and multi-label emotion evaluation\ntasks are under-explored in LLMs. In this paper, we present EthioEmo, a\nmulti-label emotion classification dataset for four Ethiopian languages,\nnamely, Amharic (amh), Afan Oromo (orm), Somali (som), and Tigrinya (tir). We\nperform extensive experiments with an additional English multi-label emotion\ndataset from SemEval 2018 Task 1. Our evaluation includes encoder-only,\nencoder-decoder, and decoder-only language models. We compare zero and few-shot\napproaches of LLMs to fine-tuning smaller language models. The results show\nthat accurate multi-label emotion classification is still insufficient even for\nhigh-resource languages such as English, and there is a large gap between the\nperformance of high-resource and low-resource languages. The results also show\nvarying performance levels depending on the language and model type. EthioEmo\nis available publicly to further improve the understanding of emotions in\nlanguage models and how people convey emotions through various languages.\n", "date": "2024-12-17", "categories": []}
{"id": "2501.01816", "title": "Uncertainty-Aware Label Refinement on Hypergraphs for Personalized\n  Federated Facial Expression Recognition", "abstract": "  Most facial expression recognition (FER) models are trained on large-scale\nexpression data with centralized learning. Unfortunately, collecting a large\namount of centralized expression data is difficult in practice due to privacy\nconcerns of facial images. In this paper, we investigate FER under the\nframework of personalized federated learning, which is a valuable and practical\ndecentralized setting for real-world applications. To this end, we develop a\nnovel uncertainty-Aware label refineMent on hYpergraphs (AMY) method. For local\ntraining, each local model consists of a backbone, an uncertainty estimation\n(UE) block, and an expression classification (EC) block. In the UE block, we\nleverage a hypergraph to model complex high-order relationships between\nexpression samples and incorporate these relationships into uncertainty\nfeatures. A personalized uncertainty estimator is then introduced to estimate\nreliable uncertainty weights of samples in the local client. In the EC block,\nwe perform label propagation on the hypergraph, obtaining high-quality refined\nlabels for retraining an expression classifier. Based on the above, we\neffectively alleviate heterogeneous sample uncertainty across clients and learn\na robust personalized FER model in each client. Experimental results on two\nchallenging real-world facial expression databases show that our proposed\nmethod consistently outperforms several state-of-the-art methods. This\nindicates the superiority of hypergraph modeling for uncertainty estimation and\nlabel refinement on the personalized federated FER task. The source code will\nbe released at https://github.com/mobei1006/AMY.\n", "date": "2025-01-03", "categories": []}
{"id": "2308.16349", "title": "Affective Visual Dialog: A Large-Scale Benchmark for Emotional Reasoning\n  Based on Visually Grounded Conversations", "abstract": "  We introduce Affective Visual Dialog, an emotion explanation and reasoning\ntask as a testbed for research on understanding the formation of emotions in\nvisually grounded conversations. The task involves three skills: (1)\nDialog-based Question Answering (2) Dialog-based Emotion Prediction and (3)\nAffective emotion explanation generation based on the dialog. Our key\ncontribution is the collection of a large-scale dataset, dubbed AffectVisDial,\nconsisting of 50K 10-turn visually grounded dialogs as well as concluding\nemotion attributions and dialog-informed textual emotion explanations,\nresulting in a total of 27,180 working hours. We explain our design decisions\nin collecting the dataset and introduce the questioner and answerer tasks that\nare associated with the participants in the conversation. We train and\ndemonstrate solid Affective Visual Dialog baselines adapted from\nstate-of-the-art models. Remarkably, the responses generated by our models show\npromising emotional reasoning abilities in response to visually grounded\nconversations. Our project page is available at\nhttps://affective-visual-dialog.github.io.\n", "date": "2023-08-30", "categories": []}
{"id": "2501.00449", "title": "Do Students with Different Personality Traits Demonstrate Different\n  Physiological Signals in Video-based Learning?", "abstract": "  Past researches show that personality trait is a strong predictor for ones\nacademic performance. Today, mature and verified marker systems for assessing\npersonality traits already exist. However, marker systems-based assessing\nmethods have their own limitations. For example, dishonest responses cannot be\navoided. In this research, the goal is to develop a method that can overcome\nthe limitations. The proposed method will rely on physiological signals for the\nassessment. Thirty participants have participated in this experiment. Based on\nthe statistical results, we found that there are correlations between students\npersonality traits and their physiological signal change when learning via\nvideos. Specifically, we found that participants degree of extraversion,\nagreeableness, conscientiousness, and openness to experiences are correlated\nwith the variance of heart rates, the variance of GSR values, and the skewness\nof voice frequencies, etc.\n", "date": "2024-12-31", "categories": []}
{"id": "2501.00995", "title": "Is It Still Fair? Investigating Gender Fairness in Cross-Corpus Speech\n  Emotion Recognition", "abstract": "  Speech emotion recognition (SER) is a vital component in various everyday\napplications. Cross-corpus SER models are increasingly recognized for their\nability to generalize performance. However, concerns arise regarding fairness\nacross demographics in diverse corpora. Existing fairness research often\nfocuses solely on corpus-specific fairness, neglecting its generalizability in\ncross-corpus scenarios. Our study focuses on this underexplored area, examining\nthe gender fairness generalizability in cross-corpus SER scenarios. We\nemphasize that the performance of cross-corpus SER models and their fairness\nare two distinct considerations. Moreover, we propose the approach of a\ncombined fairness adaptation mechanism to enhance gender fairness in the SER\ntransfer learning tasks by addressing both source and target genders. Our\nfindings bring one of the first insights into the generalizability of gender\nfairness in cross-corpus SER systems.\n", "date": "2025-01-02", "categories": []}
{"id": "2501.01103", "title": "learning discriminative features from spectrograms using center loss for\n  speech emotion recognition", "abstract": "  Identifying the emotional state from speech is essential for the natural\ninteraction of the machine with the speaker. However, extracting effective\nfeatures for emotion recognition is difficult, as emotions are ambiguous. We\npropose a novel approach to learn discriminative features from variable length\nspectrograms for emotion recognition by cooperating softmax cross-entropy loss\nand center loss together. The softmax cross-entropy loss enables features from\ndifferent emotion categories separable, and center loss efficiently pulls the\nfeatures belonging to the same emotion category to their center. By combining\nthe two losses together, the discriminative power will be highly enhanced,\nwhich leads to network learning more effective features for emotion\nrecognition. As demonstrated by the experimental results, after introducing\ncenter loss, both the unweighted accuracy and weighted accuracy are improved by\nover 3\\% on Mel-spectrogram input, and more than 4\\% on Short Time Fourier\nTransform spectrogram input.\n", "date": "2025-01-02", "categories": []}
{"id": "2501.01123", "title": "TED: Turn Emphasis with Dialogue Feature Attention for Emotion\n  Recognition in Conversation", "abstract": "  Emotion recognition in conversation (ERC) has been attracting attention by\nmethods for modeling multi-turn contexts. The multi-turn input to a pretraining\nmodel implicitly assumes that the current turn and other turns are\ndistinguished during the training process by inserting special tokens into the\ninput sequence. This paper proposes a priority-based attention method to\ndistinguish each turn explicitly by adding dialogue features into the attention\nmechanism, called Turn Emphasis with Dialogue (TED). It has a priority for each\nturn according to turn position and speaker information as dialogue features.\nIt takes multi-head self-attention between turn-based vectors for multi-turn\ninput and adjusts attention scores with the dialogue features. We evaluate TED\non four typical benchmarks. The experimental results demonstrate that TED has\nhigh overall performance in all datasets and achieves state-of-the-art\nperformance on IEMOCAP with numerous turns.\n", "date": "2025-01-02", "categories": []}
{"id": "2412.19909", "title": "Mouth Articulation-Based Anchoring for Improved Cross-Corpus Speech\n  Emotion Recognition", "abstract": "  Cross-corpus speech emotion recognition (SER) plays a vital role in numerous\npractical applications. Traditional approaches to cross-corpus emotion transfer\noften concentrate on adapting acoustic features to align with different\ncorpora, domains, or labels. However, acoustic features are inherently variable\nand error-prone due to factors like speaker differences, domain shifts, and\nrecording conditions. To address these challenges, this study adopts a novel\ncontrastive approach by focusing on emotion-specific articulatory gestures as\nthe core elements for analysis. By shifting the emphasis on the more stable and\nconsistent articulatory gestures, we aim to enhance emotion transfer learning\nin SER tasks. Our research leverages the CREMA-D and MSP-IMPROV corpora as\nbenchmarks and it reveals valuable insights into the commonality and\nreliability of these articulatory gestures. The findings highlight mouth\narticulatory gesture potential as a better constraint for improving emotion\nrecognition across different settings or domains.\n", "date": "2024-12-27", "categories": []}
{"id": "2412.20213", "title": "Decoding Emotion: Speech Perception Patterns in Individuals with\n  Self-reported Depression", "abstract": "  The current study examines the relationship between self-reported depression\nand the perception of affective speech within the Indian population. PANAS and\nPHQ-9 were used to assess current mood and depression, respectively.\nParticipants' emotional reactivity was recorded on a valence and arousal scale\nagainst the affective speech audio presented in a sequence. No significant\ndifferences between the depression and no-depression groups were observed for\nany of the emotional stimuli, except the audio file depicting neutral emotion.\nSignificantly higher PANAS scores by the depression than the no-depression\ngroup indicate the impact of pre-disposed mood on the current mood status.\nContrary to previous findings, this study did not observe reduced positive\nemotional reactivity by the depression group. However, the results demonstrated\nconsistency in emotional reactivity for speech stimuli depicting sadness and\nanger across all measures of emotion perception.\n", "date": "2024-12-28", "categories": []}
{"id": "2412.20359", "title": "EmoReg: Directional Latent Vector Modeling for Emotional Intensity\n  Regularization in Diffusion-based Voice Conversion", "abstract": "  The Emotional Voice Conversion (EVC) aims to convert the discrete emotional\nstate from the source emotion to the target for a given speech utterance while\npreserving linguistic content. In this paper, we propose regularizing emotion\nintensity in the diffusion-based EVC framework to generate precise speech of\nthe target emotion. Traditional approaches control the intensity of an\nemotional state in the utterance via emotion class probabilities or intensity\nlabels that often lead to inept style manipulations and degradations in\nquality. On the contrary, we aim to regulate emotion intensity using\nself-supervised learning-based feature representations and unsupervised\ndirectional latent vector modeling (DVM) in the emotional embedding space\nwithin a diffusion-based framework. These emotion embeddings can be modified\nbased on the given target emotion intensity and the corresponding direction\nvector. Furthermore, the updated embeddings can be fused in the reverse\ndiffusion process to generate the speech with the desired emotion and\nintensity. In summary, this paper aims to achieve high-quality emotional\nintensity regularization in the diffusion-based EVC framework, which is the\nfirst of its kind work. The effectiveness of the proposed method has been shown\nacross state-of-the-art (SOTA) baselines in terms of subjective and objective\nevaluations for the English and Hindi languages \\footnote{Demo samples are\navailable at the following URL: \\url{https://nirmesh-sony.github.io/EmoReg/}}.\n", "date": "2024-12-29", "categories": []}
{"id": "2412.20821", "title": "Enhancing Multimodal Emotion Recognition through Multi-Granularity\n  Cross-Modal Alignment", "abstract": "  Multimodal emotion recognition (MER), leveraging speech and text, has emerged\nas a pivotal domain within human-computer interaction, demanding sophisticated\nmethods for effective multimodal integration. The challenge of aligning\nfeatures across these modalities is significant, with most existing approaches\nadopting a singular alignment strategy. Such a narrow focus not only limits\nmodel performance but also fails to address the complexity and ambiguity\ninherent in emotional expressions. In response, this paper introduces a\nMulti-Granularity Cross-Modal Alignment (MGCMA) framework, distinguished by its\ncomprehensive approach encompassing distribution-based, instance-based, and\ntoken-based alignment modules. This framework enables a multi-level perception\nof emotional information across modalities. Our experiments on IEMOCAP\ndemonstrate that our proposed method outperforms current state-of-the-art\ntechniques.\n", "date": "2024-12-30", "categories": []}
{"id": "2412.16900", "title": "Speech-Based Depression Prediction Using Encoder-Weight-Only Transfer\n  Learning and a Large Corpus", "abstract": "  Speech-based algorithms have gained interest for the management of behavioral\nhealth conditions such as depression. We explore a speech-based transfer\nlearning approach that uses a lightweight encoder and that transfers only the\nencoder weights, enabling a simplified run-time model. Our study uses a large\ndata set containing roughly two orders of magnitude more speakers and sessions\nthan used in prior work. The large data set enables reliable estimation of\nimprovement from transfer learning. Results for the prediction of PHQ-8 labels\nshow up to 27% relative performance gains for binary classification; these\ngains are statistically significant with a p-value close to zero. Improvements\nwere also found for regression. Additionally, the gain from transfer learning\ndoes not appear to require strong source task performance. Results suggest that\nthis approach is flexible and offers promise for efficient implementation.\n", "date": "2024-12-22", "categories": []}
{"id": "2412.17907", "title": "A Multimodal Emotion Recognition System: Integrating Facial Expressions,\n  Body Movement, Speech, and Spoken Language", "abstract": "  Traditional psychological evaluations rely heavily on human observation and\ninterpretation, which are prone to subjectivity, bias, fatigue, and\ninconsistency. To address these limitations, this work presents a multimodal\nemotion recognition system that provides a standardised, objective, and\ndata-driven tool to support evaluators, such as psychologists, psychiatrists,\nand clinicians. The system integrates recognition of facial expressions,\nspeech, spoken language, and body movement analysis to capture subtle emotional\ncues that are often overlooked in human evaluations. By combining these\nmodalities, the system provides more robust and comprehensive emotional state\nassessment, reducing the risk of mis- and overdiagnosis. Preliminary testing in\na simulated real-world condition demonstrates the system's potential to provide\nreliable emotional insights to improve the diagnostic accuracy. This work\nhighlights the promise of automated multimodal analysis as a valuable\ncomplement to traditional psychological evaluation practices, with applications\nin clinical and therapeutic settings.\n", "date": "2024-12-23", "categories": []}
{"id": "2407.21315", "title": "Beyond Silent Letters: Amplifying LLMs in Emotion Recognition with Vocal\n  Nuances", "abstract": "  Emotion recognition in speech is a challenging multimodal task that requires\nunderstanding both verbal content and vocal nuances. This paper introduces a\nnovel approach to emotion detection using Large Language Models (LLMs), which\nhave demonstrated exceptional capabilities in natural language understanding.\nTo overcome the inherent limitation of LLMs in processing audio inputs, we\npropose SpeechCueLLM, a method that translates speech characteristics into\nnatural language descriptions, allowing LLMs to perform multimodal emotion\nanalysis via text prompts without any architectural changes. Our method is\nminimal yet impactful, outperforming baseline models that require structural\nmodifications. We evaluate SpeechCueLLM on two datasets: IEMOCAP and MELD,\nshowing significant improvements in emotion recognition accuracy, particularly\nfor high-quality audio data. We also explore the effectiveness of various\nfeature representations and fine-tuning strategies for different LLMs. Our\nexperiments demonstrate that incorporating speech descriptions yields a more\nthan 2% increase in the average weighted F1 score on IEMOCAP (from 70.111% to\n72.596%).\n", "date": "2024-07-31", "categories": []}
{"id": "2412.16444", "title": "Effective Context Modeling Framework for Emotion Recognition in\n  Conversations", "abstract": "  Emotion Recognition in Conversations (ERC) facilitates a deeper understanding\nof the emotions conveyed by speakers in each utterance within a conversation.\nRecently, Graph Neural Networks (GNNs) have demonstrated their strengths in\ncapturing data relationships, particularly in contextual information modeling\nand multimodal fusion. However, existing methods often struggle to fully\ncapture the complex interactions between multiple modalities and conversational\ncontext, limiting their expressiveness. To overcome these limitations, we\npropose ConxGNN, a novel GNN-based framework designed to capture contextual\ninformation in conversations. ConxGNN features two key parallel modules: a\nmulti-scale heterogeneous graph that captures the diverse effects of utterances\non emotional changes, and a hypergraph that models the multivariate\nrelationships among modalities and utterances. The outputs from these modules\nare integrated into a fusion layer, where a cross-modal attention mechanism is\napplied to produce a contextually enriched representation. Additionally,\nConxGNN tackles the challenge of recognizing minority or semantically similar\nemotion classes by incorporating a re-weighting scheme into the loss functions.\nExperimental results on the IEMOCAP and MELD benchmark datasets demonstrate the\neffectiveness of our method, achieving state-of-the-art performance compared to\nprevious baselines.\n", "date": "2024-12-21", "categories": []}
{"id": "2412.13168", "title": "Lifting Scheme-Based Implicit Disentanglement of Emotion-Related Facial\n  Dynamics in the Wild", "abstract": "  In-the-wild dynamic facial expression recognition (DFER) encounters a\nsignificant challenge in recognizing emotion-related expressions, which are\noften temporally and spatially diluted by emotion-irrelevant expressions and\nglobal context. Most prior DFER methods directly utilize coupled spatiotemporal\nrepresentations that may incorporate weakly relevant features with\nemotion-irrelevant context bias. Several DFER methods highlight dynamic\ninformation for DFER, but following explicit guidance that may be vulnerable to\nirrelevant motion. In this paper, we propose a novel Implicit Facial Dynamics\nDisentanglement framework (IFDD). Through expanding wavelet lifting scheme to\nfully learnable framework, IFDD disentangles emotion-related dynamic\ninformation from emotion-irrelevant global context in an implicit manner, i.e.,\nwithout exploit operations and external guidance. The disentanglement process\ncontains two stages. The first is Inter-frame Static-dynamic Splitting Module\n(ISSM) for rough disentanglement estimation, which explores inter-frame\ncorrelation to generate content-aware splitting indexes on-the-fly. We utilize\nthese indexes to split frame features into two groups, one with greater global\nsimilarity, and the other with more unique dynamic features. The second stage\nis Lifting-based Aggregation-Disentanglement Module (LADM) for further\nrefinement. LADM first aggregates two groups of features from ISSM to obtain\nfine-grained global context features by an updater, and then disentangles\nemotion-related facial dynamic features from the global context by a predictor.\nExtensive experiments on in-the-wild datasets have demonstrated that IFDD\noutperforms prior supervised DFER methods with higher recognition accuracy and\ncomparable efficiency. Code is available at\nhttps://github.com/CyberPegasus/IFDD.\n", "date": "2024-12-17", "categories": []}
{"id": "2412.13577", "title": "Bridge then Begin Anew: Generating Target-relevant Intermediate Model\n  for Source-free Visual Emotion Adaptation", "abstract": "  Visual emotion recognition (VER), which aims at understanding humans'\nemotional reactions toward different visual stimuli, has attracted increasing\nattention. Given the subjective and ambiguous characteristics of emotion,\nannotating a reliable large-scale dataset is hard. For reducing reliance on\ndata labeling, domain adaptation offers an alternative solution by adapting\nmodels trained on labeled source data to unlabeled target data. Conventional\ndomain adaptation methods require access to source data. However, due to\nprivacy concerns, source emotional data may be inaccessible. To address this\nissue, we propose an unexplored task: source-free domain adaptation (SFDA) for\nVER, which does not have access to source data during the adaptation process.\nTo achieve this, we propose a novel framework termed Bridge then Begin Anew\n(BBA), which consists of two steps: domain-bridged model generation (DMG) and\ntarget-related model adaptation (TMA). First, the DMG bridges cross-domain gaps\nby generating an intermediate model, avoiding direct alignment between two VER\ndatasets with significant differences. Then, the TMA begins training the target\nmodel anew to fit the target structure, avoiding the influence of\nsource-specific knowledge. Extensive experiments are conducted on six SFDA\nsettings for VER. The results demonstrate the effectiveness of BBA, which\nachieves remarkable performance gains compared with state-of-the-art SFDA\nmethods and outperforms representative unsupervised domain adaptation\napproaches.\n", "date": "2024-12-18", "categories": []}
{"id": "2412.12948", "title": "MOPO: Multi-Objective Prompt Optimization for Affective Text Generation", "abstract": "  How emotions are expressed depends on the context and domain. On X (formerly\nTwitter), for instance, an author might simply use the hashtag #anger, while in\na news headline, emotions are typically written in a more polite, indirect\nmanner. To enable conditional text generation models to create emotionally\nconnotated texts that fit a domain, users need to have access to a parameter\nthat allows them to choose the appropriate way to express an emotion. To\nachieve this, we introduce MOPO, a Multi-Objective Prompt Optimization\nmethodology. MOPO optimizes prompts according to multiple objectives (which\ncorrespond here to the output probabilities assigned by emotion classifiers\ntrained for different domains). In contrast to single objective optimization,\nMOPO outputs a set of prompts, each with a different weighting of the multiple\nobjectives. Users can then choose the most appropriate prompt for their\ncontext. We evaluate MOPO using three objectives, determined by various\ndomain-specific emotion classifiers. MOPO improves performance by up to 15 pp\nacross all objectives with a minimal loss (1-2 pp) for any single objective\ncompared to single-objective optimization. These minor performance losses are\noffset by a broader generalization across multiple objectives - which is not\npossible with single-objective optimization. Additionally, MOPO reduces\ncomputational requirements by simultaneously optimizing for multiple\nobjectives, eliminating separate optimization procedures for each objective.\n", "date": "2024-12-17", "categories": []}
{"id": "2411.05985", "title": "Emotional Images: Assessing Emotions in Images and Potential Biases in\n  Generative Models", "abstract": "  This paper examines potential biases and inconsistencies in emotional\nevocation of images produced by generative artificial intelligence (AI) models\nand their potential bias toward negative emotions. In particular, we assess\nthis bias by comparing the emotions evoked by an AI-produced image to the\nemotions evoked by prompts used to create those images. As a first step, the\nstudy evaluates three approaches for identifying emotions in images --\ntraditional supervised learning, zero-shot learning with vision-language\nmodels, and cross-modal auto-captioning -- using EmoSet, a large dataset of\nimage-emotion annotations that categorizes images across eight emotional types.\nResults show fine-tuned models, particularly Google's Vision Transformer (ViT),\nsignificantly outperform zero-shot and caption-based methods in recognizing\nemotions in images. For a cross-modality comparison, we then analyze the\ndifferences between emotions in text prompts -- via existing text-based\nemotion-recognition models -- and the emotions evoked in the resulting images.\nFindings indicate that AI-generated images frequently lean toward negative\nemotional content, regardless of the original prompt. This emotional skew in\ngenerative models could amplify negative affective content in digital spaces,\nperpetuating its prevalence and impact. The study advocates for a\nmultidisciplinary approach to better align AI emotion recognition with\npsychological insights and address potential biases in generative AI outputs\nacross digital media.\n", "date": "2024-11-08", "categories": []}
{"id": "2412.10460", "title": "Enriching Multimodal Sentiment Analysis through Textual Emotional\n  Descriptions of Visual-Audio Content", "abstract": "  Multimodal Sentiment Analysis (MSA) stands as a critical research frontier,\nseeking to comprehensively unravel human emotions by amalgamating text, audio,\nand visual data. Yet, discerning subtle emotional nuances within audio and\nvideo expressions poses a formidable challenge, particularly when emotional\npolarities across various segments appear similar. In this paper, our objective\nis to spotlight emotion-relevant attributes of audio and visual modalities to\nfacilitate multimodal fusion in the context of nuanced emotional shifts in\nvisual-audio scenarios. To this end, we introduce DEVA, a progressive fusion\nframework founded on textual sentiment descriptions aimed at accentuating\nemotional features of visual-audio content. DEVA employs an Emotional\nDescription Generator (EDG) to transmute raw audio and visual data into\ntextualized sentiment descriptions, thereby amplifying their emotional\ncharacteristics. These descriptions are then integrated with the source data to\nyield richer, enhanced features. Furthermore, DEVA incorporates the Text-guided\nProgressive Fusion Module (TPF), leveraging varying levels of text as a core\nmodality guide. This module progressively fuses visual-audio minor modalities\nto alleviate disparities between text and visual-audio modalities. Experimental\nresults on widely used sentiment analysis benchmark datasets, including MOSI,\nMOSEI, and CH-SIMS, underscore significant enhancements compared to\nstate-of-the-art models. Moreover, fine-grained emotion experiments corroborate\nthe robust sensitivity of DEVA to subtle emotional variations.\n", "date": "2024-12-12", "categories": []}
{"id": "2412.10469", "title": "Comparative Analysis of Mel-Frequency Cepstral Coefficients and Wavelet\n  Based Audio Signal Processing for Emotion Detection and Mental Health\n  Assessment in Spoken Speech", "abstract": "  The intersection of technology and mental health has spurred innovative\napproaches to assessing emotional well-being, particularly through\ncomputational techniques applied to audio data analysis. This study explores\nthe application of Convolutional Neural Network (CNN) and Long Short-Term\nMemory (LSTM) models on wavelet extracted features and Mel-frequency Cepstral\nCoefficients (MFCCs) for emotion detection from spoken speech. Data\naugmentation techniques, feature extraction, normalization, and model training\nwere conducted to evaluate the models' performance in classifying emotional\nstates. Results indicate that the CNN model achieved a higher accuracy of 61%\ncompared to the LSTM model's accuracy of 56%. Both models demonstrated better\nperformance in predicting specific emotions such as surprise and anger,\nleveraging distinct audio features like pitch and speed variations.\nRecommendations include further exploration of advanced data augmentation\ntechniques, combined feature extraction methods, and the integration of\nlinguistic analysis with speech characteristics for improved accuracy in mental\nhealth diagnostics. Collaboration for standardized dataset collection and\nsharing is recommended to foster advancements in affective computing and mental\nhealth care interventions.\n", "date": "2024-12-12", "categories": []}
{"id": "2412.09784", "title": "Semi-IIN: Semi-supervised Intra-inter modal Interaction Learning Network\n  for Multimodal Sentiment Analysis", "abstract": "  Despite multimodal sentiment analysis being a fertile research ground that\nmerits further investigation, current approaches take up high annotation cost\nand suffer from label ambiguity, non-amicable to high-quality labeled data\nacquisition. Furthermore, choosing the right interactions is essential because\nthe significance of intra- or inter-modal interactions can differ among various\nsamples. To this end, we propose Semi-IIN, a Semi-supervised Intra-inter modal\nInteraction learning Network for multimodal sentiment analysis. Semi-IIN\nintegrates masked attention and gating mechanisms, enabling effective dynamic\nselection after independently capturing intra- and inter-modal interactive\ninformation. Combined with the self-training approach, Semi-IIN fully utilizes\nthe knowledge learned from unlabeled data. Experimental results on two public\ndatasets, MOSI and MOSEI, demonstrate the effectiveness of Semi-IIN,\nestablishing a new state-of-the-art on several metrics. Code is available at\nhttps://github.com/flow-ljh/Semi-IIN.\n", "date": "2024-12-13", "categories": []}
{"id": "2412.10011", "title": "Enhanced Speech Emotion Recognition with Efficient Channel Attention\n  Guided Deep CNN-BiLSTM Framework", "abstract": "  Speech emotion recognition (SER) is crucial for enhancing affective computing\nand enriching the domain of human-computer interaction. However, the main\nchallenge in SER lies in selecting relevant feature representations from speech\nsignals with lower computational costs. In this paper, we propose a lightweight\nSER architecture that integrates attention-based local feature blocks (ALFBs)\nto capture high-level relevant feature vectors from speech signals. We also\nincorporate a global feature block (GFB) technique to capture sequential,\nglobal information and long-term dependencies in speech signals. By aggregating\nattention-based local and global contextual feature vectors, our model\neffectively captures the internal correlation between salient features that\nreflect complex human emotional cues. To evaluate our approach, we extracted\nfour types of spectral features from speech audio samples: mel-frequency\ncepstral coefficients, mel-spectrogram, root mean square value, and\nzero-crossing rate. Through a 5-fold cross-validation strategy, we tested the\nproposed method on five multi-lingual standard benchmark datasets: TESS,\nRAVDESS, BanglaSER, SUBESCO, and Emo-DB, and obtained a mean accuracy of\n99.65%, 94.88%, 98.12%, 97.94%, and 97.19% respectively. The results indicate\nthat our model achieves state-of-the-art (SOTA) performance compared to most\nexisting methods.\n", "date": "2024-12-13", "categories": []}
{"id": "2412.08683", "title": "Emotional Vietnamese Speech-Based Depression Diagnosis Using Dynamic\n  Attention Mechanism", "abstract": "  Major depressive disorder is a prevalent and serious mental health condition\nthat negatively impacts your emotions, thoughts, actions, and overall\nperception of the world. It is complicated to determine whether a person is\ndepressed due to the symptoms of depression not apparent. However, their voice\ncan be one of the factor from which we can acknowledge signs of depression.\nPeople who are depressed express discomfort, sadness and they may speak slowly,\ntrembly, and lose emotion in their voices. In this study, we proposed the\nDynamic Convolutional Block Attention Module (Dynamic-CBAM) to utilized with in\nan Attention-GRU Network to classify the emotions by analyzing the audio signal\nof humans. Based on the results, we can diagnose which patients are depressed\nor prone to depression then so that treatment and prevention can be started as\nsoon as possible. The research delves into the intricate computational steps\ninvolved in implementing a Attention-GRU deep learning architecture. Through\nexperimentation, the model has achieved an impressive recognition with\nUnweighted Accuracy (UA) rate of 0.87 and 0.86 Weighted Accuracy (WA) rate and\nF1 rate of 0.87 in the VNEMOS dataset. Training code is released in\nhttps://github.com/fiyud/Emotional-Vietnamese-Speech-Based-Depression-Diagnosis-Using-Dynamic-Attention-Mechanism\n", "date": "2024-12-11", "categories": []}
{"id": "2412.09317", "title": "Multimodal Sentiment Analysis based on Video and Audio Inputs", "abstract": "  Despite the abundance of current researches working on the sentiment analysis\nfrom videos and audios, finding the best model that gives the highest accuracy\nrate is still considered a challenge for researchers in this field. The main\nobjective of this paper is to prove the usability of emotion recognition models\nthat take video and audio inputs. The datasets used to train the models are the\nCREMA-D dataset for audio and the RAVDESS dataset for video. The fine-tuned\nmodels that been used are: Facebook/wav2vec2-large for audio and the\nGoogle/vivit-b-16x2-kinetics400 for video. The avarage of the probabilities for\neach emotion generated by the two previous models is utilized in the decision\nmaking framework. After disparity in the results, if one of the models gets\nmuch higher accuracy, another test framework is created. The methods used are\nthe Weighted Average method, the Confidence Level Threshold method, the Dynamic\nWeighting Based on Confidence method, and the Rule-Based Logic method. This\nlimited approach gives encouraging results that make future research into these\nmethods viable.\n", "date": "2024-12-12", "categories": []}
{"id": "2412.07906", "title": "Rethinking Emotion Annotations in the Era of Large Language Models", "abstract": "  Modern affective computing systems rely heavily on datasets with\nhuman-annotated emotion labels, for training and evaluation. However, human\nannotations are expensive to obtain, sensitive to study design, and difficult\nto quality control, because of the subjective nature of emotions. Meanwhile,\nLarge Language Models (LLMs) have shown remarkable performance on many Natural\nLanguage Understanding tasks, emerging as a promising tool for text annotation.\nIn this work, we analyze the complexities of emotion annotation in the context\nof LLMs, focusing on GPT-4 as a leading model. In our experiments, GPT-4\nachieves high ratings in a human evaluation study, painting a more positive\npicture than previous work, in which human labels served as the only ground\ntruth. On the other hand, we observe differences between human and GPT-4\nemotion perception, underscoring the importance of human input in annotation\nstudies. To harness GPT-4's strength while preserving human perspective, we\nexplore two ways of integrating GPT-4 into emotion annotation pipelines,\nshowing its potential to flag low-quality labels, reduce the workload of human\nannotators, and improve downstream model learning performance and efficiency.\nTogether, our findings highlight opportunities for new emotion labeling\npractices and suggest the use of LLMs as a promising tool to aid human\nannotation.\n", "date": "2024-12-10", "categories": []}
{"id": "2412.07116", "title": "A Review of Human Emotion Synthesis Based on Generative Technology", "abstract": "  Human emotion synthesis is a crucial aspect of affective computing. It\ninvolves using computational methods to mimic and convey human emotions through\nvarious modalities, with the goal of enabling more natural and effective\nhuman-computer interactions. Recent advancements in generative models, such as\nAutoencoders, Generative Adversarial Networks, Diffusion Models, Large Language\nModels, and Sequence-to-Sequence Models, have significantly contributed to the\ndevelopment of this field. However, there is a notable lack of comprehensive\nreviews in this field. To address this problem, this paper aims to address this\ngap by providing a thorough and systematic overview of recent advancements in\nhuman emotion synthesis based on generative models. Specifically, this review\nwill first present the review methodology, the emotion models involved, the\nmathematical principles of generative models, and the datasets used. Then, the\nreview covers the application of different generative models to emotion\nsynthesis based on a variety of modalities, including facial images, speech,\nand text. It also examines mainstream evaluation metrics. Additionally, the\nreview presents some major findings and suggests future research directions,\nproviding a comprehensive understanding of the role of generative technology in\nthe nuanced domain of emotion synthesis.\n", "date": "2024-12-10", "categories": []}
{"id": "2412.07292", "title": "Multimodal Sentiment Analysis Based on Causal Reasoning", "abstract": "  With the rapid development of multimedia, the shift from unimodal textual\nsentiment analysis to multimodal image-text sentiment analysis has obtained\nacademic and industrial attention in recent years. However, multimodal\nsentiment analysis is affected by unimodal data bias, e.g., text sentiment is\nmisleading due to explicit sentiment semantic, leading to low accuracy in the\nfinal sentiment classification. In this paper, we propose a novel\nCounterFactual Multimodal Sentiment Analysis framework (CF-MSA) using causal\ncounterfactual inference to construct multimodal sentiment causal inference.\nCF-MSA mitigates the direct effect from unimodal bias and ensures heterogeneity\nacross modalities by differentiating the treatment variables between\nmodalities. In addition, considering the information complementarity and bias\ndifferences between modalities, we propose a new optimisation objective to\neffectively integrate different modalities and reduce the inherent bias from\neach modality. Experimental results on two public datasets, MVSA-Single and\nMVSA-Multiple, demonstrate that the proposed CF-MSA has superior debiasing\ncapability and achieves new state-of-the-art performances. We will release the\ncode and datasets to facilitate future research.\n", "date": "2024-12-10", "categories": []}
{"id": "2208.11346", "title": "ICANet: A Method of Short Video Emotion Recognition Driven by Multimodal\n  Data", "abstract": "  With the fast development of artificial intelligence and short videos,\nemotion recognition in short videos has become one of the most important\nresearch topics in human-computer interaction. At present, most emotion\nrecognition methods still stay in a single modality. However, in daily life,\nhuman beings will usually disguise their real emotions, which leads to the\nproblem that the accuracy of single modal emotion recognition is relatively\nterrible. Moreover, it is not easy to distinguish similar emotions. Therefore,\nwe propose a new approach denoted as ICANet to achieve multimodal short video\nemotion recognition by employing three different modalities of audio, video and\noptical flow, making up for the lack of a single modality and then improving\nthe accuracy of emotion recognition in short videos. ICANet has a better\naccuracy of 80.77% on the IEMOCAP benchmark, exceeding the SOTA methods by\n15.89%.\n", "date": "2022-08-24", "categories": []}
{"id": "2305.01486", "title": "ARBEx: Attentive Feature Extraction with Reliability Balancing for\n  Robust Facial Expression Learning", "abstract": "  In this paper, we introduce a framework ARBEx, a novel attentive feature\nextraction framework driven by Vision Transformer with reliability balancing to\ncope against poor class distributions, bias, and uncertainty in the facial\nexpression learning (FEL) task. We reinforce several data pre-processing and\nrefinement methods along with a window-based cross-attention ViT to squeeze the\nbest of the data. We also employ learnable anchor points in the embedding space\nwith label distributions and multi-head self-attention mechanism to optimize\nperformance against weak predictions with reliability balancing, which is a\nstrategy that leverages anchor points, attention scores, and confidence values\nto enhance the resilience of label predictions. To ensure correct label\nclassification and improve the models' discriminative power, we introduce\nanchor loss, which encourages large margins between anchor points.\nAdditionally, the multi-head self-attention mechanism, which is also trainable,\nplays an integral role in identifying accurate labels. This approach provides\ncritical elements for improving the reliability of predictions and has a\nsubstantial positive effect on final prediction capabilities. Our adaptive\nmodel can be integrated with any deep neural network to forestall challenges in\nvarious recognition tasks. Our strategy outperforms current state-of-the-art\nmethodologies, according to extensive experiments conducted in a variety of\ncontexts.\n", "date": "2023-05-02", "categories": []}
{"id": "2311.17335", "title": "Towards Emotion Analysis in Short-form Videos: A Large-Scale Dataset and\n  Baseline", "abstract": "  Nowadays, short-form videos (SVs) are essential to web information\nacquisition and sharing in our daily life. The prevailing use of SVs to spread\nemotions leads to the necessity of conducting video emotion analysis (VEA)\ntowards SVs. Considering the lack of SVs emotion data, we introduce a\nlarge-scale dataset named eMotions, comprising 27,996 videos. Meanwhile, we\nalleviate the impact of subjectivities on labeling quality by emphasizing\nbetter personnel allocations and multi-stage annotations. In addition, we\nprovide the category-balanced and test-oriented variants through targeted data\nsampling. Some commonly used videos, such as facial expressions, have been well\nstudied. However, it is still challenging to analysis the emotions in SVs.\nSince the broader content diversity brings more distinct semantic gaps and\ndifficulties in learning emotion-related features, and there exists local\nbiases and collective information gaps caused by the emotion inconsistence\nunder the prevalently audio-visual co-expressions. To tackle these challenges,\nwe present an end-to-end audio-visual baseline AV-CANet which employs the video\ntransformer to better learn semantically relevant representations. We further\ndesign the Local-Global Fusion Module to progressively capture the correlations\nof audio-visual features. The EP-CE Loss is then introduced to guide model\noptimization. Extensive experimental results on seven datasets demonstrate the\neffectiveness of AV-CANet, while providing broad insights for future works.\nBesides, we investigate the key components of AV-CANet by ablation studies.\nDatasets and code will be fully open soon.\n", "date": "2023-11-29", "categories": []}
{"id": "2412.05558", "title": "WavFusion: Towards wav2vec 2.0 Multimodal Speech Emotion Recognition", "abstract": "  Speech emotion recognition (SER) remains a challenging yet crucial task due\nto the inherent complexity and diversity of human emotions. To address this\nproblem, researchers attempt to fuse information from other modalities via\nmultimodal learning. However, existing multimodal fusion techniques often\noverlook the intricacies of cross-modal interactions, resulting in suboptimal\nfeature representations. In this paper, we propose WavFusion, a multimodal\nspeech emotion recognition framework that addresses critical research problems\nin effective multimodal fusion, heterogeneity among modalities, and\ndiscriminative representation learning. By leveraging a gated cross-modal\nattention mechanism and multimodal homogeneous feature discrepancy learning,\nWavFusion demonstrates improved performance over existing state-of-the-art\nmethods on benchmark datasets. Our work highlights the importance of capturing\nnuanced cross-modal interactions and learning discriminative representations\nfor accurate multimodal SER. Experimental results on two benchmark datasets\n(IEMOCAP and MELD) demonstrate that WavFusion succeeds over the\nstate-of-the-art strategies on emotion recognition.\n", "date": "2024-12-07", "categories": []}
{"id": "2412.03625", "title": "Multimodal Sentiment Analysis Based on BERT and ResNet", "abstract": "  With the rapid development of the Internet and social media, multi-modal data\n(text and image) is increasingly important in sentiment analysis tasks.\nHowever, the existing methods are difficult to effectively fuse text and image\nfeatures, which limits the accuracy of analysis. To solve this problem, a\nmultimodal sentiment analysis framework combining BERT and ResNet was proposed.\nBERT has shown strong text representation ability in natural language\nprocessing, and ResNet has excellent image feature extraction performance in\nthe field of computer vision. Firstly, BERT is used to extract the text feature\nvector, and ResNet is used to extract the image feature representation. Then, a\nvariety of feature fusion strategies are explored, and finally the fusion model\nbased on attention mechanism is selected to make full use of the complementary\ninformation between text and image. Experimental results on the public dataset\nMAVA-single show that compared with the single-modal models that only use BERT\nor ResNet, the proposed multi-modal model improves the accuracy and F1 score,\nreaching the best accuracy of 74.5%. This study not only provides new ideas and\nmethods for multimodal sentiment analysis, but also demonstrates the\napplication potential of BERT and ResNet in cross-domain fusion. In the future,\nmore advanced feature fusion techniques and optimization strategies will be\nexplored to further improve the accuracy and generalization ability of\nmultimodal sentiment analysis.\n", "date": "2024-12-04", "categories": []}
{"id": "2403.02167", "title": "EMOVOME: A Dataset for Emotion Recognition in Spontaneous Real-Life\n  Speech", "abstract": "  Spontaneous datasets for Speech Emotion Recognition (SER) are scarce and\nfrequently derived from laboratory environments or staged scenarios, such as TV\nshows, limiting their application in real-world contexts. We developed and\npublicly released the Emotional Voice Messages (EMOVOME) dataset, including 999\nvoice messages from real conversations of 100 Spanish speakers on a messaging\napp, labeled in continuous and discrete emotions by expert and non-expert\nannotators. We evaluated speaker-independent SER models using acoustic features\nas baseline and transformer-based models. We compared the results with\nreference datasets including acted and elicited speech, and analyzed the\ninfluence of annotators and gender fairness. The pre-trained\nUniSpeech-SAT-Large model achieved the highest results, 61.64% and 55.57%\nUnweighted Accuracy (UA) for 3-class valence and arousal prediction\nrespectively on EMOVOME, a 10% improvement over baseline models. For the\nemotion categories, 42.58% UA was obtained. EMOVOME performed lower than the\nacted RAVDESS dataset. The elicited IEMOCAP dataset also outperformed EMOVOME\nin predicting emotion categories, while similar results were obtained in\nvalence and arousal. EMOVOME outcomes varied with annotator labels, showing\nbetter results and fairness when combining expert and non-expert annotations.\nThis study highlights the gap between controlled and real-life scenarios,\nsupporting further advancements in recognizing genuine emotions.\n", "date": "2024-03-04", "categories": []}
{"id": "2408.14809", "title": "GSIFN: A Graph-Structured and Interlaced-Masked Multimodal\n  Transformer-based Fusion Network for Multimodal Sentiment Analysis", "abstract": "  Multimodal Sentiment Analysis (MSA) leverages multiple data modals to analyze\nhuman sentiment. Existing MSA models generally employ cutting-edge multimodal\nfusion and representation learning-based methods to promote MSA capability.\nHowever, there are two key challenges: (i) in existing multimodal fusion\nmethods, the decoupling of modal combinations and tremendous parameter\nredundancy, lead to insufficient fusion performance and efficiency; (ii) a\nchallenging trade-off exists between representation capability and\ncomputational overhead in unimodal feature extractors and encoders. Our\nproposed GSIFN incorporates two main components to solve these problems: (i) a\ngraph-structured and interlaced-masked multimodal Transformer. It adopts the\nInterlaced Mask mechanism to construct robust multimodal graph embedding,\nachieve all-modal-in-one Transformer-based fusion, and greatly reduce the\ncomputational overhead; (ii) a self-supervised learning framework with low\ncomputational overhead and high performance, which utilizes a parallelized LSTM\nwith matrix memory to enhance non-verbal modal features for unimodal label\ngeneration. Evaluated on the MSA datasets CMU-MOSI, CMU-MOSEI, and CH-SIMS,\nGSIFN demonstrates superior performance with significantly lower computational\noverhead compared with previous state-of-the-art models.\n", "date": "2024-08-27", "categories": []}
{"id": "2412.01860", "title": "Pairwise Discernment of AffectNet Expressions with ArcFace", "abstract": "  This study takes a preliminary step toward teaching computers to recognize\nhuman emotions through Facial Emotion Recognition (FER). Transfer learning is\napplied using ResNeXt, EfficientNet models, and an ArcFace model originally\ntrained on the facial verification task, leveraging the AffectNet database, a\ncollection of human face images annotated with corresponding emotions. The\nfindings highlight the value of congruent domain transfer learning, the\nchallenges posed by imbalanced datasets in learning facial emotion patterns,\nand the effectiveness of pairwise learning in addressing class imbalances to\nenhance model performance on the FER task.\n", "date": "2024-12-01", "categories": []}
{"id": "2412.02283", "title": "VR Based Emotion Recognition Using Deep Multimodal Fusion With\n  Biosignals Across Multiple Anatomical Domains", "abstract": "  Emotion recognition is significantly enhanced by integrating multimodal\nbiosignals and IMU data from multiple domains. In this paper, we introduce a\nnovel multi-scale attention-based LSTM architecture, combined with\nSqueeze-and-Excitation (SE) blocks, by leveraging multi-domain signals from the\nhead (Meta Quest Pro VR headset), trunk (Equivital Vest), and peripheral\n(Empatica Embrace Plus) during affect elicitation via visual stimuli. Signals\nfrom 23 participants were recorded, alongside self-assessed valence and arousal\nratings after each stimulus. LSTM layers extract features from each modality,\nwhile multi-scale attention captures fine-grained temporal dependencies, and SE\nblocks recalibrate feature importance prior to classification. We assess which\ndomain's signals carry the most distinctive emotional information during VR\nexperiences, identifying key biosignals contributing to emotion detection. The\nproposed architecture, validated in a user study, demonstrates superior\nperformance in classifying valance and arousal level (high / low), showcasing\nthe efficacy of multi-domain and multi-modal fusion with biosignals (e.g.,\nTEMP, EDA) with IMU data (e.g., accelerometer) for emotion recognition in\nreal-world applications.\n", "date": "2024-12-03", "categories": []}
{"id": "2412.00429", "title": "Learner Attentiveness and Engagement Analysis in Online Education Using\n  Computer Vision", "abstract": "  In recent times, online education and the usage of video-conferencing\nplatforms have experienced massive growth. Due to the limited scope of a\nvirtual classroom, it may become difficult for instructors to analyze learners'\nattention and comprehension in real time while teaching. In the digital mode of\neducation, it would be beneficial for instructors to have an automated feedback\nmechanism to be informed regarding learners' attentiveness at any given time.\nThis research presents a novel computer vision-based approach to analyze and\nquantify learners' attentiveness, engagement, and other affective states within\nonline learning scenarios. This work presents the development of a multiclass\nmultioutput classification method using convolutional neural networks on a\npublicly available dataset - DAiSEE. A machine learning-based algorithm is\ndeveloped on top of the classification model that outputs a comprehensive\nattentiveness index of the learners. Furthermore, an end-to-end pipeline is\nproposed through which learners' live video feed is processed, providing\ndetailed attentiveness analytics of the learners to the instructors. By\ncomparing the experimental outcomes of the proposed method against those of\nprevious methods, it is demonstrated that the proposed method exhibits better\nattentiveness detection than state-of-the-art methods. The proposed system is a\ncomprehensive, practical, and real-time solution that is deployable and easy to\nuse. The experimental results also demonstrate the system's efficiency in\ngauging learners' attentiveness.\n", "date": "2024-11-30", "categories": []}
{"id": "2412.01171", "title": "Cross-Task Inconsistency Based Active Learning (CTIAL) for Emotion\n  Recognition", "abstract": "  Emotion recognition is a critical component of affective computing. Training\naccurate machine learning models for emotion recognition typically requires a\nlarge amount of labeled data. Due to the subtleness and complexity of emotions,\nmultiple evaluators are usually needed for each affective sample to obtain its\nground-truth label, which is expensive. To save the labeling cost, this paper\nproposes an inconsistency-based active learning approach for cross-task\ntransfer between emotion classification and estimation. Affective norms are\nutilized as prior knowledge to connect the label spaces of categorical and\ndimensional emotions. Then, the prediction inconsistency on the two tasks for\nthe unlabeled samples is used to guide sample selection in active learning for\nthe target task. Experiments on within-corpus and cross-corpus transfers\ndemonstrated that cross-task inconsistency could be a very valuable metric in\nactive learning. To our knowledge, this is the first work that utilizes prior\nknowledge on affective norms and data in a different task to facilitate active\nlearning for a new task, even the two tasks are from different datasets.\n", "date": "2024-12-02", "categories": []}
{"id": "2411.09189", "title": "Improvement and Implementation of a Speech Emotion Recognition Model\n  Based on Dual-Layer LSTM", "abstract": "  This paper builds upon an existing speech emotion recognition model by adding\nan additional LSTM layer to improve the accuracy and processing efficiency of\nemotion recognition from audio data. By capturing the long-term dependencies\nwithin audio sequences through a dual-layer LSTM network, the model can\nrecognize and classify complex emotional patterns more accurately. Experiments\nconducted on the RAVDESS dataset validated this approach, showing that the\nmodified dual layer LSTM model improves accuracy by 2% compared to the\nsingle-layer LSTM while significantly reducing recognition latency, thereby\nenhancing real-time performance. These results indicate that the dual-layer\nLSTM architecture is highly suitable for handling emotional features with\nlong-term dependencies, providing a viable optimization for speech emotion\nrecognition systems. This research provides a reference for practical\napplications in fields like intelligent customer service, sentiment analysis\nand human-computer interaction.\n", "date": "2024-11-14", "categories": []}
{"id": "2411.19803", "title": "A Cross-Corpus Speech Emotion Recognition Method Based on Supervised\n  Contrastive Learning", "abstract": "  Research on Speech Emotion Recognition (SER) often faces challenges such as\nthe lack of large-scale public datasets and limited generalization capability\nwhen dealing with data from different distributions. To solve this problem,\nthis paper proposes a cross-corpus speech emotion recognition method based on\nsupervised contrast learning. The method employs a two-stage fine-tuning\nprocess: first, the self-supervised speech representation model is fine-tuned\nusing supervised contrastive learning on multiple speech emotion datasets;\nthen, the classifier is fine-tuned on the target dataset. The experimental\nresults show that the WavLM-based model achieved unweighted accuracy (UA) of\n77.41% on the IEMOCAP dataset and 96.49% on the CASIA dataset, outperforming\nthe state-of-the-art results on the two datasets.\n", "date": "2024-11-25", "categories": []}
{"id": "2411.19822", "title": "SDR-GNN: Spectral Domain Reconstruction Graph Neural Network for\n  Incomplete Multimodal Learning in Conversational Emotion Recognition", "abstract": "  Multimodal Emotion Recognition in Conversations (MERC) aims to classify\nutterance emotions using textual, auditory, and visual modal features. Most\nexisting MERC methods assume each utterance has complete modalities,\noverlooking the common issue of incomplete modalities in real-world scenarios.\nRecently, graph neural networks (GNNs) have achieved notable results in\nIncomplete Multimodal Emotion Recognition in Conversations (IMERC). However,\ntraditional GNNs focus on binary relationships between nodes, limiting their\nability to capture more complex, higher-order information. Moreover, repeated\nmessage passing can cause over-smoothing, reducing their capacity to preserve\nessential high-frequency details. To address these issues, we propose a\nSpectral Domain Reconstruction Graph Neural Network (SDR-GNN) for incomplete\nmultimodal learning in conversational emotion recognition. SDR-GNN constructs\nan utterance semantic interaction graph using a sliding window based on both\nspeaker and context relationships to model emotional dependencies. To capture\nhigher-order and high-frequency information, SDR-GNN utilizes weighted\nrelationship aggregation, ensuring consistent semantic feature extraction\nacross utterances. Additionally, it performs multi-frequency aggregation in the\nspectral domain, enabling efficient recovery of incomplete modalities by\nextracting both high- and low-frequency information. Finally, multi-head\nattention is applied to fuse and optimize features for emotion recognition.\nExtensive experiments on various real-world datasets demonstrate that our\napproach is effective in incomplete multimodal learning and outperforms current\nstate-of-the-art methods.\n", "date": "2024-11-29", "categories": []}
{"id": "2411.17715", "title": "Hybrid Quantum Deep Learning Model for Emotion Detection using raw EEG\n  Signal Analysis", "abstract": "  Applications in behavioural research, human-computer interaction, and mental\nhealth depend on the ability to recognize emotions. In order to improve the\naccuracy of emotion recognition using electroencephalography (EEG) data, this\nwork presents a hybrid quantum deep learning technique. Conventional EEG-based\nemotion recognition techniques are limited by noise and high-dimensional data\ncomplexity, which make feature extraction difficult. To tackle these issues,\nour method combines traditional deep learning classification with\nquantum-enhanced feature extraction. To identify important brain wave patterns,\nBandpass filtering and Welch method are used as preprocessing techniques on EEG\ndata. Intricate inter-band interactions that are essential for determining\nemotional states are captured by mapping frequency band power attributes\n(delta, theta, alpha, and beta) to quantum representations. Entanglement and\nrotation gates are used in a hybrid quantum circuit to maximize the model's\nsensitivity to EEG patterns associated with different emotions. Promising\nresults from evaluation on a test dataset indicate the model's potential for\naccurate emotion recognition. The model will be extended for real-time\napplications and multi-class categorization in future study, which could\nimprove EEG-based mental health screening instruments. This method offers a\npromising tool for applications in adaptive human-computer systems and mental\nhealth monitoring by showcasing the possibilities of fusing traditional deep\nlearning with quantum processing for reliable, scalable emotion recognition.\n", "date": "2024-11-19", "categories": []}
{"id": "2411.18162", "title": "SentiXRL: An advanced large language Model Framework for Multilingual\n  Fine-Grained Emotion Classification in Complex Text Environment", "abstract": "  With strong expressive capabilities in Large Language Models(LLMs),\ngenerative models effectively capture sentiment structures and deep semantics,\nhowever, challenges remain in fine-grained sentiment classification across\nmulti-lingual and complex contexts. To address this, we propose the Sentiment\nCross-Lingual Recognition and Logic Framework (SentiXRL), which incorporates\ntwo modules,an emotion retrieval enhancement module to improve sentiment\nclassification accuracy in complex contexts through historical dialogue and\nlogical reasoning,and a self-circulating analysis negotiation mechanism\n(SANM)to facilitates autonomous decision-making within a single model for\nclassification tasks.We have validated SentiXRL's superiority on multiple\nstandard datasets, outperforming existing models on CPED and CH-SIMS,and\nachieving overall better performance on MELD,Emorynlp and IEMOCAP. Notably, we\nunified labels across several fine-grained sentiment annotation datasets and\nconducted category confusion experiments, revealing challenges and impacts of\nclass imbalance in standard datasets.\n", "date": "2024-11-27", "categories": []}
{"id": "2308.13382", "title": "Prompting Visual-Language Models for Dynamic Facial Expression\n  Recognition", "abstract": "  This paper presents a novel visual-language model called DFER-CLIP, which is\nbased on the CLIP model and designed for in-the-wild Dynamic Facial Expression\nRecognition (DFER). Specifically, the proposed DFER-CLIP consists of a visual\npart and a textual part. For the visual part, based on the CLIP image encoder,\na temporal model consisting of several Transformer encoders is introduced for\nextracting temporal facial expression features, and the final feature embedding\nis obtained as a learnable \"class\" token. For the textual part, we use as\ninputs textual descriptions of the facial behaviour that is related to the\nclasses (facial expressions) that we are interested in recognising -- those\ndescriptions are generated using large language models, like ChatGPT. This, in\ncontrast to works that use only the class names and more accurately captures\nthe relationship between them. Alongside the textual description, we introduce\na learnable token which helps the model learn relevant context information for\neach expression during training. Extensive experiments demonstrate the\neffectiveness of the proposed method and show that our DFER-CLIP also achieves\nstate-of-the-art results compared with the current supervised DFER methods on\nthe DFEW, FERV39k, and MAFW benchmarks. Code is publicly available at\nhttps://github.com/zengqunzhao/DFER-CLIP.\n", "date": "2023-08-25", "categories": []}
{"id": "2411.17254", "title": "Semantic Data Augmentation for Long-tailed Facial Expression Recognition", "abstract": "  Facial Expression Recognition has a wide application prospect in social\nrobotics, health care, driver fatigue monitoring, and many other practical\nscenarios. Automatic recognition of facial expressions has been extensively\nstudied by the Computer Vision research society. But Facial Expression\nRecognition in real-world is still a challenging task, partially due to the\nlong-tailed distribution of the dataset. Many recent studies use data\naugmentation for Long-Tailed Recognition tasks. In this paper, we propose a\nnovel semantic augmentation method. By introducing randomness into the encoding\nof the source data in the latent space of VAE-GAN, new samples are generated.\nThen, for facial expression recognition in RAF-DB dataset, we use our\naugmentation method to balance the long-tailed distribution. Our method can be\nused in not only FER tasks, but also more diverse data-hungry scenarios.\n", "date": "2024-11-26", "categories": []}
{"id": "2409.10545", "title": "ResEmoteNet: Bridging Accuracy and Loss Reduction in Facial Emotion\n  Recognition", "abstract": "  The human face is a silent communicator, expressing emotions and thoughts\nthrough its facial expressions. With the advancements in computer vision in\nrecent years, facial emotion recognition technology has made significant\nstrides, enabling machines to decode the intricacies of facial cues. In this\nwork, we propose ResEmoteNet, a novel deep learning architecture for facial\nemotion recognition designed with the combination of Convolutional,\nSqueeze-Excitation (SE) and Residual Networks. The inclusion of SE block\nselectively focuses on the important features of the human face, enhances the\nfeature representation and suppresses the less relevant ones. This helps in\nreducing the loss and enhancing the overall model performance. We also\nintegrate the SE block with three residual blocks that help in learning more\ncomplex representation of the data through deeper layers. We evaluated\nResEmoteNet on four open-source databases: FER2013, RAF-DB, AffectNet-7 and\nExpW, achieving accuracies of 79.79%, 94.76%, 72.39% and 75.67% respectively.\nThe proposed network outperforms state-of-the-art models across all four\ndatabases. The source code for ResEmoteNet is available at\nhttps://github.com/ArnabKumarRoy02/ResEmoteNet.\n", "date": "2024-09-01", "categories": []}
{"id": "2306.01229", "title": "Exploring the Boundaries of Semi-Supervised Facial Expression\n  Recognition using In-Distribution, Out-of-Distribution, and Unconstrained\n  Data", "abstract": "  Deep learning-based methods have been the key driving force behind much of\nthe recent success of facial expression recognition (FER) systems. However, the\nneed for large amounts of labelled data remains a challenge. Semi-supervised\nlearning offers a way to overcome this limitation, allowing models to learn\nfrom a small amount of labelled data along with a large unlabelled dataset.\nWhile semi-supervised learning has shown promise in FER, most current methods\nfrom general computer vision literature have not been explored in the context\nof FER. In this work, we present a comprehensive study on 11 of the most recent\nsemi-supervised methods, in the context of FER, namely Pi-model, Pseudo-label,\nMean Teacher, VAT, UDA, MixMatch, ReMixMatch, FlexMatch, CoMatch, and CCSSL.\nOur investigation covers semi-supervised learning from in-distribution,\nout-of-distribution, unconstrained, and very small unlabelled data. Our\nevaluation includes five FER datasets plus one large face dataset for\nunconstrained learning. Our results demonstrate that FixMatch consistently\nachieves better performance on in-distribution unlabelled data, while\nReMixMatch stands out among all methods for out-of-distribution, unconstrained,\nand scarce unlabelled data scenarios. Another significant observation is that\nwith an equal number of labelled samples, semi-supervised learning delivers a\nconsiderable improvement over supervised learning, regardless of whether the\nunlabelled data is in-distribution, out-of-distribution, or unconstrained. We\nalso conduct sensitivity analyses on critical hyper-parameters for the two best\nmethods of each setting. To facilitate reproducibility and further development,\nwe make our code publicly available at: github.com/ShuvenduRoy/SSL_FER_OOD.\n", "date": "2023-06-02", "categories": []}
{"id": "2311.03478", "title": "Multi Loss-based Feature Fusion and Top Two Voting Ensemble Decision\n  Strategy for Facial Expression Recognition in the Wild", "abstract": "  Facial expression recognition (FER) in the wild is a challenging task\naffected by the image quality and has attracted broad interest in computer\nvision. There is no research using feature fusion and ensemble strategy for FER\nsimultaneously. Different from previous studies, this paper applies both\ninternal feature fusion for a single model and feature fusion among multiple\nnetworks, as well as the ensemble strategy. This paper proposes one novel\nsingle model named R18+FAML, as well as one ensemble model named\nR18+FAML-FGA-T2V to improve the performance of the FER in the wild. Based on\nthe structure of ResNet18 (R18), R18+FAML combines internal Feature fusion and\nthree Attention blocks using Multiple Loss functions (FAML) to improve the\ndiversity of the feature extraction. To improve the performance of R18+FAML, we\npropose a Feature fusion among networks based on the Genetic Algorithm (FGA),\nwhich can fuse the convolution kernels for feature extraction of multiple\nnetworks. On the basis of R18+FAML and FGA, we propose one ensemble strategy,\ni.e., the Top Two Voting (T2V) to support the classification of FER, which can\nconsider more classification information comprehensively. Combining the above\nstrategies, R18+FAML-FGA-T2V can focus on the main expression-aware areas.\nExtensive experiments demonstrate that our single model R18+FAML and the\nensemble model R18+FAML-FGA-T2V achieve the accuracies of $\\left( 90.32, 62.17,\n65.83 \\right)\\%$ and $\\left( 91.59, 63.27, 66.63 \\right)\\%$ on three\nchallenging unbalanced FER datasets RAF-DB, AffectNet-8 and AffectNet-7\nrespectively, both outperforming the state-of-the-art results.\n", "date": "2023-11-06", "categories": []}
{"id": "2408.01728", "title": "Survey on Emotion Recognition through Posture Detection and the\n  possibility of its application in Virtual Reality", "abstract": "  A survey is presented focused on using pose estimation techniques in\nEmotional recognition using various technologies normal cameras, and depth\ncameras for real-time, and the potential use of VR and inputs including images,\nvideos, and 3-dimensional poses described in vector space. We discussed 19\nresearch papers collected from selected journals and databases highlighting\ntheir methodology, classification algorithm, and the used datasets that relate\nto emotion recognition and pose estimation. A benchmark has been made according\nto their accuracy as it was the most common performance measurement metric\nused. We concluded that the multimodal Approaches overall made the best\naccuracy and then we mentioned futuristic concerns that can improve the\ndevelopment of this research topic.\n", "date": "2024-08-03", "categories": []}
{"id": "2411.05879", "title": "Smile upon the Face but Sadness in the Eyes: Emotion Recognition based\n  on Facial Expressions and Eye Behaviors", "abstract": "  Emotion Recognition (ER) is the process of identifying human emotions from\ngiven data. Currently, the field heavily relies on facial expression\nrecognition (FER) because facial expressions contain rich emotional cues.\nHowever, it is important to note that facial expressions may not always\nprecisely reflect genuine emotions and FER-based results may yield misleading\nER. To understand and bridge this gap between FER and ER, we introduce eye\nbehaviors as an important emotional cues for the creation of a new\nEye-behavior-aided Multimodal Emotion Recognition (EMER) dataset. Different\nfrom existing multimodal ER datasets, the EMER dataset employs a stimulus\nmaterial-induced spontaneous emotion generation method to integrate\nnon-invasive eye behavior data, like eye movements and eye fixation maps, with\nfacial videos, aiming to obtain natural and accurate human emotions. Notably,\nfor the first time, we provide annotations for both ER and FER in the EMER,\nenabling a comprehensive analysis to better illustrate the gap between both\ntasks. Furthermore, we specifically design a new EMERT architecture to\nconcurrently enhance performance in both ER and FER by efficiently identifying\nand bridging the emotion gap between the two.Specifically, our EMERT employs\nmodality-adversarial feature decoupling and multi-task Transformer to augment\nthe modeling of eye behaviors, thus providing an effective complement to facial\nexpressions. In the experiment, we introduce seven multimodal benchmark\nprotocols for a variety of comprehensive evaluations of the EMER dataset. The\nresults show that the EMERT outperforms other state-of-the-art multimodal\nmethods by a great margin, revealing the importance of modeling eye behaviors\nfor robust ER. To sum up, we provide a comprehensive analysis of the importance\nof eye behaviors in ER, advancing the study on addressing the gap between FER\nand ER for more robust ER performance.\n", "date": "2024-11-08", "categories": []}
{"id": "2403.08511", "title": "A Multimodal Fusion Network For Student Emotion Recognition Based on\n  Transformer and Tensor Product", "abstract": "  This paper introduces a new multi-modal model based on the Transformer\narchitecture and tensor product fusion strategy, combining BERT's text vectors\nand ViT's image vectors to classify students' psychological conditions, with an\naccuracy of 93.65%. The purpose of the study is to accurately analyze the\nmental health status of students from various data sources. This paper\ndiscusses modal fusion methods, including early, late and intermediate fusion,\nto overcome the challenges of integrating multi-modal information. Ablation\nstudies compare the performance of different models and fusion techniques,\nshowing that the proposed model outperforms existing methods such as CLIP and\nViLBERT in terms of accuracy and inference speed. Conclusions indicate that\nwhile this model has significant advantages in emotion recognition, its\npotential to incorporate other data modalities provides areas for future\nresearch.\n", "date": "2024-03-13", "categories": []}
{"id": "2411.10464", "title": "Detecting Student Disengagement in Online Classes Using Deep Learning: A\n  Review", "abstract": "  Student disengagement in online learning has become a critical challenge,\nparticularly post-pandemic. This review explores deep learning techniques used\nto detect disengagement, emphasizing computer vision and affective computing as\neffective approaches. We examine recent studies focusing on facial expressions,\neye movements, and posture to assess student attention, along with\nnon-face-based indicators like mouse activity. A systematic review of 38\nselected studies outlines the indicators, methods, and models employed in this\nfield, providing insights for future research on real-time engagement\nmonitoring in online classrooms\n", "date": "2024-11-04", "categories": []}
{"id": "2411.10863", "title": "Improvement in Facial Emotion Recognition using Synthetic Data Generated\n  by Diffusion Model", "abstract": "  Facial Emotion Recognition (FER) plays a crucial role in computer vision,\nwith significant applications in human-computer interaction, affective\ncomputing, and areas such as mental health monitoring and personalized learning\nenvironments. However, a major challenge in FER task is the class imbalance\ncommonly found in available datasets, which can hinder both model performance\nand generalization. In this paper, we tackle the issue of data imbalance by\nincorporating synthetic data augmentation and leveraging the ResEmoteNet model\nto enhance the overall performance on facial emotion recognition task. We\nemployed Stable Diffusion 2 and Stable Diffusion 3 Medium models to generate\nsynthetic facial emotion data, augmenting the training sets of the FER2013 and\nRAF-DB benchmark datasets. Training ResEmoteNet with these augmented datasets\nresulted in substantial performance improvements, achieving accuracies of\n96.47% on FER2013 and 99.23% on RAF-DB. These findings shows an absolute\nimprovement of 16.68% in FER2013, 4.47% in RAF-DB and highlight the efficacy of\nsynthetic data augmentation in strengthening FER models and underscore the\npotential of advanced generative models in FER research and applications. The\nsource code for ResEmoteNet is available at\nhttps://github.com/ArnabKumarRoy02/ResEmoteNet\n", "date": "2024-11-16", "categories": []}
{"id": "2411.11235", "title": "MEMO-Bench: A Multiple Benchmark for Text-to-Image and Multimodal Large\n  Language Models on Human Emotion Analysis", "abstract": "  Artificial Intelligence (AI) has demonstrated significant capabilities in\nvarious fields, and in areas such as human-computer interaction (HCI), embodied\nintelligence, and the design and animation of virtual digital humans, both\npractitioners and users are increasingly concerned with AI's ability to\nunderstand and express emotion. Consequently, the question of whether AI can\naccurately interpret human emotions remains a critical challenge. To date, two\nprimary classes of AI models have been involved in human emotion analysis:\ngenerative models and Multimodal Large Language Models (MLLMs). To assess the\nemotional capabilities of these two classes of models, this study introduces\nMEMO-Bench, a comprehensive benchmark consisting of 7,145 portraits, each\ndepicting one of six different emotions, generated by 12 Text-to-Image (T2I)\nmodels. Unlike previous works, MEMO-Bench provides a framework for evaluating\nboth T2I models and MLLMs in the context of sentiment analysis. Additionally, a\nprogressive evaluation approach is employed, moving from coarse-grained to\nfine-grained metrics, to offer a more detailed and comprehensive assessment of\nthe sentiment analysis capabilities of MLLMs. The experimental results\ndemonstrate that existing T2I models are more effective at generating positive\nemotions than negative ones. Meanwhile, although MLLMs show a certain degree of\neffectiveness in distinguishing and recognizing human emotions, they fall short\nof human-level accuracy, particularly in fine-grained emotion analysis. The\nMEMO-Bench will be made publicly available to support further research in this\narea.\n", "date": "2024-11-18", "categories": []}
{"id": "2306.17799", "title": "A Low-rank Matching Attention based Cross-modal Feature Fusion Method\n  for Conversational Emotion Recognition", "abstract": "  Conversational emotion recognition (CER) is an important research topic in\nhuman-computer interactions. {Although recent advancements in transformer-based\ncross-modal fusion methods have shown promise in CER tasks, they tend to\noverlook the crucial intra-modal and inter-modal emotional interaction or\nsuffer from high computational complexity. To address this, we introduce a\nnovel and lightweight cross-modal feature fusion method called Low-Rank\nMatching Attention Method (LMAM). LMAM effectively captures contextual\nemotional semantic information in conversations while mitigating the quadratic\ncomplexity issue caused by the self-attention mechanism. Specifically, by\nsetting a matching weight and calculating inter-modal features attention scores\nrow by row, LMAM requires only one-third of the parameters of self-attention\nmethods. We also employ the low-rank decomposition method on the weights to\nfurther reduce the number of parameters in LMAM. As a result, LMAM offers a\nlightweight model while avoiding overfitting problems caused by a large number\nof parameters. Moreover, LMAM is able to fully exploit the intra-modal\nemotional contextual information within each modality and integrates\ncomplementary emotional semantic information across modalities by computing and\nfusing similarities of intra-modal and inter-modal features simultaneously.\nExperimental results verify the superiority of LMAM compared with other popular\ncross-modal fusion methods on the premise of being more lightweight. Also, LMAM\ncan be embedded into any existing state-of-the-art CER methods in a\nplug-and-play manner, and can be applied to other multi-modal recognition\ntasks, e.g., session recommendation and humour detection, demonstrating its\nremarkable generalization ability.\n", "date": "2023-06-16", "categories": []}
{"id": "2411.10018", "title": "Once More, With Feeling: Measuring Emotion of Acting Performances in\n  Contemporary American Film", "abstract": "  Narrative film is a composition of writing, cinematography, editing, and\nperformance. While much computational work has focused on the writing or visual\nstyle in film, we conduct in this paper a computational exploration of acting\nperformance. Applying speech emotion recognition models and a variationist\nsociolinguistic analytical framework to a corpus of popular, contemporary\nAmerican film, we find narrative structure, diachronic shifts, and genre- and\ndialogue-based constraints located in spoken performances.\n", "date": "2024-11-15", "categories": []}
{"id": "2411.10060", "title": "CMATH: Cross-Modality Augmented Transformer with Hierarchical\n  Variational Distillation for Multimodal Emotion Recognition in Conversation", "abstract": "  Multimodal emotion recognition in conversation (MER) aims to accurately\nidentify emotions in conversational utterances by integrating multimodal\ninformation. Previous methods usually treat multimodal information as equal\nquality and employ symmetric architectures to conduct multimodal fusion.\nHowever, in reality, the quality of different modalities usually varies\nconsiderably, and utilizing a symmetric architecture is difficult to accurately\nrecognize conversational emotions when dealing with uneven modal information.\nFurthermore, fusing multi-modality information in a single granularity may fail\nto adequately integrate modal information, exacerbating the inaccuracy in\nemotion recognition. In this paper, we propose a novel Cross-Modality Augmented\nTransformer with Hierarchical Variational Distillation, called CMATH, which\nconsists of two major components, i.e., Multimodal Interaction Fusion and\nHierarchical Variational Distillation. The former is comprised of two\nsubmodules, including Modality Reconstruction and Cross-Modality Augmented\nTransformer (CMA-Transformer), where Modality Reconstruction focuses on\nobtaining high-quality compressed representation of each modality, and\nCMA-Transformer adopts an asymmetric fusion strategy which treats one modality\nas the central modality and takes others as auxiliary modalities. The latter\nfirst designs a variational fusion network to fuse the fine-grained\nrepresentations learned by CMA- Transformer into a coarse-grained\nrepresentations. Then, it introduces a hierarchical distillation framework to\nmaintain the consistency between modality representations with different\ngranularities. Experiments on the IEMOCAP and MELD datasets demonstrate that\nour proposed model outperforms previous state-of-the-art baselines.\nImplementation codes can be available at https://github.com/ cjw-MER/CMATH.\n", "date": "2024-11-15", "categories": []}
{"id": "2410.03979", "title": "Improving Arabic Multi-Label Emotion Classification using Stacked\n  Embeddings and Hybrid Loss Function", "abstract": "  In multi-label emotion classification, particularly for low-resource\nlanguages like Arabic, the challenges of class imbalance and label correlation\nhinder model performance, especially in accurately predicting minority\nemotions. To address these issues, this study proposes a novel approach that\ncombines stacked embeddings, meta-learning, and a hybrid loss function to\nenhance multi-label emotion classification for the Arabic language. The study\nextracts contextual embeddings from three fine-tuned language\nmodels-ArabicBERT, MarBERT, and AraBERT-which are then stacked to form enriched\nembeddings. A meta-learner is trained on these stacked embeddings, and the\nresulting concatenated representations are provided as input to a Bi-LSTM\nmodel, followed by a fully connected neural network for multi-label\nclassification. To further improve performance, a hybrid loss function is\nintroduced, incorporating class weighting, label correlation matrix, and\ncontrastive learning, effectively addressing class imbalances and improving the\nhandling of label correlations. Extensive experiments validate the proposed\nmodel's performance across key metrics such as Precision, Recall, F1-Score,\nJaccard Accuracy, and Hamming Loss. The class-wise performance analysis\ndemonstrates the hybrid loss function's ability to significantly reduce\ndisparities between majority and minority classes, resulting in a more balanced\nemotion classification. An ablation study highlights the contribution of each\ncomponent, showing the superiority of the model compared to baseline approaches\nand other loss functions. This study not only advances multi-label emotion\nclassification for Arabic but also presents a generalizable framework that can\nbe adapted to other languages and domains, providing a significant step forward\nin addressing the challenges of low-resource emotion classification tasks.\n", "date": "2024-10-04", "categories": []}
{"id": "2411.08347", "title": "A Chinese Multi-label Affective Computing Dataset Based on Social Media\n  Network Users", "abstract": "  Emotion and personality are central elements in understanding human\npsychological states. Emotions reflect an individual subjective experiences,\nwhile personality reveals relatively stable behavioral and cognitive patterns.\nExisting affective computing datasets often annotate emotion and personality\ntraits separately, lacking fine-grained labeling of micro-emotions and emotion\nintensity in both single-label and multi-label classifications. Chinese emotion\ndatasets are extremely scarce, and datasets capturing Chinese user personality\ntraits are even more limited. To address these gaps, this study collected data\nfrom the major social media platform Weibo, screening 11,338 valid users from\nover 50,000 individuals with diverse MBTI personality labels and acquiring\n566,900 posts along with the user MBTI personality tags. Using the EQN method,\nwe compiled a multi-label Chinese affective computing dataset that integrates\nthe same user's personality traits with six emotions and micro-emotions, each\nannotated with intensity levels. Validation results across multiple NLP\nclassification models demonstrate the dataset strong utility. This dataset is\ndesigned to advance machine recognition of complex human emotions and provide\ndata support for research in psychology, education, marketing, finance, and\npolitics.\n", "date": "2024-11-13", "categories": []}
{"id": "2407.12196", "title": "MASIVE: Open-Ended Affective State Identification in English and Spanish", "abstract": "  In the field of emotion analysis, much NLP research focuses on identifying a\nlimited number of discrete emotion categories, often applied across languages.\nThese basic sets, however, are rarely designed with textual data in mind, and\nculture, language, and dialect can influence how particular emotions are\ninterpreted. In this work, we broaden our scope to a practically unbounded set\nof \\textit{affective states}, which includes any terms that humans use to\ndescribe their experiences of feeling. We collect and publish MASIVE, a dataset\nof Reddit posts in English and Spanish containing over 1,000 unique affective\nstates each. We then define the new problem of \\textit{affective state\nidentification} for language generation models framed as a masked span\nprediction task. On this task, we find that smaller finetuned multilingual\nmodels outperform much larger LLMs, even on region-specific Spanish affective\nstates. Additionally, we show that pretraining on MASIVE improves model\nperformance on existing emotion benchmarks. Finally, through machine\ntranslation experiments, we find that native speaker-written data is vital to\ngood performance on this task.\n", "date": "2024-07-16", "categories": []}
{"id": "2411.07708", "title": "Emotion Classification of Children Expressions", "abstract": "  This paper proposes a process for a classification model for the facial\nexpressions. The proposed process would aid in specific categorisation of\nchildren's emotions from 2 emotions namely 'Happy' and 'Sad'. Since the\nexisting emotion recognition systems algorithms primarily train on adult faces,\nthe model developed is achieved by using advanced concepts of models with\nSqueeze-andExcitation blocks, Convolutional Block Attention modules, and robust\ndata augmentation. Stable Diffusion image synthesis was used for expanding and\ndiversifying the data set generating realistic and various training samples.\nThe model designed using Batch Normalisation, Dropout, and SE Attention\nmechanisms for the classification of children's emotions achieved an accuracy\nrate of 89\\% due to these methods improving the precision of emotion\nrecognition in children. The relative importance of this issue is raised in\nthis study with an emphasis on the call for a more specific model in emotion\ndetection systems for the young generation with specific direction on how the\nyoung people can be assisted to manage emotions while online.\n", "date": "2024-11-12", "categories": []}
{"id": "2411.06326", "title": "Emotion-Aware Interaction Design in Intelligent User Interface Using\n  Multi-Modal Deep Learning", "abstract": "  In an era where user interaction with technology is ubiquitous, the\nimportance of user interface (UI) design cannot be overstated. A well-designed\nUI not only enhances usability but also fosters more natural, intuitive, and\nemotionally engaging experiences, making technology more accessible and\nimpactful in everyday life. This research addresses this growing need by\nintroducing an advanced emotion recognition system to significantly improve the\nemotional responsiveness of UI. By integrating facial expressions, speech, and\ntextual data through a multi-branch Transformer model, the system interprets\ncomplex emotional cues in real-time, enabling UIs to interact more\nempathetically and effectively with users. Using the public MELD dataset for\nvalidation, our model demonstrates substantial improvements in emotion\nrecognition accuracy and F1 scores, outperforming traditional methods. These\nfindings underscore the critical role that sophisticated emotion recognition\nplays in the evolution of UIs, making technology more attuned to user needs and\nemotions. This study highlights how enhanced emotional intelligence in UIs is\nnot only about technical innovation but also about fostering deeper, more\nmeaningful connections between users and the digital world, ultimately shaping\nhow people interact with technology in their daily lives.\n", "date": "2024-11-10", "categories": []}
{"id": "2402.18424", "title": "Emotion Classification in Low and Moderate Resource Languages", "abstract": "  It is important to be able to analyze the emotional state of people around\nthe globe. There are 7100+ active languages spoken around the world and\nbuilding emotion classification for each language is labor intensive.\nParticularly for low-resource and endangered languages, building emotion\nclassification can be quite challenging. We present a cross-lingual emotion\nclassifier, where we train an emotion classifier with resource-rich languages\n(i.e. \\textit{English} in our work) and transfer the learning to low and\nmoderate resource languages. We compare and contrast two approaches of transfer\nlearning from a high-resource language to a low or moderate-resource language.\nOne approach projects the annotation from a high-resource language to low and\nmoderate-resource language in parallel corpora and the other one uses direct\ntransfer from high-resource language to the other languages. We show the\nefficacy of our approaches on 6 languages: Farsi, Arabic, Spanish, Ilocano,\nOdia, and Azerbaijani. Our results indicate that our approaches outperform\nrandom baselines and transfer emotions across languages successfully. For all\nlanguages, the direct cross-lingual transfer of emotion yields better results.\nWe also create annotated emotion-labeled resources for four languages: Farsi,\nAzerbaijani, Ilocano and Odia.\n", "date": "2024-02-28", "categories": []}
{"id": "2410.20334", "title": "Improving Speech-based Emotion Recognition with Contextual Utterance\n  Analysis and LLMs", "abstract": "  Speech Emotion Recognition (SER) focuses on identifying emotional states from\nspoken language. The 2024 IEEE SLT-GenSEC Challenge on Post Automatic Speech\nRecognition (ASR) Emotion Recognition tasks participants to explore the\ncapabilities of large language models (LLMs) for emotion recognition using only\ntext data. We propose a novel approach that first refines all available\ntranscriptions to ensure data reliability. We then segment each complete\nconversation into smaller dialogues and use these dialogues as context to\npredict the emotion of the target utterance within the dialogue. Finally, we\ninvestigated different context lengths and prompting techniques to improve\nprediction accuracy. Our best submission exceeded the baseline by 20% in\nunweighted accuracy, achieving the best performance in the challenge. All our\nexperiments' codes, prediction results, and log files are publicly available.\n", "date": "2024-10-27", "categories": []}
{"id": "2406.16473", "title": "D2SP: Dynamic Dual-Stage Purification Framework for Dual Noise\n  Mitigation in Vision-based Affective Recognition", "abstract": "  The contemporary state-of-the-art of Dynamic Facial Expression Recognition\n(DFER) technology facilitates remarkable progress by deriving emotional\nmappings of facial expressions from video content, underpinned by training on\nvoluminous datasets. Yet, the DFER datasets encompass a substantial volume of\nnoise data. Noise arises from low-quality captures that defy logical labeling,\nand instances that suffer from mislabeling due to annotation bias, engendering\ntwo principal types of uncertainty: the uncertainty regarding data usability\nand the uncertainty concerning label reliability. Addressing the two types of\nuncertainty, we have meticulously crafted a two-stage framework aiming at\n\\textbf{S}eeking \\textbf{C}ertain data \\textbf{I}n extensive \\textbf{U}ncertain\ndata (SCIU). This initiative aims to purge the DFER datasets of these\nuncertainties, thereby ensuring that only clean, verified data is employed in\ntraining processes. To mitigate the issue of low-quality samples, we introduce\nthe Coarse-Grained Pruning (CGP) stage, which assesses sample weights and\nprunes those deemed unusable due to their low weight. For samples with\nincorrect annotations, the Fine-Grained Correction (FGC) stage evaluates\nprediction stability to rectify mislabeled data. Moreover, SCIU is conceived as\na universally compatible, plug-and-play framework, tailored to integrate\nseamlessly with prevailing DFER methodologies. Rigorous experiments across\nprevalent DFER datasets and against numerous benchmark methods substantiates\nSCIU's capacity to markedly elevate performance metrics.\n", "date": "2024-06-24", "categories": []}
{"id": "2411.02964", "title": "Speaker Emotion Recognition: Leveraging Self-Supervised Models for\n  Feature Extraction Using Wav2Vec2 and HuBERT", "abstract": "  Speech is the most natural way of expressing ourselves as humans. Identifying\nemotion from speech is a nontrivial task due to the ambiguous definition of\nemotion itself. Speaker Emotion Recognition (SER) is essential for\nunderstanding human emotional behavior. The SER task is challenging due to the\nvariety of speakers, background noise, complexity of emotions, and speaking\nstyles. It has many applications in education, healthcare, customer service,\nand Human-Computer Interaction (HCI). Previously, conventional machine learning\nmethods such as SVM, HMM, and KNN have been used for the SER task. In recent\nyears, deep learning methods have become popular, with convolutional neural\nnetworks and recurrent neural networks being used for SER tasks. The input of\nthese methods is mostly spectrograms and hand-crafted features. In this work,\nwe study the use of self-supervised transformer-based models, Wav2Vec2 and\nHuBERT, to determine the emotion of speakers from their voice. The models\nautomatically extract features from raw audio signals, which are then used for\nthe classification task. The proposed solution is evaluated on reputable\ndatasets, including RAVDESS, SHEMO, SAVEE, AESDD, and Emo-DB. The results show\nthe effectiveness of the proposed method on different datasets. Moreover, the\nmodel has been used for real-world applications like call center conversations,\nand the results demonstrate that the model accurately predicts emotions.\n", "date": "2024-11-05", "categories": []}
{"id": "2411.03966", "title": "WorryWords: Norms of Anxiety Association for over 44k English Words", "abstract": "  Anxiety, the anticipatory unease about a potential negative outcome, is a\ncommon and beneficial human emotion. However, there is still much that is not\nknown, such as how anxiety relates to our body and how it manifests in\nlanguage. This is especially pertinent given the increasing impact of\nanxiety-related disorders. In this work, we introduce WorryWords, the first\nlarge-scale repository of manually derived word--anxiety associations for over\n44,450 English words. We show that the anxiety associations are highly\nreliable. We use WorryWords to study the relationship between anxiety and other\nemotion constructs, as well as the rate at which children acquire anxiety words\nwith age. Finally, we show that using WorryWords alone, one can accurately\ntrack the change of anxiety in streams of text. The lexicon enables a wide\nvariety of anxiety-related research in psychology, NLP, public health, and\nsocial sciences. WorryWords (and its translations to over 100 languages) is\nfreely available. http://saifmohammad.com/worrywords.html\n", "date": "2024-11-06", "categories": []}
{"id": "2406.07803", "title": "EmoSphere-TTS: Emotional Style and Intensity Modeling via Spherical\n  Emotion Vector for Controllable Emotional Text-to-Speech", "abstract": "  Despite rapid advances in the field of emotional text-to-speech (TTS), recent\nstudies primarily focus on mimicking the average style of a particular emotion.\nAs a result, the ability to manipulate speech emotion remains constrained to\nseveral predefined labels, compromising the ability to reflect the nuanced\nvariations of emotion. In this paper, we propose EmoSphere-TTS, which\nsynthesizes expressive emotional speech by using a spherical emotion vector to\ncontrol the emotional style and intensity of the synthetic speech. Without any\nhuman annotation, we use the arousal, valence, and dominance pseudo-labels to\nmodel the complex nature of emotion via a Cartesian-spherical transformation.\nFurthermore, we propose a dual conditional adversarial network to improve the\nquality of generated speech by reflecting the multi-aspect characteristics. The\nexperimental results demonstrate the model ability to control emotional style\nand intensity with high-quality expressive speech.\n", "date": "2024-06-12", "categories": []}
{"id": "2411.02793", "title": "Toward Robust Incomplete Multimodal Sentiment Analysis via Hierarchical\n  Representation Learning", "abstract": "  Multimodal Sentiment Analysis (MSA) is an important research area that aims\nto understand and recognize human sentiment through multiple modalities. The\ncomplementary information provided by multimodal fusion promotes better\nsentiment analysis compared to utilizing only a single modality. Nevertheless,\nin real-world applications, many unavoidable factors may lead to situations of\nuncertain modality missing, thus hindering the effectiveness of multimodal\nmodeling and degrading the model's performance. To this end, we propose a\nHierarchical Representation Learning Framework (HRLF) for the MSA task under\nuncertain missing modalities. Specifically, we propose a fine-grained\nrepresentation factorization module that sufficiently extracts valuable\nsentiment information by factorizing modality into sentiment-relevant and\nmodality-specific representations through crossmodal translation and sentiment\nsemantic reconstruction. Moreover, a hierarchical mutual information\nmaximization mechanism is introduced to incrementally maximize the mutual\ninformation between multi-scale representations to align and reconstruct the\nhigh-level semantics in the representations. Ultimately, we propose a\nhierarchical adversarial learning mechanism that further aligns and adapts the\nlatent distribution of sentiment-relevant representations to produce robust\njoint multimodal representations. Comprehensive experiments on three datasets\ndemonstrate that HRLF significantly improves MSA performance under uncertain\nmodality missing cases.\n", "date": "2024-11-05", "categories": []}
{"id": "2403.17327", "title": "Accuracy enhancement method for speech emotion recognition from\n  spectrogram using temporal frequency correlation and positional information\n  learning through knowledge transfer", "abstract": "  In this paper, we propose a method to improve the accuracy of speech emotion\nrecognition (SER) by using vision transformer (ViT) to attend to the\ncorrelation of frequency (y-axis) with time (x-axis) in spectrogram and\ntransferring positional information between ViT through knowledge transfer. The\nproposed method has the following originality i) We use vertically segmented\npatches of log-Mel spectrogram to analyze the correlation of frequencies over\ntime. This type of patch allows us to correlate the most relevant frequencies\nfor a particular emotion with the time they were uttered. ii) We propose the\nuse of image coordinate encoding, an absolute positional encoding suitable for\nViT. By normalizing the x, y coordinates of the image to -1 to 1 and\nconcatenating them to the image, we can effectively provide valid absolute\npositional information for ViT. iii) Through feature map matching, the locality\nand location information of the teacher network is effectively transmitted to\nthe student network. Teacher network is a ViT that contains locality of\nconvolutional stem and absolute position information through image coordinate\nencoding, and student network is a structure that lacks positional encoding in\nthe basic ViT structure. In feature map matching stage, we train through the\nmean absolute error (L1 loss) to minimize the difference between the feature\nmaps of the two networks. To validate the proposed method, three emotion\ndatasets (SAVEE, EmoDB, and CREMA-D) consisting of speech were converted into\nlog-Mel spectrograms for comparison experiments. The experimental results show\nthat the proposed method significantly outperforms the state-of-the-art methods\nin terms of weighted accuracy while requiring significantly fewer floating\npoint operations (FLOPs). Overall, the proposed method offers an promising\nsolution for SER by providing improved efficiency and performance.\n", "date": "2024-03-26", "categories": []}
{"id": "2406.11161", "title": "Emotion-LLaMA: Multimodal Emotion Recognition and Reasoning with\n  Instruction Tuning", "abstract": "  Accurate emotion perception is crucial for various applications, including\nhuman-computer interaction, education, and counseling. However, traditional\nsingle-modality approaches often fail to capture the complexity of real-world\nemotional expressions, which are inherently multimodal. Moreover, existing\nMultimodal Large Language Models (MLLMs) face challenges in integrating audio\nand recognizing subtle facial micro-expressions. To address this, we introduce\nthe MERR dataset, containing 28,618 coarse-grained and 4,487 fine-grained\nannotated samples across diverse emotional categories. This dataset enables\nmodels to learn from varied scenarios and generalize to real-world\napplications. Furthermore, we propose Emotion-LLaMA, a model that seamlessly\nintegrates audio, visual, and textual inputs through emotion-specific encoders.\nBy aligning features into a shared space and employing a modified LLaMA model\nwith instruction tuning, Emotion-LLaMA significantly enhances both emotional\nrecognition and reasoning capabilities. Extensive evaluations show\nEmotion-LLaMA outperforms other MLLMs, achieving top scores in Clue Overlap\n(7.83) and Label Overlap (6.25) on EMER, an F1 score of 0.9036 on MER2023-SEMI\nchallenge, and the highest UAR (45.59) and WAR (59.37) in zero-shot evaluations\non DFEW dataset.\n", "date": "2024-06-17", "categories": []}
{"id": "2410.22023", "title": "Multi-modal Speech Emotion Recognition via Feature Distribution\n  Adaptation Network", "abstract": "  In this paper, we propose a novel deep inductive transfer learning framework,\nnamed feature distribution adaptation network, to tackle the challenging\nmulti-modal speech emotion recognition problem. Our method aims to use deep\ntransfer learning strategies to align visual and audio feature distributions to\nobtain consistent representation of emotion, thereby improving the performance\nof speech emotion recognition. In our model, the pre-trained ResNet-34 is\nutilized for feature extraction for facial expression images and acoustic Mel\nspectrograms, respectively. Then, the cross-attention mechanism is introduced\nto model the intrinsic similarity relationships of multi-modal features.\nFinally, the multi-modal feature distribution adaptation is performed\nefficiently with feed-forward network, which is extended using the local\nmaximum mean discrepancy loss. Experiments are carried out on two benchmark\ndatasets, and the results demonstrate that our model can achieve excellent\nperformance compared with existing ones.\n", "date": "2024-10-29", "categories": []}
{"id": "2411.00822", "title": "EEG-based Multimodal Representation Learning for Emotion Recognition", "abstract": "  Multimodal learning has been a popular area of research, yet integrating\nelectroencephalogram (EEG) data poses unique challenges due to its inherent\nvariability and limited availability. In this paper, we introduce a novel\nmultimodal framework that accommodates not only conventional modalities such as\nvideo, images, and audio, but also incorporates EEG data. Our framework is\ndesigned to flexibly handle varying input sizes, while dynamically adjusting\nattention to account for feature importance across modalities. We evaluate our\napproach on a recently introduced emotion recognition dataset that combines\ndata from three modalities, making it an ideal testbed for multimodal learning.\nThe experimental results provide a benchmark for the dataset and demonstrate\nthe effectiveness of the proposed framework. This work highlights the potential\nof integrating EEG into multimodal systems, paving the way for more robust and\ncomprehensive applications in emotion recognition and beyond.\n", "date": "2024-10-29", "categories": []}
{"id": "2410.22895", "title": "Combining psychoanalysis and computer science: an empirical study of the\n  relationship between emotions and the Lacanian discourses", "abstract": "  This research explores the interdisciplinary interaction between\npsychoanalysis and computer science, suggesting a mutually beneficial exchange.\nIndeed, psychoanalytic concepts can enrich technological applications involving\nunconscious, elusive aspects of the human factor, such as social media and\nother interactive digital platforms. Conversely, computer science, especially\nArtificial Intelligence (AI), can contribute quantitative concepts and methods\nto psychoanalysis, identifying patterns and emotional cues in human expression.\nIn particular, this research aims to apply computer science methods to\nestablish fundamental relationships between emotions and Lacanian discourses.\nSuch relations are discovered in our approach via empirical investigation and\nstatistical analysis, and are eventually validated in a theoretical\n(psychoanalytic) way. It is worth noting that, although emotions have been\nsporadically studied in Lacanian theory, to the best of our knowledge a\nsystematic, detailed investigation of their role is missing. Such fine-grained\nunderstanding of the role of emotions can also make the identification of\nLacanian discourses more effective and easy in practise. In particular, our\nmethods indicate the emotions with highest differentiation power in terms of\ncorresponding discourses; conversely, we identify for each discourse the most\ncharacteristic emotions it admits. As a matter of fact, we develop a method\nwhich we call Lacanian Discourse Discovery (LDD), that simplifies (via\nsystematizing) the identification of Lacanian discourses in texts. Although the\nmain contribution of this paper is inherently theoretical (psychoanalytic), it\ncan also facilitate major practical applications in the realm of interactive\ndigital systems. Indeed, our approach can be automated through Artificial\nIntelligence methods that effectively identify emotions (and corresponding\ndiscourses) in texts.\n", "date": "2024-10-30", "categories": []}
{"id": "2410.21303", "title": "VEMOCLAP: A video emotion classification web application", "abstract": "  We introduce VEMOCLAP: Video EMOtion Classifier using Pretrained features,\nthe first readily available and open-source web application that analyzes the\nemotional content of any user-provided video. We improve our previous work,\nwhich exploits open-source pretrained models that work on video frames and\naudio, and then efficiently fuse the resulting pretrained features using\nmulti-head cross-attention. Our approach increases the state-of-the-art\nclassification accuracy on the Ekman-6 video emotion dataset by 4.3% and offers\nan online application for users to run our model on their own videos or YouTube\nvideos. We invite the readers to try our application at serkansulun.com/app.\n", "date": "2024-10-22", "categories": []}
{"id": "2308.16207", "title": "MASA-TCN: Multi-anchor Space-aware Temporal Convolutional Neural\n  Networks for Continuous and Discrete EEG Emotion Recognition", "abstract": "  Emotion recognition using electroencephalogram (EEG) mainly has two\nscenarios: classification of the discrete labels and regression of the\ncontinuously tagged labels. Although many algorithms were proposed for\nclassification tasks, there are only a few methods for regression tasks. For\nemotion regression, the label is continuous in time. A natural method is to\nlearn the temporal dynamic patterns. In previous studies, long short-term\nmemory (LSTM) and temporal convolutional neural networks (TCN) were utilized to\nlearn the temporal contextual information from feature vectors of EEG. However,\nthe spatial patterns of EEG were not effectively extracted. To enable the\nspatial learning ability of TCN towards better regression and classification\nperformances, we propose a novel unified model, named MASA-TCN, for EEG emotion\nregression and classification tasks. The space-aware temporal layer enables TCN\nto additionally learn from spatial relations among EEG electrodes. Besides, a\nnovel multi-anchor block with attentive fusion is proposed to learn dynamic\ntemporal dependencies. Experiments on two publicly available datasets show\nMASA-TCN achieves higher results than the state-of-the-art methods for both EEG\nemotion regression and classification tasks. The code is available at\nhttps://github.com/yi-ding-cs/MASA-TCN.\n", "date": "2023-08-30", "categories": []}
{"id": "2311.05567", "title": "Exploring Emotion Expression Recognition in Older Adults Interacting\n  with a Virtual Coach", "abstract": "  The EMPATHIC project aimed to design an emotionally expressive virtual coach\ncapable of engaging healthy seniors to improve well-being and promote\nindependent aging. One of the core aspects of the system is its human sensing\ncapabilities, allowing for the perception of emotional states to provide a\npersonalized experience. This paper outlines the development of the emotion\nexpression recognition module of the virtual coach, encompassing data\ncollection, annotation design, and a first methodological approach, all\ntailored to the project requirements. With the latter, we investigate the role\nof various modalities, individually and combined, for discrete emotion\nexpression recognition in this context: speech from audio, and facial\nexpressions, gaze, and head dynamics from video. The collected corpus includes\nusers from Spain, France, and Norway, and was annotated separately for the\naudio and video channels with distinct emotional labels, allowing for a\nperformance comparison across cultures and label types. Results confirm the\ninformative power of the modalities studied for the emotional categories\nconsidered, with multimodal methods generally outperforming others (around 68%\naccuracy with audio labels and 72-74% with video labels). The findings are\nexpected to contribute to the limited literature on emotion recognition applied\nto older adults in conversational human-machine interaction.\n", "date": "2023-11-09", "categories": []}
{"id": "2410.19134", "title": "AlignCap: Aligning Speech Emotion Captioning to Human Preferences", "abstract": "  Speech Emotion Captioning (SEC) has gradually become an active research task.\nThe emotional content conveyed through human speech are often complex, and\nclassifying them into fixed categories may not be enough to fully capture\nspeech emotions. Describing speech emotions through natural language may be a\nmore effective approach. However, existing SEC methods often produce\nhallucinations and lose generalization on unseen speech. To overcome these\nproblems, we propose AlignCap, which Aligning Speech Emotion Captioning to\nHuman Preferences based on large language model (LLM) with two properties: 1)\nSpeech-Text Alignment, which minimizing the divergence between the LLM's\nresponse prediction distributions for speech and text inputs using knowledge\ndistillation (KD) Regularization. 2) Human Preference Alignment, where we\ndesign Preference Optimization (PO) Regularization to eliminate factuality and\nfaithfulness hallucinations. We also extract emotional clues as a prompt for\nenriching fine-grained information under KD-Regularization. Experiments\ndemonstrate that AlignCap presents stronger performance to other\nstate-of-the-art methods on Zero-shot SEC task.\n", "date": "2024-10-24", "categories": []}
{"id": "2404.03239", "title": "Exploring Emotions in Multi-componential Space using Interactive VR\n  Games", "abstract": "  Emotion understanding is a complex process that involves multiple components.\nThe ability to recognise emotions not only leads to new context awareness\nmethods but also enhances system interaction's effectiveness by perceiving and\nexpressing emotions. Despite the attention to discrete and dimensional models,\nneuroscientific evidence supports those emotions as being complex and\nmulti-faceted. One framework that resonated well with such findings is the\nComponent Process Model (CPM), a theory that considers the complexity of\nemotions with five interconnected components: appraisal, expression,\nmotivation, physiology and feeling. However, the relationship between CPM and\ndiscrete emotions has not yet been fully explored. Therefore, to better\nunderstand emotions underlying processes, we operationalised a data-driven\napproach using interactive Virtual Reality (VR) games and collected multimodal\nmeasures (self-reports, physiological and facial signals) from 39 participants.\nWe used Machine Learning (ML) methods to identify the unique contributions of\neach component to emotion differentiation. Our results showed the role of\ndifferent components in emotion differentiation, with the model including all\ncomponents demonstrating the most significant contribution. Moreover, we found\nthat at least five dimensions are needed to represent the variation of emotions\nin our dataset. These findings also have implications for using VR environments\nin emotion research and highlight the role of physiological signals in emotion\nrecognition within such environments.\n", "date": "2024-04-04", "categories": []}
{"id": "2404.08295", "title": "Study of Emotion Concept Formation by Integrating Vision, Physiology,\n  and Word Information using Multilayered Multimodal Latent Dirichlet\n  Allocation", "abstract": "  How are emotions formed? Through extensive debate and the promulgation of\ndiverse theories , the theory of constructed emotion has become prevalent in\nrecent research on emotions. According to this theory, an emotion concept\nrefers to a category formed by interoceptive and exteroceptive information\nassociated with a specific emotion. An emotion concept stores past experiences\nas knowledge and can predict unobserved information from acquired information.\nTherefore, in this study, we attempted to model the formation of emotion\nconcepts using a constructionist approach from the perspective of the\nconstructed emotion theory. Particularly, we constructed a model using\nmultilayered multimodal latent Dirichlet allocation , which is a probabilistic\ngenerative model. We then trained the model for each subject using vision,\nphysiology, and word information obtained from multiple people who experienced\ndifferent visual emotion-evoking stimuli. To evaluate the model, we verified\nwhether the formed categories matched human subjectivity and determined whether\nunobserved information could be predicted via categories. The verification\nresults exceeded chance level, suggesting that emotion concept formation can be\nexplained by the proposed model.\n", "date": "2024-04-12", "categories": []}
{"id": "2404.15213", "title": "Automatic Classification of Subjective Time Perception Using Multi-modal\n  Physiological Data of Air Traffic Controllers", "abstract": "  In high-pressure environments where human individuals must simultaneously\nmonitor multiple entities, communicate effectively, and maintain intense focus,\nthe perception of time becomes a critical factor influencing performance and\nwell-being. One indicator of well-being can be the person's subjective time\nperception. In our project $ChronoPilot$, we aim to develop a device that\nmodulates human subjective time perception. In this study, we present a method\nto automatically assess the subjective time perception of air traffic\ncontrollers, a group often faced with demanding conditions, using their\nphysiological data and eleven state-of-the-art machine learning classifiers.\nThe physiological data consist of photoplethysmogram, electrodermal activity,\nand temperature data. We find that the support vector classifier works best\nwith an accuracy of 79 % and electrodermal activity provides the most\ndescriptive biomarker. These findings are an important step towards closing the\nfeedback loop of our $ChronoPilot$-device to automatically modulate the user's\nsubjective time perception. This technological advancement may promise\nimprovements in task management, stress reduction, and overall productivity in\nhigh-stakes professions.\n", "date": "2024-03-28", "categories": []}
{"id": "2410.17740", "title": "Emotion Recognition with Facial Attention and Objective Activation\n  Functions", "abstract": "  In this paper, we study the effect of introducing channel and spatial\nattention mechanisms, namely SEN-Net, ECA-Net, and CBAM, to existing CNN\nvision-based models such as VGGNet, ResNet, and ResNetV2 to perform the Facial\nEmotion Recognition task. We show that not only attention can significantly\nimprove the performance of these models but also that combining them with a\ndifferent activation function can further help increase the performance of\nthese models.\n", "date": "2024-10-23", "categories": []}
{"id": "2408.01838", "title": "Tracking Emotional Dynamics in Chat Conversations: A Hybrid Approach\n  using DistilBERT and Emoji Sentiment Analysis", "abstract": "  Computer-mediated communication has become more important than face-to-face\ncommunication in many contexts. Tracking emotional dynamics in chat\nconversations can enhance communication, improve services, and support\nwell-being in various contexts. This paper explores a hybrid approach to\ntracking emotional dynamics in chat conversations by combining DistilBERT-based\ntext emotion detection and emoji sentiment analysis. A Twitter dataset was\nanalyzed using various machine learning algorithms, including SVM, Random\nForest, and AdaBoost. We contrasted their performance with DistilBERT. Results\nreveal DistilBERT's superior performance in emotion recognition. Our approach\naccounts for emotive expressions conveyed through emojis to better understand\nparticipants' emotions during chats. We demonstrate how this approach can\neffectively capture and analyze emotional shifts in real-time conversations.\nOur findings show that integrating text and emoji analysis is an effective way\nof tracking chat emotion, with possible applications in customer service, work\nchats, and social media interactions.\n", "date": "2024-08-03", "categories": []}
{"id": "2410.16407", "title": "Enhancing Multimodal Affective Analysis with Learned Live Comment\n  Features", "abstract": "  Live comments, also known as Danmaku, are user-generated messages that are\nsynchronized with video content. These comments overlay directly onto streaming\nvideos, capturing viewer emotions and reactions in real-time. While prior work\nhas leveraged live comments in affective analysis, its use has been limited due\nto the relative rarity of live comments across different video platforms. To\naddress this, we first construct the Live Comment for Affective Analysis\n(LCAffect) dataset which contains live comments for English and Chinese videos\nspanning diverse genres that elicit a wide spectrum of emotions. Then, using\nthis dataset, we use contrastive learning to train a video encoder to produce\nsynthetic live comment features for enhanced multimodal affective content\nanalysis. Through comprehensive experimentation on a wide range of affective\nanalysis tasks (sentiment, emotion recognition, and sarcasm detection) in both\nEnglish and Chinese, we demonstrate that these synthetic live comment features\nsignificantly improve performance over state-of-the-art methods.\n", "date": "2024-10-21", "categories": []}
{"id": "2211.09172", "title": "Deep Emotion Recognition in Textual Conversations: A Survey", "abstract": "  Emotion Recognition in Conversations (ERC) is a key step towards successful\nhuman-machine interaction. While the field has seen tremendous advancement in\nthe last few years, new applications and implementation scenarios present novel\nchallenges and opportunities. These range from leveraging the conversational\ncontext, speaker, and emotion dynamics modelling, to interpreting common sense\nexpressions, informal language, and sarcasm, addressing challenges of real-time\nERC, recognizing emotion causes, different taxonomies across datasets,\nmultilingual ERC, and interpretability. This survey starts by introducing ERC,\nelaborating on the challenges and opportunities of this task. It proceeds with\na description of the emotion taxonomies and a variety of ERC benchmark datasets\nemploying such taxonomies. This is followed by descriptions comparing the most\nprominent works in ERC with explanations of the neural architectures employed.\nThen, it provides advisable ERC practices towards better frameworks,\nelaborating on methods to deal with subjectivity in annotations and modelling\nand methods to deal with the typically unbalanced ERC datasets. Finally, it\npresents systematic review tables comparing several works regarding the methods\nused and their performance. Benchmarking these works highlights resorting to\npre-trained Transformer Language Models to extract utterance representations,\nusing Gated and Graph Neural Networks to model the interactions between these\nutterances, and leveraging Generative Large Language Models to tackle ERC\nwithin a generative framework. This survey emphasizes the advantage of\nleveraging techniques to address unbalanced data, the exploration of mixed\nemotions, and the benefits of incorporating annotation subjectivity in the\nlearning phase.\n", "date": "2022-11-16", "categories": []}
{"id": "2410.15974", "title": "Large Language Models for Cross-lingual Emotion Detection", "abstract": "  This paper presents a detailed system description of our entry for the WASSA\n2024 Task 2, focused on cross-lingual emotion detection. We utilized a\ncombination of large language models (LLMs) and their ensembles to effectively\nunderstand and categorize emotions across different languages. Our approach not\nonly outperformed other submissions with a large margin, but also demonstrated\nthe strength of integrating multiple models to enhance performance.\nAdditionally, We conducted a thorough comparison of the benefits and\nlimitations of each model used. An error analysis is included along with\nsuggested areas for future improvement. This paper aims to offer a clear and\ncomprehensive understanding of advanced techniques in emotion detection, making\nit accessible even to those new to the field.\n", "date": "2024-10-21", "categories": []}
{"id": "2410.14050", "title": "Learning Multimodal Cues of Children's Uncertainty", "abstract": "  Understanding uncertainty plays a critical role in achieving common ground\n(Clark et al.,1983). This is especially important for multimodal AI systems\nthat collaborate with users to solve a problem or guide the user through a\nchallenging concept. In this work, for the first time, we present a dataset\nannotated in collaboration with developmental and cognitive psychologists for\nthe purpose of studying nonverbal cues of uncertainty. We then present an\nanalysis of the data, studying different roles of uncertainty and its\nrelationship with task difficulty and performance. Lastly, we present a\nmultimodal machine learning model that can predict uncertainty given a\nreal-time video clip of a participant, which we find improves upon a baseline\nmultimodal transformer model. This work informs research on cognitive\ncoordination between human-human and human-AI and has broad implications for\ngesture understanding and generation. The anonymized version of our data and\ncode will be publicly available upon the completion of the required consent\nforms and data sheets.\n", "date": "2024-10-17", "categories": []}
{"id": "2410.11522", "title": "Leveraging LLM Embeddings for Cross Dataset Label Alignment and Zero\n  Shot Music Emotion Prediction", "abstract": "  In this work, we present a novel method for music emotion recognition that\nleverages Large Language Model (LLM) embeddings for label alignment across\nmultiple datasets and zero-shot prediction on novel categories. First, we\ncompute LLM embeddings for emotion labels and apply non-parametric clustering\nto group similar labels, across multiple datasets containing disjoint labels.\nWe use these cluster centers to map music features (MERT) to the LLM embedding\nspace. To further enhance the model, we introduce an alignment regularization\nthat enables dissociation of MERT embeddings from different clusters. This\nfurther enhances the model's ability to better adaptation to unseen datasets.\nWe demonstrate the effectiveness of our approach by performing zero-shot\ninference on a new dataset, showcasing its ability to generalize to unseen\nlabels without additional training.\n", "date": "2024-10-15", "categories": []}
{"id": "2410.12828", "title": "GCM-Net: Graph-enhanced Cross-Modal Infusion with a Metaheuristic-Driven\n  Network for Video Sentiment and Emotion Analysis", "abstract": "  Sentiment analysis and emotion recognition in videos are challenging tasks,\ngiven the diversity and complexity of the information conveyed in different\nmodalities. Developing a highly competent framework that effectively addresses\nthe distinct characteristics across various modalities is a primary concern in\nthis domain. Previous studies on combined multimodal sentiment and emotion\nanalysis often overlooked effective fusion for modality integration, intermodal\ncontextual congruity, optimizing concatenated feature spaces, leading to\nsuboptimal architecture. This paper presents a novel framework that leverages\nthe multi-modal contextual information from utterances and applies\nmetaheuristic algorithms to learn the contributing features for utterance-level\nsentiment and emotion prediction. Our Graph-enhanced Cross-Modal Infusion with\na Metaheuristic-Driven Network (GCM-Net) integrates graph sampling and\naggregation to recalibrate the modality features for video sentiment and\nemotion prediction. GCM-Net includes a cross-modal attention module determining\nintermodal interactions and utterance relevance. A harmonic optimization module\nemploying a metaheuristic algorithm combines attended features, allowing for\nhandling both single and multi-utterance inputs. To show the effectiveness of\nour approach, we have conducted extensive evaluations on three prominent\nmulti-modal benchmark datasets, CMU MOSI, CMU MOSEI, and IEMOCAP. The\nexperimental results demonstrate the efficacy of our proposed approach,\nshowcasing accuracies of 91.56% and 86.95% for sentiment analysis on MOSI and\nMOSEI datasets. We have performed emotion analysis for the IEMOCAP dataset\nprocuring an accuracy of 85.66% which signifies substantial performance\nenhancements over existing methods.\n", "date": "2024-10-02", "categories": []}
{"id": "2404.18327", "title": "MultiMAE-DER: Multimodal Masked Autoencoder for Dynamic Emotion\n  Recognition", "abstract": "  This paper presents a novel approach to processing multimodal data for\ndynamic emotion recognition, named as the Multimodal Masked Autoencoder for\nDynamic Emotion Recognition (MultiMAE-DER). The MultiMAE-DER leverages the\nclosely correlated representation information within spatiotemporal sequences\nacross visual and audio modalities. By utilizing a pre-trained masked\nautoencoder model, the MultiMAEDER is accomplished through simple,\nstraightforward finetuning. The performance of the MultiMAE-DER is enhanced by\noptimizing six fusion strategies for multimodal input sequences. These\nstrategies address dynamic feature correlations within cross-domain data across\nspatial, temporal, and spatiotemporal sequences. In comparison to\nstate-of-the-art multimodal supervised learning models for dynamic emotion\nrecognition, MultiMAE-DER enhances the weighted average recall (WAR) by 4.41%\non the RAVDESS dataset and by 2.06% on the CREMAD. Furthermore, when compared\nwith the state-of-the-art model of multimodal self-supervised learning,\nMultiMAE-DER achieves a 1.86% higher WAR on the IEMOCAP dataset.\n", "date": "2024-04-28", "categories": []}
{"id": "2407.19340", "title": "Integrating Large Language Models into a Tri-Modal Architecture for\n  Automated Depression Classification on the DAIC-WOZ", "abstract": "  Major Depressive Disorder (MDD) is a pervasive mental health condition that\naffects 300 million people worldwide. This work presents a novel, BiLSTM-based\ntri-modal model-level fusion architecture for the binary classification of\ndepression from clinical interview recordings. The proposed architecture\nincorporates Mel Frequency Cepstral Coefficients, Facial Action Units, and uses\na two-shot learning based GPT-4 model to process text data. This is the first\nwork to incorporate large language models into a multi-modal architecture for\nthis task. It achieves impressive results on the DAIC-WOZ AVEC 2016 Challenge\ncross-validation split and Leave-One-Subject-Out cross-validation split,\nsurpassing all baseline models and multiple state-of-the-art models. In\nLeave-One-Subject-Out testing, it achieves an accuracy of 91.01%, an F1-Score\nof 85.95%, a precision of 80%, and a recall of 92.86%.\n", "date": "2024-07-27", "categories": []}
{"id": "2410.09636", "title": "Can We Estimate Purchase Intention Based on Zero-shot Speech Emotion\n  Recognition?", "abstract": "  This paper proposes a zero-shot speech emotion recognition (SER) method that\nestimates emotions not previously defined in the SER model training.\nConventional methods are limited to recognizing emotions defined by a single\nword. Moreover, we have the motivation to recognize unknown bipolar emotions\nsuch as ``I want to buy - I do not want to buy.'' In order to allow the model\nto define classes using sentences freely and to estimate unknown bipolar\nemotions, our proposed method expands upon the contrastive language-audio\npre-training (CLAP) framework by introducing multi-class and multi-task\nsettings. We also focus on purchase intention as a bipolar emotion and\ninvestigate the model's performance to zero-shot estimate it. This study is the\nfirst attempt to estimate purchase intention from speech directly. Experiments\nconfirm that the results of zero-shot estimation by the proposed method are at\nthe same level as those of the model trained by supervised learning.\n", "date": "2024-10-12", "categories": []}
{"id": "2402.12862", "title": "Handling Ambiguity in Emotion: From Out-of-Domain Detection to\n  Distribution Estimation", "abstract": "  The subjective perception of emotion leads to inconsistent labels from human\nannotators. Typically, utterances lacking majority-agreed labels are excluded\nwhen training an emotion classifier, which cause problems when encountering\nambiguous emotional expressions during testing. This paper investigates three\nmethods to handle ambiguous emotion. First, we show that incorporating\nutterances without majority-agreed labels as an additional class in the\nclassifier reduces the classification performance of the other emotion classes.\nThen, we propose detecting utterances with ambiguous emotions as out-of-domain\nsamples by quantifying the uncertainty in emotion classification using\nevidential deep learning. This approach retains the classification accuracy\nwhile effectively detects ambiguous emotion expressions. Furthermore, to obtain\nfine-grained distinctions among ambiguous emotions, we propose representing\nemotion as a distribution instead of a single class label. The task is thus\nre-framed from classification to distribution estimation where every individual\nannotation is taken into account, not just the majority opinion. The evidential\nuncertainty measure is extended to quantify the uncertainty in emotion\ndistribution estimation. Experimental results on the IEMOCAP and CREMA-D\ndatasets demonstrate the superior capability of the proposed method in terms of\nmajority class prediction, emotion distribution estimation, and uncertainty\nestimation.\n", "date": "2024-02-20", "categories": []}
{"id": "2410.08470", "title": "DAT: Dialogue-Aware Transformer with Modality-Group Fusion for Human\n  Engagement Estimation", "abstract": "  Engagement estimation plays a crucial role in understanding human social\nbehaviors, attracting increasing research interests in fields such as affective\ncomputing and human-computer interaction. In this paper, we propose a\nDialogue-Aware Transformer framework (DAT) with Modality-Group Fusion (MGF),\nwhich relies solely on audio-visual input and is language-independent, for\nestimating human engagement in conversations. Specifically, our method employs\na modality-group fusion strategy that independently fuses audio and visual\nfeatures within each modality for each person before inferring the entire\naudio-visual content. This strategy significantly enhances the model's\nperformance and robustness. Additionally, to better estimate the target\nparticipant's engagement levels, the introduced Dialogue-Aware Transformer\nconsiders both the participant's behavior and cues from their conversational\npartners. Our method was rigorously tested in the Multi-Domain Engagement\nEstimation Challenge held by MultiMediate'24, demonstrating notable\nimprovements in engagement-level regression precision over the baseline model.\nNotably, our approach achieves a CCC score of 0.76 on the NoXi Base test set\nand an average CCC of 0.64 across the NoXi Base, NoXi-Add, and MPIIGI test\nsets.\n", "date": "2024-10-11", "categories": []}
{"id": "2207.11081", "title": "Emotion Separation and Recognition from a Facial Expression by\n  Generating the Poker Face with Vision Transformers", "abstract": "  Representation learning and feature disentanglement have garnered significant\nresearch interest in the field of facial expression recognition (FER). The\ninherent ambiguity of emotion labels poses challenges for conventional\nsupervised representation learning methods. Moreover, directly learning the\nmapping from a facial expression image to an emotion label lacks explicit\nsupervision signals for capturing fine-grained facial features. In this paper,\nwe propose a novel FER model, named Poker Face Vision Transformer or PF-ViT, to\naddress these challenges. PF-ViT aims to separate and recognize the\ndisturbance-agnostic emotion from a static facial image via generating its\ncorresponding poker face, without the need for paired images. Inspired by the\nFacial Action Coding System, we regard an expressive face as the combined\nresult of a set of facial muscle movements on one's poker face (i.e., an\nemotionless face). PF-ViT utilizes vanilla Vision Transformers, and its\ncomponents are firstly pre-trained as Masked Autoencoders on a large facial\nexpression dataset without emotion labels, yielding excellent representations.\nSubsequently, we train PF-ViT using a GAN framework. During training, the\nauxiliary task of poke face generation promotes the disentanglement between\nemotional and emotion-irrelevant components, guiding the FER model to\nholistically capture discriminative facial details. Quantitative and\nqualitative results demonstrate the effectiveness of our method, surpassing the\nstate-of-the-art methods on four popular FER datasets.\n", "date": "2022-07-22", "categories": []}
{"id": "2409.09194", "title": "Hierarchical Hypercomplex Network for Multimodal Emotion Recognition", "abstract": "  Emotion recognition is relevant in various domains, ranging from healthcare\nto human-computer interaction. Physiological signals, being beyond voluntary\ncontrol, offer reliable information for this purpose, unlike speech and facial\nexpressions which can be controlled at will. They reflect genuine emotional\nresponses, devoid of conscious manipulation, thereby enhancing the credibility\nof emotion recognition systems. Nonetheless, multimodal emotion recognition\nwith deep learning models remains a relatively unexplored field. In this paper,\nwe introduce a fully hypercomplex network with a hierarchical learning\nstructure to fully capture correlations. Specifically, at the encoder level,\nthe model learns intra-modal relations among the different channels of each\ninput signal. Then, a hypercomplex fusion module learns inter-modal relations\namong the embeddings of the different modalities. The main novelty is in\nexploiting intra-modal relations by endowing the encoders with parameterized\nhypercomplex convolutions (PHCs) that thanks to hypercomplex algebra can\ncapture inter-channel interactions within single modalities. Instead, the\nfusion module comprises parameterized hypercomplex multiplications (PHMs) that\ncan model inter-modal correlations. The proposed architecture surpasses\nstate-of-the-art models on the MAHNOB-HCI dataset for emotion recognition,\nspecifically in classifying valence and arousal from electroencephalograms\n(EEGs) and peripheral physiological signals. The code of this study is\navailable at https://github.com/ispamm/MHyEEG.\n", "date": "2024-09-13", "categories": []}
{"id": "2404.00403", "title": "UniMEEC: Towards Unified Multimodal Emotion Recognition and Emotion\n  Cause", "abstract": "  Multimodal emotion recognition in conversation (MERC) and multimodal\nemotion-cause pair extraction (MECPE) have recently garnered significant\nattention. Emotions are the expression of affect or feelings; responses to\nspecific events, or situations -- known as emotion causes. Both collectively\nexplain the causality between human emotion and intents. However, existing\nworks treat emotion recognition and emotion cause extraction as two individual\nproblems, ignoring their natural causality. In this paper, we propose a Unified\nMultimodal Emotion recognition and Emotion-Cause analysis framework (UniMEEC)\nto explore the causality between emotion and emotion cause. Concretely, UniMEEC\nreformulates the MERC and MECPE tasks as mask prediction problems and unifies\nthem with a causal prompt template. To differentiate the modal effects, UniMEEC\nproposes a multimodal causal prompt to probe the pre-trained knowledge\nspecified to modality and implements cross-task and cross-modality interactions\nunder task-oriented settings. Experiment results on four public benchmark\ndatasets verify the model performance on MERC and MECPE tasks and achieve\nconsistent improvements compared with the previous state-of-the-art methods.\n", "date": "2024-03-30", "categories": []}
{"id": "2410.05778", "title": "Song Emotion Classification of Lyrics with Out-of-Domain Data under\n  Label Scarcity", "abstract": "  Songs have been found to profoundly impact human emotions, with lyrics having\nsignificant power to stimulate emotional changes in the audience. There is a\nscarcity of large, high quality in-domain datasets for lyrics-based song\nemotion classification (Edmonds and Sedoc, 2021; Zhou, 2022). It has been noted\nthat in-domain training datasets are often difficult to acquire (Zhang and\nMiao, 2023) and that label acquisition is often limited by cost, time, and\nother factors (Azad et al., 2018). We examine the novel usage of a large\nout-of-domain dataset as a creative solution to the challenge of training data\nscarcity in the emotional classification of song lyrics. We find that CNN\nmodels trained on a large Reddit comments dataset achieve satisfactory\nperformance and generalizability to lyrical emotion classification, thus giving\ninsights into and a promising possibility in leveraging large, publicly\navailable out-of-domain datasets for domains whose in-domain data are lacking\nor costly to acquire.\n", "date": "2024-10-08", "categories": []}
{"id": "2401.15743", "title": "Real-time EEG-based Emotion Recognition Model using Principal Component\n  Analysis and Tree-based Models for Neurohumanities", "abstract": "  Within the field of Humanities, there is a recognized need for educational\ninnovation, as there are currently no reported tools available that enable\nindividuals to interact with their environment to create an enhanced learning\nexperience in the humanities (e.g., immersive spaces). This project proposes a\nsolution to address this gap by integrating technology and promoting the\ndevelopment of teaching methodologies in the humanities, specifically by\nincorporating emotional monitoring during the learning process of humanistic\ncontext inside an immersive space. In order to achieve this goal, a real-time\nemotion detection EEG-based system was developed to interpret and classify\nspecific emotions. These emotions aligned with the early proposal by Descartes\n(Passions), including admiration, love, hate, desire, joy, and sadness. This\nsystem aims to integrate emotional data into the Neurohumanities Lab\ninteractive platform, creating a comprehensive and immersive learning\nenvironment. This work developed a ML, real-time emotion detection model that\nprovided Valence, Arousal, and Dominance (VAD) estimations every 5 seconds.\nUsing PCA, PSD, RF, and Extra-Trees, the best 8 channels and their respective\nbest band powers were extracted; furthermore, multiple models were evaluated\nusing shift-based data division and cross-validations. After assessing their\nperformance, Extra-Trees achieved a general accuracy of 96%, higher than the\nreported in the literature (88% accuracy). The proposed model provided\nreal-time predictions of VAD variables and was adapted to classify Descartes'\nsix main passions. However, with the VAD values obtained, more than 15 emotions\ncan be classified (reported in the VAD emotion mapping) and extend the range of\nthis application.\n", "date": "2024-01-28", "categories": []}
{"id": "2410.03696", "title": "Improving Emotion Recognition Accuracy with Personalized Clustering", "abstract": "  Emotion recognition through artificial intelligence and smart sensing of\nphysical and physiological signals (Affective Computing) is achieving very\ninteresting results in terms of accuracy, inference times, and user-independent\nmodels. In this sense, there are applications related to the safety and\nwell-being of people (sexual aggressions, gender-based violence, children and\nelderly abuse, mental health, etc.) that require even more improvements.\nEmotion detection should be done with fast, discrete, and non-luxurious systems\nworking in real-time and real life (wearable devices, wireless communications,\nbattery-powered). Furthermore, emotional reactions to violence are not equal in\nall people. Then, large general models cannot be applied to a multiuser system\nfor people protection, and customized and simple AI models would be welcomed by\nhealth and social workers and law enforcement agents. These customized models\nwill be applicable to clusters of subjects sharing similarities in their\nemotional reactions to external stimuli. This customization requires several\nsteps: creating clusters of subjects with similar behaviors, creating AI models\nfor every cluster, continually updating these models with new data, and\nenrolling new subjects in clusters when required. A methodology for clustering\ndata compiled (physical and physiological data, together with emotional labels)\nis presented in this work, as well as the method for including new subjects\nonce the AI model is generated. Experimental results demonstrate an improvement\nof 4% in accuracy and 3% in f1-score w.r.t. the general model, along with a 14%\nreduction in variability.\n", "date": "2024-09-23", "categories": []}
{"id": "2410.04491", "title": "Knowledge-Guided Dynamic Modality Attention Fusion Framework for\n  Multimodal Sentiment Analysis", "abstract": "  Multimodal Sentiment Analysis (MSA) utilizes multimodal data to infer the\nusers' sentiment. Previous methods focus on equally treating the contribution\nof each modality or statically using text as the dominant modality to conduct\ninteraction, which neglects the situation where each modality may become\ndominant. In this paper, we propose a Knowledge-Guided Dynamic Modality\nAttention Fusion Framework (KuDA) for multimodal sentiment analysis. KuDA uses\nsentiment knowledge to guide the model dynamically selecting the dominant\nmodality and adjusting the contributions of each modality. In addition, with\nthe obtained multimodal representation, the model can further highlight the\ncontribution of dominant modality through the correlation evaluation loss.\nExtensive experiments on four MSA benchmark datasets indicate that KuDA\nachieves state-of-the-art performance and is able to adapt to different\nscenarios of dominant modality.\n", "date": "2024-10-06", "categories": []}
{"id": "2410.02804", "title": "Leveraging Retrieval Augment Approach for Multimodal Emotion Recognition\n  Under Missing Modalities", "abstract": "  Multimodal emotion recognition utilizes complete multimodal information and\nrobust multimodal joint representation to gain high performance. However, the\nideal condition of full modality integrity is often not applicable in reality\nand there always appears the situation that some modalities are missing. For\nexample, video, audio, or text data is missing due to sensor failure or network\nbandwidth problems, which presents a great challenge to MER research.\nTraditional methods extract useful information from the complete modalities and\nreconstruct the missing modalities to learn robust multimodal joint\nrepresentation. These methods have laid a solid foundation for research in this\nfield, and to a certain extent, alleviated the difficulty of multimodal emotion\nrecognition under missing modalities. However, relying solely on internal\nreconstruction and multimodal joint learning has its limitations, especially\nwhen the missing information is critical for emotion recognition. To address\nthis challenge, we propose a novel framework of Retrieval Augment for Missing\nModality Multimodal Emotion Recognition (RAMER), which introduces similar\nmultimodal emotion data to enhance the performance of emotion recognition under\nmissing modalities. By leveraging databases, that contain related multimodal\nemotion data, we can retrieve similar multimodal emotion information to fill in\nthe gaps left by missing modalities. Various experimental results demonstrate\nthat our framework is superior to existing state-of-the-art approaches in\nmissing modality MER tasks. Our whole project is publicly available on\nhttps://github.com/WooyoohL/Retrieval_Augment_MER.\n", "date": "2024-09-19", "categories": []}
{"id": "2403.17175", "title": "Engagement Measurement Based on Facial Landmarks and Spatial-Temporal\n  Graph Convolutional Networks", "abstract": "  Engagement in virtual learning is crucial for a variety of factors including\nstudent satisfaction, performance, and compliance with learning programs, but\nmeasuring it is a challenging task. There is therefore considerable interest in\nutilizing artificial intelligence and affective computing to measure engagement\nin natural settings as well as on a large scale. This paper introduces a novel,\nprivacy-preserving method for engagement measurement from videos. It uses\nfacial landmarks, which carry no personally identifiable information, extracted\nfrom videos via the MediaPipe deep learning solution. The extracted facial\nlandmarks are fed to Spatial-Temporal Graph Convolutional Networks (ST-GCNs) to\noutput the engagement level of the student in the video. To integrate the\nordinal nature of the engagement variable into the training process, ST-GCNs\nundergo training in a novel ordinal learning framework based on transfer\nlearning. Experimental results on two video student engagement measurement\ndatasets show the superiority of the proposed method compared to previous\nmethods with improved state-of-the-art on the EngageNet dataset with a 3.1%\nimprovement in four-class engagement level classification accuracy and on the\nOnline Student Engagement dataset with a 1.5% improvement in binary engagement\nclassification accuracy. Gradient-weighted Class Activation Mapping (Grad-CAM)\nwas applied to the developed ST-GCNs to interpret the engagement measurements\nobtained by the proposed method in both the spatial and temporal domains. The\nrelatively lightweight and fast ST-GCN and its integration with the real-time\nMediaPipe make the proposed approach capable of being deployed on virtual\nlearning platforms and measuring engagement in real-time.\n", "date": "2024-03-25", "categories": []}
{"id": "2410.01848", "title": "Spatial Action Unit Cues for Interpretable Deep Facial Expression\n  Recognition", "abstract": "  Although state-of-the-art classifiers for facial expression recognition (FER)\ncan achieve a high level of accuracy, they lack interpretability, an important\nfeature for end-users. Experts typically associate spatial action units (AUs)\nfrom a codebook to facial regions for the visual interpretation of expressions.\nIn this paper, the same expert steps are followed. A new learning strategy is\nproposed to explicitly incorporate AU cues into classifier training, allowing\nto train deep interpretable models. During training, this AU codebook is used,\nalong with the input image expression label, and facial landmarks, to construct\na AU heatmap that indicates the most discriminative image regions of interest\nw.r.t the facial expression. This valuable spatial cue is leveraged to train a\ndeep interpretable classifier for FER. This is achieved by constraining the\nspatial layer features of a classifier to be correlated with AU heatmaps. Using\na composite loss, the classifier is trained to correctly classify an image\nwhile yielding interpretable visual layer-wise attention correlated with AU\nmaps, simulating the expert decision process. Our strategy only relies on image\nclass expression for supervision, without additional manual annotations. Our\nnew strategy is generic, and can be applied to any deep CNN- or\ntransformer-based classifier without requiring any architectural change or\nsignificant additional training time. Our extensive evaluation on two public\nbenchmarks RAF-DB, and AffectNet datasets shows that our proposed strategy can\nimprove layer-wise interpretability without degrading classification\nperformance. In addition, we explore a common type of interpretable classifiers\nthat rely on class activation mapping (CAM) methods, and show that our approach\ncan also improve CAM interpretability.\n", "date": "2024-10-01", "categories": []}
{"id": "2410.00010", "title": "PHemoNet: A Multimodal Network for Physiological Signals", "abstract": "  Emotion recognition is essential across numerous fields, including medical\napplications and brain-computer interface (BCI). Emotional responses include\nbehavioral reactions, such as tone of voice and body movement, and changes in\nphysiological signals, such as the electroencephalogram (EEG). The latter are\ninvoluntary, thus they provide a reliable input for identifying emotions, in\ncontrast to the former which individuals can consciously control. These signals\nreveal true emotional states without intentional alteration, thus increasing\nthe accuracy of emotion recognition models. However, multimodal deep learning\nmethods from physiological signals have not been significantly investigated. In\nthis paper, we introduce PHemoNet, a fully hypercomplex network for multimodal\nemotion recognition from physiological signals. In detail, the architecture\ncomprises modality-specific encoders and a fusion module. Both encoders and\nfusion modules are defined in the hypercomplex domain through parameterized\nhypercomplex multiplications (PHMs) that can capture latent relations between\nthe different dimensions of each modality and between the modalities\nthemselves. The proposed method outperforms current state-of-the-art models on\nthe MAHNOB-HCI dataset in classifying valence and arousal using\nelectroencephalograms (EEGs) and peripheral physiological signals. The code for\nthis work is available at https://github.com/ispamm/MHyEEG.\n", "date": "2024-09-13", "categories": []}
{"id": "2409.18971", "title": "Early Joint Learning of Emotion Information Makes MultiModal Model\n  Understand You Better", "abstract": "  In this paper, we present our solutions for emotion recognition in the\nsub-challenges of Multimodal Emotion Recognition Challenge (MER2024). To\nmitigate the modal competition issue between audio and text, we adopt an early\nfusion strategy based on a large language model, where joint training of audio\nand text is conducted initially. And the joint Audio-Text modal feature will be\nlate-fused with other unimodal features. In order to solve the problems of data\ninsufficiency and class imbalance, We use multiple turns of multi-model voting\nfor data mining. Moreover, to enhance the quality of audio features, we employ\nspeech source separation to preprocess audios. Our model ranks \\textbf{2nd} in\nboth MER2024-SEMI and MER2024-NOISE, validating our method's effectiveness.\n", "date": "2024-09-12", "categories": []}
{"id": "2409.17717", "title": "Behaviour4All: in-the-wild Facial Behaviour Analysis Toolkit", "abstract": "  In this paper, we introduce Behavior4All, a comprehensive, open-source\ntoolkit for in-the-wild facial behavior analysis, integrating Face\nLocalization, Valence-Arousal Estimation, Basic Expression Recognition and\nAction Unit Detection, all within a single framework. Available in both\nCPU-only and GPU-accelerated versions, Behavior4All leverages 12 large-scale,\nin-the-wild datasets consisting of over 5 million images from diverse\ndemographic groups. It introduces a novel framework that leverages distribution\nmatching and label co-annotation to address tasks with non-overlapping\nannotations, encoding prior knowledge of their relatedness. In the largest\nstudy of its kind, Behavior4All outperforms both state-of-the-art and toolkits\nin overall performance as well as fairness across all databases and tasks. It\nalso demonstrates superior generalizability on unseen databases and on compound\nexpression recognition. Finally, Behavior4All is way times faster than other\ntoolkits.\n", "date": "2024-09-26", "categories": []}
{"id": "2409.13084", "title": "Real-time estimation of overt attention from dynamic features of the\n  face using deep-learning", "abstract": "  Students often drift in and out of focus during class. Effective teachers\nrecognize this and re-engage them when necessary. With the shift to remote\nlearning, teachers have lost the visual feedback needed to adapt to varying\nstudent engagement. We propose using readily available front-facing video to\ninfer attention levels based on movements of the eyes, head, and face. We train\na deep learning model to predict a measure of attention based on overt eye\nmovements. Specifically, we measure Inter-Subject Correlation of eye movements\nin ten-second intervals while students watch the same educational videos. In 3\ndifferent experiments (N=83) we show that the trained model predicts this\nobjective metric of attention on unseen data with $R^2$=0.38, and on unseen\nsubjects with $R^2$=0.26-0.30. The deep network relies mostly on a student's\neye movements, but to some extent also on movements of the brows, cheeks, and\nhead. In contrast to Inter-Subject Correlation of the eyes, the model can\nestimate attentional engagement from individual students' movements without\nneeding reference data from an attentive group. This enables a much broader set\nof online applications. The solution is lightweight and can operate on the\nclient side, which mitigates some of the privacy concerns associated with\nonline attention monitoring. GitHub implementation is available at\nhttps://github.com/asortubay/timeISC\n", "date": "2024-09-19", "categories": []}
{"id": "2409.15733", "title": "EvoFA: Evolvable Fast Adaptation for EEG Emotion Recognition", "abstract": "  Electroencephalography (EEG)-based emotion recognition has gained significant\ntraction due to its accuracy and objectivity. However, the non-stationary\nnature of EEG signals leads to distribution drift over time, causing severe\nperformance degradation when the model is reused. While numerous domain\nadaptation (DA) approaches have been proposed in recent years to address this\nissue, their reliance on large amounts of target data for calibration restricts\nthem to offline scenarios, rendering them unsuitable for real-time\napplications. To address this challenge, this paper proposes Evolvable Fast\nAdaptation (EvoFA), an online adaptive framework tailored for EEG data. EvoFA\norganically integrates the rapid adaptation of Few-Shot Learning (FSL) and the\ndistribution matching of Domain Adaptation (DA) through a two-stage\ngeneralization process. During the training phase, a robust base meta-learning\nmodel is constructed for strong generalization. In the testing phase, a\ndesigned evolvable meta-adaptation module iteratively aligns the marginal\ndistribution of target (testing) data with the evolving source (training) data\nwithin a model-agnostic meta-learning framework, enabling the model to learn\nthe evolving trends of testing data relative to training data and improving\nonline testing performance. Experimental results demonstrate that EvoFA\nachieves significant improvements compared to the basic FSL method and previous\nonline methods. The introduction of EvoFA paves the way for broader adoption of\nEEG-based emotion recognition in real-world applications. Our code will be\nreleased upon publication.\n", "date": "2024-09-24", "categories": []}
{"id": "2409.15936", "title": "DepMamba: Progressive Fusion Mamba for Multimodal Depression Detection", "abstract": "  Depression is a common mental disorder that affects millions of people\nworldwide. Although promising, current multimodal methods hinge on aligned or\naggregated multimodal fusion, suffering two significant limitations: (i)\ninefficient long-range temporal modeling, and (ii) sub-optimal multimodal\nfusion between intermodal fusion and intramodal processing. In this paper, we\npropose an audio-visual progressive fusion Mamba for multimodal depression\ndetection, termed DepMamba. DepMamba features two core designs: hierarchical\ncontextual modeling and progressive multimodal fusion. On the one hand,\nhierarchical modeling introduces convolution neural networks and Mamba to\nextract the local-to-global features within long-range sequences. On the other\nhand, the progressive fusion first presents a multimodal collaborative State\nSpace Model (SSM) extracting intermodal and intramodal information for each\nmodality, and then utilizes a multimodal enhanced SSM for modality cohesion.\nExtensive experimental results on two large-scale depression datasets\ndemonstrate the superior performance of our DepMamba over existing\nstate-of-the-art methods. Code is available at\nhttps://github.com/Jiaxin-Ye/DepMamba.\n", "date": "2024-09-24", "categories": []}
{"id": "2409.16081", "title": "Online Multi-level Contrastive Representation Distillation for\n  Cross-Subject fNIRS Emotion Recognition", "abstract": "  Utilizing functional near-infrared spectroscopy (fNIRS) signals for emotion\nrecognition is a significant advancement in understanding human emotions.\nHowever, due to the lack of artificial intelligence data and algorithms in this\nfield, current research faces the following challenges: 1) The portable\nwearable devices have higher requirements for lightweight models; 2) The\nobjective differences of physiology and psychology among different subjects\naggravate the difficulty of emotion recognition. To address these challenges,\nwe propose a novel cross-subject fNIRS emotion recognition method, called the\nOnline Multi-level Contrastive Representation Distillation framework (OMCRD).\nSpecifically, OMCRD is a framework designed for mutual learning among multiple\nlightweight student networks. It utilizes multi-level fNIRS feature extractor\nfor each sub-network and conducts multi-view sentimental mining using\nphysiological signals. The proposed Inter-Subject Interaction Contrastive\nRepresentation (IS-ICR) facilitates knowledge transfer for interactions between\nstudent models, enhancing cross-subject emotion recognition performance. The\noptimal student network can be selected and deployed on a wearable device. Some\nexperimental results demonstrate that OMCRD achieves state-of-the-art results\nin emotional perception and affective imagery tasks.\n", "date": "2024-09-24", "categories": []}
{"id": "2405.02151", "title": "GMP-TL: Gender-augmented Multi-scale Pseudo-label Enhanced Transfer\n  Learning for Speech Emotion Recognition", "abstract": "  The continuous evolution of pre-trained speech models has greatly advanced\nSpeech Emotion Recognition (SER). However, current research typically relies on\nutterance-level emotion labels, inadequately capturing the complexity of\nemotions within a single utterance. In this paper, we introduce GMP-TL, a novel\nSER framework that employs gender-augmented multi-scale pseudo-label (GMP)\nbased transfer learning to mitigate this gap. Specifically, GMP-TL initially\nuses the pre-trained HuBERT, implementing multi-task learning and multi-scale\nk-means clustering to acquire frame-level GMPs. Subsequently, to fully leverage\nframe-level GMPs and utterance-level emotion labels, a two-stage model\nfine-tuning approach is presented to further optimize GMP-TL. Experiments on\nIEMOCAP show that our GMP-TL attains a WAR of 80.0% and an UAR of 82.0%,\nachieving superior performance compared to state-of-the-art unimodal SER\nmethods while also yielding comparable results to multimodal SER approaches.\n", "date": "2024-05-03", "categories": []}
{"id": "2407.04279", "title": "BiosERC: Integrating Biography Speakers Supported by LLMs for ERC Tasks", "abstract": "  In the Emotion Recognition in Conversation task, recent investigations have\nutilized attention mechanisms exploring relationships among utterances from\nintra- and inter-speakers for modeling emotional interaction between them.\nHowever, attributes such as speaker personality traits remain unexplored and\npresent challenges in terms of their applicability to other tasks or\ncompatibility with diverse model architectures. Therefore, this work introduces\na novel framework named BiosERC, which investigates speaker characteristics in\na conversation. By employing Large Language Models (LLMs), we extract the\n\"biographical information\" of the speaker within a conversation as\nsupplementary knowledge injected into the model to classify emotional labels\nfor each utterance. Our proposed method achieved state-of-the-art (SOTA)\nresults on three famous benchmark datasets: IEMOCAP, MELD, and EmoryNLP,\ndemonstrating the effectiveness and generalization of our model and showcasing\nits potential for adaptation to various conversation analysis tasks. Our source\ncode is available at https://github.com/yingjie7/BiosERC.\n", "date": "2024-07-05", "categories": []}
{"id": "2409.13726", "title": "Multilingual Dyadic Interaction Corpus NoXi+J: Toward Understanding\n  Asian-European Non-verbal Cultural Characteristics and their Influences on\n  Engagement", "abstract": "  Non-verbal behavior is a central challenge in understanding the dynamics of a\nconversation and the affective states between interlocutors arising from the\ninteraction. Although psychological research has demonstrated that non-verbal\nbehaviors vary across cultures, limited computational analysis has been\nconducted to clarify these differences and assess their impact on engagement\nrecognition. To gain a greater understanding of engagement and non-verbal\nbehaviors among a wide range of cultures and language spheres, in this study we\nconduct a multilingual computational analysis of non-verbal features and\ninvestigate their role in engagement and engagement prediction. To achieve this\ngoal, we first expanded the NoXi dataset, which contains interaction data from\nparticipants living in France, Germany, and the United Kingdom, by collecting\nsession data of dyadic conversations in Japanese and Chinese, resulting in the\nenhanced dataset NoXi+J. Next, we extracted multimodal non-verbal features,\nincluding speech acoustics, facial expressions, backchanneling and gestures,\nvia various pattern recognition techniques and algorithms. Then, we conducted a\nstatistical analysis of listening behaviors and backchannel patterns to\nidentify culturally dependent and independent features in each language and\ncommon features among multiple languages. These features were also correlated\nwith the engagement shown by the interlocutors. Finally, we analyzed the\ninfluence of cultural differences in the input features of LSTM models trained\nto predict engagement for five language datasets. A SHAP analysis combined with\ntransfer learning confirmed a considerable correlation between the importance\nof input features for a language set and the significant cultural\ncharacteristics analyzed.\n", "date": "2024-09-09", "categories": []}
{"id": "2407.12927", "title": "Textualized and Feature-based Models for Compound Multimodal Emotion\n  Recognition in the Wild", "abstract": "  Systems for multimodal emotion recognition (ER) are commonly trained to\nextract features from different modalities (e.g., visual, audio, and textual)\nthat are combined to predict individual basic emotions. However, compound\nemotions often occur in real-world scenarios, and the uncertainty of\nrecognizing such complex emotions over diverse modalities is challenging for\nfeature-based models. As an alternative, emerging large language models (LLMs)\nlike BERT and LLaMA can rely on explicit non-verbal cues that may be translated\nfrom different non-textual modalities (e.g., audio and visual) into text.\nTextualization of modalities augments data with emotional cues to help the LLM\nencode the interconnections between all modalities in a shared text space. In\nsuch text-based models, prior knowledge of ER tasks is leveraged to textualize\nrelevant non-verbal cues such as audio tone from vocal expressions, and action\nunit intensity from facial expressions. Since the pre-trained weights are\npublicly available for many LLMs, training on large-scale datasets is\nunnecessary, allowing to fine-tune for downstream tasks such as compound ER\n(CER). This paper compares the potential of text- and feature-based approaches\nfor compound multimodal ER in videos. Experiments were conducted on the\nchallenging C-EXPR-DB dataset in the wild for CER, and contrasted with results\non the MELD dataset for basic ER. Our results indicate that multimodal\ntextualization provides lower accuracy than feature-based models on C-EXPR-DB,\nwhere text transcripts are captured in the wild. However, higher accuracy can\nbe achieved when the video data has rich transcripts. Our code is available.\n", "date": "2024-07-17", "categories": []}
{"id": "2409.13354", "title": "Recent Advancement of Emotion Cognition in Large Language Models", "abstract": "  Emotion cognition in large language models (LLMs) is crucial for enhancing\nperformance across various applications, such as social media, human-computer\ninteraction, and mental health assessment. We explore the current landscape of\nresearch, which primarily revolves around emotion classification, emotionally\nrich response generation, and Theory of Mind assessments, while acknowledge the\nchallenges like dependency on annotated data and complexity in emotion\nprocessing. In this paper, we present a detailed survey of recent progress in\nLLMs for emotion cognition. We explore key research studies, methodologies,\noutcomes, and resources, aligning them with Ulric Neisser's cognitive stages.\nAdditionally, we outline potential future directions for research in this\nevolving field, including unsupervised learning approaches and the development\nof more complex and interpretable emotion cognition LLMs. We also discuss\nadvanced methods such as contrastive learning used to improve LLMs' emotion\ncognition capabilities.\n", "date": "2024-09-20", "categories": []}
{"id": "2311.16114", "title": "Learning Noise-Robust Joint Representation for Multimodal Emotion\n  Recognition under Incomplete Data Scenarios", "abstract": "  Multimodal emotion recognition (MER) in practical scenarios is significantly\nchallenged by the presence of missing or incomplete data across different\nmodalities. To overcome these challenges, researchers have aimed to simulate\nincomplete conditions during the training phase to enhance the system's overall\nrobustness. Traditional methods have often involved discarding data or\nsubstituting data segments with zero vectors to approximate these\nincompletenesses. However, such approaches neither accurately represent\nreal-world conditions nor adequately address the issue of noisy data\navailability. For instance, a blurry image cannot be simply replaced with zero\nvectors, while still retaining information. To tackle this issue and develop a\nmore precise MER system, we introduce a novel noise-robust MER model that\neffectively learns robust multimodal joint representations from noisy data.\nThis approach includes two pivotal components: firstly, a noise scheduler that\nadjusts the type and level of noise in the data to emulate various realistic\nincomplete situations. Secondly, a Variational AutoEncoder (VAE)-based module\nis employed to reconstruct these robust multimodal joint representations from\nthe noisy inputs. Notably, the introduction of the noise scheduler enables the\nexploration of an entirely new type of incomplete data condition, which is\nimpossible with existing methods. Extensive experimental evaluations on the\nbenchmark datasets IEMOCAP and CMU-MOSEI demonstrate the effectiveness of the\nnoise scheduler and the excellent performance of our proposed model. Our\nproject is publicly available on https://github.com/WooyoohL/Noise-robust_MER.\n", "date": "2023-09-21", "categories": []}
{"id": "2403.13841", "title": "Integrating Wearable Sensor Data and Self-reported Diaries for\n  Personalized Affect Forecasting", "abstract": "  Emotional states, as indicators of affect, are pivotal to overall health,\nmaking their accurate prediction before onset crucial. Current studies are\nprimarily centered on immediate short-term affect detection using data from\nwearable and mobile devices. These studies typically focus on objective sensory\nmeasures, often neglecting other forms of self-reported information like\ndiaries and notes. In this paper, we propose a multimodal deep learning model\nfor affect status forecasting. This model combines a transformer encoder with a\npre-trained language model, facilitating the integrated analysis of objective\nmetrics and self-reported diaries. To validate our model, we conduct a\nlongitudinal study, enrolling college students and monitoring them over a year,\nto collect an extensive dataset including physiological, environmental, sleep,\nmetabolic, and physical activity parameters, alongside open-ended textual\ndiaries provided by the participants. Our results demonstrate that the proposed\nmodel achieves predictive accuracy of 82.50% for positive affect and 82.76% for\nnegative affect, a full week in advance. The effectiveness of our model is\nfurther elevated by its explainability.\n", "date": "2024-03-16", "categories": []}
{"id": "2409.09511", "title": "Explaining Deep Learning Embeddings for Speech Emotion Recognition by\n  Predicting Interpretable Acoustic Features", "abstract": "  Pre-trained deep learning embeddings have consistently shown superior\nperformance over handcrafted acoustic features in speech emotion recognition\n(SER). However, unlike acoustic features with clear physical meaning, these\nembeddings lack clear interpretability. Explaining these embeddings is crucial\nfor building trust in healthcare and security applications and advancing the\nscientific understanding of the acoustic information that is encoded in them.\nThis paper proposes a modified probing approach to explain deep learning\nembeddings in the SER space. We predict interpretable acoustic features (e.g.,\nf0, loudness) from (i) the complete set of embeddings and (ii) a subset of the\nembedding dimensions identified as most important for predicting each emotion.\nIf the subset of the most important dimensions better predicts a given emotion\nthan all dimensions and also predicts specific acoustic features more\naccurately, we infer those acoustic features are important for the embedding\nmodel for the given task. We conducted experiments using the WavLM embeddings\nand eGeMAPS acoustic features as audio representations, applying our method to\nthe RAVDESS and SAVEE emotional speech datasets. Based on this evaluation, we\ndemonstrate that Energy, Frequency, Spectral, and Temporal categories of\nacoustic features provide diminishing information to SER in that order,\ndemonstrating the utility of the probing classifier method to relate embeddings\nto interpretable acoustic features.\n", "date": "2024-09-14", "categories": []}
{"id": "2409.08444", "title": "Towards Unified Facial Action Unit Recognition Framework by Large\n  Language Models", "abstract": "  Facial Action Units (AUs) are of great significance in the realm of affective\ncomputing. In this paper, we propose AU-LLaVA, the first unified AU recognition\nframework based on the Large Language Model (LLM). AU-LLaVA consists of a\nvisual encoder, a linear projector layer, and a pre-trained LLM. We\nmeticulously craft the text descriptions and fine-tune the model on various AU\ndatasets, allowing it to generate different formats of AU recognition results\nfor the same input image. On the BP4D and DISFA datasets, AU-LLaVA delivers the\nmost accurate recognition results for nearly half of the AUs. Our model\nachieves improvements of F1-score up to 11.4% in specific AU recognition\ncompared to previous benchmark results. On the FEAFA dataset, our method\nachieves significant improvements over all 24 AUs compared to previous\nbenchmark results. AU-LLaVA demonstrates exceptional performance and\nversatility in AU recognition.\n", "date": "2024-09-13", "categories": []}
{"id": "2409.08907", "title": "Affective Computing Has Changed: The Foundation Model Disruption", "abstract": "  The dawn of Foundation Models has on the one hand revolutionised a wide range\nof research problems, and, on the other hand, democratised the access and use\nof AI-based tools by the general public. We even observe an incursion of these\nmodels into disciplines related to human psychology, such as the Affective\nComputing domain, suggesting their affective, emerging capabilities. In this\nwork, we aim to raise awareness of the power of Foundation Models in the field\nof Affective Computing by synthetically generating and analysing multimodal\naffective data, focusing on vision, linguistics, and speech (acoustics). We\nalso discuss some fundamental problems, such as ethical issues and regulatory\naspects, related to the use of Foundation Models in this research area.\n", "date": "2024-09-13", "categories": []}
{"id": "2409.07493", "title": "Complex Emotion Recognition System using basic emotions via Facial\n  Expression, EEG, and ECG Signals: a review", "abstract": "  The Complex Emotion Recognition System (CERS) deciphers complex emotional\nstates by examining combinations of basic emotions expressed, their\ninterconnections, and the dynamic variations. Through the utilization of\nadvanced algorithms, CERS provides profound insights into emotional dynamics,\nfacilitating a nuanced understanding and customized responses. The attainment\nof such a level of emotional recognition in machines necessitates the knowledge\ndistillation and the comprehension of novel concepts akin to human cognition.\nThe development of AI systems for discerning complex emotions poses a\nsubstantial challenge with significant implications for affective computing.\nFurthermore, obtaining a sizable dataset for CERS proves to be a daunting task\ndue to the intricacies involved in capturing subtle emotions, necessitating\nspecialized methods for data collection and processing. Incorporating\nphysiological signals such as Electrocardiogram (ECG) and Electroencephalogram\n(EEG) can notably enhance CERS by furnishing valuable insights into the user's\nemotional state, enhancing the quality of datasets, and fortifying system\ndependability. A comprehensive literature review was conducted in this study to\nassess the efficacy of machine learning, deep learning, and meta-learning\napproaches in both basic and complex emotion recognition utilizing EEG, ECG\nsignals, and facial expression datasets. The chosen research papers offer\nperspectives on potential applications, clinical implications, and results of\nCERSs, with the objective of promoting their acceptance and integration into\nclinical decision-making processes. This study highlights research gaps and\nchallenges in understanding CERSs, encouraging further investigation by\nrelevant studies and organizations. Lastly, the significance of meta-learning\napproaches in improving CERS performance and guiding future research endeavors\nis underscored.\n", "date": "2024-09-09", "categories": []}
{"id": "2409.06863", "title": "Towards Understanding Human Emotional Fluctuations with Sparse Check-In\n  Data", "abstract": "  Data sparsity is a key challenge limiting the power of AI tools across\nvarious domains. The problem is especially pronounced in domains that require\nactive user input rather than measurements derived from automated sensors. It\nis a critical barrier to harnessing the full potential of AI in domains\nrequiring active user engagement, such as self-reported mood check-ins, where\ncapturing a continuous picture of emotional states is essential. In this\ncontext, sparse data can hinder efforts to capture the nuances of individual\nemotional experiences such as causes, triggers, and contributing factors.\nExisting methods for addressing data scarcity often rely on heuristics or large\nestablished datasets, favoring deep learning models that lack adaptability to\nnew domains. This paper proposes a novel probabilistic framework that\nintegrates user-centric feedback-based learning, allowing for personalized\npredictions despite limited data. Achieving 60% accuracy in predicting user\nstates among 64 options (chance of 1/64), this framework effectively mitigates\ndata sparsity. It is versatile across various applications, bridging the gap\nbetween theoretical AI research and practical deployment.\n", "date": "2024-09-10", "categories": []}
{"id": "2409.07078", "title": "Multimodal Emotion Recognition with Vision-language Prompting and\n  Modality Dropout", "abstract": "  In this paper, we present our solution for the Second Multimodal Emotion\nRecognition Challenge Track 1(MER2024-SEMI). To enhance the accuracy and\ngeneralization performance of emotion recognition, we propose several methods\nfor Multimodal Emotion Recognition. Firstly, we introduce EmoVCLIP, a model\nfine-tuned based on CLIP using vision-language prompt learning, designed for\nvideo-based emotion recognition tasks. By leveraging prompt learning on CLIP,\nEmoVCLIP improves the performance of pre-trained CLIP on emotional videos.\nAdditionally, to address the issue of modality dependence in multimodal fusion,\nwe employ modality dropout for robust information fusion. Furthermore, to aid\nBaichuan in better extracting emotional information, we suggest using GPT-4 as\nthe prompt for Baichuan. Lastly, we utilize a self-training strategy to\nleverage unlabeled videos. In this process, we use unlabeled videos with\nhigh-confidence pseudo-labels generated by our model and incorporate them into\nthe training set. Experimental results demonstrate that our model ranks 1st in\nthe MER2024-SEMI track, achieving an accuracy of 90.15% on the test set.\n", "date": "2024-09-11", "categories": []}
{"id": "2409.07256", "title": "MRAC Track 1: 2nd Workshop on Multimodal, Generative and Responsible\n  Affective Computing", "abstract": "  With the rapid advancements in multimodal generative technology, Affective\nComputing research has provoked discussion about the potential consequences of\nAI systems equipped with emotional intelligence. Affective Computing involves\nthe design, evaluation, and implementation of Emotion AI and related\ntechnologies aimed at improving people's lives. Designing a computational model\nin affective computing requires vast amounts of multimodal data, including RGB\nimages, video, audio, text, and physiological signals. Moreover, Affective\nComputing research is deeply engaged with ethical considerations at various\nstages-from training emotionally intelligent models on large-scale human data\nto deploying these models in specific applications. Fundamentally, the\ndevelopment of any AI system must prioritize its impact on humans, aiming to\naugment and enhance human abilities rather than replace them, while drawing\ninspiration from human intelligence in a safe and responsible manner. The MRAC\n2024 Track 1 workshop seeks to extend these principles from controlled,\nsmall-scale lab environments to real-world, large-scale contexts, emphasizing\nresponsible development. The workshop also aims to highlight the potential\nimplications of generative technology, along with the ethical consequences of\nits use, to researchers and industry professionals. To the best of our\nknowledge, this is the first workshop series to comprehensively address the\nfull spectrum of multimodal, generative affective computing from a responsible\nAI perspective, and this is the second iteration of this workshop. Webpage:\nhttps://react-ws.github.io/2024/\n", "date": "2024-09-11", "categories": []}
{"id": "2312.05447", "title": "From Static to Dynamic: Adapting Landmark-Aware Image Models for Facial\n  Expression Recognition in Videos", "abstract": "  Dynamic facial expression recognition (DFER) in the wild is still hindered by\ndata limitations, e.g., insufficient quantity and diversity of pose, occlusion\nand illumination, as well as the inherent ambiguity of facial expressions. In\ncontrast, static facial expression recognition (SFER) currently shows much\nhigher performance and can benefit from more abundant high-quality training\ndata. Moreover, the appearance features and dynamic dependencies of DFER remain\nlargely unexplored. To tackle these challenges, we introduce a novel\nStatic-to-Dynamic model (S2D) that leverages existing SFER knowledge and\ndynamic information implicitly encoded in extracted facial landmark-aware\nfeatures, thereby significantly improving DFER performance. Firstly, we build\nand train an image model for SFER, which incorporates a standard Vision\nTransformer (ViT) and Multi-View Complementary Prompters (MCPs) only. Then, we\nobtain our video model (i.e., S2D), for DFER, by inserting Temporal-Modeling\nAdapters (TMAs) into the image model. MCPs enhance facial expression features\nwith landmark-aware features inferred by an off-the-shelf facial landmark\ndetector. And the TMAs capture and model the relationships of dynamic changes\nin facial expressions, effectively extending the pre-trained image model for\nvideos. Notably, MCPs and TMAs only increase a fraction of trainable parameters\n(less than +10\\%) to the original image model. Moreover, we present a novel\nEmotion-Anchors (i.e., reference samples for each emotion category) based\nSelf-Distillation Loss to reduce the detrimental influence of ambiguous emotion\nlabels, further enhancing our S2D. Experiments conducted on popular SFER and\nDFER datasets show that we achieve the state of the art.\n", "date": "2023-12-09", "categories": []}
{"id": "2403.09915", "title": "Robust Light-Weight Facial Affective Behavior Recognition with CLIP", "abstract": "  Human affective behavior analysis aims to delve into human expressions and\nbehaviors to deepen our understanding of human emotions. Basic expression\ncategories (EXPR) and Action Units (AUs) are two essential components in this\nanalysis, which categorize emotions and break down facial movements into\nelemental units, respectively. Despite advancements, existing approaches in\nexpression classification and AU detection often necessitate complex models and\nsubstantial computational resources, limiting their applicability in everyday\nsettings. In this work, we introduce the first lightweight framework adept at\nefficiently tackling both expression classification and AU detection. This\nframework employs a frozen CLIP image encoder alongside a trainable multilayer\nperceptron (MLP), enhanced with Conditional Value at Risk (CVaR) for robustness\nand a loss landscape flattening strategy for improved generalization.\nExperimental results on the Aff-wild2 dataset demonstrate superior performance\nin comparison to the baseline while maintaining minimal computational demands,\noffering a practical solution for affective behavior analysis. The code is\navailable at https://github.com/Purdue-M2/Affective_Behavior_Analysis_M2_PURDUE\n", "date": "2024-03-14", "categories": []}
{"id": "2409.04447", "title": "Leveraging Contrastive Learning and Self-Training for Multimodal Emotion\n  Recognition with Limited Labeled Samples", "abstract": "  The Multimodal Emotion Recognition challenge MER2024 focuses on recognizing\nemotions using audio, language, and visual signals. In this paper, we present\nour submission solutions for the Semi-Supervised Learning Sub-Challenge\n(MER2024-SEMI), which tackles the issue of limited annotated data in emotion\nrecognition. Firstly, to address the class imbalance, we adopt an oversampling\nstrategy. Secondly, we propose a modality representation combinatorial\ncontrastive learning (MR-CCL) framework on the trimodal input data to establish\nrobust initial models. Thirdly, we explore a self-training approach to expand\nthe training set. Finally, we enhance prediction robustness through a\nmulti-classifier weighted soft voting strategy. Our proposed method is\nvalidated to be effective on the MER2024-SEMI Challenge, achieving a weighted\naverage F-score of 88.25% and ranking 6th on the leaderboard. Our project is\navailable at https://github.com/WooyoohL/MER2024-SEMI.\n", "date": "2024-08-23", "categories": []}
{"id": "2409.04723", "title": "NapTune: Efficient Model Tuning for Mood Classification using Previous\n  Night's Sleep Measures along with Wearable Time-series", "abstract": "  Sleep is known to be a key factor in emotional regulation and overall mental\nhealth. In this study, we explore the integration of sleep measures from the\nprevious night into wearable-based mood recognition. To this end, we propose\nNapTune, a novel prompt-tuning framework that utilizes sleep-related measures\nas additional inputs to a frozen pre-trained wearable time-series encoder by\nadding and training lightweight prompt parameters to each Transformer layer.\nThrough rigorous empirical evaluation, we demonstrate that the inclusion of\nsleep data using NapTune not only improves mood recognition performance across\ndifferent wearable time-series namely ECG, PPG, and EDA, but also makes it more\nsample-efficient. Our method demonstrates significant improvements over the\nbest baselines and unimodal variants. Furthermore, we analyze the impact of\nadding sleep-related measures on recognizing different moods as well as the\ninfluence of individual sleep-related measures.\n", "date": "2024-09-07", "categories": []}
{"id": "2409.05007", "title": "Audio-Guided Fusion Techniques for Multimodal Emotion Analysis", "abstract": "  In this paper, we propose a solution for the semi-supervised learning track\n(MER-SEMI) in MER2024. First, in order to enhance the performance of the\nfeature extractor on sentiment classification tasks,we fine-tuned video and\ntext feature extractors, specifically CLIP-vit-large and Baichuan-13B, using\nlabeled data. This approach effectively preserves the original emotional\ninformation conveyed in the videos. Second, we propose an Audio-Guided\nTransformer (AGT) fusion mechanism, which leverages the robustness of\nHubert-large, showing superior effectiveness in fusing both inter-channel and\nintra-channel information. Third, To enhance the accuracy of the model, we\niteratively apply self-supervised learning by using high-confidence unlabeled\ndata as pseudo-labels. Finally, through black-box probing, we discovered an\nimbalanced data distribution between the training and test sets. Therefore, We\nadopt a prior-knowledge-based voting mechanism. The results demonstrate the\neffectiveness of our strategy, ultimately earning us third place in the\nMER-SEMI track.\n", "date": "2024-09-08", "categories": []}
{"id": "2409.05148", "title": "Better Spanish Emotion Recognition In-the-wild: Bringing Attention to\n  Deep Spectrum Voice Analysis", "abstract": "  Within the context of creating new Socially Assistive Robots, emotion\nrecognition has become a key development factor, as it allows the robot to\nadapt to the user's emotional state in the wild. In this work, we focused on\nthe analysis of two voice recording Spanish datasets: ELRA-S0329 and\nEmoMatchSpanishDB. Specifically, we centered our work in the paralanguage,\ne.~g. the vocal characteristics that go along with the message and clarifies\nthe meaning. We proposed the use of the DeepSpectrum method, which consists of\nextracting a visual representation of the audio tracks and feeding them to a\npretrained CNN model. For the classification task, DeepSpectrum is often paired\nwith a Support Vector Classifier --DS-SVC--, or a Fully-Connected deep-learning\nclassifier --DS-FC--. We compared the results of the DS-SVC and DS-FC\narchitectures with the state-of-the-art (SOTA) for ELRA-S0329 and\nEmoMatchSpanishDB. Moreover, we proposed our own classifier based upon\nAttention Mechanisms, namely DS-AM. We trained all models against both\ndatasets, and we found that our DS-AM model outperforms the SOTA models for the\ndatasets and the SOTA DeepSpectrum architectures. Finally, we trained our DS-AM\nmodel in one dataset and tested it in the other, to simulate real-world\nconditions on how biased is the model to the dataset.\n", "date": "2024-09-08", "categories": []}
{"id": "2409.05243", "title": "Mamba-Enhanced Text-Audio-Video Alignment Network for Emotion\n  Recognition in Conversations", "abstract": "  Emotion Recognition in Conversations (ERCs) is a vital area within multimodal\ninteraction research, dedicated to accurately identifying and classifying the\nemotions expressed by speakers throughout a conversation. Traditional ERC\napproaches predominantly rely on unimodal cues\\-such as text, audio, or visual\ndata\\-leading to limitations in their effectiveness. These methods encounter\ntwo significant challenges: 1) Consistency in multimodal information. Before\nintegrating various modalities, it is crucial to ensure that the data from\ndifferent sources is aligned and coherent. 2) Contextual information capture.\nSuccessfully fusing multimodal features requires a keen understanding of the\nevolving emotional tone, especially in lengthy dialogues where emotions may\nshift and develop over time. To address these limitations, we propose a novel\nMamba-enhanced Text-Audio-Video alignment network (MaTAV) for the ERC task.\nMaTAV is with the advantages of aligning unimodal features to ensure\nconsistency across different modalities and handling long input sequences to\nbetter capture contextual multimodal information. The extensive experiments on\nthe MELD and IEMOCAP datasets demonstrate that MaTAV significantly outperforms\nexisting state-of-the-art methods on the ERC task with a big margin.\n", "date": "2024-09-08", "categories": []}
{"id": "2405.15334", "title": "Detection and Positive Reconstruction of Cognitive Distortion sentences:\n  Mandarin Dataset and Evaluation", "abstract": "  This research introduces a Positive Reconstruction Framework based on\npositive psychology theory. Overcoming negative thoughts can be challenging,\nour objective is to address and reframe them through a positive\nreinterpretation. To tackle this challenge, a two-fold approach is necessary:\nidentifying cognitive distortions and suggesting a positively reframed\nalternative while preserving the original thought's meaning. Recent studies\nhave investigated the application of Natural Language Processing (NLP) models\nin English for each stage of this process. In this study, we emphasize the\ntheoretical foundation for the Positive Reconstruction Framework, grounded in\nbroaden-and-build theory. We provide a shared corpus containing 4001 instances\nfor detecting cognitive distortions and 1900 instances for positive\nreconstruction in Mandarin. Leveraging recent NLP techniques, including\ntransfer learning, fine-tuning pretrained networks, and prompt engineering, we\ndemonstrate the effectiveness of automated tools for both tasks. In summary,\nour study contributes to multilingual positive reconstruction, highlighting the\neffectiveness of NLP in cognitive distortion detection and positive\nreconstruction.\n", "date": "2024-05-24", "categories": []}
{"id": "2409.04007", "title": "Searching for Effective Preprocessing Method and CNN-based Architecture\n  with Efficient Channel Attention on Speech Emotion Recognition", "abstract": "  Speech emotion recognition (SER) classifies human emotions in speech with a\ncomputer model. Recently, performance in SER has steadily increased as deep\nlearning techniques have adapted. However, unlike many domains that use speech\ndata, data for training in the SER model is insufficient. This causes\noverfitting of training of the neural network, resulting in performance\ndegradation. In fact, successful emotion recognition requires an effective\npreprocessing method and a model structure that efficiently uses the number of\nweight parameters. In this study, we propose using eight dataset versions with\ndifferent frequency-time resolutions to search for an effective emotional\nspeech preprocessing method. We propose a 6-layer convolutional neural network\n(CNN) model with efficient channel attention (ECA) to pursue an efficient model\nstructure. In particular, the well-positioned ECA blocks can improve channel\nfeature representation with only a few parameters. With the interactive\nemotional dyadic motion capture (IEMOCAP) dataset, increasing the frequency\nresolution in preprocessing emotional speech can improve emotion recognition\nperformance. Also, ECA after the deep convolution layer can effectively\nincrease channel feature representation. Consequently, the best result (79.37UA\n79.68WA) can be obtained, exceeding the performance of previous SER models.\nFurthermore, to compensate for the lack of emotional speech data, we experiment\nwith multiple preprocessing data methods that augment trainable data\npreprocessed with all different settings from one sample. In the experiment, we\ncan achieve the highest result (80.28UA 80.46WA).\n", "date": "2024-09-06", "categories": []}
{"id": "2407.12390", "title": "Enhancing Facial Expression Recognition through Dual-Direction Attention\n  Mixed Feature Networks: Application to 7th ABAW Challenge", "abstract": "  We present our contribution to the 7th ABAW challenge at ECCV 2024, by\nutilizing a Dual-Direction Attention Mixed Feature Network (DDAMFN) for\nmultitask facial expression recognition, we achieve results far beyond the\nproposed baseline for the Multi-Task ABAW challenge. Our proposal uses the\nwell-known DDAMFN architecture as base to effectively predict valence-arousal,\nemotion recognition, and facial action units. We demonstrate the architecture\nability to handle these tasks simultaneously, providing insights into its\narchitecture and the rationale behind its design. Additionally, we compare our\nresults for a multitask solution with independent single-task performance.\n", "date": "2024-07-17", "categories": []}
{"id": "2409.02566", "title": "How Do You Perceive My Face? Recognizing Facial Expressions in\n  Multi-Modal Context by Modeling Mental Representations", "abstract": "  Facial expression perception in humans inherently relies on prior knowledge\nand contextual cues, contributing to efficient and flexible processing. For\ninstance, multi-modal emotional context (such as voice color, affective text,\nbody pose, etc.) can prompt people to perceive emotional expressions in\nobjectively neutral faces. Drawing inspiration from this, we introduce a novel\napproach for facial expression classification that goes beyond simple\nclassification tasks. Our model accurately classifies a perceived face and\nsynthesizes the corresponding mental representation perceived by a human when\nobserving a face in context. With this, our model offers visual insights into\nits internal decision-making process. We achieve this by learning two\nindependent representations of content and context using a VAE-GAN\narchitecture. Subsequently, we propose a novel attention mechanism for\ncontext-dependent feature adaptation. The adapted representation is used for\nclassification and to generate a context-augmented expression. We evaluate\nsynthesized expressions in a human study, showing that our model effectively\nproduces approximations of human mental representations. We achieve\nState-of-the-Art classification accuracies of 81.01% on the RAVDESS dataset and\n79.34% on the MEAD dataset. We make our code publicly available.\n", "date": "2024-09-04", "categories": []}
{"id": "2312.10579", "title": "DER-GCN: Dialogue and Event Relation-Aware Graph Convolutional Neural\n  Network for Multimodal Dialogue Emotion Recognition", "abstract": "  With the continuous development of deep learning (DL), the task of multimodal\ndialogue emotion recognition (MDER) has recently received extensive research\nattention, which is also an essential branch of DL. The MDER aims to identify\nthe emotional information contained in different modalities, e.g., text, video,\nand audio, in different dialogue scenes. However, existing research has focused\non modeling contextual semantic information and dialogue relations between\nspeakers while ignoring the impact of event relations on emotion. To tackle the\nabove issues, we propose a novel Dialogue and Event Relation-Aware Graph\nConvolutional Neural Network for Multimodal Emotion Recognition (DER-GCN)\nmethod. It models dialogue relations between speakers and captures latent event\nrelations information. Specifically, we construct a weighted multi-relationship\ngraph to simultaneously capture the dependencies between speakers and event\nrelations in a dialogue. Moreover, we also introduce a Self-Supervised Masked\nGraph Autoencoder (SMGAE) to improve the fusion representation ability of\nfeatures and structures. Next, we design a new Multiple Information Transformer\n(MIT) to capture the correlation between different relations, which can provide\na better fuse of the multivariate information between relations. Finally, we\npropose a loss optimization strategy based on contrastive learning to enhance\nthe representation learning ability of minority class features. We conduct\nextensive experiments on the IEMOCAP and MELD benchmark datasets, which verify\nthe effectiveness of the DER-GCN model. The results demonstrate that our model\nsignificantly improves both the average accuracy and the f1 value of emotion\nrecognition.\n", "date": "2023-12-17", "categories": []}
{"id": "2312.16778", "title": "Adversarial Representation with Intra-Modal and Inter-Modal Graph\n  Contrastive Learning for Multimodal Emotion Recognition", "abstract": "  With the release of increasing open-source emotion recognition datasets on\nsocial media platforms and the rapid development of computing resources,\nmultimodal emotion recognition tasks (MER) have begun to receive widespread\nresearch attention. The MER task extracts and fuses complementary semantic\ninformation from different modalities, which can classify the speaker's\nemotions. However, the existing feature fusion methods have usually mapped the\nfeatures of different modalities into the same feature space for information\nfusion, which can not eliminate the heterogeneity between different modalities.\nTherefore, it is challenging to make the subsequent emotion class boundary\nlearning. To tackle the above problems, we have proposed a novel Adversarial\nRepresentation with Intra-Modal and Inter-Modal Graph Contrastive for\nMultimodal Emotion Recognition (AR-IIGCN) method. Firstly, we input video,\naudio, and text features into a multi-layer perceptron (MLP) to map them into\nseparate feature spaces. Secondly, we build a generator and a discriminator for\nthe three modal features through adversarial representation, which can achieve\ninformation interaction between modalities and eliminate heterogeneity among\nmodalities. Thirdly, we introduce contrastive graph representation learning to\ncapture intra-modal and inter-modal complementary semantic information and\nlearn intra-class and inter-class boundary information of emotion categories.\nSpecifically, we construct a graph structure for three modal features and\nperform contrastive representation learning on nodes with different emotions in\nthe same modality and the same emotion in different modalities, which can\nimprove the feature representation ability of nodes. Extensive experimental\nworks show that the ARL-IIGCN method can significantly improve emotion\nrecognition accuracy on IEMOCAP and MELD datasets.\n", "date": "2023-12-28", "categories": []}
{"id": "2407.00119", "title": "Efficient Long-distance Latent Relation-aware Graph Neural Network for\n  Multi-modal Emotion Recognition in Conversations", "abstract": "  The task of multi-modal emotion recognition in conversation (MERC) aims to\nanalyze the genuine emotional state of each utterance based on the multi-modal\ninformation in the conversation, which is crucial for conversation\nunderstanding. Existing methods focus on using graph neural networks (GNN) to\nmodel conversational relationships and capture contextual latent semantic\nrelationships. However, due to the complexity of GNN, existing methods cannot\nefficiently capture the potential dependencies between long-distance\nutterances, which limits the performance of MERC. In this paper, we propose an\nEfficient Long-distance Latent Relation-aware Graph Neural Network (ELR-GNN)\nfor multi-modal emotion recognition in conversations. Specifically, we first\nuse pre-extracted text, video and audio features as input to Bi-LSTM to capture\ncontextual semantic information and obtain low-level utterance features. Then,\nwe use low-level utterance features to construct a conversational emotion\ninteraction graph. To efficiently capture the potential dependencies between\nlong-distance utterances, we use the dilated generalized forward push algorithm\nto precompute the emotional propagation between global utterances and design an\nemotional relation-aware operator to capture the potential semantic\nassociations between different utterances. Furthermore, we combine early fusion\nand adaptive late fusion mechanisms to fuse latent dependency information\nbetween speaker relationship information and context. Finally, we obtain\nhigh-level discourse features and feed them into MLP for emotion prediction.\nExtensive experimental results show that ELR-GNN achieves state-of-the-art\nperformance on the benchmark datasets IEMOCAP and MELD, with running times\nreduced by 52\\% and 35\\%, respectively.\n", "date": "2024-06-27", "categories": []}
{"id": "2407.05219", "title": "Flood of Techniques and Drought of Theories: Emotion Mining in Disasters", "abstract": "  Emotion mining has become a crucial tool for understanding human emotions\nduring disasters, leveraging the extensive data generated on social media\nplatforms. This paper aims to summarize existing research on emotion mining\nwithin disaster contexts, highlighting both significant discoveries and\npersistent issues. On the one hand, emotion mining techniques have achieved\nacceptable accuracy enabling applications such as rapid damage assessment and\nmental health surveillance. On the other hand, with many studies adopting\ndata-driven approaches, several methodological issues remain. These include\narbitrary emotion classification, ignoring biases inherent in data collection\nfrom social media, such as the overrepresentation of individuals from higher\nsocioeconomic status on Twitter, and the lack of application of theoretical\nframeworks like cross-cultural comparisons. These problems can be summarized as\na notable lack of theory-driven research and ignoring insights from social and\nbehavioral sciences. This paper underscores the need for interdisciplinary\ncollaboration between computer scientists and social scientists to develop more\nrobust and theoretically grounded approaches in emotion mining. By addressing\nthese gaps, we aim to enhance the effectiveness and reliability of emotion\nmining methodologies, ultimately contributing to improved disaster\npreparedness, response, and recovery.\n  Keywords: emotion mining, sentiment analysis, natural disasters, psychology,\ntechnological disasters\n", "date": "2024-07-07", "categories": []}
{"id": "2407.09950", "title": "PSO Fuzzy XGBoost Classifier Boosted with Neural Gas Features on EEG\n  Signals in Emotion Recognition", "abstract": "  Emotion recognition is the technology-driven process of identifying and\ncategorizing human emotions from various data sources, such as facial\nexpressions, voice patterns, body motion, and physiological signals, such as\nEEG. These physiological indicators, though rich in data, present challenges\ndue to their complexity and variability, necessitating sophisticated feature\nselection and extraction methods. NGN, an unsupervised learning algorithm,\neffectively adapts to input spaces without predefined grid structures,\nimproving feature extraction from physiological data. Furthermore, the\nincorporation of fuzzy logic enables the handling of fuzzy data by introducing\nreasoning that mimics human decision-making. The combination of PSO with\nXGBoost aids in optimizing model performance through efficient hyperparameter\ntuning and decision process optimization. This study explores the integration\nof Neural-Gas Network (NGN), XGBoost, Particle Swarm Optimization (PSO), and\nfuzzy logic to enhance emotion recognition using physiological signals. Our\nresearch addresses three critical questions concerning the improvement of\nXGBoost with PSO and fuzzy logic, NGN's effectiveness in feature selection, and\nthe performance comparison of the PSO-fuzzy XGBoost classifier with standard\nbenchmarks. Acquired results indicate that our methodologies enhance the\naccuracy of emotion recognition systems and outperform other feature selection\ntechniques using the majority of classifiers, offering significant implications\nfor both theoretical advancement and practical application in emotion\nrecognition technology.\n", "date": "2024-07-13", "categories": []}
{"id": "2408.16907", "title": "Ig3D: Integrating 3D Face Representations in Facial Expression Inference", "abstract": "  Reconstructing 3D faces with facial geometry from single images has allowed\nfor major advances in animation, generative models, and virtual reality.\nHowever, this ability to represent faces with their 3D features is not as fully\nexplored by the facial expression inference (FEI) community. This study\ntherefore aims to investigate the impacts of integrating such 3D\nrepresentations into the FEI task, specifically for facial expression\nclassification and face-based valence-arousal (VA) estimation. To accomplish\nthis, we first assess the performance of two 3D face representations (both\nbased on the 3D morphable model, FLAME) for the FEI tasks. We further explore\ntwo fusion architectures, intermediate fusion and late fusion, for integrating\nthe 3D face representations with existing 2D inference frameworks. To evaluate\nour proposed architecture, we extract the corresponding 3D representations and\nperform extensive tests on the AffectNet and RAF-DB datasets. Our experimental\nresults demonstrate that our proposed method outperforms the state-of-the-art\nAffectNet VA estimation and RAF-DB classification tasks. Moreover, our method\ncan act as a complement to other existing methods to boost performance in many\nemotion inference tasks.\n", "date": "2024-08-29", "categories": []}
{"id": "2408.17026", "title": "From Text to Emotion: Unveiling the Emotion Annotation Capabilities of\n  LLMs", "abstract": "  Training emotion recognition models has relied heavily on human annotated\ndata, which present diversity, quality, and cost challenges. In this paper, we\nexplore the potential of Large Language Models (LLMs), specifically GPT4, in\nautomating or assisting emotion annotation. We compare GPT4 with supervised\nmodels and or humans in three aspects: agreement with human annotations,\nalignment with human perception, and impact on model training. We find that\ncommon metrics that use aggregated human annotations as ground truth can\nunderestimate the performance, of GPT-4 and our human evaluation experiment\nreveals a consistent preference for GPT-4 annotations over humans across\nmultiple datasets and evaluators. Further, we investigate the impact of using\nGPT-4 as an annotation filtering process to improve model training. Together,\nour findings highlight the great potential of LLMs in emotion annotation tasks\nand underscore the need for refined evaluation methodologies.\n", "date": "2024-08-30", "categories": []}
{"id": "2309.11911", "title": "InstructERC: Reforming Emotion Recognition in Conversation with\n  Multi-task Retrieval-Augmented Large Language Models", "abstract": "  The field of emotion recognition of conversation (ERC) has been focusing on\nseparating sentence feature encoding and context modeling, lacking exploration\nin generative paradigms based on unified designs. In this study, we propose a\nnovel approach, InstructERC, to reformulate the ERC task from a discriminative\nframework to a generative framework based on Large Language Models (LLMs).\nInstructERC makes three significant contributions: (1) it introduces a simple\nyet effective retrieval template module, which helps the model explicitly\nintegrate multi-granularity dialogue supervision information. (2) We introduce\ntwo additional emotion alignment tasks, namely speaker identification and\nemotion prediction tasks, to implicitly model the dialogue role relationships\nand future emotional tendencies in conversations. (3) Pioneeringly, we unify\nemotion labels across benchmarks through the feeling wheel to fit real\napplication scenarios. InstructERC still perform impressively on this unified\ndataset. Our LLM-based plugin framework significantly outperforms all previous\nmodels and achieves comprehensive SOTA on three commonly used ERC datasets.\nExtensive analysis of parameter-efficient and data-scaling experiments provides\nempirical guidance for applying it in practical scenarios.\n", "date": "2023-09-21", "categories": []}
{"id": "2408.15255", "title": "Emotion Classification from Multi-Channel EEG Signals Using HiSTN: A\n  Hierarchical Graph-based Spatial-Temporal Approach", "abstract": "  This study introduces a parameter-efficient Hierarchical Spatial Temporal\nNetwork (HiSTN) specifically designed for the task of emotion classification\nusing multi-channel electroencephalogram data. The network incorporates a graph\nhierarchy constructed from bottom-up at various abstraction levels, offering\nthe dual advantages of enhanced task-relevant deep feature extraction and a\nlightweight design. The model's effectiveness is further amplified when used in\nconjunction with a proposed unique label smoothing method. Comprehensive\nbenchmark experiments reveal that this combined approach yields high, balanced\nperformance in terms of both quantitative and qualitative predictions. HiSTN,\nwhich has approximately 1,000 parameters, achieves mean F1 scores of 96.82%\n(valence) and 95.62% (arousal) in subject-dependent tests on the\nrarely-utilized 5-classification task problem from the DREAMER dataset. In the\nsubject-independent settings, the same model yields mean F1 scores of 78.34%\nfor valence and 81.59% for arousal. The adoption of the Sequential Top-2 Hit\nRate (Seq2HR) metric highlights the significant enhancements in terms of the\nbalance between model's quantitative and qualitative for predictions achieved\nthrough our approach when compared to training with regular one-hot labels.\nThese improvements surpass 50% in subject-dependent tasks and 30% in\nsubject-independent tasks. The study also includes relevant ablation studies\nand case explorations to further elucidate the workings of the proposed model\nand enhance its interpretability.\n", "date": "2024-08-09", "categories": []}
{"id": "2408.15276", "title": "A Survey of Deep Learning for Group-level Emotion Recognition", "abstract": "  With the advancement of artificial intelligence (AI) technology, group-level\nemotion recognition (GER) has emerged as an important area in analyzing human\nbehavior. Early GER methods are primarily relied on handcrafted features.\nHowever, with the proliferation of Deep Learning (DL) techniques and their\nremarkable success in diverse tasks, neural networks have garnered increasing\ninterest in GER. Unlike individual's emotion, group emotions exhibit diversity\nand dynamics. Presently, several DL approaches have been proposed to\neffectively leverage the rich information inherent in group-level image and\nenhance GER performance significantly. In this survey, we present a\ncomprehensive review of DL techniques applied to GER, proposing a new taxonomy\nfor the field cover all aspects of GER based on DL. The survey overviews\ndatasets, the deep GER pipeline, and performance comparisons of the\nstate-of-the-art methods past decade. Moreover, it summarizes and discuss the\nfundamental approaches and advanced developments for each aspect. Furthermore,\nwe identify outstanding challenges and suggest potential avenues for the design\nof robust GER systems. To the best of our knowledge, thus survey represents the\nfirst comprehensive review of deep GER methods, serving as a pivotal references\nfor future GER research endeavors.\n", "date": "2024-08-13", "categories": []}
{"id": "2408.15354", "title": "What Is Required for Empathic AI? It Depends, and Why That Matters for\n  AI Developers and Users", "abstract": "  Interest is growing in artificial empathy, but so is confusion about what\nartificial empathy is or needs to be. This confusion makes it challenging to\nnavigate the technical and ethical issues that accompany empathic AI\ndevelopment. Here, we outline a framework for thinking about empathic AI based\non the premise that different constellations of capabilities associated with\nempathy are important for different empathic AI applications. We describe\ndistinctions of capabilities that we argue belong under the empathy umbrella,\nand show how three medical empathic AI use cases require different sets of\nthese capabilities. We conclude by discussing why appreciation of the diverse\ncapabilities under the empathy umbrella is important for both AI creators and\nusers.\n", "date": "2024-08-27", "categories": []}
{"id": "2305.00769", "title": "Multi-scale Transformer-based Network for Emotion Recognition from Multi\n  Physiological Signals", "abstract": "  This paper presents an efficient Multi-scale Transformer-based approach for\nthe task of Emotion recognition from Physiological data, which has gained\nwidespread attention in the research community due to the vast amount of\ninformation that can be extracted from these signals using modern sensors and\nmachine learning techniques. Our approach involves applying a Multi-modal\ntechnique combined with scaling data to establish the relationship between\ninternal body signals and human emotions. Additionally, we utilize Transformer\nand Gaussian Transformation techniques to improve signal encoding effectiveness\nand overall performance. Our model achieves decent results on the CASE dataset\nof the EPiC competition, with an RMSE score of 1.45.\n", "date": "2023-05-01", "categories": []}
{"id": "2402.11886", "title": "The Colorful Future of LLMs: Evaluating and Improving LLMs as Emotional\n  Supporters for Queer Youth", "abstract": "  Queer youth face increased mental health risks, such as depression, anxiety,\nand suicidal ideation. Hindered by negative stigma, they often avoid seeking\nhelp and rely on online resources, which may provide incompatible information.\nAlthough access to a supportive environment and reliable information is\ninvaluable, many queer youth worldwide have no access to such support. However,\nthis could soon change due to the rapid adoption of Large Language Models\n(LLMs) such as ChatGPT. This paper aims to comprehensively explore the\npotential of LLMs to revolutionize emotional support for queers. To this end,\nwe conduct a qualitative and quantitative analysis of LLM's interactions with\nqueer-related content. To evaluate response quality, we develop a novel\nten-question scale that is inspired by psychological standards and expert\ninput. We apply this scale to score several LLMs and human comments to posts\nwhere queer youth seek advice and share experiences. We find that LLM responses\nare supportive and inclusive, outscoring humans. However, they tend to be\ngeneric, not empathetic enough, and lack personalization, resulting in\nnonreliable and potentially harmful advice. We discuss these challenges,\ndemonstrate that a dedicated prompt can improve the performance, and propose a\nblueprint of an LLM-supporter that actively (but sensitively) seeks user\ncontext to provide personalized, empathetic, and reliable responses. Our\nannotated dataset is available for further research.\n", "date": "2024-02-19", "categories": []}
{"id": "2408.14842", "title": "From Bias to Balance: Detecting Facial Expression Recognition Biases in\n  Large Multimodal Foundation Models", "abstract": "  This study addresses the racial biases in facial expression recognition (FER)\nsystems within Large Multimodal Foundation Models (LMFMs). Despite advances in\ndeep learning and the availability of diverse datasets, FER systems often\nexhibit higher error rates for individuals with darker skin tones. Existing\nresearch predominantly focuses on traditional FER models (CNNs, RNNs, ViTs),\nleaving a gap in understanding racial biases in LMFMs. We benchmark four\nleading LMFMs: GPT-4o, PaliGemma, Gemini, and CLIP to assess their performance\nin facial emotion detection across different racial demographics. A linear\nclassifier trained on CLIP embeddings obtains accuracies of 95.9\\% for RADIATE,\n90.3\\% for Tarr, and 99.5\\% for Chicago Face. Furthermore, we identify that\nAnger is misclassified as Disgust 2.1 times more often in Black Females than\nWhite Females. This study highlights the need for fairer FER systems and\nestablishes a foundation for developing unbiased, accurate FER technologies.\nVisit https://kvjvhub.github.io/FERRacialBias/ for further information\nregarding the biases within facial expression recognition.\n", "date": "2024-08-27", "categories": []}
{"id": "2401.08508", "title": "EmoLLMs: A Series of Emotional Large Language Models and Annotation\n  Tools for Comprehensive Affective Analysis", "abstract": "  Sentiment analysis and emotion detection are important research topics in\nnatural language processing (NLP) and benefit many downstream tasks. With the\nwidespread application of LLMs, researchers have started exploring the\napplication of LLMs based on instruction-tuning in the field of sentiment\nanalysis. However, these models only focus on single aspects of affective\nclassification tasks (e.g. sentimental polarity or categorical emotions), and\noverlook the regression tasks (e.g. sentiment strength or emotion intensity),\nwhich leads to poor performance in downstream tasks. The main reason is the\nlack of comprehensive affective instruction tuning datasets and evaluation\nbenchmarks, which cover various affective classification and regression tasks.\nMoreover, although emotional information is useful for downstream tasks,\nexisting downstream datasets lack high-quality and comprehensive affective\nannotations. In this paper, we propose EmoLLMs, the first series of\nopen-sourced instruction-following LLMs for comprehensive affective analysis\nbased on fine-tuning various LLMs with instruction data, the first multi-task\naffective analysis instruction dataset (AAID) with 234K data samples based on\nvarious classification and regression tasks to support LLM instruction tuning,\nand a comprehensive affective evaluation benchmark (AEB) with 14 tasks from\nvarious sources and domains to test the generalization ability of LLMs. We\npropose a series of EmoLLMs by fine-tuning LLMs with AAID to solve various\naffective instruction tasks. We compare our model with a variety of LLMs on\nAEB, where our models outperform all other open-sourced LLMs, and surpass\nChatGPT and GPT-4 in most tasks, which shows that the series of EmoLLMs achieve\nthe ChatGPT-level and GPT-4-level generalization capabilities on affective\nanalysis tasks, and demonstrates our models can be used as affective annotation\ntools.\n", "date": "2024-01-16", "categories": []}
{"id": "2406.10265", "title": "Improving Language Models for Emotion Analysis: Insights from Cognitive\n  Science", "abstract": "  We propose leveraging cognitive science research on emotions and\ncommunication to improve language models for emotion analysis. First, we\npresent the main emotion theories in psychology and cognitive science. Then, we\nintroduce the main methods of emotion annotation in natural language processing\nand their connections to psychological theories. We also present the two main\ntypes of analyses of emotional communication in cognitive pragmatics. Finally,\nbased on the cognitive science research presented, we propose directions for\nimproving language models for emotion analysis. We suggest that these research\nefforts pave the way for constructing new annotation schemes, methods, and a\npossible benchmark for emotional understanding, considering different facets of\nhuman emotion and communication.\n", "date": "2024-06-11", "categories": []}
{"id": "2408.13718", "title": "GPT-4 Emulates Average-Human Emotional Cognition from a Third-Person\n  Perspective", "abstract": "  This paper extends recent investigations on the emotional reasoning abilities\nof Large Language Models (LLMs). Current research on LLMs has not directly\nevaluated the distinction between how LLMs predict the self-attribution of\nemotions and the perception of others' emotions. We first look at carefully\ncrafted emotion-evoking stimuli, originally designed to find patterns of brain\nneural activity representing fine-grained inferred emotional attributions of\nothers. We show that GPT-4 is especially accurate in reasoning about such\nstimuli. This suggests LLMs agree with humans' attributions of others' emotions\nin stereotypical scenarios remarkably more than self-attributions of emotions\nin idiosyncratic situations. To further explore this, our second study utilizes\na dataset containing annotations from both the author and a third-person\nperspective. We find that GPT-4's interpretations align more closely with human\njudgments about the emotions of others than with self-assessments. Notably,\nconventional computational models of emotion primarily rely on self-reported\nground truth as the gold standard. However, an average observer's standpoint,\nwhich LLMs appear to have adopted, might be more relevant for many downstream\napplications, at least in the absence of individual information and adequate\nsafety considerations.\n", "date": "2024-08-11", "categories": []}
{"id": "2408.12844", "title": "Predicting Affective States from Screen Text Sentiment", "abstract": "  The proliferation of mobile sensing technologies has enabled the study of\nvarious physiological and behavioural phenomena through unobtrusive data\ncollection from smartphone sensors. This approach offers real-time insights\ninto individuals' physical and mental states, creating opportunities for\npersonalised treatment and interventions. However, the potential of analysing\nthe textual content viewed on smartphones to predict affective states remains\nunderexplored. To better understand how the screen text that users are exposed\nto and interact with can influence their affects, we investigated a subset of\ndata obtained from a digital phenotyping study of Australian university\nstudents conducted in 2023. We employed linear regression, zero-shot, and\nmulti-shot prompting using a large language model (LLM) to analyse\nrelationships between screen text and affective states. Our findings indicate\nthat multi-shot prompting substantially outperforms both linear regression and\nzero-shot prompting, highlighting the importance of context in affect\nprediction. We discuss the value of incorporating textual and sentiment data\nfor improving affect prediction, providing a basis for future advancements in\nunderstanding smartphone use and wellbeing.\n", "date": "2024-08-23", "categories": []}
{"id": "2408.10500", "title": "SZTU-CMU at MER2024: Improving Emotion-LLaMA with Conv-Attention for\n  Multimodal Emotion Recognition", "abstract": "  This paper presents our winning approach for the MER-NOISE and MER-OV tracks\nof the MER2024 Challenge on multimodal emotion recognition. Our system\nleverages the advanced emotional understanding capabilities of Emotion-LLaMA to\ngenerate high-quality annotations for unlabeled samples, addressing the\nchallenge of limited labeled data. To enhance multimodal fusion while\nmitigating modality-specific noise, we introduce Conv-Attention, a lightweight\nand efficient hybrid framework. Extensive experimentation vali-dates the\neffectiveness of our approach. In the MER-NOISE track, our system achieves a\nstate-of-the-art weighted average F-score of 85.30%, surpassing the second and\nthird-place teams by 1.47% and 1.65%, respectively. For the MER-OV track, our\nutilization of Emotion-LLaMA for open-vocabulary annotation yields an 8.52%\nimprovement in average accuracy and recall compared to GPT-4V, securing the\nhighest score among all participating large multimodal models. The code and\nmodel for Emotion-LLaMA are available at\nhttps://github.com/ZebangCheng/Emotion-LLaMA.\n", "date": "2024-08-20", "categories": []}
{"id": "2408.11286", "title": "Video Emotion Open-vocabulary Recognition Based on Multimodal Large\n  Language Model", "abstract": "  Multimodal emotion recognition is a task of great concern. However,\ntraditional data sets are based on fixed labels, resulting in models that often\nfocus on main emotions and ignore detailed emotional changes in complex scenes.\nThis report introduces the solution of using MLLMs technology to generate\nopen-vocabulary emotion labels from a video. The solution includes the use of\nframework, data generation and processing, training methods, results generation\nand multi-model co-judgment. In the MER-OV (Open-Word Emotion Recognition) of\nthe MER2024 challenge, our method achieved significant advantages, leading to\nits superior capabilities in complex emotion computation.\n", "date": "2024-08-21", "categories": []}
{"id": "2408.12121", "title": "Emotion-Agent: Unsupervised Deep Reinforcement Learning with\n  Distribution-Prototype Reward for Continuous Emotional EEG Analysis", "abstract": "  Continuous electroencephalography (EEG) signals are widely used in affective\nbrain-computer interface (aBCI) applications. However, not all continuously\ncollected EEG signals are relevant or meaningful to the task at hand (e.g.,\nwondering thoughts). On the other hand, manually labeling the relevant parts is\nnearly impossible due to varying engagement patterns across different tasks and\nindividuals. Therefore, effectively and efficiently identifying the important\nparts from continuous EEG recordings is crucial for downstream BCI tasks, as it\ndirectly impacts the accuracy and reliability of the results. In this paper, we\npropose a novel unsupervised deep reinforcement learning framework, called\nEmotion-Agent, to automatically identify relevant and informative emotional\nmoments from continuous EEG signals. Specifically, Emotion-Agent involves\nunsupervised deep reinforcement learning combined with a heuristic algorithm.\nWe first use the heuristic algorithm to perform an initial global search and\nform prototype representations of the EEG signals, which facilitates the\nefficient exploration of the signal space and identify potential regions of\ninterest. Then, we design distribution-prototype reward functions to estimate\nthe interactions between samples and prototypes, ensuring that the identified\nparts are both relevant and representative of the underlying emotional states.\nEmotion-Agent is trained using Proximal Policy Optimization (PPO) to achieve\nstable and efficient convergence. Our experiments compare the performance with\nand without Emotion-Agent. The results demonstrate that selecting relevant and\ninformative emotional parts before inputting them into downstream tasks\nenhances the accuracy and reliability of aBCI applications.\n", "date": "2024-08-22", "categories": []}
{"id": "2408.12124", "title": "Recording Brain Activity While Listening to Music Using Wearable EEG\n  Devices Combined with Bidirectional Long Short-Term Memory Networks", "abstract": "  Electroencephalography (EEG) signals are crucial for investigating brain\nfunction and cognitive processes. This study aims to address the challenges of\nefficiently recording and analyzing high-dimensional EEG signals while\nlistening to music to recognize emotional states. We propose a method combining\nBidirectional Long Short-Term Memory (Bi-LSTM) networks with attention\nmechanisms for EEG signal processing. Using wearable EEG devices, we collected\nbrain activity data from participants listening to music. The data was\npreprocessed, segmented, and Differential Entropy (DE) features were extracted.\nWe then constructed and trained a Bi-LSTM model to enhance key feature\nextraction and improve emotion recognition accuracy. Experiments were conducted\non the SEED and DEAP datasets. The Bi-LSTM-AttGW model achieved 98.28% accuracy\non the SEED dataset and 92.46% on the DEAP dataset in multi-class emotion\nrecognition tasks, significantly outperforming traditional models such as SVM\nand EEG-Net. This study demonstrates the effectiveness of combining Bi-LSTM\nwith attention mechanisms, providing robust technical support for applications\nin brain-computer interfaces (BCI) and affective computing. Future work will\nfocus on improving device design, incorporating multimodal data, and further\nenhancing emotion recognition accuracy, aiming to achieve practical\napplications in real-world scenarios.\n", "date": "2024-08-22", "categories": []}
{"id": "2408.11424", "title": "EMO-LLaMA: Enhancing Facial Emotion Understanding with Instruction\n  Tuning", "abstract": "  Facial expression recognition (FER) is an important research topic in\nemotional artificial intelligence. In recent decades, researchers have made\nremarkable progress. However, current FER paradigms face challenges in\ngeneralization, lack semantic information aligned with natural language, and\nstruggle to process both images and videos within a unified framework, making\ntheir application in multimodal emotion understanding and human-computer\ninteraction difficult. Multimodal Large Language Models (MLLMs) have recently\nachieved success, offering advantages in addressing these issues and\npotentially overcoming the limitations of current FER paradigms. However,\ndirectly applying pre-trained MLLMs to FER still faces several challenges. Our\nzero-shot evaluations of existing open-source MLLMs on FER indicate a\nsignificant performance gap compared to GPT-4V and current supervised\nstate-of-the-art (SOTA) methods. In this paper, we aim to enhance MLLMs'\ncapabilities in understanding facial expressions. We first generate instruction\ndata for five FER datasets with Gemini. We then propose a novel MLLM, named\nEMO-LLaMA, which incorporates facial priors from a pretrained facial analysis\nnetwork to enhance human facial information. Specifically, we design a Face\nInfo Mining module to extract both global and local facial information.\nAdditionally, we utilize a handcrafted prompt to introduce age-gender-race\nattributes, considering the emotional differences across different human\ngroups. Extensive experiments show that EMO-LLaMA achieves SOTA-comparable or\ncompetitive results across both static and dynamic FER datasets. The\ninstruction dataset and code are available at\nhttps://github.com/xxtars/EMO-LLaMA.\n", "date": "2024-08-21", "categories": []}
{"id": "2408.10248", "title": "Target-Dependent Multimodal Sentiment Analysis Via Employing Visual-to\n  Emotional-Caption Translation Network using Visual-Caption Pairs", "abstract": "  The natural language processing and multimedia field has seen a notable surge\nin interest in multimodal sentiment recognition. Hence, this study aims to\nemploy Target-Dependent Multimodal Sentiment Analysis (TDMSA) to identify the\nlevel of sentiment associated with every target (aspect) stated within a\nmultimodal post consisting of a visual-caption pair. Despite the recent\nadvancements in multimodal sentiment recognition, there has been a lack of\nexplicit incorporation of emotional clues from the visual modality,\nspecifically those pertaining to facial expressions. The challenge at hand is\nto proficiently obtain visual and emotional clues and subsequently synchronise\nthem with the textual content. In light of this fact, this study presents a\nnovel approach called the Visual-to-Emotional-Caption Translation Network\n(VECTN) technique. The primary objective of this strategy is to effectively\nacquire visual sentiment clues by analysing facial expressions. Additionally,\nit effectively aligns and blends the obtained emotional clues with the target\nattribute of the caption mode. The experimental findings demonstrate that our\nmethodology is capable of producing ground-breaking outcomes when applied to\ntwo publicly accessible multimodal Twitter datasets, namely, Twitter-2015 and\nTwitter-2017. The experimental results show that the suggested model achieves\nan accuracy of 81.23% and a macro-F1 of 80.61% on the Twitter-15 dataset, while\n77.42% and 75.19% on the Twitter-17 dataset, respectively. The observed\nimprovement in performance reveals that our model is better than others when it\ncomes to collecting target-level sentiment in multimodal data using the\nexpressions of the face.\n", "date": "2024-08-05", "categories": []}
{"id": "2408.09438", "title": "Enhancing Modal Fusion by Alignment and Label Matching for Multimodal\n  Emotion Recognition", "abstract": "  To address the limitation in multimodal emotion recognition (MER) performance\narising from inter-modal information fusion, we propose a novel MER framework\nbased on multitask learning where fusion occurs after alignment, called\nFoal-Net. The framework is designed to enhance the effectiveness of modality\nfusion and includes two auxiliary tasks: audio-video emotion alignment (AVEL)\nand cross-modal emotion label matching (MEM). First, AVEL achieves alignment of\nemotional information in audio-video representations through contrastive\nlearning. Then, a modal fusion network integrates the aligned features.\nMeanwhile, MEM assesses whether the emotions of the current sample pair are the\nsame, providing assistance for modal information fusion and guiding the model\nto focus more on emotional information. The experimental results conducted on\nIEMOCAP corpus show that Foal-Net outperforms the state-of-the-art methods and\nemotion alignment is necessary before modal fusion.\n", "date": "2024-08-18", "categories": []}
{"id": "2406.08068", "title": "Large Language Models Meet Text-Centric Multimodal Sentiment Analysis: A\n  Survey", "abstract": "  Compared to traditional sentiment analysis, which only considers text,\nmultimodal sentiment analysis needs to consider emotional signals from\nmultimodal sources simultaneously and is therefore more consistent with the way\nhow humans process sentiment in real-world scenarios. It involves processing\nemotional information from various sources such as natural language, images,\nvideos, audio, physiological signals, etc. However, although other modalities\nalso contain diverse emotional cues, natural language usually contains richer\ncontextual information and therefore always occupies a crucial position in\nmultimodal sentiment analysis. The emergence of ChatGPT has opened up immense\npotential for applying large language models (LLMs) to text-centric multimodal\ntasks. However, it is still unclear how existing LLMs can adapt better to\ntext-centric multimodal sentiment analysis tasks. This survey aims to (1)\npresent a comprehensive review of recent research in text-centric multimodal\nsentiment analysis tasks, (2) examine the potential of LLMs for text-centric\nmultimodal sentiment analysis, outlining their approaches, advantages, and\nlimitations, (3) summarize the application scenarios of LLM-based multimodal\nsentiment analysis technology, and (4) explore the challenges and potential\nresearch directions for multimodal sentiment analysis in the future.\n", "date": "2024-06-12", "categories": []}
{"id": "2408.07704", "title": "Empathic Responding for Digital Interpersonal Emotion Regulation via\n  Content Recommendation", "abstract": "  Interpersonal communication plays a key role in managing people's emotions,\nespecially on digital platforms. Studies have shown that people use social\nmedia and consume online content to regulate their emotions and find support\nfor rest and recovery. However, these platforms are not designed for emotion\nregulation, which limits their effectiveness in this regard. To address this\nissue, we propose an approach to enhance Interpersonal Emotion Regulation (IER)\non online platforms through content recommendation. The objective is to empower\nusers to regulate their emotions while actively or passively engaging in online\nplatforms by crafting media content that aligns with IER strategies,\nparticularly empathic responding. The proposed recommendation system is\nexpected to blend system-initiated and user-initiated emotion regulation,\npaving the way for real-time IER practices on digital media platforms. To\nassess the efficacy of this approach, a mixed-method research design is used,\nincluding the analysis of text-based social media data and a user survey.\nDigital applications has served as facilitators in this process, given the\nwidespread recognition of digital media applications for Digital Emotion\nRegulation (DER). The study collects 37.5K instances of user posts and\ninteractions on Reddit over a year to design a Contextual Multi-Armed Bandits\n(CMAB) based recommendation system using features from user activity and\npreferences. The experimentation shows that the empathic recommendations\ngenerated by the proposed recommendation system are preferred by users over\nwidely accepted ER strategies such as distraction and avoidance.\n", "date": "2024-08-05", "categories": []}
{"id": "2408.07891", "title": "Quantum-inspired Interpretable Deep Learning Architecture for Text\n  Sentiment Analysis", "abstract": "  Text has become the predominant form of communication on social media,\nembedding a wealth of emotional nuances. Consequently, the extraction of\nemotional information from text is of paramount importance. Despite previous\nresearch making some progress, existing text sentiment analysis models still\nface challenges in integrating diverse semantic information and lack\ninterpretability. To address these issues, we propose a quantum-inspired deep\nlearning architecture that combines fundamental principles of quantum mechanics\n(QM principles) with deep learning models for text sentiment analysis.\nSpecifically, we analyze the commonalities between text representation and QM\nprinciples to design a quantum-inspired text representation method and further\ndevelop a quantum-inspired text embedding layer. Additionally, we design a\nfeature extraction layer based on long short-term memory (LSTM) networks and\nself-attention mechanisms (SAMs). Finally, we calculate the text density matrix\nusing the quantum complex numbers principle and apply 2D-convolution neural\nnetworks (CNNs) for feature condensation and dimensionality reduction. Through\na series of visualization, comparative, and ablation experiments, we\ndemonstrate that our model not only shows significant advantages in accuracy\nand efficiency compared to previous related models but also achieves a certain\nlevel of interpretability by integrating QM principles. Our code is available\nat QISA.\n", "date": "2024-08-15", "categories": []}
{"id": "2210.14556", "title": "Multimodal Contrastive Learning via Uni-Modal Coding and Cross-Modal\n  Prediction for Multimodal Sentiment Analysis", "abstract": "  Multimodal representation learning is a challenging task in which previous\nwork mostly focus on either uni-modality pre-training or cross-modality fusion.\nIn fact, we regard modeling multimodal representation as building a skyscraper,\nwhere laying stable foundation and designing the main structure are equally\nessential. The former is like encoding robust uni-modal representation while\nthe later is like integrating interactive information among different\nmodalities, both of which are critical to learning an effective multimodal\nrepresentation. Recently, contrastive learning has been successfully applied in\nrepresentation learning, which can be utilized as the pillar of the skyscraper\nand benefit the model to extract the most important features contained in the\nmultimodal data. In this paper, we propose a novel framework named MultiModal\nContrastive Learning (MMCL) for multimodal representation to capture intra- and\ninter-modality dynamics simultaneously. Specifically, we devise uni-modal\ncontrastive coding with an efficient uni-modal feature augmentation strategy to\nfilter inherent noise contained in acoustic and visual modality and acquire\nmore robust uni-modality representations. Besides, a pseudo siamese network is\npresented to predict representation across different modalities, which\nsuccessfully captures cross-modal dynamics. Moreover, we design two contrastive\nlearning tasks, instance- and sentiment-based contrastive learning, to promote\nthe process of prediction and learn more interactive information related to\nsentiment. Extensive experiments conducted on two public datasets demonstrate\nthat our method surpasses the state-of-the-art methods.\n", "date": "2022-10-26", "categories": []}
{"id": "2408.02009", "title": "Joint Learning of Emotions in Music and Generalized Sounds", "abstract": "  In this study, we aim to determine if generalized sounds and music can share\na common emotional space, improving predictions of emotion in terms of arousal\nand valence. We propose the use of multiple datasets as a multi-domain learning\ntechnique. Our approach involves creating a common space encompassing features\nthat characterize both generalized sounds and music, as they can evoke emotions\nin a similar manner. To achieve this, we utilized two publicly available\ndatasets, namely IADS-E and PMEmo, following a standardized experimental\nprotocol. We employed a wide variety of features that capture diverse aspects\nof the audio structure including key parameters of spectrum, energy, and\nvoicing. Subsequently, we performed joint learning on the common feature space,\nleveraging heterogeneous model architectures. Interestingly, this synergistic\nscheme outperforms the state-of-the-art in both sound and music emotion\nprediction. The code enabling full replication of the presented experimental\npipeline is available at https://github.com/LIMUNIMI/MusicSoundEmotions.\n", "date": "2024-08-04", "categories": []}
{"id": "2408.05523", "title": "DeepFace-Attention: Multimodal Face Biometrics for Attention Estimation\n  with Application to e-Learning", "abstract": "  This work introduces an innovative method for estimating attention levels\n(cognitive load) using an ensemble of facial analysis techniques applied to\nwebcam videos. Our method is particularly useful, among others, in e-learning\napplications, so we trained, evaluated, and compared our approach on the mEBAL2\ndatabase, a public multi-modal database acquired in an e-learning environment.\nmEBAL2 comprises data from 60 users who performed 8 different tasks. These\ntasks varied in difficulty, leading to changes in their cognitive loads. Our\napproach adapts state-of-the-art facial analysis technologies to quantify the\nusers' cognitive load in the form of high or low attention. Several behavioral\nsignals and physiological processes related to the cognitive load are used,\nsuch as eyeblink, heart rate, facial action units, and head pose, among others.\nFurthermore, we conduct a study to understand which individual features obtain\nbetter results, the most efficient combinations, explore local and global\nfeatures, and how temporary time intervals affect attention level estimation,\namong other aspects. We find that global facial features are more appropriate\nfor multimodal systems using score-level fusion, particularly as the temporal\nwindow increases. On the other hand, local features are more suitable for\nfusion through neural network training with score-level fusion approaches. Our\nmethod outperforms existing state-of-the-art accuracies using the public mEBAL2\nbenchmark.\n", "date": "2024-08-10", "categories": []}
{"id": "2408.07694", "title": "End-to-end Semantic-centric Video-based Multimodal Affective Computing", "abstract": "  In the pathway toward Artificial General Intelligence (AGI), understanding\nhuman's affection is essential to enhance machine's cognition abilities. For\nachieving more sensual human-AI interaction, Multimodal Affective Computing\n(MAC) in human-spoken videos has attracted increasing attention. However,\nprevious methods are mainly devoted to designing multimodal fusion algorithms,\nsuffering from two issues: semantic imbalance caused by diverse pre-processing\noperations and semantic mismatch raised by inconsistent affection content\ncontained in different modalities comparing with the multimodal ground truth.\nBesides, the usage of manual features extractors make they fail in building\nend-to-end pipeline for multiple MAC downstream tasks. To address above\nchallenges, we propose a novel end-to-end framework named SemanticMAC to\ncompute multimodal semantic-centric affection for human-spoken videos. We\nfirstly employ pre-trained Transformer model in multimodal data pre-processing\nand design Affective Perceiver module to capture unimodal affective\ninformation. Moreover, we present a semantic-centric approach to unify\nmultimodal representation learning in three ways, including gated feature\ninteraction, multi-task pseudo label generation, and intra-/inter-sample\ncontrastive learning. Finally, SemanticMAC effectively learn specific- and\nshared-semantic representations in the guidance of semantic-centric labels.\nExtensive experimental results demonstrate that our approach surpass the\nstate-of-the-art methods on 7 public datasets in four MAC downstream tasks.\n", "date": "2024-08-14", "categories": []}
{"id": "2306.09626", "title": "PAtt-Lite: Lightweight Patch and Attention MobileNet for Challenging\n  Facial Expression Recognition", "abstract": "  Facial Expression Recognition (FER) is a machine learning problem that deals\nwith recognizing human facial expressions. While existing work has achieved\nperformance improvements in recent years, FER in the wild and under challenging\nconditions remains a challenge. In this paper, a lightweight patch and\nattention network based on MobileNetV1, referred to as PAtt-Lite, is proposed\nto improve FER performance under challenging conditions. A truncated\nImageNet-pre-trained MobileNetV1 is utilized as the backbone feature extractor\nof the proposed method. In place of the truncated layers is a patch extraction\nblock that is proposed for extracting significant local facial features to\nenhance the representation from MobileNetV1, especially under challenging\nconditions. An attention classifier is also proposed to improve the learning of\nthese patched feature maps from the extremely lightweight feature extractor.\nThe experimental results on public benchmark databases proved the effectiveness\nof the proposed method. PAtt-Lite achieved state-of-the-art results on CK+,\nRAF-DB, FER2013, FERPlus, and the challenging conditions subsets for RAF-DB and\nFERPlus.\n", "date": "2023-06-16", "categories": []}
{"id": "2408.03958", "title": "Optimizing Emotion Recognition with Wearable Sensor Data: Unveiling\n  Patterns in Body Movements and Heart Rate through Random Forest\n  Hyperparameter Tuning", "abstract": "  This research delves into the utilization of smartwatch sensor data and heart\nrate monitoring to discern individual emotions based on body movement and heart\nrate. Emotions play a pivotal role in human life, influencing mental\nwell-being, quality of life, and even physical and physiological responses. The\ndata were sourced from prior research by Juan C. Quiroz, PhD. The study\nenlisted 50 participants who donned smartwatches and heart rate monitors while\ncompleting a 250-meter walk. Emotions were induced through both audio-visual\nand audio stimuli, with participants' emotional states evaluated using the\nPANAS questionnaire. The study scrutinized three scenarios: viewing a movie\nbefore walking, listening to music before walking, and listening to music while\nwalking. Personal baselines were established using DummyClassifier with the\n'most_frequent' strategy from the sklearn library, and various models,\nincluding Logistic Regression and Random Forest, were employed to gauge the\nimpacts of these activities. Notably, a novel approach was undertaken by\nincorporating hyperparameter tuning to the Random Forest model using\nRandomizedSearchCV. The outcomes showcased substantial enhancements with\nhyperparameter tuning in the Random Forest model, yielding mean accuracies of\n86.63% for happy vs. sad and 76.33% for happy vs. neutral vs. sad.\n", "date": "2024-07-28", "categories": []}
{"id": "2404.01288", "title": "Large Language Models are Capable of Offering Cognitive Reappraisal, if\n  Guided", "abstract": "  Large language models (LLMs) have offered new opportunities for emotional\nsupport, and recent work has shown that they can produce empathic responses to\npeople in distress. However, long-term mental well-being requires emotional\nself-regulation, where a one-time empathic response falls short. This work\ntakes a first step by engaging with cognitive reappraisals, a strategy from\npsychology practitioners that uses language to targetedly change negative\nappraisals that an individual makes of the situation; such appraisals is known\nto sit at the root of human emotional experience. We hypothesize that\npsychologically grounded principles could enable such advanced psychology\ncapabilities in LLMs, and design RESORT which consists of a series of\nreappraisal constitutions across multiple dimensions that can be used as LLM\ninstructions. We conduct a first-of-its-kind expert evaluation (by clinical\npsychologists with M.S. or Ph.D. degrees) of an LLM's zero-shot ability to\ngenerate cognitive reappraisal responses to medium-length social media messages\nasking for support. This fine-grained evaluation showed that even LLMs at the\n7B scale guided by RESORT are capable of generating empathic responses that can\nhelp users reappraise their situations.\n", "date": "2024-04-01", "categories": []}
{"id": "2408.04235", "title": "LLDif: Diffusion Models for Low-light Emotion Recognition", "abstract": "  This paper introduces LLDif, a novel diffusion-based facial expression\nrecognition (FER) framework tailored for extremely low-light (LL) environments.\nImages captured under such conditions often suffer from low brightness and\nsignificantly reduced contrast, presenting challenges to conventional methods.\nThese challenges include poor image quality that can significantly reduce the\naccuracy of emotion recognition. LLDif addresses these issues with a novel\ntwo-stage training process that combines a Label-aware CLIP (LA-CLIP), an\nembedding prior network (PNET), and a transformer-based network adept at\nhandling the noise of low-light images. The first stage involves LA-CLIP\ngenerating a joint embedding prior distribution (EPD) to guide the LLformer in\nlabel recovery. In the second stage, the diffusion model (DM) refines the EPD\ninference, ultilising the compactness of EPD for precise predictions.\nExperimental evaluations on various LL-FER datasets have shown that LLDif\nachieves competitive performance, underscoring its potential to enhance FER\napplications in challenging lighting conditions.\n", "date": "2024-08-08", "categories": []}
{"id": "2408.04420", "title": "Recognizing Emotion Regulation Strategies from Human Behavior with Large\n  Language Models", "abstract": "  Human emotions are often not expressed directly, but regulated according to\ninternal processes and social display rules. For affective computing systems,\nan understanding of how users regulate their emotions can be highly useful, for\nexample to provide feedback in job interview training, or in psychotherapeutic\nscenarios. However, at present no method to automatically classify different\nemotion regulation strategies in a cross-user scenario exists. At the same\ntime, recent studies showed that instruction-tuned Large Language Models (LLMs)\ncan reach impressive performance across a variety of affect recognition tasks\nsuch as categorical emotion recognition or sentiment analysis. While these\nresults are promising, it remains unclear to what extent the representational\npower of LLMs can be utilized in the more subtle task of classifying users'\ninternal emotion regulation strategy. To close this gap, we make use of the\nrecently introduced \\textsc{Deep} corpus for modeling the social display of the\nemotion shame, where each point in time is annotated with one of seven\ndifferent emotion regulation classes. We fine-tune Llama2-7B as well as the\nrecently introduced Gemma model using Low-rank Optimization on prompts\ngenerated from different sources of information on the \\textsc{Deep} corpus.\nThese include verbal and nonverbal behavior, person factors, as well as the\nresults of an in-depth interview after the interaction. Our results show, that\na fine-tuned Llama2-7B LLM is able to classify the utilized emotion regulation\nstrategy with high accuracy (0.84) without needing access to data from\npost-interaction interviews. This represents a significant improvement over\nprevious approaches based on Bayesian Networks and highlights the importance of\nmodeling verbal behavior in emotion regulation.\n", "date": "2024-08-08", "categories": []}
{"id": "2408.02164", "title": "Rethinking Affect Analysis: A Protocol for Ensuring Fairness and\n  Consistency", "abstract": "  Evaluating affect analysis methods presents challenges due to inconsistencies\nin database partitioning and evaluation protocols, leading to unfair and biased\nresults. Previous studies claim continuous performance improvements, but our\nfindings challenge such assertions. Using these insights, we propose a unified\nprotocol for database partitioning that ensures fairness and comparability. We\nprovide detailed demographic annotations (in terms of race, gender and age),\nevaluation metrics, and a common framework for expression recognition, action\nunit detection and valence-arousal estimation. We also rerun the methods with\nthe new protocol and introduce a new leaderboards to encourage future research\nin affect recognition with a fairer comparison. Our annotations, code, and\npre-trained models are available on\n\\hyperlink{https://github.com/dkollias/Fair-Consistent-Affect-Analysis}{Github}.\n", "date": "2024-08-04", "categories": []}
{"id": "2408.03499", "title": "FacialPulse: An Efficient RNN-based Depression Detection via Temporal\n  Facial Landmarks", "abstract": "  Depression is a prevalent mental health disorder that significantly impacts\nindividuals' lives and well-being. Early detection and intervention are crucial\nfor effective treatment and management of depression. Recently, there are many\nend-to-end deep learning methods leveraging the facial expression features for\nautomatic depression detection. However, most current methods overlook the\ntemporal dynamics of facial expressions. Although very recent 3DCNN methods\nremedy this gap, they introduce more computational cost due to the selection of\nCNN-based backbones and redundant facial features.\n  To address the above limitations, by considering the timing correlation of\nfacial expressions, we propose a novel framework called FacialPulse, which\nrecognizes depression with high accuracy and speed. By harnessing the\nbidirectional nature and proficiently addressing long-term dependencies, the\nFacial Motion Modeling Module (FMMM) is designed in FacialPulse to fully\ncapture temporal features. Since the proposed FMMM has parallel processing\ncapabilities and has the gate mechanism to mitigate gradient vanishing, this\nmodule can also significantly boost the training speed.\n  Besides, to effectively use facial landmarks to replace original images to\ndecrease information redundancy, a Facial Landmark Calibration Module (FLCM) is\ndesigned to eliminate facial landmark errors to further improve recognition\naccuracy. Extensive experiments on the AVEC2014 dataset and MMDA dataset (a\ndepression dataset) demonstrate the superiority of FacialPulse on recognition\naccuracy and speed, with the average MAE (Mean Absolute Error) decreased by 21%\ncompared to baselines, and the recognition speed increased by 100% compared to\nstate-of-the-art methods. Codes are released at\nhttps://github.com/volatileee/FacialPulse.\n", "date": "2024-08-07", "categories": []}
{"id": "2309.12881", "title": "Affect Recognition in Conversations Using Large Language Models", "abstract": "  Affect recognition, encompassing emotions, moods, and feelings, plays a\npivotal role in human communication. In the realm of conversational artificial\nintelligence, the ability to discern and respond to human affective cues is a\ncritical factor for creating engaging and empathetic interactions. This study\ninvestigates the capacity of large language models (LLMs) to recognise human\naffect in conversations, with a focus on both open-domain chit-chat dialogues\nand task-oriented dialogues. Leveraging three diverse datasets, namely IEMOCAP\n(Busso et al., 2008), EmoWOZ (Feng et al., 2022), and DAIC-WOZ (Gratch et al.,\n2014), covering a spectrum of dialogues from casual conversations to clinical\ninterviews, we evaluate and compare LLMs' performance in affect recognition.\nOur investigation explores the zero-shot and few-shot capabilities of LLMs\nthrough in-context learning as well as their model capacities through\ntask-specific fine-tuning. Additionally, this study takes into account the\npotential impact of automatic speech recognition errors on LLM predictions.\nWith this work, we aim to shed light on the extent to which LLMs can replicate\nhuman-like affect recognition capabilities in conversations.\n", "date": "2023-09-22", "categories": []}
{"id": "2403.13731", "title": "Emotion Recognition Using Transformers with Masked Learning", "abstract": "  In recent years, deep learning has achieved innovative advancements in\nvarious fields, including the analysis of human emotions and behaviors.\nInitiatives such as the Affective Behavior Analysis in-the-wild (ABAW)\ncompetition have been particularly instrumental in driving research in this\narea by providing diverse and challenging datasets that enable precise\nevaluation of complex emotional states. This study leverages the Vision\nTransformer (ViT) and Transformer models to focus on the estimation of\nValence-Arousal (VA), which signifies the positivity and intensity of emotions,\nrecognition of various facial expressions, and detection of Action Units (AU)\nrepresenting fundamental muscle movements. This approach transcends traditional\nConvolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) based\nmethods, proposing a new Transformer-based framework that maximizes the\nunderstanding of temporal and spatial features. The core contributions of this\nresearch include the introduction of a learning technique through random frame\nmasking and the application of Focal loss adapted for imbalanced data,\nenhancing the accuracy and applicability of emotion and behavior analysis in\nreal-world settings. This approach is expected to contribute to the advancement\nof emotional computing and deep learning methodologies.\n", "date": "2024-03-19", "categories": []}
{"id": "2308.11635", "title": "Semi-Supervised Dual-Stream Self-Attentive Adversarial Graph Contrastive\n  Learning for Cross-Subject EEG-based Emotion Recognition", "abstract": "  Electroencephalography (EEG) is an objective tool for emotion recognition\nwith promising applications. However, the scarcity of labeled data remains a\nmajor challenge in this field, limiting the widespread use of EEG-based emotion\nrecognition. In this paper, a semi-supervised Dual-stream Self-Attentive\nAdversarial Graph Contrastive learning framework (termed as DS-AGC) is proposed\nto tackle the challenge of limited labeled data in cross-subject EEG-based\nemotion recognition. The DS-AGC framework includes two parallel streams for\nextracting non-structural and structural EEG features. The non-structural\nstream incorporates a semi-supervised multi-domain adaptation method to\nalleviate distribution discrepancy among labeled source domain, unlabeled\nsource domain, and unknown target domain. The structural stream develops a\ngraph contrastive learning method to extract effective graph-based feature\nrepresentation from multiple EEG channels in a semi-supervised manner. Further,\na self-attentive fusion module is developed for feature fusion, sample\nselection, and emotion recognition, which highlights EEG features more relevant\nto emotions and data samples in the labeled source domain that are closer to\nthe target domain. Extensive experiments conducted on two benchmark databases\n(SEED and SEED-IV) using a semi-supervised cross-subject leave-one-subject-out\ncross-validation evaluation scheme show that the proposed model outperforms\nexisting methods under different incomplete label conditions (with an average\nimprovement of 5.83% on SEED and 6.99% on SEED-IV), demonstrating its\neffectiveness in addressing the label scarcity problem in cross-subject\nEEG-based emotion recognition.\n", "date": "2023-08-13", "categories": []}
{"id": "2408.00780", "title": "In-Depth Analysis of Emotion Recognition through Knowledge-Based Large\n  Language Models", "abstract": "  Emotion recognition in social situations is a complex task that requires\nintegrating information from both facial expressions and the situational\ncontext. While traditional approaches to automatic emotion recognition have\nfocused on decontextualized signals, recent research emphasizes the importance\nof context in shaping emotion perceptions. This paper contributes to the\nemerging field of context-based emotion recognition by leveraging psychological\ntheories of human emotion perception to inform the design of automated methods.\nWe propose an approach that combines emotion recognition methods with Bayesian\nCue Integration (BCI) to integrate emotion inferences from decontextualized\nfacial expressions and contextual knowledge inferred via Large-language Models.\nWe test this approach in the context of interpreting facial expressions during\na social task, the prisoner's dilemma. Our results provide clear support for\nBCI across a range of automatic emotion recognition methods. The best automated\nmethod achieved results comparable to human observers, suggesting the potential\nfor this approach to advance the field of affective computing.\n", "date": "2024-07-17", "categories": []}
{"id": "2404.17100", "title": "Open-Set Video-based Facial Expression Recognition with Human\n  Expression-sensitive Prompting", "abstract": "  In Video-based Facial Expression Recognition (V-FER), models are typically\ntrained on closed-set datasets with a fixed number of known classes. However,\nthese models struggle with unknown classes common in real-world scenarios. In\nthis paper, we introduce a challenging Open-set Video-based Facial Expression\nRecognition (OV-FER) task, aiming to identify both known and new, unseen facial\nexpressions. While existing approaches use large-scale vision-language models\nlike CLIP to identify unseen classes, we argue that these methods may not\nadequately capture the subtle human expressions needed for OV-FER. To address\nthis limitation, we propose a novel Human Expression-Sensitive Prompting (HESP)\nmechanism to significantly enhance CLIP's ability to model video-based facial\nexpression details effectively. Our proposed HESP comprises three components:\n1) a textual prompting module with learnable prompts to enhance CLIP's textual\nrepresentation of both known and unknown emotions, 2) a visual prompting module\nthat encodes temporal emotional information from video frames using\nexpression-sensitive attention, equipping CLIP with a new visual modeling\nability to extract emotion-rich information, and 3) an open-set multi-task\nlearning scheme that promotes interaction between the textual and visual\nmodules, improving the understanding of novel human emotions in video\nsequences. Extensive experiments conducted on four OV-FER task settings\ndemonstrate that HESP can significantly boost CLIP's performance (a relative\nimprovement of 17.93% on AUROC and 106.18% on OSCR) and outperform other\nstate-of-the-art open-set video understanding methods by a large margin. Code\nis available at https://github.com/cosinehuang/HESP.\n", "date": "2024-04-26", "categories": []}
{"id": "2408.00030", "title": "A New Type of Foundation Model Based on Recordings of People's Emotions\n  and Physiology", "abstract": "  Foundation models have had a big impact in recent years and billions of\ndollars are being invested in them in the current AI boom. The more popular\nones, such as Chat-GPT, are trained on large amounts of data from the Internet,\nand then reinforcement learning, RAG, prompt engineering and cognitive\nmodelling are used to fine-tune and augment their behavior. This technology has\nbeen used to create models of individual people, such as Caryn Marjorie.\nHowever, these chatbots are not based on people's actual emotional and\nphysiological responses to their environment, so they are, at best,\nsurface-level approximations to the characters they are imitating. This paper\ndescribes how a new type of foundation model - a first-person foundation model\n- could be created from recordings of what a person sees and hears as well as\ntheir emotional and physiological reactions to these stimuli. A first-person\nfoundation model would map environmental stimuli to a person's emotional and\nphysiological states, and map a person's emotional and physiological states to\ntheir behavior. First-person foundation models have many exciting applications,\nincluding a new type of recommendation engine, personal assistants, generative\nadversarial networks, dating and recruitment. To obtain training data for a\nfirst-person foundation model, we have developed a recording rig that captures\nwhat the wearer is seeing and hearing as well as their emotional and\nphysiological states. This novel source of data could help to address the\nshortage of new data for building the next generation of foundation models.\n", "date": "2024-07-31", "categories": []}
{"id": "2407.17772", "title": "ERIT Lightweight Multimodal Dataset for Elderly Emotion Recognition and\n  Multimodal Fusion Evaluation", "abstract": "  ERIT is a novel multimodal dataset designed to facilitate research in a\nlightweight multimodal fusion. It contains text and image data collected from\nvideos of elderly individuals reacting to various situations, as well as seven\nemotion labels for each data sample. Because of the use of labeled images of\nelderly users reacting emotionally, it is also facilitating research on emotion\nrecognition in an underrepresented age group in machine learning visual emotion\nrecognition. The dataset is validated through comprehensive experiments\nindicating its importance in neural multimodal fusion research.\n", "date": "2024-07-25", "categories": []}
{"id": "2407.21344", "title": "Dual-Constrained Dynamical Neural ODEs for Ambiguity-aware Continuous\n  Emotion Prediction", "abstract": "  There has been a significant focus on modelling emotion ambiguity in recent\nyears, with advancements made in representing emotions as distributions to\ncapture ambiguity. However, there has been comparatively less effort devoted to\nthe consideration of temporal dependencies in emotion distributions which\nencodes ambiguity in perceived emotions that evolve smoothly over time.\nRecognizing the benefits of using constrained dynamical neural ordinary\ndifferential equations (CD-NODE) to model time series as dynamic processes, we\npropose an ambiguity-aware dual-constrained Neural ODE approach to model the\ndynamics of emotion distributions on arousal and valence. In our approach, we\nutilize ODEs parameterised by neural networks to estimate the distribution\nparameters, and we integrate additional constraints to restrict the range of\nthe system outputs to ensure the validity of predicted distributions. We\nevaluated our proposed system on the publicly available RECOLA dataset and\nobserved very promising performance across a range of evaluation metrics.\n", "date": "2024-07-31", "categories": []}
{"id": "2407.21391", "title": "Design and Development of Laughter Recognition System Based on\n  Multimodal Fusion and Deep Learning", "abstract": "  This study aims to design and implement a laughter recognition system based\non multimodal fusion and deep learning, leveraging image and audio processing\ntechnologies to achieve accurate laughter recognition and emotion analysis.\nFirst, the system loads video files and uses the OpenCV library to extract\nfacial information while employing the Librosa library to process audio\nfeatures such as MFCC. Then, multimodal fusion techniques are used to integrate\nimage and audio features, followed by training and prediction using deep\nlearning models. Evaluation results indicate that the model achieved 80%\naccuracy, precision, and recall on the test dataset, with an F1 score of 80%,\ndemonstrating robust performance and the ability to handle real-world data\nvariability. This study not only verifies the effectiveness of multimodal\nfusion methods in laughter recognition but also highlights their potential\napplications in affective computing and human-computer interaction. Future work\nwill focus on further optimizing feature extraction and model architecture to\nimprove recognition accuracy and expand application scenarios, promoting the\ndevelopment of laughter recognition technology in fields such as mental health\nmonitoring and educational activity evaluation\n", "date": "2024-07-31", "categories": []}
{"id": "2407.12258", "title": "Facial Affect Recognition based on Multi Architecture Encoder and\n  Feature Fusion for the ABAW7 Challenge", "abstract": "  In this paper, we present our approach to addressing the challenges of the\n7th ABAW competition. The competition comprises three sub-challenges: Valence\nArousal (VA) estimation, Expression (Expr) classification, and Action Unit (AU)\ndetection. To tackle these challenges, we employ state-of-the-art models to\nextract powerful visual features. Subsequently, a Transformer Encoder is\nutilized to integrate these features for the VA, Expr, and AU sub-challenges.\nTo mitigate the impact of varying feature dimensions, we introduce an affine\nmodule to align the features to a common dimension. Overall, our results\nsignificantly outperform the baselines.\n", "date": "2024-07-17", "categories": []}
{"id": "2407.18496", "title": "Towards More Accurate Prediction of Human Empathy and Emotion in Text\n  and Multi-turn Conversations by Combining Advanced NLP, Transformers-based\n  Networks, and Linguistic Methodologies", "abstract": "  Based on the WASSA 2022 Shared Task on Empathy Detection and Emotion\nClassification, we predict the level of empathic concern and personal distress\ndisplayed in essays. For the first stage of this project we implemented a\nFeed-Forward Neural Network using sentence-level embeddings as features. We\nexperimented with four different embedding models for generating the inputs to\nthe neural network. The subsequent stage builds upon the previous work and we\nhave implemented three types of revisions. The first revision focuses on the\nenhancements to the model architecture and the training approach. The second\nrevision focuses on handling class imbalance using stratified data sampling.\nThe third revision focuses on leveraging lexical resources, where we apply four\ndifferent resources to enrich the features associated with the dataset. During\nthe final stage of this project, we have created the final end-to-end system\nfor the primary task using an ensemble of models to revise primary task\nperformance. Additionally, as part of the final stage, these approaches have\nbeen adapted to the WASSA 2023 Shared Task on Empathy Emotion and Personality\nDetection in Interactions, in which the empathic concern, emotion polarity, and\nemotion intensity in dyadic text conversations are predicted.\n", "date": "2024-07-26", "categories": []}
{"id": "2407.16552", "title": "MicroEmo: Time-Sensitive Multimodal Emotion Recognition with\n  Micro-Expression Dynamics in Video Dialogues", "abstract": "  Multimodal Large Language Models (MLLMs) have demonstrated remarkable\nmultimodal emotion recognition capabilities, integrating multimodal cues from\nvisual, acoustic, and linguistic contexts in the video to recognize human\nemotional states. However, existing methods ignore capturing local facial\nfeatures of temporal dynamics of micro-expressions and do not leverage the\ncontextual dependencies of the utterance-aware temporal segments in the video,\nthereby limiting their expected effectiveness to a certain extent. In this\nwork, we propose MicroEmo, a time-sensitive MLLM aimed at directing attention\nto the local facial micro-expression dynamics and the contextual dependencies\nof utterance-aware video clips. Our model incorporates two key architectural\ncontributions: (1) a global-local attention visual encoder that integrates\nglobal frame-level timestamp-bound image features with local facial features of\ntemporal dynamics of micro-expressions; (2) an utterance-aware video Q-Former\nthat captures multi-scale and contextual dependencies by generating visual\ntoken sequences for each utterance segment and for the entire video then\ncombining them. Preliminary qualitative experiments demonstrate that in a new\nExplainable Multimodal Emotion Recognition (EMER) task that exploits\nmulti-modal and multi-faceted clues to predict emotions in an open-vocabulary\n(OV) manner, MicroEmo demonstrates its effectiveness compared with the latest\nmethods.\n", "date": "2024-07-23", "categories": []}
{"id": "2407.16714", "title": "Masked Graph Learning with Recurrent Alignment for Multimodal Emotion\n  Recognition in Conversation", "abstract": "  Since Multimodal Emotion Recognition in Conversation (MERC) can be applied to\npublic opinion monitoring, intelligent dialogue robots, and other fields, it\nhas received extensive research attention in recent years. Unlike traditional\nunimodal emotion recognition, MERC can fuse complementary semantic information\nbetween multiple modalities (e.g., text, audio, and vision) to improve emotion\nrecognition. However, previous work ignored the inter-modal alignment process\nand the intra-modal noise information before multimodal fusion but directly\nfuses multimodal features, which will hinder the model for representation\nlearning. In this study, we have developed a novel approach called Masked Graph\nLearning with Recursive Alignment (MGLRA) to tackle this problem, which uses a\nrecurrent iterative module with memory to align multimodal features, and then\nuses the masked GCN for multimodal feature fusion. First, we employ LSTM to\ncapture contextual information and use a graph attention-filtering mechanism to\neliminate noise effectively within the modality. Second, we build a recurrent\niteration module with a memory function, which can use communication between\ndifferent modalities to eliminate the gap between modalities and achieve the\npreliminary alignment of features between modalities. Then, a cross-modal\nmulti-head attention mechanism is introduced to achieve feature alignment\nbetween modalities and construct a masked GCN for multimodal feature fusion,\nwhich can perform random mask reconstruction on the nodes in the graph to\nobtain better node feature representation. Finally, we utilize a multilayer\nperceptron (MLP) for emotion recognition. Extensive experiments on two\nbenchmark datasets (i.e., IEMOCAP and MELD) demonstrate that {MGLRA}\noutperforms state-of-the-art methods.\n", "date": "2024-07-23", "categories": []}
{"id": "2407.14576", "title": "A Comparative Study of Transfer Learning for Emotion Recognition using\n  CNN and Modified VGG16 Models", "abstract": "  Emotion recognition is a critical aspect of human interaction. This topic\ngarnered significant attention in the field of artificial intelligence. In this\nstudy, we investigate the performance of convolutional neural network (CNN) and\nModified VGG16 models for emotion recognition tasks across two datasets:\nFER2013 and AffectNet. Our aim is to measure the effectiveness of these models\nin identifying emotions and their ability to generalize to different and\nbroader datasets. Our findings reveal that both models achieve reasonable\nperformance on the FER2013 dataset, with the Modified VGG16 model demonstrating\nslightly increased accuracy. When evaluated on the Affect-Net dataset,\nperformance declines for both models, with the Modified VGG16 model continuing\nto outperform the CNN. Our study emphasizes the importance of dataset diversity\nin emotion recognition and discusses open problems and future research\ndirections, including the exploration of multi-modal approaches and the\ndevelopment of more comprehensive datasets.\n", "date": "2024-07-19", "categories": []}
{"id": "2407.15590", "title": "All rivers run into the sea: Unified Modality Brain-like Emotional\n  Central Mechanism", "abstract": "  In the field of affective computing, fully leveraging information from a\nvariety of sensory modalities is essential for the comprehensive understanding\nand processing of human emotions. Inspired by the process through which the\nhuman brain handles emotions and the theory of cross-modal plasticity, we\npropose UMBEnet, a brain-like unified modal affective processing network. The\nprimary design of UMBEnet includes a Dual-Stream (DS) structure that fuses\ninherent prompts with a Prompt Pool and a Sparse Feature Fusion (SFF) module.\nThe design of the Prompt Pool is aimed at integrating information from\ndifferent modalities, while inherent prompts are intended to enhance the\nsystem's predictive guidance capabilities and effectively manage knowledge\nrelated to emotion classification. Moreover, considering the sparsity of\neffective information across different modalities, the SSF module aims to make\nfull use of all available sensory data through the sparse integration of\nmodality fusion prompts and inherent prompts, maintaining high adaptability and\nsensitivity to complex emotional states. Extensive experiments on the largest\nbenchmark datasets in the Dynamic Facial Expression Recognition (DFER) field,\nincluding DFEW, FERV39k, and MAFW, have proven that UMBEnet consistently\noutperforms the current state-of-the-art methods. Notably, in scenarios of\nModality Missingness and multimodal contexts, UMBEnet significantly surpasses\nthe leading current methods, demonstrating outstanding performance and\nadaptability in tasks that involve complex emotional understanding with rich\nmultimodal information.\n", "date": "2024-07-22", "categories": []}
{"id": "2407.15617", "title": "Norface: Improving Facial Expression Analysis by Identity Normalization", "abstract": "  Facial Expression Analysis remains a challenging task due to unexpected\ntask-irrelevant noise, such as identity, head pose, and background. To address\nthis issue, this paper proposes a novel framework, called Norface, that is\nunified for both Action Unit (AU) analysis and Facial Emotion Recognition (FER)\ntasks. Norface consists of a normalization network and a classification\nnetwork. First, the carefully designed normalization network struggles to\ndirectly remove the above task-irrelevant noise, by maintaining facial\nexpression consistency but normalizing all original images to a common identity\nwith consistent pose, and background. Then, these additional normalized images\nare fed into the classification network. Due to consistent identity and other\nfactors (e.g. head pose, background, etc.), the normalized images enable the\nclassification network to extract useful expression information more\neffectively. Additionally, the classification network incorporates a Mixture of\nExperts to refine the latent representation, including handling the input of\nfacial representations and the output of multiple (AU or emotion) labels.\nExtensive experiments validate the carefully designed framework with the\ninsight of identity normalization. The proposed method outperforms existing\nSOTA methods in multiple facial expression analysis tasks, including AU\ndetection, AU intensity estimation, and FER tasks, as well as their\ncross-dataset tasks. For the normalized datasets and code please visit\n{https://norface-fea.github.io/}.\n", "date": "2024-07-22", "categories": []}
{"id": "2407.14458", "title": "AudioInsight: Detecting Social Contexts Relevant to Social Anxiety from\n  Speech", "abstract": "  During social interactions, understanding the intricacies of the context can\nbe vital, particularly for socially anxious individuals. While previous\nresearch has found that the presence of a social interaction can be detected\nfrom ambient audio, the nuances within social contexts, which influence how\nanxiety provoking interactions are, remain largely unexplored. As an\nalternative to traditional, burdensome methods like self-report, this study\npresents a novel approach that harnesses ambient audio segments to detect\nsocial threat contexts. We focus on two key dimensions: number of interaction\npartners (dyadic vs. group) and degree of evaluative threat (explicitly\nevaluative vs. not explicitly evaluative). Building on data from a Zoom-based\nsocial interaction study (N=52 college students, of whom the majority N=45 are\nsocially anxious), we employ deep learning methods to achieve strong detection\nperformance. Under sample-wide 5-fold Cross Validation (CV), our model\ndistinguished dyadic from group interactions with 90\\% accuracy and detected\nevaluative threat at 83\\%. Using a leave-one-group-out CV, accuracies were 82\\%\nand 77\\%, respectively. While our data are based on virtual interactions due to\npandemic constraints, our method has the potential to extend to diverse\nreal-world settings. This research underscores the potential of passive sensing\nand AI to differentiate intricate social contexts, and may ultimately advance\nthe ability of context-aware digital interventions to offer personalized mental\nhealth support.\n", "date": "2024-07-19", "categories": []}
{"id": "2404.17113", "title": "MER 2024: Semi-Supervised Learning, Noise Robustness, and\n  Open-Vocabulary Multimodal Emotion Recognition", "abstract": "  Multimodal emotion recognition is an important research topic in artificial\nintelligence. Over the past few decades, researchers have made remarkable\nprogress by increasing the dataset size and building more effective algorithms.\nHowever, due to problems such as complex environments and inaccurate\nannotations, current systems are hard to meet the demands of practical\napplications. Therefore, we organize the MER series of competitions to promote\nthe development of this field. Last year, we launched MER2023, focusing on\nthree interesting topics: multi-label learning, noise robustness, and\nsemi-supervised learning. In this year's MER2024, besides expanding the dataset\nsize, we further introduce a new track around open-vocabulary emotion\nrecognition. The main purpose of this track is that existing datasets usually\nfix the label space and use majority voting to enhance the annotator\nconsistency. However, this process may lead to inaccurate annotations, such as\nignoring non-majority or non-candidate labels. In this track, we encourage\nparticipants to generate any number of labels in any category, aiming to\ndescribe emotional states as accurately as possible. Our baseline code relies\non MERTools and is available at:\nhttps://github.com/zeroQiaoba/MERTools/tree/master/MER2024.\n", "date": "2024-04-26", "categories": []}
{"id": "2407.12825", "title": "A Depression Detection Method Based on Multi-Modal Feature Fusion Using\n  Cross-Attention", "abstract": "  Depression, a prevalent and serious mental health issue, affects\napproximately 3.8\\% of the global population. Despite the existence of\neffective treatments, over 75\\% of individuals in low- and middle-income\ncountries remain untreated, partly due to the challenge in accurately\ndiagnosing depression in its early stages. This paper introduces a novel method\nfor detecting depression based on multi-modal feature fusion utilizing\ncross-attention. By employing MacBERT as a pre-training model to extract\nlexical features from text and incorporating an additional Transformer module\nto refine task-specific contextual understanding, the model's adaptability to\nthe targeted task is enhanced. Diverging from previous practices of simply\nconcatenating multimodal features, this approach leverages cross-attention for\nfeature integration, significantly improving the accuracy in depression\ndetection and enabling a more comprehensive and precise analysis of user\nemotions and behaviors. Furthermore, a Multi-Modal Feature Fusion Network based\non Cross-Attention (MFFNC) is constructed, demonstrating exceptional\nperformance in the task of depression identification. The experimental results\nindicate that our method achieves an accuracy of 0.9495 on the test dataset,\nmarking a substantial improvement over existing approaches. Moreover, it\noutlines a promising methodology for other social media platforms and tasks\ninvolving multi-modal processing. Timely identification and intervention for\nindividuals with depression are crucial for saving lives, highlighting the\nimmense potential of technology in facilitating early intervention for mental\nhealth issues.\n", "date": "2024-07-02", "categories": []}
{"id": "2407.12973", "title": "Temporal Label Hierachical Network for Compound Emotion Recognition", "abstract": "  The emotion recognition has attracted more attention in recent decades.\nAlthough significant progress has been made in the recognition technology of\nthe seven basic emotions, existing methods are still hard to tackle compound\nemotion recognition that occurred commonly in practical application. This\narticle introduces our achievements in the 7th Field Emotion Behavior Analysis\n(ABAW) competition. In the competition, we selected pre trained ResNet18 and\nTransformer, which have been widely validated, as the basic network framework.\nConsidering the continuity of emotions over time, we propose a time pyramid\nstructure network for frame level emotion prediction. Furthermore. At the same\ntime, in order to address the lack of data in composite emotion recognition, we\nutilized fine-grained labels from the DFEW database to construct training data\nfor emotion categories in competitions. Taking into account the characteristics\nof valence arousal of various complex emotions, we constructed a classification\nframework from coarse to fine in the label space.\n", "date": "2024-07-17", "categories": []}
{"id": "2407.13184", "title": "HSEmotion Team at the 7th ABAW Challenge: Multi-Task Learning and\n  Compound Facial Expression Recognition", "abstract": "  In this paper, we describe the results of the HSEmotion team in two tasks of\nthe seventh Affective Behavior Analysis in-the-wild (ABAW) competition, namely,\nmulti-task learning for simultaneous prediction of facial expression, valence,\narousal, and detection of action units, and compound expression recognition. We\npropose an efficient pipeline based on frame-level facial feature extractors\npre-trained in multi-task settings to estimate valence-arousal and basic facial\nexpressions given a facial photo. We ensure the privacy-awareness of our\ntechniques by using the lightweight architectures of neural networks, such as\nMT-EmotiDDAMFN, MT-EmotiEffNet, and MT-EmotiMobileFaceNet, that can run even on\na mobile device without the need to send facial video to a remote server. It\nwas demonstrated that a significant step in improving the overall accuracy is\nthe smoothing of neural network output scores using Gaussian or box filters. It\nwas experimentally demonstrated that such a simple post-processing of\npredictions from simple blending of two top visual models improves the F1-score\nof facial expression recognition up to 7%. At the same time, the mean\nConcordance Correlation Coefficient (CCC) of valence and arousal is increased\nby up to 1.25 times compared to each model's frame-level predictions. As a\nresult, our final performance score on the validation set from the multi-task\nlearning challenge is 4.5 times higher than the baseline (1.494 vs 0.32).\n", "date": "2024-07-18", "categories": []}
{"id": "2407.12141", "title": "Predicting Emotion Intensity in Polish Political Texts: Comparing\n  Supervised Models and Large Language Models in a Resource-Poor Language", "abstract": "  This study explores the use of large language models (LLMs) to predict\nemotion intensity in Polish political texts, a resource-poor language context.\nThe research compares the performance of several LLMs against a supervised\nmodel trained on an annotated corpus of 10,000 social media texts, evaluated\nfor the intensity of emotions by expert judges. The findings indicate that\nwhile the supervised model generally outperforms LLMs, offering higher accuracy\nand lower variance, LLMs present a viable alternative, especially given the\nhigh costs associated with data annotation. The study highlights the potential\nof LLMs in low-resource language settings and underscores the need for further\nresearch on emotion intensity prediction and its application across different\nlanguages and continuous features. The implications suggest a nuanced\ndecision-making process to choose the right approach to emotion prediction for\nresearchers and practitioners based on resource availability and the specific\nrequirements of their tasks.\n", "date": "2024-07-16", "categories": []}
{"id": "2407.11663", "title": "Affective Behavior Analysis using Task-adaptive and AU-assisted Graph\n  Network", "abstract": "  In this paper, we present our solution and experiment result for the\nMulti-Task Learning Challenge of the 7th Affective Behavior Analysis\nin-the-wild(ABAW7) Competition. This challenge consists of three tasks: action\nunit detection, facial expression recognition, and valance-arousal estimation.\nWe address the research problems of this challenge from three aspects: 1)For\nlearning robust visual feature representations, we introduce the pre-trained\nlarge model Dinov2. 2) To adaptively extract the required features of eack\ntask, we design a task-adaptive block that performs cross-attention between a\nset of learnable query vectors and pre-extracted features. 3) By proposing the\nAU-assisted Graph Convolutional Network(AU-GCN), we make full use of the\ncorrelation information between AUs to assist in solving the EXPR and VA tasks.\nFinally, we achieve the evaluation measure of \\textbf{1.2542} on the validation\nset provided by the organizers.\n", "date": "2024-07-16", "categories": []}
{"id": "2407.09911", "title": "SensEmo: Enabling Affective Learning through Real-time Emotion\n  Recognition with Smartwatches", "abstract": "  Recent research has demonstrated the capability of physiological signals to\ninfer both user emotional and attention responses. This presents an opportunity\nfor leveraging widely available physiological sensors in smartwatches, to\ndetect real-time emotional cues in users, such as stress and excitement. In\nthis paper, we introduce SensEmo, a smartwatch-based system designed for\naffective learning. SensEmo utilizes multiple physiological sensor data,\nincluding heart rate and galvanic skin response, to recognize a student's\nmotivation and concentration levels during class. This recognition is\nfacilitated by a personalized emotion recognition model that predicts emotional\nstates based on degrees of valence and arousal. With real-time emotion and\nattention feedback from students, we design a Markov decision process-based\nalgorithm to enhance student learning effectiveness and experience by by\noffering suggestions to the teacher regarding teaching content and pacing. We\nevaluate SensEmo with 22 participants in real-world classroom environments.\nEvaluation results show that SensEmo recognizes student emotion with an average\nof 88.9% accuracy. More importantly, SensEmo assists students to achieve better\nonline learning outcomes, e.g., an average of 40.0% higher grades in quizzes,\nover the traditional learning without student emotional feedback.\n", "date": "2024-07-13", "categories": []}
{"id": "2407.09913", "title": "Emotion Detection through Body Gesture and Face", "abstract": "  The project leverages advanced machine and deep learning techniques to\naddress the challenge of emotion recognition by focusing on non-facial cues,\nspecifically hands, body gestures, and gestures. Traditional emotion\nrecognition systems mainly rely on facial expression analysis and often ignore\nthe rich emotional information conveyed through body language. To bridge this\ngap, this method leverages the Aff-Wild2 and DFEW databases to train and\nevaluate a model capable of recognizing seven basic emotions (angry, disgust,\nfear, happiness, sadness, surprise, and neutral) and estimating valence and\ncontinuous scales wakeup descriptor.\n  Leverage OpenPose for pose estimation to extract detailed body posture and\nposture features from images and videos. These features serve as input to\nstate-of-the-art neural network architectures, including ResNet, and ANN for\nemotion classification, and fully connected layers for valence arousal\nregression analysis. This bifurcation strategy can solve classification and\nregression problems in the field of emotion recognition.\n  The project aims to contribute to the field of affective computing by\nenhancing the ability of machines to interpret and respond to human emotions in\na more comprehensive and nuanced way. By integrating multimodal data and\ncutting-edge computational models, I aspire to develop a system that not only\nenriches human-computer interaction but also has potential applications in\nareas as diverse as mental health support, educational technology, and\nautonomous vehicle systems.\n", "date": "2024-07-13", "categories": []}
{"id": "2404.05052", "title": "Facial Affective Behavior Analysis with Instruction Tuning", "abstract": "  Facial affective behavior analysis (FABA) is crucial for understanding human\nmental states from images. However, traditional approaches primarily deploy\nmodels to discriminate among discrete emotion categories, and lack the fine\ngranularity and reasoning capability for complex facial behaviors. The advent\nof Multi-modal Large Language Models (MLLMs) has been proven successful in\ngeneral visual understanding tasks. However, directly harnessing MLLMs for FABA\nis challenging due to the scarcity of datasets and benchmarks, neglecting\nfacial prior knowledge, and low training efficiency. To address these\nchallenges, we introduce (i) an instruction-following dataset for two FABA\ntasks, e.g., emotion and action unit recognition, (ii) a benchmark FABA-Bench\nwith a new metric considering both recognition and generation ability, and\n(iii) a new MLLM \"EmoLA\" as a strong baseline to the community. Our initiative\non the dataset and benchmarks reveal the nature and rationale of facial\naffective behaviors, i.e., fine-grained facial movement, interpretability, and\nreasoning. Moreover, to build an effective and efficient FABA MLLM, we\nintroduce a facial prior expert module with face structure knowledge and a\nlow-rank adaptation module into pre-trained MLLM. We conduct extensive\nexperiments on FABA-Bench and four commonly-used FABA datasets. The results\ndemonstrate that the proposed facial prior expert can boost the performance and\nEmoLA achieves the best results on our FABA-Bench. On commonly-used FABA\ndatasets, EmoLA is competitive rivaling task-specific state-of-the-art models.\n", "date": "2024-04-07", "categories": []}
{"id": "2407.08992", "title": "Emotion Talk: Emotional Support via Audio Messages for Psychological\n  Assistance", "abstract": "  This paper presents \"Emotion Talk,\" a system designed to provide continuous\nemotional support through audio messages for psychological assistance. The\nprimary objective is to offer consistent support to patients outside\ntraditional therapy sessions by analyzing audio messages to detect emotions and\ngenerate appropriate responses. The solution focuses on Portuguese-speaking\nusers, ensuring that the system is linguistically and culturally relevant. This\nsystem aims to complement and enhance the psychological follow-up process\nconducted by therapists, providing immediate and accessible assistance,\nespecially in emergency situations where rapid response is crucial.\nExperimental results demonstrate the effectiveness of the proposed system,\nhighlighting its potential in applications of psychological support.\n", "date": "2024-07-12", "categories": []}
{"id": "2407.09029", "title": "Enhancing Emotion Recognition in Incomplete Data: A Novel Cross-Modal\n  Alignment, Reconstruction, and Refinement Framework", "abstract": "  Multimodal emotion recognition systems rely heavily on the full availability\nof modalities, suffering significant performance declines when modal data is\nincomplete. To tackle this issue, we present the Cross-Modal Alignment,\nReconstruction, and Refinement (CM-ARR) framework, an innovative approach that\nsequentially engages in cross-modal alignment, reconstruction, and refinement\nphases to handle missing modalities and enhance emotion recognition. This\nframework utilizes unsupervised distribution-based contrastive learning to\nalign heterogeneous modal distributions, reducing discrepancies and modeling\nsemantic uncertainty effectively. The reconstruction phase applies normalizing\nflow models to transform these aligned distributions and recover missing\nmodalities. The refinement phase employs supervised point-based contrastive\nlearning to disrupt semantic correlations and accentuate emotional traits,\nthereby enriching the affective content of the reconstructed representations.\nExtensive experiments on the IEMOCAP and MSP-IMPROV datasets confirm the\nsuperior performance of CM-ARR under conditions of both missing and complete\nmodalities. Notably, averaged across six scenarios of missing modalities,\nCM-ARR achieves absolute improvements of 2.11% in WAR and 2.12% in UAR on the\nIEMOCAP dataset, and 1.71% and 1.96% in WAR and UAR, respectively, on the\nMSP-IMPROV dataset.\n", "date": "2024-07-12", "categories": []}
{"id": "2407.08240", "title": "Leveraging LLMs to Predict Affective States via Smartphone Sensor\n  Features", "abstract": "  As mental health issues for young adults present a pressing public health\nconcern, daily digital mood monitoring for early detection has become an\nimportant prospect. An active research area, digital phenotyping, involves\ncollecting and analysing data from personal digital devices such as smartphones\n(usage and sensors) and wearables to infer behaviours and mental health. Whilst\nthis data is standardly analysed using statistical and machine learning\napproaches, the emergence of large language models (LLMs) offers a new approach\nto make sense of smartphone sensing data. Despite their effectiveness across\nvarious domains, LLMs remain relatively unexplored in digital mental health,\nparticularly in integrating mobile sensor data. Our study aims to bridge this\ngap by employing LLMs to predict affect outcomes based on smartphone sensing\ndata from university students. We demonstrate the efficacy of zero-shot and\nfew-shot embedding LLMs in inferring general wellbeing. Our findings reveal\nthat LLMs can make promising predictions of affect measures using solely\nsmartphone sensing data. This research sheds light on the potential of LLMs for\naffective state prediction, emphasizing the intricate link between smartphone\nbehavioral patterns and affective states. To our knowledge, this is the first\nwork to leverage LLMs for affective state prediction and digital phenotyping\ntasks.\n", "date": "2024-07-11", "categories": []}
{"id": "2407.08607", "title": "Turn-Level Empathy Prediction Using Psychological Indicators", "abstract": "  For the WASSA 2024 Empathy and Personality Prediction Shared Task, we propose\na novel turn-level empathy detection method that decomposes empathy into six\npsychological indicators: Emotional Language, Perspective-Taking, Sympathy and\nCompassion, Extroversion, Openness, and Agreeableness. A pipeline of text\nenrichment using a Large Language Model (LLM) followed by DeBERTA fine-tuning\ndemonstrates a significant improvement in the Pearson Correlation Coefficient\nand F1 scores for empathy detection, highlighting the effectiveness of our\napproach. Our system officially ranked 7th at the CONV-turn track.\n", "date": "2024-07-11", "categories": []}
{"id": "2407.07345", "title": "Micro-Expression Recognition by Motion Feature Extraction based on\n  Pre-training", "abstract": "  Micro-expressions (MEs) are spontaneous, unconscious facial expressions that\nhave promising applications in various fields such as psychotherapy and\nnational security. Thus, micro-expression recognition (MER) has attracted more\nand more attention from researchers. Although various MER methods have emerged\nespecially with the development of deep learning techniques, the task still\nfaces several challenges, e.g. subtle motion and limited training data. To\naddress these problems, we propose a novel motion extraction strategy (MoExt)\nfor the MER task and use additional macro-expression data in the pre-training\nprocess. We primarily pretrain the feature separator and motion extractor using\nthe contrastive loss, thus enabling them to extract representative motion\nfeatures. In MoExt, shape features and texture features are first extracted\nseparately from onset and apex frames, and then motion features related to MEs\nare extracted based on the shape features of both frames. To enable the model\nto more effectively separate features, we utilize the extracted motion features\nand the texture features from the onset frame to reconstruct the apex frame.\nThrough pre-training, the module is enabled to extract inter-frame motion\nfeatures of facial expressions while excluding irrelevant information. The\nfeature separator and motion extractor are ultimately integrated into the MER\nnetwork, which is then fine-tuned using the target ME data. The effectiveness\nof proposed method is validated on three commonly used datasets, i.e., CASME\nII, SMIC, SAMM, and CAS(ME)3 dataset. The results show that our method performs\nfavorably against state-of-the-art methods.\n", "date": "2024-07-10", "categories": []}
{"id": "2405.13049", "title": "SemEval-2024 Task 3: Multimodal Emotion Cause Analysis in Conversations", "abstract": "  The ability to understand emotions is an essential component of human-like\nartificial intelligence, as emotions greatly influence human cognition,\ndecision making, and social interactions. In addition to emotion recognition in\nconversations, the task of identifying the potential causes behind an\nindividual's emotional state in conversations, is of great importance in many\napplication scenarios. We organize SemEval-2024 Task 3, named Multimodal\nEmotion Cause Analysis in Conversations, which aims at extracting all pairs of\nemotions and their corresponding causes from conversations. Under different\nmodality settings, it consists of two subtasks: Textual Emotion-Cause Pair\nExtraction in Conversations (TECPE) and Multimodal Emotion-Cause Pair\nExtraction in Conversations (MECPE). The shared task has attracted 143\nregistrations and 216 successful submissions. In this paper, we introduce the\ntask, dataset and evaluation settings, summarize the systems of the top teams,\nand discuss the findings of the participants.\n", "date": "2024-05-19", "categories": []}
{"id": "2407.03835", "title": "7th ABAW Competition: Multi-Task Learning and Compound Expression\n  Recognition", "abstract": "  This paper describes the 7th Affective Behavior Analysis in-the-wild (ABAW)\nCompetition, which is part of the respective Workshop held in conjunction with\nECCV 2024. The 7th ABAW Competition addresses novel challenges in understanding\nhuman expressions and behaviors, crucial for the development of human-centered\ntechnologies. The Competition comprises of two sub-challenges: i) Multi-Task\nLearning (the goal is to learn at the same time, in a multi-task learning\nsetting, to estimate two continuous affect dimensions, valence and arousal, to\nrecognise between the mutually exclusive classes of the 7 basic expressions and\n'other'), and to detect 12 Action Units); and ii) Compound Expression\nRecognition (the target is to recognise between the 7 mutually exclusive\ncompound expression classes). s-Aff-Wild2, which is a static version of the A/V\nAff-Wild2 database and contains annotations for valence-arousal, expressions\nand Action Units, is utilized for the purposes of the Multi-Task Learning\nChallenge; a part of C-EXPR-DB, which is an A/V in-the-wild database with\ncompound expression annotations, is utilized for the purposes of the Compound\nExpression Recognition Challenge. In this paper, we introduce the two\nchallenges, detailing their datasets and the protocols followed for each. We\nalso outline the evaluation metrics, and highlight the baseline systems and\ntheir results. Additional information about the competition can be found at\n\\url{https://affective-behavior-analysis-in-the-wild.github.io/7th}.\n", "date": "2024-07-04", "categories": []}
{"id": "2407.04712", "title": "Sensing technologies and machine learning methods for emotion\n  recognition in autism: Systematic review", "abstract": "  Background: Human Emotion Recognition (HER) has been a popular field of study\nin the past years. Despite the great progresses made so far, relatively little\nattention has been paid to the use of HER in autism. People with autism are\nknown to face problems with daily social communication and the prototypical\ninterpretation of emotional responses, which are most frequently exerted via\nfacial expressions. This poses significant practical challenges to the\napplication of regular HER systems, which are normally developed for and by\nneurotypical people. Objective: This study reviews the literature on the use of\nHER systems in autism, particularly with respect to sensing technologies and\nmachine learning methods, as to identify existing barriers and possible future\ndirections. Methods: We conducted a systematic review of articles published\nbetween January 2011 and June 2023 according to the 2020 PRISMA guidelines.\nManuscripts were identified through searching Web of Science and Scopus\ndatabases. Manuscripts were included when related to emotion recognition, used\nsensors and machine learning techniques, and involved children with autism,\nyoung, or adults. Results: The search yielded 346 articles. A total of 65\npublications met the eligibility criteria and were included in the review.\nConclusions: Studies predominantly used facial expression techniques as the\nemotion recognition method. Consequently, video cameras were the most widely\nused devices across studies, although a growing trend in the use of\nphysiological sensors was observed lately. Happiness, sadness, anger, fear,\ndisgust, and surprise were most frequently addressed. Classical supervised\nmachine learning techniques were primarily used at the expense of unsupervised\napproaches or more recent deep learning models.\n", "date": "2024-05-15", "categories": []}
{"id": "2407.04963", "title": "Towards Context-Aware Emotion Recognition Debiasing from a Causal\n  Demystification Perspective via De-confounded Training", "abstract": "  Understanding emotions from diverse contexts has received widespread\nattention in computer vision communities. The core philosophy of Context-Aware\nEmotion Recognition (CAER) is to provide valuable semantic cues for recognizing\nthe emotions of target persons by leveraging rich contextual information.\nCurrent approaches invariably focus on designing sophisticated structures to\nextract perceptually critical representations from contexts. Nevertheless, a\nlong-neglected dilemma is that a severe context bias in existing datasets\nresults in an unbalanced distribution of emotional states among different\ncontexts, causing biased visual representation learning. From a causal\ndemystification perspective, the harmful bias is identified as a confounder\nthat misleads existing models to learn spurious correlations based on\nlikelihood estimation, limiting the models' performance. To address the issue,\nwe embrace causal inference to disentangle the models from the impact of such\nbias, and formulate the causalities among variables in the CAER task via a\ncustomized causal graph. Subsequently, we present a Contextual Causal\nIntervention Module (CCIM) to de-confound the confounder, which is built upon\nbackdoor adjustment theory to facilitate seeking approximate causal effects\nduring model training. As a plug-and-play component, CCIM can easily integrate\nwith existing approaches and bring significant improvements. Systematic\nexperiments on three datasets demonstrate the effectiveness of our CCIM.\n", "date": "2024-07-06", "categories": []}
{"id": "2407.04966", "title": "A Layer-Anchoring Strategy for Enhancing Cross-Lingual Speech Emotion\n  Recognition", "abstract": "  Cross-lingual speech emotion recognition (SER) is important for a wide range\nof everyday applications. While recent SER research relies heavily on large\npretrained models for emotion training, existing studies often concentrate\nsolely on the final transformer layer of these models. However, given the\ntask-specific nature and hierarchical architecture of these models, each\ntransformer layer encapsulates different levels of information. Leveraging this\nhierarchical structure, our study focuses on the information embedded across\ndifferent layers. Through an examination of layer feature similarity across\ndifferent languages, we propose a novel strategy called a layer-anchoring\nmechanism to facilitate emotion transfer in cross-lingual SER tasks. Our\napproach is evaluated using two distinct language affective corpora\n(MSP-Podcast and BIIC-Podcast), achieving a best UAR performance of 60.21% on\nthe BIIC-podcast corpus. The analysis uncovers interesting insights into the\nbehavior of popular pretrained models.\n", "date": "2024-07-06", "categories": []}
{"id": "2407.05374", "title": "Multimodal Prompt Learning with Missing Modalities for Sentiment\n  Analysis and Emotion Recognition", "abstract": "  The development of multimodal models has significantly advanced multimodal\nsentiment analysis and emotion recognition. However, in real-world\napplications, the presence of various missing modality cases often leads to a\ndegradation in the model's performance. In this work, we propose a novel\nmultimodal Transformer framework using prompt learning to address the issue of\nmissing modalities. Our method introduces three types of prompts: generative\nprompts, missing-signal prompts, and missing-type prompts. These prompts enable\nthe generation of missing modality features and facilitate the learning of\nintra- and inter-modality information. Through prompt learning, we achieve a\nsubstantial reduction in the number of trainable parameters. Our proposed\nmethod outperforms other methods significantly across all evaluation metrics.\nExtensive experiments and ablation studies are conducted to demonstrate the\neffectiveness and robustness of our method, showcasing its ability to\neffectively handle missing modalities.\n", "date": "2024-07-07", "categories": []}
{"id": "2407.05746", "title": "MSP-Podcast SER Challenge 2024: L'antenne du Ventoux Multimodal\n  Self-Supervised Learning for Speech Emotion Recognition", "abstract": "  In this work, we detail our submission to the 2024 edition of the MSP-Podcast\nSpeech Emotion Recognition (SER) Challenge. This challenge is divided into two\ndistinct tasks: Categorical Emotion Recognition and Emotional Attribute\nPrediction. We concentrated our efforts on Task 1, which involves the\ncategorical classification of eight emotional states using data from the\nMSP-Podcast dataset. Our approach employs an ensemble of models, each trained\nindependently and then fused at the score level using a Support Vector Machine\n(SVM) classifier. The models were trained using various strategies, including\nSelf-Supervised Learning (SSL) fine-tuning across different modalities: speech\nalone, text alone, and a combined speech and text approach. This joint training\nmethodology aims to enhance the system's ability to accurately classify\nemotional states. This joint training methodology aims to enhance the system's\nability to accurately classify emotional states. Thus, the system obtained\nF1-macro of 0.35\\% on development set.\n", "date": "2024-07-08", "categories": []}
{"id": "2407.06115", "title": "Infer Induced Sentiment of Comment Response to Video: A New Task,\n  Dataset and Baseline", "abstract": "  Existing video multi-modal sentiment analysis mainly focuses on the sentiment\nexpression of people within the video, yet often neglects the induced sentiment\nof viewers while watching the videos. Induced sentiment of viewers is essential\nfor inferring the public response to videos, has broad application in analyzing\npublic societal sentiment, effectiveness of advertising and other areas. The\nmicro videos and the related comments provide a rich application scenario for\nviewers induced sentiment analysis. In light of this, we introduces a novel\nresearch task, Multi-modal Sentiment Analysis for Comment Response of Video\nInduced(MSA-CRVI), aims to inferring opinions and emotions according to the\ncomments response to micro video. Meanwhile, we manually annotate a dataset\nnamed Comment Sentiment toward to Micro Video (CSMV) to support this research.\nIt is the largest video multi-modal sentiment dataset in terms of scale and\nvideo duration to our knowledge, containing 107,267 comments and 8,210 micro\nvideos with a video duration of 68.83 hours. To infer the induced sentiment of\ncomment should leverage the video content, so we propose the Video\nContent-aware Comment Sentiment Analysis (VC-CSA) method as baseline to address\nthe challenges inherent in this new task. Extensive experiments demonstrate\nthat our method is showing significant improvements over other established\nbaselines.\n", "date": "2024-05-15", "categories": []}
{"id": "2407.02751", "title": "Emotion and Intent Joint Understanding in Multimodal Conversation: A\n  Benchmarking Dataset", "abstract": "  Emotion and Intent Joint Understanding in Multimodal Conversation (MC-EIU)\naims to decode the semantic information manifested in a multimodal\nconversational history, while inferring the emotions and intents simultaneously\nfor the current utterance. MC-EIU is enabling technology for many\nhuman-computer interfaces. However, there is a lack of available datasets in\nterms of annotation, modality, language diversity, and accessibility. In this\nwork, we propose an MC-EIU dataset, which features 7 emotion categories, 9\nintent categories, 3 modalities, i.e., textual, acoustic, and visual content,\nand two languages, i.e., English and Mandarin. Furthermore, it is completely\nopen-source for free access. To our knowledge, MC-EIU is the first\ncomprehensive and rich emotion and intent joint understanding dataset for\nmultimodal conversation. Together with the release of the dataset, we also\ndevelop an Emotion and Intent Interaction (EI$^2$) network as a reference\nsystem by modeling the deep correlation between emotion and intent in the\nmultimodal conversation. With comparative experiments and ablation studies, we\ndemonstrate the effectiveness of the proposed EI$^2$ method on the MC-EIU\ndataset. The dataset and codes will be made available at:\nhttps://github.com/MC-EIU/MC-EIU.\n", "date": "2024-07-03", "categories": []}
{"id": "2407.03640", "title": "Generative Technology for Human Emotion Recognition: A Scope Review", "abstract": "  Affective computing stands at the forefront of artificial intelligence (AI),\nseeking to imbue machines with the ability to comprehend and respond to human\nemotions. Central to this field is emotion recognition, which endeavors to\nidentify and interpret human emotional states from different modalities, such\nas speech, facial images, text, and physiological signals. In recent years,\nimportant progress has been made in generative models, including Autoencoder,\nGenerative Adversarial Network, Diffusion Model, and Large Language Model.\nThese models, with their powerful data generation capabilities, emerge as\npivotal tools in advancing emotion recognition. However, up to now, there\nremains a paucity of systematic efforts that review generative technology for\nemotion recognition. This survey aims to bridge the gaps in the existing\nliterature by conducting a comprehensive analysis of over 320 research papers\nuntil June 2024. Specifically, this survey will firstly introduce the\nmathematical principles of different generative models and the commonly used\ndatasets. Subsequently, through a taxonomy, it will provide an in-depth\nanalysis of how generative techniques address emotion recognition based on\ndifferent modalities in several aspects, including data augmentation, feature\nextraction, semi-supervised learning, cross-domain, etc. Finally, the review\nwill outline future research directions, emphasizing the potential of\ngenerative models to advance the field of emotion recognition and enhance the\nemotional intelligence of AI systems.\n", "date": "2024-07-04", "categories": []}
{"id": "2407.03818", "title": "ConText at WASSA 2024 Empathy and Personality Shared Task:\n  History-Dependent Embedding Utterance Representations for Empathy and Emotion\n  Prediction in Conversations", "abstract": "  Empathy and emotion prediction are key components in the development of\neffective and empathetic agents, amongst several other applications. The WASSA\nshared task on empathy and emotion prediction in interactions presents an\nopportunity to benchmark approaches to these tasks. Appropriately selecting and\nrepresenting the historical context is crucial in the modelling of empathy and\nemotion in conversations. In our submissions, we model empathy, emotion\npolarity and emotion intensity of each utterance in a conversation by feeding\nthe utterance to be classified together with its conversational context, i.e.,\na certain number of previous conversational turns, as input to an encoder\nPre-trained Language Model, to which we append a regression head for\nprediction. We also model perceived counterparty empathy of each interlocutor\nby feeding all utterances from the conversation and a token identifying the\ninterlocutor for which we are predicting the empathy. Our system officially\nranked $1^{st}$ at the CONV-turn track and $2^{nd}$ at the CONV-dialog track.\n", "date": "2024-07-04", "categories": []}
{"id": "2407.04560", "title": "Real Time Emotion Analysis Using Deep Learning for Education,\n  Entertainment, and Beyond", "abstract": "  The significance of emotion detection is increasing in education,\nentertainment, and various other domains. We are developing a system that can\nidentify and transform facial expressions into emojis to provide immediate\nfeedback.The project consists of two components. Initially, we will employ\nsophisticated image processing techniques and neural networks to construct a\ndeep learning model capable of precisely categorising facial expressions. Next,\nwe will develop a basic application that records live video using the camera on\nyour device. The app will utilise a sophisticated model to promptly analyse\nfacial expressions and promptly exhibit corresponding emojis.Our objective is\nto develop a dynamic tool that integrates deep learning and real-time video\nprocessing for the purposes of online education, virtual events, gaming, and\nenhancing user experience. This tool enhances interactions and introduces novel\nemotional intelligence technologies.\n", "date": "2024-07-05", "categories": []}
{"id": "2407.04592", "title": "Smell and Emotion: Recognising emotions in smell-related artworks", "abstract": "  Emotions and smell are underrepresented in digital art history. In this\nexploratory work, we show that recognising emotions from smell-related artworks\nis technically feasible but has room for improvement. Using style transfer and\nhyperparameter optimization we achieve a minor performance boost and open up\nthe field for future extensions.\n", "date": "2024-07-05", "categories": []}
{"id": "2404.15310", "title": "Automated Assessment of Encouragement and Warmth in Classrooms\n  Leveraging Multimodal Emotional Features and ChatGPT", "abstract": "  Classroom observation protocols standardize the assessment of teaching\neffectiveness and facilitate comprehension of classroom interactions. Whereas\nthese protocols offer teachers specific feedback on their teaching practices,\nthe manual coding by human raters is resource-intensive and often unreliable.\nThis has sparked interest in developing AI-driven, cost-effective methods for\nautomating such holistic coding. Our work explores a multimodal approach to\nautomatically estimating encouragement and warmth in classrooms, a key\ncomponent of the Global Teaching Insights (GTI) study's observation protocol.\nTo this end, we employed facial and speech emotion recognition with sentiment\nanalysis to extract interpretable features from video, audio, and transcript\ndata. The prediction task involved both classification and regression methods.\nAdditionally, in light of recent large language models' remarkable text\nannotation capabilities, we evaluated ChatGPT's zero-shot performance on this\nscoring task based on transcripts. We demonstrated our approach on the GTI\ndataset, comprising 367 16-minute video segments from 92 authentic lesson\nrecordings. The inferences of GPT-4 and the best-trained model yielded\ncorrelations of r = .341 and r = .441 with human ratings, respectively.\nCombining estimates from both models through averaging, an ensemble approach\nachieved a correlation of r = .513, comparable to human inter-rater\nreliability. Our model explanation analysis indicated that text sentiment\nfeatures were the primary contributors to the trained model's decisions.\nMoreover, GPT-4 could deliver logical and concrete reasoning as potential\nteacher guidelines. Our findings provide insights into using advanced,\nmultimodal techniques for automated classroom observation, aiming to foster\nteacher training through frequent and valuable feedback.\n", "date": "2024-04-01", "categories": []}
{"id": "2405.17129", "title": "TEII: Think, Explain, Interact and Iterate with Large Language Models to\n  Solve Cross-lingual Emotion Detection", "abstract": "  Cross-lingual emotion detection allows us to analyze global trends, public\nopinion, and social phenomena at scale. We participated in the Explainability\nof Cross-lingual Emotion Detection (EXALT) shared task, achieving an F1-score\nof 0.6046 on the evaluation set for the emotion detection sub-task. Our system\noutperformed the baseline by more than 0.16 F1-score absolute, and ranked\nsecond amongst competing systems. We conducted experiments using fine-tuning,\nzero-shot learning, and few-shot learning for Large Language Model (LLM)-based\nmodels as well as embedding-based BiLSTM and KNN for non-LLM-based techniques.\nAdditionally, we introduced two novel methods: the Multi-Iteration Agentic\nWorkflow and the Multi-Binary-Classifier Agentic Workflow. We found that\nLLM-based approaches provided good performance on multilingual emotion\ndetection. Furthermore, ensembles combining all our experimented models yielded\nhigher F1-scores than any single approach alone.\n", "date": "2024-05-27", "categories": []}
{"id": "2407.02474", "title": "Free Energy in a Circumplex Model of Emotion", "abstract": "  Previous active inference accounts of emotion translate fluctuations in free\nenergy to a sense of emotion, mainly focusing on valence. However, in affective\nscience, emotions are often represented as multi-dimensional. In this paper, we\npropose to adopt a Circumplex Model of emotion by mapping emotions into a\ntwo-dimensional spectrum of valence and arousal. We show how one can derive a\nvalence and arousal signal from an agent's expected free energy, relating\narousal to the entropy of posterior beliefs and valence to utility less\nexpected utility. Under this formulation, we simulate artificial agents engaged\nin a search task. We show that the manipulation of priors and object presence\nresults in commonsense variability in emotional states.\n", "date": "2024-07-02", "categories": []}
{"id": "2405.12853", "title": "Inconsistency-Aware Cross-Attention for Audio-Visual Fusion in\n  Dimensional Emotion Recognition", "abstract": "  Leveraging complementary relationships across modalities has recently drawn a\nlot of attention in multimodal emotion recognition. Most of the existing\napproaches explored cross-attention to capture the complementary relationships\nacross the modalities. However, the modalities may also exhibit weak\ncomplementary relationships, which may deteriorate the cross-attended features,\nresulting in poor multimodal feature representations. To address this problem,\nwe propose Inconsistency-Aware Cross-Attention (IACA), which can adaptively\nselect the most relevant features on-the-fly based on the strong or weak\ncomplementary relationships across audio and visual modalities. Specifically,\nwe design a two-stage gating mechanism that can adaptively select the\nappropriate relevant features to deal with weak complementary relationships.\nExtensive experiments are conducted on the challenging Aff-Wild2 dataset to\nshow the robustness of the proposed model.\n", "date": "2024-05-21", "categories": []}
{"id": "2406.16442", "title": "EmoLLM: Multimodal Emotional Understanding Meets Large Language Models", "abstract": "  Multi-modal large language models (MLLMs) have achieved remarkable\nperformance on objective multimodal perception tasks, but their ability to\ninterpret subjective, emotionally nuanced multimodal content remains largely\nunexplored. Thus, it impedes their ability to effectively understand and react\nto the intricate emotions expressed by humans through multimodal media. To\nbridge this gap, we introduce EmoBench, the first comprehensive benchmark\ndesigned specifically to evaluate the emotional capabilities of MLLMs across\nfive popular emotional tasks, using a diverse dataset of 287k images and videos\npaired with corresponding textual instructions. Meanwhile, we propose EmoLLM, a\nnovel model for multimodal emotional understanding, incorporating with two core\ntechniques. 1) Multi-perspective Visual Projection, it captures diverse\nemotional cues from visual data from multiple perspectives. 2) EmoPrompt, it\nguides MLLMs to reason about emotions in the correct direction. Experimental\nresults demonstrate that EmoLLM significantly elevates multimodal emotional\nunderstanding performance, with an average improvement of 12.1% across multiple\nfoundation models on EmoBench. Our work contributes to the advancement of MLLMs\nby facilitating a deeper and more nuanced comprehension of intricate human\nemotions, paving the way for the development of artificial emotional\nintelligence capabilities with wide-ranging applications in areas such as\nhuman-computer interaction, mental health support, and empathetic AI systems.\nCode, data, and model will be released.\n", "date": "2024-06-24", "categories": []}
{"id": "2407.00024", "title": "LMVD: A Large-Scale Multimodal Vlog Dataset for Depression Detection in\n  the Wild", "abstract": "  Depression can significantly impact many aspects of an individual's life,\nincluding their personal and social functioning, academic and work performance,\nand overall quality of life. Many researchers within the field of affective\ncomputing are adopting deep learning technology to explore potential patterns\nrelated to the detection of depression. However, because of subjects' privacy\nprotection concerns, that data in this area is still scarce, presenting a\nchallenge for the deep discriminative models used in detecting depression. To\nnavigate these obstacles, a large-scale multimodal vlog dataset (LMVD), for\ndepression recognition in the wild is built. In LMVD, which has 1823 samples\nwith 214 hours of the 1475 participants captured from four multimedia platforms\n(Sina Weibo, Bilibili, Tiktok, and YouTube). A novel architecture termed\nMDDformer to learn the non-verbal behaviors of individuals is proposed.\nExtensive validations are performed on the LMVD dataset, demonstrating superior\nperformance for depression detection. We anticipate that the LMVD will\ncontribute a valuable function to the depression detection community. The data\nand code will released at the link: https://github.com/helang818/LMVD/.\n", "date": "2024-05-09", "categories": []}
{"id": "2407.00134", "title": "A Simple Attention-Based Mechanism for Bimodal Emotion Classification", "abstract": "  Big data contain rich information for machine learning algorithms to utilize\nwhen learning important features during classification tasks. Human beings\nexpress their emotion using certain words, speech (tone, pitch, speed) or\nfacial expression. Artificial Intelligence approach to emotion classification\nare largely based on learning from textual information. However, public\ndatasets containing text and speech data provide sufficient resources to train\nmachine learning algorithms for the tack of emotion classification. In this\npaper, we present novel bimodal deep learning-based architectures enhanced with\nattention mechanism trained and tested on text and speech data for emotion\nclassification. We report details of different deep learning based\narchitectures and show the performance of each architecture including rigorous\nerror analyses. Our finding suggests that deep learning based architectures\ntrained on different types of data (text and speech) outperform architectures\ntrained only on text or speech. Our proposed attention-based bimodal\narchitecture outperforms several state-of-the-art systems in emotion\nclassification.\n", "date": "2024-06-28", "categories": []}
{"id": "2407.00581", "title": "MasonTigers at SemEval-2024 Task 10: Emotion Discovery and Flip\n  Reasoning in Conversation with Ensemble of Transformers and Prompting", "abstract": "  In this paper, we present MasonTigers' participation in SemEval-2024 Task 10,\na shared task aimed at identifying emotions and understanding the rationale\nbehind their flips within monolingual English and Hindi-English code-mixed\ndialogues. This task comprises three distinct subtasks - emotion recognition in\nconversation for Hindi-English code-mixed dialogues, emotion flip reasoning for\nHindi-English code-mixed dialogues, and emotion flip reasoning for English\ndialogues. Our team, MasonTigers, contributed to each subtask, focusing on\ndeveloping methods for accurate emotion recognition and reasoning. By\nleveraging our approaches, we attained impressive F1-scores of 0.78 for the\nfirst task and 0.79 for both the second and third tasks. This performance not\nonly underscores the effectiveness of our methods across different aspects of\nthe task but also secured us the top rank in the first and third subtasks, and\nthe 2nd rank in the second subtask. Through extensive experimentation and\nanalysis, we provide insights into our system's performance and contributions\nto each subtask.\n", "date": "2024-06-30", "categories": []}
{"id": "2407.00679", "title": "Multi-Task Learning for Affect Analysis", "abstract": "  This Project was my Undergraduate Final Year dissertation, supervised by\nDimitrios Kollias This research delves into the realm of affective computing\nfor image analysis, aiming to enhance the efficiency and effectiveness of\nmulti-task learning in the context of emotion recognition. This project\ninvestigates two primary approaches: uni-task solutions and a multi-task\napproach to the same problems. Each approach undergoes testing, exploring\nvarious formulations, variations, and initialization strategies to come up with\nthe best configuration. The project utilizes existing a neural network\narchitecture, adapting it for multi-task learning by modifying output layers\nand loss functions. Tasks encompass 7 basic emotion recognition, action unit\ndetection, and valence-arousal estimation. Comparative analyses involve\nuni-task models for each individual task, facilitating the assessment of\nmulti-task model performance. Variations within each approach, including, loss\nfunctions, and hyperparameter tuning, undergo evaluation. The impact of\ndifferent initialization strategies and pre-training techniques on model\nconvergence and accuracy is explored. The research aspires to contribute to the\nburgeoning field of affective computing, with applications spanning healthcare,\nmarketing, and human-computer interaction. By systematically exploring\nmulti-task learning formulations, this research aims to contribute to the\ndevelopment of more accurate and efficient models for recognizing and\nunderstanding emotions in images. The findings hold promise for applications in\ndiverse industries, paving the way for advancements in affective computing\n", "date": "2024-06-30", "categories": []}
{"id": "2407.00743", "title": "AIMDiT: Modality Augmentation and Interaction via Multimodal Dimension\n  Transformation for Emotion Recognition in Conversations", "abstract": "  Emotion Recognition in Conversations (ERC) is a popular task in natural\nlanguage processing, which aims to recognize the emotional state of the speaker\nin conversations. While current research primarily emphasizes contextual\nmodeling, there exists a dearth of investigation into effective multimodal\nfusion methods. We propose a novel framework called AIMDiT to solve the problem\nof multimodal fusion of deep features. Specifically, we design a Modality\nAugmentation Network which performs rich representation learning through\ndimension transformation of different modalities and parameter-efficient\ninception block. On the other hand, the Modality Interaction Network performs\ninteraction fusion of extracted inter-modal features and intra-modal features.\nExperiments conducted using our AIMDiT framework on the public benchmark\ndataset MELD reveal 2.34% and 2.87% improvements in terms of the Acc-7 and w-F1\nmetrics compared to the state-of-the-art (SOTA) models.\n", "date": "2024-04-12", "categories": []}
{"id": "2407.01143", "title": "Are you sure? Analysing Uncertainty Quantification Approaches for\n  Real-world Speech Emotion Recognition", "abstract": "  Uncertainty Quantification (UQ) is an important building block for the\nreliable use of neural networks in real-world scenarios, as it can be a useful\ntool in identifying faulty predictions. Speech emotion recognition (SER) models\ncan suffer from particularly many sources of uncertainty, such as the ambiguity\nof emotions, Out-of-Distribution (OOD) data or, in general, poor recording\nconditions. Reliable UQ methods are thus of particular interest as in many SER\napplications no prediction is better than a faulty prediction. While the\neffects of label ambiguity on uncertainty are well documented in the\nliterature, we focus our work on an evaluation of UQ methods for SER under\ncommon challenges in real-world application, such as corrupted signals, and the\nabsence of speech. We show that simple UQ methods can already give an\nindication of the uncertainty of a prediction and that training with additional\nOOD data can greatly improve the identification of such signals.\n", "date": "2024-07-01", "categories": []}
{"id": "2402.14523", "title": "Daisy-TTS: Simulating Wider Spectrum of Emotions via Prosody Embedding\n  Decomposition", "abstract": "  We often verbally express emotions in a multifaceted manner, they may vary in\ntheir intensities and may be expressed not just as a single but as a mixture of\nemotions. This wide spectrum of emotions is well-studied in the structural\nmodel of emotions, which represents variety of emotions as derivative products\nof primary emotions with varying degrees of intensity. In this paper, we\npropose an emotional text-to-speech design to simulate a wider spectrum of\nemotions grounded on the structural model. Our proposed design, Daisy-TTS,\nincorporates a prosody encoder to learn emotionally-separable prosody embedding\nas a proxy for emotion. This emotion representation allows the model to\nsimulate: (1) Primary emotions, as learned from the training samples, (2)\nSecondary emotions, as a mixture of primary emotions, (3) Intensity-level, by\nscaling the emotion embedding, and (4) Emotions polarity, by negating the\nemotion embedding. Through a series of perceptual evaluations, Daisy-TTS\ndemonstrated overall higher emotional speech naturalness and emotion\nperceiveability compared to the baseline.\n", "date": "2024-02-22", "categories": []}
{"id": "2306.09361", "title": "MFSN: Multi-perspective Fusion Search Network For Pre-training Knowledge\n  in Speech Emotion Recognition", "abstract": "  Speech Emotion Recognition (SER) is an important research topic in\nhuman-computer interaction. Many recent works focus on directly extracting\nemotional cues through pre-trained knowledge, frequently overlooking\nconsiderations of appropriateness and comprehensiveness. Therefore, we propose\na novel framework for pre-training knowledge in SER, called Multi-perspective\nFusion Search Network (MFSN). Considering comprehensiveness, we partition\nspeech knowledge into Textual-related Emotional Content (TEC) and\nSpeech-related Emotional Content (SEC), capturing cues from both semantic and\nacoustic perspectives, and we design a new architecture search space to fully\nleverage them. Considering appropriateness, we verify the efficacy of different\nmodeling approaches in capturing SEC and fills the gap in current research.\nExperimental results on multiple datasets demonstrate the superiority of MFSN.\n", "date": "2023-06-12", "categories": []}
{"id": "2406.16000", "title": "Predicting Individual Depression Symptoms from Acoustic Features During\n  Speech", "abstract": "  Current automatic depression detection systems provide predictions directly\nwithout relying on the individual symptoms/items of depression as denoted in\nthe clinical depression rating scales. In contrast, clinicians assess each item\nin the depression rating scale in a clinical setting, thus implicitly providing\na more detailed rationale for a depression diagnosis. In this work, we make a\nfirst step towards using the acoustic features of speech to predict individual\nitems of the depression rating scale before obtaining the final depression\nprediction. For this, we use convolutional (CNN) and recurrent (long short-term\nmemory (LSTM)) neural networks. We consider different approaches to learning\nthe temporal context of speech. Further, we analyze two variants of voting\nschemes for individual item prediction and depression detection. We also\ninclude an animated visualization that shows an example of item prediction over\ntime as the speech progresses.\n", "date": "2024-06-23", "categories": []}
{"id": "2406.16187", "title": "Evaluation and Comparison of Emotionally Evocative Image Augmentation\n  Methods", "abstract": "  Experiments in affective computing are based on stimulus datasets that, in\nthe process of standardization, receive metadata describing which emotions each\nstimulus evokes. In this paper, we explore an approach to creating stimulus\ndatasets for affective computing using generative adversarial networks (GANs).\nTraditional dataset preparation methods are costly and time consuming,\nprompting our investigation of alternatives. We conducted experiments with\nvarious GAN architectures, including Deep Convolutional GAN, Conditional GAN,\nAuxiliary Classifier GAN, Progressive Augmentation GAN, and Wasserstein GAN,\nalongside data augmentation and transfer learning techniques. Our findings\nhighlight promising advances in the generation of emotionally evocative\nsynthetic images, suggesting significant potential for future research and\nimprovements in this domain.\n", "date": "2024-06-23", "categories": []}
{"id": "2406.16478", "title": "EMMI -- Empathic Multimodal Motivational Interviews Dataset: Analyses\n  and Annotations", "abstract": "  The study of multimodal interaction in therapy can yield a comprehensive\nunderstanding of therapist and patient behavior that can be used to develop a\nmultimodal virtual agent supporting therapy. This investigation aims to uncover\nhow therapists skillfully blend therapy's task goal (employing classical steps\nof Motivational Interviewing) with the social goal (building a trusting\nrelationship and expressing empathy). Furthermore, we seek to categorize\npatients into various ``types'' requiring tailored therapeutic approaches. To\nthis intent, we present multimodal annotations of a corpus consisting of\nsimulated motivational interviewing conversations, wherein actors portray the\nroles of patients and therapists. We introduce EMMI, composed of two publicly\navailable MI corpora, AnnoMI and the Motivational Interviewing Dataset, for\nwhich we add multimodal annotations. We analyze these annotations to\ncharacterize functional behavior for developing a virtual agent performing\nmotivational interviews emphasizing social and empathic behaviors. Our analysis\nfound three clusters of patients expressing significant differences in behavior\nand adaptation of the therapist's behavior to those types. This shows the\nimportance of a therapist being able to adapt their behavior depending on the\ncurrent situation within the dialog and the type of user.\n", "date": "2024-06-24", "categories": []}
{"id": "2406.08931", "title": "Exploring Multilingual Unseen Speaker Emotion Recognition: Leveraging\n  Co-Attention Cues in Multitask Learning", "abstract": "  Advent of modern deep learning techniques has given rise to advancements in\nthe field of Speech Emotion Recognition (SER). However, most systems prevalent\nin the field fail to generalize to speakers not seen during training. This\nstudy focuses on handling challenges of multilingual SER, specifically on\nunseen speakers. We introduce CAMuLeNet, a novel architecture leveraging\nco-attention based fusion and multitask learning to address this problem.\nAdditionally, we benchmark pretrained encoders of Whisper, HuBERT, Wav2Vec2.0,\nand WavLM using 10-fold leave-speaker-out cross-validation on five existing\nmultilingual benchmark datasets: IEMOCAP, RAVDESS, CREMA-D, EmoDB and CaFE and,\nrelease a novel dataset for SER on the Hindi language (BhavVani). CAMuLeNet\nshows an average improvement of approximately 8% over all benchmarks on unseen\nspeakers determined by our cross-validation strategy.\n", "date": "2024-06-13", "categories": []}
{"id": "2406.14014", "title": "Feature Fusion Based on Mutual-Cross-Attention Mechanism for EEG Emotion\n  Recognition", "abstract": "  An objective and accurate emotion diagnostic reference is vital to\npsychologists, especially when dealing with patients who are difficult to\ncommunicate with for pathological reasons. Nevertheless, current systems based\non Electroencephalography (EEG) data utilized for sentiment discrimination have\nsome problems, including excessive model complexity, mediocre accuracy, and\nlimited interpretability. Consequently, we propose a novel and effective\nfeature fusion mechanism named Mutual-Cross-Attention (MCA). Combining with a\nspecially customized 3D Convolutional Neural Network (3D-CNN), this purely\nmathematical mechanism adeptly discovers the complementary relationship between\ntime-domain and frequency-domain features in EEG data. Furthermore, the new\ndesigned Channel-PSD-DE 3D feature also contributes to the high performance.\nThe proposed method eventually achieves 99.49% (valence) and 99.30% (arousal)\naccuracy on DEAP dataset.\n", "date": "2024-06-20", "categories": []}
{"id": "2310.19995", "title": "Emotional Theory of Mind: Bridging Fast Visual Processing with Slow\n  Linguistic Reasoning", "abstract": "  The emotional theory of mind problem requires facial expressions, body pose,\ncontextual information and implicit commonsense knowledge to reason about the\nperson's emotion and its causes, making it currently one of the most difficult\nproblems in affective computing. In this work, we propose multiple methods to\nincorporate the emotional reasoning capabilities by constructing \"narrative\ncaptions\" relevant to emotion perception, that includes contextual and physical\nsignal descriptors that focuses on \"Who\", \"What\", \"Where\" and \"How\" questions\nrelated to the image and emotions of the individual. We propose two distinct\nways to construct these captions using zero-shot classifiers (CLIP) and\nfine-tuning visual-language models (LLaVA) over human generated descriptors. We\nfurther utilize these captions to guide the reasoning of language (GPT-4) and\nvision-language models (LLaVa, GPT-Vision). We evaluate the use of the\nresulting models in an image-to-language-to-emotion task. Our experiments\nshowed that combining the \"Fast\" narrative descriptors and \"Slow\" reasoning of\nlanguage models is a promising way to achieve emotional theory of mind.\n", "date": "2023-10-30", "categories": []}
{"id": "2403.11879", "title": "Unimodal Multi-Task Fusion for Emotional Mimicry Intensity Prediction", "abstract": "  In this research, we introduce a novel methodology for assessing Emotional\nMimicry Intensity (EMI) as part of the 6th Workshop and Competition on\nAffective Behavior Analysis in-the-wild. Our methodology utilises the Wav2Vec\n2.0 architecture, which has been pre-trained on an extensive podcast dataset,\nto capture a wide array of audio features that include both linguistic and\nparalinguistic components. We refine our feature extraction process by\nemploying a fusion technique that combines individual features with a global\nmean vector, thereby embedding a broader contextual understanding into our\nanalysis. A key aspect of our approach is the multi-task fusion strategy that\nnot only leverages these features but also incorporates a pre-trained\nValence-Arousal-Dominance (VAD) model. This integration is designed to refine\nemotion intensity prediction by concurrently processing multiple emotional\ndimensions, thereby embedding a richer contextual understanding into our\nframework. For the temporal analysis of audio data, our feature fusion process\nutilises a Long Short-Term Memory (LSTM) network. This approach, which relies\nsolely on the provided audio data, shows marked advancements over the existing\nbaseline, offering a more comprehensive understanding of emotional mimicry in\nnaturalistic settings, achieving the second place in the EMI challenge.\n", "date": "2024-03-18", "categories": []}
{"id": "2406.10275", "title": "ExHuBERT: Enhancing HuBERT Through Block Extension and Fine-Tuning on 37\n  Emotion Datasets", "abstract": "  Foundation models have shown great promise in speech emotion recognition\n(SER) by leveraging their pre-trained representations to capture emotion\npatterns in speech signals. To further enhance SER performance across various\nlanguages and domains, we propose a novel twofold approach. First, we gather\nEmoSet++, a comprehensive multi-lingual, multi-cultural speech emotion corpus\nwith 37 datasets, 150,907 samples, and a total duration of 119.5 hours. Second,\nwe introduce ExHuBERT, an enhanced version of HuBERT achieved by backbone\nextension and fine-tuning on EmoSet++. We duplicate each encoder layer and its\nweights, then freeze the first duplicate, integrating an extra zero-initialized\nlinear layer and skip connections to preserve functionality and ensure its\nadaptability for subsequent fine-tuning. Our evaluation on unseen datasets\nshows the efficacy of ExHuBERT, setting a new benchmark for various SER tasks.\nModel and details on EmoSet++: https://huggingface.co/amiriparian/ExHuBERT.\n", "date": "2024-06-11", "categories": []}
{"id": "2406.10561", "title": "We Care: Multimodal Depression Detection and Knowledge Infused Mental\n  Health Therapeutic Response Generation", "abstract": "  The detection of depression through non-verbal cues has gained significant\nattention. Previous research predominantly centred on identifying depression\nwithin the confines of controlled laboratory environments, often with the\nsupervision of psychologists or counsellors. Unfortunately, datasets generated\nin such controlled settings may struggle to account for individual behaviours\nin real-life situations. In response to this limitation, we present the\nExtended D-vlog dataset, encompassing a collection of 1, 261 YouTube vlogs.\nAdditionally, the emergence of large language models (LLMs) like GPT3.5, and\nGPT4 has sparked interest in their potential they can act like mental health\nprofessionals. Yet, the readiness of these LLM models to be used in real-life\nsettings is still a concern as they can give wrong responses that can harm the\nusers. We introduce a virtual agent serving as an initial contact for mental\nhealth patients, offering Cognitive Behavioral Therapy (CBT)-based responses.\nIt comprises two core functions: 1. Identifying depression in individuals, and\n2. Delivering CBT-based therapeutic responses. Our Mistral model achieved\nimpressive scores of 70.1% and 30.9% for distortion assessment and\nclassification, along with a Bert score of 88.7%. Moreover, utilizing the TVLT\nmodel on our Multimodal Extended D-vlog Dataset yielded outstanding results,\nwith an impressive F1-score of 67.8%\n", "date": "2024-06-15", "categories": []}
{"id": "2406.10741", "title": "Speech Emotion Recognition Using CNN and Its Use Case in Digital\n  Healthcare", "abstract": "  The process of identifying human emotion and affective states from speech is\nknown as speech emotion recognition (SER). This is based on the observation\nthat tone and pitch in the voice frequently convey underlying emotion. Speech\nrecognition includes the ability to recognize emotions, which is becoming\nincreasingly popular and in high demand. With the help of appropriate factors\n(such modalities, emotions, intensities, repetitions, etc.) found in the data,\nmy research seeks to use the Convolutional Neural Network (CNN) to distinguish\nemotions from audio recordings and label them in accordance with the range of\ndifferent emotions. I have developed a machine learning model to identify\nemotions from supplied audio files with the aid of machine learning methods.\nThe evaluation is mostly focused on precision, recall, and F1 score, which are\ncommon machine learning metrics. To properly set up and train the machine\nlearning framework, the main objective is to investigate the influence and\ncross-relation of all input and output parameters. To improve the ability to\nrecognize intentions, a key condition for communication, I have evaluated\nemotions using my specialized machine learning algorithm via voice that would\naddress the emotional state from voice with the help of digital healthcare,\nbridging the gap between human and artificial intelligence (AI).\n", "date": "2024-06-15", "categories": []}
{"id": "2406.11135", "title": "Towards Understanding Emotions for Engaged Mental Health Conversations", "abstract": "  Providing timely support and intervention is crucial in mental health\nsettings. As the need to engage youth comfortable with texting increases,\nmental health providers are exploring and adopting text-based media such as\nchatbots, community-based forums, online therapies with licensed professionals,\nand helplines operated by trained responders. To support these text-based media\nfor mental health--particularly for crisis care--we are developing a system to\nperform passive emotion-sensing using a combination of keystroke dynamics and\nsentiment analysis. Our early studies of this system posit that the analysis of\nshort text messages and keyboard typing patterns can provide emotion\ninformation that may be used to support both clients and responders. We use our\npreliminary findings to discuss the way forward for applying AI to support\nmental health providers in providing better care.\n", "date": "2024-06-17", "categories": []}
{"id": "2406.09933", "title": "What Does it Take to Generalize SER Model Across Datasets? A\n  Comprehensive Benchmark", "abstract": "  Speech emotion recognition (SER) is essential for enhancing human-computer\ninteraction in speech-based applications. Despite improvements in specific\nemotional datasets, there is still a research gap in SER's capability to\ngeneralize across real-world situations. In this paper, we investigate\napproaches to generalize the SER system across different emotion datasets. In\nparticular, we incorporate 11 emotional speech datasets and illustrate a\ncomprehensive benchmark on the SER task. We also address the challenge of\nimbalanced data distribution using over-sampling methods when combining SER\ndatasets for training. Furthermore, we explore various evaluation protocols for\nadeptness in the generalization of SER. Building on this, we explore the\npotential of Whisper for SER, emphasizing the importance of thorough\nevaluation. Our approach is designed to advance SER technology by integrating\nspeaker-independent methods.\n", "date": "2024-06-14", "categories": []}
{"id": "2402.17496", "title": "Emotional Voice Messages (EMOVOME) database: emotion recognition in\n  spontaneous voice messages", "abstract": "  Emotional Voice Messages (EMOVOME) is a spontaneous speech dataset containing\n999 audio messages from real conversations on a messaging app from 100 Spanish\nspeakers, gender balanced. Voice messages were produced in-the-wild conditions\nbefore participants were recruited, avoiding any conscious bias due to\nlaboratory environment. Audios were labeled in valence and arousal dimensions\nby three non-experts and two experts, which were then combined to obtain a\nfinal label per dimension. The experts also provided an extra label\ncorresponding to seven emotion categories. To set a baseline for future\ninvestigations using EMOVOME, we implemented emotion recognition models using\nboth speech and audio transcriptions. For speech, we used the standard eGeMAPS\nfeature set and support vector machines, obtaining 49.27% and 44.71% unweighted\naccuracy for valence and arousal respectively. For text, we fine-tuned a\nmultilingual BERT model and achieved 61.15% and 47.43% unweighted accuracy for\nvalence and arousal respectively. This database will significantly contribute\nto research on emotion recognition in the wild, while also providing a unique\nnatural and freely accessible resource for Spanish.\n", "date": "2024-02-27", "categories": []}
{"id": "2406.07753", "title": "The MuSe 2024 Multimodal Sentiment Analysis Challenge: Social Perception\n  and Humor Recognition", "abstract": "  The Multimodal Sentiment Analysis Challenge (MuSe) 2024 addresses two\ncontemporary multimodal affect and sentiment analysis problems: In the Social\nPerception Sub-Challenge (MuSe-Perception), participants will predict 16\ndifferent social attributes of individuals such as assertiveness, dominance,\nlikability, and sincerity based on the provided audio-visual data. The\nCross-Cultural Humor Detection Sub-Challenge (MuSe-Humor) dataset expands upon\nthe Passau Spontaneous Football Coach Humor (Passau-SFCH) dataset, focusing on\nthe detection of spontaneous humor in a cross-lingual and cross-cultural\nsetting. The main objective of MuSe 2024 is to unite a broad audience from\nvarious research domains, including multimodal sentiment analysis, audio-visual\naffective computing, continuous signal processing, and natural language\nprocessing. By fostering collaboration and exchange among experts in these\nfields, the MuSe 2024 endeavors to advance the understanding and application of\nsentiment analysis and affective computing across multiple modalities. This\nbaseline paper provides details on each sub-challenge and its corresponding\ndataset, extracted features from each data modality, and discusses challenge\nbaselines. For our baseline system, we make use of a range of Transformers and\nexpert-designed features and train Gated Recurrent Unit (GRU)-Recurrent Neural\nNetwork (RNN) models on them, resulting in a competitive baseline system. On\nthe unseen test datasets of the respective sub-challenges, it achieves a mean\nPearson's Correlation Coefficient ($\\rho$) of 0.3573 for MuSe-Perception and an\nArea Under the Curve (AUC) value of 0.8682 for MuSe-Humor.\n", "date": "2024-06-11", "categories": []}
{"id": "2406.07162", "title": "EmoBox: Multilingual Multi-corpus Speech Emotion Recognition Toolkit and\n  Benchmark", "abstract": "  Speech emotion recognition (SER) is an important part of human-computer\ninteraction, receiving extensive attention from both industry and academia.\nHowever, the current research field of SER has long suffered from the following\nproblems: 1) There are few reasonable and universal splits of the datasets,\nmaking comparing different models and methods difficult. 2) No commonly used\nbenchmark covers numerous corpus and languages for researchers to refer to,\nmaking reproduction a burden. In this paper, we propose EmoBox, an\nout-of-the-box multilingual multi-corpus speech emotion recognition toolkit,\nalong with a benchmark for both intra-corpus and cross-corpus settings. For\nintra-corpus settings, we carefully designed the data partitioning for\ndifferent datasets. For cross-corpus settings, we employ a foundation SER\nmodel, emotion2vec, to mitigate annotation errors and obtain a test set that is\nfully balanced in speakers and emotions distributions. Based on EmoBox, we\npresent the intra-corpus SER results of 10 pre-trained speech models on 32\nemotion datasets with 14 languages, and the cross-corpus SER results on 4\ndatasets with the fully balanced test sets. To the best of our knowledge, this\nis the largest SER benchmark, across language scopes and quantity scales. We\nhope that our toolkit and benchmark can facilitate the research of SER in the\ncommunity.\n", "date": "2024-06-11", "categories": []}
{"id": "2404.16456", "title": "Correlation-Decoupled Knowledge Distillation for Multimodal Sentiment\n  Analysis with Incomplete Modalities", "abstract": "  Multimodal sentiment analysis (MSA) aims to understand human sentiment\nthrough multimodal data. Most MSA efforts are based on the assumption of\nmodality completeness. However, in real-world applications, some practical\nfactors cause uncertain modality missingness, which drastically degrades the\nmodel's performance. To this end, we propose a Correlation-decoupled Knowledge\nDistillation (CorrKD) framework for the MSA task under uncertain missing\nmodalities. Specifically, we present a sample-level contrastive distillation\nmechanism that transfers comprehensive knowledge containing cross-sample\ncorrelations to reconstruct missing semantics. Moreover, a category-guided\nprototype distillation mechanism is introduced to capture cross-category\ncorrelations using category prototypes to align feature distributions and\ngenerate favorable joint representations. Eventually, we design a\nresponse-disentangled consistency distillation strategy to optimize the\nsentiment decision boundaries of the student network through response\ndisentanglement and mutual information maximization. Comprehensive experiments\non three datasets indicate that our framework can achieve favorable\nimprovements compared with several baselines.\n", "date": "2024-04-25", "categories": []}
{"id": "2312.11111", "title": "The Good, The Bad, and Why: Unveiling Emotions in Generative AI", "abstract": "  Emotion significantly impacts our daily behaviors and interactions. While\nrecent generative AI models, such as large language models, have shown\nimpressive performance in various tasks, it remains unclear whether they truly\ncomprehend emotions. This paper aims to address this gap by incorporating\npsychological theories to gain a holistic understanding of emotions in\ngenerative AI models. Specifically, we propose three approaches: 1)\nEmotionPrompt to enhance AI model performance, 2) EmotionAttack to impair AI\nmodel performance, and 3) EmotionDecode to explain the effects of emotional\nstimuli, both benign and malignant. Through extensive experiments involving\nlanguage and multi-modal models on semantic understanding, logical reasoning,\nand generation tasks, we demonstrate that both textual and visual EmotionPrompt\ncan boost the performance of AI models while EmotionAttack can hinder it.\nAdditionally, EmotionDecode reveals that AI models can comprehend emotional\nstimuli akin to the mechanism of dopamine in the human brain. Our work heralds\na novel avenue for exploring psychology to enhance our understanding of\ngenerative AI models.\n", "date": "2023-12-18", "categories": []}
{"id": "2406.04758", "title": "Think out Loud: Emotion Deducing Explanation in Dialogues", "abstract": "  Humans convey emotions through daily dialogues, making emotion understanding\na crucial step of affective intelligence. To understand emotions in dialogues,\nmachines are asked to recognize the emotion for an utterance (Emotion\nRecognition in Dialogues, ERD); based on the emotion, then find causal\nutterances for the emotion (Emotion Cause Extraction in Dialogues, ECED). The\nsetting of the two tasks requires first ERD and then ECED, ignoring the mutual\ncomplement between emotion and cause. To fix this, some new tasks are proposed\nto extract them simultaneously. Although the current research on these tasks\nhas excellent achievements, simply identifying emotion-related factors by\nclassification modeling lacks realizing the specific thinking process of causes\nstimulating the emotion in an explainable way. This thinking process especially\nreflected in the reasoning ability of Large Language Models (LLMs) is\nunder-explored. To this end, we propose a new task \"Emotion Deducing\nExplanation in Dialogues\" (EDEN). EDEN recognizes emotion and causes in an\nexplicitly thinking way. That is, models need to generate an explanation text,\nwhich first summarizes the causes; analyzes the inner activities of the\nspeakers triggered by the causes using common sense; then guesses the emotion\naccordingly. To support the study of EDEN, based on the existing resources in\nECED, we construct two EDEN datasets by human effort. We further evaluate\ndifferent models on EDEN and find that LLMs are more competent than\nconventional PLMs. Besides, EDEN can help LLMs achieve better recognition of\nemotions and causes, which explores a new research direction of explainable\nemotion understanding in dialogues.\n", "date": "2024-06-07", "categories": []}
{"id": "2405.20172", "title": "Iterative Feature Boosting for Explainable Speech Emotion Recognition", "abstract": "  In speech emotion recognition (SER), using predefined features without\nconsidering their practical importance may lead to high dimensional datasets,\nincluding redundant and irrelevant information. Consequently, high-dimensional\nlearning often results in decreasing model accuracy while increasing\ncomputational complexity. Our work underlines the importance of carefully\nconsidering and analyzing features in order to build efficient SER systems. We\npresent a new supervised SER method based on an efficient feature engineering\napproach. We pay particular attention to the explainability of results to\nevaluate feature relevance and refine feature sets. This is performed\niteratively through feature evaluation loop, using Shapley values to boost\nfeature selection and improve overall framework performance. Our approach\nallows thus to balance the benefits between model performance and transparency.\nThe proposed method outperforms human-level performance (HLP) and\nstate-of-the-art machine learning methods in emotion recognition on the TESS\ndataset. The source code of this paper is publicly available at\nhttps://github.com/alaaNfissi/Iterative-Feature-Boosting-for-Explainable-Speech-Emotion-Recognition.\n", "date": "2024-05-30", "categories": []}
{"id": "2406.01624", "title": "Unveiling Hidden Factors: Explainable AI for Feature Boosting in Speech\n  Emotion Recognition", "abstract": "  Speech emotion recognition (SER) has gained significant attention due to its\nseveral application fields, such as mental health, education, and\nhuman-computer interaction. However, the accuracy of SER systems is hindered by\nhigh-dimensional feature sets that may contain irrelevant and redundant\ninformation. To overcome this challenge, this study proposes an iterative\nfeature boosting approach for SER that emphasizes feature relevance and\nexplainability to enhance machine learning model performance. Our approach\ninvolves meticulous feature selection and analysis to build efficient SER\nsystems. In addressing our main problem through model explainability, we employ\na feature evaluation loop with Shapley values to iteratively refine feature\nsets. This process strikes a balance between model performance and\ntransparency, which enables a comprehensive understanding of the model's\npredictions. The proposed approach offers several advantages, including the\nidentification and removal of irrelevant and redundant features, leading to a\nmore effective model. Additionally, it promotes explainability, facilitating\ncomprehension of the model's predictions and the identification of crucial\nfeatures for emotion determination. The effectiveness of the proposed method is\nvalidated on the SER benchmarks of the Toronto emotional speech set (TESS),\nBerlin Database of Emotional Speech (EMO-DB), Ryerson Audio-Visual Database of\nEmotional Speech and Song (RAVDESS), and Surrey Audio-Visual Expressed Emotion\n(SAVEE) datasets, outperforming state-of-the-art methods. To the best of our\nknowledge, this is the first work to incorporate model explainability into an\nSER framework. The source code of this paper is publicly available via this\nhttps://github.com/alaaNfissi/Unveiling-Hidden-Factors-Explainable-AI-for-Feature-Boosting-in-Speech-Emotion-Recognition.\n", "date": "2024-06-01", "categories": []}
{"id": "2402.01355", "title": "FindingEmo: An Image Dataset for Emotion Recognition in the Wild", "abstract": "  We introduce FindingEmo, a new image dataset containing annotations for 25k\nimages, specifically tailored to Emotion Recognition. Contrary to existing\ndatasets, it focuses on complex scenes depicting multiple people in various\nnaturalistic, social settings, with images being annotated as a whole, thereby\ngoing beyond the traditional focus on faces or single individuals. Annotated\ndimensions include Valence, Arousal and Emotion label, with annotations\ngathered using Prolific. Together with the annotations, we release the list of\nURLs pointing to the original images, as well as all associated source code.\n", "date": "2024-02-02", "categories": []}
{"id": "2406.03004", "title": "Evaluation of data inconsistency for multi-modal sentiment analysis", "abstract": "  Emotion semantic inconsistency is an ubiquitous challenge in multi-modal\nsentiment analysis (MSA). MSA involves analyzing sentiment expressed across\nvarious modalities like text, audio, and videos. Each modality may convey\ndistinct aspects of sentiment, due to subtle and nuanced expression of human\nbeings, leading to inconsistency, which may hinder the prediction of artificial\nagents. In this work, we introduce a modality conflicting test set and assess\nthe performance of both traditional multi-modal sentiment analysis models and\nmulti-modal large language models (MLLMs). Our findings reveal significant\nperformance degradation across traditional models when confronted with\nsemantically conflicting data and point out the drawbacks of MLLMs when\nhandling multi-modal emotion analysis. Our research presents a new challenge\nand offer valuable insights for the future development of sentiment analysis\nsystems.\n", "date": "2024-06-05", "categories": []}
{"id": "2406.02251", "title": "Modeling Emotional Trajectories in Written Stories Utilizing\n  Transformers and Weakly-Supervised Learning", "abstract": "  Telling stories is an integral part of human communication which can evoke\nemotions and influence the affective states of the audience. Automatically\nmodeling emotional trajectories in stories has thus attracted considerable\nscholarly interest. However, as most existing works have been limited to\nunsupervised dictionary-based approaches, there is no benchmark for this task.\nWe address this gap by introducing continuous valence and arousal labels for an\nexisting dataset of children's stories originally annotated with discrete\nemotion categories. We collect additional annotations for this data and map the\ncategorical labels to the continuous valence and arousal space. For predicting\nthe thus obtained emotionality signals, we fine-tune a DeBERTa model and\nimprove upon this baseline via a weakly supervised learning approach. The best\nconfiguration achieves a Concordance Correlation Coefficient (CCC) of $.8221$\nfor valence and $.7125$ for arousal on the test set, demonstrating the efficacy\nof our proposed approach. A detailed analysis shows the extent to which the\nresults vary depending on factors such as the author, the individual story, or\nthe section within the story. In addition, we uncover the weaknesses of our\napproach by investigating examples that prove to be difficult to predict.\n", "date": "2024-06-04", "categories": []}
{"id": "2308.00264", "title": "Multimodal Multi-loss Fusion Network for Sentiment Analysis", "abstract": "  This paper investigates the optimal selection and fusion of feature encoders\nacross multiple modalities and combines these in one neural network to improve\nsentiment detection. We compare different fusion methods and examine the impact\nof multi-loss training within the multi-modality fusion network, identifying\nsurprisingly important findings relating to subnet performance. We have also\nfound that integrating context significantly enhances model performance. Our\nbest model achieves state-of-the-art performance for three datasets (CMU-MOSI,\nCMU-MOSEI and CH-SIMS). These results suggest a roadmap toward an optimized\nfeature selection and fusion approach for enhancing sentiment detection in\nneural networks.\n", "date": "2023-08-01", "categories": []}
{"id": "2403.05963", "title": "Robust Emotion Recognition in Context Debiasing", "abstract": "  Context-aware emotion recognition (CAER) has recently boosted the practical\napplications of affective computing techniques in unconstrained environments.\nMainstream CAER methods invariably extract ensemble representations from\ndiverse contexts and subject-centred characteristics to perceive the target\nperson's emotional state. Despite advancements, the biggest challenge remains\ndue to context bias interference. The harmful bias forces the models to rely on\nspurious correlations between background contexts and emotion labels in\nlikelihood estimation, causing severe performance bottlenecks and confounding\nvaluable context priors. In this paper, we propose a counterfactual emotion\ninference (CLEF) framework to address the above issue. Specifically, we first\nformulate a generalized causal graph to decouple the causal relationships among\nthe variables in CAER. Following the causal graph, CLEF introduces a\nnon-invasive context branch to capture the adverse direct effect caused by the\ncontext bias. During the inference, we eliminate the direct context effect from\nthe total causal effect by comparing factual and counterfactual outcomes,\nresulting in bias mitigation and robust prediction. As a model-agnostic\nframework, CLEF can be readily integrated into existing methods, bringing\nconsistent performance gains.\n", "date": "2024-03-09", "categories": []}
{"id": "2401.11017", "title": "Revealing Emotional Clusters in Speaker Embeddings: A Contrastive\n  Learning Strategy for Speech Emotion Recognition", "abstract": "  Speaker embeddings carry valuable emotion-related information, which makes\nthem a promising resource for enhancing speech emotion recognition (SER),\nespecially with limited labeled data. Traditionally, it has been assumed that\nemotion information is indirectly embedded within speaker embeddings, leading\nto their under-utilization. Our study reveals a direct and useful link between\nemotion and state-of-the-art speaker embeddings in the form of intra-speaker\nclusters. By conducting a thorough clustering analysis, we demonstrate that\nemotion information can be readily extracted from speaker embeddings. In order\nto leverage this information, we introduce a novel contrastive pretraining\napproach applied to emotion-unlabeled data for speech emotion recognition. The\nproposed approach involves the sampling of positive and the negative examples\nbased on the intra-speaker clusters of speaker embeddings. The proposed\nstrategy, which leverages extensive emotion-unlabeled data, leads to a\nsignificant improvement in SER performance, whether employed as a standalone\npretraining task or integrated into a multi-task pretraining setting.\n", "date": "2024-01-19", "categories": []}
{"id": "2405.20600", "title": "Multi-label Class Incremental Emotion Decoding with Augmented Emotional\n  Semantics Learning", "abstract": "  Emotion decoding plays an important role in affective human-computer\ninteraction. However, previous studies ignored the dynamic real-world scenario,\nwhere human experience a blend of multiple emotions which are incrementally\nintegrated into the model, leading to the multi-label class incremental\nlearning (MLCIL) problem. Existing methods have difficulty in solving MLCIL\nissue due to notorious catastrophic forgetting caused by partial label problem\nand inadequate label semantics mining. In this paper, we propose an augmented\nemotional semantics learning framework for multi-label class incremental\nemotion decoding. Specifically, we design an augmented emotional relation graph\nmodule with label disambiguation to handle the past-missing partial label\nproblem. Then, we leverage domain knowledge from affective dimension space to\nalleviate future-missing partial label problem by knowledge distillation.\nBesides, an emotional semantics learning module is constructed with a graph\nautoencoder to obtain emotion embeddings in order to guide the\nsemantic-specific feature decoupling for better multi-label learning. Extensive\nexperiments on three datasets show the superiority of our method for improving\nemotion decoding performance and mitigating forgetting on MLCIL problem.\n", "date": "2024-05-31", "categories": []}
{"id": "2405.19373", "title": "Multi-modal Mood Reader: Pre-trained Model Empowers Cross-Subject\n  Emotion Recognition", "abstract": "  Emotion recognition based on Electroencephalography (EEG) has gained\nsignificant attention and diversified development in fields such as neural\nsignal processing and affective computing. However, the unique brain anatomy of\nindividuals leads to non-negligible natural differences in EEG signals across\nsubjects, posing challenges for cross-subject emotion recognition. While recent\nstudies have attempted to address these issues, they still face limitations in\npractical effectiveness and model framework unity. Current methods often\nstruggle to capture the complex spatial-temporal dynamics of EEG signals and\nfail to effectively integrate multimodal information, resulting in suboptimal\nperformance and limited generalizability across subjects. To overcome these\nlimitations, we develop a Pre-trained model based Multimodal Mood Reader for\ncross-subject emotion recognition that utilizes masked brain signal modeling\nand interlinked spatial-temporal attention mechanism. The model learns\nuniversal latent representations of EEG signals through pre-training on large\nscale dataset, and employs Interlinked spatial-temporal attention mechanism to\nprocess Differential Entropy(DE) features extracted from EEG data.\nSubsequently, a multi-level fusion layer is proposed to integrate the\ndiscriminative features, maximizing the advantages of features across different\ndimensions and modalities. Extensive experiments on public datasets demonstrate\nMood Reader's superior performance in cross-subject emotion recognition tasks,\noutperforming state-of-the-art methods. Additionally, the model is dissected\nfrom attention perspective, providing qualitative analysis of emotion-related\nbrain areas, offering valuable insights for affective research in neural signal\nprocessing.\n", "date": "2024-05-28", "categories": []}
{"id": "2405.18510", "title": "Improved Emotional Alignment of AI and Humans: Human Ratings of Emotions\n  Expressed by Stable Diffusion v1, DALL-E 2, and DALL-E 3", "abstract": "  Generative AI systems are increasingly capable of expressing emotions via\ntext and imagery. Effective emotional expression will likely play a major role\nin the efficacy of AI systems -- particularly those designed to support human\nmental health and wellbeing. This motivates our present research to better\nunderstand the alignment of AI expressed emotions with the human perception of\nemotions. When AI tries to express a particular emotion, how might we assess\nwhether they are successful? To answer this question, we designed a survey to\nmeasure the alignment between emotions expressed by generative AI and human\nperceptions. Three generative image models (DALL-E 2, DALL-E 3 and Stable\nDiffusion v1) were used to generate 240 examples of images, each of which was\nbased on a prompt designed to express five positive and five negative emotions\nacross both humans and robots. 24 participants recruited from the Prolific\nwebsite rated the alignment of AI-generated emotional expressions with a text\nprompt used to generate the emotion (i.e., \"A robot expressing the emotion\namusement\"). The results of our evaluation suggest that generative AI models\nare indeed capable of producing emotional expressions that are well-aligned\nwith a range of human emotions; however, we show that the alignment\nsignificantly depends upon the AI model used and the emotion itself. We analyze\nvariations in the performance of these systems to identify gaps for future\nimprovement. We conclude with a discussion of the implications for future AI\nsystems designed to support mental health and wellbeing.\n", "date": "2024-05-28", "categories": []}
{"id": "2405.18769", "title": "OUS: Scene-Guided Dynamic Facial Expression Recognition", "abstract": "  Dynamic Facial Expression Recognition (DFER) is crucial for affective\ncomputing but often overlooks the impact of scene context. We have identified a\nsignificant issue in current DFER tasks: human annotators typically integrate\nemotions from various angles, including environmental cues and body language,\nwhereas existing DFER methods tend to consider the scene as noise that needs to\nbe filtered out, focusing solely on facial information. We refer to this as the\nRigid Cognitive Problem. The Rigid Cognitive Problem can lead to discrepancies\nbetween the cognition of annotators and models in some samples. To align more\nclosely with the human cognitive paradigm of emotions, we propose an Overall\nUnderstanding of the Scene DFER method (OUS). OUS effectively integrates scene\nand facial features, combining scene-specific emotional knowledge for DFER.\nExtensive experiments on the two largest datasets in the DFER field, DFEW and\nFERV39k, demonstrate that OUS significantly outperforms existing methods. By\nanalyzing the Rigid Cognitive Problem, OUS successfully understands the complex\nrelationship between scene context and emotional expression, closely aligning\nwith human emotional understanding in real-world scenarios.\n", "date": "2024-05-29", "categories": []}
{"id": "2401.13260", "title": "MF-AED-AEC: Speech Emotion Recognition by Leveraging Multimodal Fusion,\n  Asr Error Detection, and Asr Error Correction", "abstract": "  The prevalent approach in speech emotion recognition (SER) involves\nintegrating both audio and textual information to comprehensively identify the\nspeaker's emotion, with the text generally obtained through automatic speech\nrecognition (ASR). An essential issue of this approach is that ASR errors from\nthe text modality can worsen the performance of SER. Previous studies have\nproposed using an auxiliary ASR error detection task to adaptively assign\nweights of each word in ASR hypotheses. However, this approach has limited\nimprovement potential because it does not address the coherence of semantic\ninformation in the text. Additionally, the inherent heterogeneity of different\nmodalities leads to distribution gaps between their representations, making\ntheir fusion challenging. Therefore, in this paper, we incorporate two\nauxiliary tasks, ASR error detection (AED) and ASR error correction (AEC), to\nenhance the semantic coherence of ASR text, and further introduce a novel\nmulti-modal fusion (MF) method to learn shared representations across\nmodalities. We refer to our method as MF-AED-AEC. Experimental results indicate\nthat MF-AED-AEC significantly outperforms the baseline model by a margin of\n4.1\\%.\n", "date": "2024-01-24", "categories": []}
{"id": "2405.17900", "title": "Enhancing Emotion Recognition in Conversation through Emotional\n  Cross-Modal Fusion and Inter-class Contrastive Learning", "abstract": "  The purpose of emotion recognition in conversation (ERC) is to identify the\nemotion category of an utterance based on contextual information. Previous ERC\nmethods relied on simple connections for cross-modal fusion and ignored the\ninformation differences between modalities, resulting in the model being unable\nto focus on modality-specific emotional information. At the same time, the\nshared information between modalities was not processed to generate emotions.\nInformation redundancy problem. To overcome these limitations, we propose a\ncross-modal fusion emotion prediction network based on vector connections. The\nnetwork mainly includes two stages: the multi-modal feature fusion stage based\non connection vectors and the emotion classification stage based on fused\nfeatures. Furthermore, we design a supervised inter-class contrastive learning\nmodule based on emotion labels. Experimental results confirm the effectiveness\nof the proposed method, demonstrating excellent performance on the IEMOCAP and\nMELD datasets.\n", "date": "2024-05-28", "categories": []}
{"id": "2405.16701", "title": "Detail-Enhanced Intra- and Inter-modal Interaction for Audio-Visual\n  Emotion Recognition", "abstract": "  Capturing complex temporal relationships between video and audio modalities\nis vital for Audio-Visual Emotion Recognition (AVER). However, existing methods\nlack attention to local details, such as facial state changes between video\nframes, which can reduce the discriminability of features and thus lower\nrecognition accuracy. In this paper, we propose a Detail-Enhanced Intra- and\nInter-modal Interaction network(DE-III) for AVER, incorporating several novel\naspects. We introduce optical flow information to enrich video representations\nwith texture details that better capture facial state changes. A fusion module\nintegrates the optical flow estimation with the corresponding video frames to\nenhance the representation of facial texture variations. We also design\nattentive intra- and inter-modal feature enhancement modules to further improve\nthe richness and discriminability of video and audio representations. A\ndetailed quantitative evaluation shows that our proposed model outperforms all\nexisting methods on three benchmark datasets for both concrete and continuous\nemotion recognition. To encourage further research and ensure replicability, we\nwill release our full code upon acceptance.\n", "date": "2024-05-26", "categories": []}
{"id": "2405.15708", "title": "EmpathicStories++: A Multimodal Dataset for Empathy towards Personal\n  Experiences", "abstract": "  Modeling empathy is a complex endeavor that is rooted in interpersonal and\nexperiential dimensions of human interaction, and remains an open problem\nwithin AI. Existing empathy datasets fall short in capturing the richness of\nempathy responses, often being confined to in-lab or acted scenarios, lacking\nlongitudinal data, and missing self-reported labels. We introduce a new\nmultimodal dataset for empathy during personal experience sharing: the\nEmpathicStories++ dataset\n(https://mitmedialab.github.io/empathic-stories-multimodal/) containing 53\nhours of video, audio, and text data of 41 participants sharing vulnerable\nexperiences and reading empathically resonant stories with an AI agent.\nEmpathicStories++ is the first longitudinal dataset on empathy, collected over\na month-long deployment of social robots in participants' homes, as\nparticipants engage in natural, empathic storytelling interactions with AI\nagents. We then introduce a novel task of predicting individuals' empathy\ntoward others' stories based on their personal experiences, evaluated in two\ncontexts: participants' own personal shared story context and their reflections\non stories they read. We benchmark this task using state-of-the-art models to\npave the way for future improvements in contextualized and longitudinal empathy\nmodeling. Our work provides a valuable resource for further research in\ndeveloping empathetic AI systems and understanding the intricacies of human\nempathy within genuine, real-world settings.\n", "date": "2024-05-24", "categories": []}
{"id": "2405.13903", "title": "ST-Gait++: Leveraging spatio-temporal convolutions for gait-based\n  emotion recognition on videos", "abstract": "  Emotion recognition is relevant for human behaviour understanding, where\nfacial expression and speech recognition have been widely explored by the\ncomputer vision community. Literature in the field of behavioural psychology\nindicates that gait, described as the way a person walks, is an additional\nindicator of emotions. In this work, we propose a deep framework for emotion\nrecognition through the analysis of gait. More specifically, our model is\ncomposed of a sequence of spatial-temporal Graph Convolutional Networks that\nproduce a robust skeleton-based representation for the task of emotion\nclassification. We evaluate our proposed framework on the E-Gait dataset,\ncomposed of a total of 2177 samples. The results obtained represent an\nimprovement of approximately 5% in accuracy compared to the state of the art.\nIn addition, during training we observed a faster convergence of our model\ncompared to the state-of-the-art methodologies.\n", "date": "2024-05-22", "categories": []}
{"id": "2405.14385", "title": "Emotion Identification for French in Written Texts: Considering their\n  Modes of Expression as a Step Towards Text Complexity Analysis", "abstract": "  The objective of this paper is to predict (A) whether a sentence in a written\ntext expresses an emotion, (B) the mode(s) in which it is expressed, (C)\nwhether it is basic or complex, and (D) its emotional category.\n  One of our major contributions, through a dataset and a model, is to\nintegrate the fact that an emotion can be expressed in different modes: from a\ndirect mode, essentially lexicalized, to a more indirect mode, where emotions\nwill only be suggested, a mode that NLP approaches generally don't take into\naccount.\n  Another originality is that the scope is on written texts, as opposed usual\nwork focusing on conversational (often multi-modal) data. In this context,\nmodes of expression are seen as a factor towards the automatic analysis of\ncomplexity in texts.\n  Experiments on French texts show acceptable results compared to the human\nannotators' agreement, and outperforming results compared to using a large\nlanguage model with in-context learning (i.e. no fine-tuning).\n", "date": "2024-05-23", "categories": []}
{"id": "2405.10860", "title": "ECR-Chain: Advancing Generative Language Models to Better Emotion-Cause\n  Reasoners through Reasoning Chains", "abstract": "  Understanding the process of emotion generation is crucial for analyzing the\ncauses behind emotions. Causal Emotion Entailment (CEE), an\nemotion-understanding task, aims to identify the causal utterances in a\nconversation that stimulate the emotions expressed in a target utterance.\nHowever, current works in CEE mainly focus on modeling semantic and emotional\ninteractions in conversations, neglecting the exploration of the\nemotion-generation process. This hinders the models from deeply understanding\nemotions, restricting their ability to produce explainable predictions. In this\nwork, inspired by the emotion generation process of\n\"stimulus-appraisal-emotion\" in the cognitive appraisal theory, we introduce a\nstep-by-step reasoning method, Emotion-Cause Reasoning Chain (ECR-Chain), to\ninfer the stimulus from the target emotional expressions in conversations.\nSpecifically, we first introduce the ECR-Chain to ChatGPT via few-shot\nprompting, which significantly improves its performance on the CEE task. We\nfurther propose an automated construction process to utilize ChatGPT in\nbuilding an ECR-Chain set, which can enhance the reasoning abilities of smaller\nmodels through supervised training and assist the Vicuna-7B model in achieving\nstate-of-the-art CEE performance. Moreover, our methods can enable these\ngenerative language models to effectively perform emotion-cause reasoning in an\nexplainable manner. Our code, data and more details are at\nhttps://github.com/hzp3517/ECR-Chain.\n", "date": "2024-05-17", "categories": []}
{"id": "2405.06841", "title": "Bridging the Gap: Protocol Towards Fair and Consistent Affect Analysis", "abstract": "  The increasing integration of machine learning algorithms in daily life\nunderscores the critical need for fairness and equity in their deployment. As\nthese technologies play a pivotal role in decision-making, addressing biases\nacross diverse subpopulation groups, including age, gender, and race, becomes\nparamount. Automatic affect analysis, at the intersection of physiology,\npsychology, and machine learning, has seen significant development. However,\nexisting databases and methodologies lack uniformity, leading to biased\nevaluations. This work addresses these issues by analyzing six affective\ndatabases, annotating demographic attributes, and proposing a common protocol\nfor database partitioning. Emphasis is placed on fairness in evaluations.\nExtensive experiments with baseline and state-of-the-art methods demonstrate\nthe impact of these changes, revealing the inadequacy of prior assessments. The\nfindings underscore the importance of considering demographic attributes in\naffect analysis research and provide a foundation for more equitable\nmethodologies. Our annotations, code and pre-trained models are available at:\nhttps://github.com/dkollias/Fair-Consistent-Affect-Analysis\n", "date": "2024-05-10", "categories": []}
{"id": "2404.04394", "title": "Analyzing Participants' Engagement during Online Meetings Using\n  Unsupervised Remote Photoplethysmography with Behavioral Features", "abstract": "  Engagement measurement finds application in healthcare, education, services.\nThe use of physiological and behavioral features is viable, but the\nimpracticality of traditional physiological measurement arises due to the need\nfor contact sensors. We demonstrate the feasibility of unsupervised remote\nphotoplethysmography (rPPG) as an alternative for contact sensors in deriving\nheart rate variability (HRV) features, then fusing these with behavioral\nfeatures to measure engagement in online group meetings. Firstly, a unique\nEngagement Dataset of online interactions among social workers is collected\nwith granular engagement labels, offering insight into virtual meeting\ndynamics. Secondly, a pre-trained rPPG model is customized to reconstruct rPPG\nsignals from video meetings in an unsupervised manner, enabling the calculation\nof HRV features. Thirdly, the feasibility of estimating engagement from HRV\nfeatures using short observation windows, with a notable enhancement when using\nlonger observation windows of two to four minutes, is demonstrated. Fourthly,\nthe effectiveness of behavioral cues is evaluated when fused with physiological\ndata, which further enhances engagement estimation performance. An accuracy of\n94% is achieved when only HRV features are used, eliminating the need for\ncontact sensors or ground truth signals; use of behavioral cues raises the\naccuracy to 96%. Facial analysis offers precise engagement measurement,\nbeneficial for future applications.\n", "date": "2024-04-05", "categories": []}
{"id": "2405.06922", "title": "EmoMix-3L: A Code-Mixed Dataset for Bangla-English-Hindi Emotion\n  Detection", "abstract": "  Code-mixing is a well-studied linguistic phenomenon that occurs when two or\nmore languages are mixed in text or speech. Several studies have been conducted\non building datasets and performing downstream NLP tasks on code-mixed data.\nAlthough it is not uncommon to observe code-mixing of three or more languages,\nmost available datasets in this domain contain code-mixed data from only two\nlanguages. In this paper, we introduce EmoMix-3L, a novel multi-label emotion\ndetection dataset containing code-mixed data from three different languages. We\nexperiment with several models on EmoMix-3L and we report that MuRIL\noutperforms other models on this dataset.\n", "date": "2024-05-11", "categories": []}
{"id": "2405.01828", "title": "FER-YOLO-Mamba: Facial Expression Detection and Classification Based on\n  Selective State Space", "abstract": "  Facial Expression Recognition (FER) plays a pivotal role in understanding\nhuman emotional cues. However, traditional FER methods based on visual\ninformation have some limitations, such as preprocessing, feature extraction,\nand multi-stage classification procedures. These not only increase\ncomputational complexity but also require a significant amount of computing\nresources. Considering Convolutional Neural Network (CNN)-based FER schemes\nfrequently prove inadequate in identifying the deep, long-distance dependencies\nembedded within facial expression images, and the Transformer's inherent\nquadratic computational complexity, this paper presents the FER-YOLO-Mamba\nmodel, which integrates the principles of Mamba and YOLO technologies to\nfacilitate efficient coordination in facial expression image recognition and\nlocalization. Within the FER-YOLO-Mamba model, we further devise a FER-YOLO-VSS\ndual-branch module, which combines the inherent strengths of convolutional\nlayers in local feature extraction with the exceptional capability of State\nSpace Models (SSMs) in revealing long-distance dependencies. To the best of our\nknowledge, this is the first Vision Mamba model designed for facial expression\ndetection and classification. To evaluate the performance of the proposed\nFER-YOLO-Mamba model, we conducted experiments on two benchmark datasets,\nRAF-DB and SFEW. The experimental results indicate that the FER-YOLO-Mamba\nmodel achieved better results compared to other models. The code is available\nfrom https://github.com/SwjtuMa/FER-YOLO-Mamba.\n", "date": "2024-05-03", "categories": []}
{"id": "2405.06319", "title": "Decoding Emotions in Abstract Art: Cognitive Plausibility of CLIP in\n  Recognizing Color-Emotion Associations", "abstract": "  This study investigates the cognitive plausibility of a pretrained multimodal\nmodel, CLIP, in recognizing emotions evoked by abstract visual art. We employ a\ndataset comprising images with associated emotion labels and textual rationales\nof these labels provided by human annotators. We perform linguistic analyses of\nrationales, zero-shot emotion classification of images and rationales, apply\nsimilarity-based prediction of emotion, and investigate color-emotion\nassociations. The relatively low, yet above baseline, accuracy in recognizing\nemotion for abstract images and rationales suggests that CLIP decodes emotional\ncomplexities in a manner not well aligned with human cognitive processes.\nFurthermore, we explore color-emotion interactions in images and rationales.\nExpected color-emotion associations, such as red relating to anger, are\nidentified in images and texts annotated with emotion labels by both humans and\nCLIP, with the latter showing even stronger interactions. Our results highlight\nthe disparity between human processing and machine processing when connecting\nimage features and emotions.\n", "date": "2024-05-10", "categories": []}
{"id": "2405.06483", "title": "LyS at SemEval-2024 Task 3: An Early Prototype for End-to-End Multimodal\n  Emotion Linking as Graph-Based Parsing", "abstract": "  This paper describes our participation in SemEval 2024 Task 3, which focused\non Multimodal Emotion Cause Analysis in Conversations. We developed an early\nprototype for an end-to-end system that uses graph-based methods from\ndependency parsing to identify causal emotion relations in multi-party\nconversations. Our model comprises a neural transformer-based encoder for\ncontextualizing multimodal conversation data and a graph-based decoder for\ngenerating the adjacency matrix scores of the causal graph. We ranked 7th out\nof 15 valid and official submissions for Subtask 1, using textual inputs only.\nWe also discuss our participation in Subtask 2 during post-evaluation using\nmulti-modal inputs.\n", "date": "2024-05-10", "categories": []}
{"id": "2405.04897", "title": "Machine Learning-based NLP for Emotion Classification on a Cholera X\n  Dataset", "abstract": "  Recent social media posts on the cholera outbreak in Hammanskraal have\nhighlighted the diverse range of emotions people experienced in response to\nsuch an event. The extent of people's opinions varies greatly depending on\ntheir level of knowledge and information about the disease. The documented\nre-search about Cholera lacks investigations into the classification of\nemotions. This study aims to examine the emotions expressed in social media\nposts about Chol-era. A dataset of 23,000 posts was extracted and\npre-processed. The Python Nat-ural Language Toolkit (NLTK) sentiment analyzer\nlibrary was applied to deter-mine the emotional significance of each text.\nAdditionally, Machine Learning (ML) models were applied for emotion\nclassification, including Long short-term memory (LSTM), Logistic regression,\nDecision trees, and the Bidirectional En-coder Representations from\nTransformers (BERT) model. The results of this study demonstrated that LSTM\nachieved the highest accuracy of 75%. Emotion classification presents a\npromising tool for gaining a deeper understanding of the impact of Cholera on\nsociety. The findings of this study might contribute to the development of\neffective interventions in public health strategies.\n", "date": "2024-05-08", "categories": []}
{"id": "2405.04900", "title": "Self-supervised Gait-based Emotion Representation Learning from\n  Selective Strongly Augmented Skeleton Sequences", "abstract": "  Emotion recognition is an important part of affective computing. Extracting\nemotional cues from human gaits yields benefits such as natural interaction, a\nnonintrusive nature, and remote detection. Recently, the introduction of\nself-supervised learning techniques offers a practical solution to the issues\narising from the scarcity of labeled data in the field of gait-based emotion\nrecognition. However, due to the limited diversity of gaits and the\nincompleteness of feature representations for skeletons, the existing\ncontrastive learning methods are usually inefficient for the acquisition of\ngait emotions. In this paper, we propose a contrastive learning framework\nutilizing selective strong augmentation (SSA) for self-supervised gait-based\nemotion representation, which aims to derive effective representations from\nlimited labeled gait data. First, we propose an SSA method for the gait emotion\nrecognition task, which includes upper body jitter and random spatiotemporal\nmask. The goal of SSA is to generate more diverse and targeted positive samples\nand prompt the model to learn more distinctive and robust feature\nrepresentations. Then, we design a complementary feature fusion network (CFFN)\nthat facilitates the integration of cross-domain information to acquire\ntopological structural and global adaptive features. Finally, we implement the\ndistributional divergence minimization loss to supervise the representation\nlearning of the generally and strongly augmented queries. Our approach is\nvalidated on the Emotion-Gait (E-Gait) and Emilya datasets and outperforms the\nstate-of-the-art methods under different evaluation protocols.\n", "date": "2024-05-08", "categories": []}
{"id": "2405.03960", "title": "ESIHGNN: Event-State Interactions Infused Heterogeneous Graph Neural\n  Network for Conversational Emotion Recognition", "abstract": "  Conversational Emotion Recognition (CER) aims to predict the emotion\nexpressed by an utterance (referred to as an ``event'') during a conversation.\nExisting graph-based methods mainly focus on event interactions to comprehend\nthe conversational context, while overlooking the direct influence of the\nspeaker's emotional state on the events. In addition, real-time modeling of the\nconversation is crucial for real-world applications but is rarely considered.\nToward this end, we propose a novel graph-based approach, namely Event-State\nInteractions infused Heterogeneous Graph Neural Network (ESIHGNN), which\nincorporates the speaker's emotional state and constructs a heterogeneous\nevent-state interaction graph to model the conversation. Specifically, a\nheterogeneous directed acyclic graph neural network is employed to dynamically\nupdate and enhance the representations of events and emotional states at each\nturn, thereby improving conversational coherence and consistency. Furthermore,\nto further improve the performance of CER, we enrich the graph's edges with\nexternal knowledge. Experimental results on four publicly available CER\ndatasets show the superiority of our approach and the effectiveness of the\nintroduced heterogeneous event-state interaction graph.\n", "date": "2024-05-07", "categories": []}
{"id": "2405.04128", "title": "Fine-grained Speech Sentiment Analysis in Chinese Psychological Support\n  Hotlines Based on Large-scale Pre-trained Model", "abstract": "  Suicide and suicidal behaviors remain significant challenges for public\npolicy and healthcare. In response, psychological support hotlines have been\nestablished worldwide to provide immediate help to individuals in mental\ncrises. The effectiveness of these hotlines largely depends on accurately\nidentifying callers' emotional states, particularly underlying negative\nemotions indicative of increased suicide risk. However, the high demand for\npsychological interventions often results in a shortage of professional\noperators, highlighting the need for an effective speech emotion recognition\nmodel. This model would automatically detect and analyze callers' emotions,\nfacilitating integration into hotline services. Additionally, it would enable\nlarge-scale data analysis of psychological support hotline interactions to\nexplore psychological phenomena and behaviors across populations. Our study\nutilizes data from the Beijing psychological support hotline, the largest\nsuicide hotline in China. We analyzed speech data from 105 callers containing\n20,630 segments and categorized them into 11 types of negative emotions. We\ndeveloped a negative emotion recognition model and a fine-grained multi-label\nclassification model using a large-scale pre-trained model. Our experiments\nindicate that the negative emotion recognition model achieves a maximum\nF1-score of 76.96%. However, it shows limited efficacy in the fine-grained\nmulti-label classification task, with the best model achieving only a 41.74%\nweighted F1-score. We conducted an error analysis for this task, discussed\npotential future improvements, and considered the clinical application\npossibilities of our study. All the codes are public available.\n", "date": "2024-05-07", "categories": []}
{"id": "2405.03202", "title": "Hierarchical Space-Time Attention for Micro-Expression Recognition", "abstract": "  Micro-expression recognition (MER) aims to recognize the short and subtle\nfacial movements from the Micro-expression (ME) video clips, which reveal real\nemotions. Recent MER methods mostly only utilize special frames from ME video\nclips or extract optical flow from these special frames. However, they neglect\nthe relationship between movements and space-time, while facial cues are hidden\nwithin these relationships. To solve this issue, we propose the Hierarchical\nSpace-Time Attention (HSTA). Specifically, we first process ME video frames and\nspecial frames or data parallelly by our cascaded Unimodal Space-Time Attention\n(USTA) to establish connections between subtle facial movements and specific\nfacial areas. Then, we design Crossmodal Space-Time Attention (CSTA) to achieve\na higher-quality fusion for crossmodal data. Finally, we hierarchically\nintegrate USTA and CSTA to grasp the deeper facial cues. Our model emphasizes\ntemporal modeling without neglecting the processing of special data, and it\nfuses the contents in different modalities while maintaining their respective\nuniqueness. Extensive experiments on the four benchmarks show the effectiveness\nof our proposed HSTA. Specifically, compared with the latest method on the\nCASME3 dataset, it achieves about 3% score improvement in seven-category\nclassification.\n", "date": "2024-05-06", "categories": []}
{"id": "2404.17858", "title": "Revisiting Multi-modal Emotion Learning with Broad State Space Models\n  and Probability-guidance Fusion", "abstract": "  Multi-modal Emotion Recognition in Conversation (MERC) has received\nconsiderable attention in various fields, e.g., human-computer interaction and\nrecommendation systems. Most existing works perform feature disentanglement and\nfusion to extract emotional contextual information from multi-modal features\nand emotion classification. After revisiting the characteristic of MERC, we\nargue that long-range contextual semantic information should be extracted in\nthe feature disentanglement stage and the inter-modal semantic information\nconsistency should be maximized in the feature fusion stage. Inspired by recent\nState Space Models (SSMs), Mamba can efficiently model long-distance\ndependencies. Therefore, in this work, we fully consider the above insights to\nfurther improve the performance of MERC. Specifically, on the one hand, in the\nfeature disentanglement stage, we propose a Broad Mamba, which does not rely on\na self-attention mechanism for sequence modeling, but uses state space models\nto compress emotional representation, and utilizes broad learning systems to\nexplore the potential data distribution in broad space. Different from previous\nSSMs, we design a bidirectional SSM convolution to extract global context\ninformation. On the other hand, we design a multi-modal fusion strategy based\non probability guidance to maximize the consistency of information between\nmodalities. Experimental results show that the proposed method can overcome the\ncomputational and memory limitations of Transformer when modeling long-distance\ncontexts, and has great potential to become a next-generation general\narchitecture in MERC.\n", "date": "2024-04-27", "categories": []}
{"id": "2404.17862", "title": "Revisiting Multimodal Emotion Recognition in Conversation from the\n  Perspective of Graph Spectrum", "abstract": "  Efficiently capturing consistent and complementary semantic features in a\nmultimodal conversation context is crucial for Multimodal Emotion Recognition\nin Conversation (MERC). Existing methods mainly use graph structures to model\ndialogue context semantic dependencies and employ Graph Neural Networks (GNN)\nto capture multimodal semantic features for emotion recognition. However, these\nmethods are limited by some inherent characteristics of GNN, such as\nover-smoothing and low-pass filtering, resulting in the inability to learn\nlong-distance consistency information and complementary information\nefficiently. Since consistency and complementarity information correspond to\nlow-frequency and high-frequency information, respectively, this paper revisits\nthe problem of multimodal emotion recognition in conversation from the\nperspective of the graph spectrum. Specifically, we propose a\nGraph-Spectrum-based Multimodal Consistency and Complementary collaborative\nlearning framework GS-MCC. First, GS-MCC uses a sliding window to construct a\nmultimodal interaction graph to model conversational relationships and uses\nefficient Fourier graph operators to extract long-distance high-frequency and\nlow-frequency information, respectively. Then, GS-MCC uses contrastive learning\nto construct self-supervised signals that reflect complementarity and\nconsistent semantic collaboration with high and low-frequency signals, thereby\nimproving the ability of high and low-frequency information to reflect real\nemotions. Finally, GS-MCC inputs the collaborative high and low-frequency\ninformation into the MLP network and softmax function for emotion prediction.\nExtensive experiments have proven the superiority of the GS-MCC architecture\nproposed in this paper on two benchmark data sets.\n", "date": "2024-04-27", "categories": []}
{"id": "2405.00307", "title": "Active Learning with Task Adaptation Pre-training for Speech Emotion\n  Recognition", "abstract": "  Speech emotion recognition (SER) has garnered increasing attention due to its\nwide range of applications in various fields, including human-machine\ninteraction, virtual assistants, and mental health assistance. However,\nexisting SER methods often overlook the information gap between the\npre-training speech recognition task and the downstream SER task, resulting in\nsub-optimal performance. Moreover, current methods require much time for\nfine-tuning on each specific speech dataset, such as IEMOCAP, which limits\ntheir effectiveness in real-world scenarios with large-scale noisy data. To\naddress these issues, we propose an active learning (AL)-based fine-tuning\nframework for SER, called \\textsc{After}, that leverages task adaptation\npre-training (TAPT) and AL methods to enhance performance and efficiency.\nSpecifically, we first use TAPT to minimize the information gap between the\npre-training speech recognition task and the downstream speech emotion\nrecognition task. Then, AL methods are employed to iteratively select a subset\nof the most informative and diverse samples for fine-tuning, thereby reducing\ntime consumption. Experiments demonstrate that our proposed method\n\\textsc{After}, using only 20\\% of samples, improves accuracy by 8.45\\% and\nreduces time consumption by 79\\%. The additional extension of \\textsc{After}\nand ablation studies further confirm its effectiveness and applicability to\nvarious real-world scenarios. Our source code is available on Github for\nreproducibility. (https://github.com/Clearloveyuan/AFTER).\n", "date": "2024-05-01", "categories": []}
{"id": "2404.14606", "title": "Cross-Task Multi-Branch Vision Transformer for Facial Expression and\n  Mask Wearing Classification", "abstract": "  With wearing masks becoming a new cultural norm, facial expression\nrecognition (FER) while taking masks into account has become a significant\nchallenge. In this paper, we propose a unified multi-branch vision transformer\nfor facial expression recognition and mask wearing classification tasks. Our\napproach extracts shared features for both tasks using a dual-branch\narchitecture that obtains multi-scale feature representations. Furthermore, we\npropose a cross-task fusion phase that processes tokens for each task with\nseparate branches, while exchanging information using a cross attention module.\nOur proposed framework reduces the overall complexity compared with using\nseparate networks for both tasks by the simple yet effective cross-task fusion\nphase. Extensive experiments demonstrate that our proposed model performs\nbetter than or on par with different state-of-the-art methods on both facial\nexpression recognition and facial mask wearing classification task.\n", "date": "2024-04-22", "categories": []}
{"id": "2312.05632", "title": "Subject-Based Domain Adaptation for Facial Expression Recognition", "abstract": "  Adapting a deep learning model to a specific target individual is a\nchallenging facial expression recognition (FER) task that may be achieved using\nunsupervised domain adaptation (UDA) methods. Although several UDA methods have\nbeen proposed to adapt deep FER models across source and target data sets,\nmultiple subject-specific source domains are needed to accurately represent the\nintra- and inter-person variability in subject-based adaption. This paper\nconsiders the setting where domains correspond to individuals, not entire\ndatasets. Unlike UDA, multi-source domain adaptation (MSDA) methods can\nleverage multiple source datasets to improve the accuracy and robustness of the\ntarget model. However, previous methods for MSDA adapt image classification\nmodels across datasets and do not scale well to a more significant number of\nsource domains. This paper introduces a new MSDA method for subject-based\ndomain adaptation in FER. It efficiently leverages information from multiple\nsource subjects (labeled source domain data) to adapt a deep FER model to a\nsingle target individual (unlabeled target domain data). During adaptation, our\nsubject-based MSDA first computes a between-source discrepancy loss to mitigate\nthe domain shift among data from several source subjects. Then, a new strategy\nis employed to generate augmented confident pseudo-labels for the target\nsubject, allowing a reduction in the domain shift between source and target\nsubjects. Experiments performed on the challenging BioVid heat and pain dataset\nwith 87 subjects and the UNBC-McMaster shoulder pain dataset with 25 subjects\nshow that our subject-based MSDA can outperform state-of-the-art methods yet\nscale well to multiple subject-based source domains.\n", "date": "2023-12-09", "categories": []}
{"id": "2401.15489", "title": "Distilling Privileged Multimodal Information for Expression Recognition\n  using Optimal Transport", "abstract": "  Deep learning models for multimodal expression recognition have reached\nremarkable performance in controlled laboratory environments because of their\nability to learn complementary and redundant semantic information. However,\nthese models struggle in the wild, mainly because of the unavailability and\nquality of modalities used for training. In practice, only a subset of the\ntraining-time modalities may be available at test time. Learning with\nprivileged information enables models to exploit data from additional\nmodalities that are only available during training. State-of-the-art knowledge\ndistillation (KD) methods have been proposed to distill information from\nmultiple teacher models (each trained on a modality) to a common student model.\nThese privileged KD methods typically utilize point-to-point matching, yet have\nno explicit mechanism to capture the structural information in the teacher\nrepresentation space formed by introducing the privileged modality. Experiments\nwere performed on two challenging problems - pain estimation on the Biovid\ndataset (ordinal classification) and arousal-valance prediction on the Affwild2\ndataset (regression). Results show that our proposed method can outperform\nstate-of-the-art privileged KD methods on these problems. The diversity among\nmodalities and fusion architectures indicates that PKDOT is modality- and\nmodel-agnostic.\n", "date": "2024-01-27", "categories": []}
{"id": "2404.11447", "title": "Research on emotionally intelligent dialogue generation based on\n  automatic dialogue system", "abstract": "  Automated dialogue systems are important applications of artificial\nintelligence, and traditional systems struggle to understand user emotions and\nprovide empathetic feedback. This study integrates emotional intelligence\ntechnology into automated dialogue systems and creates a dialogue generation\nmodel with emotional intelligence through deep learning and natural language\nprocessing techniques. The model can detect and understand a wide range of\nemotions and specific pain signals in real time, enabling the system to provide\nempathetic interaction. By integrating the results of the study \"Can artificial\nintelligence detect pain and express pain empathy?\", the model's ability to\nunderstand the subtle elements of pain empathy has been enhanced, setting\nhigher standards for emotional intelligence dialogue systems. The project aims\nto provide theoretical understanding and practical suggestions to integrate\nadvanced emotional intelligence capabilities into dialogue systems, thereby\nimproving user experience and interaction quality.\n", "date": "2024-04-17", "categories": []}
{"id": "2404.17205", "title": "Two in One Go: Single-stage Emotion Recognition with Decoupled\n  Subject-context Transformer", "abstract": "  Emotion recognition aims to discern the emotional state of subjects within an\nimage, relying on subject-centric and contextual visual cues. Current\napproaches typically follow a two-stage pipeline: first localize subjects by\noff-the-shelf detectors, then perform emotion classification through the late\nfusion of subject and context features. However, the complicated paradigm\nsuffers from disjoint training stages and limited interaction between\nfine-grained subject-context elements. To address the challenge, we present a\nsingle-stage emotion recognition approach, employing a Decoupled\nSubject-Context Transformer (DSCT), for simultaneous subject localization and\nemotion classification. Rather than compartmentalizing training stages, we\njointly leverage box and emotion signals as supervision to enrich\nsubject-centric feature learning. Furthermore, we introduce DSCT to facilitate\ninteractions between fine-grained subject-context cues in a decouple-then-fuse\nmanner. The decoupled query token--subject queries and context\nqueries--gradually intertwine across layers within DSCT, during which spatial\nand semantic relations are exploited and aggregated. We evaluate our\nsingle-stage framework on two widely used context-aware emotion recognition\ndatasets, CAER-S and EMOTIC. Our approach surpasses two-stage alternatives with\nfewer parameter numbers, achieving a 3.39% accuracy improvement and a 6.46%\naverage precision gain on CAER-S and EMOTIC datasets, respectively.\n", "date": "2024-04-26", "categories": []}
{"id": "2404.16670", "title": "EmoVIT: Revolutionizing Emotion Insights with Visual Instruction Tuning", "abstract": "  Visual Instruction Tuning represents a novel learning paradigm involving the\nfine-tuning of pre-trained language models using task-specific instructions.\nThis paradigm shows promising zero-shot results in various natural language\nprocessing tasks but is still unexplored in vision emotion understanding. In\nthis work, we focus on enhancing the model's proficiency in understanding and\nadhering to instructions related to emotional contexts. Initially, we identify\nkey visual clues critical to visual emotion recognition. Subsequently, we\nintroduce a novel GPT-assisted pipeline for generating emotion visual\ninstruction data, effectively addressing the scarcity of annotated instruction\ndata in this domain. Expanding on the groundwork established by InstructBLIP,\nour proposed EmoVIT architecture incorporates emotion-specific instruction\ndata, leveraging the powerful capabilities of Large Language Models to enhance\nperformance. Through extensive experiments, our model showcases its proficiency\nin emotion classification, adeptness in affective reasoning, and competence in\ncomprehending humor. The comparative analysis provides a robust benchmark for\nEmotion Visual Instruction Tuning in the era of LLMs, providing valuable\ninsights and opening avenues for future exploration in this domain. Our code is\navailable at \\url{https://github.com/aimmemotion/EmoVIT}.\n", "date": "2024-04-25", "categories": []}
{"id": "2404.15351", "title": "Integrating Physiological Data with Large Language Models for Empathic\n  Human-AI Interaction", "abstract": "  This paper explores enhancing empathy in Large Language Models (LLMs) by\nintegrating them with physiological data. We propose a physiological computing\napproach that includes developing deep learning models that use physiological\ndata for recognizing psychological states and integrating the predicted states\nwith LLMs for empathic interaction. We showcase the application of this\napproach in an Empathic LLM (EmLLM) chatbot for stress monitoring and control.\nWe also discuss the results of a pilot study that evaluates this EmLLM chatbot\nbased on its ability to accurately predict user stress, provide human-like\nresponses, and assess the therapeutic alliance with the user.\n", "date": "2024-04-14", "categories": []}
{"id": "2404.15373", "title": "Robust EEG-based Emotion Recognition Using an Inception and Two-sided\n  Perturbation Model", "abstract": "  Automated emotion recognition using electroencephalogram (EEG) signals has\ngained substantial attention. Although deep learning approaches exhibit strong\nperformance, they often suffer from vulnerabilities to various perturbations,\nlike environmental noise and adversarial attacks. In this paper, we propose an\nInception feature generator and two-sided perturbation (INC-TSP) approach to\nenhance emotion recognition in brain-computer interfaces. INC-TSP integrates\nthe Inception module for EEG data analysis and employs two-sided perturbation\n(TSP) as a defensive mechanism against input perturbations. TSP introduces\nworst-case perturbations to the model's weights and inputs, reinforcing the\nmodel's elasticity against adversarial attacks. The proposed approach addresses\nthe challenge of maintaining accurate emotion recognition in the presence of\ninput uncertainties. We validate INC-TSP in a subject-independent three-class\nemotion recognition scenario, demonstrating robust performance.\n", "date": "2024-04-21", "categories": []}
{"id": "2309.15877", "title": "Neuro-Inspired Hierarchical Multimodal Learning", "abstract": "  Integrating and processing information from various sources or modalities are\ncritical for obtaining a comprehensive and accurate perception of the real\nworld. Drawing inspiration from neuroscience, we develop the\nInformation-Theoretic Hierarchical Perception (ITHP) model, which utilizes the\nconcept of information bottleneck. Distinct from most traditional fusion models\nthat aim to incorporate all modalities as input, our model designates the prime\nmodality as input, while the remaining modalities act as detectors in the\ninformation pathway. Our proposed perception model focuses on constructing an\neffective and compact information flow by achieving a balance between the\nminimization of mutual information between the latent state and the input modal\nstate, and the maximization of mutual information between the latent states and\nthe remaining modal states. This approach leads to compact latent state\nrepresentations that retain relevant information while minimizing redundancy,\nthereby substantially enhancing the performance of downstream tasks.\nExperimental evaluations on both the MUStARD and CMU-MOSI datasets demonstrate\nthat our model consistently distills crucial information in multimodal learning\nscenarios, outperforming state-of-the-art benchmarks.\n", "date": "2023-09-27", "categories": []}
{"id": "2404.14975", "title": "CAGE: Circumplex Affect Guided Expression Inference", "abstract": "  Understanding emotions and expressions is a task of interest across multiple\ndisciplines, especially for improving user experiences. Contrary to the common\nperception, it has been shown that emotions are not discrete entities but\ninstead exist along a continuum. People understand discrete emotions\ndifferently due to a variety of factors, including cultural background,\nindividual experiences, and cognitive biases. Therefore, most approaches to\nexpression understanding, particularly those relying on discrete categories,\nare inherently biased. In this paper, we present a comparative in-depth\nanalysis of two common datasets (AffectNet and EMOTIC) equipped with the\ncomponents of the circumplex model of affect. Further, we propose a model for\nthe prediction of facial expressions tailored for lightweight applications.\nUsing a small-scaled MaxViT-based model architecture, we evaluate the impact of\ndiscrete expression category labels in training with the continuous valence and\narousal labels. We show that considering valence and arousal in addition to\ndiscrete category labels helps to significantly improve expression inference.\nThe proposed model outperforms the current state-of-the-art models on\nAffectNet, establishing it as the best-performing model for inferring valence\nand arousal achieving a 7% lower RMSE. Training scripts and trained weights to\nreproduce our results can be found here:\nhttps://github.com/wagner-niklas/CAGE_expression_inference.\n", "date": "2024-04-23", "categories": []}
{"id": "2403.10488", "title": "Joint Multimodal Transformer for Emotion Recognition in the Wild", "abstract": "  Multimodal emotion recognition (MMER) systems typically outperform unimodal\nsystems by leveraging the inter- and intra-modal relationships between, e.g.,\nvisual, textual, physiological, and auditory modalities. This paper proposes an\nMMER method that relies on a joint multimodal transformer (JMT) for fusion with\nkey-based cross-attention. This framework can exploit the complementary nature\nof diverse modalities to improve predictive accuracy. Separate backbones\ncapture intra-modal spatiotemporal dependencies within each modality over video\nsequences. Subsequently, our JMT fusion architecture integrates the individual\nmodality embeddings, allowing the model to effectively capture inter- and\nintra-modal relationships. Extensive experiments on two challenging expression\nrecognition tasks -- (1) dimensional emotion recognition on the Affwild2\ndataset (with face and voice) and (2) pain estimation on the Biovid dataset\n(with face and biosensors) -- indicate that our JMT fusion can provide a\ncost-effective solution for MMER. Empirical results show that MMER systems with\nour proposed fusion allow us to outperform relevant baseline and\nstate-of-the-art methods.\n", "date": "2024-03-15", "categories": []}
{"id": "2404.13509", "title": "MFHCA: Enhancing Speech Emotion Recognition Via Multi-Spatial Fusion and\n  Hierarchical Cooperative Attention", "abstract": "  Speech emotion recognition is crucial in human-computer interaction, but\nextracting and using emotional cues from audio poses challenges. This paper\nintroduces MFHCA, a novel method for Speech Emotion Recognition using\nMulti-Spatial Fusion and Hierarchical Cooperative Attention on spectrograms and\nraw audio. We employ the Multi-Spatial Fusion module (MF) to efficiently\nidentify emotion-related spectrogram regions and integrate Hubert features for\nhigher-level acoustic information. Our approach also includes a Hierarchical\nCooperative Attention module (HCA) to merge features from various auditory\nlevels. We evaluate our method on the IEMOCAP dataset and achieve 2.6\\% and\n1.87\\% improvements on the weighted accuracy and unweighted accuracy,\nrespectively. Extensive experiments demonstrate the effectiveness of the\nproposed method.\n", "date": "2024-04-21", "categories": []}
{"id": "2404.13764", "title": "Using Adaptive Empathetic Responses for Teaching English", "abstract": "  Existing English-teaching chatbots rarely incorporate empathy explicitly in\ntheir feedback, but empathetic feedback could help keep students engaged and\nreduce learner anxiety. Toward this end, we propose the task of negative\nemotion detection via audio, for recognizing empathetic feedback opportunities\nin language learning. We then build the first spoken English-teaching chatbot\nwith adaptive, empathetic feedback. This feedback is synthesized through\nautomatic prompt optimization of ChatGPT and is evaluated with English\nlearners. We demonstrate the effectiveness of our system through a preliminary\nuser study.\n", "date": "2024-04-21", "categories": []}
{"id": "2404.13778", "title": "Multi-channel Emotion Analysis for Consensus Reaching in Group Movie\n  Recommendation Systems", "abstract": "  Watching movies is one of the social activities typically done in groups.\nEmotion is the most vital factor that affects movie viewers' preferences. So,\nthe emotional aspect of the movie needs to be determined and analyzed for\nfurther recommendations. It can be challenging to choose a movie that appeals\nto the emotions of a diverse group. Reaching an agreement for a group can be\ndifficult due to the various genres and choices. This paper proposes a novel\napproach to group movie suggestions by examining emotions from three different\nchannels: movie descriptions (text), soundtracks (audio), and posters (image).\nWe employ the Jaccard similarity index to match each participant's emotional\npreferences to prospective movie choices, followed by a fuzzy inference\ntechnique to determine group consensus. We use a weighted integration process\nfor the fusion of emotion scores from diverse data types. Then, group movie\nrecommendation is based on prevailing emotions and viewers' best-loved movies.\nAfter determining the recommendations, the group's consensus level is\ncalculated using a fuzzy inference system, taking participants' feedback as\ninput. Participants (n=130) in the survey were provided with different emotion\ncategories and asked to select the emotions best suited for particular movies\n(n=12). Comparison results between predicted and actual scores demonstrate the\nefficiency of using emotion detection for this problem (Jaccard similarity\nindex = 0.76). We explored the relationship between induced emotions and movie\npopularity as an additional experiment, analyzing emotion distribution in 100\npopular movies from the TMDB database. Such systems can potentially improve the\naccuracy of movie recommendation systems and achieve a high level of consensus\namong participants with diverse preferences.\n", "date": "2024-04-21", "categories": []}
{"id": "2404.12642", "title": "Cooperative Sentiment Agents for Multimodal Sentiment Analysis", "abstract": "  In this paper, we propose a new Multimodal Representation Learning (MRL)\nmethod for Multimodal Sentiment Analysis (MSA), which facilitates the adaptive\ninteraction between modalities through Cooperative Sentiment Agents, named\nCo-SA. Co-SA comprises two critical components: the Sentiment Agents\nEstablishment (SAE) phase and the Sentiment Agents Cooperation (SAC) phase.\nDuring the SAE phase, each sentiment agent deals with an unimodal signal and\nhighlights explicit dynamic sentiment variations within the modality via the\nModality-Sentiment Disentanglement (MSD) and Deep Phase Space Reconstruction\n(DPSR) modules. Subsequently, in the SAC phase, Co-SA meticulously designs\ntask-specific interaction mechanisms for sentiment agents so that coordinating\nmultimodal signals to learn the joint representation. Specifically, Co-SA\nequips an independent policy model for each sentiment agent that captures\nsignificant properties within the modality. These policies are optimized\nmutually through the unified reward adaptive to downstream tasks. Benefitting\nfrom the rewarding mechanism, Co-SA transcends the limitation of pre-defined\nfusion modes and adaptively captures unimodal properties for MRL in the\nmultimodal interaction setting. To demonstrate the effectiveness of Co-SA, we\napply it to address Multimodal Sentiment Analysis (MSA) and Multimodal Emotion\nRecognition (MER) tasks. Our comprehensive experimental results demonstrate\nthat Co-SA excels at discovering diverse cross-modal features, encompassing\nboth common and complementary aspects. The code can be available at\nhttps://github.com/smwanghhh/Co-SA.\n", "date": "2024-04-19", "categories": []}
{"id": "2307.10757", "title": "Vesper: A Compact and Effective Pretrained Model for Speech Emotion\n  Recognition", "abstract": "  This paper presents a paradigm that adapts general large-scale pretrained\nmodels (PTMs) to speech emotion recognition task. Although PTMs shed new light\non artificial general intelligence, they are constructed with general tasks in\nmind, and thus, their efficacy for specific tasks can be further improved.\nAdditionally, employing PTMs in practical applications can be challenging due\nto their considerable size. Above limitations spawn another research direction,\nnamely, optimizing large-scale PTMs for specific tasks to generate\ntask-specific PTMs that are both compact and effective. In this paper, we focus\non the speech emotion recognition task and propose an improved emotion-specific\npretrained encoder called Vesper. Vesper is pretrained on a speech dataset\nbased on WavLM and takes into account emotional characteristics. To enhance\nsensitivity to emotional information, Vesper employs an emotion-guided masking\nstrategy to identify the regions that need masking. Subsequently, Vesper\nemploys hierarchical and cross-layer self-supervision to improve its ability to\ncapture acoustic and semantic representations, both of which are crucial for\nemotion recognition. Experimental results on the IEMOCAP, MELD, and CREMA-D\ndatasets demonstrate that Vesper with 4 layers outperforms WavLM Base with 12\nlayers, and the performance of Vesper with 12 layers surpasses that of WavLM\nLarge with 24 layers.\n", "date": "2023-07-20", "categories": []}
{"id": "2404.12251", "title": "Dynamic Modality and View Selection for Multimodal Emotion Recognition\n  with Missing Modalities", "abstract": "  The study of human emotions, traditionally a cornerstone in fields like\npsychology and neuroscience, has been profoundly impacted by the advent of\nartificial intelligence (AI). Multiple channels, such as speech (voice) and\nfacial expressions (image), are crucial in understanding human emotions.\nHowever, AI's journey in multimodal emotion recognition (MER) is marked by\nsubstantial technical challenges. One significant hurdle is how AI models\nmanage the absence of a particular modality - a frequent occurrence in\nreal-world situations. This study's central focus is assessing the performance\nand resilience of two strategies when confronted with the lack of one modality:\na novel multimodal dynamic modality and view selection and a cross-attention\nmechanism. Results on the RECOLA dataset show that dynamic selection-based\nmethods are a promising approach for MER. In the missing modalities scenarios,\nall dynamic selection-based methods outperformed the baseline. The study\nconcludes by emphasizing the intricate interplay between audio and video\nmodalities in emotion prediction, showcasing the adaptability of dynamic\nselection methods in handling missing modalities.\n", "date": "2024-04-18", "categories": []}
{"id": "2404.11141", "title": "Context-Aware Siamese Networks for Efficient Emotion Recognition in\n  Conversation", "abstract": "  The advent of deep learning models has made a considerable contribution to\nthe achievement of Emotion Recognition in Conversation (ERC). However, this\ntask still remains an important challenge due to the plurality and subjectivity\nof human emotions. Previous work on ERC provides predictive models using mostly\ngraph-based conversation representations. In this work, we propose a way to\nmodel the conversational context that we incorporate into a metric learning\ntraining strategy, with a two-step process. This allows us to perform ERC in a\nflexible classification scenario and to end up with a lightweight yet efficient\nmodel. Using metric learning through a Siamese Network architecture, we achieve\n57.71 in macro F1 score for emotion classification in conversation on\nDailyDialog dataset, which outperforms the related work. This state-of-the-art\nresult is promising regarding the use of metric learning for emotion\nrecognition, yet perfectible compared to the microF1 score obtained.\n", "date": "2024-04-17", "categories": []}
{"id": "2307.15432", "title": "CFN-ESA: A Cross-Modal Fusion Network with Emotion-Shift Awareness for\n  Dialogue Emotion Recognition", "abstract": "  Multimodal emotion recognition in conversation (ERC) has garnered growing\nattention from research communities in various fields. In this paper, we\npropose a Cross-modal Fusion Network with Emotion-Shift Awareness (CFN-ESA) for\nERC. Extant approaches employ each modality equally without distinguishing the\namount of emotional information in these modalities, rendering it hard to\nadequately extract complementary information from multimodal data. To cope with\nthis problem, in CFN-ESA, we treat textual modality as the primary source of\nemotional information, while visual and acoustic modalities are taken as the\nsecondary sources. Besides, most multimodal ERC models ignore emotion-shift\ninformation and overfocus on contextual information, leading to the failure of\nemotion recognition under emotion-shift scenario. We elaborate an emotion-shift\nmodule to address this challenge. CFN-ESA mainly consists of unimodal encoder\n(RUME), cross-modal encoder (ACME), and emotion-shift module (LESM). RUME is\napplied to extract conversation-level contextual emotional cues while pulling\ntogether data distributions between modalities; ACME is utilized to perform\nmultimodal interaction centered on textual modality; LESM is used to model\nemotion shift and capture emotion-shift information, thereby guiding the\nlearning of the main task. Experimental results demonstrate that CFN-ESA can\neffectively promote performance for ERC and remarkably outperform\nstate-of-the-art models.\n", "date": "2023-07-28", "categories": []}
{"id": "2402.09055", "title": "Comment-aided Video-Language Alignment via Contrastive Pre-training for\n  Short-form Video Humor Detection", "abstract": "  The growing importance of multi-modal humor detection within affective\ncomputing correlates with the expanding influence of short-form video sharing\non social media platforms. In this paper, we propose a novel two-branch\nhierarchical model for short-form video humor detection (SVHD), named\nComment-aided Video-Language Alignment (CVLA) via data-augmented multi-modal\ncontrastive pre-training. Notably, our CVLA not only operates on raw signals\nacross various modal channels but also yields an appropriate multi-modal\nrepresentation by aligning the video and language components within a\nconsistent semantic space. The experimental results on two humor detection\ndatasets, including DY11k and UR-FUNNY, demonstrate that CVLA dramatically\noutperforms state-of-the-art and several competitive baseline approaches. Our\ndataset, code and model release at https://github.com/yliu-cs/CVLA.\n", "date": "2024-02-14", "categories": []}
{"id": "2403.13659", "title": "Recursive Joint Cross-Modal Attention for Multimodal Fusion in\n  Dimensional Emotion Recognition", "abstract": "  Though multimodal emotion recognition has achieved significant progress over\nrecent years, the potential of rich synergic relationships across the\nmodalities is not fully exploited. In this paper, we introduce Recursive Joint\nCross-Modal Attention (RJCMA) to effectively capture both intra- and\ninter-modal relationships across audio, visual, and text modalities for\ndimensional emotion recognition. In particular, we compute the attention\nweights based on cross-correlation between the joint audio-visual-text feature\nrepresentations and the feature representations of individual modalities to\nsimultaneously capture intra- and intermodal relationships across the\nmodalities. The attended features of the individual modalities are again fed as\ninput to the fusion model in a recursive mechanism to obtain more refined\nfeature representations. We have also explored Temporal Convolutional Networks\n(TCNs) to improve the temporal modeling of the feature representations of\nindividual modalities. Extensive experiments are conducted to evaluate the\nperformance of the proposed fusion model on the challenging Affwild2 dataset.\nBy effectively capturing the synergic intra- and inter-modal relationships\nacross audio, visual, and text modalities, the proposed fusion model achieves a\nConcordance Correlation Coefficient (CCC) of 0.585 (0.542) and 0.674 (0.619)\nfor valence and arousal respectively on the validation set(test set). This\nshows a significant improvement over the baseline of 0.240 (0.211) and 0.200\n(0.191) for valence and arousal, respectively, in the validation set (test\nset), achieving second place in the valence-arousal challenge of the 6th\nAffective Behavior Analysis in-the-Wild (ABAW) competition.\n", "date": "2024-03-20", "categories": []}
{"id": "2404.09042", "title": "Improving Personalisation in Valence and Arousal Prediction using Data\n  Augmentation", "abstract": "  In the field of emotion recognition and Human-Machine Interaction (HMI),\npersonalised approaches have exhibited their efficacy in capturing\nindividual-specific characteristics and enhancing affective prediction\naccuracy. However, personalisation techniques often face the challenge of\nlimited data for target individuals. This paper presents our work on an\nenhanced personalisation strategy, that leverages data augmentation to develop\ntailored models for continuous valence and arousal prediction. Our proposed\napproach, Distance Weighting Augmentation (DWA), employs a weighting-based\naugmentation method that expands a target individual's dataset, leveraging\ndistance metrics to identify similar samples at the segment-level. Experimental\nresults on the MuSe-Personalisation 2023 Challenge dataset demonstrate that our\nmethod significantly improves the performance of features sets which have low\nbaseline performance, on the test set. This improvement in poor-performing\nfeatures comes without sacrificing performance on high-performing features. In\nparticular, our method achieves a maximum combined testing CCC of 0.78,\ncompared to the reported baseline score of 0.76 (reproduced at 0.72). It also\nachieved a peak arousal and valence scores of 0.81 and 0.76, compared to\nreproduced baseline scores of 0.76 and 0.67 respectively. Through this work, we\nmake significant contributions to the advancement of personalised affective\ncomputing models, enhancing the practicality and adaptability of data-level\npersonalisation in real world contexts.\n", "date": "2024-04-13", "categories": []}
{"id": "2404.00511", "title": "MIPS at SemEval-2024 Task 3: Multimodal Emotion-Cause Pair Extraction in\n  Conversations with Multimodal Language Models", "abstract": "  This paper presents our winning submission to Subtask 2 of SemEval 2024 Task\n3 on multimodal emotion cause analysis in conversations. We propose a novel\nMultimodal Emotion Recognition and Multimodal Emotion Cause Extraction\n(MER-MCE) framework that integrates text, audio, and visual modalities using\nspecialized emotion encoders. Our approach sets itself apart from\ntop-performing teams by leveraging modality-specific features for enhanced\nemotion understanding and causality inference. Experimental evaluation\ndemonstrates the advantages of our multimodal approach, with our submission\nachieving a competitive weighted F1 score of 0.3435, ranking third with a\nmargin of only 0.0339 behind the 1st team and 0.0025 behind the 2nd team.\nProject: https://github.com/MIPS-COLT/MER-MCE.git\n", "date": "2024-03-31", "categories": []}
{"id": "2404.07484", "title": "Multimodal Emotion Recognition by Fusing Video Semantic in MOOC Learning\n  Scenarios", "abstract": "  In the Massive Open Online Courses (MOOC) learning scenario, the semantic\ninformation of instructional videos has a crucial impact on learners' emotional\nstate. Learners mainly acquire knowledge by watching instructional videos, and\nthe semantic information in the videos directly affects learners' emotional\nstates. However, few studies have paid attention to the potential influence of\nthe semantic information of instructional videos on learners' emotional states.\nTo deeply explore the impact of video semantic information on learners'\nemotions, this paper innovatively proposes a multimodal emotion recognition\nmethod by fusing video semantic information and physiological signals. We\ngenerate video descriptions through a pre-trained large language model (LLM) to\nobtain high-level semantic information about instructional videos. Using the\ncross-attention mechanism for modal interaction, the semantic information is\nfused with the eye movement and PhotoPlethysmoGraphy (PPG) signals to obtain\nthe features containing the critical information of the three modes. The\naccurate recognition of learners' emotional states is realized through the\nemotion classifier. The experimental results show that our method has\nsignificantly improved emotion recognition performance, providing a new\nperspective and efficient method for emotion recognition research in MOOC\nlearning scenarios. The method proposed in this paper not only contributes to a\ndeeper understanding of the impact of instructional videos on learners'\nemotional states but also provides a beneficial reference for future research\non emotion recognition in MOOC learning scenarios.\n", "date": "2024-04-11", "categories": []}
{"id": "2403.05916", "title": "GPT as Psychologist? Preliminary Evaluations for GPT-4V on Visual\n  Affective Computing", "abstract": "  Multimodal large language models (MLLMs) are designed to process and\nintegrate information from multiple sources, such as text, speech, images, and\nvideos. Despite its success in language understanding, it is critical to\nevaluate the performance of downstream tasks for better human-centric\napplications. This paper assesses the application of MLLMs with 5 crucial\nabilities for affective computing, spanning from visual affective tasks and\nreasoning tasks. The results show that \\gpt has high accuracy in facial action\nunit recognition and micro-expression detection while its general facial\nexpression recognition performance is not accurate. We also highlight the\nchallenges of achieving fine-grained micro-expression recognition and the\npotential for further study and demonstrate the versatility and potential of\n\\gpt for handling advanced tasks in emotion recognition and related fields by\nintegrating with task-related agents for more complex tasks, such as heart rate\nestimation through signal processing. In conclusion, this paper provides\nvaluable insights into the potential applications and challenges of MLLMs in\nhuman-centric computing. Our interesting examples are at\nhttps://github.com/EnVision-Research/GPT4Affectivity.\n", "date": "2024-03-09", "categories": []}
{"id": "2404.06812", "title": "Emotion-cause pair extraction method based on multi-granularity\n  information and multi-module interaction", "abstract": "  The purpose of emotion-cause pair extraction is to extract the pair of\nemotion clauses and cause clauses. On the one hand, the existing methods do not\ntake fully into account the relationship between the emotion extraction of two\nauxiliary tasks. On the other hand, the existing two-stage model has the\nproblem of error propagation. In addition, existing models do not adequately\naddress the emotion and cause-induced locational imbalance of samples. To solve\nthese problems, an end-to-end multitasking model (MM-ECPE) based on shared\ninteraction between GRU, knowledge graph and transformer modules is proposed.\nFurthermore, based on MM-ECPE, in order to use the encoder layer to better\nsolve the problem of imbalanced distribution of clause distances between\nclauses and emotion clauses, we propose a novel encoding based on BERT,\nsentiment lexicon, and position-aware interaction module layer of emotion motif\npair retrieval model (MM-ECPE(BERT)). The model first fully models the\ninteraction between different tasks through the multi-level sharing module, and\nmines the shared information between emotion-cause pair extraction and the\nemotion extraction and cause extraction. Second, to solve the imbalanced\ndistribution of emotion clauses and cause clauses problem, suitable labels are\nscreened out according to the knowledge graph path length and task-specific\nfeatures are constructed so that the model can focus on extracting pairs with\ncorresponding emotion-cause relationships. Experimental results on the ECPE\nbenchmark dataset show that the proposed model achieves good performance,\nespecially on position-imbalanced samples.\n", "date": "2024-04-10", "categories": []}
{"id": "2311.18518", "title": "Color-Emotion Associations in Art: Fuzzy Approach", "abstract": "  Art objects can evoke certain emotions. Color is a fundamental element of\nvisual art and plays a significant role in how art is perceived. This paper\nintroduces a novel approach to classifying emotions in art using Fuzzy Sets. We\nemploy a fuzzy approach because it aligns well with human judgments' imprecise\nand subjective nature. Extensive fuzzy colors (n=120) and a broad emotional\nspectrum (n=10) allow for a more human-consistent and context-aware exploration\nof emotions inherent in paintings. First, we introduce the fuzzy color\nrepresentation model. Then, at the fuzzification stage, we process the Wiki Art\nDataset of paintings tagged with emotions, extracting fuzzy dominant colors\nlinked to specific emotions. This results in fuzzy color distributions for ten\nemotions. Finally, we convert them back to a crisp domain, obtaining a\nknowledge base of color-emotion associations in primary colors. Our findings\nreveal strong associations between specific emotions and colors; for instance,\ngratitude strongly correlates with green, brown, and orange. Other noteworthy\nassociations include brown and anger, orange with shame, yellow with happiness,\nand gray with fear. Using these associations and Jaccard similarity, we can\nfind the emotions in the arbitrary untagged image. We conducted a 2AFC\nexperiment involving human subjects to evaluate the proposed method. The\naverage hit rate of 0.77 indicates a significant correlation between the\nmethod's predictions and human perception. The proposed method is simple to\nadapt to art painting retrieval systems. The study contributes to the\ntheoretical understanding of color-emotion associations in art, offering\nvaluable insights for various practical applications besides art, like\nmarketing, design, and psychology.\n", "date": "2023-11-30", "categories": []}
{"id": "2403.06108", "title": "Large Language Models on Fine-grained Emotion Detection Dataset with\n  Data Augmentation and Transfer Learning", "abstract": "  This paper delves into enhancing the classification performance on the\nGoEmotions dataset, a large, manually annotated dataset for emotion detection\nin text. The primary goal of this paper is to address the challenges of\ndetecting subtle emotions in text, a complex issue in Natural Language\nProcessing (NLP) with significant practical applications. The findings offer\nvaluable insights into addressing the challenges of emotion detection in text\nand suggest directions for future research, including the potential for a\nsurvey paper that synthesizes methods and performances across various datasets\nin this domain.\n", "date": "2024-03-10", "categories": []}
{"id": "2404.06292", "title": "nEMO: Dataset of Emotional Speech in Polish", "abstract": "  Speech emotion recognition has become increasingly important in recent years\ndue to its potential applications in healthcare, customer service, and\npersonalization of dialogue systems. However, a major issue in this field is\nthe lack of datasets that adequately represent basic emotional states across\nvarious language families. As datasets covering Slavic languages are rare,\nthere is a need to address this research gap. This paper presents the\ndevelopment of nEMO, a novel corpus of emotional speech in Polish. The dataset\ncomprises over 3 hours of samples recorded with the participation of nine\nactors portraying six emotional states: anger, fear, happiness, sadness,\nsurprise, and a neutral state. The text material used was carefully selected to\nrepresent the phonetics of the Polish language adequately. The corpus is freely\navailable under the terms of a Creative Commons license (CC BY-NC-SA 4.0).\n", "date": "2024-04-09", "categories": []}
{"id": "2306.05680", "title": "Emotion Detection from EEG using Transfer Learning", "abstract": "  The detection of emotions using an Electroencephalogram (EEG) is a crucial\narea in brain-computer interfaces and has valuable applications in fields such\nas rehabilitation and medicine. In this study, we employed transfer learning to\novercome the challenge of limited data availability in EEG-based emotion\ndetection. The base model used in this study was Resnet50. Additionally, we\nemployed a novel feature combination in EEG-based emotion detection. The input\nto the model was in the form of an image matrix, which comprised Mean Phase\nCoherence (MPC) and Magnitude Squared Coherence (MSC) in the upper-triangular\nand lower-triangular matrices, respectively. We further improved the technique\nby incorporating features obtained from the Differential Entropy (DE) into the\ndiagonal, which previously held little to no useful information for classifying\nemotions. The dataset used in this study, SEED EEG (62 channel EEG), comprises\nthree classes (Positive, Neutral, and Negative). We calculated both\nsubject-independent and subject-dependent accuracy. The subject-dependent\naccuracy was obtained using a 10-fold cross-validation method and was 93.1%,\nwhile the subject-independent classification was performed by employing the\nleave-one-subject-out (LOSO) strategy. The accuracy obtained in\nsubject-independent classification was 71.6%. Both of these accuracies are at\nleast twice better than the chance accuracy of classifying 3 classes. The study\nfound the use of MSC and MPC in EEG-based emotion detection promising for\nemotion classification. The future scope of this work includes the use of data\naugmentation techniques, enhanced classifiers, and better features for emotion\nclassification.\n", "date": "2023-06-09", "categories": []}
{"id": "2401.16119", "title": "Triple Disentangled Representation Learning for Multimodal Affective\n  Analysis", "abstract": "  Multimodal learning has exhibited a significant advantage in affective\nanalysis tasks owing to the comprehensive information of various modalities,\nparticularly the complementary information. Thus, many emerging studies focus\non disentangling the modality-invariant and modality-specific representations\nfrom input data and then fusing them for prediction. However, our study shows\nthat modality-specific representations may contain information that is\nirrelevant or conflicting with the tasks, which downgrades the effectiveness of\nlearned multimodal representations. We revisit the disentanglement issue, and\npropose a novel triple disentanglement approach, TriDiRA, which disentangles\nthe modality-invariant, effective modality-specific and ineffective\nmodality-specific representations from input data. By fusing only the\nmodality-invariant and effective modality-specific representations, TriDiRA can\nsignificantly alleviate the impact of irrelevant and conflicting information\nacross modalities during model training. Extensive experiments conducted on\nfour benchmark datasets demonstrate the effectiveness and generalization of our\ntriple disentanglement, which outperforms SOTA methods.\n", "date": "2024-01-29", "categories": []}
{"id": "2404.04525", "title": "IITK at SemEval-2024 Task 10: Who is the speaker? Improving Emotion\n  Recognition and Flip Reasoning in Conversations via Speaker Embeddings", "abstract": "  This paper presents our approach for the SemEval-2024 Task 10: Emotion\nDiscovery and Reasoning its Flip in Conversations. For the Emotion Recognition\nin Conversations (ERC) task, we utilize a masked-memory network along with\nspeaker participation. We propose a transformer-based speaker-centric model for\nthe Emotion Flip Reasoning (EFR) task. We also introduce Probable Trigger Zone,\na region of the conversation that is more likely to contain the utterances\ncausing the emotion to flip. For sub-task 3, the proposed approach achieves a\n5.9 (F1 score) improvement over the task baseline. The ablation study results\nhighlight the significance of various design choices in the proposed method.\n", "date": "2024-04-06", "categories": []}
{"id": "2404.04654", "title": "Music Recommendation Based on Facial Emotion Recognition", "abstract": "  Introduction: Music provides an incredible avenue for individuals to express\ntheir thoughts and emotions, while also serving as a delightful mode of\nentertainment for enthusiasts and music lovers. Objectives: This paper presents\na comprehensive approach to enhancing the user experience through the\nintegration of emotion recognition, music recommendation, and explainable AI\nusing GRAD-CAM. Methods: The proposed methodology utilizes a ResNet50 model\ntrained on the Facial Expression Recognition (FER) dataset, consisting of real\nimages of individuals expressing various emotions. Results: The system achieves\nan accuracy of 82% in emotion classification. By leveraging GRAD-CAM, the model\nprovides explanations for its predictions, allowing users to understand the\nreasoning behind the system's recommendations. The model is trained on both FER\nand real user datasets, which include labelled facial expressions, and real\nimages of individuals expressing various emotions. The training process\ninvolves pre-processing the input images, extracting features through\nconvolutional layers, reasoning with dense layers, and generating emotion\npredictions through the output layer. Conclusion: The proposed methodology,\nleveraging the Resnet50 model with ROI-based analysis and explainable AI\ntechniques, offers a robust and interpretable solution for facial emotion\ndetection paper.\n", "date": "2024-04-06", "categories": []}
{"id": "2404.05071", "title": "Test-Time Training for Depression Detection", "abstract": "  Previous works on depression detection use datasets collected in similar\nenvironments to train and test the models. In practice, however, the train and\ntest distributions cannot be guaranteed to be identical. Distribution shifts\ncan be introduced due to variations such as recording environment (e.g.,\nbackground noise) and demographics (e.g., gender, age, etc). Such\ndistributional shifts can surprisingly lead to severe performance degradation\nof the depression detection models. In this paper, we analyze the application\nof test-time training (TTT) to improve robustness of models trained for\ndepression detection. When compared to regular testing of the models, we find\nTTT can significantly improve the robustness of the model under a variety of\ndistributional shifts introduced due to: (a) background-noise, (b) gender-bias,\nand (c) data collection and curation procedure (i.e., train and test samples\nare from separate datasets).\n", "date": "2024-04-07", "categories": []}
{"id": "2404.03361", "title": "nicolay-r at SemEval-2024 Task 3: Using Flan-T5 for Reasoning Emotion\n  Cause in Conversations with Chain-of-Thought on Emotion States", "abstract": "  Emotion expression is one of the essential traits of conversations. It may be\nself-related or caused by another speaker. The variety of reasons may serve as\na source of the further emotion causes: conversation history, speaker's\nemotional state, etc. Inspired by the most recent advances in Chain-of-Thought,\nin this work, we exploit the existing three-hop reasoning approach (THOR) to\nperform large language model instruction-tuning for answering: emotion states\n(THOR-state), and emotion caused by one speaker to the other (THOR-cause). We\nequip THOR-cause with the reasoning revision (rr) for devising a reasoning path\nin fine-tuning. In particular, we rely on the annotated speaker emotion states\nto revise reasoning path. Our final submission, based on Flan-T5-base (250M)\nand the rule-based span correction technique, preliminary tuned with THOR-state\nand fine-tuned with THOR-cause-rr on competition training data, results in 3rd\nand 4th places (F1-proportional) and 5th place (F1-strict) among 15\nparticipating teams. Our THOR implementation fork is publicly available:\nhttps://github.com/nicolay-r/THOR-ECAC\n", "date": "2024-04-04", "categories": []}
{"id": "2403.04798", "title": "JMI at SemEval 2024 Task 3: Two-step approach for multimodal ECAC using\n  in-context learning with GPT and instruction-tuned Llama models", "abstract": "  This paper presents our system development for SemEval-2024 Task 3: \"The\nCompetition of Multimodal Emotion Cause Analysis in Conversations\". Effectively\ncapturing emotions in human conversations requires integrating multiple\nmodalities such as text, audio, and video. However, the complexities of these\ndiverse modalities pose challenges for developing an efficient multimodal\nemotion cause analysis (ECA) system. Our proposed approach addresses these\nchallenges by a two-step framework. We adopt two different approaches in our\nimplementation. In Approach 1, we employ instruction-tuning with two separate\nLlama 2 models for emotion and cause prediction. In Approach 2, we use GPT-4V\nfor conversation-level video description and employ in-context learning with\nannotated conversation using GPT 3.5. Our system wins rank 4, and system\nablation experiments demonstrate that our proposed solutions achieve\nsignificant performance gains. All the experimental codes are available on\nGithub.\n", "date": "2024-03-05", "categories": []}
{"id": "2404.01805", "title": "Improved Text Emotion Prediction Using Combined Valence and Arousal\n  Ordinal Classification", "abstract": "  Emotion detection in textual data has received growing interest in recent\nyears, as it is pivotal for developing empathetic human-computer interaction\nsystems. This paper introduces a method for categorizing emotions from text,\nwhich acknowledges and differentiates between the diversified similarities and\ndistinctions of various emotions. Initially, we establish a baseline by\ntraining a transformer-based model for standard emotion classification,\nachieving state-of-the-art performance. We argue that not all\nmisclassifications are of the same importance, as there are perceptual\nsimilarities among emotional classes. We thus redefine the emotion labeling\nproblem by shifting it from a traditional classification model to an ordinal\nclassification one, where discrete emotions are arranged in a sequential order\naccording to their valence levels. Finally, we propose a method that performs\nordinal classification in the two-dimensional emotion space, considering both\nvalence and arousal scales. The results show that our approach not only\npreserves high accuracy in emotion prediction but also significantly reduces\nthe magnitude of errors in cases of misclassification.\n", "date": "2024-04-02", "categories": []}
{"id": "2404.02088", "title": "LastResort at SemEval-2024 Task 3: Exploring Multimodal Emotion Cause\n  Pair Extraction as Sequence Labelling Task", "abstract": "  Conversation is the most natural form of human communication, where each\nutterance can range over a variety of possible emotions. While significant work\nhas been done towards the detection of emotions in text, relatively little work\nhas been done towards finding the cause of the said emotions, especially in\nmultimodal settings. SemEval 2024 introduces the task of Multimodal Emotion\nCause Analysis in Conversations, which aims to extract emotions reflected in\nindividual utterances in a conversation involving multiple modalities (textual,\naudio, and visual modalities) along with the corresponding utterances that were\nthe cause for the emotion. In this paper, we propose models that tackle this\ntask as an utterance labeling and a sequence labeling problem and perform a\ncomparative study of these models, involving baselines using different\nencoders, using BiLSTM for adding contextual information of the conversation,\nand finally adding a CRF layer to try to model the inter-dependencies between\nadjacent utterances more effectively. In the official leaderboard for the task,\nour architecture was ranked 8th, achieving an F1-score of 0.1759 on the\nleaderboard.\n", "date": "2024-04-02", "categories": []}
{"id": "2211.04834", "title": "Distribution-based Emotion Recognition in Conversation", "abstract": "  Automatic emotion recognition in conversation (ERC) is crucial for\nemotion-aware conversational artificial intelligence. This paper proposes a\ndistribution-based framework that formulates ERC as a sequence-to-sequence\nproblem for emotion distribution estimation. The inherent ambiguity of emotions\nand the subjectivity of human perception lead to disagreements in emotion\nlabels, which is handled naturally in our framework from the perspective of\nuncertainty estimation in emotion distributions. A Bayesian training loss is\nintroduced to improve the uncertainty estimation by conditioning each emotional\nstate on an utterance-specific Dirichlet prior distribution. Experimental\nresults on the IEMOCAP dataset show that ERC outperformed the\nsingle-utterance-based system, and the proposed distribution-based ERC methods\nhave not only better classification accuracy, but also show improved\nuncertainty estimation.\n", "date": "2022-11-09", "categories": []}
{"id": "2306.06760", "title": "Estimating the Uncertainty in Emotion Attributes using Deep Evidential\n  Regression", "abstract": "  In automatic emotion recognition (AER), labels assigned by different human\nannotators to the same utterance are often inconsistent due to the inherent\ncomplexity of emotion and the subjectivity of perception. Though deterministic\nlabels generated by averaging or voting are often used as the ground truth, it\nignores the intrinsic uncertainty revealed by the inconsistent labels. This\npaper proposes a Bayesian approach, deep evidential emotion regression (DEER),\nto estimate the uncertainty in emotion attributes. Treating the emotion\nattribute labels of an utterance as samples drawn from an unknown Gaussian\ndistribution, DEER places an utterance-specific normal-inverse gamma prior over\nthe Gaussian likelihood and predicts its hyper-parameters using a deep neural\nnetwork model. It enables a joint estimation of emotion attributes along with\nthe aleatoric and epistemic uncertainties. AER experiments on the widely used\nMSP-Podcast and IEMOCAP datasets showed DEER produced state-of-the-art results\nfor both the mean values and the distribution of emotion attributes.\n", "date": "2023-06-11", "categories": []}
{"id": "2401.05698", "title": "HiCMAE: Hierarchical Contrastive Masked Autoencoder for Self-Supervised\n  Audio-Visual Emotion Recognition", "abstract": "  Audio-Visual Emotion Recognition (AVER) has garnered increasing attention in\nrecent years for its critical role in creating emotion-ware intelligent\nmachines. Previous efforts in this area are dominated by the supervised\nlearning paradigm. Despite significant progress, supervised learning is meeting\nits bottleneck due to the longstanding data scarcity issue in AVER. Motivated\nby recent advances in self-supervised learning, we propose Hierarchical\nContrastive Masked Autoencoder (HiCMAE), a novel self-supervised framework that\nleverages large-scale self-supervised pre-training on vast unlabeled\naudio-visual data to promote the advancement of AVER. Following prior arts in\nself-supervised audio-visual representation learning, HiCMAE adopts two primary\nforms of self-supervision for pre-training, namely masked data modeling and\ncontrastive learning. Unlike them which focus exclusively on top-layer\nrepresentations while neglecting explicit guidance of intermediate layers,\nHiCMAE develops a three-pronged strategy to foster hierarchical audio-visual\nfeature learning and improve the overall quality of learned representations. To\nverify the effectiveness of HiCMAE, we conduct extensive experiments on 9\ndatasets covering both categorical and dimensional AVER tasks. Experimental\nresults show that our method significantly outperforms state-of-the-art\nsupervised and self-supervised audio-visual methods, which indicates that\nHiCMAE is a powerful audio-visual emotion representation learner. Codes and\nmodels will be publicly available at https://github.com/sunlicai/HiCMAE.\n", "date": "2024-01-11", "categories": []}
{"id": "2401.12987", "title": "TelME: Teacher-leading Multimodal Fusion Network for Emotion Recognition\n  in Conversation", "abstract": "  Emotion Recognition in Conversation (ERC) plays a crucial role in enabling\ndialogue systems to effectively respond to user requests. The emotions in a\nconversation can be identified by the representations from various modalities,\nsuch as audio, visual, and text. However, due to the weak contribution of\nnon-verbal modalities to recognize emotions, multimodal ERC has always been\nconsidered a challenging task. In this paper, we propose Teacher-leading\nMultimodal fusion network for ERC (TelME). TelME incorporates cross-modal\nknowledge distillation to transfer information from a language model acting as\nthe teacher to the non-verbal students, thereby optimizing the efficacy of the\nweak modalities. We then combine multimodal features using a shifting fusion\napproach in which student networks support the teacher. TelME achieves\nstate-of-the-art performance in MELD, a multi-speaker conversation dataset for\nERC. Finally, we demonstrate the effectiveness of our components through\nadditional experiments.\n", "date": "2024-01-16", "categories": []}
{"id": "2402.11747", "title": "Parameter Efficient Finetuning for Speech Emotion Recognition and Domain\n  Adaptation", "abstract": "  Foundation models have shown superior performance for speech emotion\nrecognition (SER). However, given the limited data in emotion corpora,\nfinetuning all parameters of large pre-trained models for SER can be both\nresource-intensive and susceptible to overfitting. This paper investigates\nparameter-efficient finetuning (PEFT) for SER. Various PEFT adaptors are\nsystematically studied for both classification of discrete emotion categories\nand prediction of dimensional emotional attributes. The results demonstrate\nthat the combination of PEFT methods surpasses full finetuning with a\nsignificant reduction in the number of trainable parameters. Furthermore, a\ntwo-stage adaptation strategy is proposed to adapt models trained on acted\nemotion data, which is more readily available, to make the model more adept at\ncapturing natural emotional expressions. Both intra- and cross-corpus\nexperiments validate the efficacy of the proposed approach in enhancing the\nperformance on both the source and target domains.\n", "date": "2024-02-19", "categories": []}
{"id": "2403.17683", "title": "Solution for Emotion Prediction Competition of Workshop on Emotionally\n  and Culturally Intelligent AI", "abstract": "  This report provide a detailed description of the method that we explored and\nproposed in the WECIA Emotion Prediction Competition (EPC), which predicts a\nperson's emotion through an artistic work with a comment. The dataset of this\ncompetition is ArtELingo, designed to encourage work on diversity across\nlanguages and cultures. The dataset has two main challenges, namely modal\nimbalance problem and language-cultural differences problem. In order to\naddress this issue, we propose a simple yet effective approach called\nsingle-multi modal with Emotion-Cultural specific prompt(ECSP), which focuses\non using the single modal message to enhance the performance of multimodal\nmodels and a well-designed prompt to reduce cultural differences problem. To\nclarify, our approach contains two main blocks:\n(1)XLM-R\\cite{conneau2019unsupervised} based unimodal model and\nX$^2$-VLM\\cite{zeng2022x} based multimodal model (2) Emotion-Cultural specific\nprompt. Our approach ranked first in the final test with a score of 0.627.\n", "date": "2024-03-26", "categories": []}
{"id": "2404.00526", "title": "The Emotional Impact of Game Duration: A Framework for Understanding\n  Player Emotions in Extended Gameplay Sessions", "abstract": "  Video games have played a crucial role in entertainment since their\ndevelopment in the 1970s, becoming even more prominent during the lockdown\nperiod when people were looking for ways to entertain them. However, at that\ntime, players were unaware of the significant impact that playtime could have\non their feelings. This has made it challenging for designers and developers to\ncreate new games since they have to control the emotional impact that these\ngames will take on players. Thus, the purpose of this study is to look at how a\nplayer's emotions are affected by the duration of the game. In order to achieve\nthis goal, a framework for emotion detection is created. According to the\nexperiment's results, the volunteers' general ability to express emotions\nincreased from 20 to 60 minutes. In comparison to shorter gameplay sessions,\nthe experiment found that extended gameplay sessions did significantly affect\nthe player's emotions. According to the results, it was recommended that in\norder to lessen the potential emotional impact that playing computer and video\ngames may have in the future, game producers should think about creating\nshorter, entertaining games.\n", "date": "2024-03-31", "categories": []}
{"id": "2404.01243", "title": "A Unified and Interpretable Emotion Representation and Expression\n  Generation", "abstract": "  Canonical emotions, such as happy, sad, and fearful, are easy to understand\nand annotate. However, emotions are often compound, e.g. happily surprised, and\ncan be mapped to the action units (AUs) used for expressing emotions, and\ntrivially to the canonical ones. Intuitively, emotions are continuous as\nrepresented by the arousal-valence (AV) model. An interpretable unification of\nthese four modalities - namely, Canonical, Compound, AUs, and AV - is highly\ndesirable, for a better representation and understanding of emotions. However,\nsuch unification remains to be unknown in the current literature. In this work,\nwe propose an interpretable and unified emotion model, referred as C2A2. We\nalso develop a method that leverages labels of the non-unified models to\nannotate the novel unified one. Finally, we modify the text-conditional\ndiffusion models to understand continuous numbers, which are then used to\ngenerate continuous expressions using our unified emotion model. Through\nquantitative and qualitative experiments, we show that our generated images are\nrich and capture subtle expressions. Our work allows a fine-grained generation\nof expressions in conjunction with other textual inputs and offers a new label\nspace for emotions at the same time.\n", "date": "2024-04-01", "categories": []}
{"id": "2403.12687", "title": "Audio-Visual Compound Expression Recognition Method based on Late\n  Modality Fusion and Rule-based Decision", "abstract": "  This paper presents the results of the SUN team for the Compound Expressions\nRecognition Challenge of the 6th ABAW Competition. We propose a novel\naudio-visual method for compound expression recognition. Our method relies on\nemotion recognition models that fuse modalities at the emotion probability\nlevel, while decisions regarding the prediction of compound expressions are\nbased on predefined rules. Notably, our method does not use any training data\nspecific to the target task. Thus, the problem is a zero-shot classification\ntask. The method is evaluated in multi-corpus training and cross-corpus\nvalidation setups. Using our proposed method is achieved an F1-score value\nequals to 22.01% on the C-EXPR-DB test subset. Our findings from the challenge\ndemonstrate that the proposed method can potentially form a basis for\ndeveloping intelligent tools for annotating audio-visual data in the context of\nhuman's basic and compound emotions.\n", "date": "2024-03-19", "categories": []}
{"id": "2403.20289", "title": "Emotion-Anchored Contrastive Learning Framework for Emotion Recognition\n  in Conversation", "abstract": "  Emotion Recognition in Conversation (ERC) involves detecting the underlying\nemotion behind each utterance within a conversation. Effectively generating\nrepresentations for utterances remains a significant challenge in this task.\nRecent works propose various models to address this issue, but they still\nstruggle with differentiating similar emotions such as excitement and\nhappiness. To alleviate this problem, We propose an Emotion-Anchored\nContrastive Learning (EACL) framework that can generate more distinguishable\nutterance representations for similar emotions. To achieve this, we utilize\nlabel encodings as anchors to guide the learning of utterance representations\nand design an auxiliary loss to ensure the effective separation of anchors for\nsimilar emotions. Moreover, an additional adaptation process is proposed to\nadapt anchors to serve as effective classifiers to improve classification\nperformance. Across extensive experiments, our proposed EACL achieves\nstate-of-the-art emotion recognition performance and exhibits superior\nperformance on similar emotions. Our code is available at\nhttps://github.com/Yu-Fangxu/EACL.\n", "date": "2024-03-29", "categories": []}
{"id": "2403.19014", "title": "Thelxino\\\"e: Recognizing Human Emotions Using Pupillometry and Machine\n  Learning", "abstract": "  In this study, we present a method for emotion recognition in Virtual Reality\n(VR) using pupillometry. We analyze pupil diameter responses to both visual and\nauditory stimuli via a VR headset and focus on extracting key features in the\ntime-domain, frequency-domain, and time-frequency domain from VR generated\ndata. Our approach utilizes feature selection to identify the most impactful\nfeatures using Maximum Relevance Minimum Redundancy (mRMR). By applying a\nGradient Boosting model, an ensemble learning technique using stacked decision\ntrees, we achieve an accuracy of 98.8% with feature engineering, compared to\n84.9% without it. This research contributes significantly to the Thelxino\\\"e\nframework, aiming to enhance VR experiences by integrating multiple sensor data\nfor realistic and emotionally resonant touch interactions. Our findings open\nnew avenues for developing more immersive and interactive VR environments,\npaving the way for future advancements in virtual touch technology.\n", "date": "2024-03-27", "categories": []}
{"id": "2403.19554", "title": "Cross-Attention is Not Always Needed: Dynamic Cross-Attention for\n  Audio-Visual Dimensional Emotion Recognition", "abstract": "  In video-based emotion recognition, audio and visual modalities are often\nexpected to have a complementary relationship, which is widely explored using\ncross-attention. However, they may also exhibit weak complementary\nrelationships, resulting in poor representations of audio-visual features, thus\ndegrading the performance of the system. To address this issue, we propose\nDynamic Cross-Attention (DCA) that can dynamically select cross-attended or\nunattended features on the fly based on their strong or weak complementary\nrelationship with each other, respectively. Specifically, a simple yet\nefficient gating layer is designed to evaluate the contribution of the\ncross-attention mechanism and choose cross-attended features only when they\nexhibit a strong complementary relationship, otherwise unattended features. We\nevaluate the performance of the proposed approach on the challenging RECOLA and\nAff-Wild2 datasets. We also compare the proposed approach with other variants\nof cross-attention and show that the proposed model consistently improves the\nperformance on both datasets.\n", "date": "2024-03-28", "categories": []}
{"id": "2403.18635", "title": "Fusion approaches for emotion recognition from speech using acoustic and\n  text-based features", "abstract": "  In this paper, we study different approaches for classifying emotions from\nspeech using acoustic and text-based features. We propose to obtain\ncontextualized word embeddings with BERT to represent the information contained\nin speech transcriptions and show that this results in better performance than\nusing Glove embeddings. We also propose and compare different strategies to\ncombine the audio and text modalities, evaluating them on IEMOCAP and\nMSP-PODCAST datasets. We find that fusing acoustic and text-based systems is\nbeneficial on both datasets, though only subtle differences are observed across\nthe evaluated fusion approaches. Finally, for IEMOCAP, we show the large effect\nthat the criteria used to define the cross-validation folds have on results. In\nparticular, the standard way of creating folds for this dataset results in a\nhighly optimistic estimation of performance for the text-based system,\nsuggesting that some previous works may overestimate the advantage of\nincorporating transcriptions.\n", "date": "2024-03-27", "categories": []}
{"id": "2311.09602", "title": "Language Models (Mostly) Do Not Consider Emotion Triggers When\n  Predicting Emotion", "abstract": "  Situations and events evoke emotions in humans, but to what extent do they\ninform the prediction of emotion detection models? This work investigates how\nwell human-annotated emotion triggers correlate with features that models\ndeemed salient in their prediction of emotions. First, we introduce a novel\ndataset EmoTrigger, consisting of 900 social media posts sourced from three\ndifferent datasets; these were annotated by experts for emotion triggers with\nhigh agreement. Using EmoTrigger, we evaluate the ability of large language\nmodels (LLMs) to identify emotion triggers, and conduct a comparative analysis\nof the features considered important for these tasks between LLMs and\nfine-tuned models. Our analysis reveals that emotion triggers are largely not\nconsidered salient features for emotion prediction models, instead there is\nintricate interplay between various features and the task of emotion detection.\n", "date": "2023-11-16", "categories": []}
{"id": "2403.17379", "title": "Exploring and Applying Audio-Based Sentiment Analysis in Music", "abstract": "  Sentiment analysis is a continuously explored area of text processing that\ndeals with the computational analysis of opinions, sentiments, and subjectivity\nof text. However, this idea is not limited to text and speech, in fact, it\ncould be applied to other modalities. In reality, humans do not express\nthemselves in text as deeply as they do in music. The ability of a\ncomputational model to interpret musical emotions is largely unexplored and\ncould have implications and uses in therapy and musical queuing. In this paper,\ntwo individual tasks are addressed. This study seeks to (1) predict the emotion\nof a musical clip over time and (2) determine the next emotion value after the\nmusic in a time series to ensure seamless transitions. Utilizing data from the\nEmotions in Music Database, which contains clips of songs selected from the\nFree Music Archive annotated with levels of valence and arousal as reported on\nRussel's circumplex model of affect by multiple volunteers, models are trained\nfor both tasks. Overall, the performance of these models reflected that they\nwere able to perform the tasks they were designed for effectively and\naccurately.\n", "date": "2024-02-22", "categories": []}
{"id": "2403.15574", "title": "SensoryT5: Infusing Sensorimotor Norms into T5 for Enhanced Fine-grained\n  Emotion Classification", "abstract": "  In traditional research approaches, sensory perception and emotion\nclassification have traditionally been considered separate domains. Yet, the\nsignificant influence of sensory experiences on emotional responses is\nundeniable. The natural language processing (NLP) community has often missed\nthe opportunity to merge sensory knowledge with emotion classification. To\naddress this gap, we propose SensoryT5, a neuro-cognitive approach that\nintegrates sensory information into the T5 (Text-to-Text Transfer Transformer)\nmodel, designed specifically for fine-grained emotion classification. This\nmethodology incorporates sensory cues into the T5's attention mechanism,\nenabling a harmonious balance between contextual understanding and sensory\nawareness. The resulting model amplifies the richness of emotional\nrepresentations. In rigorous tests across various detailed emotion\nclassification datasets, SensoryT5 showcases improved performance, surpassing\nboth the foundational T5 model and current state-of-the-art works. Notably,\nSensoryT5's success signifies a pivotal change in the NLP domain, highlighting\nthe potential influence of neuro-cognitive data in refining machine learning\nmodels' emotional sensitivity.\n", "date": "2024-03-22", "categories": []}
{"id": "2403.16263", "title": "Emotion Recognition from the perspective of Activity Recognition", "abstract": "  Applications of an efficient emotion recognition system can be found in\nseveral domains such as medicine, driver fatigue surveillance, social robotics,\nand human-computer interaction. Appraising human emotional states, behaviors,\nand reactions displayed in real-world settings can be accomplished using latent\ncontinuous dimensions. Continuous dimensional models of human affect, such as\nthose based on valence and arousal are more accurate in describing a broad\nrange of spontaneous everyday emotions than more traditional models of discrete\nstereotypical emotion categories (e.g. happiness, surprise). Most of the prior\nwork on estimating valence and arousal considers laboratory settings and acted\ndata. But, for emotion recognition systems to be deployed and integrated into\nreal-world mobile and computing devices, we need to consider data collected in\nthe world. Action recognition is a domain of Computer Vision that involves\ncapturing complementary information on appearance from still frames and motion\nbetween frames. In this paper, we treat emotion recognition from the\nperspective of action recognition by exploring the application of deep learning\narchitectures specifically designed for action recognition, for continuous\naffect recognition. We propose a novel three-stream end-to-end deep learning\nregression pipeline with an attention mechanism, which is an ensemble design\nbased on sub-modules of multiple state-of-the-art action recognition systems.\nThe pipeline constitutes a novel data pre-processing approach with a spatial\nself-attention mechanism to extract keyframes. The optical flow of\nhigh-attention regions of the face is extracted to capture temporal context.\nAFEW-VA in-the-wild dataset has been used to conduct comparative experiments.\nQuantitative analysis shows that the proposed model outperforms multiple\nstandard baselines of both emotion recognition and action recognition models.\n", "date": "2024-03-24", "categories": []}
{"id": "2308.04025", "title": "MSAC: Multiple Speech Attribute Control Method for Reliable Speech\n  Emotion Recognition", "abstract": "  Despite notable progress, speech emotion recognition (SER) remains\nchallenging due to the intricate and ambiguous nature of speech emotion,\nparticularly in wild world. While current studies primarily focus on\nrecognition and generalization abilities, our research pioneers an\ninvestigation into the reliability of SER methods in the presence of semantic\ndata shifts and explores how to exert fine-grained control over various\nattributes inherent in speech signals to enhance speech emotion modeling. In\nthis paper, we first introduce MSAC-SERNet, a novel unified SER framework\ncapable of simultaneously handling both single-corpus and cross-corpus SER.\nSpecifically, concentrating exclusively on the speech emotion attribute, a\nnovel CNN-based SER model is presented to extract discriminative emotional\nrepresentations, guided by additive margin softmax loss. Considering\ninformation overlap between various speech attributes, we propose a novel\nlearning paradigm based on correlations of different speech attributes, termed\nMultiple Speech Attribute Control (MSAC), which empowers the proposed SER model\nto simultaneously capture fine-grained emotion-related features while\nmitigating the negative impact of emotion-agnostic representations.\nFurthermore, we make a first attempt to examine the reliability of the\nMSAC-SERNet framework using out-of-distribution detection methods. Experiments\non both single-corpus and cross-corpus SER scenarios indicate that MSAC-SERNet\nnot only consistently outperforms the baseline in all aspects, but achieves\nsuperior performance compared to state-of-the-art SER approaches.\n", "date": "2023-08-08", "categories": []}
{"id": "2403.15044", "title": "Multimodal Fusion with Pre-Trained Model Features in Affective Behaviour\n  Analysis In-the-wild", "abstract": "  Multimodal fusion is a significant method for most multimodal tasks. With the\nrecent surge in the number of large pre-trained models, combining both\nmultimodal fusion methods and pre-trained model features can achieve\noutstanding performance in many multimodal tasks. In this paper, we present our\napproach, which leverages both advantages for addressing the task of Expression\n(Expr) Recognition and Valence-Arousal (VA) Estimation. We evaluate the\nAff-Wild2 database using pre-trained models, then extract the final hidden\nlayers of the models as features. Following preprocessing and interpolation or\nconvolution to align the extracted features, different models are employed for\nmodal fusion. Our code is available at GitHub - FulgenceWen/ABAW6th.\n", "date": "2024-03-22", "categories": []}
{"id": "2403.14006", "title": "On Prompt Sensitivity of ChatGPT in Affective Computing", "abstract": "  Recent studies have demonstrated the emerging capabilities of foundation\nmodels like ChatGPT in several fields, including affective computing. However,\naccessing these emerging capabilities is facilitated through prompt\nengineering. Despite the existence of some prompting techniques, the field is\nstill rapidly evolving and many prompting ideas still require investigation. In\nthis work, we introduce a method to evaluate and investigate the sensitivity of\nthe performance of foundation models based on different prompts or generation\nparameters. We perform our evaluation on ChatGPT within the scope of affective\ncomputing on three major problems, namely sentiment analysis, toxicity\ndetection, and sarcasm detection. First, we carry out a sensitivity analysis on\npivotal parameters in auto-regressive text generation, specifically the\ntemperature parameter $T$ and the top-$p$ parameter in Nucleus sampling,\ndictating how conservative or creative the model should be during generation.\nFurthermore, we explore the efficacy of several prompting ideas, where we\nexplore how giving different incentives or structures affect the performance.\nOur evaluation takes into consideration performance measures on the affective\ncomputing tasks, and the effectiveness of the model to follow the stated\ninstructions, hence generating easy-to-parse responses to be smoothly used in\ndownstream applications.\n", "date": "2024-03-20", "categories": []}
{"id": "2303.11117", "title": "EmotionIC: emotional inertia and contagion-driven dependency modeling\n  for emotion recognition in conversation", "abstract": "  Emotion Recognition in Conversation (ERC) has attracted growing attention in\nrecent years as a result of the advancement and implementation of\nhuman-computer interface technologies. In this paper, we propose an emotional\ninertia and contagion-driven dependency modeling approach (EmotionIC) for ERC\ntask. Our EmotionIC consists of three main components, i.e., Identity Masked\nMulti-Head Attention (IMMHA), Dialogue-based Gated Recurrent Unit (DiaGRU), and\nSkip-chain Conditional Random Field (SkipCRF). Compared to previous ERC models,\nEmotionIC can model a conversation more thoroughly at both the\nfeature-extraction and classification levels. The proposed model attempts to\nintegrate the advantages of attention- and recurrence-based methods at the\nfeature-extraction level. Specifically, IMMHA is applied to capture\nidentity-based global contextual dependencies, while DiaGRU is utilized to\nextract speaker- and temporal-aware local contextual information. At the\nclassification level, SkipCRF can explicitly mine complex emotional flows from\nhigher-order neighboring utterances in the conversation. Experimental results\nshow that our method can significantly outperform the state-of-the-art models\non four benchmark datasets. The ablation studies confirm that our modules can\neffectively model emotional inertia and contagion.\n", "date": "2023-03-20", "categories": []}
{"id": "2403.11757", "title": "Efficient Feature Extraction and Late Fusion Strategy for Audiovisual\n  Emotional Mimicry Intensity Estimation", "abstract": "  In this paper, we present the solution to the Emotional Mimicry Intensity\n(EMI) Estimation challenge, which is part of 6th Affective Behavior Analysis\nin-the-wild (ABAW) Competition.The EMI Estimation challenge task aims to\nevaluate the emotional intensity of seed videos by assessing them from a set of\npredefined emotion categories (i.e., \"Admiration\", \"Amusement\",\n\"Determination\", \"Empathic Pain\", \"Excitement\" and \"Joy\"). To tackle this\nchallenge, we extracted rich dual-channel visual features based on ResNet18 and\nAUs for the video modality and effective single-channel features based on\nWav2Vec2.0 for the audio modality. This allowed us to obtain comprehensive\nemotional features for the audiovisual modality. Additionally, leveraging a\nlate fusion strategy, we averaged the predictions of the visual and acoustic\nmodels, resulting in a more accurate estimation of audiovisual emotional\nmimicry intensity. Experimental results validate the effectiveness of our\napproach, with the average Pearson's correlation Coefficient($\\rho$) across the\n6 emotion dimensionson the validation set achieving 0.3288.\n", "date": "2024-03-18", "categories": []}
{"id": "2403.12425", "title": "Multimodal Fusion Method with Spatiotemporal Sequences and Relationship\n  Learning for Valence-Arousal Estimation", "abstract": "  This paper presents our approach for the VA (Valence-Arousal) estimation task\nin the ABAW6 competition. We devised a comprehensive model by preprocessing\nvideo frames and audio segments to extract visual and audio features. Through\nthe utilization of Temporal Convolutional Network (TCN) modules, we effectively\ncaptured the temporal and spatial correlations between these features.\nSubsequently, we employed a Transformer encoder structure to learn long-range\ndependencies, thereby enhancing the model's performance and generalization\nability. Our method leverages a multimodal data fusion approach, integrating\npre-trained audio and video backbones for feature extraction, followed by\nTCN-based spatiotemporal encoding and Transformer-based temporal information\ncapture. Experimental results demonstrate the effectiveness of our approach,\nachieving competitive performance in VA estimation on the AffWild2 dataset.\n", "date": "2024-03-19", "categories": []}
{"id": "2305.08473", "title": "Shared and Private Information Learning in Multimodal Sentiment Analysis\n  with Deep Modal Alignment and Self-supervised Multi-Task Learning", "abstract": "  Designing an effective representation learning method for multimodal\nsentiment analysis tasks is a crucial research direction. The challenge lies in\nlearning both shared and private information in a complete modal\nrepresentation, which is difficult with uniform multimodal labels and a raw\nfeature fusion approach. In this work, we propose a deep modal shared\ninformation learning module based on the covariance matrix to capture the\nshared information between modalities. Additionally, we use a label generation\nmodule based on a self-supervised learning strategy to capture the private\ninformation of the modalities. Our module is plug-and-play in multimodal tasks,\nand by changing the parameterization, it can adjust the information exchange\nrelationship between the modes and learn the private or shared information\nbetween the specified modes. We also employ a multi-task learning strategy to\nhelp the model focus its attention on the modal differentiation training data.\nWe provide a detailed formulation derivation and feasibility proof for the\ndesign of the deep modal shared information learning module. We conduct\nextensive experiments on three common multimodal sentiment analysis baseline\ndatasets, and the experimental results validate the reliability of our model.\nFurthermore, we explore more combinatorial techniques for the use of the\nmodule. Our approach outperforms current state-of-the-art methods on most of\nthe metrics of the three public datasets.\n", "date": "2023-05-15", "categories": []}
{"id": "2403.12609", "title": "SUN Team's Contribution to ABAW 2024 Competition: Audio-visual\n  Valence-Arousal Estimation and Expression Recognition", "abstract": "  As emotions play a central role in human communication, automatic emotion\nrecognition has attracted increasing attention in the last two decades. While\nmultimodal systems enjoy high performances on lab-controlled data, they are\nstill far from providing ecological validity on non-lab-controlled, namely\n'in-the-wild' data. This work investigates audiovisual deep learning approaches\nfor emotion recognition in-the-wild problem. We particularly explore the\neffectiveness of architectures based on fine-tuned Convolutional Neural\nNetworks (CNN) and Public Dimensional Emotion Model (PDEM), for video and audio\nmodality, respectively. We compare alternative temporal modeling and fusion\nstrategies using the embeddings from these multi-stage trained\nmodality-specific Deep Neural Networks (DNN). We report results on the AffWild2\ndataset under Affective Behavior Analysis in-the-Wild 2024 (ABAW'24) challenge\nprotocol.\n", "date": "2024-03-19", "categories": []}
{"id": "2312.04293", "title": "GPT-4V with Emotion: A Zero-shot Benchmark for Generalized Emotion\n  Recognition", "abstract": "  Recently, GPT-4 with Vision (GPT-4V) has demonstrated remarkable visual\ncapabilities across various tasks, but its performance in emotion recognition\nhas not been fully evaluated. To bridge this gap, we present the quantitative\nevaluation results of GPT-4V on 21 benchmark datasets covering 6 tasks: visual\nsentiment analysis, tweet sentiment analysis, micro-expression recognition,\nfacial emotion recognition, dynamic facial emotion recognition, and multimodal\nemotion recognition. This paper collectively refers to these tasks as\n``Generalized Emotion Recognition (GER)''. Through experimental analysis, we\nobserve that GPT-4V exhibits strong visual understanding capabilities in GER\ntasks. Meanwhile, GPT-4V shows the ability to integrate multimodal clues and\nexploit temporal information, which is also critical for emotion recognition.\nHowever, it's worth noting that GPT-4V is primarily designed for general\ndomains and cannot recognize micro-expressions that require specialized\nknowledge. To the best of our knowledge, this paper provides the first\nquantitative assessment of GPT-4V for GER tasks. We have open-sourced the code\nand encourage subsequent researchers to broaden the evaluation scope by\nincluding more tasks and datasets. Our code and evaluation results are\navailable at: https://github.com/zeroQiaoba/gpt4v-emotion.\n", "date": "2023-12-07", "categories": []}
{"id": "2403.01222", "title": "Emotion Analysis in NLP: Trends, Gaps and Roadmap for Future Directions", "abstract": "  Emotions are a central aspect of communication. Consequently, emotion\nanalysis (EA) is a rapidly growing field in natural language processing (NLP).\nHowever, there is no consensus on scope, direction, or methods. In this paper,\nwe conduct a thorough review of 154 relevant NLP publications from the last\ndecade. Based on this review, we address four different questions: (1) How are\nEA tasks defined in NLP? (2) What are the most prominent emotion frameworks and\nwhich emotions are modeled? (3) Is the subjectivity of emotions considered in\nterms of demographics and cultural factors? and (4) What are the primary NLP\napplications for EA? We take stock of trends in EA and tasks, emotion\nframeworks used, existing datasets, methods, and applications. We then discuss\nfour lacunae: (1) the absence of demographic and cultural aspects does not\naccount for the variation in how emotions are perceived, but instead assumes\nthey are universally experienced in the same manner; (2) the poor fit of\nemotion categories from the two main emotion theories to the task; (3) the lack\nof standardized EA terminology hinders gap identification, comparison, and\nfuture goals; and (4) the absence of interdisciplinary research isolates EA\nfrom insights in other fields. Our work will enable more focused research into\nEA and a more holistic approach to modeling emotions in NLP.\n", "date": "2024-03-02", "categories": []}
{"id": "2403.10764", "title": "ECRC: Emotion-Causality Recognition in Korean Conversation for GCN", "abstract": "  In this multi-task learning study on simultaneous analysis of emotions and\ntheir underlying causes in conversational contexts, deep neural network methods\nwere employed to effectively process and train large labeled datasets. However,\nthese approaches are typically limited to conducting context analyses across\nthe entire corpus because they rely on one of the two methods: word- or\nsentence-level embedding. The former struggles with polysemy and homonyms,\nwhereas the latter causes information loss when processing long sentences. In\nthis study, we overcome the limitations of previous embeddings by utilizing\nboth word- and sentence-level embeddings. Furthermore, we propose the\nemotion-causality recognition in conversation (ECRC) model, which is based on a\nnovel graph structure, thereby leveraging the strengths of both embedding\nmethods. This model uniquely integrates the bidirectional long short-term\nmemory (Bi-LSTM) and graph neural network (GCN) models for Korean conversation\nanalysis. Compared with models that rely solely on one embedding method, the\nproposed model effectively structures abstract concepts, such as language\nfeatures and relationships, thereby minimizing information loss. To assess\nmodel performance, we compared the multi-task learning results of three deep\nneural network models with varying graph structures. Additionally, we evaluated\nthe proposed model using Korean and English datasets. The experimental results\nshow that the proposed model performs better in emotion and causality\nmulti-task learning (74.62% and 75.30%, respectively) when node and edge\ncharacteristics are incorporated into the graph structure. Similar results were\nrecorded for the Korean ECC and Wellness datasets (74.62% and 73.44%,\nrespectively) with 71.35% on the IEMOCAP English dataset.\n", "date": "2024-03-16", "categories": []}
{"id": "2403.10825", "title": "Affective Behaviour Analysis via Integrating Multi-Modal Knowledge", "abstract": "  Affective Behavior Analysis aims to facilitate technology emotionally smart,\ncreating a world where devices can understand and react to our emotions as\nhumans do. To comprehensively evaluate the authenticity and applicability of\nemotional behavior analysis techniques in natural environments, the 6th\ncompetition on Affective Behavior Analysis in-the-wild (ABAW) utilizes the\nAff-Wild2, Hume-Vidmimic2, and C-EXPR-DB datasets to set up five competitive\ntracks, i.e., Valence-Arousal (VA) Estimation, Expression (EXPR) Recognition,\nAction Unit (AU) Detection, Compound Expression (CE) Recognition, and Emotional\nMimicry Intensity (EMI) Estimation. In this paper, we present our method\ndesigns for the five tasks. Specifically, our design mainly includes three\naspects: 1) Utilizing a transformer-based feature fusion module to fully\nintegrate emotional information provided by audio signals, visual images, and\ntranscripts, offering high-quality expression features for the downstream\ntasks. 2) To achieve high-quality facial feature representations, we employ\nMasked-Auto Encoder as the visual features extraction model and fine-tune it\nwith our facial dataset. 3) Considering the complexity of the video collection\nscenes, we conduct a more detailed dataset division based on scene\ncharacteristics and train the classifier for each scene. Extensive experiments\ndemonstrate the superiority of our designs.\n", "date": "2024-03-16", "categories": []}
{"id": "2403.11440", "title": "Boosting Continuous Emotion Recognition with Self-Pretraining using\n  Masked Autoencoders, Temporal Convolutional Networks, and Transformers", "abstract": "  Human emotion recognition holds a pivotal role in facilitating seamless\nhuman-computer interaction. This paper delineates our methodology in tackling\nthe Valence-Arousal (VA) Estimation Challenge, Expression (Expr) Classification\nChallenge, and Action Unit (AU) Detection Challenge within the ambit of the 6th\nWorkshop and Competition on Affective Behavior Analysis in-the-wild (ABAW). Our\nstudy advocates a novel approach aimed at refining continuous emotion\nrecognition. We achieve this by initially harnessing pre-training with Masked\nAutoencoders (MAE) on facial datasets, followed by fine-tuning on the aff-wild2\ndataset annotated with expression (Expr) labels. The pre-trained model serves\nas an adept visual feature extractor, thereby enhancing the model's robustness.\nFurthermore, we bolster the performance of continuous emotion recognition by\nintegrating Temporal Convolutional Network (TCN) modules and Transformer\nEncoder modules into our framework.\n", "date": "2024-03-18", "categories": []}
{"id": "2403.11450", "title": "Zero-shot Compound Expression Recognition with Visual Language Model at\n  the 6th ABAW Challenge", "abstract": "  Conventional approaches to facial expression recognition primarily focus on\nthe classification of six basic facial expressions. Nevertheless, real-world\nsituations present a wider range of complex compound expressions that consist\nof combinations of these basics ones due to limited availability of\ncomprehensive training datasets. The 6th Workshop and Competition on Affective\nBehavior Analysis in-the-wild (ABAW) offered unlabeled datasets containing\ncompound expressions. In this study, we propose a zero-shot approach for\nrecognizing compound expressions by leveraging a pretrained visual language\nmodel integrated with some traditional CNN networks.\n", "date": "2024-03-18", "categories": []}
{"id": "2403.11590", "title": "HSEmotion Team at the 6th ABAW Competition: Facial Expressions,\n  Valence-Arousal and Emotion Intensity Prediction", "abstract": "  This article presents our results for the sixth Affective Behavior Analysis\nin-the-wild (ABAW) competition. To improve the trustworthiness of facial\nanalysis, we study the possibility of using pre-trained deep models that\nextract reliable emotional features without the need to fine-tune the neural\nnetworks for a downstream task. In particular, we introduce several lightweight\nmodels based on MobileViT, MobileFaceNet, EfficientNet, and DDAMFN\narchitectures trained in multi-task scenarios to recognize facial expressions,\nvalence, and arousal on static photos. These neural networks extract\nframe-level features fed into a simple classifier, e.g., linear feed-forward\nneural network, to predict emotion intensity, compound expressions, action\nunits, facial expressions, and valence/arousal. Experimental results for five\ntasks from the sixth ABAW challenge demonstrate that our approach lets us\nsignificantly improve quality metrics on validation sets compared to existing\nnon-ensemble techniques.\n", "date": "2024-03-18", "categories": []}
{"id": "2309.04849", "title": "Speech Emotion Recognition with Distilled Prosodic and Linguistic Affect\n  Representations", "abstract": "  We propose EmoDistill, a novel speech emotion recognition (SER) framework\nthat leverages cross-modal knowledge distillation during training to learn\nstrong linguistic and prosodic representations of emotion from speech. During\ninference, our method only uses a stream of speech signals to perform unimodal\nSER thus reducing computation overhead and avoiding run-time transcription and\nprosodic feature extraction errors. During training, our method distills\ninformation at both embedding and logit levels from a pair of pre-trained\nProsodic and Linguistic teachers that are fine-tuned for SER. Experiments on\nthe IEMOCAP benchmark demonstrate that our method outperforms other unimodal\nand multimodal techniques by a considerable margin, and achieves\nstate-of-the-art performance of 77.49% unweighted accuracy and 78.91% weighted\naccuracy. Detailed ablation studies demonstrate the impact of each component of\nour method.\n", "date": "2023-09-09", "categories": []}
{"id": "2209.08614", "title": "Deep Adaptation of Adult-Child Facial Expressions by Fusing Landmark\n  Features", "abstract": "  Imaging of facial affects may be used to measure psychophysiological\nattributes of children through their adulthood for applications in education,\nhealthcare, and entertainment, among others. Deep convolutional neural networks\nshow promising results in classifying facial expressions of adults. However,\nclassifier models trained with adult benchmark data are unsuitable for learning\nchild expressions due to discrepancies in psychophysical development.\nSimilarly, models trained with child data perform poorly in adult expression\nclassification. We propose domain adaptation to concurrently align\ndistributions of adult and child expressions in a shared latent space for\nrobust classification of either domain. Furthermore, age variations in facial\nimages are studied in age-invariant face recognition yet remain unleveraged in\nadult-child expression classification. We take inspiration from multiple fields\nand propose deep adaptive FACial Expressions fusing BEtaMix SElected Landmark\nFeatures (FACE-BE-SELF) for adult-child expression classification. For the\nfirst time in the literature, a mixture of Beta distributions is used to\ndecompose and select facial features based on correlations with expression,\ndomain, and identity factors. We evaluate FACE-BE-SELF using 5-fold cross\nvalidation for two pairs of adult-child data sets. Our proposed FACE-BE-SELF\napproach outperforms transfer learning and other baseline domain adaptation\nmethods in aligning latent representations of adult and child expressions.\n", "date": "2022-09-18", "categories": []}
{"id": "2403.09451", "title": "M&M: Multimodal-Multitask Model Integrating Audiovisual Cues in\n  Cognitive Load Assessment", "abstract": "  This paper introduces the M&M model, a novel multimodal-multitask learning\nframework, applied to the AVCAffe dataset for cognitive load assessment (CLA).\nM&M uniquely integrates audiovisual cues through a dual-pathway architecture,\nfeaturing specialized streams for audio and video inputs. A key innovation lies\nin its cross-modality multihead attention mechanism, fusing the different\nmodalities for synchronized multitasking. Another notable feature is the\nmodel's three specialized branches, each tailored to a specific cognitive load\nlabel, enabling nuanced, task-specific analysis. While it shows modest\nperformance compared to the AVCAffe's single-task baseline, M\\&M demonstrates a\npromising framework for integrated multimodal processing. This work paves the\nway for future enhancements in multimodal-multitask learning systems,\nemphasizing the fusion of diverse data types for complex task handling.\n", "date": "2024-03-14", "categories": []}
{"id": "2402.19344", "title": "The 6th Affective Behavior Analysis in-the-wild (ABAW) Competition", "abstract": "  This paper describes the 6th Affective Behavior Analysis in-the-wild (ABAW)\nCompetition, which is part of the respective Workshop held in conjunction with\nIEEE CVPR 2024. The 6th ABAW Competition addresses contemporary challenges in\nunderstanding human emotions and behaviors, crucial for the development of\nhuman-centered technologies. In more detail, the Competition focuses on affect\nrelated benchmarking tasks and comprises of five sub-challenges: i)\nValence-Arousal Estimation (the target is to estimate two continuous affect\ndimensions, valence and arousal), ii) Expression Recognition (the target is to\nrecognise between the mutually exclusive classes of the 7 basic expressions and\n'other'), iii) Action Unit Detection (the target is to detect 12 action units),\niv) Compound Expression Recognition (the target is to recognise between the 7\nmutually exclusive compound expression classes), and v) Emotional Mimicry\nIntensity Estimation (the target is to estimate six continuous emotion\ndimensions). In the paper, we present these Challenges, describe their\nrespective datasets and challenge protocols (we outline the evaluation metrics)\nand present the baseline systems as well as their obtained performance. More\ninformation for the Competition can be found in:\nhttps://affective-behavior-analysis-in-the-wild.github.io/6th.\n", "date": "2024-02-29", "categories": []}
{"id": "2403.04789", "title": "TopicDiff: A Topic-enriched Diffusion Approach for Multimodal\n  Conversational Emotion Detection", "abstract": "  Multimodal Conversational Emotion (MCE) detection, generally spanning across\nthe acoustic, vision and language modalities, has attracted increasing interest\nin the multimedia community. Previous studies predominantly focus on learning\ncontextual information in conversations with only a few considering the topic\ninformation in single language modality, while always neglecting the acoustic\nand vision topic information. On this basis, we propose a model-agnostic\nTopic-enriched Diffusion (TopicDiff) approach for capturing multimodal topic\ninformation in MCE tasks. Particularly, we integrate the diffusion model into\nneural topic model to alleviate the diversity deficiency problem of neural\ntopic model in capturing topic information. Detailed evaluations demonstrate\nthe significant improvements of TopicDiff over the state-of-the-art MCE\nbaselines, justifying the importance of multimodal topic information to MCE and\nthe effectiveness of TopicDiff in capturing such information. Furthermore, we\nobserve an interesting finding that the topic information in acoustic and\nvision is more discriminative and robust compared to the language.\n", "date": "2024-03-04", "categories": []}
{"id": "2402.17269", "title": "Curriculum Learning Meets Directed Acyclic Graph for Multimodal Emotion\n  Recognition", "abstract": "  Emotion recognition in conversation (ERC) is a crucial task in natural\nlanguage processing and affective computing. This paper proposes MultiDAG+CL, a\nnovel approach for Multimodal Emotion Recognition in Conversation (ERC) that\nemploys Directed Acyclic Graph (DAG) to integrate textual, acoustic, and visual\nfeatures within a unified framework. The model is enhanced by Curriculum\nLearning (CL) to address challenges related to emotional shifts and data\nimbalance. Curriculum learning facilitates the learning process by gradually\npresenting training samples in a meaningful order, thereby improving the\nmodel's performance in handling emotional variations and data imbalance.\nExperimental results on the IEMOCAP and MELD datasets demonstrate that the\nMultiDAG+CL models outperform baseline models. We release the code for\nMultiDAG+CL and experiments: https://github.com/vanntc711/MultiDAG-CL\n", "date": "2024-02-27", "categories": []}
{"id": "2403.00632", "title": "Metamorpheus: Interactive, Affective, and Creative Dream Narration\n  Through Metaphorical Visual Storytelling", "abstract": "  Human emotions are essentially molded by lived experiences, from which we\nconstruct personalised meaning. The engagement in such meaning-making process\nhas been practiced as an intervention in various psychotherapies to promote\nwellness. Nevertheless, to support recollecting and recounting lived\nexperiences in everyday life remains under explored in HCI. It also remains\nunknown how technologies such as generative AI models can facilitate the\nmeaning making process, and ultimately support affective mindfulness. In this\npaper we present Metamorpheus, an affective interface that engages users in a\ncreative visual storytelling of emotional experiences during dreams.\nMetamorpheus arranges the storyline based on a dream's emotional arc, and\nprovokes self-reflection through the creation of metaphorical images and text\ndepictions. The system provides metaphor suggestions, and generates visual\nmetaphors and text depictions using generative AI models, while users can apply\ngenerations to recolour and re-arrange the interface to be visually affective.\nOur experience-centred evaluation manifests that, by interacting with\nMetamorpheus, users can recall their dreams in vivid detail, through which they\nrelive and reflect upon their experiences in a meaningful way.\n", "date": "2024-03-01", "categories": []}
{"id": "2403.00887", "title": "SEGAA: A Unified Approach to Predicting Age, Gender, and Emotion in\n  Speech", "abstract": "  The interpretation of human voices holds importance across various\napplications. This study ventures into predicting age, gender, and emotion from\nvocal cues, a field with vast applications. Voice analysis tech advancements\nspan domains, from improving customer interactions to enhancing healthcare and\nretail experiences. Discerning emotions aids mental health, while age and\ngender detection are vital in various contexts. Exploring deep learning models\nfor these predictions involves comparing single, multi-output, and sequential\nmodels highlighted in this paper. Sourcing suitable data posed challenges,\nresulting in the amalgamation of the CREMA-D and EMO-DB datasets. Prior work\nshowed promise in individual predictions, but limited research considered all\nthree variables simultaneously. This paper identifies flaws in an individual\nmodel approach and advocates for our novel multi-output learning architecture\nSpeech-based Emotion Gender and Age Analysis (SEGAA) model. The experiments\nsuggest that Multi-output models perform comparably to individual models,\nefficiently capturing the intricate relationships between variables and speech\ninputs, all while achieving improved runtime.\n", "date": "2024-03-01", "categories": []}
{"id": "2403.02112", "title": "A New Perspective on Smiling and Laughter Detection: Intensity Levels\n  Matter", "abstract": "  Smiles and laughs detection systems have attracted a lot of attention in the\npast decade contributing to the improvement of human-agent interaction systems.\nBut very few considered these expressions as distinct, although no prior work\nclearly proves them to belong to the same category or not. In this work, we\npresent a deep learning-based multimodal smile and laugh classification system,\nconsidering them as two different entities. We compare the use of audio and\nvision-based models as well as a fusion approach. We show that, as expected,\nthe fusion leads to a better generalization on unseen data. We also present an\nin-depth analysis of the behavior of these models on the smiles and laughs\nintensity levels. The analyses on the intensity levels show that the\nrelationship between smiles and laughs might not be as simple as a binary one\nor even grouping them in a single category, and so, a more complex approach\nshould be taken when dealing with them. We also tackle the problem of limited\nresources by showing that transfer learning allows the models to improve the\ndetection of confusing intensity levels.\n", "date": "2024-03-04", "categories": []}
{"id": "2403.00366", "title": "Exploring the dynamic interplay of cognitive load and emotional arousal\n  by using multimodal measurements: Correlation of pupil diameter and emotional\n  arousal in emotionally engaging tasks", "abstract": "  Multimodal data analysis and validation based on streams from\nstate-of-the-art sensor technology such as eye-tracking or emotion recognition\nusing the Facial Action Coding System (FACTs) with deep learning allows\neducational researchers to study multifaceted learning and problem-solving\nprocesses and to improve educational experiences. This study aims to\ninvestigate the correlation between two continuous sensor streams, pupil\ndiameter as an indicator of cognitive workload and FACTs with deep learning as\nan indicator of emotional arousal (RQ 1a), specifically for epochs of high,\nmedium, and low arousal (RQ 1b). Furthermore, the time lag between emotional\narousal and pupil diameter data will be analyzed (RQ 2). 28 participants worked\non three cognitively demanding and emotionally engaging everyday moral dilemmas\nwhile eye-tracking and emotion recognition data were collected. The data were\npre-processed in Phyton (synchronization, blink control, downsampling) and\nanalyzed using correlation analysis and Granger causality tests. The results\nshow negative and statistically significant correlations between the data\nstreams for emotional arousal and pupil diameter. However, the correlation is\nnegative and significant only for epochs of high arousal, while positive but\nnon-significant relationships were found for epochs of medium or low arousal.\nThe average time lag for the relationship between arousal and pupil diameter\nwas 2.8 ms. In contrast to previous findings without a multimodal approach\nsuggesting a positive correlation between the constructs, the results\ncontribute to the state of research by highlighting the importance of\nmultimodal data validation and research on convergent vagility. Future research\nshould consider emotional regulation strategies and emotional valence.\n", "date": "2024-03-01", "categories": []}
{"id": "2402.18944", "title": "SemEval 2024 -- Task 10: Emotion Discovery and Reasoning its Flip in\n  Conversation (EDiReF)", "abstract": "  We present SemEval-2024 Task 10, a shared task centred on identifying\nemotions and finding the rationale behind their flips within monolingual\nEnglish and Hindi-English code-mixed dialogues. This task comprises three\ndistinct subtasks - emotion recognition in conversation for code-mixed\ndialogues, emotion flip reasoning for code-mixed dialogues, and emotion flip\nreasoning for English dialogues. Participating systems were tasked to\nautomatically execute one or more of these subtasks. The datasets for these\ntasks comprise manually annotated conversations focusing on emotions and\ntriggers for emotion shifts (The task data is available at\nhttps://github.com/LCS2-IIITD/EDiReF-SemEval2024.git). A total of 84\nparticipants engaged in this task, with the most adept systems attaining\nF1-scores of 0.70, 0.79, and 0.76 for the respective subtasks. This paper\nsummarises the results and findings from 24 teams alongside their system\ndescriptions.\n", "date": "2024-02-29", "categories": []}
{"id": "2312.11576", "title": "Emotion Based Prediction in the Context of Optimized Trajectory Planning\n  for Immersive Learning", "abstract": "  In the virtual elements of immersive learning, the use of Google Expedition\nand touch-screen-based emotion are examined. The objective is to investigate\npossible ways to combine these technologies to enhance virtual learning\nenvironments and learners emotional engagement. Pedagogical application,\naffordances, and cognitive load are the corresponding measures that are\ninvolved. Students will gain insight into the reason behind their significantly\nhigher post-assessment Prediction Systems scores compared to preassessment\nscores through this work that leverages technology. This suggests that it is\neffective to include emotional elements in immersive learning scenarios. The\nresults of this study may help develop new strategies by leveraging the\nfeatures of immersive learning technology in educational technologies to\nimprove virtual reality and augmented reality experiences. Furthermore, the\neffectiveness of immersive learning environments can be raised by utilizing\nmagnetic, optical, or hybrid trackers that considerably improve object\ntracking.\n", "date": "2023-12-18", "categories": []}
{"id": "2402.17792", "title": "EGNN-C+: Interpretable Evolving Granular Neural Network and Application\n  in Classification of Weakly-Supervised EEG Data Streams", "abstract": "  We introduce a modified incremental learning algorithm for evolving Granular\nNeural Network Classifiers (eGNN-C+). We use double-boundary hyper-boxes to\nrepresent granules, and customize the adaptation procedures to enhance the\nrobustness of outer boxes for data coverage and noise suppression, while\nensuring that inner boxes remain flexible to capture drifts. The classifier\nevolves from scratch, incorporates new classes on the fly, and performs local\nincremental feature weighting. As an application, we focus on the\nclassification of emotion-related patterns within electroencephalogram (EEG)\nsignals. Emotion recognition is crucial for enhancing the realism and\ninteractivity of computer systems. We extract features from the Fourier\nspectrum of EEG signals obtained from 28 individuals engaged in playing\ncomputer games -- a public dataset. Each game elicits a different predominant\nemotion: boredom, calmness, horror, or joy. We analyze individual electrodes,\ntime window lengths, and frequency bands to assess the accuracy and\ninterpretability of resulting user-independent neural models. The findings\nindicate that both brain hemispheres assist classification, especially\nelectrodes on the temporal (T8) and parietal (P7) areas, alongside\ncontributions from frontal and occipital electrodes. While patterns may\nmanifest in any band, the Alpha (8-13Hz), Delta (1-4Hz), and Theta (4-8Hz)\nbands, in this order, exhibited higher correspondence with the emotion classes.\nThe eGNN-C+ demonstrates effectiveness in learning EEG data. It achieves an\naccuracy of 81.7% and a 0.0029 II interpretability using 10-second time\nwindows, even in face of a highly-stochastic time-varying 4-class\nclassification problem.\n", "date": "2024-02-26", "categories": []}
{"id": "2402.18241", "title": "Affective State Detection using fNIRs and Machine Learning", "abstract": "  Affective states regulate our day to day to function and has a tremendous\neffect on mental and physical health. Detection of affective states is of\nutmost importance for mental health monitoring, smart entertainment selection\nand dynamic workload management. In this paper, we discussed relevant\nliterature on affective state detection using physiology data, the benefits and\nlimitations of different sensors and methods used for collecting physiology\ndata, and our rationale for selecting functional near-infrared spectroscopy. We\npresent the design of an experiment involving nine subjects to evoke the\naffective states of meditation, amusement and cognitive load and the results of\nthe attempt to classify using machine learning. A mean accuracy of 83.04% was\nachieved in three class classification with an individual model; 84.39%\naccuracy was achieved for a group model and 60.57% accuracy was achieved for\nsubject independent model using leave one out cross validation. It was found\nthat prediction accuracy for cognitive load was higher (evoked using a pen and\npaper task) than the other two classes (evoked using computer bases tasks). To\nverify that this discrepancy was not due to motor skills involved in the pen\nand paper task, a second experiment was conducted using four participants and\nthe results of that experiment has also been presented in the paper.\n", "date": "2024-02-28", "categories": []}
{"id": "2402.17437", "title": "Exploiting Emotion-Semantic Correlations for Empathetic Response\n  Generation", "abstract": "  Empathetic response generation aims to generate empathetic responses by\nunderstanding the speaker's emotional feelings from the language of dialogue.\nRecent methods capture emotional words in the language of communicators and\nconstruct them as static vectors to perceive nuanced emotions. However,\nlinguistic research has shown that emotional words in language are dynamic and\nhave correlations with other grammar semantic roles, i.e., words with semantic\nmeanings, in grammar. Previous methods overlook these two characteristics,\nwhich easily lead to misunderstandings of emotions and neglect of key\nsemantics. To address this issue, we propose a dynamical Emotion-Semantic\nCorrelation Model (ESCM) for empathetic dialogue generation tasks. ESCM\nconstructs dynamic emotion-semantic vectors through the interaction of context\nand emotions. We introduce dependency trees to reflect the correlations between\nemotions and semantics. Based on dynamic emotion-semantic vectors and\ndependency trees, we propose a dynamic correlation graph convolutional network\nto guide the model in learning context meanings in dialogue and generating\nempathetic responses. Experimental results on the EMPATHETIC-DIALOGUES dataset\nshow that ESCM understands semantics and emotions more accurately and expresses\nfluent and informative empathetic responses. Our analysis results also indicate\nthat the correlations between emotions and semantics are frequently used in\ndialogues, which is of great significance for empathetic perception and\nexpression.\n", "date": "2024-02-27", "categories": []}
{"id": "2402.15513", "title": "Investigating the Generalizability of Physiological Characteristics of\n  Anxiety", "abstract": "  Recent works have demonstrated the effectiveness of machine learning (ML)\ntechniques in detecting anxiety and stress using physiological signals, but it\nis unclear whether ML models are learning physiological features specific to\nstress. To address this ambiguity, we evaluated the generalizability of\nphysiological features that have been shown to be correlated with anxiety and\nstress to high-arousal emotions. Specifically, we examine features extracted\nfrom electrocardiogram (ECG) and electrodermal (EDA) signals from the following\nthree datasets: Anxiety Phases Dataset (APD), Wearable Stress and Affect\nDetection (WESAD), and the Continuously Annotated Signals of Emotion (CASE)\ndataset. We aim to understand whether these features are specific to anxiety or\ngeneral to other high-arousal emotions through a statistical regression\nanalysis, in addition to a within-corpus, cross-corpus, and\nleave-one-corpus-out cross-validation across instances of stress and arousal.\nWe used the following classifiers: Support Vector Machines, LightGBM, Random\nForest, XGBoost, and an ensemble of the aforementioned models. We found that\nmodels trained on an arousal dataset perform relatively well on a previously\nunseen stress dataset, and vice versa. Our experimental results suggest that\nthe evaluated models may be identifying emotional arousal instead of stress.\nThis work is the first cross-corpus evaluation across stress and arousal from\nECG and EDA signals, contributing new findings about the generalizability of\nstress detection.\n", "date": "2024-01-23", "categories": []}
{"id": "2402.15662", "title": "GiMeFive: Towards Interpretable Facial Emotion Classification", "abstract": "  Deep convolutional neural networks have been shown to successfully recognize\nfacial emotions for the past years in the realm of computer vision. However,\nthe existing detection approaches are not always reliable or explainable, we\nhere propose our model GiMeFive with interpretations, i.e., via layer\nactivations and gradient-weighted class activation mapping. We compare against\nthe state-of-the-art methods to classify the six facial emotions. Empirical\nresults show that our model outperforms the previous methods in terms of\naccuracy on two Facial Emotion Recognition (FER) benchmarks and our aggregated\nFER GiMeFive. Furthermore, we explain our work in real-world image and video\nexamples, as well as real-time live camera streams. Our code and supplementary\nmaterial are available at https: //github.com/werywjw/SEP-CVDL.\n", "date": "2024-02-24", "categories": []}
{"id": "2402.16194", "title": "ASEM: Enhancing Empathy in Chatbot through Attention-based Sentiment and\n  Emotion Modeling", "abstract": "  Effective feature representations play a critical role in enhancing the\nperformance of text generation models that rely on deep neural networks.\nHowever, current approaches suffer from several drawbacks, such as the\ninability to capture the deep semantics of language and sensitivity to minor\ninput variations, resulting in significant changes in the generated text. In\nthis paper, we present a novel solution to these challenges by employing a\nmixture of experts, multiple encoders, to offer distinct perspectives on the\nemotional state of the user's utterance while simultaneously enhancing\nperformance. We propose an end-to-end model architecture called ASEM that\nperforms emotion analysis on top of sentiment analysis for open-domain\nchatbots, enabling the generation of empathetic responses that are fluent and\nrelevant. In contrast to traditional attention mechanisms, the proposed model\nemploys a specialized attention strategy that uniquely zeroes in on sentiment\nand emotion nuances within the user's utterance. This ensures the generation of\ncontext-rich representations tailored to the underlying emotional tone and\nsentiment intricacies of the text. Our approach outperforms existing methods\nfor generating empathetic embeddings, providing empathetic and diverse\nresponses. The performance of our proposed model significantly exceeds that of\nexisting models, enhancing emotion detection accuracy by 6.2% and lexical\ndiversity by 1.4%.\n", "date": "2024-02-25", "categories": []}
{"id": "2402.14879", "title": "Driving Generative Agents With Their Personality", "abstract": "  This research explores the potential of Large Language Models (LLMs) to\nutilize psychometric values, specifically personality information, within the\ncontext of video game character development. Affective Computing (AC) systems\nquantify a Non-Player character's (NPC) psyche, and an LLM can take advantage\nof the system's information by using the values for prompt generation. The\nresearch shows an LLM can consistently represent a given personality profile,\nthereby enhancing the human-like characteristics of game characters.\nRepurposing a human examination, the International Personality Item Pool (IPIP)\nquestionnaire, to evaluate an LLM shows that the model can accurately generate\ncontent concerning the personality provided. Results show that the improvement\nof LLM, such as the latest GPT-4 model, can consistently utilize and interpret\na personality to represent behavior.\n", "date": "2024-02-21", "categories": []}
{"id": "2402.15195", "title": "The AffectToolbox: Affect Analysis for Everyone", "abstract": "  In the field of affective computing, where research continually advances at a\nrapid pace, the demand for user-friendly tools has become increasingly\napparent. In this paper, we present the AffectToolbox, a novel software system\nthat aims to support researchers in developing affect-sensitive studies and\nprototypes. The proposed system addresses the challenges posed by existing\nframeworks, which often require profound programming knowledge and cater\nprimarily to power-users or skilled developers. Aiming to facilitate ease of\nuse, the AffectToolbox requires no programming knowledge and offers its\nfunctionality to reliably analyze the affective state of users through an\naccessible graphical user interface. The architecture encompasses a variety of\nmodels for emotion recognition on multiple affective channels and modalities,\nas well as an elaborate fusion system to merge multi-modal assessments into a\nunified result. The entire system is open-sourced and will be publicly\navailable to ensure easy integration into more complex applications through a\nwell-structured, Python-based code base - therefore marking a substantial\ncontribution toward advancing affective computing research and fostering a more\ncollaborative and inclusive environment within this interdisciplinary field.\n", "date": "2024-02-23", "categories": []}
{"id": "2402.13955", "title": "BEE-NET: A deep neural network to identify in-the-wild Bodily Expression\n  of Emotions", "abstract": "  In this study, we investigate how environmental factors, specifically the\nscenes and objects involved, can affect the expression of emotions through body\nlanguage. To this end, we introduce a novel multi-stream deep convolutional\nneural network named BEE-NET. We also propose a new late fusion strategy that\nincorporates meta-information on places and objects as prior knowledge in the\nlearning process. Our proposed probabilistic pooling model leverages this\ninformation to generate a joint probability distribution of both available and\nanticipated non-available contextual information in latent space. Importantly,\nour fusion strategy is differentiable, allowing for end-to-end training and\ncapturing of hidden associations among data points without requiring further\npost-processing or regularisation. To evaluate our deep model, we use the Body\nLanguage Database (BoLD), which is currently the largest available database for\nthe Automatic Identification of the in-the-wild Bodily Expression of Emotions\n(AIBEE). Our experimental results demonstrate that our proposed approach\nsurpasses the current state-of-the-art in AIBEE by a margin of 2.07%, achieving\nan Emotional Recognition Score of 66.33%.\n", "date": "2024-02-21", "categories": []}
{"id": "2401.06659", "title": "WisdoM: Improving Multimodal Sentiment Analysis by Fusing Contextual\n  World Knowledge", "abstract": "  Sentiment analysis is rapidly advancing by utilizing various data modalities\n(e.g., text, image). However, most previous works relied on superficial\ninformation, neglecting the incorporation of contextual world knowledge (e.g.,\nbackground information derived from but beyond the given image and text pairs)\nand thereby restricting their ability to achieve better multimodal sentiment\nanalysis (MSA). In this paper, we proposed a plug-in framework named WisdoM, to\nleverage the contextual world knowledge induced from the large vision-language\nmodels (LVLMs) for enhanced MSA. WisdoM utilizes LVLMs to comprehensively\nanalyze both images and corresponding texts, simultaneously generating\npertinent context. To reduce the noise in the context, we also introduce a\ntraining-free contextual fusion mechanism. Experiments across diverse\ngranularities of MSA tasks consistently demonstrate that our approach has\nsubstantial improvements (brings an average +1.96% F1 score among five advanced\nmethods) over several state-of-the-art methods.\n", "date": "2024-01-12", "categories": []}
{"id": "2402.10921", "title": "AM^2-EmoJE: Adaptive Missing-Modality Emotion Recognition in\n  Conversation via Joint Embedding Learning", "abstract": "  Human emotion can be presented in different modes i.e., audio, video, and\ntext. However, the contribution of each mode in exhibiting each emotion is not\nuniform. Furthermore, the availability of complete mode-specific details may\nnot always be guaranteed in the test time. In this work, we propose AM^2-EmoJE,\na model for Adaptive Missing-Modality Emotion Recognition in Conversation via\nJoint Embedding Learning model that is grounded on two-fold contributions:\nFirst, a query adaptive fusion that can automatically learn the relative\nimportance of its mode-specific representations in a query-specific manner. By\nthis the model aims to prioritize the mode-invariant spatial query details of\nthe emotion patterns, while also retaining its mode-exclusive aspects within\nthe learned multimodal query descriptor. Second the multimodal joint embedding\nlearning module that explicitly addresses various missing modality scenarios in\ntest-time. By this, the model learns to emphasize on the correlated patterns\nacross modalities, which may help align the cross-attended mode-specific\ndescriptors pairwise within a joint-embedding space and thereby compensate for\nmissing modalities during inference. By leveraging the spatio-temporal details\nat the dialogue level, the proposed AM^2-EmoJE not only demonstrates superior\nperformance compared to the best-performing state-of-the-art multimodal\nmethods, by effectively leveraging body language in place of face expression,\nit also exhibits an enhanced privacy feature. By reporting around 2-5%\nimprovement in the weighted-F1 score, the proposed multimodal joint embedding\nmodule facilitates an impressive performance gain in a variety of\nmissing-modality query scenarios during test time.\n", "date": "2024-01-26", "categories": []}
{"id": "2402.08837", "title": "Learning to Generate Context-Sensitive Backchannel Smiles for Embodied\n  AI Agents with Applications in Mental Health Dialogues", "abstract": "  Addressing the critical shortage of mental health resources for effective\nscreening, diagnosis, and treatment remains a significant challenge. This\nscarcity underscores the need for innovative solutions, particularly in\nenhancing the accessibility and efficacy of therapeutic support. Embodied\nagents with advanced interactive capabilities emerge as a promising and\ncost-effective supplement to traditional caregiving methods. Crucial to these\nagents' effectiveness is their ability to simulate non-verbal behaviors, like\nbackchannels, that are pivotal in establishing rapport and understanding in\ntherapeutic contexts but remain under-explored. To improve the rapport-building\ncapabilities of embodied agents we annotated backchannel smiles in videos of\nintimate face-to-face conversations over topics such as mental health, illness,\nand relationships. We hypothesized that both speaker and listener behaviors\naffect the duration and intensity of backchannel smiles. Using cues from speech\nprosody and language along with the demographics of the speaker and listener,\nwe found them to contain significant predictors of the intensity of backchannel\nsmiles. Based on our findings, we introduce backchannel smile production in\nembodied agents as a generation problem. Our attention-based generative model\nsuggests that listener information offers performance improvements over the\nbaseline speaker-centric generation approach. Conditioned generation using the\nsignificant predictors of smile intensity provides statistically significant\nimprovements in empirical measures of generation quality. Our user study by\ntransferring generated smiles to an embodied agent suggests that agent with\nbackchannel smiles is perceived to be more human-like and is an attractive\nalternative for non-personal conversations over agent without backchannel\nsmiles.\n", "date": "2024-02-13", "categories": []}
{"id": "2402.07326", "title": "Persian Speech Emotion Recognition by Fine-Tuning Transformers", "abstract": "  Given the significance of speech emotion recognition, numerous methods have\nbeen developed in recent years to create effective and efficient systems in\nthis domain. One of these methods involves the use of pretrained transformers,\nfine-tuned to address this specific problem, resulting in high accuracy.\nDespite extensive discussions and global-scale efforts to enhance these\nsystems, the application of this innovative and effective approach has received\nless attention in the context of Persian speech emotion recognition. In this\narticle, we review the field of speech emotion recognition and its background,\nwith an emphasis on the importance of employing transformers in this context.\nWe present two models, one based on spectrograms and the other on the audio\nitself, fine-tuned using the shEMO dataset. These models significantly enhance\nthe accuracy of previous systems, increasing it from approximately 65% to 80%\non the mentioned dataset. Subsequently, to investigate the effect of\nmultilinguality on the fine-tuning process, these same models are fine-tuned\ntwice. First, they are fine-tuned using the English IEMOCAP dataset, and then\nthey are fine-tuned with the Persian shEMO dataset. This results in an improved\naccuracy of 82% for the Persian emotion recognition system. Keywords: Persian\nSpeech Emotion Recognition, shEMO, Self-Supervised Learning\n", "date": "2024-02-11", "categories": []}
{"id": "2402.07327", "title": "Multi-Modal Emotion Recognition by Text, Speech and Video Using\n  Pretrained Transformers", "abstract": "  Due to the complex nature of human emotions and the diversity of emotion\nrepresentation methods in humans, emotion recognition is a challenging field.\nIn this research, three input modalities, namely text, audio (speech), and\nvideo, are employed to generate multimodal feature vectors. For generating\nfeatures for each of these modalities, pre-trained Transformer models with\nfine-tuning are utilized. In each modality, a Transformer model is used with\ntransfer learning to extract feature and emotional structure. These features\nare then fused together, and emotion recognition is performed using a\nclassifier. To select an appropriate fusion method and classifier, various\nfeature-level and decision-level fusion techniques have been experimented with,\nand ultimately, the best model, which combines feature-level fusion by\nconcatenating feature vectors and classification using a Support Vector Machine\non the IEMOCAP multimodal dataset, achieves an accuracy of 75.42%. Keywords:\nMultimodal Emotion Recognition, IEMOCAP, Self-Supervised Learning, Transfer\nLearning, Transformer.\n", "date": "2024-02-11", "categories": []}
{"id": "2402.02184", "title": "Sentiment analysis in non-fixed length audios using a Fully\n  Convolutional Neural Network", "abstract": "  In this work, a sentiment analysis method that is capable of accepting audio\nof any length, without being fixed a priori, is proposed. Mel spectrogram and\nMel Frequency Cepstral Coefficients are used as audio description methods and a\nFully Convolutional Neural Network architecture is proposed as a classifier.\nThe results have been validated using three well known datasets: EMODB,\nRAVDESS, and TESS. The results obtained were promising, outperforming the\nstate-of-the-art methods. Also, thanks to the fact that the proposed method\nadmits audios of any size, it allows a sentiment analysis to be made in near\nreal time, which is very interesting for a wide range of fields such as call\ncenters, medical consultations, or financial brokers.\n", "date": "2024-02-03", "categories": []}
{"id": "2402.04232", "title": "Can Generative Agents Predict Emotion?", "abstract": "  Large Language Models (LLMs) have demonstrated a number of human-like\nabilities, however the empathic understanding and emotional state of LLMs is\nyet to be aligned to that of humans. In this work, we investigate how the\nemotional state of generative LLM agents evolves as they perceive new events,\nintroducing a novel architecture in which new experiences are compared to past\nmemories. Through this comparison, the agent gains the ability to understand\nnew experiences in context, which according to the appraisal theory of emotion\nis vital in emotion creation. First, the agent perceives new experiences as\ntime series text data. After perceiving each new input, the agent generates a\nsummary of past relevant memories, referred to as the norm, and compares the\nnew experience to this norm. Through this comparison we can analyse how the\nagent reacts to the new experience in context. The PANAS, a test of affect, is\nadministered to the agent, capturing the emotional state of the agent after the\nperception of the new event. Finally, the new experience is then added to the\nagents memory to be used in the creation of future norms. By creating multiple\nexperiences in natural language from emotionally charged situations, we test\nthe proposed architecture on a wide range of scenarios. The mixed results\nsuggests that introducing context can occasionally improve the emotional\nalignment of the agent, but further study and comparison with human evaluators\nis necessary. We hope that this paper is another step towards the alignment of\ngenerative agents.\n", "date": "2024-02-06", "categories": []}
{"id": "2312.09043", "title": "Topic Bias in Emotion Classification", "abstract": "  Emotion corpora are typically sampled based on keyword/hashtag search or by\nasking study participants to generate textual instances. In any case, these\ncorpora are not uniform samples representing the entirety of a domain. We\nhypothesize that this practice of data acquisition leads to unrealistic\ncorrelations between overrepresented topics in these corpora that harm the\ngeneralizability of models. Such topic bias could lead to wrong predictions for\ninstances like \"I organized the service for my aunt's funeral.\" when funeral\nevents are over-represented for instances labeled with sadness, despite the\nemotion of pride being more appropriate here. In this paper, we study this\ntopic bias both from the data and the modeling perspective. We first label a\nset of emotion corpora automatically via topic modeling and show that emotions\nin fact correlate with specific topics. Further, we see that emotion\nclassifiers are confounded by such topics. Finally, we show that the\nestablished debiasing method of adversarial correction via gradient reversal\nmitigates the issue. Our work points out issues with existing emotion corpora\nand that more representative resources are required for fair evaluation of\nmodels predicting affective concepts from text.\n", "date": "2023-12-14", "categories": []}
{"id": "2402.01422", "title": "EmoSpeaker: One-shot Fine-grained Emotion-Controlled Talking Face\n  Generation", "abstract": "  Implementing fine-grained emotion control is crucial for emotion generation\ntasks because it enhances the expressive capability of the generative model,\nallowing it to accurately and comprehensively capture and express various\nnuanced emotional states, thereby improving the emotional quality and\npersonalization of generated content. Generating fine-grained facial animations\nthat accurately portray emotional expressions using only a portrait and an\naudio recording presents a challenge. In order to address this challenge, we\npropose a visual attribute-guided audio decoupler. This enables the obtention\nof content vectors solely related to the audio content, enhancing the stability\nof subsequent lip movement coefficient predictions. To achieve more precise\nemotional expression, we introduce a fine-grained emotion coefficient\nprediction module. Additionally, we propose an emotion intensity control method\nusing a fine-grained emotion matrix. Through these, effective control over\nemotional expression in the generated videos and finer classification of\nemotion intensity are accomplished. Subsequently, a series of 3DMM coefficient\ngeneration networks are designed to predict 3D coefficients, followed by the\nutilization of a rendering network to generate the final video. Our\nexperimental results demonstrate that our proposed method, EmoSpeaker,\noutperforms existing emotional talking face generation methods in terms of\nexpression variation and lip synchronization. Project page:\nhttps://peterfanfan.github.io/EmoSpeaker/\n", "date": "2024-02-02", "categories": []}
{"id": "2402.01592", "title": "Towards Sustainable Workplace Mental Health: A Novel Approach to Early\n  Intervention and Support", "abstract": "  Employee well-being is a critical concern in the contemporary workplace, as\nhighlighted by the American Psychological Association's 2021 report, indicating\nthat 71% of employees experience stress or tension. This stress contributes\nsignificantly to workplace attrition and absenteeism, with 61% of attrition and\n16% of sick days attributed to poor mental health. A major challenge for\nemployers is that employees often remain unaware of their mental health issues\nuntil they reach a crisis point, resulting in limited utilization of corporate\nwell-being benefits. This research addresses this challenge by presenting a\ngroundbreaking stress detection algorithm that provides real-time support\npreemptively. Leveraging automated chatbot technology, the algorithm\nobjectively measures mental health levels by analyzing chat conversations,\noffering personalized treatment suggestions in real-time based on linguistic\nbiomarkers. The study explores the feasibility of integrating these innovations\ninto practical learning applications within real-world contexts and introduces\na chatbot-style system integrated into the broader employee experience\nplatform. This platform, encompassing various features, aims to enhance overall\nemployee well-being, detect stress in real time, and proactively engage with\nindividuals to improve support effectiveness, demonstrating a 22% increase when\nassistance is provided early. Overall, the study emphasizes the importance of\nfostering a supportive workplace environment for employees' mental health.\n", "date": "2024-02-02", "categories": []}
{"id": "2311.04507", "title": "Conversation Understanding using Relational Temporal Graph Neural\n  Networks with Auxiliary Cross-Modality Interaction", "abstract": "  Emotion recognition is a crucial task for human conversation understanding.\nIt becomes more challenging with the notion of multimodal data, e.g., language,\nvoice, and facial expressions. As a typical solution, the global- and the local\ncontext information are exploited to predict the emotional label for every\nsingle sentence, i.e., utterance, in the dialogue. Specifically, the global\nrepresentation could be captured via modeling of cross-modal interactions at\nthe conversation level. The local one is often inferred using the temporal\ninformation of speakers or emotional shifts, which neglects vital factors at\nthe utterance level. Additionally, most existing approaches take fused features\nof multiple modalities in an unified input without leveraging modality-specific\nrepresentations. Motivating from these problems, we propose the Relational\nTemporal Graph Neural Network with Auxiliary Cross-Modality Interaction\n(CORECT), an novel neural network framework that effectively captures\nconversation-level cross-modality interactions and utterance-level temporal\ndependencies with the modality-specific manner for conversation understanding.\nExtensive experiments demonstrate the effectiveness of CORECT via its\nstate-of-the-art results on the IEMOCAP and CMU-MOSEI datasets for the\nmultimodal ERC task.\n", "date": "2023-11-08", "categories": []}
{"id": "2401.16541", "title": "GuReT: Distinguishing Guilt and Regret related Text", "abstract": "  The intricate relationship between human decision-making and emotions,\nparticularly guilt and regret, has significant implications on behavior and\nwell-being. Yet, these emotions subtle distinctions and interplay are often\noverlooked in computational models. This paper introduces a dataset tailored to\ndissect the relationship between guilt and regret and their unique textual\nmarkers, filling a notable gap in affective computing research. Our approach\ntreats guilt and regret recognition as a binary classification task and employs\nthree machine learning and six transformer-based deep learning techniques to\nbenchmark the newly created dataset. The study further implements innovative\nreasoning methods like chain-of-thought and tree-of-thought to assess the\nmodels interpretive logic. The results indicate a clear performance edge for\ntransformer-based models, achieving a 90.4% macro F1 score compared to the\n85.3% scored by the best machine learning classifier, demonstrating their\nsuperior capability in distinguishing complex emotional states.\n", "date": "2024-01-29", "categories": []}
{"id": "2401.15164", "title": "AMuSE: Adaptive Multimodal Analysis for Speaker Emotion Recognition in\n  Group Conversations", "abstract": "  Analyzing individual emotions during group conversation is crucial in\ndeveloping intelligent agents capable of natural human-machine interaction.\nWhile reliable emotion recognition techniques depend on different modalities\n(text, audio, video), the inherent heterogeneity between these modalities and\nthe dynamic cross-modal interactions influenced by an individual's unique\nbehavioral patterns make the task of emotion recognition very challenging. This\ndifficulty is compounded in group settings, where the emotion and its temporal\nevolution are not only influenced by the individual but also by external\ncontexts like audience reaction and context of the ongoing conversation. To\nmeet this challenge, we propose a Multimodal Attention Network that captures\ncross-modal interactions at various levels of spatial abstraction by jointly\nlearning its interactive bunch of mode-specific Peripheral and Central\nnetworks. The proposed MAN injects cross-modal attention via its Peripheral\nkey-value pairs within each layer of a mode-specific Central query network. The\nresulting cross-attended mode-specific descriptors are then combined using an\nAdaptive Fusion technique that enables the model to integrate the\ndiscriminative and complementary mode-specific data patterns within an\ninstance-specific multimodal descriptor. Given a dialogue represented by a\nsequence of utterances, the proposed AMuSE model condenses both spatial and\ntemporal features into two dense descriptors: speaker-level and\nutterance-level. This helps not only in delivering better classification\nperformance (3-5% improvement in Weighted-F1 and 5-7% improvement in Accuracy)\nin large-scale public datasets but also helps the users in understanding the\nreasoning behind each emotion prediction made by the model via its Multimodal\nExplainability Visualization module.\n", "date": "2024-01-26", "categories": []}
{"id": "2308.06450", "title": "ERNetCL: A novel emotion recognition network in textual conversation\n  based on curriculum learning strategy", "abstract": "  Emotion recognition in conversation (ERC) has emerged as a research hotspot\nin domains such as conversational robots and question-answer systems. How to\nefficiently and adequately retrieve contextual emotional cues has been one of\nthe key challenges in the ERC task. Existing efforts do not fully model the\ncontext and employ complex network structures, resulting in limited performance\ngains. In this paper, we propose a novel emotion recognition network based on\ncurriculum learning strategy (ERNetCL). The proposed ERNetCL primarily consists\nof temporal encoder (TE), spatial encoder (SE), and curriculum learning (CL)\nloss. We utilize TE and SE to combine the strengths of previous methods in a\nsimplistic manner to efficiently capture temporal and spatial contextual\ninformation in the conversation. To ease the harmful influence resulting from\nemotion shift and simulate the way humans learn curriculum from easy to hard,\nwe apply the idea of CL to the ERC task to progressively optimize the network\nparameters. At the beginning of training, we assign lower learning weights to\ndifficult samples. As the epoch increases, the learning weights for these\nsamples are gradually raised. Extensive experiments on four datasets exhibit\nthat our proposed method is effective and dramatically beats other baseline\nmodels.\n", "date": "2023-08-12", "categories": []}
{"id": "2401.13697", "title": "Toward Robust Multimodal Learning using Multimodal Foundational Models", "abstract": "  Existing multimodal sentiment analysis tasks are highly rely on the\nassumption that the training and test sets are complete multimodal data, while\nthis assumption can be difficult to hold: the multimodal data are often\nincomplete in real-world scenarios. Therefore, a robust multimodal model in\nscenarios with randomly missing modalities is highly preferred. Recently,\nCLIP-based multimodal foundational models have demonstrated impressive\nperformance on numerous multimodal tasks by learning the aligned cross-modal\nsemantics of image and text pairs, but the multimodal foundational models are\nalso unable to directly address scenarios involving modality absence. To\nalleviate this issue, we propose a simple and effective framework, namely TRML,\nToward Robust Multimodal Learning using Multimodal Foundational Models. TRML\nemploys generated virtual modalities to replace missing modalities, and aligns\nthe semantic spaces between the generated and missing modalities. Concretely,\nwe design a missing modality inference module to generate virtual modaliites\nand replace missing modalities. We also design a semantic matching learning\nmodule to align semantic spaces generated and missing modalities. Under the\nprompt of complete modality, our model captures the semantics of missing\nmodalities by leveraging the aligned cross-modal semantic space. Experiments\ndemonstrate the superiority of our approach on three multimodal sentiment\nanalysis benchmark datasets, CMU-MOSI, CMU-MOSEI, and MELD.\n", "date": "2024-01-20", "categories": []}
{"id": "2401.13722", "title": "Proactive Emotion Tracker: AI-Driven Continuous Mood and Emotion\n  Monitoring", "abstract": "  This research project aims to tackle the growing mental health challenges in\ntoday's digital age. It employs a modified pre-trained BERT model to detect\ndepressive text within social media and users' web browsing data, achieving an\nimpressive 93% test accuracy. Simultaneously, the project aims to incorporate\nphysiological signals from wearable devices, such as smartwatches and EEG\nsensors, to provide long-term tracking and prognosis of mood disorders and\nemotional states. This comprehensive approach holds promise for enhancing early\ndetection of depression and advancing overall mental health outcomes.\n", "date": "2024-01-24", "categories": []}
{"id": "2308.03022", "title": "SAPIEN: Affective Virtual Agents Powered by Large Language Models", "abstract": "  In this demo paper, we introduce SAPIEN, a platform for high-fidelity virtual\nagents driven by large language models that can hold open domain conversations\nwith users in 13 different languages, and display emotions through facial\nexpressions and voice. The platform allows users to customize their virtual\nagent's personality, background, and conversation premise, thus providing a\nrich, immersive interaction experience. Furthermore, after the virtual meeting,\nthe user can choose to get the conversation analyzed and receive actionable\nfeedback on their communication skills. This paper illustrates an overview of\nthe platform and discusses the various application domains of this technology,\nranging from entertainment to mental health, communication training, language\nlearning, education, healthcare, and beyond. Additionally, we consider the\nethical implications of such realistic virtual agent representations and the\npotential challenges in ensuring responsible use.\n", "date": "2023-08-06", "categories": []}
{"id": "2401.10946", "title": "Self context-aware emotion perception on human-robot interaction", "abstract": "  Emotion recognition plays a crucial role in various domains of human-robot\ninteraction. In long-term interactions with humans, robots need to respond\ncontinuously and accurately, however, the mainstream emotion recognition\nmethods mostly focus on short-term emotion recognition, disregarding the\ncontext in which emotions are perceived. Humans consider that contextual\ninformation and different contexts can lead to completely different emotional\nexpressions. In this paper, we introduce self context-aware model (SCAM) that\nemploys a two-dimensional emotion coordinate system for anchoring and\nre-labeling distinct emotions. Simultaneously, it incorporates its distinctive\ninformation retention structure and contextual loss. This approach has yielded\nsignificant improvements across audio, video, and multimodal. In the auditory\nmodality, there has been a notable enhancement in accuracy, rising from 63.10%\nto 72.46%. Similarly, the visual modality has demonstrated improved accuracy,\nincreasing from 77.03% to 80.82%. In the multimodal, accuracy has experienced\nan elevation from 77.48% to 78.93%. In the future, we will validate the\nreliability and usability of SCAM on robots through psychology experiments.\n", "date": "2024-01-18", "categories": []}
{"id": "2309.00138", "title": "Fuzzy Approach for Audio-Video Emotion Recognition in Computer Games for\n  Children", "abstract": "  Computer games are widespread nowadays and enjoyed by people of all ages. But\nwhen it comes to kids, playing these games can be more than just fun, it is a\nway for them to develop important skills and build emotional intelligence.\nFacial expressions and sounds that kids produce during gameplay reflect their\nfeelings, thoughts, and moods. In this paper, we propose a novel framework that\nintegrates a fuzzy approach for the recognition of emotions through the\nanalysis of audio and video data. Our focus lies within the specific context of\ncomputer games tailored for children, aiming to enhance their overall user\nexperience. We use the FER dataset to detect facial emotions in video frames\nrecorded from the screen during the game. For the audio emotion recognition of\nsounds a kid produces during the game, we use CREMA-D, TESS, RAVDESS, and Savee\ndatasets. Next, a fuzzy inference system is used for the fusion of results.\nBesides this, our system can detect emotion stability and emotion diversity\nduring gameplay, which, together with prevailing emotion report, can serve as\nvaluable information for parents worrying about the effect of certain games on\ntheir kids. The proposed approach has shown promising results in the\npreliminary experiments we conducted, involving 3 different video games, namely\nfighting, racing, and logic games, and providing emotion-tracking results for\nkids in each game. Our study can contribute to the advancement of\nchild-oriented game development, which is not only engaging but also accounts\nfor children's cognitive and emotional states.\n", "date": "2023-08-31", "categories": []}
{"id": "2401.10536", "title": "Speech Swin-Transformer: Exploring a Hierarchical Transformer with\n  Shifted Windows for Speech Emotion Recognition", "abstract": "  Swin-Transformer has demonstrated remarkable success in computer vision by\nleveraging its hierarchical feature representation based on Transformer. In\nspeech signals, emotional information is distributed across different scales of\nspeech features, e.\\,g., word, phrase, and utterance. Drawing above\ninspiration, this paper presents a hierarchical speech Transformer with shifted\nwindows to aggregate multi-scale emotion features for speech emotion\nrecognition (SER), called Speech Swin-Transformer. Specifically, we first\ndivide the speech spectrogram into segment-level patches in the time domain,\ncomposed of multiple frame patches. These segment-level patches are then\nencoded using a stack of Swin blocks, in which a local window Transformer is\nutilized to explore local inter-frame emotional information across frame\npatches of each segment patch. After that, we also design a shifted window\nTransformer to compensate for patch correlations near the boundaries of segment\npatches. Finally, we employ a patch merging operation to aggregate\nsegment-level emotional features for hierarchical speech representation by\nexpanding the receptive field of Transformer from frame-level to segment-level.\nExperimental results demonstrate that our proposed Speech Swin-Transformer\noutperforms the state-of-the-art methods.\n", "date": "2024-01-19", "categories": []}
{"id": "2310.17212", "title": "Affective Video Content Analysis: Decade Review and New Perspectives", "abstract": "  Video content is rich in semantics and has the ability to evoke various\nemotions in viewers. In recent years, with the rapid development of affective\ncomputing and the explosive growth of visual data, affective video content\nanalysis (AVCA) as an essential branch of affective computing has become a\nwidely researched topic. In this study, we comprehensively review the\ndevelopment of AVCA over the past decade, particularly focusing on the most\nadvanced methods adopted to address the three major challenges of video feature\nextraction, expression subjectivity, and multimodal feature fusion. We first\nintroduce the widely used emotion representation models in AVCA and describe\ncommonly used datasets. We summarize and compare representative methods in the\nfollowing aspects: (1) unimodal AVCA models, including facial expression\nrecognition and posture emotion recognition; (2) multimodal AVCA models,\nincluding feature fusion, decision fusion, and attention-based multimodal\nmodels; (3) model performance evaluation standards. Finally, we discuss future\nchallenges and promising research directions, such as emotion recognition and\npublic opinion analysis, human-computer interaction, and emotional\nintelligence.\n", "date": "2023-10-26", "categories": []}
{"id": "2312.10201", "title": "CARAT: Contrastive Feature Reconstruction and Aggregation for\n  Multi-Modal Multi-Label Emotion Recognition", "abstract": "  Multi-modal multi-label emotion recognition (MMER) aims to identify relevant\nemotions from multiple modalities. The challenge of MMER is how to effectively\ncapture discriminative features for multiple labels from heterogeneous data.\nRecent studies are mainly devoted to exploring various fusion strategies to\nintegrate multi-modal information into a unified representation for all labels.\nHowever, such a learning scheme not only overlooks the specificity of each\nmodality but also fails to capture individual discriminative features for\ndifferent labels. Moreover, dependencies of labels and modalities cannot be\neffectively modeled. To address these issues, this paper presents ContrAstive\nfeature Reconstruction and AggregaTion (CARAT) for the MMER task. Specifically,\nwe devise a reconstruction-based fusion mechanism to better model fine-grained\nmodality-to-label dependencies by contrastively learning modal-separated and\nlabel-specific features. To further exploit the modality complementarity, we\nintroduce a shuffle-based aggregation strategy to enrich co-occurrence\ncollaboration among labels. Experiments on two benchmark datasets CMU-MOSEI and\nM3ED demonstrate the effectiveness of CARAT over state-of-the-art methods. Code\nis available at https://github.com/chengzju/CARAT.\n", "date": "2023-12-15", "categories": []}
{"id": "2401.06957", "title": "EVOKE: Emotion Enabled Virtual Avatar Mapping Using Optimized Knowledge\n  Distillation", "abstract": "  As virtual environments continue to advance, the demand for immersive and\nemotionally engaging experiences has grown. Addressing this demand, we\nintroduce Emotion enabled Virtual avatar mapping using Optimized KnowledgE\ndistillation (EVOKE), a lightweight emotion recognition framework designed for\nthe seamless integration of emotion recognition into 3D avatars within virtual\nenvironments. Our approach leverages knowledge distillation involving\nmulti-label classification on the publicly available DEAP dataset, which covers\nvalence, arousal, and dominance as primary emotional classes. Remarkably, our\ndistilled model, a CNN with only two convolutional layers and 18 times fewer\nparameters than the teacher model, achieves competitive results, boasting an\naccuracy of 87% while demanding far less computational resources. This\nequilibrium between performance and deployability positions our framework as an\nideal choice for virtual environment systems. Furthermore, the multi-label\nclassification outcomes are utilized to map emotions onto custom-designed 3D\navatars.\n", "date": "2024-01-13", "categories": []}
{"id": "2401.07336", "title": "Construction and Evaluation of Mandarin Multimodal Emotional Speech\n  Database", "abstract": "  A multi-modal emotional speech Mandarin database including articulatory\nkinematics, acoustics, glottal and facial micro-expressions is designed and\nestablished, which is described in detail from the aspects of corpus design,\nsubject selection, recording details and data processing. Where signals are\nlabeled with discrete emotion labels (neutral, happy, pleasant, indifferent,\nangry, sad, grief) and dimensional emotion labels (pleasure, arousal,\ndominance). In this paper, the validity of dimension annotation is verified by\nstatistical analysis of dimension annotation data. The SCL-90 scale data of\nannotators are verified and combined with PAD annotation data for analysis, so\nas to explore the internal relationship between the outlier phenomenon in\nannotation and the psychological state of annotators. In order to verify the\nspeech quality and emotion discrimination of the database, this paper uses 3\nbasic models of SVM, CNN and DNN to calculate the recognition rate of these\nseven emotions. The results show that the average recognition rate of seven\nemotions is about 82% when using acoustic data alone. When using glottal data\nalone, the average recognition rate is about 72%. Using kinematics data alone,\nthe average recognition rate also reaches 55.7%. Therefore, the database is of\nhigh quality and can be used as an important source for speech analysis\nresearch, especially for the task of multimodal emotional speech analysis.\n", "date": "2024-01-14", "categories": []}
{"id": "2401.07414", "title": "Leveraging the power of transformers for guilt detection in text", "abstract": "  In recent years, language models and deep learning techniques have\nrevolutionized natural language processing tasks, including emotion detection.\nHowever, the specific emotion of guilt has received limited attention in this\nfield. In this research, we explore the applicability of three\ntransformer-based language models for detecting guilt in text and compare their\nperformance for general emotion detection and guilt detection. Our proposed\nmodel outformed BERT and RoBERTa models by two and one points respectively.\nAdditionally, we analyze the challenges in developing accurate guilt-detection\nmodels and evaluate our model's effectiveness in detecting related emotions\nlike \"shame\" through qualitative analysis of results.\n", "date": "2024-01-15", "categories": []}
{"id": "2401.05166", "title": "REACT 2024: the Second Multiple Appropriate Facial Reaction Generation\n  Challenge", "abstract": "  In dyadic interactions, humans communicate their intentions and state of mind\nusing verbal and non-verbal cues, where multiple different facial reactions\nmight be appropriate in response to a specific speaker behaviour. Then, how to\ndevelop a machine learning (ML) model that can automatically generate multiple\nappropriate, diverse, realistic and synchronised human facial reactions from an\npreviously unseen speaker behaviour is a challenging task. Following the\nsuccessful organisation of the first REACT challenge (REACT 2023), this edition\nof the challenge (REACT 2024) employs a subset used by the previous challenge,\nwhich contains segmented 30-secs dyadic interaction clips originally recorded\nas part of the NOXI and RECOLA datasets, encouraging participants to develop\nand benchmark Machine Learning (ML) models that can generate multiple\nappropriate facial reactions (including facial image sequences and their\nattributes) given an input conversational partner's stimulus under various\ndyadic video conference scenarios. This paper presents: (i) the guidelines of\nthe REACT 2024 challenge; (ii) the dataset utilized in the challenge; and (iii)\nthe performance of the baseline systems on the two proposed sub-challenges:\nOffline Multiple Appropriate Facial Reaction Generation and Online Multiple\nAppropriate Facial Reaction Generation, respectively. The challenge baseline\ncode is publicly available at\nhttps://github.com/reactmultimodalchallenge/baseline_react2024.\n", "date": "2024-01-10", "categories": []}
{"id": "2304.06910", "title": "HCAM -- Hierarchical Cross Attention Model for Multi-modal Emotion\n  Recognition", "abstract": "  Emotion recognition in conversations is challenging due to the multi-modal\nnature of the emotion expression. We propose a hierarchical cross-attention\nmodel (HCAM) approach to multi-modal emotion recognition using a combination of\nrecurrent and co-attention neural network models. The input to the model\nconsists of two modalities, i) audio data, processed through a learnable\nwav2vec approach and, ii) text data represented using a bidirectional encoder\nrepresentations from transformers (BERT) model. The audio and text\nrepresentations are processed using a set of bi-directional recurrent neural\nnetwork layers with self-attention that converts each utterance in a given\nconversation to a fixed dimensional embedding. In order to incorporate\ncontextual knowledge and the information across the two modalities, the audio\nand text embeddings are combined using a co-attention layer that attempts to\nweigh the utterance level embeddings relevant to the task of emotion\nrecognition. The neural network parameters in the audio layers, text layers as\nwell as the multi-modal co-attention layers, are hierarchically trained for the\nemotion classification task. We perform experiments on three established\ndatasets namely, IEMOCAP, MELD and CMU-MOSI, where we illustrate that the\nproposed model improves significantly over other benchmarks and helps achieve\nstate-of-art results on all these datasets.\n", "date": "2023-04-14", "categories": []}
{"id": "2401.04511", "title": "Zero Shot Audio to Audio Emotion Transfer With Speaker Disentanglement", "abstract": "  The problem of audio-to-audio (A2A) style transfer involves replacing the\nstyle features of the source audio with those from the target audio while\npreserving the content related attributes of the source audio. In this paper,\nwe propose an efficient approach, termed as Zero-shot Emotion Style Transfer\n(ZEST), that allows the transfer of emotional content present in the given\nsource audio with the one embedded in the target audio while retaining the\nspeaker and speech content from the source. The proposed system builds upon\ndecomposing speech into semantic tokens, speaker representations and emotion\nembeddings. Using these factors, we propose a framework to reconstruct the\npitch contour of the given speech signal and train a decoder that reconstructs\nthe speech signal. The model is trained using a self-supervision based\nreconstruction loss. During conversion, the emotion embedding is alone derived\nfrom the target audio, while rest of the factors are derived from the source\naudio. In our experiments, we show that, even without using parallel training\ndata or labels from the source or target audio, we illustrate zero shot emotion\ntransfer capabilities of the proposed ZEST model using objective and subjective\nquality evaluations.\n", "date": "2024-01-09", "categories": []}
{"id": "2312.07507", "title": "NAC-TCN: Temporal Convolutional Networks with Causal Dilated\n  Neighborhood Attention for Emotion Understanding", "abstract": "  In the task of emotion recognition from videos, a key improvement has been to\nfocus on emotions over time rather than a single frame. There are many\narchitectures to address this task such as GRUs, LSTMs, Self-Attention,\nTransformers, and Temporal Convolutional Networks (TCNs). However, these\nmethods suffer from high memory usage, large amounts of operations, or poor\ngradients. We propose a method known as Neighborhood Attention with\nConvolutions TCN (NAC-TCN) which incorporates the benefits of attention and\nTemporal Convolutional Networks while ensuring that causal relationships are\nunderstood which results in a reduction in computation and memory cost. We\naccomplish this by introducing a causal version of Dilated Neighborhood\nAttention while incorporating it with convolutions. Our model achieves\ncomparable, better, or state-of-the-art performance over TCNs, TCAN, LSTMs, and\nGRUs while requiring fewer parameters on standard emotion recognition datasets.\nWe publish our code online for easy reproducibility and use in other projects.\n", "date": "2023-12-12", "categories": []}
{"id": "2401.02995", "title": "CANAMRF: An Attention-Based Model for Multimodal Depression Detection", "abstract": "  Multimodal depression detection is an important research topic that aims to\npredict human mental states using multimodal data. Previous methods treat\ndifferent modalities equally and fuse each modality by na\\\"ive mathematical\noperations without measuring the relative importance between them, which cannot\nobtain well-performed multimodal representations for downstream depression\ntasks. In order to tackle the aforementioned concern, we present a Cross-modal\nAttention Network with Adaptive Multi-modal Recurrent Fusion (CANAMRF) for\nmultimodal depression detection. CANAMRF is constructed by a multimodal feature\nextractor, an Adaptive Multimodal Recurrent Fusion module, and a Hybrid\nAttention Module. Through experimentation on two benchmark datasets, CANAMRF\ndemonstrates state-of-the-art performance, underscoring the effectiveness of\nour proposed approach.\n", "date": "2024-01-04", "categories": []}
{"id": "2401.03000", "title": "Bridging Modalities: Knowledge Distillation and Masked Training for\n  Translating Multi-Modal Emotion Recognition to Uni-Modal, Speech-Only Emotion\n  Recognition", "abstract": "  This paper presents an innovative approach to address the challenges of\ntranslating multi-modal emotion recognition models to a more practical and\nresource-efficient uni-modal counterpart, specifically focusing on speech-only\nemotion recognition. Recognizing emotions from speech signals is a critical\ntask with applications in human-computer interaction, affective computing, and\nmental health assessment. However, existing state-of-the-art models often rely\non multi-modal inputs, incorporating information from multiple sources such as\nfacial expressions and gestures, which may not be readily available or feasible\nin real-world scenarios. To tackle this issue, we propose a novel framework\nthat leverages knowledge distillation and masked training techniques.\n", "date": "2024-01-04", "categories": []}
{"id": "2401.02746", "title": "Reading Between the Frames: Multi-Modal Depression Detection in Videos\n  from Non-Verbal Cues", "abstract": "  Depression, a prominent contributor to global disability, affects a\nsubstantial portion of the population. Efforts to detect depression from social\nmedia texts have been prevalent, yet only a few works explored depression\ndetection from user-generated video content. In this work, we address this\nresearch gap by proposing a simple and flexible multi-modal temporal model\ncapable of discerning non-verbal depression cues from diverse modalities in\nnoisy, real-world videos. We show that, for in-the-wild videos, using\nadditional high-level non-verbal cues is crucial to achieving good performance,\nand we extracted and processed audio speech embeddings, face emotion\nembeddings, face, body and hand landmarks, and gaze and blinking information.\nThrough extensive experiments, we show that our model achieves state-of-the-art\nresults on three key benchmark datasets for depression detection from video by\na substantial margin. Our code is publicly available on GitHub.\n", "date": "2024-01-05", "categories": []}
{"id": "2401.02344", "title": "Multi-Source Domain Adaptation with Transformer-based Feature Generation\n  for Subject-Independent EEG-based Emotion Recognition", "abstract": "  Although deep learning-based algorithms have demonstrated excellent\nperformance in automated emotion recognition via electroencephalogram (EEG)\nsignals, variations across brain signal patterns of individuals can diminish\nthe model's effectiveness when applied across different subjects. While\ntransfer learning techniques have exhibited promising outcomes, they still\nencounter challenges related to inadequate feature representations and may\noverlook the fact that source subjects themselves can possess distinct\ncharacteristics. In this work, we propose a multi-source domain adaptation\napproach with a transformer-based feature generator (MSDA-TF) designed to\nleverage information from multiple sources. The proposed feature generator\nretains convolutional layers to capture shallow spatial, temporal, and spectral\nEEG data representations, while self-attention mechanisms extract global\ndependencies within these features. During the adaptation process, we group the\nsource subjects based on correlation values and aim to align the moments of the\ntarget subject with each source as well as within the sources. MSDA-TF is\nvalidated on the SEED dataset and is shown to yield promising results.\n", "date": "2024-01-04", "categories": []}
{"id": "2401.01495", "title": "A Two-Stage Multimodal Emotion Recognition Model Based on Graph\n  Contrastive Learning", "abstract": "  In terms of human-computer interaction, it is becoming more and more\nimportant to correctly understand the user's emotional state in a conversation,\nso the task of multimodal emotion recognition (MER) started to receive more\nattention. However, existing emotion classification methods usually perform\nclassification only once. Sentences are likely to be misclassified in a single\nround of classification. Previous work usually ignores the similarities and\ndifferences between different morphological features in the fusion process. To\naddress the above issues, we propose a two-stage emotion recognition model\nbased on graph contrastive learning (TS-GCL). First, we encode the original\ndataset with different preprocessing modalities. Second, a graph contrastive\nlearning (GCL) strategy is introduced for these three modal data with other\nstructures to learn similarities and differences within and between modalities.\nFinally, we use MLP twice to achieve the final emotion classification. This\nstaged classification method can help the model to better focus on different\nlevels of emotional information, thereby improving the performance of the\nmodel. Extensive experiments show that TS-GCL has superior performance on\nIEMOCAP and MELD datasets compared with previous methods.\n", "date": "2024-01-03", "categories": []}
{"id": "2211.08843", "title": "Improving Speech Emotion Recognition with Unsupervised Speaking Style\n  Transfer", "abstract": "  Humans can effortlessly modify various prosodic attributes, such as the\nplacement of stress and the intensity of sentiment, to convey a specific\nemotion while maintaining consistent linguistic content. Motivated by this\ncapability, we propose EmoAug, a novel style transfer model designed to enhance\nemotional expression and tackle the data scarcity issue in speech emotion\nrecognition tasks. EmoAug consists of a semantic encoder and a paralinguistic\nencoder that represent verbal and non-verbal information respectively.\nAdditionally, a decoder reconstructs speech signals by conditioning on the\naforementioned two information flows in an unsupervised fashion. Once training\nis completed, EmoAug enriches expressions of emotional speech with different\nprosodic attributes, such as stress, rhythm and intensity, by feeding different\nstyles into the paralinguistic encoder. EmoAug enables us to generate similar\nnumbers of samples for each class to tackle the data imbalance issue as well.\nExperimental results on the IEMOCAP dataset demonstrate that EmoAug can\nsuccessfully transfer different speaking styles while retaining the speaker\nidentity and semantic content. Furthermore, we train a SER model with data\naugmented by EmoAug and show that the augmented model not only surpasses the\nstate-of-the-art supervised and self-supervised methods but also overcomes\noverfitting problems caused by data imbalance. Some audio samples can be found\non our demo website.\n", "date": "2022-11-16", "categories": []}
{"id": "2312.16180", "title": "Investigating salient representations and label Variance in Dimensional\n  Speech Emotion Analysis", "abstract": "  Representations derived from models such as BERT (Bidirectional Encoder\nRepresentations from Transformers) and HuBERT (Hidden units BERT), have helped\nto achieve state-of-the-art performance in dimensional speech emotion\nrecognition. Despite their large dimensionality, and even though these\nrepresentations are not tailored for emotion recognition tasks, they are\nfrequently used to train large speech emotion models with high memory and\ncomputational costs. In this work, we show that there exist lower-dimensional\nsubspaces within the these pre-trained representational spaces that offer a\nreduction in downstream model complexity without sacrificing performance on\nemotion estimation. In addition, we model label uncertainty in the form of\ngrader opinion variance, and demonstrate that such information can improve the\nmodels generalization capacity and robustness. Finally, we compare the\nrobustness of the emotion models against acoustic degradations and observed\nthat the reduced dimensional representations were able to retain the\nperformance similar to the full-dimensional representations without significant\nregression in dimensional emotion performance.\n", "date": "2023-12-17", "categories": []}
{"id": "2312.16383", "title": "Frame-level emotional state alignment method for speech emotion\n  recognition", "abstract": "  Speech emotion recognition (SER) systems aim to recognize human emotional\nstate during human-computer interaction. Most existing SER systems are trained\nbased on utterance-level labels. However, not all frames in an audio have\naffective states consistent with utterance-level label, which makes it\ndifficult for the model to distinguish the true emotion of the audio and\nperform poorly. To address this problem, we propose a frame-level emotional\nstate alignment method for SER. First, we fine-tune HuBERT model to obtain a\nSER system with task-adaptive pretraining (TAPT) method, and extract embeddings\nfrom its transformer layers to form frame-level pseudo-emotion labels with\nclustering. Then, the pseudo labels are used to pretrain HuBERT. Hence, the\neach frame output of HuBERT has corresponding emotional information. Finally,\nwe fine-tune the above pretrained HuBERT for SER by adding an attention layer\non the top of it, which can focus only on those frames that are emotionally\nmore consistent with utterance-level label. The experimental results performed\non IEMOCAP indicate that our proposed method performs better than\nstate-of-the-art (SOTA) methods.\n", "date": "2023-12-27", "categories": []}
{"id": "2305.19379", "title": "Inter Subject Emotion Recognition Using Spatio-Temporal Features From\n  EEG Signal", "abstract": "  Inter-subject or subject-independent emotion recognition has been a\nchallenging task in affective computing. This work is about an\neasy-to-implement emotion recognition model that classifies emotions from EEG\nsignals subject independently. It is based on the famous EEGNet architecture,\nwhich is used in EEG-related BCIs. We used the Dataset on Emotion using\nNaturalistic Stimuli (DENS) dataset. The dataset contains the Emotional Events\n-- the precise information of the emotion timings that participants felt. The\nmodel is a combination of regular, depthwise and separable convolution layers\nof CNN to classify the emotions. The model has the capacity to learn the\nspatial features of the EEG channels and the temporal features of the EEG\nsignals variability with time. The model is evaluated for the valence space\nratings. The model achieved an accuracy of 73.04%.\n", "date": "2023-05-27", "categories": []}
{"id": "2310.09596", "title": "RethinkingTMSC: An Empirical Study for Target-Oriented Multimodal\n  Sentiment Classification", "abstract": "  Recently, Target-oriented Multimodal Sentiment Classification (TMSC) has\ngained significant attention among scholars. However, current multimodal models\nhave reached a performance bottleneck. To investigate the causes of this\nproblem, we perform extensive empirical evaluation and in-depth analysis of the\ndatasets to answer the following questions: Q1: Are the modalities equally\nimportant for TMSC? Q2: Which multimodal fusion modules are more effective? Q3:\nDo existing datasets adequately support the research? Our experiments and\nanalyses reveal that the current TMSC systems primarily rely on the textual\nmodality, as most of targets' sentiments can be determined solely by text.\nConsequently, we point out several directions to work on for the TMSC task in\nterms of model design and dataset construction. The code and data can be found\nin https://github.com/Junjie-Ye/RethinkingTMSC.\n", "date": "2023-10-14", "categories": []}
{"id": "2312.15185", "title": "emotion2vec: Self-Supervised Pre-Training for Speech Emotion\n  Representation", "abstract": "  We propose emotion2vec, a universal speech emotion representation model.\nemotion2vec is pre-trained on open-source unlabeled emotion data through\nself-supervised online distillation, combining utterance-level loss and\nframe-level loss during pre-training. emotion2vec outperforms state-of-the-art\npre-trained universal models and emotion specialist models by only training\nlinear layers for the speech emotion recognition task on the mainstream IEMOCAP\ndataset. In addition, emotion2vec shows consistent improvements among 10\ndifferent languages of speech emotion recognition datasets. emotion2vec also\nshows excellent results on other emotion tasks, such as song emotion\nrecognition, emotion prediction in conversation, and sentiment analysis.\nComparison experiments, ablation experiments, and visualization comprehensively\ndemonstrate the universal capability of the proposed emotion2vec. To the best\nof our knowledge, emotion2vec is the first universal representation model in\nvarious emotion-related tasks, filling a gap in the field.\n", "date": "2023-12-23", "categories": []}
{"id": "2312.15593", "title": "DSNet: Disentangled Siamese Network with Neutral Calibration for Speech\n  Emotion Recognition", "abstract": "  One persistent challenge in deep learning based speech emotion recognition\n(SER) is the unconscious encoding of emotion-irrelevant factors (e.g., speaker\nor phonetic variability), which limits the generalization of SER in practical\nuse. In this paper, we propose DSNet, a Disentangled Siamese Network with\nneutral calibration, to meet the demand for a more robust and explainable SER\nmodel. Specifically, we introduce an orthogonal feature disentanglement module\nto explicitly project the high-level representation into two distinct\nsubspaces. Later, we propose a novel neutral calibration mechanism to encourage\none subspace to capture sufficient emotion-irrelevant information. In this way,\nthe other one can better isolate and emphasize the emotion-relevant information\nwithin speech signals. Experimental results on two popular benchmark datasets\ndemonstrate the superiority of DSNet over various state-of-the-art methods for\nspeaker-independent SER.\n", "date": "2023-12-25", "categories": []}
{"id": "2312.15848", "title": "Modality-Collaborative Transformer with Hybrid Feature Reconstruction\n  for Robust Emotion Recognition", "abstract": "  As a vital aspect of affective computing, Multimodal Emotion Recognition has\nbeen an active research area in the multimedia community. Despite recent\nprogress, this field still confronts two major challenges in real-world\napplications: 1) improving the efficiency of constructing joint representations\nfrom unaligned multimodal features, and 2) relieving the performance decline\ncaused by random modality feature missing. In this paper, we propose a unified\nframework, Modality-Collaborative Transformer with Hybrid Feature\nReconstruction (MCT-HFR), to address these issues. The crucial component of MCT\nis a novel attention-based encoder which concurrently extracts and dynamically\nbalances the intra- and inter-modality relations for all associated modalities.\nWith additional modality-wise parameter sharing, a more compact representation\ncan be encoded with less time and space complexity. To improve the robustness\nof MCT, we further introduce HFR which consists of two modules: Local Feature\nImagination (LFI) and Global Feature Alignment (GFA). During model training,\nLFI leverages complete features as supervisory signals to recover local missing\nfeatures, while GFA is designed to reduce the global semantic gap between\npairwise complete and incomplete representations. Experimental evaluations on\ntwo popular benchmark datasets demonstrate that our proposed method\nconsistently outperforms advanced baselines in both complete and incomplete\ndata scenarios.\n", "date": "2023-12-26", "categories": []}
{"id": "2312.16036", "title": "Ensemble Learning to Assess Dynamics of Affective Experience Ratings and\n  Physiological Change", "abstract": "  The congruence between affective experiences and physiological changes has\nbeen a debated topic for centuries. Recent technological advances in\nmeasurement and data analysis provide hope to solve this epic challenge. Open\nscience and open data practices, together with data analysis challenges open to\nthe academic community, are also promising tools for solving this problem. In\nthis entry to the Emotion Physiology and Experience Collaboration (EPiC)\nchallenge, we propose a data analysis solution that combines theoretical\nassumptions with data-driven methodologies. We used feature engineering and\nensemble selection. Each predictor was trained on subsets of the training data\nthat would maximize the information available for training. Late fusion was\nused with an averaging step. We chose to average considering a ``wisdom of\ncrowds'' strategy. This strategy yielded an overall RMSE of 1.19 in the test\nset. Future work should carefully explore if our assumptions are correct and\nthe potential of weighted fusion.\n", "date": "2023-12-26", "categories": []}
{"id": "2312.11503", "title": "Speech and Text-Based Emotion Recognizer", "abstract": "  Affective computing is a field of study that focuses on developing systems\nand technologies that can understand, interpret, and respond to human emotions.\nSpeech Emotion Recognition (SER), in particular, has got a lot of attention\nfrom researchers in the recent past. However, in many cases, the publicly\navailable datasets, used for training and evaluation, are scarce and imbalanced\nacross the emotion labels. In this work, we focused on building a balanced\ncorpus from these publicly available datasets by combining these datasets as\nwell as employing various speech data augmentation techniques. Furthermore, we\nexperimented with different architectures for speech emotion recognition. Our\nbest system, a multi-modal speech, and text-based model, provides a performance\nof UA(Unweighed Accuracy) + WA (Weighed Accuracy) of 157.57 compared to the\nbaseline algorithm performance of 119.66\n", "date": "2023-12-10", "categories": []}
{"id": "2312.11542", "title": "FER-C: Benchmarking Out-of-Distribution Soft Calibration for Facial\n  Expression Recognition", "abstract": "  We present a soft benchmark for calibrating facial expression recognition\n(FER). While prior works have focused on identifying affective states, we find\nthat FER models are uncalibrated. This is particularly true when\nout-of-distribution (OOD) shifts further exacerbate the ambiguity of facial\nexpressions. While most OOD benchmarks provide hard labels, we argue that the\nground-truth labels for evaluating FER models should be soft in order to better\nreflect the ambiguity behind facial behaviours. Our framework proposes soft\nlabels that closely approximates the average information loss based on\ndifferent types of OOD shifts. Finally, we show the benefits of calibration on\nfive state-of-the-art FER algorithms tested on our benchmark.\n", "date": "2023-12-16", "categories": []}
{"id": "2312.12334", "title": "PowMix: A Versatile Regularizer for Multimodal Sentiment Analysis", "abstract": "  Multimodal sentiment analysis (MSA) leverages heterogeneous data sources to\ninterpret the complex nature of human sentiments. Despite significant progress\nin multimodal architecture design, the field lacks comprehensive regularization\nmethods. This paper introduces PowMix, a versatile embedding space regularizer\nthat builds upon the strengths of unimodal mixing-based regularization\napproaches and introduces novel algorithmic components that are specifically\ntailored to multimodal tasks. PowMix is integrated before the fusion stage of\nmultimodal architectures and facilitates intra-modal mixing, such as mixing\ntext with text, to act as a regularizer. PowMix consists of five components: 1)\na varying number of generated mixed examples, 2) mixing factor reweighting, 3)\nanisotropic mixing, 4) dynamic mixing, and 5) cross-modal label mixing.\nExtensive experimentation across benchmark MSA datasets and a broad spectrum of\ndiverse architectural designs demonstrate the efficacy of PowMix, as evidenced\nby consistent performance improvements over baselines and existing mixing\nmethods. An in-depth ablation study highlights the critical contribution of\neach PowMix component and how they synergistically enhance performance.\nFurthermore, algorithmic analysis demonstrates how PowMix behaves in different\nscenarios, particularly comparing early versus late fusion architectures.\nNotably, PowMix enhances overall performance without sacrificing model\nrobustness or magnifying text dominance. It also retains its strong performance\nin situations of limited data. Our findings position PowMix as a promising\nversatile regularization strategy for MSA. Code will be made available.\n", "date": "2023-12-19", "categories": []}
{"id": "2312.10818", "title": "Facial Emotion Recognition using CNN in PyTorch", "abstract": "  In this project, we have implemented a model to recognize real-time facial\nemotions given the camera images. Current approaches would read all data and\ninput it into their model, which has high space complexity. Our model is based\non the Convolutional Neural Network utilizing the PyTorch library. We believe\nour implementation will significantly improve the space complexity and provide\na useful contribution to facial emotion recognition. Our motivation is to\nunderstanding clearly about deep learning, particularly in CNNs, and analysis\nreal-life scenarios. Therefore, we tunned the hyper parameter of model such as\nlearning rate, batch size, and number of epochs to meet our needs. In addition,\nwe also used techniques to optimize the networks, such as activation function,\ndropout and max pooling. Finally, we analyzed the result from two optimizer to\nobserve the relationship between number of epochs and accuracy.\n", "date": "2023-12-17", "categories": []}
{"id": "2312.10937", "title": "An Extended Variational Mode Decomposition Algorithm Developed Speech\n  Emotion Recognition Performance", "abstract": "  Emotion recognition (ER) from speech signals is a robust approach since it\ncannot be imitated like facial expression or text based sentiment analysis.\nValuable information underlying the emotions are significant for human-computer\ninteractions enabling intelligent machines to interact with sensitivity in the\nreal world. Previous ER studies through speech signal processing have focused\nexclusively on associations between different signal mode decomposition methods\nand hidden informative features. However, improper decomposition parameter\nselections lead to informative signal component losses due to mode duplicating\nand mixing. In contrast, the current study proposes VGG-optiVMD, an empowered\nvariational mode decomposition algorithm, to distinguish meaningful speech\nfeatures and automatically select the number of decomposed modes and optimum\nbalancing parameter for the data fidelity constraint by assessing their effects\non the VGG16 flattening output layer. Various feature vectors were employed to\ntrain the VGG16 network on different databases and assess VGG-optiVMD\nreproducibility and reliability. One, two, and three-dimensional feature\nvectors were constructed by concatenating Mel-frequency cepstral coefficients,\nChromagram, Mel spectrograms, Tonnetz diagrams, and spectral centroids. Results\nconfirmed a synergistic relationship between the fine-tuning of the signal\nsample rate and decomposition parameters with classification accuracy,\nachieving state-of-the-art 96.09% accuracy in predicting seven emotions on the\nBerlin EMO-DB database.\n", "date": "2023-12-18", "categories": []}
{"id": "2312.10949", "title": "Leveraged Mel spectrograms using Harmonic and Percussive Components in\n  Speech Emotion Recognition", "abstract": "  Speech Emotion Recognition (SER) affective technology enables the intelligent\nembedded devices to interact with sensitivity. Similarly, call centre employees\nrecognise customers' emotions from their pitch, energy, and tone of voice so as\nto modify their speech for a high-quality interaction with customers. This work\nexplores, for the first time, the effects of the harmonic and percussive\ncomponents of Mel spectrograms in SER. We attempt to leverage the Mel\nspectrogram by decomposing distinguishable acoustic features for exploitation\nin our proposed architecture, which includes a novel feature map generator\nalgorithm, a CNN-based network feature extractor and a multi-layer perceptron\n(MLP) classifier. This study specifically focuses on effective data\naugmentation techniques for building an enriched hybrid-based feature map. This\nprocess results in a function that outputs a 2D image so that it can be used as\ninput data for a pre-trained CNN-VGG16 feature extractor. Furthermore, we also\ninvestigate other acoustic features such as MFCCs, chromagram, spectral\ncontrast, and the tonnetz to assess our proposed framework. A test accuracy of\n92.79% on the Berlin EMO-DB database is achieved. Our result is higher than\nprevious works using CNN-VGG16.\n", "date": "2023-12-18", "categories": []}
{"id": "2310.05804", "title": "Learning Language-guided Adaptive Hyper-modality Representation for\n  Multimodal Sentiment Analysis", "abstract": "  Though Multimodal Sentiment Analysis (MSA) proves effective by utilizing rich\ninformation from multiple sources (e.g., language, video, and audio), the\npotential sentiment-irrelevant and conflicting information across modalities\nmay hinder the performance from being further improved. To alleviate this, we\npresent Adaptive Language-guided Multimodal Transformer (ALMT), which\nincorporates an Adaptive Hyper-modality Learning (AHL) module to learn an\nirrelevance/conflict-suppressing representation from visual and audio features\nunder the guidance of language features at different scales. With the obtained\nhyper-modality representation, the model can obtain a complementary and joint\nrepresentation through multimodal fusion for effective MSA. In practice, ALMT\nachieves state-of-the-art performance on several popular datasets (e.g., MOSI,\nMOSEI and CH-SIMS) and an abundance of ablation demonstrates the validity and\nnecessity of our irrelevance/conflict suppression mechanism.\n", "date": "2023-10-09", "categories": []}
{"id": "2311.14483", "title": "SER_AMPEL: a multi-source dataset for speech emotion recognition of\n  Italian older adults", "abstract": "  In this paper, SER_AMPEL, a multi-source dataset for speech emotion\nrecognition (SER) is presented. The peculiarity of the dataset is that it is\ncollected with the aim of providing a reference for speech emotion recognition\nin case of Italian older adults. The dataset is collected following different\nprotocols, in particular considering acted conversations, extracted from movies\nand TV series, and recording natural conversations where the emotions are\nelicited by proper questions. The evidence of the need for such a dataset\nemerges from the analysis of the state of the art. Preliminary considerations\non the critical issues of SER are reported analyzing the classification results\non a subset of the proposed dataset.\n", "date": "2023-11-24", "categories": []}
{"id": "2312.08998", "title": "Design, construction and evaluation of emotional multimodal pathological\n  speech database", "abstract": "  The lack of an available emotion pathology database is one of the key\nobstacles in studying the emotion expression status of patients with\ndysarthria. The first Chinese multimodal emotional pathological speech database\ncontaining multi-perspective information is constructed in this paper. It\nincludes 29 controls and 39 patients with different degrees of motor\ndysarthria, expressing happy, sad, angry and neutral emotions. All emotional\nspeech was labeled for intelligibility, types and discrete dimensional emotions\nby developed WeChat mini-program. The subjective analysis justifies from\nemotion discrimination accuracy, speech intelligibility, valence-arousal\nspatial distribution, and correlation between SCL-90 and disease severity. The\nautomatic recognition tested on speech and glottal data, with average accuracy\nof 78% for controls and 60% for patients in audio, while 51% for controls and\n38% for patients in glottal data, indicating an influence of the disease on\nemotional expression.\n", "date": "2023-12-14", "categories": []}
{"id": "2312.07627", "title": "Multimodal Sentiment Analysis: Perceived vs Induced Sentiments", "abstract": "  Social media has created a global network where people can easily access and\nexchange vast information. This information gives rise to a variety of\nopinions, reflecting both positive and negative viewpoints. GIFs stand out as a\nmultimedia format offering a visually engaging way for users to communicate. In\nthis research, we propose a multimodal framework that integrates visual and\ntextual features to predict the GIF sentiment. It also incorporates attributes\nincluding face emotion detection and OCR generated captions to capture the\nsemantic aspects of the GIF. The developed classifier achieves an accuracy of\n82.7% on Twitter GIFs, which is an improvement over state-of-the-art models.\nMoreover, we have based our research on the ReactionGIF dataset, analysing the\nvariance in sentiment perceived by the author and sentiment induced in the\nreader\n", "date": "2023-12-12", "categories": []}
{"id": "2312.06925", "title": "Facial Emotion Recognition in VR Games", "abstract": "  Emotion detection is a crucial component of Games User Research (GUR), as it\nallows game developers to gain insights into players' emotional experiences and\ntailor their games accordingly. However, detecting emotions in Virtual Reality\n(VR) games is challenging due to the Head-Mounted Display (HMD) that covers the\ntop part of the player's face, namely, their eyes and eyebrows, which provide\ncrucial information for recognizing the impression. To tackle this we used a\nConvolutional Neural Network (CNN) to train a model to predict emotions in\nfull-face images where the eyes and eyebrows are covered. We used the FER2013\ndataset, which we modified to cover eyes and eyebrows in images. The model in\nthese images can accurately recognize seven different emotions which are anger,\nhappiness, disgust, fear, impartiality, sadness and surprise.\n  We assessed the model's performance by testing it on two VR games and using\nit to detect players' emotions. We collected self-reported emotion data from\nthe players after the gameplay sessions. We analyzed the data collected from\nour experiment to understand which emotions players experience during the\ngameplay. We found that our approach has the potential to enhance gameplay\nanalysis by enabling the detection of players' emotions in VR games, which can\nhelp game developers create more engaging and immersive game experiences.\n", "date": "2023-12-12", "categories": []}
{"id": "2312.07500", "title": "Multi-Branch Network for Imagery Emotion Prediction", "abstract": "  For a long time, images have proved perfect at both storing and conveying\nrich semantics, especially human emotions. A lot of research has been conducted\nto provide machines with the ability to recognize emotions in photos of people.\nPrevious methods mostly focus on facial expressions but fail to consider the\nscene context, meanwhile scene context plays an important role in predicting\nemotions, leading to more accurate results. In addition,\nValence-Arousal-Dominance (VAD) values offer a more precise quantitative\nunderstanding of continuous emotions, yet there has been less emphasis on\npredicting them compared to discrete emotional categories. In this paper, we\npresent a novel Multi-Branch Network (MBN), which utilizes various source\ninformation, including faces, bodies, and scene contexts to predict both\ndiscrete and continuous emotions in an image. Experimental results on EMOTIC\ndataset, which contains large-scale images of people in unconstrained\nsituations labeled with 26 discrete categories of emotions and VAD values, show\nthat our proposed method significantly outperforms state-of-the-art methods\nwith 28.4% in mAP and 0.93 in MAE. The results highlight the importance of\nutilizing multiple contextual information in emotion prediction and illustrate\nthe potential of our proposed method in a wide range of applications, such as\neffective computing, human-computer interaction, and social robotics. Source\ncode:\nhttps://github.com/BaoNinh2808/Multi-Branch-Network-for-Imagery-Emotion-Prediction\n", "date": "2023-12-12", "categories": []}
{"id": "2312.05265", "title": "Multimodal Group Emotion Recognition In-the-wild Using Privacy-Compliant\n  Features", "abstract": "  This paper explores privacy-compliant group-level emotion recognition\n''in-the-wild'' within the EmotiW Challenge 2023. Group-level emotion\nrecognition can be useful in many fields including social robotics,\nconversational agents, e-coaching and learning analytics. This research imposes\nitself using only global features avoiding individual ones, i.e. all features\nthat can be used to identify or track people in videos (facial landmarks, body\nposes, audio diarization, etc.). The proposed multimodal model is composed of a\nvideo and an audio branches with a cross-attention between modalities. The\nvideo branch is based on a fine-tuned ViT architecture. The audio branch\nextracts Mel-spectrograms and feed them through CNN blocks into a transformer\nencoder. Our training paradigm includes a generated synthetic dataset to\nincrease the sensitivity of our model on facial expression within the image in\na data-driven way. The extensive experiments show the significance of our\nmethodology. Our privacy-compliant proposal performs fairly on the EmotiW\nchallenge, with 79.24% and 75.13% of accuracy respectively on validation and\ntest set for the best models. Noticeably, our findings highlight that it is\npossible to reach this accuracy level with privacy-compliant features using\nonly 5 frames uniformly distributed on the video.\n", "date": "2023-12-06", "categories": []}
{"id": "2312.05907", "title": "Hypergraph-Guided Disentangled Spectrum Transformer Networks for\n  Near-Infrared Facial Expression Recognition", "abstract": "  With the strong robusticity on illumination variations, near-infrared (NIR)\ncan be an effective and essential complement to visible (VIS) facial expression\nrecognition in low lighting or complete darkness conditions. However, facial\nexpression recognition (FER) from NIR images presents more challenging problem\nthan traditional FER due to the limitations imposed by the data scale and the\ndifficulty of extracting discriminative features from incomplete visible\nlighting contents. In this paper, we give the first attempt to deep NIR facial\nexpression recognition and proposed a novel method called near-infrared facial\nexpression transformer (NFER-Former). Specifically, to make full use of the\nabundant label information in the field of VIS, we introduce a Self-Attention\nOrthogonal Decomposition mechanism that disentangles the expression information\nand spectrum information from the input image, so that the expression features\ncan be extracted without the interference of spectrum variation. We also\npropose a Hypergraph-Guided Feature Embedding method that models some key\nfacial behaviors and learns the structure of the complex correlations between\nthem, thereby alleviating the interference of inter-class similarity.\nAdditionally, we have constructed a large NIR-VIS Facial Expression dataset\nthat includes 360 subjects to better validate the efficiency of NFER-Former.\nExtensive experiments and ablation studies show that NFER-Former significantly\nimproves the performance of NIR FER and achieves state-of-the-art results on\nthe only two available NIR FER datasets, Oulu-CASIA and Large-HFE.\n", "date": "2023-12-10", "categories": []}
{"id": "2312.06025", "title": "Stress Management Using Virtual Reality-Based Attention Training", "abstract": "  In this research, we are concerned with the applicability of virtual\nreality-based attention training as a tool for stress management. Mental stress\nis a worldwide challenge that is still far from being fully managed. This has\nmaintained a remarkable research attention on developing and validating tools\nfor detecting and managing stress. Technology-based tools have been at the\nheart of these endeavors, including virtual reality (VR) technology.\nNevertheless, the potential of VR lies, to a large part, in the nature of the\ncontent being consumed through such technology. In this study, we investigate\nthe impact of a special type of content, namely, attention training, on the\nfeasibility of using VR for stress management. On a group of fourteen\nundergraduate engineering students, we conducted a study in which the\nparticipants got exposed twice to a stress inducer while their EEG signals were\nbeing recorded. The first iteration involved VR-based attention training before\nstarting the stress task while the second time did not. Using multiple features\nand various machine learning models, we show that VR-based attention training\nhas consistently resulted in reducing the number of recognized stress instances\nin the recorded EEG signals. This research gives preliminary insights on\nadopting VR-based attention training for managing stress, and future studies\nare required to replicate the results in larger samples.\n", "date": "2023-12-10", "categories": []}
{"id": "2312.06337", "title": "Deep Imbalanced Learning for Multimodal Emotion Recognition in\n  Conversations", "abstract": "  The main task of Multimodal Emotion Recognition in Conversations (MERC) is to\nidentify the emotions in modalities, e.g., text, audio, image and video, which\nis a significant development direction for realizing machine intelligence.\nHowever, many data in MERC naturally exhibit an imbalanced distribution of\nemotion categories, and researchers ignore the negative impact of imbalanced\ndata on emotion recognition. To tackle this problem, we systematically analyze\nit from three aspects: data augmentation, loss sensitivity, and sampling\nstrategy, and propose the Class Boundary Enhanced Representation Learning\n(CBERL) model. Concretely, we first design a multimodal generative adversarial\nnetwork to address the imbalanced distribution of {emotion} categories in raw\ndata. Secondly, a deep joint variational autoencoder is proposed to fuse\ncomplementary semantic information across modalities and obtain discriminative\nfeature representations. Finally, we implement a multi-task graph neural\nnetwork with mask reconstruction and classification optimization to solve the\nproblem of overfitting and underfitting in class boundary learning, and achieve\ncross-modal emotion recognition. We have conducted extensive experiments on the\nIEMOCAP and MELD benchmark datasets, and the results show that CBERL has\nachieved a certain performance improvement in the effectiveness of emotion\nrecognition. Especially on the minority class fear and disgust emotion labels,\nour model improves the accuracy and F1 value by 10% to 20%.\n", "date": "2023-12-11", "categories": []}
{"id": "2310.06536", "title": "EmoTwiCS: A Corpus for Modelling Emotion Trajectories in Dutch Customer\n  Service Dialogues on Twitter", "abstract": "  Due to the rise of user-generated content, social media is increasingly\nadopted as a channel to deliver customer service. Given the public character of\nthese online platforms, the automatic detection of emotions forms an important\napplication in monitoring customer satisfaction and preventing negative\nword-of-mouth. This paper introduces EmoTwiCS, a corpus of 9,489 Dutch customer\nservice dialogues on Twitter that are annotated for emotion trajectories. In\nour business-oriented corpus, we view emotions as dynamic attributes of the\ncustomer that can change at each utterance of the conversation. The term\n`emotion trajectory' refers therefore not only to the fine-grained emotions\nexperienced by customers (annotated with 28 labels and\nvalence-arousal-dominance scores), but also to the event happening prior to the\nconversation and the responses made by the human operator (both annotated with\n8 categories). Inter-annotator agreement (IAA) scores on the resulting dataset\nare substantial and comparable with related research, underscoring its high\nquality. Given the interplay between the different layers of annotated\ninformation, we perform several in-depth analyses to investigate (i) static\nemotions in isolated tweets, (ii) dynamic emotions and their shifts in\ntrajectory, and (iii) the role of causes and response strategies in emotion\ntrajectories. We conclude by listing the advantages and limitations of our\ndataset, after which we give some suggestions on the different types of\npredictive modelling tasks and open research questions to which EmoTwiCS can be\napplied. The dataset is available upon request and will be made publicly\navailable upon acceptance of the paper.\n", "date": "2023-10-10", "categories": []}
{"id": "2310.02298", "title": "Prompting Audios Using Acoustic Properties For Emotion Representation", "abstract": "  Emotions lie on a continuum, but current models treat emotions as a finite\nvalued discrete variable. This representation does not capture the diversity in\nthe expression of emotion. To better represent emotions we propose the use of\nnatural language descriptions (or prompts). In this work, we address the\nchallenge of automatically generating these prompts and training a model to\nbetter learn emotion representations from audio and prompt pairs. We use\nacoustic properties that are correlated to emotion like pitch, intensity,\nspeech rate, and articulation rate to automatically generate prompts i.e.\n'acoustic prompts'. We use a contrastive learning objective to map speech to\ntheir respective acoustic prompts. We evaluate our model on Emotion Audio\nRetrieval and Speech Emotion Recognition. Our results show that the acoustic\nprompts significantly improve the model's performance in EAR, in various\nPrecision@K metrics. In SER, we observe a 3.8% relative accuracy improvement on\nthe Ravdess dataset.\n", "date": "2023-10-03", "categories": []}
{"id": "2312.03756", "title": "LineConGraphs: Line Conversation Graphs for Effective Emotion\n  Recognition using Graph Neural Networks", "abstract": "  Emotion Recognition in Conversations (ERC) is a critical aspect of affective\ncomputing, and it has many practical applications in healthcare, education,\nchatbots, and social media platforms. Earlier approaches for ERC analysis\ninvolved modeling both speaker and long-term contextual information using graph\nneural network architectures. However, it is ideal to deploy\nspeaker-independent models for real-world applications. Additionally, long\ncontext windows can potentially create confusion in recognizing the emotion of\nan utterance in a conversation. To overcome these limitations, we propose novel\nline conversation graph convolutional network (LineConGCN) and graph attention\n(LineConGAT) models for ERC analysis. These models are speaker-independent and\nbuilt using a graph construction strategy for conversations -- line\nconversation graphs (LineConGraphs). The conversational context in\nLineConGraphs is short-term -- limited to one previous and future utterance,\nand speaker information is not part of the graph. We evaluate the performance\nof our proposed models on two benchmark datasets, IEMOCAP and MELD, and show\nthat our LineConGAT model outperforms the state-of-the-art methods with an\nF1-score of 64.58% and 76.50%. Moreover, we demonstrate that embedding\nsentiment shift information into line conversation graphs further enhances the\nERC performance in the case of GCN models.\n", "date": "2023-12-04", "categories": []}
{"id": "2312.02227", "title": "Improving Multimodal Sentiment Analysis: Supervised Angular Margin-based\n  Contrastive Learning for Enhanced Fusion Representation", "abstract": "  The effectiveness of a model is heavily reliant on the quality of the fusion\nrepresentation of multiple modalities in multimodal sentiment analysis.\nMoreover, each modality is extracted from raw input and integrated with the\nrest to construct a multimodal representation. Although previous methods have\nproposed multimodal representations and achieved promising results, most of\nthem focus on forming positive and negative pairs, neglecting the variation in\nsentiment scores within the same class. Additionally, they fail to capture the\nsignificance of unimodal representations in the fusion vector. To address these\nlimitations, we introduce a framework called Supervised Angular-based\nContrastive Learning for Multimodal Sentiment Analysis. This framework aims to\nenhance discrimination and generalizability of the multimodal representation\nand overcome biases in the fusion vector's modality. Our experimental results,\nalong with visualizations on two widely used datasets, demonstrate the\neffectiveness of our approach.\n", "date": "2023-12-04", "categories": []}
{"id": "2312.02578", "title": "Empathy and Distress Detection using Ensembles of Transformer Models", "abstract": "  This paper presents our approach for the WASSA 2023 Empathy, Emotion and\nPersonality Shared Task. Empathy and distress are human feelings that are\nimplicitly expressed in natural discourses. Empathy and distress detection are\ncrucial challenges in Natural Language Processing that can aid our\nunderstanding of conversations. The provided dataset consists of several\nlong-text examples in the English language, with each example associated with a\nnumeric score for empathy and distress. We experiment with several BERT-based\nmodels as a part of our approach. We also try various ensemble methods. Our\nfinal submission has a Pearson's r score of 0.346, placing us third in the\nempathy and distress detection subtask.\n", "date": "2023-12-05", "categories": []}
{"id": "2306.07848", "title": "GEmo-CLAP: Gender-Attribute-Enhanced Contrastive Language-Audio\n  Pretraining for Accurate Speech Emotion Recognition", "abstract": "  Contrastive cross-modality pretraining has recently exhibited impressive\nsuccess in diverse fields, whereas there is limited research on their merits in\nspeech emotion recognition (SER). In this paper, we propose GEmo-CLAP, a kind\nof gender-attribute-enhanced contrastive language-audio pretraining (CLAP)\nmethod for SER. Specifically, we first construct an effective emotion CLAP\n(Emo-CLAP) for SER, using pre-trained text and audio encoders. Second, given\nthe significance of gender information in SER, two novel multi-task learning\nbased GEmo-CLAP (ML-GEmo-CLAP) and soft label based GEmo-CLAP (SL-GEmo-CLAP)\nmodels are further proposed to incorporate gender information of speech\nsignals, forming more reasonable objectives. Experiments on IEMOCAP indicate\nthat our proposed two GEmo-CLAPs consistently outperform Emo-CLAP with\ndifferent pre-trained models. Remarkably, the proposed WavLM-based SL-GEmo-CLAP\nobtains the best WAR of 83.16\\%, which performs better than state-of-the-art\nSER methods.\n", "date": "2023-06-13", "categories": []}
{"id": "2310.11753", "title": "Bias in Emotion Recognition with ChatGPT", "abstract": "  This technical report explores the ability of ChatGPT in recognizing emotions\nfrom text, which can be the basis of various applications like interactive\nchatbots, data annotation, and mental health analysis. While prior research has\nshown ChatGPT's basic ability in sentiment analysis, its performance in more\nnuanced emotion recognition is not yet explored. Here, we conducted experiments\nto evaluate its performance of emotion recognition across different datasets\nand emotion labels. Our findings indicate a reasonable level of reproducibility\nin its performance, with noticeable improvement through fine-tuning. However,\nthe performance varies with different emotion labels and datasets, highlighting\nan inherent instability and possible bias. The choice of dataset and emotion\nlabels significantly impacts ChatGPT's emotion recognition performance. This\npaper sheds light on the importance of dataset and label selection, and the\npotential of fine-tuning in enhancing ChatGPT's emotion recognition\ncapabilities, providing a groundwork for better integration of emotion analysis\nin applications using ChatGPT.\n", "date": "2023-10-18", "categories": []}
{"id": "2312.01335", "title": "Facial Emotion Recognition Under Mask Coverage Using a Data Augmentation\n  Technique", "abstract": "  Identifying human emotions using AI-based computer vision systems, when\nindividuals wear face masks, presents a new challenge in the current Covid-19\npandemic. In this study, we propose a facial emotion recognition system capable\nof recognizing emotions from individuals wearing different face masks. A novel\ndata augmentation technique was utilized to improve the performance of our\nmodel using four mask types for each face image. We evaluated the effectiveness\nof four convolutional neural networks, Alexnet, Squeezenet, Resnet50 and\nVGGFace2 that were trained using transfer learning. The experimental findings\nrevealed that our model works effectively in multi-mask mode compared to\nsingle-mask mode. The VGGFace2 network achieved the highest accuracy rate, with\n97.82% for the person-dependent mode and 74.21% for the person-independent mode\nusing the JAFFE dataset. However, we evaluated our proposed model using the\nUIBVFED dataset. The Resnet50 has demonstrated superior performance, with\naccuracies of 73.68% for the person-dependent mode and 59.57% for the\nperson-independent mode. Moreover, we employed metrics such as precision,\nsensitivity, specificity, AUC, F1 score, and confusion matrix to measure our\nsystem's efficiency in detail. Additionally, the LIME algorithm was used to\nvisualize CNN's decision-making strategy.\n", "date": "2023-12-03", "categories": []}
{"id": "2310.09932", "title": "\"Reading Between the Heat\": Co-Teaching Body Thermal Signatures for\n  Non-intrusive Stress Detection", "abstract": "  Stress impacts our physical and mental health as well as our social life. A\npassive and contactless indoor stress monitoring system can unlock numerous\nimportant applications such as workplace productivity assessment, smart homes,\nand personalized mental health monitoring. While the thermal signatures from a\nuser's body captured by a thermal camera can provide important information\nabout the \"fight-flight\" response of the sympathetic and parasympathetic\nnervous system, relying solely on thermal imaging for training a stress\nprediction model often lead to overfitting and consequently a suboptimal\nperformance. This paper addresses this challenge by introducing ThermaStrain, a\nnovel co-teaching framework that achieves high-stress prediction performance by\ntransferring knowledge from the wearable modality to the contactless thermal\nmodality. During training, ThermaStrain incorporates a wearable electrodermal\nactivity (EDA) sensor to generate stress-indicative representations from\nthermal videos, emulating stress-indicative representations from a wearable EDA\nsensor. During testing, only thermal sensing is used, and stress-indicative\npatterns from thermal data and emulated EDA representations are extracted to\nimprove stress assessment. The study collected a comprehensive dataset with\nthermal video and EDA data under various stress conditions and distances.\nThermaStrain achieves an F1 score of 0.8293 in binary stress classification,\noutperforming the thermal-only baseline approach by over 9%. Extensive\nevaluations highlight ThermaStrain's effectiveness in recognizing\nstress-indicative attributes, its adaptability across distances and stress\nscenarios, real-time executability on edge platforms, its applicability to\nmulti-individual sensing, ability to function on limited visibility and\nunfamiliar conditions, and the advantages of its co-teaching approach.\n", "date": "2023-10-15", "categories": []}
{"id": "2311.17204", "title": "Optimal EEG Electrode Set for Emotion Recognition From Brain Signals: An\n  Empirical Quest", "abstract": "  The human brain is a complex organ, still completely undiscovered, that\ncontrols almost all the parts of the body. Apart from survival, the human brain\nstimulates emotions. Recent research indicates that brain signals can be very\neffective for emotion recognition. However, which parts of the brain exhibit\nmost of the emotions is still under-explored. In this study, we empirically\nanalyze the contribution of each part of the brain in exhibiting emotions. We\nuse the DEAP dataset to find the most optimal electrode set which eventually\nleads to the effective brain part associated with emotions. We use Fast Fourier\nTransformation for effective feature extraction and a 1D-CNN with residual\nconnection for classification. Though 32 electrodes from the DEAP dataset got\nan accuracy of 97.34%, only 12 electrodes (F7, P8, O1, F8, C4, T7, PO3, Fp1,\nFp2, O2, P3, and Fz) achieve 95.81% accuracy. This study also shows that adding\nmore than 10 electrodes does not improve performance significantly. Moreover,\nthe frontal lobe is the most important for recognizing emotion.\n", "date": "2023-11-28", "categories": []}
{"id": "2311.14674", "title": "Emotion-Oriented Behavior Model Using Deep Learning", "abstract": "  Emotions, as a fundamental ingredient of any social interaction, lead to\nbehaviors that represent the effectiveness of the interaction through facial\nexpressions and gestures in humans. Hence an agent must possess the social and\ncognitive abilities to understand human social parameters and behave\naccordingly. However, no such emotion-oriented behavior model is presented yet\nin the existing research. The emotion prediction may generate appropriate\nagents' behaviors for effective interaction using conversation modality.\nConsidering the importance of emotions, and behaviors, for an agent's social\ninteraction, an Emotion-based Behavior model is presented in this paper for\nSocio-cognitive artificial agents. The proposed model is implemented using\ntweets data trained on multiple models like Long Short-Term Memory (LSTM),\nConvolution Neural Network (CNN) and Bidirectional Encoder Representations from\nTransformers (BERT) for emotion prediction with an average accuracy of 92%, and\n55% respectively. Further, using emotion predictions from CNN-LSTM, the\nbehavior module responds using facial expressions and gestures using Behavioral\nMarkup Language (BML). The accuracy of emotion-based behavior predictions is\nstatistically validated using the 2-tailed Pearson correlation on the data\ncollected from human users through questionnaires. Analysis shows that all\nemotion-based behaviors accurately depict human-like gestures and facial\nexpressions based on the significant correlation at the 0.01 and 0.05 levels.\nThis study is a steppingstone to a multi-faceted artificial agent interaction\nbased on emotion-oriented behaviors. Cognition has significance regarding\nsocial interaction among humans.\n", "date": "2023-10-28", "categories": []}
{"id": "2207.12261", "title": "GraphCFC: A Directed Graph Based Cross-Modal Feature Complementation\n  Approach for Multimodal Conversational Emotion Recognition", "abstract": "  Emotion Recognition in Conversation (ERC) plays a significant part in\nHuman-Computer Interaction (HCI) systems since it can provide empathetic\nservices. Multimodal ERC can mitigate the drawbacks of uni-modal approaches.\nRecently, Graph Neural Networks (GNNs) have been widely used in a variety of\nfields due to their superior performance in relation modeling. In multimodal\nERC, GNNs are capable of extracting both long-distance contextual information\nand inter-modal interactive information. Unfortunately, since existing methods\nsuch as MMGCN directly fuse multiple modalities, redundant information may be\ngenerated and diverse information may be lost. In this work, we present a\ndirected Graph based Cross-modal Feature Complementation (GraphCFC) module that\ncan efficiently model contextual and interactive information. GraphCFC\nalleviates the problem of heterogeneity gap in multimodal fusion by utilizing\nmultiple subspace extractors and Pair-wise Cross-modal Complementary (PairCC)\nstrategy. We extract various types of edges from the constructed graph for\nencoding, thus enabling GNNs to extract crucial contextual and interactive\ninformation more accurately when performing message passing. Furthermore, we\ndesign a GNN structure called GAT-MLP, which can provide a new unified network\nframework for multimodal learning. The experimental results on two benchmark\ndatasets show that our GraphCFC outperforms the state-of-the-art (SOTA)\napproaches.\n", "date": "2022-07-06", "categories": []}
{"id": "2307.00449", "title": "A Dual-Stream Recurrence-Attention Network With Global-Local Awareness\n  for Emotion Recognition in Textual Dialog", "abstract": "  In real-world dialog systems, the ability to understand the user's emotions\nand interact anthropomorphically is of great significance. Emotion Recognition\nin Conversation (ERC) is one of the key ways to accomplish this goal and has\nattracted growing attention. How to model the context in a conversation is a\ncentral aspect and a major challenge of ERC tasks. Most existing approaches\nstruggle to adequately incorporate both global and local contextual\ninformation, and their network structures are overly sophisticated. For this\nreason, we propose a simple and effective Dual-stream Recurrence-Attention\nNetwork (DualRAN), which is based on Recurrent Neural Network (RNN) and\nMulti-head ATtention network (MAT). DualRAN eschews the complex components of\ncurrent methods and focuses on combining recurrence-based methods with\nattention-based ones. DualRAN is a dual-stream structure mainly consisting of\nlocal- and global-aware modules, modeling a conversation simultaneously from\ndistinct perspectives. In addition, we develop two single-stream network\nvariants for DualRAN, i.e., SingleRANv1 and SingleRANv2. According to the\nexperimental findings, DualRAN boosts the weighted F1 scores by 1.43% and 0.64%\non the IEMOCAP and MELD datasets, respectively, in comparison to the strongest\nbaseline. On two other datasets (i.e., EmoryNLP and DailyDialog), our method\nalso attains competitive results.\n", "date": "2023-07-02", "categories": []}
{"id": "2311.12812", "title": "Personalization of Affective Models to Enable Neuropsychiatric Digital\n  Precision Health Interventions: A Feasibility Study", "abstract": "  Mobile digital therapeutics for autism spectrum disorder (ASD) often target\nemotion recognition and evocation, which is a challenge for children with ASD.\nWhile such mobile applications often use computer vision machine learning (ML)\nmodels to guide the adaptive nature of the digital intervention, a single model\nis usually deployed and applied to all children. Here, we explore the potential\nof model personalization, or training a single emotion recognition model per\nperson, to improve the performance of these underlying emotion recognition\nmodels used to guide digital health therapies for children with ASD. We\nconducted experiments on the Emognition dataset, a video dataset of human\nsubjects evoking a series of emotions. For a subset of 10 individuals in the\ndataset with a sufficient representation of at least two ground truth emotion\nlabels, we trained a personalized version of three classical ML models on a set\nof 51 features extracted from each video frame. We measured the importance of\neach facial feature for all personalized models and observed differing ranked\nlists of top features across subjects, motivating the need for model\npersonalization. We then compared the personalized models against a generalized\nmodel trained using data from all 10 participants. The mean F1-scores achieved\nby the personalized models were 90.48%, 92.66%, and 86.40%, respectively. By\ncontrast, the mean F1-scores reached by non-personalized models trained on\ndifferent human subjects and evaluated using the same test set were 88.55%,\n91.78%, and 80.42%, respectively. The personalized models outperformed the\ngeneralized models for 7 out of 10 participants. PCA analyses on the remaining\n3 participants revealed relatively facial configuration differences between\nemotion labels within each subject, suggesting that personalized ML will fail\nwhen the variation among data points within a subjects data is too low.\n", "date": "2023-09-21", "categories": []}
{"id": "2311.11009", "title": "Joyful: Joint Modality Fusion and Graph Contrastive Learning for\n  Multimodal Emotion Recognition", "abstract": "  Multimodal emotion recognition aims to recognize emotions for each utterance\nof multiple modalities, which has received increasing attention for its\napplication in human-machine interaction. Current graph-based methods fail to\nsimultaneously depict global contextual features and local diverse uni-modal\nfeatures in a dialogue. Furthermore, with the number of graph layers\nincreasing, they easily fall into over-smoothing. In this paper, we propose a\nmethod for joint modality fusion and graph contrastive learning for multimodal\nemotion recognition (Joyful), where multimodality fusion, contrastive learning,\nand emotion recognition are jointly optimized. Specifically, we first design a\nnew multimodal fusion mechanism that can provide deep interaction and fusion\nbetween the global contextual and uni-modal specific features. Then, we\nintroduce a graph contrastive learning framework with inter-view and intra-view\ncontrastive losses to learn more distinguishable representations for samples\nwith different sentiments. Extensive experiments on three benchmark datasets\nindicate that Joyful achieved state-of-the-art (SOTA) performance compared to\nall baselines.\n", "date": "2023-11-18", "categories": []}
{"id": "2311.11116", "title": "Utilizing Speech Emotion Recognition and Recommender Systems for\n  Negative Emotion Handling in Therapy Chatbots", "abstract": "  Emotional well-being significantly influences mental health and overall\nquality of life. As therapy chatbots become increasingly prevalent, their\nability to comprehend and respond empathetically to users' emotions remains\nlimited. This paper addresses this limitation by proposing an approach to\nenhance therapy chatbots with auditory perception, enabling them to understand\nusers' feelings and provide human-like empathy. The proposed method\nincorporates speech emotion recognition (SER) techniques using Convolutional\nNeural Network (CNN) models and the ShEMO dataset to accurately detect and\nclassify negative emotions, including anger, fear, and sadness. The SER model\nachieves a validation accuracy of 88%, demonstrating its effectiveness in\nrecognizing emotional states from speech signals. Furthermore, a recommender\nsystem is developed, leveraging the SER model's output to generate personalized\nrecommendations for managing negative emotions, for which a new bilingual\ndataset was generated as well since there is no such dataset available for this\ntask. The recommender model achieves an accuracy of 98% by employing a\ncombination of global vectors for word representation (GloVe) and LSTM models.\nTo provide a more immersive and empathetic user experience, a text-to-speech\nmodel called GlowTTS is integrated, enabling the therapy chatbot to audibly\ncommunicate the generated recommendations to users in both English and Persian.\nThe proposed approach offers promising potential to enhance therapy chatbots by\nproviding them with the ability to recognize and respond to users' emotions,\nultimately improving the delivery of mental health support for both English and\nPersian-speaking users.\n", "date": "2023-11-18", "categories": []}
{"id": "2311.11237", "title": "Implementation of AI Deep Learning Algorithm For Multi-Modal Sentiment\n  Analysis", "abstract": "  A multi-modal emotion recognition method was established by combining\ntwo-channel convolutional neural network with ring network. This method can\nextract emotional information effectively and improve learning efficiency. The\nwords were vectorized with GloVe, and the word vector was input into the\nconvolutional neural network. Combining attention mechanism and maximum pool\nconverter BiSRU channel, the local deep emotion and pre-post sequential emotion\nsemantics are obtained. Finally, multiple features are fused and input as the\npolarity of emotion, so as to achieve the emotion analysis of the target.\nExperiments show that the emotion analysis method based on feature fusion can\neffectively improve the recognition accuracy of emotion data set and reduce the\nlearning time. The model has a certain generalization.\n", "date": "2023-11-19", "categories": []}
{"id": "2311.10119", "title": "Accommodating Missing Modalities in Time-Continuous Multimodal Emotion\n  Recognition", "abstract": "  Decades of research indicate that emotion recognition is more effective when\ndrawing information from multiple modalities. But what if some modalities are\nsometimes missing? To address this problem, we propose a novel\nTransformer-based architecture for recognizing valence and arousal in a\ntime-continuous manner even with missing input modalities. We use a coupling of\ncross-attention and self-attention mechanisms to emphasize relationships\nbetween modalities during time and enhance the learning process on weak salient\ninputs. Experimental results on the Ulm-TSST dataset show that our model\nexhibits an improvement of the concordance correlation coefficient evaluation\nof 37% when predicting arousal values and 30% when predicting valence values,\ncompared to a late-fusion baseline approach.\n", "date": "2023-11-16", "categories": []}
{"id": "2311.10343", "title": "Enhancing Student Engagement in Online Learning through Facial\n  Expression Analysis and Complex Emotion Recognition using Deep Learning", "abstract": "  In response to the COVID-19 pandemic, traditional physical classrooms have\ntransitioned to online environments, necessitating effective strategies to\nensure sustained student engagement. A significant challenge in online teaching\nis the absence of real-time feedback from teachers on students learning\nprogress. This paper introduces a novel approach employing deep learning\ntechniques based on facial expressions to assess students engagement levels\nduring online learning sessions. Human emotions cannot be adequately conveyed\nby a student using only the basic emotions, including anger, disgust, fear,\njoy, sadness, surprise, and neutrality. To address this challenge, proposed a\ngeneration of four complex emotions such as confusion, satisfaction,\ndisappointment, and frustration by combining the basic emotions. These complex\nemotions are often experienced simultaneously by students during the learning\nsession. To depict these emotions dynamically,utilized a continuous stream of\nimage frames instead of discrete images. The proposed work utilized a\nConvolutional Neural Network (CNN) model to categorize the fundamental\nemotional states of learners accurately. The proposed CNN model demonstrates\nstrong performance, achieving a 95% accuracy in precise categorization of\nlearner emotions.\n", "date": "2023-11-17", "categories": []}
{"id": "2304.12470", "title": "Vision-based Estimation of Fatigue and Engagement in Cognitive Training\n  Sessions", "abstract": "  Computerized cognitive training (CCT) is a scalable, well-tolerated\nintervention that has promise for slowing cognitive decline. Outcomes from CCT\nare limited by a lack of effective engagement, which is decreased by factors\nsuch as mental fatigue, particularly in older adults at risk for dementia.\nThere is a need for scalable, automated measures that can monitor mental\nfatigue during CCT. Here, we develop and validate a novel Recurrent Video\nTransformer (RVT) method for monitoring real-time mental fatigue in older\nadults with mild cognitive impairment from video-recorded facial gestures\nduring CCT. The RVT model achieved the highest balanced accuracy(78%) and\nprecision (0.82) compared to the prior state-of-the-art models for binary and\nmulti-class classification of mental fatigue and was additionally validated via\nsignificant association (p=0.023) with CCT reaction time. By leveraging dynamic\ntemporal information, the RVT model demonstrates the potential to accurately\nmeasure real-time mental fatigue, laying the foundation for future personalized\nCCT that increase effective engagement.\n", "date": "2023-04-24", "categories": []}
{"id": "2311.08607", "title": "Towards Generalizable SER: Soft Labeling and Data Augmentation for\n  Modeling Temporal Emotion Shifts in Large-Scale Multilingual Speech", "abstract": "  Recognizing emotions in spoken communication is crucial for advanced\nhuman-machine interaction. Current emotion detection methodologies often\ndisplay biases when applied cross-corpus. To address this, our study\namalgamates 16 diverse datasets, resulting in 375 hours of data across\nlanguages like English, Chinese, and Japanese. We propose a soft labeling\nsystem to capture gradational emotional intensities. Using the Whisper encoder\nand data augmentation methods inspired by contrastive learning, our method\nemphasizes the temporal dynamics of emotions. Our validation on four\nmultilingual datasets demonstrates notable zero-shot generalization. We publish\nour open source model weights and initial promising results after fine-tuning\non Hume-Prosody.\n", "date": "2023-11-15", "categories": []}
{"id": "2305.13583", "title": "Cross-Attention is Not Enough: Incongruity-Aware Dynamic Hierarchical\n  Fusion for Multimodal Affect Recognition", "abstract": "  Fusing multiple modalities has proven effective for multimodal information\nprocessing. However, the incongruity between modalities poses a challenge for\nmultimodal fusion, especially in affect recognition. In this study, we first\nanalyze how the salient affective information in one modality can be affected\nby the other, and demonstrate that inter-modal incongruity exists latently in\ncrossmodal attention. Based on this finding, we propose the Hierarchical\nCrossmodal Transformer with Dynamic Modality Gating (HCT-DMG), a lightweight\nincongruity-aware model, which dynamically chooses the primary modality in each\ntraining batch and reduces fusion times by leveraging the learned hierarchy in\nthe latent space to alleviate incongruity. The experimental evaluation on five\nbenchmark datasets: CMU-MOSI, CMU-MOSEI, and IEMOCAP (sentiment and emotion),\nwhere incongruity implicitly lies in hard samples, as well as UR-FUNNY (humour)\nand MUStaRD (sarcasm), where incongruity is common, verifies the efficacy of\nour approach, showing that HCT-DMG: 1) outperforms previous multimodal models\nwith a reduced size of approximately 0.8M parameters; 2) recognizes hard\nsamples where incongruity makes affect recognition difficult; 3) mitigates the\nincongruity at the latent level in crossmodal attention.\n", "date": "2023-05-23", "categories": []}
{"id": "2311.05513", "title": "From Learning Management System to Affective Tutoring system: a\n  preliminary study", "abstract": "  In this study, we investigate the combination of indicators, including\nperformance, behavioral engagement, and emotional engagement, to identify\nstudents experiencing difficulties. We analyzed data from two primary sources:\ndigital traces extracted from th e Learning Management System (LMS) and images\ncaptured by students' webcams. The digital traces provided insights into\nstudents' interactions with the educational content, while the images were\nutilized to analyze their emotional expressions during learnin g activities. By\nutilizing real data collected from students at a French engineering school,\nrecorded during the 2022 2023 academic year, we observed a correlation between\npositive emotional states and improved academic outcomes. These preliminary\nfindings support the notion that emotions play a crucial role in\ndifferentiating between high achieving and low achieving students.\n", "date": "2023-11-09", "categories": []}
{"id": "2311.03385", "title": "Intelligent Stress Assessment for e-Coaching", "abstract": "  This paper considers the adaptation of the e-coaching concept at times of\nemergencies and disasters, through aiding the e-coaching with intelligent tools\nfor monitoring humans' affective state. The states such as anxiety, panic,\navoidance, and stress, if properly detected, can be mitigated using the\ne-coaching tactic and strategy. In this work, we focus on a stress monitoring\nassistant tool developed on machine learning techniques. We provide the results\nof an experimental study using the proposed method.\n", "date": "2023-11-03", "categories": []}
{"id": "2311.03606", "title": "Multimodal Stress Detection Using Facial Landmarks and Biometric Signals", "abstract": "  The development of various sensing technologies is improving measurements of\nstress and the well-being of individuals. Although progress has been made with\nsingle signal modalities like wearables and facial emotion recognition,\nintegrating multiple modalities provides a more comprehensive understanding of\nstress, given that stress manifests differently across different people.\nMulti-modal learning aims to capitalize on the strength of each modality rather\nthan relying on a single signal. Given the complexity of processing and\nintegrating high-dimensional data from limited subjects, more research is\nneeded. Numerous research efforts have been focused on fusing stress and\nemotion signals at an early stage, e.g., feature-level fusion using basic\nmachine learning methods and 1D-CNN Methods. This paper proposes a multi-modal\nlearning approach for stress detection that integrates facial landmarks and\nbiometric signals. We test this multi-modal integration with various\nearly-fusion and late-fusion techniques to integrate the 1D-CNN model from\nbiometric signals and 2-D CNN using facial landmarks. We evaluate these\narchitectures using a rigorous test of models' generalizability using the\nleave-one-subject-out mechanism, i.e., all samples related to a single subject\nare left out to train the model. Our findings show that late-fusion achieved\n94.39\\% accuracy, and early-fusion surpassed it with a 98.38\\% accuracy rate.\nThis research contributes valuable insights into enhancing stress detection\nthrough a multi-modal approach. The proposed research offers important\nknowledge in improving stress detection using a multi-modal approach.\n", "date": "2023-11-06", "categories": []}
{"id": "2305.18907", "title": "Multitask learning for recognizing stress and depression in social media", "abstract": "  Stress and depression are prevalent nowadays across people of all ages due to\nthe quick paces of life. People use social media to express their feelings.\nThus, social media constitute a valuable form of information for the early\ndetection of stress and depression. Although many research works have been\nintroduced targeting the early recognition of stress and depression, there are\nstill limitations. There have been proposed multi-task learning settings, which\nuse depression and emotion (or figurative language) as the primary and\nauxiliary tasks respectively. However, although stress is inextricably linked\nwith depression, researchers face these two tasks as two separate tasks. To\naddress these limitations, we present the first study, which exploits two\ndifferent datasets collected under different conditions, and introduce two\nmultitask learning frameworks, which use depression and stress as the main and\nauxiliary tasks respectively. Specifically, we use a depression dataset and a\nstressful dataset including stressful posts from ten subreddits of five\ndomains. In terms of the first approach, each post passes through a shared BERT\nlayer, which is updated by both tasks. Next, two separate BERT encoder layers\nare exploited, which are updated by each task separately. Regarding the second\napproach, it consists of shared and task-specific layers weighted by attention\nfusion networks. We conduct a series of experiments and compare our approaches\nwith existing research initiatives, single-task learning, and transfer\nlearning. Experiments show multiple advantages of our approaches over\nstate-of-the-art ones.\n", "date": "2023-05-30", "categories": []}
{"id": "2306.02213", "title": "Evaluating Emotion Arcs Across Languages: Bridging the Global Divide in\n  Sentiment Analysis", "abstract": "  Emotion arcs capture how an individual (or a population) feels over time.\nThey are widely used in industry and research; however, there is little work on\nevaluating the automatically generated arcs. This is because of the difficulty\nof establishing the true (gold) emotion arc. Our work, for the first time,\nsystematically and quantitatively evaluates automatically generated emotion\narcs. We also compare two common ways of generating emotion arcs:\nMachine-Learning (ML) models and Lexicon-Only (LexO) methods. By running\nexperiments on 18 diverse datasets in 9 languages, we show that despite being\nmarkedly poor at instance level emotion classification, LexO methods are highly\naccurate at generating emotion arcs when aggregating information from hundreds\nof instances. We also show, through experiments on six indigenous African\nlanguages, as well as Arabic, and Spanish, that automatic translations of\nEnglish emotion lexicons can be used to generate high-quality emotion arcs in\nless-resource languages. This opens up avenues for work on emotions in\nlanguages from around the world; which is crucial for commerce, public policy,\nand health research in service of speakers often left behind. Code and\nresources: https://github.com/dteodore/EmotionArcs\n", "date": "2023-06-03", "categories": []}
{"id": "2311.02647", "title": "New Approach for an Affective Computing-Driven Quality of Experience\n  (QoE) Prediction", "abstract": "  In human interactions, emotion recognition is crucial. For this reason, the\ntopic of computer-vision approaches for automatic emotion recognition is\ncurrently being extensively researched. Processing multi-channel\nelectroencephalogram (EEG) information is one of the most researched methods\nfor automatic emotion recognition. This paper presents a new model for an\naffective computing-driven Quality of Experience (QoE) prediction. In order to\nvalidate the proposed model, a publicly available dataset is used. The dataset\ncontains EEG, ECG, and respiratory data and is focused on a multimedia QoE\nassessment context. The EEG data are retained on which the differential entropy\nand the power spectral density are calculated with an observation window of\nthree seconds. These two features were extracted to train several deep-learning\nmodels to investigate the possibility of predicting QoE with five different\nfactors. The performance of these models is compared, and the best model is\noptimized to improve the results. The best results were obtained with an\nLSTM-based model, presenting an F1-score from 68% to 78%. An analysis of the\nmodel and its features shows that the Delta frequency band is the least\nnecessary, that two electrodes have a higher importance, and that two other\nelectrodes have a very low impact on the model's performances.\n", "date": "2023-11-05", "categories": []}
{"id": "2311.02910", "title": "Benchmarking Deep Facial Expression Recognition: An Extensive Protocol\n  with Balanced Dataset in the Wild", "abstract": "  Facial expression recognition (FER) is a crucial part of human-computer\ninteraction. Existing FER methods achieve high accuracy and generalization\nbased on different open-source deep models and training approaches. However,\nthe performance of these methods is not always good when encountering practical\nsettings, which are seldom explored. In this paper, we collected a new\nin-the-wild facial expression dataset for cross-domain validation. Twenty-three\ncommonly used network architectures were implemented and evaluated following a\nuniform protocol. Moreover, various setups, in terms of input resolutions,\nclass balance management, and pre-trained strategies, were verified to show the\ncorresponding performance contribution. Based on extensive experiments on three\nlarge-scale FER datasets and our practical cross-validation, we ranked network\narchitectures and summarized a set of recommendations on deploying deep FER\nmethods in real scenarios. In addition, potential ethical rules, privacy\nissues, and regulations were discussed in practical FER applications such as\nmarketing, education, and entertainment business.\n", "date": "2023-11-06", "categories": []}
{"id": "2308.16741", "title": "Socratis: Are large multimodal models emotionally aware?", "abstract": "  Existing emotion prediction benchmarks contain coarse emotion labels which do\nnot consider the diversity of emotions that an image and text can elicit in\nhumans due to various reasons. Learning diverse reactions to multimodal content\nis important as intelligent machines take a central role in generating and\ndelivering content to society. To address this gap, we propose Socratis, a\nsocietal reactions benchmark, where each image-caption (IC) pair is annotated\nwith multiple emotions and the reasons for feeling them. Socratis contains 18K\nfree-form reactions for 980 emotions on 2075 image-caption pairs from 5\nwidely-read news and image-caption (IC) datasets. We benchmark the capability\nof state-of-the-art multimodal large language models to generate the reasons\nfor feeling an emotion given an IC pair. Based on a preliminary human study, we\nobserve that humans prefer human-written reasons over 2 times more often than\nmachine-generated ones. This shows our task is harder than standard generation\ntasks because it starkly contrasts recent findings where humans cannot tell\napart machine vs human-written news articles, for instance. We further see that\ncurrent captioning metrics based on large vision-language models also fail to\ncorrelate with human preferences. We hope that these findings and our benchmark\nwill inspire further research on training emotionally aware models.\n", "date": "2023-08-31", "categories": []}
{"id": "2311.00273", "title": "SoulChat: Improving LLMs' Empathy, Listening, and Comfort Abilities\n  through Fine-tuning with Multi-turn Empathy Conversations", "abstract": "  Large language models (LLMs) have been widely applied in various fields due\nto their excellent capability for memorizing knowledge and chain of thought\n(CoT). When these language models are applied in the field of psychological\ncounseling, they often rush to provide universal advice. However, when users\nseek psychological support, they need to gain empathy, trust, understanding and\ncomfort, rather than just reasonable advice. To this end, we constructed a\nmulti-turn empathetic conversation dataset of more than 2 million samples, in\nwhich the input is the multi-turn conversation context, and the target is\nempathetic responses that cover expressions such as questioning, comfort,\nrecognition, listening, trust, emotional support, etc. Experiments have shown\nthat the empathy ability of LLMs can be significantly enhanced when finetuning\nby using multi-turn dialogue history and responses that are closer to the\nexpression of a psychological consultant.\n", "date": "2023-11-01", "categories": []}
{"id": "2305.17422", "title": "Understanding Emotion Valence is a Joint Deep Learning Task", "abstract": "  The valence analysis of speakers' utterances or written posts helps to\nunderstand the activation and variations of the emotional state throughout the\nconversation. More recently, the concept of Emotion Carriers (EC) has been\nintroduced to explain the emotion felt by the speaker and its manifestations.\nIn this work, we investigate the natural inter-dependency of valence and ECs\nvia a multi-task learning approach. We experiment with Pre-trained Language\nModels (PLM) for single-task, two-step, and joint settings for the valence and\nEC prediction tasks. We compare and evaluate the performance of generative\n(GPT-2) and discriminative (BERT) architectures in each setting. We observed\nthat providing the ground truth label of one task improves the prediction\nperformance of the models in the other task. We further observed that the\ndiscriminative model achieves the best trade-off of valence and EC prediction\ntasks in the joint prediction setting. As a result, we attain a single model\nthat performs both tasks, thus, saving computation resources at training and\ninference times.\n", "date": "2023-05-27", "categories": []}
{"id": "2310.20494", "title": "A Transformer-Based Model With Self-Distillation for Multimodal Emotion\n  Recognition in Conversations", "abstract": "  Emotion recognition in conversations (ERC), the task of recognizing the\nemotion of each utterance in a conversation, is crucial for building empathetic\nmachines. Existing studies focus mainly on capturing context- and\nspeaker-sensitive dependencies on the textual modality but ignore the\nsignificance of multimodal information. Different from emotion recognition in\ntextual conversations, capturing intra- and inter-modal interactions between\nutterances, learning weights between different modalities, and enhancing modal\nrepresentations play important roles in multimodal ERC. In this paper, we\npropose a transformer-based model with self-distillation (SDT) for the task.\nThe transformer-based model captures intra- and inter-modal interactions by\nutilizing intra- and inter-modal transformers, and learns weights between\nmodalities dynamically by designing a hierarchical gated fusion strategy.\nFurthermore, to learn more expressive modal representations, we treat soft\nlabels of the proposed model as extra training supervision. Specifically, we\nintroduce self-distillation to transfer knowledge of hard and soft labels from\nthe proposed model to each modality. Experiments on IEMOCAP and MELD datasets\ndemonstrate that SDT outperforms previous state-of-the-art baselines.\n", "date": "2023-10-31", "categories": []}
{"id": "2310.18322", "title": "What's Next in Affective Modeling? Large Language Models", "abstract": "  Large Language Models (LLM) have recently been shown to perform well at\nvarious tasks from language understanding, reasoning, storytelling, and\ninformation search to theory of mind. In an extension of this work, we explore\nthe ability of GPT-4 to solve tasks related to emotion prediction. GPT-4\nperforms well across multiple emotion tasks; it can distinguish emotion\ntheories and come up with emotional stories. We show that by prompting GPT-4 to\nidentify key factors of an emotional experience, it is able to manipulate the\nemotional intensity of its own stories. Furthermore, we explore GPT-4's ability\non reverse appraisals by asking it to predict either the goal, belief, or\nemotion of a person using the other two. In general, GPT-4 can make the correct\ninferences. We suggest that LLMs could play an important role in affective\nmodeling; however, they will not fully replace works that attempt to model the\nmechanisms underlying emotion-related processes.\n", "date": "2023-10-03", "categories": []}
{"id": "2310.18363", "title": "A Contextualized Real-Time Multimodal Emotion Recognition for\n  Conversational Agents using Graph Convolutional Networks in Reinforcement\n  Learning", "abstract": "  Owing to the recent developments in Generative Artificial Intelligence\n(GenAI) and Large Language Models (LLM), conversational agents are becoming\nincreasingly popular and accepted. They provide a human touch by interacting in\nways familiar to us and by providing support as virtual companions. Therefore,\nit is important to understand the user's emotions in order to respond\nconsiderately. Compared to the standard problem of emotion recognition,\nconversational agents face an additional constraint in that recognition must be\nreal-time. Studies on model architectures using audio, visual, and textual\nmodalities have mainly focused on emotion classification using full video\nsequences that do not provide online features. In this work, we present a novel\nparadigm for contextualized Emotion Recognition using Graph Convolutional\nNetwork with Reinforcement Learning (conER-GRL). Conversations are partitioned\ninto smaller groups of utterances for effective extraction of contextual\ninformation. The system uses Gated Recurrent Units (GRU) to extract multimodal\nfeatures from these groups of utterances. More importantly, Graph Convolutional\nNetworks (GCN) and Reinforcement Learning (RL) agents are cascade trained to\ncapture the complex dependencies of emotion features in interactive scenarios.\nComparing the results of the conER-GRL model with other state-of-the-art models\non the benchmark dataset IEMOCAP demonstrates the advantageous capabilities of\nthe conER-GRL architecture in recognizing emotions in real-time from multimodal\nconversational signals.\n", "date": "2023-10-24", "categories": []}
{"id": "2310.18930", "title": "Retrofitting Light-weight Language Models for Emotions using Supervised\n  Contrastive Learning", "abstract": "  We present a novel retrofitting method to induce emotion aspects into\npre-trained language models (PLMs) such as BERT and RoBERTa. Our method updates\npre-trained network weights using contrastive learning so that the text\nfragments exhibiting similar emotions are encoded nearby in the representation\nspace, and the fragments with different emotion content are pushed apart. While\ndoing so, it also ensures that the linguistic knowledge already present in PLMs\nis not inadvertently perturbed. The language models retrofitted by our method,\ni.e., BERTEmo and RoBERTaEmo, produce emotion-aware text representations, as\nevaluated through different clustering and retrieval metrics. For the\ndownstream tasks on sentiment analysis and sarcasm detection, they perform\nbetter than their pre-trained counterparts (about 1% improvement in F1-score)\nand other existing approaches. Additionally, a more significant boost in\nperformance is observed for the retrofitted models over pre-trained ones in\nfew-shot learning setting.\n", "date": "2023-10-29", "categories": []}
{"id": "2310.15189", "title": "Towards Subject Agnostic Affective Emotion Recognition", "abstract": "  This paper focuses on affective emotion recognition, aiming to perform in the\nsubject-agnostic paradigm based on EEG signals. However, EEG signals manifest\nsubject instability in subject-agnostic affective Brain-computer interfaces\n(aBCIs), which led to the problem of distributional shift. Furthermore, this\nproblem is alleviated by approaches such as domain generalisation and domain\nadaptation. Typically, methods based on domain adaptation confer comparatively\nbetter results than the domain generalisation methods but demand more\ncomputational resources given new subjects. We propose a novel framework,\nmeta-learning based augmented domain adaptation for subject-agnostic aBCIs. Our\ndomain adaptation approach is augmented through meta-learning, which consists\nof a recurrent neural network, a classifier, and a distributional shift\ncontroller based on a sum-decomposable function. Also, we present that a neural\nnetwork explicating a sum-decomposable function can effectively estimate the\ndivergence between varied domains. The network setting for augmented domain\nadaptation follows meta-learning and adversarial learning, where the controller\npromptly adapts to new domains employing the target data via a few\nself-adaptation steps in the test phase. Our proposed approach is shown to be\neffective in experiments on a public aBICs dataset and achieves similar\nperformance to state-of-the-art domain adaptation methods while avoiding the\nuse of additional computational resources.\n", "date": "2023-10-20", "categories": []}
{"id": "2310.13711", "title": "Automatic Sensor-free Affect Detection: A Systematic Literature Review", "abstract": "  Emotions and other affective states play a pivotal role in cognition and,\nconsequently, the learning process. It is well-established that computer-based\nlearning environments (CBLEs) that can detect and adapt to students' affective\nstates can enhance learning outcomes. However, practical constraints often pose\nchallenges to the deployment of sensor-based affect detection in CBLEs,\nparticularly for large-scale or long-term applications. As a result,\nsensor-free affect detection, which exclusively relies on logs of students'\ninteractions with CBLEs, emerges as a compelling alternative. This paper\nprovides a comprehensive literature review on sensor-free affect detection. It\ndelves into the most frequently identified affective states, the methodologies\nand techniques employed for sensor development, the defining attributes of\nCBLEs and data samples, as well as key research trends. Despite the field's\nevident maturity, demonstrated by the consistent performance of the models and\nthe application of advanced machine learning techniques, there is ample scope\nfor future research. Potential areas for further exploration include enhancing\nthe performance of sensor-free detection models, amassing more samples of\nunderrepresented emotions, and identifying additional emotions. There is also a\nneed to refine model development practices and methods. This could involve\ncomparing the accuracy of various data collection techniques, determining the\noptimal granularity of duration, establishing a shared database of action logs\nand emotion labels, and making the source code of these models publicly\naccessible. Future research should also prioritize the integration of models\ninto CBLEs for real-time detection, the provision of meaningful interventions\nbased on detected emotions, and a deeper understanding of the impact of\nemotions on learning.\n", "date": "2023-10-11", "categories": []}
{"id": "2310.14143", "title": "MMTF-DES: A Fusion of Multimodal Transformer Models for Desire, Emotion,\n  and Sentiment Analysis of Social Media Data", "abstract": "  Desire is a set of human aspirations and wishes that comprise verbal and\ncognitive aspects that drive human feelings and behaviors, distinguishing\nhumans from other animals. Understanding human desire has the potential to be\none of the most fascinating and challenging research domains. It is tightly\ncoupled with sentiment analysis and emotion recognition tasks. It is beneficial\nfor increasing human-computer interactions, recognizing human emotional\nintelligence, understanding interpersonal relationships, and making decisions.\nHowever, understanding human desire is challenging and under-explored because\nways of eliciting desire might be different among humans. The task gets more\ndifficult due to the diverse cultures, countries, and languages. Prior studies\noverlooked the use of image-text pairwise feature representation, which is\ncrucial for the task of human desire understanding. In this research, we have\nproposed a unified multimodal transformer-based framework with image-text pair\nsettings to identify human desire, sentiment, and emotion. The core of our\nproposed method lies in the encoder module, which is built using two\nstate-of-the-art multimodal transformer models. These models allow us to\nextract diverse features. To effectively extract visual and contextualized\nembedding features from social media image and text pairs, we conducted joint\nfine-tuning of two pre-trained multimodal transformer models:\nVision-and-Language Transformer (ViLT) and Vision-and-Augmented-Language\nTransformer (VAuLT). Subsequently, we use an early fusion strategy on these\nembedding features to obtain combined diverse feature representations of the\nimage-text pair. This consolidation incorporates diverse information about this\ntask, enabling us to robustly perceive the context and image pair from multiple\nperspectives.\n", "date": "2023-10-22", "categories": []}
{"id": "2310.14225", "title": "Customising General Large Language Models for Specialised Emotion\n  Recognition Tasks", "abstract": "  The advent of large language models (LLMs) has gained tremendous attention\nover the past year. Previous studies have shown the astonishing performance of\nLLMs not only in other tasks but also in emotion recognition in terms of\naccuracy, universality, explanation, robustness, few/zero-shot learning, and\nothers. Leveraging the capability of LLMs inevitably becomes an essential\nsolution for emotion recognition. To this end, we further comprehensively\ninvestigate how LLMs perform in linguistic emotion recognition if we\nconcentrate on this specific task. Specifically, we exemplify a publicly\navailable and widely used LLM -- Chat General Language Model, and customise it\nfor our target by using two different modal adaptation techniques, i.e., deep\nprompt tuning and low-rank adaptation. The experimental results obtained on six\nwidely used datasets present that the adapted LLM can easily outperform other\nstate-of-the-art but specialised deep models. This indicates the strong\ntransferability and feasibility of LLMs in the field of emotion recognition.\n", "date": "2023-10-22", "categories": []}
{"id": "2310.14389", "title": "Evaluating Subjective Cognitive Appraisals of Emotions from Large\n  Language Models", "abstract": "  The emotions we experience involve complex processes; besides physiological\naspects, research in psychology has studied cognitive appraisals where people\nassess their situations subjectively, according to their own values (Scherer,\n2005). Thus, the same situation can often result in different emotional\nexperiences. While the detection of emotion is a well-established task, there\nis very limited work so far on the automatic prediction of cognitive\nappraisals. This work fills the gap by presenting CovidET-Appraisals, the most\ncomprehensive dataset to-date that assesses 24 appraisal dimensions, each with\na natural language rationale, across 241 Reddit posts. CovidET-Appraisals\npresents an ideal testbed to evaluate the ability of large language models --\nexcelling at a wide range of NLP tasks -- to automatically assess and explain\ncognitive appraisals. We found that while the best models are performant,\nopen-sourced LLMs fall short at this task, presenting a new challenge in the\nfuture development of emotionally intelligent models. We release our dataset at\nhttps://github.com/honglizhan/CovidET-Appraisals-Public.\n", "date": "2023-10-22", "categories": []}
{"id": "2310.14614", "title": "Efficient Cross-Task Prompt Tuning for Few-Shot Conversational Emotion\n  Recognition", "abstract": "  Emotion Recognition in Conversation (ERC) has been widely studied due to its\nimportance in developing emotion-aware empathetic machines. The rise of\npre-trained language models (PLMs) has further pushed the limit of ERC\nperformance. However, most recent works on ERC using PLMs are heavily\ndata-driven, and requires fine-tuning the entire PLMs. To improve both sample\nand computational efficiency, we propose a derivative-free optimization method\ncalled Cross-Task Prompt Tuning (CTPT) for few-shot conversational emotion\nrecognition. Unlike existing methods that learn independent knowledge from\nindividual tasks, CTPT leverages sharable cross-task knowledge by exploiting\nexternal knowledge from other source tasks to improve learning performance\nunder the few-shot setting. Moreover, CTPT only needs to optimize a vector\nunder the low intrinsic dimensionality without gradient, which is highly\nparameter-efficient compared with existing approaches. Experiments on five\ndifferent contextual conversation datasets demonstrate that our CTPT method has\nsuperior results on both few-shot scenarios and zero-shot transfers.\n", "date": "2023-10-23", "categories": []}
{"id": "2310.13080", "title": "From Multilingual Complexity to Emotional Clarity: Leveraging\n  Commonsense to Unveil Emotions in Code-Mixed Dialogues", "abstract": "  Understanding emotions during conversation is a fundamental aspect of human\ncommunication, driving NLP research for Emotion Recognition in Conversation\n(ERC). While considerable research has focused on discerning emotions of\nindividual speakers in monolingual dialogues, understanding the emotional\ndynamics in code-mixed conversations has received relatively less attention.\nThis motivates our undertaking of ERC for code-mixed conversations in this\nstudy. Recognizing that emotional intelligence encompasses a comprehension of\nworldly knowledge, we propose an innovative approach that integrates\ncommonsense information with dialogue context to facilitate a deeper\nunderstanding of emotions. To achieve this, we devise an efficient pipeline\nthat extracts relevant commonsense from existing knowledge graphs based on the\ncode-mixed input. Subsequently, we develop an advanced fusion technique that\nseamlessly combines the acquired commonsense information with the dialogue\nrepresentation obtained from a dedicated dialogue understanding module. Our\ncomprehensive experimentation showcases the substantial performance improvement\nobtained through the systematic incorporation of commonsense in ERC. Both\nquantitative assessments and qualitative analyses further corroborate the\nvalidity of our hypothesis, reaffirming the pivotal role of commonsense\nintegration in enhancing ERC.\n", "date": "2023-10-19", "categories": []}
{"id": "2310.12459", "title": "Affective Conversational Agents: Understanding Expectations and Personal\n  Influences", "abstract": "  The rise of AI conversational agents has broadened opportunities to enhance\nhuman capabilities across various domains. As these agents become more\nprevalent, it is crucial to investigate the impact of different affective\nabilities on their performance and user experience. In this study, we surveyed\n745 respondents to understand the expectations and preferences regarding\naffective skills in various applications. Specifically, we assessed preferences\nconcerning AI agents that can perceive, respond to, and simulate emotions\nacross 32 distinct scenarios. Our results indicate a preference for scenarios\nthat involve human interaction, emotional support, and creative tasks, with\ninfluences from factors such as emotional reappraisal and personality traits.\nOverall, the desired affective skills in AI agents depend largely on the\napplication's context and nature, emphasizing the need for adaptability and\ncontext-awareness in the design of affective AI conversational agents.\n", "date": "2023-10-19", "categories": []}
{"id": "2310.12851", "title": "EmoDiarize: Speaker Diarization and Emotion Identification from Speech\n  Signals using Convolutional Neural Networks", "abstract": "  In the era of advanced artificial intelligence and human-computer\ninteraction, identifying emotions in spoken language is paramount. This\nresearch explores the integration of deep learning techniques in speech emotion\nrecognition, offering a comprehensive solution to the challenges associated\nwith speaker diarization and emotion identification. It introduces a framework\nthat combines a pre-existing speaker diarization pipeline and an emotion\nidentification model built on a Convolutional Neural Network (CNN) to achieve\nhigher precision. The proposed model was trained on data from five speech\nemotion datasets, namely, RAVDESS, CREMA-D, SAVEE, TESS, and Movie Clips, out\nof which the latter is a speech emotion dataset created specifically for this\nresearch. The features extracted from each sample include Mel Frequency\nCepstral Coefficients (MFCC), Zero Crossing Rate (ZCR), Root Mean Square (RMS),\nand various data augmentation algorithms like pitch, noise, stretch, and shift.\nThis feature extraction approach aims to enhance prediction accuracy while\nreducing computational complexity. The proposed model yields an unweighted\naccuracy of 63%, demonstrating remarkable efficiency in accurately identifying\nemotional states within speech signals.\n", "date": "2023-10-19", "categories": []}
{"id": "2310.12062", "title": "On the use of Vision-Language models for Visual Sentiment Analysis: a\n  study on CLIP", "abstract": "  This work presents a study on how to exploit the CLIP embedding space to\nperform Visual Sentiment Analysis. We experiment with two architectures built\non top of the CLIP embedding space, which we denote by CLIP-E. We train the\nCLIP-E models with WEBEmo, the largest publicly available and manually labeled\nbenchmark for Visual Sentiment Analysis, and perform two sets of experiments.\nFirst, we test on WEBEmo and compare the CLIP-E architectures with\nstate-of-the-art (SOTA) models and with CLIP Zero-Shot. Second, we perform\ncross dataset evaluation, and test the CLIP-E architectures trained with WEBEmo\non other Visual Sentiment Analysis benchmarks. Our results show that the CLIP-E\napproaches outperform SOTA models in WEBEmo fine grained categorization, and\nthey also generalize better when tested on datasets that have not been seen\nduring training. Interestingly, we observed that for the FI dataset, CLIP\nZero-Shot produces better accuracies than SOTA models and CLIP-E trained on\nWEBEmo. These results motivate several questions that we discuss in this paper,\nsuch as how we should design new benchmarks and evaluate Visual Sentiment\nAnalysis, and whether we should keep designing tailored Deep Learning models\nfor Visual Sentiment Analysis or focus our efforts on better using the\nknowledge encoded in large vision-language models such as CLIP for this task.\n", "date": "2023-10-18", "categories": []}
{"id": "2206.05833", "title": "COLD Fusion: Calibrated and Ordinal Latent Distribution Fusion for\n  Uncertainty-Aware Multimodal Emotion Recognition", "abstract": "  Automatically recognising apparent emotions from face and voice is hard, in\npart because of various sources of uncertainty, including in the input data and\nthe labels used in a machine learning framework. This paper introduces an\nuncertainty-aware audiovisual fusion approach that quantifies modality-wise\nuncertainty towards emotion prediction. To this end, we propose a novel fusion\nframework in which we first learn latent distributions over audiovisual\ntemporal context vectors separately, and then constrain the variance vectors of\nunimodal latent distributions so that they represent the amount of information\neach modality provides w.r.t. emotion recognition. In particular, we impose\nCalibration and Ordinal Ranking constraints on the variance vectors of\naudiovisual latent distributions. When well-calibrated, modality-wise\nuncertainty scores indicate how much their corresponding predictions may differ\nfrom the ground truth labels. Well-ranked uncertainty scores allow the ordinal\nranking of different frames across the modalities. To jointly impose both these\nconstraints, we propose a softmax distributional matching loss. In both\nclassification and regression settings, we compare our uncertainty-aware fusion\nmodel with standard model-agnostic fusion baselines. Our evaluation on two\nemotion recognition corpora, AVEC 2019 CES and IEMOCAP, shows that audiovisual\nemotion recognition can considerably benefit from well-calibrated and\nwell-ranked latent uncertainty measures.\n", "date": "2022-06-12", "categories": []}
{"id": "2212.07055", "title": "Most Important Person-guided Dual-branch Cross-Patch Attention for Group\n  Affect Recognition", "abstract": "  Group affect refers to the subjective emotion that is evoked by an external\nstimulus in a group, which is an important factor that shapes group behavior\nand outcomes. Recognizing group affect involves identifying important\nindividuals and salient objects among a crowd that can evoke emotions. However,\nmost existing methods lack attention to affective meaning in group dynamics and\nfail to account for the contextual relevance of faces and objects in\ngroup-level images. In this work, we propose a solution by incorporating the\npsychological concept of the Most Important Person (MIP), which represents the\nmost noteworthy face in a crowd and has affective semantic meaning. We present\nthe Dual-branch Cross-Patch Attention Transformer (DCAT) which uses global\nimage and MIP together as inputs. Specifically, we first learn the informative\nfacial regions produced by the MIP and the global context separately. Then, the\nCross-Patch Attention module is proposed to fuse the features of MIP and global\ncontext together to complement each other. Our proposed method outperforms\nstate-of-the-art methods on GAF 3.0, GroupEmoW, and HECO datasets. Moreover, we\ndemonstrate the potential for broader applications by showing that our proposed\nmodel can be transferred to another group affect task, group cohesion, and\nachieve comparable results.\n", "date": "2022-12-14", "categories": []}
{"id": "2310.10673", "title": "Towards Emotion-Based Synthetic Consciousness: Using LLMs to Estimate\n  Emotion Probability Vectors", "abstract": "  This paper shows how LLMs (Large Language Models) may be used to estimate a\nsummary of the emotional state associated with piece of text. The summary of\nemotional state is a dictionary of words used to describe emotion together with\nthe probability of the word appearing after a prompt comprising the original\ntext and an emotion eliciting tail. Through emotion analysis of Amazon product\nreviews we demonstrate emotion descriptors can be mapped into a PCA type space.\nIt was hoped that text descriptions of actions to improve a current text\ndescribed state could also be elicited through a tail prompt. Experiment seemed\nto indicate that this is not straightforward to make work. This failure put our\nhoped for selection of action via choosing the best predict ed outcome via\ncomparing emotional responses out of reach for the moment.\n", "date": "2023-10-09", "categories": []}
{"id": "2309.02092", "title": "Where are We in Event-centric Emotion Analysis? Bridging Emotion Role\n  Labeling and Appraisal-based Approaches", "abstract": "  The term emotion analysis in text subsumes various natural language\nprocessing tasks which have in common the goal to enable computers to\nunderstand emotions. Most popular is emotion classification in which one or\nmultiple emotions are assigned to a predefined textual unit. While such setting\nis appropriate for identifying the reader's or author's emotion, emotion role\nlabeling adds the perspective of mentioned entities and extracts text spans\nthat correspond to the emotion cause. The underlying emotion theories agree on\none important point; that an emotion is caused by some internal or external\nevent and comprises several subcomponents, including the subjective feeling and\na cognitive evaluation. We therefore argue that emotions and events are related\nin two ways. (1) Emotions are events; and this perspective is the fundament in\nnatural language processing for emotion role labeling. (2) Emotions are caused\nby events; a perspective that is made explicit with research how to incorporate\npsychological appraisal theories in NLP models to interpret events. These two\nresearch directions, role labeling and (event-focused) emotion classification,\nhave by and large been tackled separately. In this paper, we contextualize both\nperspectives and discuss open research questions.\n", "date": "2023-09-05", "categories": []}
{"id": "2310.07648", "title": "Hypercomplex Multimodal Emotion Recognition from EEG and Peripheral\n  Physiological Signals", "abstract": "  Multimodal emotion recognition from physiological signals is receiving an\nincreasing amount of attention due to the impossibility to control them at will\nunlike behavioral reactions, thus providing more reliable information. Existing\ndeep learning-based methods still rely on extracted handcrafted features, not\ntaking full advantage of the learning ability of neural networks, and often\nadopt a single-modality approach, while human emotions are inherently expressed\nin a multimodal way. In this paper, we propose a hypercomplex multimodal\nnetwork equipped with a novel fusion module comprising parameterized\nhypercomplex multiplications. Indeed, by operating in a hypercomplex domain the\noperations follow algebraic rules which allow to model latent relations among\nlearned feature dimensions for a more effective fusion step. We perform\nclassification of valence and arousal from electroencephalogram (EEG) and\nperipheral physiological signals, employing the publicly available database\nMAHNOB-HCI surpassing a multimodal state-of-the-art network. The code of our\nwork is freely available at https://github.com/ispamm/MHyEEG.\n", "date": "2023-10-11", "categories": []}
{"id": "2310.00068", "title": "Emotional Listener Portrait: Neural Listener Head Generation with\n  Emotion", "abstract": "  Listener head generation centers on generating non-verbal behaviors (e.g.,\nsmile) of a listener in reference to the information delivered by a speaker. A\nsignificant challenge when generating such responses is the non-deterministic\nnature of fine-grained facial expressions during a conversation, which varies\ndepending on the emotions and attitudes of both the speaker and the listener.\nTo tackle this problem, we propose the Emotional Listener Portrait (ELP), which\ntreats each fine-grained facial motion as a composition of several discrete\nmotion-codewords and explicitly models the probability distribution of the\nmotions under different emotion in conversation. Benefiting from the\n``explicit'' and ``discrete'' design, our ELP model can not only automatically\ngenerate natural and diverse responses toward a given speaker via sampling from\nthe learned distribution but also generate controllable responses with a\npredetermined attitude. Under several quantitative metrics, our ELP exhibits\nsignificant improvements compared to previous methods.\n", "date": "2023-09-29", "categories": []}
{"id": "2310.04456", "title": "Multimodal Prompt Transformer with Hybrid Contrastive Learning for\n  Emotion Recognition in Conversation", "abstract": "  Emotion Recognition in Conversation (ERC) plays an important role in driving\nthe development of human-machine interaction. Emotions can exist in multiple\nmodalities, and multimodal ERC mainly faces two problems: (1) the noise problem\nin the cross-modal information fusion process, and (2) the prediction problem\nof less sample emotion labels that are semantically similar but different\ncategories. To address these issues and fully utilize the features of each\nmodality, we adopted the following strategies: first, deep emotion cues\nextraction was performed on modalities with strong representation ability, and\nfeature filters were designed as multimodal prompt information for modalities\nwith weak representation ability. Then, we designed a Multimodal Prompt\nTransformer (MPT) to perform cross-modal information fusion. MPT embeds\nmultimodal fusion information into each attention layer of the Transformer,\nallowing prompt information to participate in encoding textual features and\nbeing fused with multi-level textual information to obtain better multimodal\nfusion features. Finally, we used the Hybrid Contrastive Learning (HCL)\nstrategy to optimize the model's ability to handle labels with few samples.\nThis strategy uses unsupervised contrastive learning to improve the\nrepresentation ability of multimodal fusion and supervised contrastive learning\nto mine the information of labels with few samples. Experimental results show\nthat our proposed model outperforms state-of-the-art models in ERC on two\nbenchmark datasets.\n", "date": "2023-10-04", "categories": []}
{"id": "2310.04481", "title": "Acoustic and linguistic representations for speech continuous emotion\n  recognition in call center conversations", "abstract": "  The goal of our research is to automatically retrieve the satisfaction and\nthe frustration in real-life call-center conversations. This study focuses an\nindustrial application in which the customer satisfaction is continuously\ntracked down to improve customer services. To compensate the lack of large\nannotated emotional databases, we explore the use of pre-trained speech\nrepresentations as a form of transfer learning towards AlloSat corpus.\nMoreover, several studies have pointed out that emotion can be detected not\nonly in speech but also in facial trait, in biological response or in textual\ninformation. In the context of telephone conversations, we can break down the\naudio information into acoustic and linguistic by using the speech signal and\nits transcription. Our experiments confirms the large gain in performance\nobtained with the use of pre-trained features. Surprisingly, we found that the\nlinguistic content is clearly the major contributor for the prediction of\nsatisfaction and best generalizes to unseen data. Our experiments conclude to\nthe definitive advantage of using CamemBERT representations, however the\nbenefit of the fusion of acoustic and linguistic modalities is not as obvious.\nWith models learnt on individual annotations, we found that fusion approaches\nare more robust to the subjectivity of the annotation task. This study also\ntackles the problem of performances variability and intends to estimate this\nvariability from different views: weights initialization, confidence intervals\nand annotation subjectivity. A deep analysis on the linguistic content\ninvestigates interpretable factors able to explain the high contribution of the\nlinguistic modality for this task.\n", "date": "2023-10-06", "categories": []}
{"id": "2310.04043", "title": "In the Blink of an Eye: Event-based Emotion Recognition", "abstract": "  We introduce a wearable single-eye emotion recognition device and a real-time\napproach to recognizing emotions from partial observations of an emotion that\nis robust to changes in lighting conditions. At the heart of our method is a\nbio-inspired event-based camera setup and a newly designed lightweight Spiking\nEye Emotion Network (SEEN). Compared to conventional cameras, event-based\ncameras offer a higher dynamic range (up to 140 dB vs. 80 dB) and a higher\ntemporal resolution. Thus, the captured events can encode rich temporal cues\nunder challenging lighting conditions. However, these events lack texture\ninformation, posing problems in decoding temporal information effectively. SEEN\ntackles this issue from two different perspectives. First, we adopt\nconvolutional spiking layers to take advantage of the spiking neural network's\nability to decode pertinent temporal information. Second, SEEN learns to\nextract essential spatial cues from corresponding intensity frames and\nleverages a novel weight-copy scheme to convey spatial attention to the\nconvolutional spiking layers during training and inference. We extensively\nvalidate and demonstrate the effectiveness of our approach on a specially\ncollected Single-eye Event-based Emotion (SEE) dataset. To the best of our\nknowledge, our method is the first eye-based emotion recognition method that\nleverages event-based cameras and spiking neural network.\n", "date": "2023-10-06", "categories": []}
{"id": "2310.02281", "title": "End-to-End Continuous Speech Emotion Recognition in Real-life Customer\n  Service Call Center Conversations", "abstract": "  Speech Emotion recognition (SER) in call center conversations has emerged as\na valuable tool for assessing the quality of interactions between clients and\nagents. In contrast to controlled laboratory environments, real-life\nconversations take place under uncontrolled conditions and are subject to\ncontextual factors that influence the expression of emotions. In this paper, we\npresent our approach to constructing a large-scale reallife dataset (CusEmo)\nfor continuous SER in customer service call center conversations. We adopted\nthe dimensional emotion annotation approach to capture the subtlety,\ncomplexity, and continuity of emotions in real-life call center conversations,\nwhile annotating contextual information. The study also addresses the\nchallenges encountered during the application of the End-to-End (E2E) SER\nsystem to the dataset, including determining the appropriate label sampling\nrate and input segment length, as well as integrating contextual information\n(interlocutor's gender and empathy level) with different weights using\nmultitask learning. The result shows that incorporating the empathy level\ninformation improved the model's performance.\n", "date": "2023-10-02", "categories": []}
{"id": "2211.08233", "title": "Temporal Modeling Matters: A Novel Temporal Emotional Modeling Approach\n  for Speech Emotion Recognition", "abstract": "  Speech emotion recognition (SER) plays a vital role in improving the\ninteractions between humans and machines by inferring human emotion and\naffective states from speech signals. Whereas recent works primarily focus on\nmining spatiotemporal information from hand-crafted features, we explore how to\nmodel the temporal patterns of speech emotions from dynamic temporal scales.\nTowards that goal, we introduce a novel temporal emotional modeling approach\nfor SER, termed Temporal-aware bI-direction Multi-scale Network (TIM-Net),\nwhich learns multi-scale contextual affective representations from various time\nscales. Specifically, TIM-Net first employs temporal-aware blocks to learn\ntemporal affective representation, then integrates complementary information\nfrom the past and the future to enrich contextual representations, and finally,\nfuses multiple time scale features for better adaptation to the emotional\nvariation. Extensive experimental results on six benchmark SER datasets\ndemonstrate the superior performance of TIM-Net, gaining 2.34% and 2.61%\nimprovements of the average UAR and WAR over the second-best on each corpus.\nThe source code is available at https://github.com/Jiaxin-Ye/TIM-Net_SER.\n", "date": "2022-11-14", "categories": []}
{"id": "2303.09892", "title": "Memotion 3: Dataset on Sentiment and Emotion Analysis of Codemixed\n  Hindi-English Memes", "abstract": "  Memes are the new-age conveyance mechanism for humor on social media sites.\nMemes often include an image and some text. Memes can be used to promote\ndisinformation or hatred, thus it is crucial to investigate in details. We\nintroduce Memotion 3, a new dataset with 10,000 annotated memes. Unlike other\nprevalent datasets in the domain, including prior iterations of Memotion,\nMemotion 3 introduces Hindi-English Codemixed memes while prior works in the\narea were limited to only the English memes. We describe the Memotion task, the\ndata collection and the dataset creation methodologies. We also provide a\nbaseline for the task. The baseline code and dataset will be made available at\nhttps://github.com/Shreyashm16/Memotion-3.0\n", "date": "2023-03-17", "categories": []}
{"id": "2310.00010", "title": "Artificial Empathy Classification: A Survey of Deep Learning Techniques,\n  Datasets, and Evaluation Scales", "abstract": "  From the last decade, researchers in the field of machine learning (ML) and\nassistive developmental robotics (ADR) have taken an interest in artificial\nempathy (AE) as a possible future paradigm for human-robot interaction (HRI).\nHumans learn empathy since birth, therefore, it is challenging to instill this\nsense in robots and intelligent machines. Nevertheless, by training over a vast\namount of data and time, imitating empathy, to a certain extent, can be\npossible for robots. Training techniques for AE, along with findings from the\nfield of empathetic AI research, are ever-evolving. The standard workflow for\nartificial empathy consists of three stages: 1) Emotion Recognition (ER) using\nthe retrieved features from video or textual data, 2) analyzing the perceived\nemotion or degree of empathy to choose the best course of action, and 3)\ncarrying out a response action. Recent studies that show AE being used with\nvirtual agents or robots often include Deep Learning (DL) techniques. For\ninstance, models like VGGFace are used to conduct ER. Semi-supervised models\nlike Autoencoders generate the corresponding emotional states and behavioral\nresponses. However, there has not been any study that presents an independent\napproach for evaluating AE, or the degree to which a reaction was empathetic.\nThis paper aims to investigate and evaluate existing works for measuring and\nevaluating empathy, as well as the datasets that have been collected and used\nso far. Our goal is to highlight and facilitate the use of state-of-the-art\nmethods in the area of AE by comparing their performance. This will aid\nresearchers in the area of AE in selecting their approaches with precision.\n", "date": "2023-09-04", "categories": []}
{"id": "2310.00283", "title": "Active Learning Based Fine-Tuning Framework for Speech Emotion\n  Recognition", "abstract": "  Speech emotion recognition (SER) has drawn increasing attention for its\napplications in human-machine interaction. However, existing SER methods ignore\nthe information gap between the pre-training speech recognition task and the\ndownstream SER task, leading to sub-optimal performance. Moreover, they require\nmuch time to fine-tune on each specific speech dataset, restricting their\neffectiveness in real-world scenes with large-scale noisy data. To address\nthese issues, we propose an active learning (AL) based Fine-Tuning framework\nfor SER that leverages task adaptation pre-training (TAPT) and AL methods to\nenhance performance and efficiency. Specifically, we first use TAPT to minimize\nthe information gap between the pre-training and the downstream task. Then, AL\nmethods are used to iteratively select a subset of the most informative and\ndiverse samples for fine-tuning, reducing time consumption. Experiments\ndemonstrate that using only 20\\%pt. samples improves 8.45\\%pt. accuracy and\nreduces 79\\%pt. time consumption.\n", "date": "2023-09-30", "categories": []}
{"id": "2310.00287", "title": "InFER: A Multi-Ethnic Indian Facial Expression Recognition Dataset", "abstract": "  The rapid advancement in deep learning over the past decade has transformed\nFacial Expression Recognition (FER) systems, as newer methods have been\nproposed that outperform the existing traditional handcrafted techniques.\nHowever, such a supervised learning approach requires a sufficiently large\ntraining dataset covering all the possible scenarios. And since most people\nexhibit facial expressions based upon their age group, gender, and ethnicity, a\ndiverse facial expression dataset is needed. This becomes even more crucial\nwhile developing a FER system for the Indian subcontinent, which comprises of a\ndiverse multi-ethnic population. In this work, we present InFER, a real-world\nmulti-ethnic Indian Facial Expression Recognition dataset consisting of 10,200\nimages and 4,200 short videos of seven basic facial expressions. The dataset\nhas posed expressions of 600 human subjects, and spontaneous/acted expressions\nof 6000 images crowd-sourced from the internet. To the best of our knowledge\nInFER is the first of its kind consisting of images from 600 subjects from very\ndiverse ethnicity of the Indian Subcontinent. We also present the experimental\nresults of baseline & deep FER methods on our dataset to substantiate its\nusability in real-world practical applications.\n", "date": "2023-09-30", "categories": []}
{"id": "2308.14359", "title": "Effect of Attention and Self-Supervised Speech Embeddings on\n  Non-Semantic Speech Tasks", "abstract": "  Human emotion understanding is pivotal in making conversational technology\nmainstream. We view speech emotion understanding as a perception task which is\na more realistic setting. With varying contexts (languages, demographics, etc.)\ndifferent share of people perceive the same speech segment as a non-unanimous\nemotion. As part of the ACM Multimedia 2023 Computational Paralinguistics\nChallengE (ComParE) in the EMotion Share track, we leverage their rich dataset\nof multilingual speakers and multi-label regression target of 'emotion share'\nor perception of that emotion. We demonstrate that the training scheme of\ndifferent foundation models dictates their effectiveness for tasks beyond\nspeech recognition, especially for non-semantic speech tasks like emotion\nunderstanding. This is a very complex task due to multilingual speakers,\nvariability in the target labels, and inherent imbalance in the regression\ndataset. Our results show that HuBERT-Large with a self-attention-based\nlight-weight sequence model provides 4.6% improvement over the reported\nbaseline.\n", "date": "2023-08-28", "categories": []}
{"id": "2309.13136", "title": "Contextual Emotion Estimation from Image Captions", "abstract": "  Emotion estimation in images is a challenging task, typically using computer\nvision methods to directly estimate people's emotions using face, body pose and\ncontextual cues. In this paper, we explore whether Large Language Models (LLMs)\ncan support the contextual emotion estimation task, by first captioning images,\nthen using an LLM for inference. First, we must understand: how well do LLMs\nperceive human emotions? And which parts of the information enable them to\ndetermine emotions? One initial challenge is to construct a caption that\ndescribes a person within a scene with information relevant for emotion\nperception. Towards this goal, we propose a set of natural language descriptors\nfor faces, bodies, interactions, and environments. We use them to manually\ngenerate captions and emotion annotations for a subset of 331 images from the\nEMOTIC dataset. These captions offer an interpretable representation for\nemotion estimation, towards understanding how elements of a scene affect\nemotion perception in LLMs and beyond. Secondly, we test the capability of a\nlarge language model to infer an emotion from the resulting image captions. We\nfind that GPT-3.5, specifically the text-davinci-003 model, provides\nsurprisingly reasonable emotion predictions consistent with human annotations,\nbut accuracy can depend on the emotion concept. Overall, the results suggest\npromise in the image captioning and LLM approach.\n", "date": "2023-09-22", "categories": []}
{"id": "2309.13214", "title": "Assessing the Impact of Personality on Affective States from Video Game\n  Communication", "abstract": "  Individual differences in personality determine our preferences, traits and\nvalues, which should similarly hold for the way we express ourselves. With\ncurrent advancements and transformations of technology and society, text-based\ncommunication has become ordinary and often even surpasses natural voice\nconversations -- with distinct challenges and opportunities. In this\nexploratory work, we investigate the impact of personality on the tendency how\nplayers of a team-based collaborative alternate reality game express themselves\naffectively. We collected chat logs from eleven players over two weeks, labeled\nthem according to their affective state, and assessed the connection between\nthem and the five-factor personality domains and facets. After applying\nmulti-linear regression, we found a series of reasonable correlations between\n(combinations of) personality variables and expressed affect -- as increased\nconfusion could be predicted by lower self-competence (C1), personal annoyance\nby vulnerability to stress (N6) and expressing anger occured more often in\nplayers that are prone to anxiety (N1), less humble and modest (A5), think less\ncarefully before they act (C6) and have higher neuroticism (N). Expanding the\ndata set, sample size and input modalities in subsequent work, we aim to\nconfirm these findings and reveal even more interesting connections that could\ninform affective computing and games user research equally.\n", "date": "2023-09-22", "categories": []}
{"id": "2309.14104", "title": "Affective Game Computing: A Survey", "abstract": "  This paper surveys the current state of the art in affective computing\nprinciples, methods and tools as applied to games. We review this emerging\nfield, namely affective game computing, through the lens of the four core\nphases of the affective loop: game affect elicitation, game affect sensing,\ngame affect detection and game affect adaptation. In addition, we provide a\ntaxonomy of terms, methods and approaches used across the four phases of the\naffective game loop and situate the field within this taxonomy. We continue\nwith a comprehensive review of available affect data collection methods with\nregards to gaming interfaces, sensors, annotation protocols, and available\ncorpora. The paper concludes with a discussion on the current limitations of\naffective game computing and our vision for the most promising future research\ndirections in the field.\n", "date": "2023-09-25", "categories": []}
{"id": "2309.12714", "title": "Unsupervised Representations Improve Supervised Learning in Speech\n  Emotion Recognition", "abstract": "  Speech Emotion Recognition (SER) plays a pivotal role in enhancing\nhuman-computer interaction by enabling a deeper understanding of emotional\nstates across a wide range of applications, contributing to more empathetic and\neffective communication. This study proposes an innovative approach that\nintegrates self-supervised feature extraction with supervised classification\nfor emotion recognition from small audio segments. In the preprocessing step,\nto eliminate the need of crafting audio features, we employed a self-supervised\nfeature extractor, based on the Wav2Vec model, to capture acoustic features\nfrom audio data. Then, the output featuremaps of the preprocessing step are fed\nto a custom designed Convolutional Neural Network (CNN)-based model to perform\nemotion classification. Utilizing the ShEMO dataset as our testing ground, the\nproposed method surpasses two baseline methods, i.e. support vector machine\nclassifier and transfer learning of a pretrained CNN. comparing the propose\nmethod to the state-of-the-art methods in SER task indicates the superiority of\nthe proposed method. Our findings underscore the pivotal role of deep\nunsupervised feature learning in elevating the landscape of SER, offering\nenhanced emotional comprehension in the realm of human-computer interactions.\n", "date": "2023-09-22", "categories": []}
{"id": "2305.19184", "title": "Leveraging Semantic Information for Efficient Self-Supervised Emotion\n  Recognition with Audio-Textual Distilled Models", "abstract": "  In large part due to their implicit semantic modeling, self-supervised\nlearning (SSL) methods have significantly increased the performance of valence\nrecognition in speech emotion recognition (SER) systems. Yet, their large size\nmay often hinder practical implementations. In this work, we take HuBERT as an\nexample of an SSL model and analyze the relevance of each of its layers for\nSER. We show that shallow layers are more important for arousal recognition\nwhile deeper layers are more important for valence. This observation motivates\nthe importance of additional textual information for accurate valence\nrecognition, as the distilled framework lacks the depth of its large-scale SSL\nteacher. Thus, we propose an audio-textual distilled SSL framework that, while\nhaving only ~20% of the trainable parameters of a large SSL model, achieves on\npar performance across the three emotion dimensions (arousal, valence,\ndominance) on the MSP-Podcast v1.10 dataset.\n", "date": "2023-05-30", "categories": []}
{"id": "2309.09799", "title": "Watch the Speakers: A Hybrid Continuous Attribution Network for Emotion\n  Recognition in Conversation With Emotion Disentanglement", "abstract": "  Emotion Recognition in Conversation (ERC) has attracted widespread attention\nin the natural language processing field due to its enormous potential for\npractical applications. Existing ERC methods face challenges in achieving\ngeneralization to diverse scenarios due to insufficient modeling of context,\nambiguous capture of dialogue relationships and overfitting in speaker\nmodeling. In this work, we present a Hybrid Continuous Attributive Network\n(HCAN) to address these issues in the perspective of emotional continuation and\nemotional attribution. Specifically, HCAN adopts a hybrid recurrent and\nattention-based module to model global emotion continuity. Then a novel\nEmotional Attribution Encoding (EAE) is proposed to model intra- and\ninter-emotional attribution for each utterance. Moreover, aiming to enhance the\nrobustness of the model in speaker modeling and improve its performance in\ndifferent scenarios, A comprehensive loss function emotional cognitive loss\n$\\mathcal{L}_{\\rm EC}$ is proposed to alleviate emotional drift and overcome\nthe overfitting of the model to speaker modeling. Our model achieves\nstate-of-the-art performance on three datasets, demonstrating the superiority\nof our work. Another extensive comparative experiments and ablation studies on\nthree benchmarks are conducted to provided evidence to support the efficacy of\neach module. Further exploration of generalization ability experiments shows\nthe plug-and-play nature of the EAE module in our method.\n", "date": "2023-09-18", "categories": []}
{"id": "2309.06745", "title": "VEATIC: Video-based Emotion and Affect Tracking in Context Dataset", "abstract": "  Human affect recognition has been a significant topic in psychophysics and\ncomputer vision. However, the currently published datasets have many\nlimitations. For example, most datasets contain frames that contain only\ninformation about facial expressions. Due to the limitations of previous\ndatasets, it is very hard to either understand the mechanisms for affect\nrecognition of humans or generalize well on common cases for computer vision\nmodels trained on those datasets. In this work, we introduce a brand new large\ndataset, the Video-based Emotion and Affect Tracking in Context Dataset\n(VEATIC), that can conquer the limitations of the previous datasets. VEATIC has\n124 video clips from Hollywood movies, documentaries, and home videos with\ncontinuous valence and arousal ratings of each frame via real-time annotation.\nAlong with the dataset, we propose a new computer vision task to infer the\naffect of the selected character via both context and character information in\neach video frame. Additionally, we propose a simple model to benchmark this new\ncomputer vision task. We also compare the performance of the pretrained model\nusing our dataset with other similar datasets. Experiments show the competing\nresults of our pretrained model via VEATIC, indicating the generalizability of\nVEATIC. Our dataset is available at https://veatic.github.io.\n", "date": "2023-09-13", "categories": []}
{"id": "2309.07925", "title": "Hierarchical Audio-Visual Information Fusion with Multi-label Joint\n  Decoding for MER 2023", "abstract": "  In this paper, we propose a novel framework for recognizing both discrete and\ndimensional emotions. In our framework, deep features extracted from foundation\nmodels are used as robust acoustic and visual representations of raw video.\nThree different structures based on attention-guided feature gathering (AFG)\nare designed for deep feature fusion. Then, we introduce a joint decoding\nstructure for emotion classification and valence regression in the decoding\nstage. A multi-task loss based on uncertainty is also designed to optimize the\nwhole process. Finally, by combining three different structures on the\nposterior probability level, we obtain the final predictions of discrete and\ndimensional emotions. When tested on the dataset of multimodal emotion\nrecognition challenge (MER 2023), the proposed framework yields consistent\nimprovements in both emotion classification and valence regression. Our final\nsystem achieves state-of-the-art performance and ranks third on the leaderboard\non MER-MULTI sub-challenge.\n", "date": "2023-09-11", "categories": []}
{"id": "2207.06767", "title": "Semi-supervised cross-lingual speech emotion recognition", "abstract": "  Performance in Speech Emotion Recognition (SER) on a single language has\nincreased greatly in the last few years thanks to the use of deep learning\ntechniques. However, cross-lingual SER remains a challenge in real-world\napplications due to two main factors: the first is the big gap among the source\nand the target domain distributions; the second factor is the major\navailability of unlabeled utterances in contrast to the labeled ones for the\nnew language. Taking into account previous aspects, we propose a\nSemi-Supervised Learning (SSL) method for cross-lingual emotion recognition\nwhen only few labeled examples in the target domain (i.e. the new language) are\navailable. Our method is based on a Transformer and it adapts to the new domain\nby exploiting a pseudo-labeling strategy on the unlabeled utterances. In\nparticular, the use of a hard and soft pseudo-labels approach is investigated.\nWe thoroughly evaluate the performance of the proposed method in a\nspeaker-independent setup on both the source and the new language and show its\nrobustness across five languages belonging to different linguistic strains. The\nexperimental findings indicate that the unweighted accuracy is increased by an\naverage of 40% compared to state-of-the-art methods.\n", "date": "2022-07-14", "categories": []}
{"id": "2304.08981", "title": "MER 2023: Multi-label Learning, Modality Robustness, and Semi-Supervised\n  Learning", "abstract": "  The first Multimodal Emotion Recognition Challenge (MER 2023) was\nsuccessfully held at ACM Multimedia. The challenge focuses on system robustness\nand consists of three distinct tracks: (1) MER-MULTI, where participants are\nrequired to recognize both discrete and dimensional emotions; (2) MER-NOISE, in\nwhich noise is added to test videos for modality robustness evaluation; (3)\nMER-SEMI, which provides a large amount of unlabeled samples for\nsemi-supervised learning. In this paper, we introduce the motivation behind\nthis challenge, describe the benchmark dataset, and provide some statistics\nabout participants. To continue using this dataset after MER 2023, please sign\na new End User License Agreement and send it to our official email address\nmerchallenge.contact@gmail.com. We believe this high-quality dataset can become\na new benchmark in multimodal emotion recognition, especially for the Chinese\nresearch community.\n", "date": "2023-04-18", "categories": []}
{"id": "2309.06517", "title": "Overview of Memotion 3: Sentiment and Emotion Analysis of Codemixed\n  Hinglish Memes", "abstract": "  Analyzing memes on the internet has emerged as a crucial endeavor due to the\nimpact this multi-modal form of content wields in shaping online discourse.\nMemes have become a powerful tool for expressing emotions and sentiments,\npossibly even spreading hate and misinformation, through humor and sarcasm. In\nthis paper, we present the overview of the Memotion 3 shared task, as part of\nthe DeFactify 2 workshop at AAAI-23. The task released an annotated dataset of\nHindi-English code-mixed memes based on their Sentiment (Task A), Emotion (Task\nB), and Emotion intensity (Task C). Each of these is defined as an individual\ntask and the participants are ranked separately for each task. Over 50 teams\nregistered for the shared task and 5 made final submissions to the test set of\nthe Memotion 3 dataset. CLIP, BERT modifications, ViT etc. were the most\npopular models among the participants along with approaches such as\nStudent-Teacher model, Fusion, and Ensembling. The best final F1 score for Task\nA is 34.41, Task B is 79.77 and Task C is 59.82.\n", "date": "2023-09-12", "categories": []}
{"id": "2309.05076", "title": "An Appraisal-Based Chain-Of-Emotion Architecture for Affective Language\n  Model Game Agents", "abstract": "  The development of believable, natural, and interactive digital artificial\nagents is a field of growing interest. Theoretical uncertainties and technical\nbarriers present considerable challenges to the field, particularly with\nregards to developing agents that effectively simulate human emotions. Large\nlanguage models (LLMs) might address these issues by tapping common patterns in\nsituational appraisal. In three empirical experiments, this study tests the\ncapabilities of LLMs to solve emotional intelligence tasks and to simulate\nemotions. It presents and evaluates a new chain-of-emotion architecture for\nemotion simulation within video games, based on psychological appraisal\nresearch. Results show that it outperforms standard LLM architectures on a\nrange of user experience and content analysis metrics. This study therefore\nprovides early evidence of how to construct and test affective agents based on\ncognitive processes represented in language models.\n", "date": "2023-09-10", "categories": []}
{"id": "2309.03978", "title": "LanSER: Language-Model Supported Speech Emotion Recognition", "abstract": "  Speech emotion recognition (SER) models typically rely on costly\nhuman-labeled data for training, making scaling methods to large speech\ndatasets and nuanced emotion taxonomies difficult. We present LanSER, a method\nthat enables the use of unlabeled data by inferring weak emotion labels via\npre-trained large language models through weakly-supervised learning. For\ninferring weak labels constrained to a taxonomy, we use a textual entailment\napproach that selects an emotion label with the highest entailment score for a\nspeech transcript extracted via automatic speech recognition. Our experimental\nresults show that models pre-trained on large datasets with this weak\nsupervision outperform other baseline models on standard SER datasets when\nfine-tuned, and show improved label efficiency. Despite being pre-trained on\nlabels derived only from text, we show that the resulting representations\nappear to model the prosodic content of speech.\n", "date": "2023-09-07", "categories": []}
{"id": "2309.04292", "title": "Fuzzy Fingerprinting Transformer Language-Models for Emotion Recognition\n  in Conversations", "abstract": "  Fuzzy Fingerprints have been successfully used as an interpretable text\nclassification technique, but, like most other techniques, have been largely\nsurpassed in performance by Large Pre-trained Language Models, such as BERT or\nRoBERTa. These models deliver state-of-the-art results in several Natural\nLanguage Processing tasks, namely Emotion Recognition in Conversations (ERC),\nbut suffer from the lack of interpretability and explainability. In this paper,\nwe propose to combine the two approaches to perform ERC, as a means to obtain\nsimpler and more interpretable Large Language Models-based classifiers. We\npropose to feed the utterances and their previous conversational turns to a\npre-trained RoBERTa, obtaining contextual embedding utterance representations,\nthat are then supplied to an adapted Fuzzy Fingerprint classification module.\nWe validate our approach on the widely used DailyDialog ERC benchmark dataset,\nin which we obtain state-of-the-art level results using a much lighter model.\n", "date": "2023-09-08", "categories": []}
{"id": "2309.03238", "title": "Implicit Design Choices and Their Impact on Emotion Recognition Model\n  Development and Evaluation", "abstract": "  Emotion recognition is a complex task due to the inherent subjectivity in\nboth the perception and production of emotions. The subjectivity of emotions\nposes significant challenges in developing accurate and robust computational\nmodels. This thesis examines critical facets of emotion recognition, beginning\nwith the collection of diverse datasets that account for psychological factors\nin emotion production.\n  To handle the challenge of non-representative training data, this work\ncollects the Multimodal Stressed Emotion dataset, which introduces controlled\nstressors during data collection to better represent real-world influences on\nemotion production. To address issues with label subjectivity, this research\ncomprehensively analyzes how data augmentation techniques and annotation\nschemes impact emotion perception and annotator labels. It further handles\nnatural confounding variables and variations by employing adversarial networks\nto isolate key factors like stress from learned emotion representations during\nmodel training. For tackling concerns about leakage of sensitive demographic\nvariables, this work leverages adversarial learning to strip sensitive\ndemographic information from multimodal encodings. Additionally, it proposes\noptimized sociological evaluation metrics aligned with cost-effective,\nreal-world needs for model testing.\n  This research advances robust, practical emotion recognition through\nmultifaceted studies of challenges in datasets, labels, modeling, demographic\nand membership variable encoding in representations, and evaluation. The\ngroundwork has been laid for cost-effective, generalizable emotion recognition\nmodels that are less likely to encode sensitive demographic information.\n", "date": "2023-09-06", "categories": []}
{"id": "2303.08356", "title": "Leveraging TCN and Transformer for effective visual-audio fusion in\n  continuous emotion recognition", "abstract": "  Human emotion recognition plays an important role in human-computer\ninteraction. In this paper, we present our approach to the Valence-Arousal (VA)\nEstimation Challenge, Expression (Expr) Classification Challenge, and Action\nUnit (AU) Detection Challenge of the 5th Workshop and Competition on Affective\nBehavior Analysis in-the-wild (ABAW). Specifically, we propose a novel\nmulti-modal fusion model that leverages Temporal Convolutional Networks (TCN)\nand Transformer to enhance the performance of continuous emotion recognition.\nOur model aims to effectively integrate visual and audio information for\nimproved accuracy in recognizing emotions. Our model outperforms the baseline\nand ranks 3 in the Expression Classification challenge.\n", "date": "2023-03-15", "categories": []}
{"id": "2308.04032", "title": "Measure of Uncertainty in Human Emotions", "abstract": "  Many research explore how well computers are able to examine emotions\ndisplayed by humans and use that data to perform different tasks. However,\nthere have been very few research which evaluate the computers ability to\ngenerate emotion classification information in an attempt to help the user make\ndecisions or perform tasks. This is a crucial area to explore as it is\nparamount to the two way communication between humans and computers. This\nresearch conducted an experiment to investigate the impact of different\nuncertainty information displays of emotion classification on the human\ndecision making process. Results show that displaying more uncertainty\ninformation can help users to be more confident when making decisions.\n", "date": "2023-08-08", "categories": []}
{"id": "2309.01164", "title": "Noise robust speech emotion recognition with signal-to-noise ratio\n  adapting speech enhancement", "abstract": "  Speech emotion recognition (SER) often experiences reduced performance due to\nbackground noise. In addition, making a prediction on signals with only\nbackground noise could undermine user trust in the system. In this study, we\npropose a Noise Robust Speech Emotion Recognition system, NRSER. NRSER employs\nspeech enhancement (SE) to effectively reduce the noise in input signals. Then,\nthe signal-to-noise-ratio (SNR)-level detection structure and waveform\nreconstitution strategy are introduced to reduce the negative impact of SE on\nspeech signals with no or little background noise. Our experimental results\nshow that NRSER can effectively improve the noise robustness of the SER system,\nincluding preventing the system from making emotion recognition on signals\nconsisting solely of background noise. Moreover, the proposed SNR-level\ndetection structure can be used individually for tasks such as data selection.\n", "date": "2023-09-03", "categories": []}
{"id": "2309.01339", "title": "UniSA: Unified Generative Framework for Sentiment Analysis", "abstract": "  Sentiment analysis is a crucial task that aims to understand people's\nemotional states and predict emotional categories based on multimodal\ninformation. It consists of several subtasks, such as emotion recognition in\nconversation (ERC), aspect-based sentiment analysis (ABSA), and multimodal\nsentiment analysis (MSA). However, unifying all subtasks in sentiment analysis\npresents numerous challenges, including modality alignment, unified\ninput/output forms, and dataset bias. To address these challenges, we propose a\nTask-Specific Prompt method to jointly model subtasks and introduce a\nmultimodal generative framework called UniSA. Additionally, we organize the\nbenchmark datasets of main subtasks into a new Sentiment Analysis Evaluation\nbenchmark, SAEval. We design novel pre-training tasks and training methods to\nenable the model to learn generic sentiment knowledge among subtasks to improve\nthe model's multimodal sentiment perception ability. Our experimental results\nshow that UniSA performs comparably to the state-of-the-art on all subtasks and\ngeneralizes well to various subtasks in sentiment analysis.\n", "date": "2023-09-04", "categories": []}
{"id": "2309.01664", "title": "Fine-grained Affective Processing Capabilities Emerging from Large\n  Language Models", "abstract": "  Large language models, in particular generative pre-trained transformers\n(GPTs), show impressive results on a wide variety of language-related tasks. In\nthis paper, we explore ChatGPT's zero-shot ability to perform affective\ncomputing tasks using prompting alone. We show that ChatGPT a) performs\nmeaningful sentiment analysis in the Valence, Arousal and Dominance dimensions,\nb) has meaningful emotion representations in terms of emotion categories and\nthese affective dimensions, and c) can perform basic appraisal-based emotion\nelicitation of situations based on a prompt-based computational implementation\nof the OCC appraisal model. These findings are highly relevant: First, they\nshow that the ability to solve complex affect processing tasks emerges from\nlanguage-based token prediction trained on extensive data sets. Second, they\nshow the potential of large language models for simulating, processing and\nanalyzing human emotions, which has important implications for various\napplications such as sentiment analysis, socially interactive agents, and\nsocial robotics.\n", "date": "2023-09-04", "categories": []}
{"id": "2309.02106", "title": "Leveraging Label Information for Multimodal Emotion Recognition", "abstract": "  Multimodal emotion recognition (MER) aims to detect the emotional status of a\ngiven expression by combining the speech and text information. Intuitively,\nlabel information should be capable of helping the model locate the salient\ntokens/frames relevant to the specific emotion, which finally facilitates the\nMER task. Inspired by this, we propose a novel approach for MER by leveraging\nlabel information. Specifically, we first obtain the representative label\nembeddings for both text and speech modalities, then learn the label-enhanced\ntext/speech representations for each utterance via label-token and label-frame\ninteractions. Finally, we devise a novel label-guided attentive fusion module\nto fuse the label-aware text and speech representations for emotion\nclassification. Extensive experiments were conducted on the public IEMOCAP\ndataset, and experimental results demonstrate that our proposed approach\noutperforms existing baselines and achieves new state-of-the-art performance.\n", "date": "2023-09-05", "categories": []}
{"id": "2308.11029", "title": "RBA-GCN: Relational Bilevel Aggregation Graph Convolutional Network for\n  Emotion Recognition", "abstract": "  Emotion recognition in conversation (ERC) has received increasing attention\nfrom researchers due to its wide range of applications.As conversation has a\nnatural graph structure,numerous approaches used to model ERC based on graph\nconvolutional networks (GCNs) have yielded significant results.However,the\naggregation approach of traditional GCNs suffers from the node information\nredundancy problem,leading to node discriminant information\nloss.Additionally,single-layer GCNs lack the capacity to capture long-range\ncontextual information from the graph. Furthermore,the majority of approaches\nare based on textual modality or stitching together different modalities,\nresulting in a weak ability to capture interactions between modalities. To\naddress these problems, we present the relational bilevel aggregation graph\nconvolutional network (RBA-GCN), which consists of three modules: the graph\ngeneration module (GGM), similarity-based cluster building module (SCBM) and\nbilevel aggregation module (BiAM). First, GGM constructs a novel graph to\nreduce the redundancy of target node information.Then,SCBM calculates the node\nsimilarity in the target node and its structural neighborhood, where noisy\ninformation with low similarity is filtered out to preserve the discriminant\ninformation of the node. Meanwhile, BiAM is a novel aggregation method that can\npreserve the information of nodes during the aggregation process. This module\ncan construct the interaction between different modalities and capture\nlong-range contextual information based on similarity clusters. On both the\nIEMOCAP and MELD datasets, the weighted average F1 score of RBA-GCN has a\n2.17$\\sim$5.21\\% improvement over that of the most advanced method.Our code is\navailable at https://github.com/luftmenscher/RBA-GCN and our article with the\nsame name has been published in IEEE/ACM Transactions on Audio,Speech,and\nLanguage Processing,vol.31,2023\n", "date": "2023-08-18", "categories": []}
{"id": "2308.14894", "title": "Multiscale Contextual Learning for Speech Emotion Recognition in\n  Emergency Call Center Conversations", "abstract": "  Emotion recognition in conversations is essential for ensuring advanced\nhuman-machine interactions. However, creating robust and accurate emotion\nrecognition systems in real life is challenging, mainly due to the scarcity of\nemotion datasets collected in the wild and the inability to take into account\nthe dialogue context. The CEMO dataset, composed of conversations between\nagents and patients during emergency calls to a French call center, fills this\ngap. The nature of these interactions highlights the role of the emotional flow\nof the conversation in predicting patient emotions, as context can often make a\ndifference in understanding actual feelings. This paper presents a multi-scale\nconversational context learning approach for speech emotion recognition, which\ntakes advantage of this hypothesis. We investigated this approach on both\nspeech transcriptions and acoustic segments. Experimentally, our method uses\nthe previous or next information of the targeted segment. In the text domain,\nwe tested the context window using a wide range of tokens (from 10 to 100) and\nat the speech turns level, considering inputs from both the same and opposing\nspeakers. According to our tests, the context derived from previous tokens has\na more significant influence on accurate prediction than the following tokens.\nFurthermore, taking the last speech turn of the same speaker in the\nconversation seems useful. In the acoustic domain, we conducted an in-depth\nanalysis of the impact of the surrounding emotions on the prediction. While\nmulti-scale conversational context learning using Transformers can enhance\nperformance in the textual modality for emergency call recordings,\nincorporating acoustic context is more challenging.\n", "date": "2023-08-28", "categories": []}
{"id": "2308.13911", "title": "A Wide Evaluation of ChatGPT on Affective Computing Tasks", "abstract": "  With the rise of foundation models, a new artificial intelligence paradigm\nhas emerged, by simply using general purpose foundation models with prompting\nto solve problems instead of training a separate machine learning model for\neach problem. Such models have been shown to have emergent properties of\nsolving problems that they were not initially trained on. The studies for the\neffectiveness of such models are still quite limited. In this work, we widely\nstudy the capabilities of the ChatGPT models, namely GPT-4 and GPT-3.5, on 13\naffective computing problems, namely aspect extraction, aspect polarity\nclassification, opinion extraction, sentiment analysis, sentiment intensity\nranking, emotions intensity ranking, suicide tendency detection, toxicity\ndetection, well-being assessment, engagement measurement, personality\nassessment, sarcasm detection, and subjectivity detection. We introduce a\nframework to evaluate the ChatGPT models on regression-based problems, such as\nintensity ranking problems, by modelling them as pairwise ranking\nclassification. We compare ChatGPT against more traditional NLP methods, such\nas end-to-end recurrent neural networks and transformers. The results\ndemonstrate the emergent abilities of the ChatGPT models on a wide range of\naffective computing problems, where GPT-3.5 and especially GPT-4 have shown\nstrong performance on many problems, particularly the ones related to\nsentiment, emotions, or toxicity. The ChatGPT models fell short for problems\nwith implicit signals, such as engagement measurement and subjectivity\ndetection.\n", "date": "2023-08-26", "categories": []}
{"id": "2308.14160", "title": "A Unified Transformer-based Network for multimodal Emotion Recognition", "abstract": "  The development of transformer-based models has resulted in significant\nadvances in addressing various vision and NLP-based research challenges.\nHowever, the progress made in transformer-based methods has not been\neffectively applied to biosensing research. This paper presents a novel Unified\nBiosensor-Vision Multi-modal Transformer-based (UBVMT) method to classify\nemotions in an arousal-valence space by combining a 2D representation of an\nECG/PPG signal with the face information. To achieve this goal, we first\ninvestigate and compare the unimodal emotion recognition performance of three\nimage-based representations of the ECG/PPG signal. We then present our UBVMT\nnetwork which is trained to perform emotion recognition by combining the 2D\nimage-based representation of the ECG/PPG signal and the facial expression\nfeatures. Our unified transformer model consists of homogeneous transformer\nblocks that take as an input the 2D representation of the ECG/PPG signal and\nthe corresponding face frame for emotion representation learning with minimal\nmodality-specific design. Our UBVMT model is trained by reconstructing masked\npatches of video frames and 2D images of ECG/PPG signals, and contrastive\nmodeling to align face and ECG/PPG data. Extensive experiments on the\nMAHNOB-HCI and DEAP datasets show that our Unified UBVMT-based model produces\ncomparable results to the state-of-the-art techniques.\n", "date": "2023-08-27", "categories": []}
{"id": "2308.14245", "title": "A Comparison of Personalized and Generalized Approaches to Emotion\n  Recognition Using Consumer Wearable Devices: Machine Learning Study", "abstract": "  Background: Studies have shown the potential adverse health effects, ranging\nfrom headaches to cardiovascular disease, associated with long-term negative\nemotions and chronic stress. Since many indicators of stress are imperceptible\nto observers, the early detection and intervention of stress remains a pressing\nmedical need. Physiological signals offer a non-invasive method of monitoring\nemotions and are easily collected by smartwatches. Existing research primarily\nfocuses on developing generalized machine learning-based models for emotion\nclassification. Objective: We aim to study the differences between personalized\nand generalized machine learning models for three-class emotion classification\n(neutral, stress, and amusement) using wearable biosignal data. Methods: We\ndeveloped a convolutional encoder for the three-class emotion classification\nproblem using data from WESAD, a multimodal dataset with physiological signals\nfor 15 subjects. We compared the results between a subject-exclusive\ngeneralized, subject-inclusive generalized, and personalized model. Results:\nFor the three-class classification problem, our personalized model achieved an\naverage accuracy of 95.06% and F1-score of 91.71, our subject-inclusive\ngeneralized model achieved an average accuracy of 66.95% and F1-score of 42.50,\nand our subject-exclusive generalized model achieved an average accuracy of\n67.65% and F1-score of 43.05. Conclusions: Our results emphasize the need for\nincreased research in personalized emotion recognition models given that they\noutperform generalized models in certain contexts. We also demonstrate that\npersonalized machine learning models for emotion classification are viable and\ncan achieve high performance.\n", "date": "2023-08-28", "categories": []}
{"id": "2308.10713", "title": "LibreFace: An Open-Source Toolkit for Deep Facial Expression Analysis", "abstract": "  Facial expression analysis is an important tool for human-computer\ninteraction. In this paper, we introduce LibreFace, an open-source toolkit for\nfacial expression analysis. This open-source toolbox offers real-time and\noffline analysis of facial behavior through deep learning models, including\nfacial action unit (AU) detection, AU intensity estimation, and facial\nexpression recognition. To accomplish this, we employ several techniques,\nincluding the utilization of a large-scale pre-trained network, feature-wise\nknowledge distillation, and task-specific fine-tuning. These approaches are\ndesigned to effectively and accurately analyze facial expressions by leveraging\nvisual information, thereby facilitating the implementation of real-time\ninteractive applications. In terms of Action Unit (AU) intensity estimation, we\nachieve a Pearson Correlation Coefficient (PCC) of 0.63 on DISFA, which is 7%\nhigher than the performance of OpenFace 2.0 while maintaining highly-efficient\ninference that runs two times faster than OpenFace 2.0. Despite being compact,\nour model also demonstrates competitive performance to state-of-the-art facial\nexpression analysis methods on AffecNet, FFHQ, and RAF-DB. Our code will be\nreleased at https://github.com/ihp-lab/LibreFace\n", "date": "2023-08-18", "categories": []}
{"id": "2308.13421", "title": "Exploiting Diverse Feature for Multimodal Sentiment Analysis", "abstract": "  In this paper, we present our solution to the MuSe-Personalisation\nsub-challenge in the MuSe 2023 Multimodal Sentiment Analysis Challenge. The\ntask of MuSe-Personalisation aims to predict the continuous arousal and valence\nvalues of a participant based on their audio-visual, language, and\nphysiological signal modalities data. Considering different people have\npersonal characteristics, the main challenge of this task is how to build\nrobustness feature presentation for sentiment prediction. To address this\nissue, we propose exploiting diverse features. Specifically, we proposed a\nseries of feature extraction methods to build a robust representation and model\nensemble. We empirically evaluate the performance of the utilized method on the\nofficially provided dataset. \\textbf{As a result, we achieved 3rd place in the\nMuSe-Personalisation sub-challenge.} Specifically, we achieve the results of\n0.8492 and 0.8439 for MuSe-Personalisation in terms of arousal and valence CCC.\n", "date": "2023-08-25", "categories": []}
{"id": "2308.11673", "title": "WEARS: Wearable Emotion AI with Real-time Sensor data", "abstract": "  Emotion prediction is the field of study to understand human emotions.\nExisting methods focus on modalities like text, audio, facial expressions,\netc., which could be private to the user. Emotion can be derived from the\nsubject's psychological data as well. Various approaches that employ\ncombinations of physiological sensors for emotion recognition have been\nproposed. Yet, not all sensors are simple to use and handy for individuals in\ntheir daily lives. Thus, we propose a system to predict user emotion using\nsmartwatch sensors. We design a framework to collect ground truth in real-time\nutilizing a mix of English and regional language-based videos to invoke\nemotions in participants and collect the data. Further, we modeled the problem\nas binary classification due to the limited dataset size and experimented with\nmultiple machine-learning models. We also did an ablation study to understand\nthe impact of features including Heart Rate, Accelerometer, and Gyroscope\nsensor data on mood. From the experimental results, Multi-Layer Perceptron has\nshown a maximum accuracy of 93.75 percent for pleasant-unpleasant (high/low\nvalence classification) moods.\n", "date": "2023-08-22", "categories": []}
{"id": "2308.12547", "title": "Hybrid Models for Facial Emotion Recognition in Children", "abstract": "  This paper focuses on the use of emotion recognition techniques to assist\npsychologists in performing children's therapy through remotely robot operated\nsessions. In the field of psychology, the use of agent-mediated therapy is\ngrowing increasingly given recent advances in robotics and computer science.\nSpecifically, the use of Embodied Conversational Agents (ECA) as an\nintermediary tool can help professionals connect with children who face social\nchallenges such as Attention Deficit Hyperactivity Disorder (ADHD), Autism\nSpectrum Disorder (ASD) or even who are physically unavailable due to being in\nregions of armed conflict, natural disasters, or other circumstances. In this\ncontext, emotion recognition represents an important feedback for the\npsychotherapist. In this article, we initially present the result of a\nbibliographical research associated with emotion recognition in children. This\nresearch revealed an initial overview on algorithms and datasets widely used by\nthe community. Then, based on the analysis carried out on the results of the\nbibliographical research, we used the technique of dense optical flow features\nto improve the ability of identifying emotions in children in uncontrolled\nenvironments. From the output of a hybrid model of Convolutional Neural\nNetwork, two intermediary features are fused before being processed by a final\nclassifier. The proposed architecture was called HybridCNNFusion. Finally, we\npresent the initial results achieved in the recognition of children's emotions\nusing a dataset of Brazilian children.\n", "date": "2023-08-24", "categories": []}
{"id": "2308.12648", "title": "From Chatter to Matter: Addressing Critical Steps of Emotion Recognition\n  Learning in Task-oriented Dialogue", "abstract": "  Emotion recognition in conversations (ERC) is a crucial task for building\nhuman-like conversational agents. While substantial efforts have been devoted\nto ERC for chit-chat dialogues, the task-oriented counterpart is largely left\nunattended. Directly applying chit-chat ERC models to task-oriented dialogues\n(ToDs) results in suboptimal performance as these models overlook key features\nsuch as the correlation between emotions and task completion in ToDs. In this\npaper, we propose a framework that turns a chit-chat ERC model into a\ntask-oriented one, addressing three critical aspects: data, features and\nobjective. First, we devise two ways of augmenting rare emotions to improve ERC\nperformance. Second, we use dialogue states as auxiliary features to\nincorporate key information from the goal of the user. Lastly, we leverage a\nmulti-aspect emotion definition in ToDs to devise a multi-task learning\nobjective and a novel emotion-distance weighted loss function. Our framework\nyields significant improvements for a range of chit-chat ERC models on EmoWOZ,\na large-scale dataset for user emotion in ToDs. We further investigate the\ngeneralisability of the best resulting model to predict user satisfaction in\ndifferent ToD datasets. A comparison with supervised baselines shows a strong\nzero-shot capability, highlighting the potential usage of our framework in\nwider scenarios.\n", "date": "2023-08-24", "categories": []}
{"id": "2308.12156", "title": "Multimodal Latent Emotion Recognition from Micro-expression and\n  Physiological Signals", "abstract": "  This paper discusses the benefits of incorporating multimodal data for\nimproving latent emotion recognition accuracy, focusing on micro-expression\n(ME) and physiological signals (PS). The proposed approach presents a novel\nmultimodal learning framework that combines ME and PS, including a 1D separable\nand mixable depthwise inception network, a standardised normal distribution\nweighted feature fusion method, and depth/physiology guided attention modules\nfor multimodal learning. Experimental results show that the proposed approach\noutperforms the benchmark method, with the weighted fusion method and guided\nattention modules both contributing to enhanced performance.\n", "date": "2023-08-23", "categories": []}
{"id": "2308.11578", "title": "Refashioning Emotion Recognition Modelling: The Advent of Generalised\n  Large Models", "abstract": "  After the inception of emotion recognition or affective computing, it has\nincreasingly become an active research topic due to its broad applications.\nOver the past couple of decades, emotion recognition models have gradually\nmigrated from statistically shallow models to neural network-based deep models,\nwhich can significantly boost the performance of emotion recognition models and\nconsistently achieve the best results on different benchmarks. Therefore, in\nrecent years, deep models have always been considered the first option for\nemotion recognition. However, the debut of large language models (LLMs), such\nas ChatGPT, has remarkably astonished the world due to their emerged\ncapabilities of zero/few-shot learning, in-context learning, chain-of-thought,\nand others that are never shown in previous deep models. In the present paper,\nwe comprehensively investigate how the LLMs perform in emotion recognition in\nterms of diverse aspects, including in-context learning, few-short learning,\naccuracy, generalisation, and explanation. Moreover, we offer some insights and\npose other potential challenges, hoping to ignite broader discussions about\nenhancing emotion recognition in the new era of advanced and generalised large\nmodels.\n", "date": "2023-08-21", "categories": []}
{"id": "2308.11584", "title": "Building Emotional Support Chatbots in the Era of LLMs", "abstract": "  The integration of emotional support into various conversational scenarios\npresents profound societal benefits, such as social interactions, mental health\ncounseling, and customer service. However, there are unsolved challenges that\nhinder real-world applications in this field, including limited data\navailability and the absence of well-accepted model training paradigms. This\nwork endeavors to navigate these challenges by harnessing the capabilities of\nLarge Language Models (LLMs). We introduce an innovative methodology that\nsynthesizes human insights with the computational prowess of LLMs to curate an\nextensive emotional support dialogue dataset. Our approach is initiated with a\nmeticulously designed set of dialogues spanning diverse scenarios as generative\nseeds. By utilizing the in-context learning potential of ChatGPT, we\nrecursively generate an ExTensible Emotional Support dialogue dataset, named\nExTES. Following this, we deploy advanced tuning techniques on the LLaMA model,\nexamining the impact of diverse training strategies, ultimately yielding an LLM\nmeticulously optimized for emotional support interactions. An exhaustive\nassessment of the resultant model showcases its proficiency in offering\nemotional support, marking a pivotal step in the realm of emotional support\nbots and paving the way for subsequent research and implementations.\n", "date": "2023-08-17", "categories": []}
{"id": "2306.13592", "title": "TACOformer:Token-channel compounded Cross Attention for Multimodal\n  Emotion Recognition", "abstract": "  Recently, emotion recognition based on physiological signals has emerged as a\nfield with intensive research. The utilization of multi-modal, multi-channel\nphysiological signals has significantly improved the performance of emotion\nrecognition systems, due to their complementarity. However, effectively\nintegrating emotion-related semantic information from different modalities and\ncapturing inter-modal dependencies remains a challenging issue. Many existing\nmultimodal fusion methods ignore either token-to-token or channel-to-channel\ncorrelations of multichannel signals from different modalities, which limits\nthe classification capability of the models to some extent. In this paper, we\npropose a comprehensive perspective of multimodal fusion that integrates\nchannel-level and token-level cross-modal interactions. Specifically, we\nintroduce a unified cross attention module called Token-chAnnel COmpound (TACO)\nCross Attention to perform multimodal fusion, which simultaneously models\nchannel-level and token-level dependencies between modalities. Additionally, we\npropose a 2D position encoding method to preserve information about the spatial\ndistribution of EEG signal channels, then we use two transformer encoders ahead\nof the fusion module to capture long-term temporal dependencies from the EEG\nsignal and the peripheral physiological signal, respectively.\nSubject-independent experiments on emotional dataset DEAP and Dreamer\ndemonstrate that the proposed model achieves state-of-the-art performance.\n", "date": "2023-06-23", "categories": []}
{"id": "2308.10869", "title": "A Novel Loss Function Utilizing Wasserstein Distance to Reduce\n  Subject-Dependent Noise for Generalizable Models in Affective Computing", "abstract": "  Emotions are an essential part of human behavior that can impact thinking,\ndecision-making, and communication skills. Thus, the ability to accurately\nmonitor and identify emotions can be useful in many human-centered applications\nsuch as behavioral training, tracking emotional well-being, and development of\nhuman-computer interfaces. The correlation between patterns in physiological\ndata and affective states has allowed for the utilization of deep learning\ntechniques which can accurately detect the affective states of a person.\nHowever, the generalisability of existing models is often limited by the\nsubject-dependent noise in the physiological data due to variations in a\nsubject's reactions to stimuli. Hence, we propose a novel cost function that\nemploys Optimal Transport Theory, specifically Wasserstein Distance, to scale\nthe importance of subject-dependent data such that higher importance is\nassigned to patterns in data that are common across all participants while\ndecreasing the importance of patterns that result from subject-dependent noise.\nThe performance of the proposed cost function is demonstrated through an\nautoencoder with a multi-class classifier attached to the latent space and\ntrained simultaneously to detect different affective states. An autoencoder\nwith a state-of-the-art loss function i.e., Mean Squared Error, is used as a\nbaseline for comparison with our model across four different commonly used\ndatasets. Centroid and minimum distance between different classes are used as a\nmetrics to indicate the separation between different classes in the latent\nspace. An average increase of 14.75% and 17.75% (from benchmark to proposed\nloss function) was found for minimum and centroid euclidean distance\nrespectively over all datasets.\n", "date": "2023-08-17", "categories": []}
{"id": "2302.00431", "title": "Do I Have Your Attention: A Large Scale Engagement Prediction Dataset\n  and Baselines", "abstract": "  The degree of concentration, enthusiasm, optimism, and passion displayed by\nindividual(s) while interacting with a machine is referred to as `user\nengagement'. Engagement comprises of behavioral, cognitive, and affect related\ncues. To create engagement prediction systems that can work in real-world\nconditions, it is quintessential to learn from rich, diverse datasets. To this\nend, a large scale multi-faceted engagement in the wild dataset EngageNet is\nproposed. 31 hours duration data of 127 participants representing different\nillumination conditions are recorded. Thorough experiments are performed\nexploring the applicability of different features, action units, eye gaze, head\npose, and MARLIN. Data from user interactions (question-answer) are analyzed to\nunderstand the relationship between effective learning and user engagement. To\nfurther validate the rich nature of the dataset, evaluation is also performed\non the EngageWild dataset. The experiments show the usefulness of the proposed\ndataset. The code, models, and dataset link are publicly available at\nhttps://github.com/engagenet/engagenet_baselines.\n", "date": "2023-02-01", "categories": []}
{"id": "2306.06979", "title": "A Weakly Supervised Approach to Emotion-change Prediction and Improved\n  Mood Inference", "abstract": "  Whilst a majority of affective computing research focuses on inferring\nemotions, examining mood or understanding the \\textit{mood-emotion interplay}\nhas received significantly less attention. Building on prior work, we (a)\ndeduce and incorporate emotion-change ($\\Delta$) information for inferring\nmood, without resorting to annotated labels, and (b) attempt mood prediction\nfor long duration video clips, in alignment with the characterisation of mood.\nWe generate the emotion-change ($\\Delta$) labels via metric learning from a\npre-trained Siamese Network, and use these in addition to mood labels for mood\nclassification. Experiments evaluating \\textit{unimodal} (training only using\nmood labels) vs \\textit{multimodal} (training using mood plus $\\Delta$ labels)\nmodels show that mood prediction benefits from the incorporation of\nemotion-change information, emphasising the importance of modelling the\nmood-emotion interplay for effective mood inference.\n", "date": "2023-06-12", "categories": []}
{"id": "2308.08713", "title": "Decoding Emotions: A comprehensive Multilingual Study of Speech Models\n  for Speech Emotion Recognition", "abstract": "  Recent advancements in transformer-based speech representation models have\ngreatly transformed speech processing. However, there has been limited research\nconducted on evaluating these models for speech emotion recognition (SER)\nacross multiple languages and examining their internal representations. This\narticle addresses these gaps by presenting a comprehensive benchmark for SER\nwith eight speech representation models and six different languages. We\nconducted probing experiments to gain insights into inner workings of these\nmodels for SER. We find that using features from a single optimal layer of a\nspeech model reduces the error rate by 32\\% on average across seven datasets\nwhen compared to systems where features from all layers of speech models are\nused. We also achieve state-of-the-art results for German and Persian\nlanguages. Our probing results indicate that the middle layers of speech models\ncapture the most important emotional information for speech emotion\nrecognition.\n", "date": "2023-08-17", "categories": []}
{"id": "2211.15377", "title": "Whose Emotion Matters? Speaking Activity Localisation without Prior\n  Knowledge", "abstract": "  The task of emotion recognition in conversations (ERC) benefits from the\navailability of multiple modalities, as provided, for example, in the\nvideo-based Multimodal EmotionLines Dataset (MELD). However, only a few\nresearch approaches use both acoustic and visual information from the MELD\nvideos. There are two reasons for this: First, label-to-video alignments in\nMELD are noisy, making those videos an unreliable source of emotional speech\ndata. Second, conversations can involve several people in the same scene, which\nrequires the localisation of the utterance source. In this paper, we introduce\nMELD with Fixed Audiovisual Information via Realignment (MELD-FAIR) by using\nrecent active speaker detection and automatic speech recognition models, we are\nable to realign the videos of MELD and capture the facial expressions from\nspeakers in 96.92% of the utterances provided in MELD. Experiments with a\nself-supervised voice recognition model indicate that the realigned MELD-FAIR\nvideos more closely match the transcribed utterances given in the MELD dataset.\nFinally, we devise a model for emotion recognition in conversations trained on\nthe realigned MELD-FAIR videos, which outperforms state-of-the-art models for\nERC based on vision alone. This indicates that localising the source of\nspeaking activities is indeed effective for extracting facial expressions from\nthe uttering speakers and that faces provide more informative visual cues than\nthe visual features state-of-the-art models have been using so far. The\nMELD-FAIR realignment data, and the code of the realignment procedure and of\nthe emotional recognition, are available at\nhttps://github.com/knowledgetechnologyuhh/MELD-FAIR.\n", "date": "2022-11-23", "categories": []}
{"id": "2308.07871", "title": "Emotion Embeddings $\\unicode{x2014}$ Learning Stable and Homogeneous\n  Abstractions from Heterogeneous Affective Datasets", "abstract": "  Human emotion is expressed in many communication modalities and media formats\nand so their computational study is equally diversified into natural language\nprocessing, audio signal analysis, computer vision, etc. Similarly, the large\nvariety of representation formats used in previous research to describe\nemotions (polarity scales, basic emotion categories, dimensional approaches,\nappraisal theory, etc.) have led to an ever proliferating diversity of\ndatasets, predictive models, and software tools for emotion analysis. Because\nof these two distinct types of heterogeneity, at the expressional and\nrepresentational level, there is a dire need to unify previous work on\nincreasingly diverging data and label types. This article presents such a\nunifying computational model. We propose a training procedure that learns a\nshared latent representation for emotions, so-called emotion embeddings,\nindependent of different natural languages, communication modalities, media or\nrepresentation label formats, and even disparate model architectures.\nExperiments on a wide range of heterogeneous affective datasets indicate that\nthis approach yields the desired interoperability for the sake of reusability,\ninterpretability and flexibility, without penalizing prediction quality. Code\nand data are archived under https://doi.org/10.5281/zenodo.7405327 .\n", "date": "2023-08-15", "categories": []}
{"id": "2208.00847", "title": "MAFW: A Large-scale, Multi-modal, Compound Affective Database for\n  Dynamic Facial Expression Recognition in the Wild", "abstract": "  Dynamic facial expression recognition (FER) databases provide important data\nsupport for affective computing and applications. However, most FER databases\nare annotated with several basic mutually exclusive emotional categories and\ncontain only one modality, e.g., videos. The monotonous labels and modality\ncannot accurately imitate human emotions and fulfill applications in the real\nworld. In this paper, we propose MAFW, a large-scale multi-modal compound\naffective database with 10,045 video-audio clips in the wild. Each clip is\nannotated with a compound emotional category and a couple of sentences that\ndescribe the subjects' affective behaviors in the clip. For the compound\nemotion annotation, each clip is categorized into one or more of the 11\nwidely-used emotions, i.e., anger, disgust, fear, happiness, neutral, sadness,\nsurprise, contempt, anxiety, helplessness, and disappointment. To ensure high\nquality of the labels, we filter out the unreliable annotations by an\nExpectation Maximization (EM) algorithm, and then obtain 11 single-label\nemotion categories and 32 multi-label emotion categories. To the best of our\nknowledge, MAFW is the first in-the-wild multi-modal database annotated with\ncompound emotion annotations and emotion-related captions. Additionally, we\nalso propose a novel Transformer-based expression snippet feature learning\nmethod to recognize the compound emotions leveraging the expression-change\nrelations among different emotions and modalities. Extensive experiments on\nMAFW database show the advantages of the proposed method over other\nstate-of-the-art methods for both uni- and multi-modal FER. Our MAFW database\nis publicly available from https://mafw-database.github.io/MAFW.\n", "date": "2022-08-01", "categories": []}
{"id": "2308.04502", "title": "Revisiting Disentanglement and Fusion on Modality and Context in\n  Conversational Multimodal Emotion Recognition", "abstract": "  It has been a hot research topic to enable machines to understand human\nemotions in multimodal contexts under dialogue scenarios, which is tasked with\nmultimodal emotion analysis in conversation (MM-ERC). MM-ERC has received\nconsistent attention in recent years, where a diverse range of methods has been\nproposed for securing better task performance. Most existing works treat MM-ERC\nas a standard multimodal classification problem and perform multimodal feature\ndisentanglement and fusion for maximizing feature utility. Yet after revisiting\nthe characteristic of MM-ERC, we argue that both the feature multimodality and\nconversational contextualization should be properly modeled simultaneously\nduring the feature disentanglement and fusion steps. In this work, we target\nfurther pushing the task performance by taking full consideration of the above\ninsights. On the one hand, during feature disentanglement, based on the\ncontrastive learning technique, we devise a Dual-level Disentanglement\nMechanism (DDM) to decouple the features into both the modality space and\nutterance space. On the other hand, during the feature fusion stage, we propose\na Contribution-aware Fusion Mechanism (CFM) and a Context Refusion Mechanism\n(CRM) for multimodal and context integration, respectively. They together\nschedule the proper integrations of multimodal and context features.\nSpecifically, CFM explicitly manages the multimodal feature contributions\ndynamically, while CRM flexibly coordinates the introduction of dialogue\ncontexts. On two public MM-ERC datasets, our system achieves new\nstate-of-the-art performance consistently. Further analyses demonstrate that\nall our proposed mechanisms greatly facilitate the MM-ERC task by making full\nuse of the multimodal and context features adaptively. Note that our proposed\nmethods have the great potential to facilitate a broader range of other\nconversational multimodal tasks.\n", "date": "2023-08-08", "categories": []}
{"id": "2308.06549", "title": "Human Behavior-based Personalized Meal Recommendation and Menu Planning\n  Social System", "abstract": "  The traditional dietary recommendation systems are basically nutrition or\nhealth-aware where the human feelings on food are ignored. Human affects vary\nwhen it comes to food cravings, and not all foods are appealing in all moods. A\nquestionnaire-based and preference-aware meal recommendation system can be a\nsolution. However, automated recognition of social affects on different foods\nand planning the menu considering nutritional demand and social-affect has some\nsignificant benefits of the questionnaire-based and preference-aware meal\nrecommendations. A patient with severe illness, a person in a coma, or patients\nwith locked-in syndrome and amyotrophic lateral sclerosis (ALS) cannot express\ntheir meal preferences. Therefore, the proposed framework includes a\nsocial-affective computing module to recognize the affects of different meals\nwhere the person's affect is detected using electroencephalography signals. EEG\nallows to capture the brain signals and analyze them to anticipate affective\ntoward a food. In this study, we have used a 14-channel wireless Emotive Epoc+\nto measure affectivity for different food items. A hierarchical ensemble method\nis applied to predict affectivity upon multiple feature extraction methods and\nTOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) is\nused to generate a food list based on the predicted affectivity. In addition to\nthe meal recommendation, an automated menu planning approach is also proposed\nconsidering a person's energy intake requirement, affectivity, and nutritional\nvalues of the different menus. The bin-packing algorithm is used for the\npersonalized menu planning of breakfast, lunch, dinner, and snacks. The\nexperimental findings reveal that the suggested affective computing, meal\nrecommendation, and menu planning algorithms perform well across a variety of\nassessment parameters.\n", "date": "2023-08-12", "categories": []}
{"id": "2308.05767", "title": "EEG-based Emotion Style Transfer Network for Cross-dataset Emotion\n  Recognition", "abstract": "  As the key to realizing aBCIs, EEG emotion recognition has been widely\nstudied by many researchers. Previous methods have performed well for\nintra-subject EEG emotion recognition. However, the style mismatch between\nsource domain (training data) and target domain (test data) EEG samples caused\nby huge inter-domain differences is still a critical problem for EEG emotion\nrecognition. To solve the problem of cross-dataset EEG emotion recognition, in\nthis paper, we propose an EEG-based Emotion Style Transfer Network (E2STN) to\nobtain EEG representations that contain the content information of source\ndomain and the style information of target domain, which is called stylized\nemotional EEG representations. The representations are helpful for\ncross-dataset discriminative prediction. Concretely, E2STN consists of three\nmodules, i.e., transfer module, transfer evaluation module, and discriminative\nprediction module. The transfer module encodes the domain-specific information\nof source and target domains and then re-constructs the source domain's\nemotional pattern and the target domain's statistical characteristics into the\nnew stylized EEG representations. In this process, the transfer evaluation\nmodule is adopted to constrain the generated representations that can more\nprecisely fuse two kinds of complementary information from source and target\ndomains and avoid distorting. Finally, the generated stylized EEG\nrepresentations are fed into the discriminative prediction module for final\nclassification. Extensive experiments show that the E2STN can achieve the\nstate-of-the-art performance on cross-dataset EEG emotion recognition tasks.\n", "date": "2023-08-09", "categories": []}
{"id": "2308.04517", "title": "Capturing Spectral and Long-term Contextual Information for Speech\n  Emotion Recognition Using Deep Learning Techniques", "abstract": "  Traditional approaches in speech emotion recognition, such as LSTM, CNN, RNN,\nSVM, and MLP, have limitations such as difficulty capturing long-term\ndependencies in sequential data, capturing the temporal dynamics, and\nstruggling to capture complex patterns and relationships in multimodal data.\nThis research addresses these shortcomings by proposing an ensemble model that\ncombines Graph Convolutional Networks (GCN) for processing textual data and the\nHuBERT transformer for analyzing audio signals. We found that GCNs excel at\ncapturing Long-term contextual dependencies and relationships within textual\ndata by leveraging graph-based representations of text and thus detecting the\ncontextual meaning and semantic relationships between words. On the other hand,\nHuBERT utilizes self-attention mechanisms to capture long-range dependencies,\nenabling the modeling of temporal dynamics present in speech and capturing\nsubtle nuances and variations that contribute to emotion recognition. By\ncombining GCN and HuBERT, our ensemble model can leverage the strengths of both\napproaches. This allows for the simultaneous analysis of multimodal data, and\nthe fusion of these modalities enables the extraction of complementary\ninformation, enhancing the discriminative power of the emotion recognition\nsystem. The results indicate that the combined model can overcome the\nlimitations of traditional methods, leading to enhanced accuracy in recognizing\nemotions from speech.\n", "date": "2023-08-04", "categories": []}
{"id": "2308.04674", "title": "Addressing Racial Bias in Facial Emotion Recognition", "abstract": "  Fairness in deep learning models trained with high-dimensional inputs and\nsubjective labels remains a complex and understudied area. Facial emotion\nrecognition, a domain where datasets are often racially imbalanced, can lead to\nmodels that yield disparate outcomes across racial groups. This study focuses\non analyzing racial bias by sub-sampling training sets with varied racial\ndistributions and assessing test performance across these simulations. Our\nfindings indicate that smaller datasets with posed faces improve on both\nfairness and performance metrics as the simulations approach racial balance.\nNotably, the F1-score increases by $27.2\\%$ points, and demographic parity\nincreases by $15.7\\%$ points on average across the simulations. However, in\nlarger datasets with greater facial variation, fairness metrics generally\nremain constant, suggesting that racial balance by itself is insufficient to\nachieve parity in test performance across different racial groups.\n", "date": "2023-08-09", "categories": []}
{"id": "2308.04811", "title": "A Bipartite Graph is All We Need for Enhancing Emotional Reasoning with\n  Commonsense Knowledge", "abstract": "  The context-aware emotional reasoning ability of AI systems, especially in\nconversations, is of vital importance in applications such as online opinion\nmining from social media and empathetic dialogue systems. Due to the implicit\nnature of conveying emotions in many scenarios, commonsense knowledge is widely\nutilized to enrich utterance semantics and enhance conversation modeling.\nHowever, most previous knowledge infusion methods perform empirical knowledge\nfiltering and design highly customized architectures for knowledge interaction\nwith the utterances, which can discard useful knowledge aspects and limit their\ngeneralizability to different knowledge sources. Based on these observations,\nwe propose a Bipartite Heterogeneous Graph (BHG) method for enhancing emotional\nreasoning with commonsense knowledge. In BHG, the extracted context-aware\nutterance representations and knowledge representations are modeled as\nheterogeneous nodes. Two more knowledge aggregation node types are proposed to\nperform automatic knowledge filtering and interaction. BHG-based knowledge\ninfusion can be directly generalized to multi-type and multi-grained knowledge\nsources. In addition, we propose a Multi-dimensional Heterogeneous Graph\nTransformer (MHGT) to perform graph reasoning, which can retain unchanged\nfeature spaces and unequal dimensions for heterogeneous node types during\ninference to prevent unnecessary loss of information. Experiments show that\nBHG-based methods significantly outperform state-of-the-art knowledge infusion\nmethods and show generalized knowledge infusion ability with higher efficiency.\nFurther analysis proves that previous empirical knowledge filtering methods do\nnot guarantee to provide the most useful knowledge information. Our code is\navailable at: https://github.com/SteveKGYang/BHG.\n", "date": "2023-08-09", "categories": []}
{"id": "2307.02227", "title": "MAE-DFER: Efficient Masked Autoencoder for Self-supervised Dynamic\n  Facial Expression Recognition", "abstract": "  Dynamic facial expression recognition (DFER) is essential to the development\nof intelligent and empathetic machines. Prior efforts in this field mainly fall\ninto supervised learning paradigm, which is severely restricted by the limited\nlabeled data in existing datasets. Inspired by recent unprecedented success of\nmasked autoencoders (e.g., VideoMAE), this paper proposes MAE-DFER, a novel\nself-supervised method which leverages large-scale self-supervised pre-training\non abundant unlabeled data to largely advance the development of DFER. Since\nthe vanilla Vision Transformer (ViT) employed in VideoMAE requires substantial\ncomputation during fine-tuning, MAE-DFER develops an efficient local-global\ninteraction Transformer (LGI-Former) as the encoder. Moreover, in addition to\nthe standalone appearance content reconstruction in VideoMAE, MAE-DFER also\nintroduces explicit temporal facial motion modeling to encourage LGI-Former to\nexcavate both static appearance and dynamic motion information. Extensive\nexperiments on six datasets show that MAE-DFER consistently outperforms\nstate-of-the-art supervised methods by significant margins (e.g., +6.30\\% UAR\non DFEW and +8.34\\% UAR on MAFW), verifying that it can learn powerful dynamic\nfacial representations via large-scale self-supervised pre-training. Besides,\nit has comparable or even better performance than VideoMAE, while largely\nreducing the computational cost (about 38\\% FLOPs). We believe MAE-DFER has\npaved a new way for the advancement of DFER and can inspire more relevant\nresearch in this field and even other related tasks. Codes and models are\npublicly available at https://github.com/sunlicai/MAE-DFER.\n", "date": "2023-07-05", "categories": []}
{"id": "2308.02778", "title": "Unveiling Emotions from EEG: A GRU-Based Approach", "abstract": "  One of the most important study areas in affective computing is emotion\nidentification using EEG data. In this study, the Gated Recurrent Unit (GRU)\nalgorithm, which is a type of Recurrent Neural Networks (RNNs), is tested to\nsee if it can use EEG signals to predict emotional states. Our publicly\naccessible dataset consists of resting neutral data as well as EEG recordings\nfrom people who were exposed to stimuli evoking happy, neutral, and negative\nemotions. For the best feature extraction, we pre-process the EEG data using\nartifact removal, bandpass filters, and normalization methods. With 100%\naccuracy on the validation set, our model produced outstanding results by\nutilizing the GRU's capacity to capture temporal dependencies. When compared to\nother machine learning techniques, our GRU model's Extreme Gradient Boosting\nClassifier had the highest accuracy. Our investigation of the confusion matrix\nrevealed insightful information about the performance of the model, enabling\nprecise emotion classification. This study emphasizes the potential of deep\nlearning models like GRUs for emotion recognition and advances in affective\ncomputing. Our findings open up new possibilities for interacting with\ncomputers and comprehending how emotions are expressed through brainwave\nactivity.\n", "date": "2023-07-20", "categories": []}
{"id": "2308.03000", "title": "StyleEDL: Style-Guided High-order Attention Network for Image Emotion\n  Distribution Learning", "abstract": "  Emotion distribution learning has gained increasing attention with the\ntendency to express emotions through images. As for emotion ambiguity arising\nfrom humans' subjectivity, substantial previous methods generally focused on\nlearning appropriate representations from the holistic or significant part of\nimages. However, they rarely consider establishing connections with the\nstylistic information although it can lead to a better understanding of images.\nIn this paper, we propose a style-guided high-order attention network for image\nemotion distribution learning termed StyleEDL, which interactively learns\nstylistic-aware representations of images by exploring the hierarchical\nstylistic information of visual contents. Specifically, we consider exploring\nthe intra- and inter-layer correlations among GRAM-based stylistic\nrepresentations, and meanwhile exploit an adversary-constrained high-order\nattention mechanism to capture potential interactions between subtle visual\nparts. In addition, we introduce a stylistic graph convolutional network to\ndynamically generate the content-dependent emotion representations to benefit\nthe final emotion distribution learning. Extensive experiments conducted on\nseveral benchmark datasets demonstrate the effectiveness of our proposed\nStyleEDL compared to state-of-the-art methods. The implementation is released\nat: https://github.com/liuxianyi/StyleEDL.\n", "date": "2023-08-06", "categories": []}
{"id": "2308.03150", "title": "\"We care\": Improving Code Mixed Speech Emotion Recognition in\n  Customer-Care Conversations", "abstract": "  Speech Emotion Recognition (SER) is the task of identifying the emotion\nexpressed in a spoken utterance. Emotion recognition is essential in building\nrobust conversational agents in domains such as law, healthcare, education, and\ncustomer support. Most of the studies published on SER use datasets created by\nemploying professional actors in a noise-free environment. In natural settings\nsuch as a customer care conversation, the audio is often noisy with speakers\nregularly switching between different languages as they see fit. We have worked\nin collaboration with a leading unicorn in the Conversational AI sector to\ndevelop Natural Speech Emotion Dataset (NSED). NSED is a natural code-mixed\nspeech emotion dataset where each utterance in a conversation is annotated with\nemotion, sentiment, valence, arousal, and dominance (VAD) values. In this\npaper, we show that by incorporating word-level VAD value we improve on the\ntask of SER by 2%, for negative emotions, over the baseline value for NSED.\nHigh accuracy for negative emotion recognition is essential because customers\nexpressing negative opinions/views need to be pacified with urgency, lest\ncomplaints and dissatisfaction snowball and get out of hand. Escalation of\nnegative opinions speedily is crucial for business interests. Our study then\ncan be utilized to develop conversational agents which are more polite and\nempathetic in such situations.\n", "date": "2023-08-06", "categories": []}
{"id": "2308.01919", "title": "Emotion recognition based on multi-modal electrophysiology multi-head\n  attention Contrastive Learning", "abstract": "  Emotion recognition is an important research direction in artificial\nintelligence, helping machines understand and adapt to human emotional states.\nMultimodal electrophysiological(ME) signals, such as EEG, GSR,\nrespiration(Resp), and temperature(Temp), are effective biomarkers for\nreflecting changes in human emotions. However, using electrophysiological\nsignals for emotion recognition faces challenges such as data scarcity,\ninconsistent labeling, and difficulty in cross-individual generalization. To\naddress these issues, we propose ME-MHACL, a self-supervised contrastive\nlearning-based multimodal emotion recognition method that can learn meaningful\nfeature representations from unlabeled electrophysiological signals and use\nmulti-head attention mechanisms for feature fusion to improve recognition\nperformance. Our method includes two stages: first, we use the Meiosis method\nto group sample and augment unlabeled electrophysiological signals and design a\nself-supervised contrastive learning task; second, we apply the trained feature\nextractor to labeled electrophysiological signals and use multi-head attention\nmechanisms for feature fusion. We conducted experiments on two public datasets,\nDEAP and MAHNOB-HCI, and our method outperformed existing benchmark methods in\nemotion recognition tasks and had good cross-individual generalization ability.\n", "date": "2023-07-12", "categories": []}
{"id": "2308.02173", "title": "Efficient Labelling of Affective Video Datasets via Few-Shot &\n  Multi-Task Contrastive Learning", "abstract": "  Whilst deep learning techniques have achieved excellent emotion prediction,\nthey still require large amounts of labelled training data, which are (a)\nonerous and tedious to compile, and (b) prone to errors and biases. We propose\nMulti-Task Contrastive Learning for Affect Representation (\\textbf{MT-CLAR})\nfor few-shot affect inference. MT-CLAR combines multi-task learning with a\nSiamese network trained via contrastive learning to infer from a pair of\nexpressive facial images (a) the (dis)similarity between the facial\nexpressions, and (b) the difference in valence and arousal levels of the two\nfaces. We further extend the image-based MT-CLAR framework for automated video\nlabelling where, given one or a few labelled video frames (termed\n\\textit{support-set}), MT-CLAR labels the remainder of the video for valence\nand arousal. Experiments are performed on the AFEW-VA dataset with multiple\nsupport-set configurations; moreover, supervised learning on representations\nlearnt via MT-CLAR are used for valence, arousal and categorical emotion\nprediction on the AffectNet and AFEW-VA datasets. The results show that valence\nand arousal predictions via MT-CLAR are very comparable to the state-of-the-art\n(SOTA), and we significantly outperform SOTA with a support-set $\\approx$6\\%\nthe size of the video dataset.\n", "date": "2023-08-04", "categories": []}
{"id": "2308.02190", "title": "Emo-DNA: Emotion Decoupling and Alignment Learning for Cross-Corpus\n  Speech Emotion Recognition", "abstract": "  Cross-corpus speech emotion recognition (SER) seeks to generalize the ability\nof inferring speech emotion from a well-labeled corpus to an unlabeled one,\nwhich is a rather challenging task due to the significant discrepancy between\ntwo corpora. Existing methods, typically based on unsupervised domain\nadaptation (UDA), struggle to learn corpus-invariant features by global\ndistribution alignment, but unfortunately, the resulting features are mixed\nwith corpus-specific features or not class-discriminative. To tackle these\nchallenges, we propose a novel Emotion Decoupling aNd Alignment learning\nframework (EMO-DNA) for cross-corpus SER, a novel UDA method to learn\nemotion-relevant corpus-invariant features. The novelties of EMO-DNA are\ntwo-fold: contrastive emotion decoupling and dual-level emotion alignment. On\none hand, our contrastive emotion decoupling achieves decoupling learning via a\ncontrastive decoupling loss to strengthen the separability of emotion-relevant\nfeatures from corpus-specific ones. On the other hand, our dual-level emotion\nalignment introduces an adaptive threshold pseudo-labeling to select confident\ntarget samples for class-level alignment, and performs corpus-level alignment\nto jointly guide model for learning class-discriminative corpus-invariant\nfeatures across corpora. Extensive experimental results demonstrate the\nsuperior performance of EMO-DNA over the state-of-the-art methods in several\ncross-corpus scenarios. Source code is available at\nhttps://github.com/Jiaxin-Ye/Emo-DNA.\n", "date": "2023-08-04", "categories": []}
{"id": "2303.10741", "title": "Computer Vision Estimation of Emotion Reaction Intensity in the Wild", "abstract": "  Emotions play an essential role in human communication. Developing computer\nvision models for automatic recognition of emotion expression can aid in a\nvariety of domains, including robotics, digital behavioral healthcare, and\nmedia analytics. There are three types of emotional representations which are\ntraditionally modeled in affective computing research: Action Units, Valence\nArousal (VA), and Categorical Emotions. As part of an effort to move beyond\nthese representations towards more fine-grained labels, we describe our\nsubmission to the newly introduced Emotional Reaction Intensity (ERI)\nEstimation challenge in the 5th competition for Affective Behavior Analysis\nin-the-Wild (ABAW). We developed four deep neural networks trained in the\nvisual domain and a multimodal model trained with both visual and audio\nfeatures to predict emotion reaction intensity. Our best performing model on\nthe Hume-Reaction dataset achieved an average Pearson correlation coefficient\nof 0.4080 on the test set using a pre-trained ResNet50 model. This work\nprovides a first step towards the development of production-grade models which\npredict emotion reaction intensities rather than discrete emotion categories.\n", "date": "2023-03-19", "categories": []}
{"id": "2308.01700", "title": "Bees Local Phase Quantization Feature Selection for RGB-D Facial\n  Expressions Recognition", "abstract": "  Feature selection could be defined as an optimization problem and solved by\nbio-inspired algorithms. Bees Algorithm (BA) shows decent performance in\nfeature selection optimization tasks. On the other hand, Local Phase\nQuantization (LPQ) is a frequency domain feature which has excellent\nperformance on Depth images. Here, after extracting LPQ features out of RGB\n(colour) and Depth images from the Iranian Kinect Face Database (IKFDB), the\nBees feature selection algorithm applies to select the desired number of\nfeatures for final classification tasks. IKFDB is recorded with Kinect sensor\nV.2 and contains colour and depth images for facial and facial\nmicro-expressions recognition purposes. Here five facial expressions of Anger,\nJoy, Surprise, Disgust and Fear are used for final validation. The proposed\nBees LPQ method is compared with Particle Swarm Optimization (PSO) LPQ, PCA\nLPQ, Lasso LPQ, and just LPQ features for classification tasks with Support\nVector Machines (SVM), K-Nearest Neighbourhood (KNN), Shallow Neural Network\nand Ensemble Subspace KNN. Returned results, show a decent performance of the\nproposed algorithm (99 % accuracy) in comparison with others.\n", "date": "2023-08-03", "categories": []}
{"id": "2211.06607", "title": "Few-shot Multimodal Sentiment Analysis based on Multimodal Probabilistic\n  Fusion Prompts", "abstract": "  Multimodal sentiment analysis has gained significant attention due to the\nproliferation of multimodal content on social media. However, existing studies\nin this area rely heavily on large-scale supervised data, which is\ntime-consuming and labor-intensive to collect. Thus, there is a need to address\nthe challenge of few-shot multimodal sentiment analysis. To tackle this\nproblem, we propose a novel method called Multimodal Probabilistic Fusion\nPrompts (MultiPoint) that leverages diverse cues from different modalities for\nmultimodal sentiment detection in the few-shot scenario. Specifically, we start\nby introducing a Consistently Distributed Sampling approach called CDS, which\nensures that the few-shot dataset has the same category distribution as the\nfull dataset. Unlike previous approaches primarily using prompts based on the\ntext modality, we design unified multimodal prompts to reduce discrepancies\nbetween different modalities and dynamically incorporate multimodal\ndemonstrations into the context of each multimodal instance. To enhance the\nmodel's robustness, we introduce a probabilistic fusion method to fuse output\npredictions from multiple diverse prompts for each input. Our extensive\nexperiments on six datasets demonstrate the effectiveness of our approach.\nFirst, our method outperforms strong baselines in the multimodal few-shot\nsetting. Furthermore, under the same amount of data (1% of the full dataset),\nour CDS-based experimental results significantly outperform those based on\npreviously sampled datasets constructed from the same number of instances of\neach class.\n", "date": "2022-11-12", "categories": []}
{"id": "2308.00228", "title": "Using Scene and Semantic Features for Multi-modal Emotion Recognition", "abstract": "  Automatic emotion recognition is a hot topic with a wide range of\napplications. Much work has been done in the area of automatic emotion\nrecognition in recent years. The focus has been mainly on using the\ncharacteristics of a person such as speech, facial expression and pose for this\npurpose. However, the processing of scene and semantic features for emotion\nrecognition has had limited exploration. In this paper, we propose to use\ncombined scene and semantic features, along with personal features, for\nmulti-modal emotion recognition. Scene features will describe the environment\nor context in which the target person is operating. The semantic feature can\ninclude objects that are present in the environment, as well as their\nattributes and relationships with the target person. In addition, we use a\nmodified EmbraceNet to extract features from the images, which is trained to\nlearn both the body and pose features simultaneously. By fusing both body and\npose features, the EmbraceNet can improve the accuracy and robustness of the\nmodel, particularly when dealing with partially missing data. This is because\nhaving both body and pose features provides a more complete representation of\nthe subject in the images, which can help the model to make more accurate\npredictions even when some parts of body are missing. We demonstrate the\nefficiency of our method on the benchmark EMOTIC dataset. We report an average\nprecision of 40.39\\% across the 26 emotion categories, which is a 5\\%\nimprovement over previous approaches.\n", "date": "2023-08-01", "categories": []}
{"id": "2308.00246", "title": "EEG-based Cognitive Load Classification using Feature Masked\n  Autoencoding and Emotion Transfer Learning", "abstract": "  Cognitive load, the amount of mental effort required for task completion,\nplays an important role in performance and decision-making outcomes, making its\nclassification and analysis essential in various sensitive domains. In this\npaper, we present a new solution for the classification of cognitive load using\nelectroencephalogram (EEG). Our model uses a transformer architecture employing\ntransfer learning between emotions and cognitive load. We pre-train our model\nusing self-supervised masked autoencoding on emotion-related EEG datasets and\nuse transfer learning with both frozen weights and fine-tuning to perform\ndownstream cognitive load classification. To evaluate our method, we carry out\na series of experiments utilizing two publicly available EEG-based emotion\ndatasets, namely SEED and SEED-IV, for pre-training, while we use the CL-Drive\ndataset for downstream cognitive load classification. The results of our\nexperiments show that our proposed approach achieves strong results and\noutperforms conventional single-stage fully supervised learning. Moreover, we\nperform detailed ablation and sensitivity studies to evaluate the impact of\ndifferent aspects of our proposed solution. This research contributes to the\ngrowing body of literature in affective computing with a focus on cognitive\nload, and opens up new avenues for future research in the field of cross-domain\ntransfer learning using self-supervised pre-training.\n", "date": "2023-08-01", "categories": []}
{"id": "2307.15862", "title": "Catching Elusive Depression via Facial Micro-Expression Recognition", "abstract": "  Depression is a common mental health disorder that can cause consequential\nsymptoms with continuously depressed mood that leads to emotional distress. One\ncategory of depression is Concealed Depression, where patients intentionally or\nunintentionally hide their genuine emotions through exterior optimism, thereby\ncomplicating and delaying diagnosis and treatment and leading to unexpected\nsuicides. In this paper, we propose to diagnose concealed depression by using\nfacial micro-expressions (FMEs) to detect and recognize underlying true\nemotions. However, the extremely low intensity and subtle nature of FMEs make\ntheir recognition a tough task. We propose a facial landmark-based\nRegion-of-Interest (ROI) approach to address the challenge, and describe a\nlow-cost and privacy-preserving solution that enables self-diagnosis using\nportable mobile devices in a personal setting (e.g., at home). We present\nresults and findings that validate our method, and discuss other technical\nchallenges and future directions in applying such techniques to real clinical\nsettings.\n", "date": "2023-07-29", "categories": []}
{"id": "2307.07961", "title": "EmoSet: A Large-scale Visual Emotion Dataset with Rich Attributes", "abstract": "  Visual Emotion Analysis (VEA) aims at predicting people's emotional responses\nto visual stimuli. This is a promising, yet challenging, task in affective\ncomputing, which has drawn increasing attention in recent years. Most of the\nexisting work in this area focuses on feature design, while little attention\nhas been paid to dataset construction. In this work, we introduce EmoSet, the\nfirst large-scale visual emotion dataset annotated with rich attributes, which\nis superior to existing datasets in four aspects: scale, annotation richness,\ndiversity, and data balance. EmoSet comprises 3.3 million images in total, with\n118,102 of these images carefully labeled by human annotators, making it five\ntimes larger than the largest existing dataset. EmoSet includes images from\nsocial networks, as well as artistic images, and it is well balanced between\ndifferent emotion categories. Motivated by psychological studies, in addition\nto emotion category, each image is also annotated with a set of describable\nemotion attributes: brightness, colorfulness, scene type, object class, facial\nexpression, and human action, which can help understand visual emotions in a\nprecise and interpretable way. The relevance of these emotion attributes is\nvalidated by analyzing the correlations between them and visual emotion, as\nwell as by designing an attribute module to help visual emotion recognition. We\nbelieve EmoSet will bring some key insights and encourage further research in\nvisual emotion analysis and understanding. Project page:\nhttps://vcc.tech/EmoSet.\n", "date": "2023-07-16", "categories": []}
{"id": "2307.15164", "title": "VISU at WASSA 2023 Shared Task: Detecting Emotions in Reaction to News\n  Stories Leveraging BERT and Stacked Embeddings", "abstract": "  Our system, VISU, participated in the WASSA 2023 Shared Task (3) of Emotion\nClassification from essays written in reaction to news articles. Emotion\ndetection from complex dialogues is challenging and often requires\ncontext/domain understanding. Therefore in this research, we have focused on\ndeveloping deep learning (DL) models using the combination of word embedding\nrepresentations with tailored prepossessing strategies to capture the nuances\nof emotions expressed. Our experiments used static and contextual embeddings\n(individual and stacked) with Bidirectional Long short-term memory (BiLSTM) and\nTransformer based models. We occupied rank tenth in the emotion detection task\nby scoring a Macro F1-Score of 0.2717, validating the efficacy of our\nimplemented approaches for small and imbalanced datasets with mixed categories\nof target emotions.\n", "date": "2023-07-27", "categories": []}
{"id": "2305.16731", "title": "Automatic Emotion Experiencer Recognition", "abstract": "  The most prominent subtask in emotion analysis is emotion classification; to\nassign a category to a textual unit, for instance a social media post. Many\nresearch questions from the social sciences do, however, not only require the\ndetection of the emotion of an author of a post but to understand who is\nascribed an emotion in text. This task is tackled by emotion role labeling\nwhich aims at extracting who is described in text to experience an emotion,\nwhy, and towards whom. This could, however, be considered overly sophisticated\nif the main question to answer is who feels which emotion. A targeted approach\nfor such setup is to classify emotion experiencer mentions (aka \"emoters\")\nregarding the emotion they presumably perceive. This task is similar to named\nentity recognition of person names with the difference that not every mentioned\nentity name is an emoter. While, very recently, data with emoter annotations\nhas been made available, no experiments have yet been performed to detect such\nmentions. With this paper, we provide baseline experiments to understand how\nchallenging the task is. We further evaluate the impact on experiencer-specific\nemotion categorization and appraisal detection in a pipeline, when gold\nmentions are not available. We show that experiencer detection in text is a\nchallenging task, with a precision of .82 and a recall of .56 (F1 =.66). These\nresults motivate future work of jointly modeling emoter spans and\nemotion/appraisal predictions.\n", "date": "2023-05-26", "categories": []}
{"id": "2307.14637", "title": "HTNet for micro-expression recognition", "abstract": "  Facial expression is related to facial muscle contractions and different\nmuscle movements correspond to different emotional states. For micro-expression\nrecognition, the muscle movements are usually subtle, which has a negative\nimpact on the performance of current facial emotion recognition algorithms.\nMost existing methods use self-attention mechanisms to capture relationships\nbetween tokens in a sequence, but they do not take into account the inherent\nspatial relationships between facial landmarks. This can result in sub-optimal\nperformance on micro-expression recognition tasks.Therefore, learning to\nrecognize facial muscle movements is a key challenge in the area of\nmicro-expression recognition. In this paper, we propose a Hierarchical\nTransformer Network (HTNet) to identify critical areas of facial muscle\nmovement. HTNet includes two major components: a transformer layer that\nleverages the local temporal features and an aggregation layer that extracts\nlocal and global semantical facial features. Specifically, HTNet divides the\nface into four different facial areas: left lip area, left eye area, right eye\narea and right lip area. The transformer layer is used to focus on representing\nlocal minor muscle movement with local self-attention in each area. The\naggregation layer is used to learn the interactions between eye areas and lip\nareas. The experiments on four publicly available micro-expression datasets\nshow that the proposed approach outperforms previous methods by a large margin.\nThe codes and models are available at:\n\\url{https://github.com/wangzhifengharrison/HTNet}\n", "date": "2023-07-27", "categories": []}
{"id": "2307.13706", "title": "Introducing CALMED: Multimodal Annotated Dataset for Emotion Detection\n  in Children with Autism", "abstract": "  Automatic Emotion Detection (ED) aims to build systems to identify users'\nemotions automatically. This field has the potential to enhance HCI, creating\nan individualised experience for the user. However, ED systems tend to perform\npoorly on people with Autism Spectrum Disorder (ASD). Hence, the need to create\nED systems tailored to how people with autism express emotions. Previous works\nhave created ED systems tailored for children with ASD but did not share the\nresulting dataset. Sharing annotated datasets is essential to enable the\ndevelopment of more advanced computer models for ED within the research\ncommunity. In this paper, we describe our experience establishing a process to\ncreate a multimodal annotated dataset featuring children with a level 1\ndiagnosis of autism. In addition, we introduce CALMED (Children, Autism,\nMultimodal, Emotion, Detection), the resulting multimodal emotion detection\ndataset featuring children with autism aged 8-12. CALMED includes audio and\nvideo features extracted from recording files of study sessions with\nparticipants, together with annotations provided by their parents into four\ntarget classes. The generated dataset includes a total of 57,012 examples, with\neach example representing a time window of 200ms (0.2s). Our experience and\nmethods described here, together with the dataset shared, aim to contribute to\nfuture research applications of affective computing in ASD, which has the\npotential to create systems to improve the lives of people with ASD.\n", "date": "2023-07-24", "categories": []}
{"id": "2307.13779", "title": "Is GPT a Computational Model of Emotion? Detailed Analysis", "abstract": "  This paper investigates the emotional reasoning abilities of the GPT family\nof large language models via a component perspective. The paper first examines\nhow the model reasons about autobiographical memories. Second, it\nsystematically varies aspects of situations to impact emotion intensity and\ncoping tendencies. Even without the use of prompt engineering, it is shown that\nGPT's predictions align significantly with human-provided appraisals and\nemotional labels. However, GPT faces difficulties predicting emotion intensity\nand coping responses. GPT-4 showed the highest performance in the initial study\nbut fell short in the second, despite providing superior results after minor\nprompt engineering. This assessment brings up questions on how to effectively\nemploy the strong points and address the weak areas of these models,\nparticularly concerning response variability. These studies underscore the\nmerits of evaluating models from a componential perspective.\n", "date": "2023-07-25", "categories": []}
{"id": "2307.14004", "title": "Affective Natural Language Generation of Event Descriptions through\n  Fine-grained Appraisal Conditions", "abstract": "  Models for affective text generation have shown a remarkable progress, but\nthey commonly rely only on basic emotion theories or valance/arousal values as\nconditions. This is appropriate when the goal is to create explicit emotion\nstatements (\"The kid is happy.\"). Emotions are, however, commonly communicated\nimplicitly. For instance, the emotional interpretation of an event (\"Their dog\ndied.\") does often not require an explicit emotion statement. In psychology,\nappraisal theories explain the link between a cognitive evaluation of an event\nand the potentially developed emotion. They put the assessment of the situation\non the spot, for instance regarding the own control or the responsibility for\nwhat happens. We hypothesize and subsequently show that including appraisal\nvariables as conditions in a generation framework comes with two advantages.\n(1) The generation model is informed in greater detail about what makes a\nspecific emotion and what properties it has. This leads to text generation that\nbetter fulfills the condition. (2) The variables of appraisal allow a user to\nperform a more fine-grained control of the generated text, by stating\nproperties of a situation instead of only providing the emotion category. Our\nBart and T5-based experiments with 7 emotions (Anger, Disgust, Fear, Guilt,\nJoy, Sadness, Shame), and 7 appraisals (Attention, Responsibility, Control,\nCircumstance, Pleasantness, Effort, Certainty) show that (1) adding appraisals\nduring training improves the accurateness of the generated texts by 10 pp in\nF1. Further, (2) the texts with appraisal variables are longer and contain more\ndetails. This exemplifies the greater control for users.\n", "date": "2023-07-26", "categories": []}
{"id": "2307.13463", "title": "Unlocking the Emotional World of Visual Media: An Overview of the\n  Science, Research, and Impact of Understanding Emotion", "abstract": "  The emergence of artificial emotional intelligence technology is\nrevolutionizing the fields of computers and robotics, allowing for a new level\nof communication and understanding of human behavior that was once thought\nimpossible. While recent advancements in deep learning have transformed the\nfield of computer vision, automated understanding of evoked or expressed\nemotions in visual media remains in its infancy. This foundering stems from the\nabsence of a universally accepted definition of \"emotion\", coupled with the\ninherently subjective nature of emotions and their intricate nuances. In this\narticle, we provide a comprehensive, multidisciplinary overview of the field of\nemotion analysis in visual media, drawing on insights from psychology,\nengineering, and the arts. We begin by exploring the psychological foundations\nof emotion and the computational principles that underpin the understanding of\nemotions from images and videos. We then review the latest research and systems\nwithin the field, accentuating the most promising approaches. We also discuss\nthe current technological challenges and limitations of emotion analysis,\nunderscoring the necessity for continued investigation and innovation. We\ncontend that this represents a \"Holy Grail\" research problem in computing and\ndelineate pivotal directions for future inquiry. Finally, we examine the\nethical ramifications of emotion-understanding technologies and contemplate\ntheir potential societal impacts. Overall, this article endeavors to equip\nreaders with a deeper understanding of the domain of emotion analysis in visual\nmedia and to inspire further research and development in this captivating and\nrapidly evolving field.\n", "date": "2023-07-25", "categories": []}
{"id": "2307.12221", "title": "FATRER: Full-Attention Topic Regularizer for Accurate and Robust\n  Conversational Emotion Recognition", "abstract": "  This paper concentrates on the understanding of interlocutors' emotions\nevoked in conversational utterances. Previous studies in this literature mainly\nfocus on more accurate emotional predictions, while ignoring model robustness\nwhen the local context is corrupted by adversarial attacks. To maintain\nrobustness while ensuring accuracy, we propose an emotion recognizer augmented\nby a full-attention topic regularizer, which enables an emotion-related global\nview when modeling the local context in a conversation. A joint topic modeling\nstrategy is introduced to implement regularization from both representation and\nloss perspectives. To avoid over-regularization, we drop the constraints on\nprior distributions that exist in traditional topic modeling and perform\nprobabilistic approximations based entirely on attention alignment. Experiments\nshow that our models obtain more favorable results than state-of-the-art\nmodels, and gain convincing robustness under three types of adversarial\nattacks.\n", "date": "2023-07-23", "categories": []}
{"id": "2307.12241", "title": "Explainable Depression Detection via Head Motion Patterns", "abstract": "  While depression has been studied via multimodal non-verbal behavioural cues,\nhead motion behaviour has not received much attention as a biomarker. This\nstudy demonstrates the utility of fundamental head-motion units, termed\n\\emph{kinemes}, for depression detection by adopting two distinct approaches,\nand employing distinctive features: (a) discovering kinemes from head motion\ndata corresponding to both depressed patients and healthy controls, and (b)\nlearning kineme patterns only from healthy controls, and computing statistics\nderived from reconstruction errors for both the patient and control classes.\nEmploying machine learning methods, we evaluate depression classification\nperformance on the \\emph{BlackDog} and \\emph{AVEC2013} datasets. Our findings\nindicate that: (1) head motion patterns are effective biomarkers for detecting\ndepressive symptoms, and (2) explanatory kineme patterns consistent with prior\nfindings can be observed for the two classes. Overall, we achieve peak F1\nscores of 0.79 and 0.82, respectively, over BlackDog and AVEC2013 for binary\nclassification over episodic \\emph{thin-slices}, and a peak F1 of 0.72 over\nvideos for AVEC2013.\n", "date": "2023-07-23", "categories": []}
{"id": "2307.12343", "title": "Self-Supervised Learning for Audio-Based Emotion Recognition", "abstract": "  Emotion recognition models using audio input data can enable the development\nof interactive systems with applications in mental healthcare, marketing,\ngaming, and social media analysis. While the field of affective computing using\naudio data is rich, a major barrier to achieve consistently high-performance\nmodels is the paucity of available training labels. Self-supervised learning\n(SSL) is a family of methods which can learn despite a scarcity of supervised\nlabels by predicting properties of the data itself. To understand the utility\nof self-supervised learning for audio-based emotion recognition, we have\napplied self-supervised learning pre-training to the classification of emotions\nfrom the CMU- MOSEI's acoustic modality. Unlike prior papers that have\nexperimented with raw acoustic data, our technique has been applied to encoded\nacoustic data. Our model is first pretrained to uncover the randomly-masked\ntimestamps of the acoustic data. The pre-trained model is then fine-tuned using\na small sample of annotated data. The performance of the final model is then\nevaluated via several evaluation metrics against a baseline deep learning model\nwith an identical backbone architecture. We find that self-supervised learning\nconsistently improves the performance of the model across all metrics. This\nwork shows the utility of self-supervised learning for affective computing,\ndemonstrating that self-supervised learning is most useful when the number of\ntraining examples is small, and that the effect is most pronounced for emotions\nwhich are easier to classify such as happy, sad and anger. This work further\ndemonstrates that self-supervised learning works when applied to embedded\nfeature representations rather than the traditional approach of pre-training on\nthe raw input space.\n", "date": "2023-07-23", "categories": []}
{"id": "2307.11404", "title": "Latent-OFER: Detect, Mask, and Reconstruct with Latent Vectors for\n  Occluded Facial Expression Recognition", "abstract": "  Most research on facial expression recognition (FER) is conducted in highly\ncontrolled environments, but its performance is often unacceptable when applied\nto real-world situations. This is because when unexpected objects occlude the\nface, the FER network faces difficulties extracting facial features and\naccurately predicting facial expressions. Therefore, occluded FER (OFER) is a\nchallenging problem. Previous studies on occlusion-aware FER have typically\nrequired fully annotated facial images for training. However, collecting facial\nimages with various occlusions and expression annotations is time-consuming and\nexpensive. Latent-OFER, the proposed method, can detect occlusions, restore\noccluded parts of the face as if they were unoccluded, and recognize them,\nimproving FER accuracy. This approach involves three steps: First, the vision\ntransformer (ViT)-based occlusion patch detector masks the occluded position by\ntraining only latent vectors from the unoccluded patches using the support\nvector data description algorithm. Second, the hybrid reconstruction network\ngenerates the masking position as a complete image using the ViT and\nconvolutional neural network (CNN). Last, the expression-relevant latent vector\nextractor retrieves and uses expression-related information from all latent\nvectors by applying a CNN-based class activation map. This mechanism has a\nsignificant advantage in preventing performance degradation from occlusion by\nunseen objects. The experimental results on several databases demonstrate the\nsuperiority of the proposed method over state-of-the-art methods.\n", "date": "2023-07-21", "categories": []}
{"id": "2307.11584", "title": "A Change of Heart: Improving Speech Emotion Recognition through\n  Speech-to-Text Modality Conversion", "abstract": "  Speech Emotion Recognition (SER) is a challenging task. In this paper, we\nintroduce a modality conversion concept aimed at enhancing emotion recognition\nperformance on the MELD dataset. We assess our approach through two\nexperiments: first, a method named Modality-Conversion that employs automatic\nspeech recognition (ASR) systems, followed by a text classifier; second, we\nassume perfect ASR output and investigate the impact of modality conversion on\nSER, this method is called Modality-Conversion++. Our findings indicate that\nthe first method yields substantial results, while the second method\noutperforms state-of-the-art (SOTA) speech-based approaches in terms of SER\nweighted-F1 (WF1) score on the MELD dataset. This research highlights the\npotential of modality conversion for tasks that can be conducted in alternative\nmodalities.\n", "date": "2023-07-21", "categories": []}
{"id": "2307.09023", "title": "LA-Net: Landmark-Aware Learning for Reliable Facial Expression\n  Recognition under Label Noise", "abstract": "  Facial expression recognition (FER) remains a challenging task due to the\nambiguity of expressions. The derived noisy labels significantly harm the\nperformance in real-world scenarios. To address this issue, we present a new\nFER model named Landmark-Aware Net~(LA-Net), which leverages facial landmarks\nto mitigate the impact of label noise from two perspectives. Firstly, LA-Net\nuses landmark information to suppress the uncertainty in expression space and\nconstructs the label distribution of each sample by neighborhood aggregation,\nwhich in turn improves the quality of training supervision. Secondly, the model\nincorporates landmark information into expression representations using the\ndevised expression-landmark contrastive loss. The enhanced expression feature\nextractor can be less susceptible to label noise. Our method can be integrated\nwith any deep neural network for better training supervision without\nintroducing extra inference costs. We conduct extensive experiments on both\nin-the-wild datasets and synthetic noisy datasets and demonstrate that LA-Net\nachieves state-of-the-art performance.\n", "date": "2023-07-18", "categories": []}
{"id": "2307.10814", "title": "Cross-Corpus Multilingual Speech Emotion Recognition: Amharic vs. Other\n  Languages", "abstract": "  In a conventional Speech emotion recognition (SER) task, a classifier for a\ngiven language is trained on a pre-existing dataset for that same language.\nHowever, where training data for a language does not exist, data from other\nlanguages can be used instead. We experiment with cross-lingual and\nmultilingual SER, working with Amharic, English, German and URDU. For Amharic,\nwe use our own publicly-available Amharic Speech Emotion Dataset (ASED). For\nEnglish, German and Urdu we use the existing RAVDESS, EMO-DB and URDU datasets.\nWe followed previous research in mapping labels for all datasets to just two\nclasses, positive and negative. Thus we can compare performance on different\nlanguages directly, and combine languages for training and testing. In\nExperiment 1, monolingual SER trials were carried out using three classifiers,\nAlexNet, VGGE (a proposed variant of VGG), and ResNet50. Results averaged for\nthe three models were very similar for ASED and RAVDESS, suggesting that\nAmharic and English SER are equally difficult. Similarly, German SER is more\ndifficult, and Urdu SER is easier. In Experiment 2, we trained on one language\nand tested on another, in both directions for each pair: Amharic<->German,\nAmharic<->English, and Amharic<->Urdu. Results with Amharic as target suggested\nthat using English or German as source will give the best result. In Experiment\n3, we trained on several non-Amharic languages and then tested on Amharic. The\nbest accuracy obtained was several percent greater than the best accuracy in\nExperiment 2, suggesting that a better result can be obtained when using two or\nthree non-Amharic languages for training than when using just one non-Amharic\nlanguage. Overall, the results suggest that cross-lingual and multilingual\ntraining can be an effective strategy for training a SER classifier when\nresources for a language are scarce.\n", "date": "2023-07-20", "categories": []}
{"id": "2306.13804", "title": "Cross-Language Speech Emotion Recognition Using Multimodal Dual\n  Attention Transformers", "abstract": "  Despite the recent progress in speech emotion recognition (SER),\nstate-of-the-art systems are unable to achieve improved performance in\ncross-language settings. In this paper, we propose a Multimodal Dual Attention\nTransformer (MDAT) model to improve cross-language SER. Our model utilises\npre-trained models for multimodal feature extraction and is equipped with a\ndual attention mechanism including graph attention and co-attention to capture\ncomplex dependencies across different modalities and achieve improved\ncross-language SER results using minimal target language data. In addition, our\nmodel also exploits a transformer encoder layer for high-level feature\nrepresentation to improve emotion classification accuracy. In this way, MDAT\nperforms refinement of feature representation at various stages and provides\nemotional salient features to the classification layer. This novel approach\nalso ensures the preservation of modality-specific emotional information while\nenhancing cross-modality and cross-language interactions. We assess our model's\nperformance on four publicly available SER datasets and establish its superior\neffectiveness compared to recent approaches and baseline models.\n", "date": "2023-06-23", "categories": []}
{"id": "2307.05375", "title": "Emotion Analysis on EEG Signal Using Machine Learning and Neural Network", "abstract": "  Emotion has a significant influence on how one thinks and interacts with\nothers. It serves as a link between how a person feels and the actions one\ntakes, or it could be said that it influences one's life decisions on occasion.\nSince the patterns of emotions and their reflections vary from person to\nperson, their inquiry must be based on approaches that are effective over a\nwide range of population regions. To extract features and enhance accuracy,\nemotion recognition using brain waves or EEG signals requires the\nimplementation of efficient signal processing techniques. Various approaches to\nhuman-machine interaction technologies have been ongoing for a long time, and\nin recent years, researchers have had great success in automatically\nunderstanding emotion using brain signals. In our research, several emotional\nstates were classified and tested on EEG signals collected from a well-known\npublicly available dataset, the DEAP Dataset, using SVM (Support Vector\nMachine), KNN (K-Nearest Neighbor), and an advanced neural network model, RNN\n(Recurrent Neural Network), trained with LSTM (Long Short Term Memory). The\nmain purpose of this study is to improve ways to improve emotion recognition\nperformance using brain signals. Emotions, on the other hand, can change with\ntime. As a result, the changes in emotion over time are also examined in our\nresearch.\n", "date": "2023-07-09", "categories": []}
{"id": "2307.05383", "title": "Human Emotion Recognition Based On Galvanic Skin Response signal Feature\n  Selection and SVM", "abstract": "  A novel human emotion recognition method based on automatically selected\nGalvanic Skin Response (GSR) signal features and SVM is proposed in this paper.\nGSR signals were acquired by e-Health Sensor Platform V2.0. Then, the data is\nde-noised by wavelet function and normalized to get rid of the individual\ndifference. 30 features are extracted from the normalized data, however,\ndirectly using of these features will lead to a low recognition rate. In order\nto gain the optimized features, a covariance based feature selection is\nemployed in our method. Finally, a SVM with input of the optimized features is\nutilized to achieve the human emotion recognition. The experimental results\nindicate that the proposed method leads to good human emotion recognition, and\nthe recognition accuracy is more than 66.67%.\n", "date": "2023-07-04", "categories": []}
{"id": "2307.04648", "title": "Can ChatGPT's Responses Boost Traditional Natural Language Processing?", "abstract": "  The employment of foundation models is steadily expanding, especially with\nthe launch of ChatGPT and the release of other foundation models. These models\nhave shown the potential of emerging capabilities to solve problems, without\nbeing particularly trained to solve. A previous work demonstrated these\nemerging capabilities in affective computing tasks; the performance quality was\nsimilar to traditional Natural Language Processing (NLP) techniques, but\nfalling short of specialised trained models, like fine-tuning of the RoBERTa\nlanguage model. In this work, we extend this by exploring if ChatGPT has novel\nknowledge that would enhance existing specialised models when they are fused\ntogether. We achieve this by investigating the utility of verbose responses\nfrom ChatGPT about solving a downstream task, in addition to studying the\nutility of fusing that with existing NLP methods. The study is conducted on\nthree affective computing problems, namely sentiment analysis, suicide tendency\ndetection, and big-five personality assessment. The results conclude that\nChatGPT has indeed novel knowledge that can improve existing NLP techniques by\nway of fusion, be it early or late fusion.\n", "date": "2023-07-06", "categories": []}
{"id": "2307.02773", "title": "SeLiNet: Sentiment enriched Lightweight Network for Emotion Recognition\n  in Images", "abstract": "  In this paper, we propose a sentiment-enriched lightweight network SeLiNet\nand an end-to-end on-device pipeline for contextual emotion recognition in\nimages. SeLiNet model consists of body feature extractor, image aesthetics\nfeature extractor, and learning-based fusion network which jointly estimates\ndiscrete emotion and human sentiments tasks. On the EMOTIC dataset, the\nproposed approach achieves an Average Precision (AP) score of 27.17 in\ncomparison to the baseline AP score of 27.38 while reducing the model size by\n>85%. In addition, we report an on-device AP score of 26.42 with reduction in\nmodel size by >93% when compared to the baseline.\n", "date": "2023-07-06", "categories": []}
{"id": "2307.02820", "title": "Evaluating raw waveforms with deep learning frameworks for speech\n  emotion recognition", "abstract": "  Speech emotion recognition is a challenging task in speech processing field.\nFor this reason, feature extraction process has a crucial importance to\ndemonstrate and process the speech signals. In this work, we represent a model,\nwhich feeds raw audio files directly into the deep neural networks without any\nfeature extraction stage for the recognition of emotions utilizing six\ndifferent data sets, EMO-DB, RAVDESS, TESS, CREMA, SAVEE, and TESS+RAVDESS. To\ndemonstrate the contribution of proposed model, the performance of traditional\nfeature extraction techniques namely, mel-scale spectogram, mel-frequency\ncepstral coefficients, are blended with machine learning algorithms, ensemble\nlearning methods, deep and hybrid deep learning techniques. Support vector\nmachine, decision tree, naive Bayes, random forests models are evaluated as\nmachine learning algorithms while majority voting and stacking methods are\nassessed as ensemble learning techniques. Moreover, convolutional neural\nnetworks, long short-term memory networks, and hybrid CNN- LSTM model are\nevaluated as deep learning techniques and compared with machine learning and\nensemble learning methods. To demonstrate the effectiveness of proposed model,\nthe comparison with state-of-the-art studies are carried out. Based on the\nexperiment results, CNN model excels existent approaches with 95.86% of\naccuracy for TESS+RAVDESS data set using raw audio files, thence determining\nthe new state-of-the-art. The proposed model performs 90.34% of accuracy for\nEMO-DB with CNN model, 90.42% of accuracy for RAVDESS with CNN model, 99.48% of\naccuracy for TESS with LSTM model, 69.72% of accuracy for CREMA with CNN model,\n85.76% of accuracy for SAVEE with CNN model in speaker-independent audio\ncategorization problems.\n", "date": "2023-07-06", "categories": []}
{"id": "2305.07611", "title": "Multimodal Sentiment Analysis: A Survey", "abstract": "  Multimodal sentiment analysis has become an important research area in the\nfield of artificial intelligence. With the latest advances in deep learning,\nthis technology has reached new heights. It has great potential for both\napplication and research, making it a popular research topic. This review\nprovides an overview of the definition, background, and development of\nmultimodal sentiment analysis. It also covers recent datasets and advanced\nmodels, emphasizing the challenges and future prospects of this technology.\nFinally, it looks ahead to future research directions. It should be noted that\nthis review provides constructive suggestions for promising research directions\nand building better performing multimodal sentiment analysis models, which can\nhelp researchers in this field.\n", "date": "2023-05-12", "categories": []}
{"id": "2306.17500", "title": "Empirical Interpretation of the Relationship Between Speech Acoustic\n  Context and Emotion Recognition", "abstract": "  Speech emotion recognition (SER) is vital for obtaining emotional\nintelligence and understanding the contextual meaning of speech. Variations of\nconsonant-vowel (CV) phonemic boundaries can enrich acoustic context with\nlinguistic cues, which impacts SER. In practice, speech emotions are treated as\nsingle labels over an acoustic segment for a given time duration. However,\nphone boundaries within speech are not discrete events, therefore the perceived\nemotion state should also be distributed over potentially continuous\ntime-windows.\n  This research explores the implication of acoustic context and phone\nboundaries on local markers for SER using an attention-based approach. The\nbenefits of using a distributed approach to speech emotion understanding are\nsupported by the results of cross-corpora analysis experiments. Experiments\nwhere phones and words are mapped to the attention vectors along with the\nfundamental frequency to observe the overlapping distributions and thereby the\nrelationship between acoustic context and emotion. This work aims to bridge\npsycholinguistic theory research with computational modelling for SER.\n", "date": "2023-06-30", "categories": []}
{"id": "2306.15796", "title": "ConKI: Contrastive Knowledge Injection for Multimodal Sentiment Analysis", "abstract": "  Multimodal Sentiment Analysis leverages multimodal signals to detect the\nsentiment of a speaker. Previous approaches concentrate on performing\nmultimodal fusion and representation learning based on general knowledge\nobtained from pretrained models, which neglects the effect of domain-specific\nknowledge. In this paper, we propose Contrastive Knowledge Injection (ConKI)\nfor multimodal sentiment analysis, where specific-knowledge representations for\neach modality can be learned together with general knowledge representations\nvia knowledge injection based on an adapter architecture. In addition, ConKI\nuses a hierarchical contrastive learning procedure performed between knowledge\ntypes within every single modality, across modalities within each sample, and\nacross samples to facilitate the effective learning of the proposed\nrepresentations, hence improving multimodal sentiment predictions. The\nexperiments on three popular multimodal sentiment analysis benchmarks show that\nConKI outperforms all prior methods on a variety of performance metrics.\n", "date": "2023-06-27", "categories": []}
{"id": "2306.15376", "title": "Exploiting Pseudo Future Contexts for Emotion Recognition in\n  Conversations", "abstract": "  With the extensive accumulation of conversational data on the Internet,\nemotion recognition in conversations (ERC) has received increasing attention.\nPrevious efforts of this task mainly focus on leveraging contextual and\nspeaker-specific features, or integrating heterogeneous external commonsense\nknowledge. Among them, some heavily rely on future contexts, which, however,\nare not always available in real-life scenarios. This fact inspires us to\ngenerate pseudo future contexts to improve ERC. Specifically, for an utterance,\nwe generate its future context with pre-trained language models, potentially\ncontaining extra beneficial knowledge in a conversational form homogeneous with\nthe historical ones. These characteristics make pseudo future contexts easily\nfused with historical contexts and historical speaker-specific contexts,\nyielding a conceptually simple framework systematically integrating\nmulti-contexts. Experimental results on four ERC datasets demonstrate our\nmethod's superiority. Further in-depth analyses reveal that pseudo future\ncontexts can rival real ones to some extent, especially in relatively\ncontext-independent conversations.\n", "date": "2023-06-27", "categories": []}
{"id": "2306.13959", "title": "Emotion Flip Reasoning in Multiparty Conversations", "abstract": "  In a conversational dialogue, speakers may have different emotional states\nand their dynamics play an important role in understanding dialogue's emotional\ndiscourse. However, simply detecting emotions is not sufficient to entirely\ncomprehend the speaker-specific changes in emotion that occur during a\nconversation. To understand the emotional dynamics of speakers in an efficient\nmanner, it is imperative to identify the rationale or instigator behind any\nchanges or flips in emotion expressed by the speaker. In this paper, we explore\nthe task called Instigator based Emotion Flip Reasoning (EFR), which aims to\nidentify the instigator behind a speaker's emotion flip within a conversation.\nFor example, an emotion flip from joy to anger could be caused by an instigator\nlike threat. To facilitate this task, we present MELD-I, a dataset that\nincludes ground-truth EFR instigator labels, which are in line with emotional\npsychology. To evaluate the dataset, we propose a novel neural architecture\ncalled TGIF, which leverages Transformer encoders and stacked GRUs to capture\nthe dialogue context, speaker dynamics, and emotion sequence in a conversation.\nOur evaluation demonstrates state-of-the-art performance (+4-12% increase in\nF1-score) against five baselines used for the task. Further, we establish the\ngeneralizability of TGIF on an unseen dataset in a zero-shot setting.\nAdditionally, we provide a detailed analysis of the competing models,\nhighlighting the advantages and limitations of our neural architecture.\n", "date": "2023-06-24", "categories": []}
{"id": "2306.14517", "title": "Cross-Lingual Cross-Age Group Adaptation for Low-Resource Elderly Speech\n  Emotion Recognition", "abstract": "  Speech emotion recognition plays a crucial role in human-computer\ninteractions. However, most speech emotion recognition research is biased\ntoward English-speaking adults, which hinders its applicability to other\ndemographic groups in different languages and age groups. In this work, we\nanalyze the transferability of emotion recognition across three different\nlanguages--English, Mandarin Chinese, and Cantonese; and 2 different age\ngroups--adults and the elderly. To conduct the experiment, we develop an\nEnglish-Mandarin speech emotion benchmark for adults and the elderly, BiMotion,\nand a Cantonese speech emotion dataset, YueMotion. This study concludes that\ndifferent language and age groups require specific speech features, thus making\ncross-lingual inference an unsuitable method. However, cross-group data\naugmentation is still beneficial to regularize the model, with linguistic\ndistance being a significant influence on cross-lingual transferability. We\nrelease publicly release our code at https://github.com/HLTCHKUST/elderly_ser.\n", "date": "2023-06-26", "categories": []}
{"id": "2306.14822", "title": "Label-Aware Hyperbolic Embeddings for Fine-grained Emotion\n  Classification", "abstract": "  Fine-grained emotion classification (FEC) is a challenging task.\nSpecifically, FEC needs to handle subtle nuance between labels, which can be\ncomplex and confusing. Most existing models only address text classification\nproblem in the euclidean space, which we believe may not be the optimal\nsolution as labels of close semantic (e.g., afraid and terrified) may not be\ndifferentiated in such space, which harms the performance. In this paper, we\npropose HypEmo, a novel framework that can integrate hyperbolic embeddings to\nimprove the FEC task. First, we learn label embeddings in the hyperbolic space\nto better capture their hierarchical structure, and then our model projects\ncontextualized representations to the hyperbolic space to compute the distance\nbetween samples and labels. Experimental results show that incorporating such\ndistance to weight cross entropy loss substantially improves the performance\nwith significantly higher efficiency. We evaluate our proposed model on two\nbenchmark datasets and found 4.8% relative improvement compared to the previous\nstate of the art with 43.2% fewer parameters and 76.9% less training time. Code\nis available at https: //github.com/dinobby/HypEmo.\n", "date": "2023-06-26", "categories": []}
{"id": "2306.13076", "title": "A Comparison of Time-based Models for Multimodal Emotion Recognition", "abstract": "  Emotion recognition has become an important research topic in the field of\nhuman-computer interaction. Studies on sound and videos to understand emotions\nfocused mainly on analyzing facial expressions and classified 6 basic emotions.\nIn this study, the performance of different sequence models in multi-modal\nemotion recognition was compared. The sound and images were first processed by\nmulti-layered CNN models, and the outputs of these models were fed into various\nsequence models. The sequence model is GRU, Transformer, LSTM and Max Pooling.\nAccuracy, precision, and F1 Score values of all models were calculated. The\nmulti-modal CREMA-D dataset was used in the experiments. As a result of the\ncomparison of the CREMA-D dataset, GRU-based architecture with 0.640 showed the\nbest result in F1 score, LSTM-based architecture with 0.699 in precision\nmetric, while sensitivity showed the best results over time with Max\nPooling-based architecture with 0.620. As a result, it has been observed that\nthe sequence models compare performances close to each other.\n", "date": "2023-06-22", "categories": []}
{"id": "2306.10397", "title": "Enhancing the Prediction of Emotional Experience in Movies using Deep\n  Neural Networks: The Significance of Audio and Language", "abstract": "  Our paper focuses on making use of deep neural network models to accurately\npredict the range of human emotions experienced during watching movies. In this\ncertain setup, there exist three clear-cut input modalities that considerably\ninfluence the experienced emotions: visual cues derived from RGB video frames,\nauditory components encompassing sounds, speech, and music, and linguistic\nelements encompassing actors' dialogues. Emotions are commonly described using\na two-factor model including valence (ranging from happy to sad) and arousal\n(indicating the intensity of the emotion). In this regard, a Plethora of works\nhave presented a multitude of models aiming to predict valence and arousal from\nvideo content. However, non of these models contain all three modalities, with\nlanguage being consistently eliminated across all of them. In this study, we\ncomprehensively combine all modalities and conduct an analysis to ascertain the\nimportance of each in predicting valence and arousal. Making use of pre-trained\nneural networks, we represent each input modality in our study. In order to\nprocess visual input, we employ pre-trained convolutional neural networks to\nrecognize scenes[1], objects[2], and actions[3,4]. For audio processing, we\nutilize a specialized neural network designed for handling sound-related tasks,\nnamely SoundNet[5]. Finally, Bidirectional Encoder Representations from\nTransformers (BERT) models are used to extract linguistic features[6] in our\nanalysis. We report results on the COGNIMUSE dataset[7], where our proposed\nmodel outperforms the current state-of-the-art approaches. Surprisingly, our\nfindings reveal that language significantly influences the experienced arousal,\nwhile sound emerges as the primary determinant for predicting valence. In\ncontrast, the visual modality exhibits the least impact among all modalities in\npredicting emotions.\n", "date": "2023-06-17", "categories": []}
{"id": "2306.09372", "title": "SAFER: Situation Aware Facial Emotion Recognition", "abstract": "  In this paper, we present SAFER, a novel system for emotion recognition from\nfacial expressions. It employs state-of-the-art deep learning techniques to\nextract various features from facial images and incorporates contextual\ninformation, such as background and location type, to enhance its performance.\nThe system has been designed to operate in an open-world setting, meaning it\ncan adapt to unseen and varied facial expressions, making it suitable for\nreal-world applications. An extensive evaluation of SAFER against existing\nworks in the field demonstrates improved performance, achieving an accuracy of\n91.4% on the CAER-S dataset. Additionally, the study investigates the effect of\nnovelty such as face masks during the Covid-19 pandemic on facial emotion\nrecognition and critically examines the limitations of mainstream facial\nexpressions datasets. To address these limitations, a novel dataset for facial\nemotion recognition is proposed. The proposed dataset and the system are\nexpected to be useful for various applications such as human-computer\ninteraction, security, and surveillance.\n", "date": "2023-06-14", "categories": []}
{"id": "2306.09385", "title": "Employing Multimodal Machine Learning for Stress Detection", "abstract": "  In the current age, human lifestyle has become more knowledge oriented\nleading to generation of sedentary employment. This has given rise to a number\nof health and mental disorders. Mental wellness is one of the most neglected\nbut crucial aspects of today's world. Mental health issues can, both directly\nand indirectly, affect other sections of human physiology and impede an\nindividual's day-to-day activities and performance. However, identifying the\nstress and finding the stress trend for an individual leading to serious mental\nailments is challenging and involves multiple factors. Such identification can\nbe achieved accurately by fusing these multiple modalities (due to various\nfactors) arising from behavioral patterns. Certain techniques are identified in\nthe literature for this purpose; however, very few machine learning-based\nmethods are proposed for such multimodal fusion tasks. In this work, a\nmultimodal AI-based framework is proposed to monitor a person's working\nbehavior and stress levels. We propose a methodology for efficiently detecting\nstress due to workload by concatenating heterogeneous raw sensor data streams\n(e.g., face expressions, posture, heart rate, computer interaction). This data\ncan be securely stored and analyzed to understand and discover personalized\nunique behavioral patterns leading to mental strain and fatigue. The\ncontribution of this work is twofold; proposing a multimodal AI-based strategy\nfor fusion to detect stress and its level and secondly identify a stress\npattern over a period of time. We were able to achieve 96.09% accuracy on the\ntest set in stress detection and classification. Further, we reduce the stress\nscale prediction model loss to 0.036 using these modalities. This work can\nprove important for the community at large, specifically those working\nsedentary jobs to monitor and identify stress levels, especially in current\ntimes of COVID-19.\n", "date": "2023-06-15", "categories": []}
{"id": "2306.08657", "title": "EMERSK -- Explainable Multimodal Emotion Recognition with Situational\n  Knowledge", "abstract": "  Automatic emotion recognition has recently gained significant attention due\nto the growing popularity of deep learning algorithms. One of the primary\nchallenges in emotion recognition is effectively utilizing the various cues\n(modalities) available in the data. Another challenge is providing a proper\nexplanation of the outcome of the learning.To address these challenges, we\npresent Explainable Multimodal Emotion Recognition with Situational Knowledge\n(EMERSK), a generalized and modular system for human emotion recognition and\nexplanation using visual information. Our system can handle multiple\nmodalities, including facial expressions, posture, and gait, in a flexible and\nmodular manner. The network consists of different modules that can be added or\nremoved depending on the available data. We utilize a two-stream network\narchitecture with convolutional neural networks (CNNs) and encoder-decoder\nstyle attention mechanisms to extract deep features from face images.\nSimilarly, CNNs and recurrent neural networks (RNNs) with Long Short-term\nMemory (LSTM) are employed to extract features from posture and gait data. We\nalso incorporate deep features from the background as contextual information\nfor the learning process. The deep features from each module are fused using an\nearly fusion network. Furthermore, we leverage situational knowledge derived\nfrom the location type and adjective-noun pair (ANP) extracted from the scene,\nas well as the spatio-temporal average distribution of emotions, to generate\nexplanations. Ablation studies demonstrate that each sub-network can\nindependently perform emotion recognition, and combining them in a multimodal\napproach significantly improves overall recognition performance. Extensive\nexperiments conducted on various benchmark datasets, including GroupWalk,\nvalidate the superior performance of our approach compared to other\nstate-of-the-art methods.\n", "date": "2023-06-14", "categories": []}
{"id": "2306.08733", "title": "Continuous Learning Based Novelty Aware Emotion Recognition System", "abstract": "  Current works in human emotion recognition follow the traditional closed\nlearning approach governed by rigid rules without any consideration of novelty.\nClassification models are trained on some collected datasets and expected to\nhave the same data distribution in the real-world deployment. Due to the fluid\nand constantly changing nature of the world we live in, it is possible to have\nunexpected and novel sample distribution which can lead the model to fail.\nHence, in this work, we propose a continuous learning based approach to deal\nwith novelty in the automatic emotion recognition task.\n", "date": "2023-06-14", "categories": []}
{"id": "2209.15449", "title": "End-to-End Label Uncertainty Modeling in Speech Emotion Recognition\n  using Bayesian Neural Networks and Label Distribution Learning", "abstract": "  To train machine learning algorithms to predict emotional expressions in\nterms of arousal and valence, annotated datasets are needed. However, as\ndifferent people perceive others' emotional expressions differently, their\nannotations are subjective. To account for this, annotations are typically\ncollected from multiple annotators and averaged to obtain ground-truth labels.\nHowever, when exclusively trained on this averaged ground-truth, the model is\nagnostic to the inherent subjectivity in emotional expressions. In this work,\nwe therefore propose an end-to-end Bayesian neural network capable of being\ntrained on a distribution of annotations to also capture the subjectivity-based\nlabel uncertainty. Instead of a Gaussian, we model the annotation distribution\nusing Student's t-distribution, which also accounts for the number of\nannotations available. We derive the corresponding Kullback-Leibler divergence\nloss and use it to train an estimator for the annotation distribution, from\nwhich the mean and uncertainty can be inferred. We validate the proposed method\nusing two in-the-wild datasets. We show that the proposed t-distribution based\napproach achieves state-of-the-art uncertainty modeling results in speech\nemotion recognition, and also consistent results in cross-corpora evaluations.\nFurthermore, analyses reveal that the advantage of a t-distribution over a\nGaussian grows with increasing inter-annotator correlation and a decreasing\nnumber of annotations available.\n", "date": "2022-09-30", "categories": []}
{"id": "2210.06405", "title": "Transformer-based Text Classification on Unified Bangla Multi-class\n  Emotion Corpus", "abstract": "  In this research, we propose a complete set of approaches for identifying and\nextracting emotions from Bangla texts. We provide a Bangla emotion classifier\nfor six classes: anger, disgust, fear, joy, sadness, and surprise, from Bangla\nwords using transformer-based models, which exhibit phenomenal results in\nrecent days, especially for high-resource languages. The Unified Bangla\nMulti-class Emotion Corpus (UBMEC) is used to assess the performance of our\nmodels. UBMEC is created by combining two previously released manually labeled\ndatasets of Bangla comments on six emotion classes with fresh manually labeled\nBangla comments created by us. The corpus dataset and code we used in this work\nare publicly available.\n", "date": "2022-10-12", "categories": []}
{"id": "2306.06583", "title": "REACT2023: the first Multi-modal Multiple Appropriate Facial Reaction\n  Generation Challenge", "abstract": "  The Multi-modal Multiple Appropriate Facial Reaction Generation Challenge\n(REACT2023) is the first competition event focused on evaluating multimedia\nprocessing and machine learning techniques for generating human-appropriate\nfacial reactions in various dyadic interaction scenarios, with all participants\ncompeting strictly under the same conditions. The goal of the challenge is to\nprovide the first benchmark test set for multi-modal information processing and\nto foster collaboration among the audio, visual, and audio-visual affective\ncomputing communities, to compare the relative merits of the approaches to\nautomatic appropriate facial reaction generation under different spontaneous\ndyadic interaction conditions. This paper presents: (i) novelties,\ncontributions and guidelines of the REACT2023 challenge; (ii) the dataset\nutilized in the challenge; and (iii) the performance of baseline systems on the\ntwo proposed sub-challenges: Offline Multiple Appropriate Facial Reaction\nGeneration and Online Multiple Appropriate Facial Reaction Generation,\nrespectively. The challenge baseline code is publicly available at\n\\url{https://github.com/reactmultimodalchallenge/baseline_react2023}.\n", "date": "2023-06-11", "categories": []}
{"id": "2306.06601", "title": "Mimicking the Thinking Process for Emotion Recognition in Conversation\n  with Prompts and Paraphrasing", "abstract": "  Emotion recognition in conversation, which aims to predict the emotion for\nall utterances, has attracted considerable research attention in recent years.\nIt is a challenging task since the recognition of the emotion in one utterance\ninvolves many complex factors, such as the conversational context, the\nspeaker's background, and the subtle difference between emotion labels. In this\npaper, we propose a novel framework which mimics the thinking process when\nmodeling these factors. Specifically, we first comprehend the conversational\ncontext with a history-oriented prompt to selectively gather information from\npredecessors of the target utterance. We then model the speaker's background\nwith an experience-oriented prompt to retrieve the similar utterances from all\nconversations. We finally differentiate the subtle label semantics with a\nparaphrasing mechanism to elicit the intrinsic label related knowledge. We\nconducted extensive experiments on three benchmarks. The empirical results\ndemonstrate the superiority of our proposed framework over the state-of-the-art\nbaselines.\n", "date": "2023-06-11", "categories": []}
{"id": "2306.06696", "title": "Toward Fair Facial Expression Recognition with Improved Distribution\n  Alignment", "abstract": "  We present a novel approach to mitigate bias in facial expression recognition\n(FER) models. Our method aims to reduce sensitive attribute information such as\ngender, age, or race, in the embeddings produced by FER models. We employ a\nkernel mean shrinkage estimator to estimate the kernel mean of the\ndistributions of the embeddings associated with different sensitive attribute\ngroups, such as young and old, in the Hilbert space. Using this estimation, we\ncalculate the maximum mean discrepancy (MMD) distance between the distributions\nand incorporate it in the classifier loss along with an adversarial loss, which\nis then minimized through the learning process to improve the distribution\nalignment. Our method makes sensitive attributes less recognizable for the\nmodel, which in turn promotes fairness. Additionally, for the first time, we\nanalyze the notion of attractiveness as an important sensitive attribute in FER\nmodels and demonstrate that FER models can indeed exhibit biases towards more\nattractive faces. To prove the efficacy of our model in reducing bias regarding\ndifferent sensitive attributes (including the newly proposed attractiveness\nattribute), we perform several experiments on two widely used datasets, CelebA\nand RAF-DB. The results in terms of both accuracy and fairness measures\noutperform the state-of-the-art in most cases, demonstrating the effectiveness\nof the proposed method.\n", "date": "2023-06-11", "categories": []}
{"id": "2306.07115", "title": "Exploring Attention Mechanisms for Multimodal Emotion Recognition in an\n  Emergency Call Center Corpus", "abstract": "  The emotion detection technology to enhance human decision-making is an\nimportant research issue for real-world applications, but real-life emotion\ndatasets are relatively rare and small. The experiments conducted in this paper\nuse the CEMO, which was collected in a French emergency call center. Two\npre-trained models based on speech and text were fine-tuned for speech emotion\nrecognition. Using pre-trained Transformer encoders mitigates our data's\nlimited and sparse nature. This paper explores the different fusion strategies\nof these modality-specific models. In particular, fusions with and without\ncross-attention mechanisms were tested to gather the most relevant information\nfrom both the speech and text encoders. We show that multimodal fusion brings\nan absolute gain of 4-9% with respect to either single modality and that the\nSymmetric multi-headed cross-attention mechanism performed better than late\nclassical fusion approaches. Our experiments also suggest that for the\nreal-life CEMO corpus, the audio component encodes more emotive information\nthan the textual one.\n", "date": "2023-06-12", "categories": []}
{"id": "2306.05709", "title": "Learning Emotional Representations from Imbalanced Speech Data for\n  Speech Emotion Recognition and Emotional Text-to-Speech", "abstract": "  Effective speech emotional representations play a key role in Speech Emotion\nRecognition (SER) and Emotional Text-To-Speech (TTS) tasks. However, emotional\nspeech samples are more difficult and expensive to acquire compared with\nNeutral style speech, which causes one issue that most related works\nunfortunately neglect: imbalanced datasets. Models might overfit to the\nmajority Neutral class and fail to produce robust and effective emotional\nrepresentations. In this paper, we propose an Emotion Extractor to address this\nissue. We use augmentation approaches to train the model and enable it to\nextract effective and generalizable emotional representations from imbalanced\ndatasets. Our empirical results show that (1) for the SER task, the proposed\nEmotion Extractor surpasses the state-of-the-art baseline on three imbalanced\ndatasets; (2) the produced representations from our Emotion Extractor benefit\nthe TTS model, and enable it to synthesize more expressive speech.\n", "date": "2023-06-09", "categories": []}
{"id": "2306.04340", "title": "Co-evolving Graph Reasoning Network for Emotion-Cause Pair Extraction", "abstract": "  Emotion-Cause Pair Extraction (ECPE) aims to extract all emotion clauses and\ntheir corresponding cause clauses from a document. Existing approaches tackle\nthis task through multi-task learning (MTL) framework in which the two subtasks\nprovide indicative clues for ECPE. However, the previous MTL framework\nconsiders only one round of multi-task reasoning and ignores the reverse\nfeedbacks from ECPE to the subtasks. Besides, its multi-task reasoning only\nrelies on semantics-level interactions, which cannot capture the explicit\ndependencies, and both the encoder sharing and multi-task hidden states\nconcatenations can hardly capture the causalities. To solve these issues, we\nfirst put forward a new MTL framework based on Co-evolving Reasoning. It (1)\nmodels the bidirectional feedbacks between ECPE and its subtasks; (2) allows\nthe three tasks to evolve together and prompt each other recurrently; (3)\nintegrates prediction-level interactions to capture explicit dependencies. Then\nwe propose a novel multi-task relational graph (MRG) to sufficiently exploit\nthe causal relations. Finally, we propose a Co-evolving Graph Reasoning Network\n(CGR-Net) that implements our MTL framework and conducts Co-evolving Reasoning\non MRG. Experimental results show that our model achieves new state-of-the-art\nperformance, and further analysis confirms the advantages of our method.\n", "date": "2023-06-07", "categories": []}
{"id": "2305.03506", "title": "SI-LSTM: Speaker Hybrid Long-short Term Memory and Cross Modal Attention\n  for Emotion Recognition in Conversation", "abstract": "  Emotion Recognition in Conversation~(ERC) across modalities is of vital\nimportance for a variety of applications, including intelligent healthcare,\nartificial intelligence for conversation, and opinion mining over chat history.\nThe crux of ERC is to model both cross-modality and cross-time interactions\nthroughout the conversation. Previous methods have made progress in learning\nthe time series information of conversation while lacking the ability to trace\ndown the different emotional states of each speaker in a conversation. In this\npaper, we propose a recurrent structure called Speaker Information Enhanced\nLong-Short Term Memory (SI-LSTM) for the ERC task, where the emotional states\nof the distinct speaker can be tracked in a sequential way to enhance the\nlearning of the emotion in conversation. Further, to improve the learning of\nmultimodal features in ERC, we utilize a cross-modal attention component to\nfuse the features between different modalities and model the interaction of the\nimportant information from different modalities. Experimental results on two\nbenchmark datasets demonstrate the superiority of the proposed SI-LSTM against\nthe state-of-the-art baseline methods in the ERC task on multimodal data.\n", "date": "2023-05-04", "categories": []}
{"id": "2306.03112", "title": "Synthesizing Affective Neurophysiological Signals Using Generative\n  Models: A Review Paper", "abstract": "  The integration of emotional intelligence in machines is an important step in\nadvancing human-computer interaction. This demands the development of reliable\nend-to-end emotion recognition systems. However, the scarcity of public\naffective datasets presents a challenge. In this literature review, we\nemphasize the use of generative models to address this issue in\nneurophysiological signals, particularly Electroencephalogram (EEG) and\nFunctional Near-Infrared Spectroscopy (fNIRS). We provide a comprehensive\nanalysis of different generative models used in the field, examining their\ninput formulation, deployment strategies, and methodologies for evaluating the\nquality of synthesized data. This review serves as a comprehensive overview,\noffering insights into the advantages, challenges, and promising future\ndirections in the application of generative models in emotion recognition\nsystems. Through this review, we aim to facilitate the progression of\nneurophysiological data augmentation, thereby supporting the development of\nmore efficient and reliable emotion recognition systems.\n", "date": "2023-06-05", "categories": []}
{"id": "2306.03650", "title": "A Quantum Probability Driven Framework for Joint Multi-Modal Sarcasm,\n  Sentiment and Emotion Analysis", "abstract": "  Sarcasm, sentiment, and emotion are three typical kinds of spontaneous\naffective responses of humans to external events and they are tightly\nintertwined with each other. Such events may be expressed in multiple\nmodalities (e.g., linguistic, visual and acoustic), e.g., multi-modal\nconversations. Joint analysis of humans' multi-modal sarcasm, sentiment, and\nemotion is an important yet challenging topic, as it is a complex cognitive\nprocess involving both cross-modality interaction and cross-affection\ncorrelation. From the probability theory perspective, cross-affection\ncorrelation also means that the judgments on sarcasm, sentiment, and emotion\nare incompatible. However, this exposed phenomenon cannot be sufficiently\nmodelled by classical probability theory due to its assumption of\ncompatibility. Neither do the existing approaches take it into consideration.\nIn view of the recent success of quantum probability (QP) in modeling human\ncognition, particularly contextual incompatible decision making, we take the\nfirst step towards introducing QP into joint multi-modal sarcasm, sentiment,\nand emotion analysis. Specifically, we propose a QUantum probabIlity driven\nmulti-modal sarcasm, sEntiment and emoTion analysis framework, termed QUIET.\nExtensive experiments on two datasets and the results show that the\neffectiveness and advantages of QUIET in comparison with a wide range of the\nstate-of-the-art baselines. We also show the great potential of QP in\nmulti-affect analysis.\n", "date": "2023-06-06", "categories": []}
{"id": "2303.10644", "title": "Spatio-Temporal AU Relational Graph Representation Learning For Facial\n  Action Units Detection", "abstract": "  This paper presents our Facial Action Units (AUs) detection submission to the\nfifth Affective Behavior Analysis in-the-wild Competition (ABAW). Our approach\nconsists of three main modules: (i) a pre-trained facial representation encoder\nwhich produce a strong facial representation from each input face image in the\ninput sequence; (ii) an AU-specific feature generator that specifically learns\na set of AU features from each facial representation; and (iii) a\nspatio-temporal graph learning module that constructs a spatio-temporal graph\nrepresentation. This graph representation describes AUs contained in all frames\nand predicts the occurrence of each AU based on both the modeled spatial\ninformation within the corresponding face and the learned temporal dynamics\namong frames. The experimental results show that our approach outperformed the\nbaseline and the spatio-temporal graph representation learning allows our model\nto generate the best results among all ablated systems. Our model ranks at the\n4th place in the AU recognition track at the 5th ABAW Competition. Our code is\npublicly available at https://github.com/wzh125/ABAW-5.\n", "date": "2023-03-19", "categories": []}
{"id": "2304.08216", "title": "Context-Dependent Embedding Utterance Representations for Emotion\n  Recognition in Conversations", "abstract": "  Emotion Recognition in Conversations (ERC) has been gaining increasing\nimportance as conversational agents become more and more common. Recognizing\nemotions is key for effective communication, being a crucial component in the\ndevelopment of effective and empathetic conversational agents. Knowledge and\nunderstanding of the conversational context are extremely valuable for\nidentifying the emotions of the interlocutor. We thus approach Emotion\nRecognition in Conversations leveraging the conversational context, i.e.,\ntaking into attention previous conversational turns. The usual approach to\nmodel the conversational context has been to produce context-independent\nrepresentations of each utterance and subsequently perform contextual modeling\nof these. Here we propose context-dependent embedding representations of each\nutterance by leveraging the contextual representational power of pre-trained\ntransformer language models. In our approach, we feed the conversational\ncontext appended to the utterance to be classified as input to the RoBERTa\nencoder, to which we append a simple classification module, thus discarding the\nneed to deal with context after obtaining the embeddings since these constitute\nalready an efficient representation of such context. We also investigate how\nthe number of introduced conversational turns influences our model performance.\nThe effectiveness of our approach is validated on the open-domain DailyDialog\ndataset and on the task-oriented EmoWOZ dataset.\n", "date": "2023-04-17", "categories": []}
{"id": "2306.01916", "title": "In-the-wild Speech Emotion Conversion Using Disentangled Self-Supervised\n  Representations and Neural Vocoder-based Resynthesis", "abstract": "  Speech emotion conversion aims to convert the expressed emotion of a spoken\nutterance to a target emotion while preserving the lexical information and the\nspeaker's identity. In this work, we specifically focus on in-the-wild emotion\nconversion where parallel data does not exist, and the problem of disentangling\nlexical, speaker, and emotion information arises. In this paper, we introduce a\nmethodology that uses self-supervised networks to disentangle the lexical,\nspeaker, and emotional content of the utterance, and subsequently uses a\nHiFiGAN vocoder to resynthesise the disentangled representations to a speech\nsignal of the targeted emotion. For better representation and to achieve\nemotion intensity control, we specifically focus on the aro\\-usal dimension of\ncontinuous representations, as opposed to performing emotion conversion on\ncategorical representations. We test our methodology on the large in-the-wild\nMSP-Podcast dataset. Results reveal that the proposed approach is aptly\nconditioned on the emotional content of input speech and is capable of\nsynthesising natural-sounding speech for a target emotion. Results further\nreveal that the methodology better synthesises speech for mid-scale arousal (2\nto 6) than for extreme arousal (1 and 7).\n", "date": "2023-06-02", "categories": []}
{"id": "2306.02193", "title": "LDEB -- Label Digitization with Emotion Binarization and Machine\n  Learning for Emotion Recognition in Conversational Dialogues", "abstract": "  Emotion recognition in conversations (ERC) is vital to the advancements of\nconversational AI and its applications. Therefore, the development of an\nautomated ERC model using the concepts of machine learning (ML) would be\nbeneficial. However, the conversational dialogues present a unique problem\nwhere each dialogue depicts nested emotions that entangle the association\nbetween the emotional feature descriptors and emotion type (or label). This\nentanglement that can be multiplied with the presence of data paucity is an\nobstacle for a ML model. To overcome this problem, we proposed a novel approach\ncalled Label Digitization with Emotion Binarization (LDEB) that disentangles\nthe twists by utilizing the text normalization and 7-bit digital encoding\ntechniques and constructs a meaningful feature space for a ML model to be\ntrained. We also utilized the publicly available dataset called the\nFETA-DailyDialog dataset for feature learning and developed a hierarchical ERC\nmodel using random forest (RF) and artificial neural network (ANN) classifiers.\nSimulations showed that the ANN-based ERC model was able to predict emotion\nwith the best accuracy and precision scores of about 74% and 76%, respectively.\nSimulations also showed that the ANN-model could reach a training accuracy\nscore of about 98% with 60 epochs. On the other hand, the RF-based ERC model\nwas able to predict emotions with the best accuracy and precision scores of\nabout 78% and 75%, respectively.\n", "date": "2023-06-03", "categories": []}
{"id": "2306.02845", "title": "Interpretable Multimodal Emotion Recognition using Facial Features and\n  Physiological Signals", "abstract": "  This paper aims to demonstrate the importance and feasibility of fusing\nmultimodal information for emotion recognition. It introduces a multimodal\nframework for emotion understanding by fusing the information from visual\nfacial features and rPPG signals extracted from the input videos. An\ninterpretability technique based on permutation feature importance analysis has\nalso been implemented to compute the contributions of rPPG and visual\nmodalities toward classifying a given input video into a particular emotion\nclass. The experiments on IEMOCAP dataset demonstrate that the emotion\nclassification performance improves by combining the complementary information\nfrom multiple modalities.\n", "date": "2023-06-05", "categories": []}
{"id": "2305.02572", "title": "High-fidelity Generalized Emotional Talking Face Generation with\n  Multi-modal Emotion Space Learning", "abstract": "  Recently, emotional talking face generation has received considerable\nattention. However, existing methods only adopt one-hot coding, image, or audio\nas emotion conditions, thus lacking flexible control in practical applications\nand failing to handle unseen emotion styles due to limited semantics. They\neither ignore the one-shot setting or the quality of generated faces. In this\npaper, we propose a more flexible and generalized framework. Specifically, we\nsupplement the emotion style in text prompts and use an Aligned Multi-modal\nEmotion encoder to embed the text, image, and audio emotion modality into a\nunified space, which inherits rich semantic prior from CLIP. Consequently,\neffective multi-modal emotion space learning helps our method support arbitrary\nemotion modality during testing and could generalize to unseen emotion styles.\nBesides, an Emotion-aware Audio-to-3DMM Convertor is proposed to connect the\nemotion condition and the audio sequence to structural representation. A\nfollowed style-based High-fidelity Emotional Face generator is designed to\ngenerate arbitrary high-resolution realistic identities. Our texture generator\nhierarchically learns flow fields and animated faces in a residual manner.\nExtensive experiments demonstrate the flexibility and generalization of our\nmethod in emotion control and the effectiveness of high-quality face synthesis.\n", "date": "2023-05-04", "categories": []}
{"id": "2207.01113", "title": "Are 3D Face Shapes Expressive Enough for Recognising Continuous Emotions\n  and Action Unit Intensities?", "abstract": "  Recognising continuous emotions and action unit (AU) intensities from face\nvideos requires a spatial and temporal understanding of expression dynamics.\nExisting works primarily rely on 2D face appearances to extract such dynamics.\nThis work focuses on a promising alternative based on parametric 3D face shape\nalignment models, which disentangle different factors of variation, including\nexpression-induced shape variations. We aim to understand how expressive 3D\nface shapes are in estimating valence-arousal and AU intensities compared to\nthe state-of-the-art 2D appearance-based models. We benchmark four recent 3D\nface alignment models: ExpNet, 3DDFA-V2, DECA, and EMOCA. In valence-arousal\nestimation, expression features of 3D face models consistently surpassed\nprevious works and yielded an average concordance correlation of .739 and .574\non SEWA and AVEC 2019 CES corpora, respectively. We also study how 3D face\nshapes performed on AU intensity estimation on BP4D and DISFA datasets, and\nreport that 3D face features were on par with 2D appearance features in AUs 4,\n6, 10, 12, and 25, but not the entire set of AUs. To understand this\ndiscrepancy, we conduct a correspondence analysis between valence-arousal and\nAUs, which points out that accurate prediction of valence-arousal may require\nthe knowledge of only a few AUs.\n", "date": "2022-07-03", "categories": []}
{"id": "2305.14071", "title": "Disentangled Variational Autoencoder for Emotion Recognition in\n  Conversations", "abstract": "  In Emotion Recognition in Conversations (ERC), the emotions of target\nutterances are closely dependent on their context. Therefore, existing works\ntrain the model to generate the response of the target utterance, which aims to\nrecognise emotions leveraging contextual information. However, adjacent\nresponse generation ignores long-range dependencies and provides limited\naffective information in many cases. In addition, most ERC models learn a\nunified distributed representation for each utterance, which lacks\ninterpretability and robustness. To address these issues, we propose a\nVAD-disentangled Variational AutoEncoder (VAD-VAE), which first introduces a\ntarget utterance reconstruction task based on Variational Autoencoder, then\ndisentangles three affect representations Valence-Arousal-Dominance (VAD) from\nthe latent space. We also enhance the disentangled representations by\nintroducing VAD supervision signals from a sentiment lexicon and minimising the\nmutual information between VAD distributions. Experiments show that VAD-VAE\noutperforms the state-of-the-art model on two datasets. Further analysis proves\nthe effectiveness of each proposed module and the quality of disentangled VAD\nrepresentations. The code is available at\nhttps://github.com/SteveKGYang/VAD-VAE.\n", "date": "2023-05-23", "categories": []}
{"id": "2305.16065", "title": "ASR and Emotional Speech: A Word-Level Investigation of the Mutual\n  Impact of Speech and Emotion Recognition", "abstract": "  In Speech Emotion Recognition (SER), textual data is often used alongside\naudio signals to address their inherent variability. However, the reliance on\nhuman annotated text in most research hinders the development of practical SER\nsystems. To overcome this challenge, we investigate how Automatic Speech\nRecognition (ASR) performs on emotional speech by analyzing the ASR performance\non emotion corpora and examining the distribution of word errors and confidence\nscores in ASR transcripts to gain insight into how emotion affects ASR. We\nutilize four ASR systems, namely Kaldi ASR, wav2vec2, Conformer, and Whisper,\nand three corpora: IEMOCAP, MOSI, and MELD to ensure generalizability.\nAdditionally, we conduct text-based SER on ASR transcripts with increasing word\nerror rates to investigate how ASR affects SER. The objective of this study is\nto uncover the relationship and mutual impact of ASR and SER, in order to\nfacilitate ASR adaptation to emotional speech and the use of SER in real world.\n", "date": "2023-05-25", "categories": []}
{"id": "2305.17580", "title": "ArPanEmo: An Open-Source Dataset for Fine-Grained Emotion Recognition in\n  Arabic Online Content during COVID-19 Pandemic", "abstract": "  Emotion recognition is a crucial task in Natural Language Processing (NLP)\nthat enables machines to comprehend the feelings conveyed in the text. The\napplications of emotion recognition are diverse, including mental health\ndiagnosis, student support, and the detection of online suspicious behavior.\nDespite the substantial amount of literature available on emotion recognition\nin various languages, Arabic emotion recognition has received relatively little\nattention, leading to a scarcity of emotion-annotated corpora. This paper\npresents the ArPanEmo dataset, a novel dataset for fine-grained emotion\nrecognition of online posts in Arabic. The dataset comprises 11,128 online\nposts manually labeled for ten emotion categories or neutral, with Fleiss'\nkappa of 0.71. It targets a specific Arabic dialect and addresses topics\nrelated to the COVID-19 pandemic, making it the first and largest of its kind.\nPython's packages were utilized to collect online posts related to the COVID-19\npandemic from three sources: Twitter, YouTube, and online newspaper comments\nbetween March 2020 and March 2022. Upon collection of the online posts, each\none underwent a semi-automatic classification process using a lexicon of\nemotion-related terms to determine whether it belonged to the neutral or\nemotional category. Subsequently, manual labeling was conducted to further\ncategorize the emotional data into fine-grained emotion categories.\n", "date": "2023-05-27", "categories": []}
{"id": "2305.17895", "title": "ReSup: Reliable Label Noise Suppression for Facial Expression\n  Recognition", "abstract": "  Because of the ambiguous and subjective property of the facial expression\nrecognition (FER) task, the label noise is widely existing in the FER dataset.\nFor this problem, in the training phase, current FER methods often directly\npredict whether the label of the input image is noised or not, aiming to reduce\nthe contribution of the noised data in training. However, we argue that this\nkind of method suffers from the low reliability of such noise data decision\noperation. It makes that some mistakenly abounded clean data are not utilized\nsufficiently and some mistakenly kept noised data disturbing the model learning\nprocess. In this paper, we propose a more reliable noise-label suppression\nmethod called ReSup (Reliable label noise Suppression for FER). First, instead\nof directly predicting noised or not, ReSup makes the noise data decision by\nmodeling the distribution of noise and clean labels simultaneously according to\nthe disagreement between the prediction and the target. Specifically, to\nachieve optimal distribution modeling, ReSup models the similarity distribution\nof all samples. To further enhance the reliability of our noise decision\nresults, ReSup uses two networks to jointly achieve noise suppression.\nSpecifically, ReSup utilize the property that two networks are less likely to\nmake the same mistakes, making two networks swap decisions and tending to trust\ndecisions with high agreement. Extensive experiments on three popular\nbenchmarks show that the proposed method significantly outperforms\nstate-of-the-art noisy label FER methods by 3.01% on FERPlus becnmarks. Code:\nhttps://github.com/purpleleaves007/FERDenoise\n", "date": "2023-05-29", "categories": []}
{"id": "2210.12078", "title": "Experiencer-Specific Emotion and Appraisal Prediction", "abstract": "  Emotion classification in NLP assigns emotions to texts, such as sentences or\nparagraphs. With texts like \"I felt guilty when he cried\", focusing on the\nsentence level disregards the standpoint of each participant in the situation:\nthe writer (\"I\") and the other entity (\"he\") could in fact have different\naffective states. The emotions of different entities have been considered only\npartially in emotion semantic role labeling, a task that relates semantic roles\nto emotion cue words. Proposing a related task, we narrow the focus on the\nexperiencers of events, and assign an emotion (if any holds) to each of them.\nTo this end, we represent each emotion both categorically and with appraisal\nvariables, as a psychological access to explaining why a person develops a\nparticular emotion. On an event description corpus, our experiencer-aware\nmodels of emotions and appraisals outperform the experiencer-agnostic\nbaselines, showing that disregarding event participants is an\noversimplification for the emotion detection task.\n", "date": "2022-10-21", "categories": []}
{"id": "2302.10021", "title": "Medical Face Masks and Emotion Recognition from the Body: Insights from\n  a Deep Learning Perspective", "abstract": "  The COVID-19 pandemic has undoubtedly changed the standards and affected all\naspects of our lives, especially social communication. It has forced people to\nextensively wear medical face masks, in order to prevent transmission. This\nface occlusion can strongly irritate emotional reading from the face and urges\nus to incorporate the whole body as an emotional cue. In this paper, we conduct\ninsightful studies about the effect of face occlusion on emotion recognition\nperformance, and showcase the superiority of full body input over the plain\nmasked face. We utilize a deep learning model based on the Temporal Segment\nNetwork framework, and aspire to fully overcome the face mask consequences.\nAlthough facial and bodily features can be learned from a single input, this\nmay lead to irrelevant information confusion. By processing those features\nseparately and fusing their prediction scores, we are more effectively taking\nadvantage of both modalities. This framework also naturally supports temporal\nmodeling, by mingling information among neighboring frames. In combination,\nthese techniques form an effective system capable of tackling emotion\nrecognition difficulties, caused by safety protocols applied in crucial areas.\n", "date": "2023-02-20", "categories": []}
{"id": "2305.12540", "title": "On the Efficacy and Noise-Robustness of Jointly Learned Speech Emotion\n  and Automatic Speech Recognition", "abstract": "  New-age conversational agent systems perform both speech emotion recognition\n(SER) and automatic speech recognition (ASR) using two separate and often\nindependent approaches for real-world application in noisy environments. In\nthis paper, we investigate a joint ASR-SER multitask learning approach in a\nlow-resource setting and show that improvements are observed not only in SER,\nbut also in ASR. We also investigate the robustness of such jointly trained\nmodels to the presence of background noise, babble, and music. Experimental\nresults on the IEMOCAP dataset show that joint learning can improve ASR word\nerror rate (WER) and SER classification accuracy by 10.7% and 2.3% respectively\nin clean scenarios. In noisy scenarios, results on data augmented with MUSAN\nshow that the joint approach outperforms the independent ASR and SER approaches\nacross many noisy conditions. Overall, the joint ASR-SER approach yielded more\nnoise-resistant models than the independent ASR and SER approaches.\n", "date": "2023-05-21", "categories": []}
{"id": "2305.08112", "title": "Quantum Operation of Affective Artificial Intelligence", "abstract": "  The review analyzes the fundamental principles which Artificial Intelligence\nshould be based on in order to imitate the realistic process of taking\ndecisions by humans experiencing emotions. Two approaches are compared, one\nbased on quantum theory and the other employing classical terms. Both these\napproaches have a number of similarities, being principally probabilistic. The\nanalogies between quantum measurements under intrinsic noise and affective\ndecision making are elucidated. It is shown that cognitive processes have many\nfeatures that are formally similar to quantum measurements. This, however, in\nno way means that for the imitation of human decision making Affective\nArtificial Intelligence has necessarily to rely on the functioning of quantum\nsystems. Appreciating the common features between quantum measurements and\ndecision making helps for the formulation of an axiomatic approach employing\nonly classical notions. Artificial Intelligence, following this approach,\noperates similarly to humans, by taking into account the utility of the\nconsidered alternatives as well as their emotional attractiveness. Affective\nArtificial Intelligence, whose operation takes account of the cognition-emotion\nduality, avoids numerous behavioural paradoxes of traditional decision making.\nA society of intelligent agents, interacting through the repeated multistep\nexchange of information, forms a network accomplishing dynamic decision making.\nThe considered intelligent networks can characterize the operation of either a\nhuman society of affective decision makers, or the brain composed of neurons,\nor a typical probabilistic network of an artificial intelligence.\n", "date": "2023-05-14", "categories": []}
{"id": "2305.13500", "title": "Learning Emotion Representations from Verbal and Nonverbal Communication", "abstract": "  Emotion understanding is an essential but highly challenging component of\nartificial general intelligence. The absence of extensively annotated datasets\nhas significantly impeded advancements in this field. We present EmotionCLIP,\nthe first pre-training paradigm to extract visual emotion representations from\nverbal and nonverbal communication using only uncurated data. Compared to\nnumerical labels or descriptions used in previous methods, communication\nnaturally contains emotion information. Furthermore, acquiring emotion\nrepresentations from communication is more congruent with the human learning\nprocess. We guide EmotionCLIP to attend to nonverbal emotion cues through\nsubject-aware context encoding and verbal emotion cues using sentiment-guided\ncontrastive learning. Extensive experiments validate the effectiveness and\ntransferability of EmotionCLIP. Using merely linear-probe evaluation protocol,\nEmotionCLIP outperforms the state-of-the-art supervised visual emotion\nrecognition methods and rivals many multimodal approaches across various\nbenchmarks. We anticipate that the advent of EmotionCLIP will address the\nprevailing issue of data scarcity in emotion understanding, thereby fostering\nprogress in related domains. The code and pre-trained models are available at\nhttps://github.com/Xeaver/EmotionCLIP.\n", "date": "2023-05-22", "categories": []}
{"id": "2208.07589", "title": "Efficient Multimodal Transformer with Dual-Level Feature Restoration for\n  Robust Multimodal Sentiment Analysis", "abstract": "  With the proliferation of user-generated online videos, Multimodal Sentiment\nAnalysis (MSA) has attracted increasing attention recently. Despite significant\nprogress, there are still two major challenges on the way towards robust MSA:\n1) inefficiency when modeling cross-modal interactions in unaligned multimodal\ndata; and 2) vulnerability to random modality feature missing which typically\noccurs in realistic settings. In this paper, we propose a generic and unified\nframework to address them, named Efficient Multimodal Transformer with\nDual-Level Feature Restoration (EMT-DLFR). Concretely, EMT employs\nutterance-level representations from each modality as the global multimodal\ncontext to interact with local unimodal features and mutually promote each\nother. It not only avoids the quadratic scaling cost of previous local-local\ncross-modal interaction methods but also leads to better performance. To\nimprove model robustness in the incomplete modality setting, on the one hand,\nDLFR performs low-level feature reconstruction to implicitly encourage the\nmodel to learn semantic information from incomplete data. On the other hand, it\ninnovatively regards complete and incomplete data as two different views of one\nsample and utilizes siamese representation learning to explicitly attract their\nhigh-level representations. Comprehensive experiments on three popular datasets\ndemonstrate that our method achieves superior performance in both complete and\nincomplete modality settings.\n", "date": "2022-08-16", "categories": []}
{"id": "2305.12369", "title": "HIINT: Historical, Intra- and Inter- personal Dynamics Modeling with\n  Cross-person Memory Transformer", "abstract": "  Accurately modeling affect dynamics, which refers to the changes and\nfluctuations in emotions and affective displays during human conversations, is\ncrucial for understanding human interactions. By analyzing affect dynamics, we\ncan gain insights into how people communicate, respond to different situations,\nand form relationships. However, modeling affect dynamics is challenging due to\ncontextual factors, such as the complex and nuanced nature of interpersonal\nrelationships, the situation, and other factors that influence affective\ndisplays. To address this challenge, we propose a Cross-person Memory\nTransformer (CPM-T) framework which is able to explicitly model affective\ndynamics (intrapersonal and interpersonal influences) by identifying verbal and\nnon-verbal cues, and with a large language model to utilize the pre-trained\nknowledge and perform verbal reasoning. The CPM-T framework maintains memory\nmodules to store and update the contexts within the conversation window,\nenabling the model to capture dependencies between earlier and later parts of a\nconversation. Additionally, our framework employs cross-modal attention to\neffectively align information from multi-modalities and leverage cross-person\nattention to align behaviors in multi-party interactions. We evaluate the\neffectiveness and generalizability of our approach on three publicly available\ndatasets for joint engagement, rapport, and human beliefs prediction tasks.\nRemarkably, the CPM-T framework outperforms baseline models in average\nF1-scores by up to 7.3%, 9.3%, and 2.0% respectively. Finally, we demonstrate\nthe importance of each component in the framework via ablation studies with\nrespect to multimodal temporal behavior.\n", "date": "2023-05-21", "categories": []}
{"id": "2305.10446", "title": "Emotion Recognition based on Psychological Components in Guided\n  Narratives for Emotion Regulation", "abstract": "  Emotion regulation is a crucial element in dealing with emotional events and\nhas positive effects on mental health. This paper aims to provide a more\ncomprehensive understanding of emotional events by introducing a new French\ncorpus of emotional narratives collected using a questionnaire for emotion\nregulation. We follow the theoretical framework of the Component Process Model\nwhich considers emotions as dynamic processes composed of four interrelated\ncomponents (behavior, feeling, thinking and territory). Each narrative is\nrelated to a discrete emotion and is structured based on all emotion components\nby the writers. We study the interaction of components and their impact on\nemotion classification with machine learning methods and pre-trained language\nmodels. Our results show that each component improves prediction performance,\nand that the best results are achieved by jointly considering all components.\nOur results also show the effectiveness of pre-trained language models in\npredicting discrete emotion from certain components, which reveal differences\nin how emotion components are expressed.\n", "date": "2023-05-15", "categories": []}
{"id": "2305.10827", "title": "Expanding the Role of Affective Phenomena in Multimodal Interaction\n  Research", "abstract": "  In recent decades, the field of affective computing has made substantial\nprogress in advancing the ability of AI systems to recognize and express\naffective phenomena, such as affect and emotions, during human-human and\nhuman-machine interactions. This paper describes our examination of research at\nthe intersection of multimodal interaction and affective computing, with the\nobjective of observing trends and identifying understudied areas. We examined\nover 16,000 papers from selected conferences in multimodal interaction,\naffective computing, and natural language processing: ACM International\nConference on Multimodal Interaction, AAAC International Conference on\nAffective Computing and Intelligent Interaction, Annual Meeting of the\nAssociation for Computational Linguistics, and Conference on Empirical Methods\nin Natural Language Processing. We identified 910 affect-related papers and\npresent our analysis of the role of affective phenomena in these papers. We\nfind that this body of research has primarily focused on enabling machines to\nrecognize and express affect and emotion. However, we find limited research on\nhow affect and emotion predictions might be used by AI systems to enhance\nmachine understanding of human social behaviors and cognitive states. Based on\nour analysis, we discuss directions to expand the role of affective phenomena\nin multimodal interaction research.\n", "date": "2023-05-18", "categories": []}
{"id": "2305.10919", "title": "From the Lab to the Wild: Affect Modeling via Privileged Information", "abstract": "  How can we reliably transfer affect models trained in controlled laboratory\nconditions (in-vitro) to uncontrolled real-world settings (in-vivo)? The\ninformation gap between in-vitro and in-vivo applications defines a core\nchallenge of affective computing. This gap is caused by limitations related to\naffect sensing including intrusiveness, hardware malfunctions and availability\nof sensors. As a response to these limitations, we introduce the concept of\nprivileged information for operating affect models in real-world scenarios (in\nthe wild). Privileged information enables affect models to be trained across\nmultiple modalities available in a lab, and ignore, without significant\nperformance drops, those modalities that are not available when they operate in\nthe wild. Our approach is tested in two multimodal affect databases one of\nwhich is designed for testing models of affect in the wild. By training our\naffect models using all modalities and then using solely raw footage frames for\ntesting the models, we reach the performance of models that fuse all available\nmodalities for both training and testing. The results are robust across both\nclassification and regression affect modeling tasks which are dominant\nparadigms in affective computing. Our findings make a decisive step towards\nrealizing affect interaction in the wild.\n", "date": "2023-05-18", "categories": []}
{"id": "2304.00860", "title": "Designing and Evaluating Speech Emotion Recognition Systems: A reality\n  check case study with IEMOCAP", "abstract": "  There is an imminent need for guidelines and standard test sets to allow\ndirect and fair comparisons of speech emotion recognition (SER). While\nresources, such as the Interactive Emotional Dyadic Motion Capture (IEMOCAP)\ndatabase, have emerged as widely-adopted reference corpora for researchers to\ndevelop and test models for SER, published work reveals a wide range of\nassumptions and variety in its use that challenge reproducibility and\ngeneralization. Based on a critical review of the latest advances in SER using\nIEMOCAP as the use case, our work aims at two contributions: First, using an\nanalysis of the recent literature, including assumptions made and metrics used\ntherein, we provide a set of SER evaluation guidelines. Second, using recent\npublications with open-sourced implementations, we focus on reproducibility\nassessment in SER.\n", "date": "2023-04-03", "categories": []}
{"id": "2305.07665", "title": "A Comprehensive Survey on Affective Computing; Challenges, Trends,\n  Applications, and Future Directions", "abstract": "  As the name suggests, affective computing aims to recognize human emotions,\nsentiments, and feelings. There is a wide range of fields that study affective\ncomputing, including languages, sociology, psychology, computer science, and\nphysiology. However, no research has ever been done to determine how machine\nlearning (ML) and mixed reality (XR) interact together. This paper discusses\nthe significance of affective computing, as well as its ideas, conceptions,\nmethods, and outcomes. By using approaches of ML and XR, we survey and discuss\nrecent methodologies in affective computing. We survey the state-of-the-art\napproaches along with current affective data resources. Further, we discuss\nvarious applications where affective computing has a significant impact, which\nwill aid future scholars in gaining a better understanding of its significance\nand practical relevance.\n", "date": "2023-05-08", "categories": []}
{"id": "2305.06162", "title": "Interpretable multimodal sentiment analysis based on textual modality\n  descriptions by using large-scale language models", "abstract": "  Multimodal sentiment analysis is an important area for understanding the\nuser's internal states. Deep learning methods were effective, but the problem\nof poor interpretability has gradually gained attention. Previous works have\nattempted to use attention weights or vector distributions to provide\ninterpretability. However, their explanations were not intuitive and can be\ninfluenced by different trained models. This study proposed a novel approach to\nprovide interpretability by converting nonverbal modalities into text\ndescriptions and by using large-scale language models for sentiment\npredictions. This provides an intuitive approach to directly interpret what\nmodels depend on with respect to making decisions from input texts, thus\nsignificantly improving interpretability. Specifically, we convert descriptions\nbased on two feature patterns for the audio modality and discrete action units\nfor the facial modality. Experimental results on two sentiment analysis tasks\ndemonstrated that the proposed approach maintained, or even improved\neffectiveness for sentiment analysis compared to baselines using conventional\nfeatures, with the highest improvement of 2.49% on the F1 score. The results\nalso showed that multimodal descriptions have similar characteristics on fusing\nmodalities as those of conventional fusion methods. The results demonstrated\nthat the proposed approach is interpretable and effective for multimodal\nsentiment analysis.\n", "date": "2023-05-07", "categories": []}
{"id": "2305.07446", "title": "A Lightweight Domain Adversarial Neural Network Based on Knowledge\n  Distillation for EEG-based Cross-subject Emotion Recognition", "abstract": "  Individual differences of Electroencephalogram (EEG) could cause the domain\nshift which would significantly degrade the performance of cross-subject\nstrategy. The domain adversarial neural networks (DANN), where the\nclassification loss and domain loss jointly update the parameters of feature\nextractor, are adopted to deal with the domain shift. However, limited EEG data\nquantity and strong individual difference are challenges for the DANN with\ncumbersome feature extractor. In this work, we propose knowledge distillation\n(KD) based lightweight DANN to enhance cross-subject EEG-based emotion\nrecognition. Specifically, the teacher model with strong context learning\nability is utilized to learn complex temporal dynamics and spatial correlations\nof EEG, and robust lightweight student model is guided by the teacher model to\nlearn more difficult domain-invariant features. In the feature-based KD\nframework, a transformer-based hierarchical temporalspatial learning model is\nserved as the teacher model. The student model, which is composed of Bi-LSTM\nunits, is a lightweight version of the teacher model. Hence, the student model\ncould be supervised to mimic the robust feature representations of teacher\nmodel by leveraging complementary latent temporal features and spatial\nfeatures. In the DANN-based cross-subject emotion recognition, we combine the\nobtained student model and a lightweight temporal-spatial feature interaction\nmodule as the feature extractor. And the feature aggregation is fed to the\nemotion classifier and domain classifier for domain-invariant feature learning.\nTo verify the effectiveness of the proposed method, we conduct the\nsubject-independent experiments on the public dataset DEAP with arousal and\nvalence classification. The outstanding performance and t-SNE visualization of\nlatent features verify the advantage and effectiveness of the proposed method.\n", "date": "2023-05-12", "categories": []}
{"id": "2305.06842", "title": "Emotion Recognition for Challenged People Facial Appearance in Social\n  using Neural Network", "abstract": "  Human communication is the vocal and non verbal signal to communicate with\nothers. Human expression is a significant biometric object in picture and\nrecord databases of surveillance systems. Face appreciation has a serious role\nin biometric methods and is good-looking for plentiful applications, including\nvisual scrutiny and security. Facial expressions are a form of nonverbal\ncommunication; recognizing them helps improve the human machine interaction.\nThis paper proposes an idea for face and enlightenment invariant credit of\nfacial expressions by the images. In order on, the person's face can be\ncomputed. Face expression is used in CNN classifier to categorize the acquired\npicture into different emotion categories. It is a deep, feed-forward\nartificial neural network. Outcome surpasses human presentation and shows poses\nalternate performance. Varying lighting conditions can influence the fitting\nprocess and reduce recognition precision. Results illustrate that dependable\nfacial appearance credited with changing lighting conditions for separating\nreasonable facial terminology display emotions is an efficient representation\nof clean and assorted moving expressions. This process can also manage the\nproportions of dissimilar basic affecting expressions of those mixed jointly to\nproduce sensible emotional facial expressions. Our system contains a\npre-defined data set, which was residential by a statistics scientist and\nincludes all pure and varied expressions. On average, a data set has achieved\n92.4% exact validation of the expressions synthesized by our technique. These\nfacial expressions are compared through the pre-defined data-position inside\nour system. If it recognizes the person in an abnormal condition, an alert will\nbe passed to the nearby hospital/doctor seeing that a message.\n", "date": "2023-05-11", "categories": []}
{"id": "2305.06273", "title": "Learning Robust Self-attention Features for Speech Emotion Recognition\n  with Label-adaptive Mixup", "abstract": "  Speech Emotion Recognition (SER) is to recognize human emotions in a natural\nverbal interaction scenario with machines, which is considered as a challenging\nproblem due to the ambiguous human emotions. Despite the recent progress in\nSER, state-of-the-art models struggle to achieve a satisfactory performance. We\npropose a self-attention based method with combined use of label-adaptive mixup\nand center loss. By adapting label probabilities in mixup and fitting center\nloss to the mixup training scheme, our proposed method achieves a superior\nperformance to the state-of-the-art methods.\n", "date": "2023-05-07", "categories": []}
{"id": "2305.05523", "title": "RMES: Real-Time Micro-Expression Spotting Using Phase From Riesz Pyramid", "abstract": "  Micro-expressions (MEs) are involuntary and subtle facial expressions that\nare thought to reveal feelings people are trying to hide. ME spotting detects\nthe temporal intervals containing MEs in videos. Detecting such quick and\nsubtle motions from long videos is difficult. Recent works leverage detailed\nfacial motion representations, such as the optical flow, and deep learning\nmodels, leading to high computational complexity. To reduce computational\ncomplexity and achieve real-time operation, we propose RMES, a real-time ME\nspotting framework. We represent motion using phase computed by Riesz Pyramid,\nand feed this motion representation into a three-stream shallow CNN, which\npredicts the likelihood of each frame belonging to an ME. In comparison to\noptical flow, phase provides more localized motion estimates, which are\nessential for ME spotting, resulting in higher performance. Using phase also\nreduces the required computation of the ME spotting pipeline by 77.8%. Despite\nits relative simplicity and low computational complexity, our framework\nachieves state-of-the-art performance on two public datasets: CAS(ME)2 and SAMM\nLong Videos.\n", "date": "2023-05-09", "categories": []}
{"id": "2305.05534", "title": "Integrating Holistic and Local Information to Estimate Emotional\n  Reaction Intensity", "abstract": "  Video-based Emotional Reaction Intensity (ERI) estimation measures the\nintensity of subjects' reactions to stimuli along several emotional dimensions\nfrom videos of the subject as they view the stimuli. We propose a multi-modal\narchitecture for video-based ERI combining video and audio information. Video\ninput is encoded spatially first, frame-by-frame, combining features encoding\nholistic aspects of the subjects' facial expressions and features encoding\nspatially localized aspects of their expressions. Input is then combined across\ntime: from frame-to-frame using gated recurrent units (GRUs), then globally by\na transformer. We handle variable video length with a regression token that\naccumulates information from all frames into a fixed-dimensional vector\nindependent of video length. Audio information is handled similarly: spectral\ninformation extracted within each frame is integrated across time by a cascade\nof GRUs and a transformer with regression token. The video and audio regression\ntokens' outputs are merged by concatenation, then input to a final fully\nconnected layer producing intensity estimates. Our architecture achieved\nexcellent performance on the Hume-Reaction dataset in the ERI Esimation\nChallenge of the Fifth Competition on Affective Behavior Analysis in-the-Wild\n(ABAW5). The Pearson Correlation Coefficients between estimated and subject\nself-reported scores, averaged across all emotions, were 0.455 on the\nvalidation dataset and 0.4547 on the test dataset, well above the baselines.\nThe transformer's self-attention mechanism enables our architecture to focus on\nthe most critical video frames regardless of length. Ablation experiments\nestablish the advantages of combining holistic/local features and of\nmulti-modal integration. Code available at https://github.com/HKUST-NISL/ABAW5.\n", "date": "2023-05-09", "categories": []}
{"id": "2305.04796", "title": "The Application of Affective Measures in Text-based Emotion Aware\n  Recommender Systems", "abstract": "  This paper presents an innovative approach to address the problems\nresearchers face in Emotion Aware Recommender Systems (EARS): the difficulty\nand cumbersome collecting voluminously good quality emotion-tagged datasets and\nan effective way to protect users' emotional data privacy. Without enough\ngood-quality emotion-tagged datasets, researchers cannot conduct repeatable\naffective computing research in EARS that generates personalized\nrecommendations based on users' emotional preferences. Similarly, if we fail to\nfully protect users' emotional data privacy, users could resist engaging with\nEARS services. This paper introduced a method that detects affective features\nin subjective passages using the Generative Pre-trained Transformer Technology,\nforming the basis of the Affective Index and Affective Index Indicator (AII).\nEliminate the need for users to build an affective feature detection mechanism.\nThe paper advocates for a separation of responsibility approach where users\nprotect their emotional profile data while EARS service providers refrain from\nretaining or storing it. Service providers can update users' Affective Indices\nin memory without saving their privacy data, providing Affective Aware\nrecommendations without compromising user privacy. This paper offers a solution\nto the subjectivity and variability of emotions, data privacy concerns, and\nevaluation metrics and benchmarks, paving the way for future EARS research.\n", "date": "2023-05-04", "categories": []}
{"id": "2305.03369", "title": "The MuSe 2023 Multimodal Sentiment Analysis Challenge: Mimicked\n  Emotions, Cross-Cultural Humour, and Personalisation", "abstract": "  The MuSe 2023 is a set of shared tasks addressing three different\ncontemporary multimodal affect and sentiment analysis problems: In the Mimicked\nEmotions Sub-Challenge (MuSe-Mimic), participants predict three continuous\nemotion targets. This sub-challenge utilises the Hume-Vidmimic dataset\ncomprising of user-generated videos. For the Cross-Cultural Humour Detection\nSub-Challenge (MuSe-Humour), an extension of the Passau Spontaneous Football\nCoach Humour (Passau-SFCH) dataset is provided. Participants predict the\npresence of spontaneous humour in a cross-cultural setting. The Personalisation\nSub-Challenge (MuSe-Personalisation) is based on the Ulm-Trier Social Stress\nTest (Ulm-TSST) dataset, featuring recordings of subjects in a stressed\nsituation. Here, arousal and valence signals are to be predicted, whereas parts\nof the test labels are made available in order to facilitate personalisation.\nMuSe 2023 seeks to bring together a broad audience from different research\ncommunities such as audio-visual emotion recognition, natural language\nprocessing, signal processing, and health informatics. In this baseline paper,\nwe introduce the datasets, sub-challenges, and provided feature sets. As a\ncompetitive baseline system, a Gated Recurrent Unit (GRU)-Recurrent Neural\nNetwork (RNN) is employed. On the respective sub-challenges' test datasets, it\nachieves a mean (across three continuous intensity targets) Pearson's\nCorrelation Coefficient of .4727 for MuSe-Mimic, an Area Under the Curve (AUC)\nvalue of .8310 for MuSe-Humor and Concordance Correlation Coefficient (CCC)\nvalues of .7482 for arousal and .7827 for valence in the MuSe-Personalisation\nsub-challenge.\n", "date": "2023-05-05", "categories": []}
{"id": "2305.03500", "title": "High-Level Context Representation for Emotion Recognition in Images", "abstract": "  Emotion recognition is the task of classifying perceived emotions in people.\nPrevious works have utilized various nonverbal cues to extract features from\nimages and correlate them to emotions. Of these cues, situational context is\nparticularly crucial in emotion perception since it can directly influence the\nemotion of a person. In this paper, we propose an approach for high-level\ncontext representation extraction from images. The model relies on a single cue\nand a single encoding stream to correlate this representation with emotions.\nOur model competes with the state-of-the-art, achieving an mAP of 0.3002 on the\nEMOTIC dataset while also being capable of execution on consumer-grade hardware\nat approximately 90 frames per second. Overall, our approach is more efficient\nthan previous models and can be easily deployed to address real-world problems\nrelated to emotion recognition.\n", "date": "2023-05-05", "categories": []}
{"id": "2305.02629", "title": "Integrating Psychometrics and Computing Perspectives on Bias and\n  Fairness in Affective Computing: A Case Study of Automated Video Interviews", "abstract": "  We provide a psychometric-grounded exposition of bias and fairness as applied\nto a typical machine learning pipeline for affective computing. We expand on an\ninterpersonal communication framework to elucidate how to identify sources of\nbias that may arise in the process of inferring human emotions and other\npsychological constructs from observed behavior. Various methods and metrics\nfor measuring fairness and bias are discussed along with pertinent implications\nwithin the United States legal context. We illustrate how to measure some types\nof bias and fairness in a case study involving automatic personality and\nhireability inference from multimodal data collected in video interviews for\nmock job applications. We encourage affective computing researchers and\npractitioners to encapsulate bias and fairness in their research processes and\nproducts and to consider their role, agency, and responsibility in promoting\nequitable and just systems.\n", "date": "2023-05-04", "categories": []}
{"id": "2305.02814", "title": "Noise-Resistant Multimodal Transformer for Emotion Recognition", "abstract": "  Multimodal emotion recognition identifies human emotions from various data\nmodalities like video, text, and audio. However, we found that this task can be\neasily affected by noisy information that does not contain useful semantics. To\nthis end, we present a novel paradigm that attempts to extract noise-resistant\nfeatures in its pipeline and introduces a noise-aware learning scheme to\neffectively improve the robustness of multimodal emotion understanding. Our new\npipeline, namely Noise-Resistant Multimodal Transformer (NORM-TR), mainly\nintroduces a Noise-Resistant Generic Feature (NRGF) extractor and a Transformer\nfor the multimodal emotion recognition task. In particular, we make the NRGF\nextractor learn a generic and disturbance-insensitive representation so that\nconsistent and meaningful semantics can be obtained. Furthermore, we apply a\nTransformer to incorporate Multimodal Features (MFs) of multimodal inputs based\non their relations to the NRGF. Therefore, the possible insensitive but useful\ninformation of NRGF could be complemented by MFs that contain more details. To\ntrain the NORM-TR properly, our proposed noise-aware learning scheme\ncomplements normal emotion recognition losses by enhancing the learning against\nnoises. Our learning scheme explicitly adds noises to either all the modalities\nor a specific modality at random locations of a multimodal input sequence. We\ncorrespondingly introduce two adversarial losses to encourage the NRGF\nextractor to learn to extract the NRGFs invariant to the added noises, thus\nfacilitating the NORM-TR to achieve more favorable multimodal emotion\nrecognition performance. In practice, on several popular multimodal datasets,\nour NORM-TR achieves state-of-the-art performance and outperforms existing\nmethods by a large margin, which demonstrates that the ability to resist noisy\ninformation is important for effective emotion recognition.\n", "date": "2023-05-04", "categories": []}
{"id": "2305.02842", "title": "MEDIC: A Multimodal Empathy Dataset in Counseling", "abstract": "  Although empathic interaction between counselor and client is fundamental to\nsuccess in the psychotherapeutic process, there are currently few datasets to\naid a computational approach to empathy understanding. In this paper, we\nconstruct a multimodal empathy dataset collected from face-to-face\npsychological counseling sessions. The dataset consists of 771 video clips. We\nalso propose three labels (i.e., expression of experience, emotional reaction,\nand cognitive reaction) to describe the degree of empathy between counselors\nand their clients. Expression of experience describes whether the client has\nexpressed experiences that can trigger empathy, and emotional and cognitive\nreactions indicate the counselor's empathic reactions. As an elementary\nassessment of the usability of the constructed multimodal empathy dataset, an\ninterrater reliability analysis of annotators' subjective evaluations for video\nclips is conducted using the intraclass correlation coefficient and Fleiss'\nKappa. Results prove that our data annotation is reliable. Furthermore, we\nconduct empathy prediction using three typical methods, including the tensor\nfusion network, the sentimental words aware fusion network, and a simple\nconcatenation model. The experimental results show that empathy can be well\npredicted on our dataset. Our dataset is available for research purposes.\n", "date": "2023-05-04", "categories": []}
{"id": "2305.01884", "title": "Class adaptive threshold and negative class guided noisy annotation\n  robust Facial Expression Recognition", "abstract": "  The hindering problem in facial expression recognition (FER) is the presence\nof inaccurate annotations referred to as noisy annotations in the datasets.\nThese noisy annotations are present in the datasets inherently because the\nlabeling is subjective to the annotator, clarity of the image, etc. Recent\nworks use sample selection methods to solve this noisy annotation problem in\nFER. In our work, we use a dynamic adaptive threshold to separate confident\nsamples from non-confident ones so that our learning won't be hampered due to\nnon-confident samples. Instead of discarding the non-confident samples, we\nimpose consistency in the negative classes of those non-confident samples to\nguide the model to learn better in the positive class. Since FER datasets\nusually come with 7 or 8 classes, we can correctly guess a negative class by\n85% probability even by choosing randomly. By learning \"which class a sample\ndoesn't belong to\", the model can learn \"which class it belongs to\" in a better\nmanner. We demonstrate proposed framework's effectiveness using quantitative as\nwell as qualitative results. Our method performs better than the baseline by a\nmargin of 4% to 28% on RAFDB and 3.3% to 31.4% on FERPlus for various levels of\nsynthetic noisy labels in the aforementioned datasets.\n", "date": "2023-05-03", "categories": []}
{"id": "2304.14882", "title": "The ACM Multimedia 2023 Computational Paralinguistics Challenge: Emotion\n  Share & Requests", "abstract": "  The ACM Multimedia 2023 Computational Paralinguistics Challenge addresses two\ndifferent problems for the first time in a research competition under\nwell-defined conditions: In the Emotion Share Sub-Challenge, a regression on\nspeech has to be made; and in the Requests Sub-Challenges, requests and\ncomplaints need to be detected. We describe the Sub-Challenges, baseline\nfeature extraction, and classifiers based on the usual ComPaRE features, the\nauDeep toolkit, and deep feature extraction from pre-trained CNNs using the\nDeepSpectRum toolkit; in addition, wav2vec2 models are used.\n", "date": "2023-04-28", "categories": []}
{"id": "2304.14714", "title": "SGED: A Benchmark dataset for Performance Evaluation of Spiking Gesture\n  Emotion Recognition", "abstract": "  In the field of affective computing, researchers in the community have\npromoted the performance of models and algorithms by using the complementarity\nof multimodal information. However, the emergence of more and more modal\ninformation makes the development of datasets unable to keep up with the\nprogress of existing modal sensing equipment. Collecting and studying\nmultimodal data is a complex and significant work. In order to supplement the\nchallenge of partial missing of community data. We collected and labeled a new\nhomogeneous multimodal gesture emotion recognition dataset based on the\nanalysis of the existing data sets. This data set complements the defects of\nhomogeneous multimodal data and provides a new research direction for emotion\nrecognition. Moreover, we propose a pseudo dual-flow network based on this\ndataset, and verify the application potential of this dataset in the affective\ncomputing community. The experimental results demonstrate that it is feasible\nto use the traditional visual information and spiking visual information based\non homogeneous multimodal data for visual emotion recognition.The dataset is\navailable at \\url{https://github.com/201528014227051/SGED}\n", "date": "2023-04-28", "categories": []}
{"id": "2304.14942", "title": "The Emotions of the Crowd: Learning Image Sentiment from Tweets via\n  Cross-modal Distillation", "abstract": "  Trends and opinion mining in social media increasingly focus on novel\ninteractions involving visual media, like images and short videos, in addition\nto text. In this work, we tackle the problem of visual sentiment analysis of\nsocial media images -- specifically, the prediction of image sentiment\npolarity. While previous work relied on manually labeled training sets, we\npropose an automated approach for building sentiment polarity classifiers based\non a cross-modal distillation paradigm; starting from scraped multimodal (text\n+ images) data, we train a student model on the visual modality based on the\noutputs of a textual teacher model that analyses the sentiment of the\ncorresponding textual modality. We applied our method to randomly collected\nimages crawled from Twitter over three months and produced, after automatic\ncleaning, a weakly-labeled dataset of $\\sim$1.5 million images. Despite\nexploiting noisy labeled samples, our training pipeline produces classifiers\nshowing strong generalization capabilities and outperforming the current state\nof the art on five manually labeled benchmarks for image sentiment polarity\nprediction.\n", "date": "2023-04-28", "categories": []}
{"id": "2304.13324", "title": "A Portrait of Emotion: Empowering Self-Expression through AI-Generated\n  Art", "abstract": "  We investigated the potential and limitations of generative artificial\nintelligence (AI) in reflecting the authors' cognitive processes through\ncreative expression. The focus is on the AI-generated artwork's ability to\nunderstand human intent (alignment) and visually represent emotions based on\ncriteria such as creativity, aesthetic, novelty, amusement, and depth. Results\nshow a preference for images based on the descriptions of the authors' emotions\nover the main events. We also found that images that overrepresent specific\nelements or stereotypes negatively impact AI alignment. Our findings suggest\nthat AI could facilitate creativity and the self-expression of emotions. Our\nresearch framework with generative AIs can help design AI-based interventions\nin related fields (e.g., mental health education, therapy, and counseling).\n", "date": "2023-04-26", "categories": []}
{"id": "2304.10973", "title": "LEIA: Linguistic Embeddings for the Identification of Affect", "abstract": "  The wealth of text data generated by social media has enabled new kinds of\nanalysis of emotions with language models. These models are often trained on\nsmall and costly datasets of text annotations produced by readers who guess the\nemotions expressed by others in social media posts. This affects the quality of\nemotion identification methods due to training data size limitations and noise\nin the production of labels used in model development. We present LEIA, a model\nfor emotion identification in text that has been trained on a dataset of more\nthan 6 million posts with self-annotated emotion labels for happiness,\naffection, sadness, anger, and fear. LEIA is based on a word masking method\nthat enhances the learning of emotion words during model pre-training. LEIA\nachieves macro-F1 values of approximately 73 on three in-domain test datasets,\noutperforming other supervised and unsupervised methods in a strong benchmark\nthat shows that LEIA generalizes across posts, users, and time periods. We\nfurther perform an out-of-domain evaluation on five different datasets of\nsocial media and other sources, showing LEIA's robust performance across media,\ndata collection methods, and annotation schemes. Our results show that LEIA\ngeneralizes its classification of anger, happiness, and sadness beyond the\ndomain it was trained on. LEIA can be applied in future research to provide\nbetter identification of emotions in text from the perspective of the writer.\nThe models produced for this article are publicly available at\nhttps://huggingface.co/LEIA\n", "date": "2023-04-21", "categories": []}
{"id": "2304.11040", "title": "Emotional Expression Detection in Spoken Language Employing Machine\n  Learning Algorithms", "abstract": "  There are a variety of features of the human voice that can be classified as\npitch, timbre, loudness, and vocal tone. It is observed in numerous incidents\nthat human expresses their feelings using different vocal qualities when they\nare speaking. The primary objective of this research is to recognize different\nemotions of human beings such as anger, sadness, fear, neutrality, disgust,\npleasant surprise, and happiness by using several MATLAB functions namely,\nspectral descriptors, periodicity, and harmonicity. To accomplish the work, we\nanalyze the CREMA-D (Crowd-sourced Emotional Multimodal Actors Data) & TESS\n(Toronto Emotional Speech Set) datasets of human speech. The audio file\ncontains data that have various characteristics (e.g., noisy, speedy, slow)\nthereby the efficiency of the ML (Machine Learning) models increases\nsignificantly. The EMD (Empirical Mode Decomposition) is utilized for the\nprocess of signal decomposition. Then, the features are extracted through the\nuse of several techniques such as the MFCC, GTCC, spectral centroid, roll-off\npoint, entropy, spread, flux, harmonic ratio, energy, skewness, flatness, and\naudio delta. The data is trained using some renowned ML models namely, Support\nVector Machine, Neural Network, Ensemble, and KNN. The algorithms show an\naccuracy of 67.7%, 63.3%, 61.6%, and 59.0% respectively for the test data and\n77.7%, 76.1%, 99.1%, and 61.2% for the training data. We have conducted\nexperiments using Matlab and the result shows that our model is very prominent\nand flexible than existing similar works.\n", "date": "2023-04-20", "categories": []}
{"id": "2304.11046", "title": "Affective social anthropomorphic intelligent system", "abstract": "  Human conversational styles are measured by the sense of humor, personality,\nand tone of voice. These characteristics have become essential for\nconversational intelligent virtual assistants. However, most of the\nstate-of-the-art intelligent virtual assistants (IVAs) are failed to interpret\nthe affective semantics of human voices. This research proposes an\nanthropomorphic intelligent system that can hold a proper human-like\nconversation with emotion and personality. A voice style transfer method is\nalso proposed to map the attributes of a specific emotion. Initially, the\nfrequency domain data (Mel-Spectrogram) is created by converting the temporal\naudio wave data, which comprises discrete patterns for audio features such as\nnotes, pitch, rhythm, and melody. A collateral CNN-Transformer-Encoder is used\nto predict seven different affective states from voice. The voice is also fed\nparallelly to the deep-speech, an RNN model that generates the text\ntranscription from the spectrogram. Then the transcripted text is transferred\nto the multi-domain conversation agent using blended skill talk,\ntransformer-based retrieve-and-generate generation strategy, and beam-search\ndecoding, and an appropriate textual response is generated. The system learns\nan invertible mapping of data to a latent space that can be manipulated and\ngenerates a Mel-spectrogram frame based on previous Mel-spectrogram frames to\nvoice synthesize and style transfer. Finally, the waveform is generated using\nWaveGlow from the spectrogram. The outcomes of the studies we conducted on\nindividual models were auspicious. Furthermore, users who interacted with the\nsystem provided positive feedback, demonstrating the system's effectiveness.\n", "date": "2023-04-19", "categories": []}
{"id": "2304.11117", "title": "A vector quantized masked autoencoder for speech emotion recognition", "abstract": "  Recent years have seen remarkable progress in speech emotion recognition\n(SER), thanks to advances in deep learning techniques. However, the limited\navailability of labeled data remains a significant challenge in the field.\nSelf-supervised learning has recently emerged as a promising solution to\naddress this challenge. In this paper, we propose the vector quantized masked\nautoencoder for speech (VQ-MAE-S), a self-supervised model that is fine-tuned\nto recognize emotions from speech signals. The VQ-MAE-S model is based on a\nmasked autoencoder (MAE) that operates in the discrete latent space of a\nvector-quantized variational autoencoder. Experimental results show that the\nproposed VQ-MAE-S model, pre-trained on the VoxCeleb2 dataset and fine-tuned on\nemotional speech data, outperforms an MAE working on the raw spectrogram\nrepresentation and other state-of-the-art methods in SER.\n", "date": "2023-04-21", "categories": []}
{"id": "2304.09928", "title": "Personalized State Anxiety Detection: An Empirical Study with Linguistic\n  Biomarkers and A Machine Learning Pipeline", "abstract": "  Individuals high in social anxiety symptoms often exhibit elevated state\nanxiety in social situations. Research has shown it is possible to detect state\nanxiety by leveraging digital biomarkers and machine learning techniques.\nHowever, most existing work trains models on an entire group of participants,\nfailing to capture individual differences in their psychological and behavioral\nresponses to social contexts. To address this concern, in Study 1, we collected\nlinguistic data from N=35 high socially anxious participants in a variety of\nsocial contexts, finding that digital linguistic biomarkers significantly\ndiffer between evaluative vs. non-evaluative social contexts and between\nindividuals having different trait psychological symptoms, suggesting the\nlikely importance of personalized approaches to detect state anxiety. In Study\n2, we used the same data and results from Study 1 to model a multilayer\npersonalized machine learning pipeline to detect state anxiety that considers\ncontextual and individual differences. This personalized model outperformed the\nbaseline F1-score by 28.0%. Results suggest that state anxiety can be more\naccurately detected with personalized machine learning approaches, and that\nlinguistic biomarkers hold promise for identifying periods of state anxiety in\nan unobtrusive way.\n", "date": "2023-04-19", "categories": []}
{"id": "2304.10211", "title": "Spiking-Fer: Spiking Neural Network for Facial Expression Recognition\n  With Event Cameras", "abstract": "  Facial Expression Recognition (FER) is an active research domain that has\nshown great progress recently, notably thanks to the use of large deep learning\nmodels. However, such approaches are particularly energy intensive, which makes\ntheir deployment difficult for edge devices. To address this issue, Spiking\nNeural Networks (SNNs) coupled with event cameras are a promising alternative,\ncapable of processing sparse and asynchronous events with lower energy\nconsumption. In this paper, we establish the first use of event cameras for\nFER, named \"Event-based FER\", and propose the first related benchmarks by\nconverting popular video FER datasets to event streams. To deal with this new\ntask, we propose \"Spiking-FER\", a deep convolutional SNN model, and compare it\nagainst a similar Artificial Neural Network (ANN). Experiments show that the\nproposed approach achieves comparable performance to the ANN architecture,\nwhile consuming less energy by orders of magnitude (up to 65.39x). In addition,\nan experimental study of various event-based data augmentation techniques is\nperformed to provide insights into the efficient transformations specific to\nevent-based FER.\n", "date": "2023-04-20", "categories": []}
{"id": "2304.09582", "title": "Is ChatGPT Equipped with Emotional Dialogue Capabilities?", "abstract": "  This report presents a study on the emotional dialogue capability of ChatGPT,\nan advanced language model developed by OpenAI. The study evaluates the\nperformance of ChatGPT on emotional dialogue understanding and generation\nthrough a series of experiments on several downstream tasks. Our findings\nindicate that while ChatGPT's performance on emotional dialogue understanding\nmay still lag behind that of supervised models, it exhibits promising results\nin generating emotional responses. Furthermore, the study suggests potential\navenues for future research directions.\n", "date": "2023-04-19", "categories": []}
{"id": "2211.11425", "title": "Data Leakage and Evaluation Issues in Micro-Expression Analysis", "abstract": "  Micro-expressions have drawn increasing interest lately due to various\npotential applications. The task is, however, difficult as it incorporates many\nchallenges from the fields of computer vision, machine learning and emotional\nsciences. Due to the spontaneous and subtle characteristics of\nmicro-expressions, the available training and testing data are limited, which\nmake evaluation complex. We show that data leakage and fragmented evaluation\nprotocols are issues among the micro-expression literature. We find that fixing\ndata leaks can drastically reduce model performance, in some cases even making\nthe models perform similarly to a random classifier. To this end, we go through\ncommon pitfalls, propose a new standardized evaluation protocol using facial\naction units with over 2000 micro-expression samples, and provide an open\nsource library that implements the evaluation protocols in a standardized\nmanner. Code is publicly available in \\url{https://github.com/tvaranka/meb}.\n", "date": "2022-11-21", "categories": []}
{"id": "2304.07958", "title": "Recursive Joint Attention for Audio-Visual Fusion in Regression based\n  Emotion Recognition", "abstract": "  In video-based emotion recognition (ER), it is important to effectively\nleverage the complementary relationship among audio (A) and visual (V)\nmodalities, while retaining the intra-modal characteristics of individual\nmodalities. In this paper, a recursive joint attention model is proposed along\nwith long short-term memory (LSTM) modules for the fusion of vocal and facial\nexpressions in regression-based ER. Specifically, we investigated the\npossibility of exploiting the complementary nature of A and V modalities using\na joint cross-attention model in a recursive fashion with LSTMs to capture the\nintra-modal temporal dependencies within the same modalities as well as among\nthe A-V feature representations. By integrating LSTMs with recursive joint\ncross-attention, our model can efficiently leverage both intra- and inter-modal\nrelationships for the fusion of A and V modalities. The results of extensive\nexperiments performed on the challenging Affwild2 and Fatigue (private)\ndatasets indicate that the proposed A-V fusion model can significantly\noutperform state-of-art-methods.\n", "date": "2023-04-17", "categories": []}
{"id": "2303.09164", "title": "Multimodal Feature Extraction and Fusion for Emotional Reaction\n  Intensity Estimation and Expression Classification in Videos with\n  Transformers", "abstract": "  In this paper, we present our advanced solutions to the two sub-challenges of\nAffective Behavior Analysis in the wild (ABAW) 2023: the Emotional Reaction\nIntensity (ERI) Estimation Challenge and Expression (Expr) Classification\nChallenge. ABAW 2023 aims to tackle the challenge of affective behavior\nanalysis in natural contexts, with the ultimate goal of creating intelligent\nmachines and robots that possess the ability to comprehend human emotions,\nfeelings, and behaviors. For the Expression Classification Challenge, we\npropose a streamlined approach that handles the challenges of classification\neffectively. However, our main contribution lies in our use of diverse models\nand tools to extract multimodal features such as audio and video cues from the\nHume-Reaction dataset. By studying, analyzing, and combining these features, we\nsignificantly enhance the model's accuracy for sentiment prediction in a\nmultimodal context. Furthermore, our method achieves outstanding results on the\nEmotional Reaction Intensity (ERI) Estimation Challenge, surpassing the\nbaseline method by an impressive 84\\% increase, as measured by the Pearson\nCoefficient, on the validation dataset.\n", "date": "2023-03-16", "categories": []}
{"id": "2304.05634", "title": "How you feelin'? Learning Emotions and Mental States in Movie Scenes", "abstract": "  Movie story analysis requires understanding characters' emotions and mental\nstates. Towards this goal, we formulate emotion understanding as predicting a\ndiverse and multi-label set of emotions at the level of a movie scene and for\neach character. We propose EmoTx, a multimodal Transformer-based architecture\nthat ingests videos, multiple characters, and dialog utterances to make joint\npredictions. By leveraging annotations from the MovieGraphs dataset, we aim to\npredict classic emotions (e.g. happy, angry) and other mental states (e.g.\nhonest, helpful). We conduct experiments on the most frequently occurring 10\nand 25 labels, and a mapping that clusters 181 labels to 26. Ablation studies\nand comparison against adapted state-of-the-art emotion recognition approaches\nshows the effectiveness of EmoTx. Analyzing EmoTx's self-attention scores\nreveals that expressive emotions often look at character tokens while other\nmental states rely on video and dialog cues.\n", "date": "2023-04-12", "categories": []}
{"id": "2303.10849", "title": "Multi-modal Facial Affective Analysis based on Masked Autoencoder", "abstract": "  Human affective behavior analysis focuses on analyzing human expressions or\nother behaviors to enhance the understanding of human psychology. The CVPR 2023\nCompetition on Affective Behavior Analysis in-the-wild (ABAW) is dedicated to\nproviding high-quality and large-scale Aff-wild2 for the recognition of\ncommonly used emotion representations, such as Action Units (AU), basic\nexpression categories(EXPR), and Valence-Arousal (VA). The competition is\ncommitted to making significant strides in improving the accuracy and\npracticality of affective analysis research in real-world scenarios. In this\npaper, we introduce our submission to the CVPR 2023: ABAW5. Our approach\ninvolves several key components. First, we utilize the visual information from\na Masked Autoencoder(MAE) model that has been pre-trained on a large-scale face\nimage dataset in a self-supervised manner. Next, we finetune the MAE encoder on\nthe image frames from the Aff-wild2 for AU, EXPR and VA tasks, which can be\nregarded as a static and uni-modal training. Additionally, we leverage the\nmulti-modal and temporal information from the videos and implement a\ntransformer-based framework to fuse the multi-modal features. Our approach\nachieves impressive results in the ABAW5 competition, with an average F1 score\nof 55.49\\% and 41.21\\% in the AU and EXPR tracks, respectively, and an average\nCCC of 0.6372 in the VA track. Our approach ranks first in the EXPR and AU\ntracks, and second in the VA track. Extensive quantitative experiments and\nablation studies demonstrate the effectiveness of our proposed method.\n", "date": "2023-03-20", "categories": []}
{"id": "2304.04915", "title": "AffectMachine-Classical: A novel system for generating affective\n  classical music", "abstract": "  This work introduces a new music generation system, called\nAffectMachine-Classical, that is capable of generating affective Classic music\nin real-time. AffectMachine was designed to be incorporated into biofeedback\nsystems (such as brain-computer-interfaces) to help users become aware of, and\nultimately mediate, their own dynamic affective states. That is, this system\nwas developed for music-based MedTech to support real-time emotion\nself-regulation in users. We provide an overview of the rule-based,\nprobabilistic system architecture, describing the main aspects of the system\nand how they are novel. We then present the results of a listener study that\nwas conducted to validate the ability of the system to reliably convey target\nemotions to listeners. The findings indicate that AffectMachine-Classical is\nvery effective in communicating various levels of Arousal ($R^2 = .96$) to\nlisteners, and is also quite convincing in terms of Valence (R^2 = .90). Future\nwork will embed AffectMachine-Classical into biofeedback systems, to leverage\nthe efficacy of the affective music for emotional well-being in listeners.\n", "date": "2023-04-11", "categories": []}
{"id": "2304.03867", "title": "Masked Student Dataset of Expressions", "abstract": "  Facial expression recognition (FER) algorithms work well in constrained\nenvironments with little or no occlusion of the face. However, real-world face\nocclusion is prevalent, most notably with the need to use a face mask in the\ncurrent Covid-19 scenario. While there are works on the problem of occlusion in\nFER, little has been done before on the particular face mask scenario.\nMoreover, the few works in this area largely use synthetically created masked\nFER datasets. Motivated by these challenges posed by the pandemic to FER, we\npresent a novel dataset, the Masked Student Dataset of Expressions or MSD-E,\nconsisting of 1,960 real-world non-masked and masked facial expression images\ncollected from 142 individuals. Along with the issue of obfuscated facial\nfeatures, we illustrate how other subtler issues in masked FER are represented\nin our dataset. We then provide baseline results using ResNet-18, finding that\nits performance dips in the non-masked case when trained for FER in the\npresence of masks. To tackle this, we test two training paradigms: contrastive\nlearning and knowledge distillation, and find that they increase the model's\nperformance in the masked scenario while maintaining its non-masked\nperformance. We further visualise our results using t-SNE plots and Grad-CAM,\ndemonstrating that these paradigms capitalise on the limited features available\nin the masked scenario. Finally, we benchmark SOTA methods on MSD-E.\n", "date": "2023-04-07", "categories": []}
{"id": "2304.03899", "title": "An Empirical Study and Improvement for Speech Emotion Recognition", "abstract": "  Multimodal speech emotion recognition aims to detect speakers' emotions from\naudio and text. Prior works mainly focus on exploiting advanced networks to\nmodel and fuse different modality information to facilitate performance, while\nneglecting the effect of different fusion strategies on emotion recognition. In\nthis work, we consider a simple yet important problem: how to fuse audio and\ntext modality information is more helpful for this multimodal task. Further, we\npropose a multimodal emotion recognition model improved by perspective loss.\nEmpirical results show our method obtained new state-of-the-art results on the\nIEMOCAP dataset. The in-depth analysis explains why the improved model can\nachieve improvements and outperforms baselines.\n", "date": "2023-04-08", "categories": []}
{"id": "2301.00508", "title": "EmoGator: A New Open Source Vocal Burst Dataset with Baseline Machine\n  Learning Classification Methodologies", "abstract": "  Vocal Bursts -- short, non-speech vocalizations that convey emotions, such as\nlaughter, cries, sighs, moans, and groans -- are an often-overlooked aspect of\nspeech emotion recognition, but an important aspect of human vocal\ncommunication. One barrier to study of these interesting vocalizations is a\nlack of large datasets. I am pleased to introduce the EmoGator dataset, which\nconsists of 32,130 samples from 357 speakers, 16.9654 hours of audio; each\nsample classified into one of 30 distinct emotion categories by the speaker.\nSeveral different approaches to construct classifiers to identify emotion\ncategories will be discussed, and directions for future research will be\nsuggested. Data set is available for download from\nhttps://github.com/fredbuhl/EmoGator.\n", "date": "2023-01-02", "categories": []}
{"id": "2304.03064", "title": "An experimental study in Real-time Facial Emotion Recognition on new 3RL\n  dataset", "abstract": "  Although real-time facial emotion recognition is a hot topic research domain\nin the field of human-computer interaction, state-of the-art available datasets\nstill suffer from various problems, such as some unrelated photos such as\ndocument photos, unbalanced numbers of photos in each class, and misleading\nimages that can negatively affect correct classification. The 3RL dataset was\ncreated, which contains approximately 24K images and will be publicly\navailable, to overcome previously available dataset problems. The 3RL dataset\nis labelled with five basic emotions: happiness, fear, sadness, disgust, and\nanger. Moreover, we compared the 3RL dataset with other famous state-of-the-art\ndatasets (FER dataset, CK+ dataset), and we applied the most commonly used\nalgorithms in previous works, SVM and CNN. The results show a noticeable\nimprovement in generalization on the 3RL dataset. Experiments have shown an\naccuracy of up to 91.4% on 3RL dataset using CNN where results on FER2013, CK+\nare, respectively (approximately from 60% to 85%).\n", "date": "2023-04-06", "categories": []}
{"id": "2304.03195", "title": "Micron-BERT: BERT-based Facial Micro-Expression Recognition", "abstract": "  Micro-expression recognition is one of the most challenging topics in\naffective computing. It aims to recognize tiny facial movements difficult for\nhumans to perceive in a brief period, i.e., 0.25 to 0.5 seconds. Recent\nadvances in pre-training deep Bidirectional Transformers (BERT) have\nsignificantly improved self-supervised learning tasks in computer vision.\nHowever, the standard BERT in vision problems is designed to learn only from\nfull images or videos, and the architecture cannot accurately detect details of\nfacial micro-expressions. This paper presents Micron-BERT ($\\mu$-BERT), a novel\napproach to facial micro-expression recognition. The proposed method can\nautomatically capture these movements in an unsupervised manner based on two\nkey ideas. First, we employ Diagonal Micro-Attention (DMA) to detect tiny\ndifferences between two frames. Second, we introduce a new Patch of Interest\n(PoI) module to localize and highlight micro-expression interest regions and\nsimultaneously reduce noisy backgrounds and distractions. By incorporating\nthese components into an end-to-end deep network, the proposed $\\mu$-BERT\nsignificantly outperforms all previous work in various micro-expression tasks.\n$\\mu$-BERT can be trained on a large-scale unlabeled dataset, i.e., up to 8\nmillion images, and achieves high accuracy on new unseen facial\nmicro-expression datasets. Empirical experiments show $\\mu$-BERT consistently\noutperforms state-of-the-art performance on four micro-expression benchmarks,\nincluding SAMM, CASME II, SMIC, and CASME3, by significant margins. Code will\nbe available at \\url{https://github.com/uark-cviu/Micron-BERT}\n", "date": "2023-04-06", "categories": []}
{"id": "2304.02187", "title": "Bodily expressed emotion understanding through integrating Laban\n  movement analysis", "abstract": "  Body movements carry important information about a person's emotions or\nmental state and are essential in daily communication. Enhancing the ability of\nmachines to understand emotions expressed through body language can improve the\ncommunication of assistive robots with children and elderly users, provide\npsychiatric professionals with quantitative diagnostic and prognostic\nassistance, and aid law enforcement in identifying deception. This study\ndevelops a high-quality human motor element dataset based on the Laban Movement\nAnalysis movement coding system and utilizes that to jointly learn about motor\nelements and emotions. Our long-term ambition is to integrate knowledge from\ncomputing, psychology, and performing arts to enable automated understanding\nand analysis of emotion and mental state through body language. This work\nserves as a launchpad for further research into recognizing emotions through\nanalysis of human movement.\n", "date": "2023-04-05", "categories": []}
{"id": "2303.17608", "title": "Visual Response to Emotional State of User Interaction", "abstract": "  This work proposes an interactive art installation \"Mood spRing\" designed to\nreflect the mood of the environment through interpretation of language and\ntone. Mood spRing consists of an AI program that controls an immersive 3D\nanimation of the seasons. If the AI program perceives the language and tone of\nthe users as pleasant, the animation progresses through idealized renditions of\nseasons. Otherwise, it slips into unpleasant weather and natural disasters of\nthe season. To interpret the language and tone of the user interaction, hybrid\nstate-of-the-art emotion detection methods are applied to the user audio and\ntext inputs. The emotional states detected separately from tone and language\nare fused by a novel approach that aims at minimizing the possible model\ndisparity across diverse demographic groups.\n", "date": "2023-03-27", "categories": []}
{"id": "2303.17611", "title": "Transformer-based Self-supervised Multimodal Representation Learning for\n  Wearable Emotion Recognition", "abstract": "  Recently, wearable emotion recognition based on peripheral physiological\nsignals has drawn massive attention due to its less invasive nature and its\napplicability in real-life scenarios. However, how to effectively fuse\nmultimodal data remains a challenging problem. Moreover, traditional\nfully-supervised based approaches suffer from overfitting given limited labeled\ndata. To address the above issues, we propose a novel self-supervised learning\n(SSL) framework for wearable emotion recognition, where efficient multimodal\nfusion is realized with temporal convolution-based modality-specific encoders\nand a transformer-based shared encoder, capturing both intra-modal and\ninter-modal correlations. Extensive unlabeled data is automatically assigned\nlabels by five signal transforms, and the proposed SSL model is pre-trained\nwith signal transformation recognition as a pretext task, allowing the\nextraction of generalized multimodal representations for emotion-related\ndownstream tasks. For evaluation, the proposed SSL model was first pre-trained\non a large-scale self-collected physiological dataset and the resulting encoder\nwas subsequently frozen or fine-tuned on three public supervised emotion\nrecognition datasets. Ultimately, our SSL-based method achieved\nstate-of-the-art results in various emotion classification tasks. Meanwhile,\nthe proposed model proved to be more accurate and robust compared to\nfully-supervised methods on low data regimes.\n", "date": "2023-03-29", "categories": []}
{"id": "2303.11921", "title": "Context De-confounded Emotion Recognition", "abstract": "  Context-Aware Emotion Recognition (CAER) is a crucial and challenging task\nthat aims to perceive the emotional states of the target person with contextual\ninformation. Recent approaches invariably focus on designing sophisticated\narchitectures or mechanisms to extract seemingly meaningful representations\nfrom subjects and contexts. However, a long-overlooked issue is that a context\nbias in existing datasets leads to a significantly unbalanced distribution of\nemotional states among different context scenarios. Concretely, the harmful\nbias is a confounder that misleads existing models to learn spurious\ncorrelations based on conventional likelihood estimation, significantly\nlimiting the models' performance. To tackle the issue, this paper provides a\ncausality-based perspective to disentangle the models from the impact of such\nbias, and formulate the causalities among variables in the CAER task via a\ntailored causal graph. Then, we propose a Contextual Causal Intervention Module\n(CCIM) based on the backdoor adjustment to de-confound the confounder and\nexploit the true causal effect for model training. CCIM is plug-in and\nmodel-agnostic, which improves diverse state-of-the-art approaches by\nconsiderable margins. Extensive experiments on three benchmark datasets\ndemonstrate the effectiveness of our CCIM and the significance of causal\ninsight.\n", "date": "2023-03-21", "categories": []}
{"id": "2303.14708", "title": "Exploring Multimodal Sentiment Analysis via CBAM Attention and\n  Double-layer BiLSTM Architecture", "abstract": "  Because multimodal data contains more modal information, multimodal sentiment\nanalysis has become a recent research hotspot. However, redundant information\nis easily involved in feature fusion after feature extraction, which has a\ncertain impact on the feature representation after fusion. Therefore, in this\npapaer, we propose a new multimodal sentiment analysis model. In our model, we\nuse BERT + BiLSTM as new feature extractor to capture the long-distance\ndependencies in sentences and consider the position information of input\nsequences to obtain richer text features. To remove redundant information and\nmake the network pay more attention to the correlation between image and text\nfeatures, CNN and CBAM attention are added after splicing text features and\npicture features, to improve the feature representation ability. On the\nMVSA-single dataset and HFM dataset, compared with the baseline model, the ACC\nof our model is improved by 1.78% and 1.91%, and the F1 value is enhanced by\n3.09% and 2.0%, respectively. The experimental results show that our model\nachieves a sound effect, similar to the advanced model.\n", "date": "2023-03-26", "categories": []}
{"id": "2303.13802", "title": "Decoupled Multimodal Distilling for Emotion Recognition", "abstract": "  Human multimodal emotion recognition (MER) aims to perceive human emotions\nvia language, visual and acoustic modalities. Despite the impressive\nperformance of previous MER approaches, the inherent multimodal heterogeneities\nstill haunt and the contribution of different modalities varies significantly.\nIn this work, we mitigate this issue by proposing a decoupled multimodal\ndistillation (DMD) approach that facilitates flexible and adaptive crossmodal\nknowledge distillation, aiming to enhance the discriminative features of each\nmodality. Specially, the representation of each modality is decoupled into two\nparts, i.e., modality-irrelevant/-exclusive spaces, in a self-regression\nmanner. DMD utilizes a graph distillation unit (GD-Unit) for each decoupled\npart so that each GD can be performed in a more specialized and effective\nmanner. A GD-Unit consists of a dynamic graph where each vertice represents a\nmodality and each edge indicates a dynamic knowledge distillation. Such GD\nparadigm provides a flexible knowledge transfer manner where the distillation\nweights can be automatically learned, thus enabling diverse crossmodal\nknowledge transfer patterns. Experimental results show DMD consistently obtains\nsuperior performance than state-of-the-art MER methods. Visualization results\nshow the graph edges in DMD exhibit meaningful distributional patterns w.r.t.\nthe modality-irrelevant/-exclusive feature spaces. Codes are released at\n\\url{https://github.com/mdswyz/DMD}.\n", "date": "2023-03-24", "categories": []}
{"id": "2303.13364", "title": "Reevaluating Data Partitioning for Emotion Detection in EmoWOZ", "abstract": "  This paper focuses on the EmoWoz dataset, an extension of MultiWOZ that\nprovides emotion labels for the dialogues. MultiWOZ was partitioned initially\nfor another purpose, resulting in a distributional shift when considering the\nnew purpose of emotion recognition. The emotion tags in EmoWoz are highly\nimbalanced and unevenly distributed across the partitions, which causes\nsub-optimal performance and poor comparison of models. We propose a stratified\nsampling scheme based on emotion tags to address this issue, improve the\ndataset's distribution, and reduce dataset shift. We also introduce a special\ntechnique to handle conversation (sequential) data with many emotional tags.\nUsing our proposed sampling method, models built upon EmoWoz can perform\nbetter, making it a more reliable resource for training conversational agents\nwith emotional intelligence. We recommend that future researchers use this new\npartitioning to ensure consistent and accurate performance evaluations.\n", "date": "2023-03-15", "categories": []}
{"id": "2303.11421", "title": "Improving EEG-based Emotion Recognition by Fusing Time-frequency And\n  Spatial Representations", "abstract": "  Using deep learning methods to classify EEG signals can accurately identify\npeople's emotions. However, existing studies have rarely considered the\napplication of the information in another domain's representations to feature\nselection in the time-frequency domain. We propose a classification network of\nEEG signals based on the cross-domain feature fusion method, which makes the\nnetwork more focused on the features most related to brain activities and\nthinking changes by using the multi-domain attention mechanism. In addition, we\npropose a two-step fusion method and apply these methods to the EEG emotion\nrecognition network. Experimental results show that our proposed network, which\ncombines multiple representations in the time-frequency domain and spatial\ndomain, outperforms previous methods on public datasets and achieves\nstate-of-the-art at present.\n", "date": "2023-03-14", "categories": []}
{"id": "2303.11840", "title": "Self-Paced Neutral Expression-Disentangled Learning for Facial\n  Expression Recognition", "abstract": "  The accuracy of facial expression recognition is typically affected by the\nfollowing factors: high similarities across different expressions, disturbing\nfactors, and micro-facial movement of rapid and subtle changes. One potentially\nviable solution for addressing these barriers is to exploit the neutral\ninformation concealed in neutral expression images. To this end, in this paper\nwe propose a self-Paced Neutral Expression-Disentangled Learning (SPNDL) model.\nSPNDL disentangles neutral information from facial expressions, making it\neasier to extract key and deviation features. Specifically, it allows to\ncapture discriminative information among similar expressions and perceive\nmicro-facial movements. In order to better learn these neutral\nexpression-disentangled features (NDFs) and to alleviate the non-convex\noptimization problem, a self-paced learning (SPL) strategy based on NDFs is\nproposed in the training stage. SPL learns samples from easy to complex by\nincreasing the number of samples selected into the training process, which\nenables to effectively suppress the negative impacts introduced by low-quality\nsamples and inconsistently distributed NDFs. Experiments on three popular\ndatabases (i.e., CK+, Oulu-CASIA, and RAF-DB) show the effectiveness of our\nproposed method.\n", "date": "2023-03-21", "categories": []}
{"id": "2303.01498", "title": "ABAW: Valence-Arousal Estimation, Expression Recognition, Action Unit\n  Detection & Emotional Reaction Intensity Estimation Challenges", "abstract": "  The fifth Affective Behavior Analysis in-the-wild (ABAW) Competition is part\nof the respective ABAW Workshop which will be held in conjunction with IEEE\nComputer Vision and Pattern Recognition Conference (CVPR), 2023. The 5th ABAW\nCompetition is a continuation of the Competitions held at ECCV 2022, IEEE CVPR\n2022, ICCV 2021, IEEE FG 2020 and CVPR 2017 Conferences, and is dedicated at\nautomatically analyzing affect. For this year's Competition, we feature two\ncorpora: i) an extended version of the Aff-Wild2 database and ii) the\nHume-Reaction dataset. The former database is an audiovisual one of around 600\nvideos of around 3M frames and is annotated with respect to:a) two continuous\naffect dimensions -valence (how positive/negative a person is) and arousal (how\nactive/passive a person is)-; b) basic expressions (e.g. happiness, sadness,\nneutral state); and c) atomic facial muscle actions (i.e., action units). The\nlatter dataset is an audiovisual one in which reactions of individuals to\nemotional stimuli have been annotated with respect to seven emotional\nexpression intensities. Thus the 5th ABAW Competition encompasses four\nChallenges: i) uni-task Valence-Arousal Estimation, ii) uni-task Expression\nClassification, iii) uni-task Action Unit Detection, and iv) Emotional Reaction\nIntensity Estimation. In this paper, we present these Challenges, along with\ntheir corpora, we outline the evaluation metrics, we present the baseline\nsystems and illustrate their obtained performance.\n", "date": "2023-03-02", "categories": []}
{"id": "2303.08419", "title": "Multi Modal Facial Expression Recognition with Transformer-Based Fusion\n  Networks and Dynamic Sampling", "abstract": "  Facial expression recognition is an essential task for various applications,\nincluding emotion detection, mental health analysis, and human-machine\ninteractions. In this paper, we propose a multi-modal facial expression\nrecognition method that exploits audio information along with facial images to\nprovide a crucial clue to differentiate some ambiguous facial expressions.\nSpecifically, we introduce a Modal Fusion Module (MFM) to fuse audio-visual\ninformation, where image and audio features are extracted from Swin\nTransformer. Additionally, we tackle the imbalance problem in the dataset by\nemploying dynamic data resampling. Our model has been evaluated in the\nAffective Behavior in-the-wild (ABAW) challenge of CVPR 2023.\n", "date": "2023-03-15", "categories": []}
{"id": "2303.09158", "title": "Facial Affect Recognition based on Transformer Encoder and Audiovisual\n  Fusion for the ABAW5 Challenge", "abstract": "  In this paper, we present our solutions for the 5th Workshop and Competition\non Affective Behavior Analysis in-the-wild (ABAW), which includes four\nsub-challenges of Valence-Arousal (VA) Estimation, Expression (Expr)\nClassification, Action Unit (AU) Detection and Emotional Reaction Intensity\n(ERI) Estimation. The 5th ABAW competition focuses on facial affect recognition\nutilizing different modalities and datasets. In our work, we extract powerful\naudio and visual features using a large number of sota models. These features\nare fused by Transformer Encoder and TEMMA. Besides, to avoid the possible\nimpact of large dimensional differences between various features, we design an\nAffine Module to align different features to the same dimension. Extensive\nexperiments demonstrate that the superiority of the proposed method. For the VA\nEstimation sub-challenge, our method obtains the mean Concordance Correlation\nCoefficient (CCC) of 0.6066. For the Expression Classification sub-challenge,\nthe average F1 Score is 0.4055. For the AU Detection sub-challenge, the average\nF1 Score is 0.5296. For the Emotional Reaction Intensity Estimation\nsub-challenge, the average pearson's correlations coefficient on the validation\nset is 0.3968. All of the results of four sub-challenges outperform the\nbaseline with a large margin.\n", "date": "2023-03-16", "categories": []}
{"id": "2303.09293", "title": "A transformer-based approach to video frame-level prediction in\n  Affective Behaviour Analysis In-the-wild", "abstract": "  In recent years, transformer architecture has been a dominating paradigm in\nmany applications, including affective computing. In this report, we propose\nour transformer-based model to handle Emotion Classification Task in the 5th\nAffective Behavior Analysis In-the-wild Competition. By leveraging the\nattentive model and the synthetic dataset, we attain a score of 0.4775 on the\nvalidation set of Aff-Wild2, the dataset provided by the organizer.\n", "date": "2023-03-16", "categories": []}
{"id": "2303.10421", "title": "Mutilmodal Feature Extraction and Attention-based Fusion for Emotion\n  Estimation in Videos", "abstract": "  The continuous improvement of human-computer interaction technology makes it\npossible to compute emotions. In this paper, we introduce our submission to the\nCVPR 2023 Competition on Affective Behavior Analysis in-the-wild (ABAW).\nSentiment analysis in human-computer interaction should, as far as possible\nStart with multiple dimensions, fill in the single imperfect emotion channel,\nand finally determine the emotion tendency by fitting multiple results.\nTherefore, We exploited multimodal features extracted from video of different\nlengths from the competition dataset, including audio, pose and images.\nWell-informed emotion representations drive us to propose a Attention-based\nmultimodal framework for emotion estimation. Our system achieves the\nperformance of 0.361 on the validation dataset. The code is available at\n[https://github.com/xkwangcn/ABAW-5th-RT-IAI].\n", "date": "2023-03-18", "categories": []}
{"id": "2303.10511", "title": "Exploring Expression-related Self-supervised Learning for Affective\n  Behaviour Analysis", "abstract": "  This paper explores an expression-related self-supervised learning (SSL)\nmethod (ContraWarping) to perform expression classification in the 5th\nAffective Behavior Analysis in-the-wild (ABAW) competition. Affective datasets\nare expensive to annotate, and SSL methods could learn from large-scale\nunlabeled data, which is more suitable for this task. By evaluating on the\nAff-Wild2 dataset, we demonstrate that ContraWarping outperforms most existing\nsupervised methods and shows great application potential in the affective\nanalysis area. Codes will be released on:\nhttps://github.com/youqingxiaozhua/ABAW5.\n", "date": "2023-03-18", "categories": []}
{"id": "2303.10561", "title": "Spatial-temporal Transformer for Affective Behavior Analysis", "abstract": "  The in-the-wild affective behavior analysis has been an important study. In\nthis paper, we submit our solutions for the 5th Workshop and Competition on\nAffective Behavior Analysis in-the-wild (ABAW), which includes V-A Estimation,\nFacial Expression Classification and AU Detection Sub-challenges. We propose a\nTransformer Encoder with Multi-Head Attention framework to learn the\ndistribution of both the spatial and temporal features. Besides, there are\nvirious effective data augmentation strategies employed to alleviate the\nproblems of sample imbalance during model training. The results fully\ndemonstrate the effectiveness of our proposed model based on the Aff-Wild2\ndataset.\n", "date": "2023-03-19", "categories": []}
{"id": "2303.10033", "title": "Multi-modal Expression Recognition with Ensemble Method", "abstract": "  This paper presents our submission to the Expression Classification Challenge\nof the fifth Affective Behavior Analysis in-the-wild (ABAW) Competition. In our\nmethod, multimodal feature combinations extracted by several different\npre-trained models are applied to capture more effective emotional information.\nFor these combinations of visual and audio modal features, we utilize two\ntemporal encoders to explore the temporal contextual information in the data.\nIn addition, we employ several ensemble strategies for different experimental\nsettings to obtain the most accurate expression recognition results. Our system\nachieves the average F1 Score of 0.45774 on the validation set.\n", "date": "2023-03-17", "categories": []}
{"id": "2303.09034", "title": "Unsupervised Facial Expression Representation Learning with Contrastive\n  Local Warping", "abstract": "  This paper investigates unsupervised representation learning for facial\nexpression analysis. We think Unsupervised Facial Expression Representation\n(UFER) deserves exploration and has the potential to address some key\nchallenges in facial expression analysis, such as scaling, annotation bias, the\ndiscrepancy between discrete labels and continuous emotions, and model\npre-training. Such motivated, we propose a UFER method with contrastive local\nwarping (ContraWarping), which leverages the insight that the emotional\nexpression is robust to current global transformation (affine transformation,\ncolor jitter, etc.) but can be easily changed by random local warping.\nTherefore, given a facial image, ContraWarping employs some global\ntransformations and local warping to generate its positive and negative samples\nand sets up a novel contrastive learning framework. Our in-depth investigation\nshows that: 1) the positive pairs from global transformations may be exploited\nwith general self-supervised learning (e.g., BYOL) and already bring some\ninformative features, and 2) the negative pairs from local warping explicitly\nintroduce expression-related variation and further bring substantial\nimprovement. Based on ContraWarping, we demonstrate the benefit of UFER under\ntwo facial expression analysis scenarios: facial expression recognition and\nimage retrieval. For example, directly using ContraWarping features for linear\nprobing achieves 79.14% accuracy on RAF-DB, significantly reducing the gap\ntowards the full-supervised counterpart (88.92% / 84.81% with/without\npre-training).\n", "date": "2023-03-16", "categories": []}
{"id": "2303.09145", "title": "Facial Affective Behavior Analysis Method for 5th ABAW Competition", "abstract": "  Facial affective behavior analysis is important for human-computer\ninteraction. 5th ABAW competition includes three challenges from Aff-Wild2\ndatabase. Three common facial affective analysis tasks are involved, i.e.\nvalence-arousal estimation, expression classification, action unit recognition.\nFor the three challenges, we construct three different models to solve the\ncorresponding problems to improve the results, such as data unbalance and data\nnoise. For the experiments of three challenges, we train the models on the\nprovided training data and validate the models on the validation data.\n", "date": "2023-03-16", "categories": []}
{"id": "2303.09162", "title": "EmotiEffNet Facial Features in Uni-task Emotion Recognition in Video at\n  ABAW-5 competition", "abstract": "  In this article, the results of our team for the fifth Affective Behavior\nAnalysis in-the-wild (ABAW) competition are presented. The usage of the\npre-trained convolutional networks from the EmotiEffNet family for frame-level\nfeature extraction is studied. In particular, we propose an ensemble of a\nmulti-layered perceptron and the LightAutoML-based classifier. The\npost-processing by smoothing the results for sequential frames is implemented.\nExperimental results for the large-scale Aff-Wild2 database demonstrate that\nour model achieves a much greater macro-averaged F1-score for facial expression\nrecognition and action unit detection and concordance correlation coefficients\nfor valence/arousal estimation when compared to baseline.\n", "date": "2023-03-16", "categories": []}
{"id": "2303.09167", "title": "Emotional Reaction Intensity Estimation Based on Multimodal Data", "abstract": "  This paper introduces our method for the Emotional Reaction Intensity (ERI)\nEstimation Challenge, in CVPR 2023: 5th Workshop and Competition on Affective\nBehavior Analysis in-the-wild (ABAW). Based on the multimodal data provided by\nthe originazers, we extract acoustic and visual features with different\npretrained models. The multimodal features are mixed together by Transformer\nEncoders with cross-modal attention mechnism. In this paper, 1. better features\nare extracted with the SOTA pretrained models. 2. Compared with the baseline,\nwe improve the Pearson's Correlations Coefficient a lot. 3. We process the data\nwith some special skills to enhance performance ability of our model.\n", "date": "2023-03-16", "categories": []}
{"id": "2303.09210", "title": "A Dual Branch Network for Emotional Reaction Intensity Estimation", "abstract": "  Emotional Reaction Intensity(ERI) estimation is an important task in\nmultimodal scenarios, and has fundamental applications in medicine, safe\ndriving and other fields. In this paper, we propose a solution to the ERI\nchallenge of the fifth Affective Behavior Analysis in-the-wild(ABAW), a\ndual-branch based multi-output regression model. The spatial attention is used\nto better extract visual features, and the Mel-Frequency Cepstral Coefficients\ntechnology extracts acoustic features, and a method named modality dropout is\nadded to fusion multimodal features. Our method achieves excellent results on\nthe official validation set.\n", "date": "2023-03-16", "categories": []}
{"id": "2303.09364", "title": "Tollywood Emotions: Annotation of Valence-Arousal in Telugu Song Lyrics", "abstract": "  Emotion recognition from a given music track has heavily relied on acoustic\nfeatures, social tags, and metadata but is seldom focused on lyrics. There are\nno datasets of Indian language songs that contain both valence and arousal\nmanual ratings of lyrics. We present a new manually annotated dataset of Telugu\nsongs' lyrics collected from Spotify with valence and arousal annotated on a\ndiscrete scale. A fairly high inter-annotator agreement was observed for both\nvalence and arousal. Subsequently, we create two music emotion recognition\nmodels by using two classification techniques to identify valence, arousal and\nrespective emotion quadrant from lyrics. Support vector machine (SVM) with term\nfrequency-inverse document frequency (TF-IDF) features and fine-tuning the\npre-trained XLMRoBERTa (XLM-R) model were used for valence, arousal and\nquadrant classification tasks. Fine-tuned XLMRoBERTa performs better than the\nSVM by improving macro-averaged F1-scores of 54.69%, 67.61%, 34.13% to 77.90%,\n80.71% and 58.33% for valence, arousal and quadrant classifications,\nrespectively, on 10-fold cross-validation. In addition, we compare our lyrics\nannotations with Spotify's annotations of valence and energy (same as arousal),\nwhich are based on entire music tracks. The implications of our findings are\ndiscussed. Finally, we make the dataset publicly available with lyrics,\nannotations and Spotify IDs.\n", "date": "2023-03-16", "categories": []}
{"id": "2303.07742", "title": "ForDigitStress: A multi-modal stress dataset employing a digital job\n  interview scenario", "abstract": "  We present a multi-modal stress dataset that uses digital job interviews to\ninduce stress. The dataset provides multi-modal data of 40 participants\nincluding audio, video (motion capturing, facial recognition, eye tracking) as\nwell as physiological information (photoplethysmography, electrodermal\nactivity). In addition to that, the dataset contains time-continuous\nannotations for stress and occurred emotions (e.g. shame, anger, anxiety,\nsurprise). In order to establish a baseline, five different machine learning\nclassifiers (Support Vector Machine, K-Nearest Neighbors, Random Forest,\nLong-Short-Term Memory Network) have been trained and evaluated on the proposed\ndataset for a binary stress classification task. The best-performing classifier\nachieved an accuracy of 88.3% and an F1-score of 87.5%.\n", "date": "2023-03-14", "categories": []}
{"id": "2303.07831", "title": "Quaternion Orthogonal Transformer for Facial Expression Recognition in\n  the Wild", "abstract": "  Facial expression recognition (FER) is a challenging topic in artificial\nintelligence. Recently, many researchers have attempted to introduce Vision\nTransformer (ViT) to the FER task. However, ViT cannot fully utilize emotional\nfeatures extracted from raw images and requires a lot of computing resources.\nTo overcome these problems, we propose a quaternion orthogonal transformer\n(QOT) for FER. Firstly, to reduce redundancy among features extracted from\npre-trained ResNet-50, we use the orthogonal loss to decompose and compact\nthese features into three sets of orthogonal sub-features. Secondly, three\northogonal sub-features are integrated into a quaternion matrix, which\nmaintains the correlations between different orthogonal components. Finally, we\ndevelop a quaternion vision transformer (Q-ViT) for feature classification. The\nQ-ViT adopts quaternion operations instead of the original operations in ViT,\nwhich improves the final accuracies with fewer parameters. Experimental results\non three in-the-wild FER datasets show that the proposed QOT outperforms\nseveral state-of-the-art models and reduces the computations.\n", "date": "2023-03-14", "categories": []}
{"id": "2303.08027", "title": "A Hierarchical Regression Chain Framework for Affective Vocal Burst\n  Recognition", "abstract": "  As a common way of emotion signaling via non-linguistic vocalizations, vocal\nburst (VB) plays an important role in daily social interaction. Understanding\nand modeling human vocal bursts are indispensable for developing robust and\ngeneral artificial intelligence. Exploring computational approaches for\nunderstanding vocal bursts is attracting increasing research attention. In this\nwork, we propose a hierarchical framework, based on chain regression models,\nfor affective recognition from VBs, that explicitly considers multiple\nrelationships: (i) between emotional states and diverse cultures; (ii) between\nlow-dimensional (arousal & valence) and high-dimensional (10 emotion classes)\nemotion spaces; and (iii) between various emotion classes within the\nhigh-dimensional space. To address the challenge of data sparsity, we also use\nself-supervised learning (SSL) representations with layer-wise and temporal\naggregation modules. The proposed systems participated in the ACII Affective\nVocal Burst (A-VB) Challenge 2022 and ranked first in the \"TWO'' and \"CULTURE''\ntasks. Experimental results based on the ACII Challenge 2022 dataset\ndemonstrate the superior performance of the proposed system and the\neffectiveness of considering multiple relationships using hierarchical\nregression chain models.\n", "date": "2023-03-14", "categories": []}
{"id": "2210.03538", "title": "An Overview of Affective Speech Synthesis and Conversion in the Deep\n  Learning Era", "abstract": "  Speech is the fundamental mode of human communication, and its synthesis has\nlong been a core priority in human-computer interaction research. In recent\nyears, machines have managed to master the art of generating speech that is\nunderstandable by humans. But the linguistic content of an utterance\nencompasses only a part of its meaning. Affect, or expressivity, has the\ncapacity to turn speech into a medium capable of conveying intimate thoughts,\nfeelings, and emotions -- aspects that are essential for engaging and\nnaturalistic interpersonal communication. While the goal of imparting\nexpressivity to synthesised utterances has so far remained elusive, following\nrecent advances in text-to-speech synthesis, a paradigm shift is well under way\nin the fields of affective speech synthesis and conversion as well. Deep\nlearning, as the technology which underlies most of the recent advances in\nartificial intelligence, is spearheading these efforts. In the present\noverview, we outline ongoing trends and summarise state-of-the-art approaches\nin an attempt to provide a comprehensive overview of this exciting field.\n", "date": "2022-10-06", "categories": []}
{"id": "2211.00171", "title": "Using Emotion Embeddings to Transfer Knowledge Between Emotions,\n  Languages, and Annotation Formats", "abstract": "  The need for emotional inference from text continues to diversify as more and\nmore disciplines integrate emotions into their theories and applications. These\nneeds include inferring different emotion types, handling multiple languages,\nand different annotation formats. A shared model between different\nconfigurations would enable the sharing of knowledge and a decrease in training\ncosts, and would simplify the process of deploying emotion recognition models\nin novel environments. In this work, we study how we can build a single model\nthat can transition between these different configurations by leveraging\nmultilingual models and Demux, a transformer-based model whose input includes\nthe emotions of interest, enabling us to dynamically change the emotions\npredicted by the model. Demux also produces emotion embeddings, and performing\noperations on them allows us to transition to clusters of emotions by pooling\nthe embeddings of each cluster. We show that Demux can simultaneously transfer\nknowledge in a zero-shot manner to a new language, to a novel annotation format\nand to unseen emotions. Code is available at\nhttps://github.com/gchochla/Demux-MEmo .\n", "date": "2022-10-31", "categories": []}
{"id": "2303.06904", "title": "Contextually-rich human affect perception using multimodal scene\n  information", "abstract": "  The process of human affect understanding involves the ability to infer\nperson specific emotional states from various sources including images, speech,\nand language. Affect perception from images has predominantly focused on\nexpressions extracted from salient face crops. However, emotions perceived by\nhumans rely on multiple contextual cues including social settings, foreground\ninteractions, and ambient visual scenes. In this work, we leverage pretrained\nvision-language (VLN) models to extract descriptions of foreground context from\nimages. Further, we propose a multimodal context fusion (MCF) module to combine\nforeground cues with the visual scene and person-based contextual information\nfor emotion prediction. We show the effectiveness of our proposed modular\ndesign on two datasets associated with natural scenes and TV shows.\n", "date": "2023-03-13", "categories": []}
{"id": "2208.06411", "title": "SFF-DA: Sptialtemporal Feature Fusion for Detecting Anxiety\n  Nonintrusively", "abstract": "  Early detection of anxiety is crucial for reducing the suffering of\nindividuals with mental disorders and improving treatment outcomes. Utilizing\nan mHealth platform for anxiety screening can be particularly practical in\nimproving screening efficiency and reducing costs. However, the effectiveness\nof existing methods has been hindered by differences in mobile devices used to\ncapture subjects' physical and mental evaluations, as well as by the\nvariability in data quality and small sample size problems encountered in\nreal-world settings. To address these issues, we propose a framework with\nspatiotemporal feature fusion for detecting anxiety nonintrusively. We use a\nfeature extraction network based on a 3D convolutional network and long\nshort-term memory (\"3DCNN+LSTM\") to fuse the spatiotemporal features of facial\nbehavior and noncontact physiology, which reduces the impact of uneven data\nquality. Additionally, we design a similarity assessment strategy to address\nthe issue of deteriorating model accuracy due to small sample sizes. Our\nframework is validated with a crew dataset from the real world and two public\ndatasets: the University of Burgundy Franche-Comt\\'e Psychophysiological\n(UBFC-Phys) dataset and the Smart Reasoning for Well-being at Home and at Work\nfor Knowledge Work (SWELL-KW) dataset. The experimental results indicate that\nour framework outperforms the comparison methods.\n", "date": "2022-08-12", "categories": []}
{"id": "2302.09856", "title": "Knowledge-aware Bayesian Co-attention for Multimodal Emotion Recognition", "abstract": "  Multimodal emotion recognition is a challenging research area that aims to\nfuse different modalities to predict human emotion. However, most existing\nmodels that are based on attention mechanisms have difficulty in learning\nemotionally relevant parts on their own. To solve this problem, we propose to\nincorporate external emotion-related knowledge in the co-attention based fusion\nof pre-trained models. To effectively incorporate this knowledge, we enhance\nthe co-attention model with a Bayesian attention module (BAM) where a prior\ndistribution is estimated using the emotion-related knowledge. Experimental\nresults on the IEMOCAP dataset show that the proposed approach can outperform\nseveral state-of-the-art approaches by at least 0.7% unweighted accuracy (UA).\n", "date": "2023-02-20", "categories": []}
{"id": "2303.03510", "title": "Guilt Detection in Text: A Step Towards Understanding Complex Emotions", "abstract": "  We introduce a novel Natural Language Processing (NLP) task called Guilt\ndetection, which focuses on detecting guilt in text. We identify guilt as a\ncomplex and vital emotion that has not been previously studied in NLP, and we\naim to provide a more fine-grained analysis of it. To address the lack of\npublicly available corpora for guilt detection, we created VIC, a dataset\ncontaining 4622 texts from three existing emotion detection datasets that we\nbinarized into guilt and no-guilt classes. We experimented with traditional\nmachine learning methods using bag-of-words and term frequency-inverse document\nfrequency features, achieving a 72% f1 score with the highest-performing model.\nOur study provides a first step towards understanding guilt in text and opens\nthe door for future research in this area.\n", "date": "2023-03-06", "categories": []}
{"id": "2303.03186", "title": "Will Affective Computing Emerge from Foundation Models and General AI? A\n  First Evaluation on ChatGPT", "abstract": "  ChatGPT has shown the potential of emerging general artificial intelligence\ncapabilities, as it has demonstrated competent performance across many natural\nlanguage processing tasks. In this work, we evaluate the capabilities of\nChatGPT to perform text classification on three affective computing problems,\nnamely, big-five personality prediction, sentiment analysis, and suicide\ntendency detection. We utilise three baselines, a robust language model\n(RoBERTa-base), a legacy word model with pretrained embeddings (Word2Vec), and\na simple bag-of-words baseline (BoW). Results show that the RoBERTa trained for\na specific downstream task generally has a superior performance. On the other\nhand, ChatGPT provides decent results, and is relatively comparable to the\nWord2Vec and BoW baselines. ChatGPT further shows robustness against noisy\ndata, where Word2Vec models achieve worse results due to noise. Results\nindicate that ChatGPT is a good generalist model that is capable of achieving\ngood results across various problems without any specialised training, however,\nit is not as good as a specialised model for a downstream task.\n", "date": "2023-03-03", "categories": []}
{"id": "2303.01694", "title": "DWFormer: Dynamic Window transFormer for Speech Emotion Recognition", "abstract": "  Speech emotion recognition is crucial to human-computer interaction. The\ntemporal regions that represent different emotions scatter in different parts\nof the speech locally. Moreover, the temporal scales of important information\nmay vary over a large range within and across speech segments. Although\ntransformer-based models have made progress in this field, the existing models\ncould not precisely locate important regions at different temporal scales. To\naddress the issue, we propose Dynamic Window transFormer (DWFormer), a new\narchitecture that leverages temporal importance by dynamically splitting\nsamples into windows. Self-attention mechanism is applied within windows for\ncapturing temporal important information locally in a fine-grained way.\nCross-window information interaction is also taken into account for global\ncommunication. DWFormer is evaluated on both the IEMOCAP and the MELD datasets.\nExperimental results show that the proposed model achieves better performance\nthan the previous state-of-the-art methods.\n", "date": "2023-03-03", "categories": []}
{"id": "2303.01776", "title": "Prior Information based Decomposition and Reconstruction Learning for\n  Micro-Expression Recognition", "abstract": "  Micro-expression recognition (MER) draws intensive research interest as\nmicro-expressions (MEs) can infer genuine emotions. Prior information can guide\nthe model to learn discriminative ME features effectively. However, most works\nfocus on researching the general models with a stronger representation ability\nto adaptively aggregate ME movement information in a holistic way, which may\nignore the prior information and properties of MEs. To solve this issue, driven\nby the prior information that the category of ME can be inferred by the\nrelationship between the actions of facial different components, this work\ndesigns a novel model that can conform to this prior information and learn ME\nmovement features in an interpretable way. Specifically, this paper proposes a\nDecomposition and Reconstruction-based Graph Representation Learning (DeRe-GRL)\nmodel to effectively learn high-level ME features. DeRe-GRL includes two\nmodules: Action Decomposition Module (ADM) and Relation Reconstruction Module\n(RRM), where ADM learns action features of facial key components and RRM\nexplores the relationship between these action features. Based on facial key\ncomponents, ADM divides the geometric movement features extracted by the graph\nmodel-based backbone into several sub-features, and learns the map matrix to\nmap these sub-features into multiple action features; then, RRM learns weights\nto weight all action features to build the relationship between action\nfeatures. The experimental results demonstrate the effectiveness of the\nproposed modules, and the proposed method achieves competitive performance.\n", "date": "2023-03-03", "categories": []}
{"id": "2303.01795", "title": "PAGE: A Position-Aware Graph-Based Model for Emotion Cause Entailment in\n  Conversation", "abstract": "  Conversational Causal Emotion Entailment (C2E2) is a task that aims at\nrecognizing the causes corresponding to a target emotion in a conversation. The\norder of utterances in the conversation affects the causal inference. However,\nmost current position encoding strategies ignore the order relation among\nutterances and speakers. To address the issue, we devise a novel position-aware\ngraph to encode the entire conversation, fully modeling causal relations among\nutterances. The comprehensive experiments show that our method consistently\nachieves state-of-the-art performance on two challenging test sets, proving the\neffectiveness of our model. Our source code is available on Github:\nhttps://github.com/XiaojieGu/PAGE.\n", "date": "2023-03-03", "categories": []}
{"id": "2211.07711", "title": "Multilevel Transformer For Multimodal Emotion Recognition", "abstract": "  Multimodal emotion recognition has attracted much attention recently. Fusing\nmultiple modalities effectively with limited labeled data is a challenging\ntask. Considering the success of pre-trained model and fine-grained nature of\nemotion expression, it is reasonable to take these two aspects into\nconsideration. Unlike previous methods that mainly focus on one aspect, we\nintroduce a novel multi-granularity framework, which combines fine-grained\nrepresentation with pre-trained utterance-level representation. Inspired by\nTransformer TTS, we propose a multilevel transformer model to perform\nfine-grained multimodal emotion recognition. Specifically, we explore different\nmethods to incorporate phoneme-level embedding with word-level embedding. To\nperform multi-granularity learning, we simply combine multilevel transformer\nmodel with Albert. Extensive experimental results show that both our multilevel\ntransformer model and multi-granularity model outperform previous\nstate-of-the-art approaches on IEMOCAP dataset with text transcripts and speech\nsignal.\n", "date": "2022-10-26", "categories": []}
{"id": "2303.00193", "title": "CLIPER: A Unified Vision-Language Framework for In-the-Wild Facial\n  Expression Recognition", "abstract": "  Facial expression recognition (FER) is an essential task for understanding\nhuman behaviors. As one of the most informative behaviors of humans, facial\nexpressions are often compound and variable, which is manifested by the fact\nthat different people may express the same expression in very different ways.\nHowever, most FER methods still use one-hot or soft labels as the supervision,\nwhich lack sufficient semantic descriptions of facial expressions and are less\ninterpretable. Recently, contrastive vision-language pre-training (VLP) models\n(e.g., CLIP) use text as supervision and have injected new vitality into\nvarious computer vision tasks, benefiting from the rich semantics in text.\nTherefore, in this work, we propose CLIPER, a unified framework for both static\nand dynamic facial Expression Recognition based on CLIP. Besides, we introduce\nmultiple expression text descriptors (METD) to learn fine-grained expression\nrepresentations that make CLIPER more interpretable. We conduct extensive\nexperiments on several popular FER benchmarks and achieve state-of-the-art\nperformance, which demonstrates the effectiveness of CLIPER.\n", "date": "2023-03-01", "categories": []}
{"id": "2302.14727", "title": "Automatically Classifying Emotions based on Text: A Comparative\n  Exploration of Different Datasets", "abstract": "  Emotion Classification based on text is a task with many applications which\nhas received growing interest in recent years. This paper presents a\npreliminary study with the goal to help researchers and practitioners gain\ninsight into relatively new datasets as well as emotion classification in\ngeneral. We focus on three datasets that were recently presented in the related\nliterature, and we explore the performance of traditional as well as\nstate-of-the-art deep learning models in the presence of different\ncharacteristics in the data. We also explore the use of data augmentation in\norder to improve performance. Our experimental work shows that state-of-the-art\nmodels such as RoBERTa perform the best for all cases. We also provide\nobservations and discussion that highlight the complexity of emotion\nclassification in these datasets and test out the applicability of the models\nto actual social media posts we collected and labeled.\n", "date": "2023-02-28", "categories": []}
{"id": "2210.12531", "title": "Why Do You Feel This Way? Summarizing Triggers of Emotions in Social\n  Media Posts", "abstract": "  Crises such as the COVID-19 pandemic continuously threaten our world and\nemotionally affect billions of people worldwide in distinct ways. Understanding\nthe triggers leading to people's emotions is of crucial importance. Social\nmedia posts can be a good source of such analysis, yet these texts tend to be\ncharged with multiple emotions, with triggers scattering across multiple\nsentences. This paper takes a novel angle, namely, emotion detection and\ntrigger summarization, aiming to both detect perceived emotions in text, and\nsummarize events and their appraisals that trigger each emotion. To support\nthis goal, we introduce CovidET (Emotions and their Triggers during Covid-19),\na dataset of ~1,900 English Reddit posts related to COVID-19, which contains\nmanual annotations of perceived emotions and abstractive summaries of their\ntriggers described in the post. We develop strong baselines to jointly detect\nemotions and summarize emotion triggers. Our analyses show that CovidET\npresents new challenges in emotion-specific summarization, as well as\nmulti-emotion detection in long social media posts.\n", "date": "2022-10-22", "categories": []}
{"id": "2302.13661", "title": "Using Auxiliary Tasks In Multimodal Fusion Of Wav2vec 2.0 And BERT For\n  Multimodal Emotion Recognition", "abstract": "  The lack of data and the difficulty of multimodal fusion have always been\nchallenges for multimodal emotion recognition (MER). In this paper, we propose\nto use pretrained models as upstream network, wav2vec 2.0 for audio modality\nand BERT for text modality, and finetune them in downstream task of MER to cope\nwith the lack of data. For the difficulty of multimodal fusion, we use a\nK-layer multi-head attention mechanism as a downstream fusion module. Starting\nfrom the MER task itself, we design two auxiliary tasks to alleviate the\ninsufficient fusion between modalities and guide the network to capture and\nalign emotion-related features. Compared to the previous state-of-the-art\nmodels, we achieve a better performance by 78.42% Weighted Accuracy (WA) and\n79.71% Unweighted Accuracy (UA) on the IEMOCAP dataset.\n", "date": "2023-02-27", "categories": []}
{"id": "2302.14021", "title": "Quantifying Valence and Arousal in Text with Multilingual Pre-trained\n  Transformers", "abstract": "  The analysis of emotions expressed in text has numerous applications. In\ncontrast to categorical analysis, focused on classifying emotions according to\na pre-defined set of common classes, dimensional approaches can offer a more\nnuanced way to distinguish between different emotions. Still, dimensional\nmethods have been less studied in the literature. Considering a valence-arousal\ndimensional space, this work assesses the use of pre-trained Transformers to\npredict these two dimensions on a continuous scale, with input texts from\nmultiple languages and domains. We specifically combined multiple annotated\ndatasets from previous studies, corresponding to either emotional lexica or\nshort text documents, and evaluated models of multiple sizes and trained under\ndifferent settings. Our results show that model size can have a significant\nimpact on the quality of predictions, and that by fine-tuning a large model we\ncan confidently predict valence and arousal in multiple languages. We make\navailable the code, models, and supporting data.\n", "date": "2023-02-27", "categories": []}
{"id": "2302.09817", "title": "Explainable Human-centered Traits from Head Motion and Facial Expression\n  Dynamics", "abstract": "  We explore the efficacy of multimodal behavioral cues for explainable\nprediction of personality and interview-specific traits. We utilize elementary\nhead-motion units named kinemes, atomic facial movements termed action units\nand speech features to estimate these human-centered traits. Empirical results\nconfirm that kinemes and action units enable discovery of multiple\ntrait-specific behaviors while also enabling explainability in support of the\npredictions. For fusing cues, we explore decision and feature-level fusion, and\nan additive attention-based fusion strategy which quantifies the relative\nimportance of the three modalities for trait prediction. Examining various\nlong-short term memory (LSTM) architectures for classification and regression\non the MIT Interview and First Impressions Candidate Screening (FICS) datasets,\nwe note that: (1) Multimodal approaches outperform unimodal counterparts; (2)\nEfficient trait predictions and plausible explanations are achieved with both\nunimodal and multimodal approaches, and (3) Following the thin-slice approach,\neffective trait prediction is achieved even from two-second behavioral\nsnippets.\n", "date": "2023-02-20", "categories": []}
{"id": "2210.09381", "title": "Learning Diversified Feature Representations for Facial Expression\n  Recognition in the Wild", "abstract": "  Diversity of the features extracted by deep neural networks is important for\nenhancing the model generalization ability and accordingly its performance in\ndifferent learning tasks. Facial expression recognition in the wild has\nattracted interest in recent years due to the challenges existing in this area\nfor extracting discriminative and informative features from occluded images in\nreal-world scenarios. In this paper, we propose a mechanism to diversify the\nfeatures extracted by CNN layers of state-of-the-art facial expression\nrecognition architectures for enhancing the model capacity in learning\ndiscriminative features. To evaluate the effectiveness of the proposed\napproach, we incorporate this mechanism in two state-of-the-art models to (i)\ndiversify local/global features in an attention-based model and (ii) diversify\nfeatures extracted by different learners in an ensemble-based model.\nExperimental results on three well-known facial expression recognition\nin-the-wild datasets, AffectNet, FER+, and RAF-DB, show the effectiveness of\nour method, achieving the state-of-the-art performance of 89.99% on RAF-DB,\n89.34% on FER+ and the competitive accuracy of 60.02% on AffectNet dataset.\n", "date": "2022-10-17", "categories": []}
{"id": "2302.09070", "title": "Empathetic AI for Empowering Resilience in Games", "abstract": "  Failure and resilience are important aspects of gameplay. This is especially\nimportant for serious and competitive games, where players need to adapt and\ncope with failure frequently. In such situations, emotion regulation -- the\nactive process of modulating ones' emotions to cope and adapt to challenging\nsituations -- becomes essential. It is one of the prominent aspects of human\nintelligence and promotes mental health and well-being. While there has been\nwork on developing artificial emotional regulation assistants to help users\ncope with emotion regulation in the field of Intelligent Tutoring systems,\nlittle is done to incorporate such systems or ideas into (serious) video games.\nIn this paper, we introduce a data-driven 6-phase approach to establish\nempathetic artificial intelligence (EAI), which operates on raw chat log data\nto detect key affective states, identify common sequences and emotion\nregulation strategies and generalizes these to make them applicable for\nintervention systems.\n", "date": "2023-02-16", "categories": []}
{"id": "2302.09742", "title": "Affect-Conditioned Image Generation", "abstract": "  In creativity support and computational co-creativity contexts, the task of\ndiscovering appropriate prompts for use with text-to-image generative models\nremains difficult. In many cases the creator wishes to evoke a certain\nimpression with the image, but the task of conferring that succinctly in a text\nprompt poses a challenge: affective language is nuanced, complex, and\nmodel-specific. In this work we introduce a method for generating images\nconditioned on desired affect, quantified using a psychometrically validated\nthree-component approach, that can be combined with conditioning on text\ndescriptions. We first train a neural network for estimating the affect content\nof text and images from semantic embeddings, and then demonstrate how this can\nbe used to exert control over a variety of generative models. We show examples\nof how affect modifies the outputs, provide quantitative and qualitative\nanalysis of its capabilities, and discuss possible extensions and use cases.\n", "date": "2023-02-20", "categories": []}
{"id": "2302.08650", "title": "Gaussian-smoothed Imbalance Data Improves Speech Emotion Recognition", "abstract": "  In speech emotion recognition tasks, models learn emotional representations\nfrom datasets. We find the data distribution in the IEMOCAP dataset is very\nimbalanced, which may harm models to learn a better representation. To address\nthis issue, we propose a novel Pairwise-emotion Data Distribution Smoothing\n(PDDS) method. PDDS considers that the distribution of emotional data should be\nsmooth in reality, then applies Gaussian smoothing to emotion-pairs for\nconstructing a new training set with a smoother distribution. The required new\ndata are complemented using the mixup augmentation. As PDDS is model and\nmodality agnostic, it is evaluated with three SOTA models on the IEMOCAP\ndataset. The experimental results show that these models are improved by 0.2\\%\n- 4.8\\% and 1.5\\% - 5.9\\% in terms of WA and UA. In addition, an ablation study\ndemonstrates that the key advantage of PDDS is the reasonable data distribution\nrather than a simple data augmentation.\n", "date": "2023-02-17", "categories": []}
{"id": "2211.09496", "title": "EmoDiff: Intensity Controllable Emotional Text-to-Speech with Soft-Label\n  Guidance", "abstract": "  Although current neural text-to-speech (TTS) models are able to generate\nhigh-quality speech, intensity controllable emotional TTS is still a\nchallenging task. Most existing methods need external optimizations for\nintensity calculation, leading to suboptimal results or degraded quality. In\nthis paper, we propose EmoDiff, a diffusion-based TTS model where emotion\nintensity can be manipulated by a proposed soft-label guidance technique\nderived from classifier guidance. Specifically, instead of being guided with a\none-hot vector for the specified emotion, EmoDiff is guided with a soft label\nwhere the value of the specified emotion and \\textit{Neutral} is set to\n$\\alpha$ and $1-\\alpha$ respectively. The $\\alpha$ here represents the emotion\nintensity and can be chosen from 0 to 1. Our experiments show that EmoDiff can\nprecisely control the emotion intensity while maintaining high voice quality.\nMoreover, diverse speech with specified emotion intensity can be generated by\nsampling in the reverse denoising process.\n", "date": "2022-11-17", "categories": []}
{"id": "2302.08326", "title": "NUAA-QMUL-AIIT at Memotion 3: Multi-modal Fusion with\n  Squeeze-and-Excitation for Internet Meme Emotion Analysis", "abstract": "  This paper describes the participation of our NUAA-QMUL-AIIT team in the\nMemotion 3 shared task on meme emotion analysis. We propose a novel multi-modal\nfusion method, Squeeze-and-Excitation Fusion (SEFusion), and embed it into our\nsystem for emotion classification in memes. SEFusion is a simple fusion method\nthat employs fully connected layers, reshaping, and matrix multiplication.\nSEFusion learns a weight for each modality and then applies it to its own\nmodality feature. We evaluate the performance of our system on the three\nMemotion 3 sub-tasks. Among all participating systems in this Memotion 3 shared\ntask, our system ranked first on task A, fifth on task B, and second on task C.\nOur proposed SEFusion provides the flexibility to fuse any features from\ndifferent modalities. The source code for our method is published on\nhttps://github.com/xxxxxxxxy/memotion3-SEFusion.\n", "date": "2023-02-16", "categories": []}
{"id": "2302.08343", "title": "Cluster-based Deep Ensemble Learning for Emotion Classification in\n  Internet Memes", "abstract": "  Memes have gained popularity as a means to share visual ideas through the\nInternet and social media by mixing text, images and videos, often for humorous\npurposes. Research enabling automated analysis of memes has gained attention in\nrecent years, including among others the task of classifying the emotion\nexpressed in memes. In this paper, we propose a novel model, cluster-based deep\nensemble learning (CDEL), for emotion classification in memes. CDEL is a hybrid\nmodel that leverages the benefits of a deep learning model in combination with\na clustering algorithm, which enhances the model with additional information\nafter clustering memes with similar facial features. We evaluate the\nperformance of CDEL on a benchmark dataset for emotion classification, proving\nits effectiveness by outperforming a wide range of baseline models and\nachieving state-of-the-art performance. Further evaluation through ablated\nmodels demonstrates the effectiveness of the different components of CDEL.\n", "date": "2023-02-16", "categories": []}
{"id": "2302.06078", "title": "NYCU-TWO at Memotion 3: Good Foundation, Good Teacher, then you have\n  Good Meme Analysis", "abstract": "  This paper presents a robust solution to the Memotion 3.0 Shared Task. The\ngoal of this task is to classify the emotion and the corresponding intensity\nexpressed by memes, which are usually in the form of images with short captions\non social media. Understanding the multi-modal features of the given memes will\nbe the key to solving the task. In this work, we use CLIP to extract aligned\nimage-text features and propose a novel meme sentiment analysis framework,\nconsisting of a Cooperative Teaching Model (CTM) for Task A and a Cascaded\nEmotion Classifier (CEC) for Tasks B&C. CTM is based on the idea of knowledge\ndistillation, and can better predict the sentiment of a given meme in Task A;\nCEC can leverage the emotion intensity suggestion from the prediction of Task C\nto classify the emotion more precisely in Task B. Experiments show that we\nachieved the 2nd place ranking for both Task A and Task B and the 4th place\nranking for Task C, with weighted F1-scores of 0.342, 0.784, and 0.535\nrespectively. The results show the robustness and effectiveness of our\nframework. Our code is released at github.\n", "date": "2023-02-13", "categories": []}
{"id": "2301.12149", "title": "POSTER++: A simpler and stronger facial expression recognition network", "abstract": "  Facial expression recognition (FER) plays an important role in a variety of\nreal-world applications such as human-computer interaction. POSTER achieves the\nstate-of-the-art (SOTA) performance in FER by effectively combining facial\nlandmark and image features through two-stream pyramid cross-fusion design.\nHowever, the architecture of POSTER is undoubtedly complex. It causes expensive\ncomputational costs. In order to relieve the computational pressure of POSTER,\nin this paper, we propose POSTER++. It improves POSTER in three directions:\ncross-fusion, two-stream, and multi-scale feature extraction. In cross-fusion,\nwe use window-based cross-attention mechanism replacing vanilla cross-attention\nmechanism. We remove the image-to-landmark branch in the two-stream design. For\nmulti-scale feature extraction, POSTER++ combines images with landmark's\nmulti-scale features to replace POSTER's pyramid design. Extensive experiments\non several standard datasets show that our POSTER++ achieves the SOTA FER\nperformance with the minimum computational cost. For example, POSTER++ reached\n92.21% on RAF-DB, 67.49% on AffectNet (7 cls) and 63.77% on AffectNet (8 cls),\nrespectively, using only 8.4G floating point operations (FLOPs) and 43.7M\nparameters (Param). This demonstrates the effectiveness of our improvements.\n", "date": "2023-01-28", "categories": []}
{"id": "2302.06029", "title": "Emotion Detection in Unfix-length-Context Conversation", "abstract": "  We leverage different context windows when predicting the emotion of\ndifferent utterances. New modules are included to realize variable-length\ncontext: 1) two speaker-aware units, which explicitly model inner- and\ninter-speaker dependencies to form distilled conversational context, and 2) a\ntop-k normalization layer, which determines the most proper context windows\nfrom the conversational context to predict emotion. Experiments and ablation\nstudies show that our approach outperforms several strong baselines on three\npublic datasets.\n", "date": "2023-02-13", "categories": []}
{"id": "2210.07206", "title": "Best Practices in the Creation and Use of Emotion Lexicons", "abstract": "  Words play a central role in how we express ourselves. Lexicons of\nword-emotion associations are widely used in research and real-world\napplications for sentiment analysis, tracking emotions associated with products\nand policies, studying health disorders, tracking emotional arcs of stories,\nand so on. However, inappropriate and incorrect use of these lexicons can lead\nto not just sub-optimal results, but also inferences that are directly harmful\nto people. This paper brings together ideas from Affective Computing and AI\nEthics to present, some of the practical and ethical considerations involved in\nthe creation and use of emotion lexicons -- best practices. The goal is to\nprovide a comprehensive set of relevant considerations, so that readers\n(especially those new to work with emotions) can find relevant information in\none place. We hope this work will facilitate more thoughtfulness when one is\ndeciding on what emotions to work on, how to create an emotion lexicon, how to\nuse an emotion lexicon, how to draw meaningful inferences, and how to judge\nsuccess.\n", "date": "2022-10-13", "categories": []}
{"id": "2302.03508", "title": "Cluster-Level Contrastive Learning for Emotion Recognition in\n  Conversations", "abstract": "  A key challenge for Emotion Recognition in Conversations (ERC) is to\ndistinguish semantically similar emotions. Some works utilise Supervised\nContrastive Learning (SCL) which uses categorical emotion labels as supervision\nsignals and contrasts in high-dimensional semantic space. However, categorical\nlabels fail to provide quantitative information between emotions. ERC is also\nnot equally dependent on all embedded features in the semantic space, which\nmakes the high-dimensional SCL inefficient. To address these issues, we propose\na novel low-dimensional Supervised Cluster-level Contrastive Learning (SCCL)\nmethod, which first reduces the high-dimensional SCL space to a\nthree-dimensional affect representation space Valence-Arousal-Dominance (VAD),\nthen performs cluster-level contrastive learning to incorporate measurable\nemotion prototypes. To help modelling the dialogue and enriching the context,\nwe leverage the pre-trained knowledge adapters to infuse linguistic and factual\nknowledge. Experiments show that our method achieves new state-of-the-art\nresults with 69.81% on IEMOCAP, 65.7% on MELD, and 62.51% on DailyDialog\ndatasets. The analysis also proves that the VAD space is not only suitable for\nERC but also interpretable, with VAD prototypes enhancing its performance and\nstabilising the training of SCCL. In addition, the pre-trained knowledge\nadapters benefit the performance of the utterance encoder and SCCL. Our code is\navailable at: https://github.com/SteveKGYang/SCCL\n", "date": "2023-02-07", "categories": []}
{"id": "2302.04108", "title": "Triplet Loss-less Center Loss Sampling Strategies in Facial Expression\n  Recognition Scenarios", "abstract": "  Facial expressions convey massive information and play a crucial role in\nemotional expression. Deep neural network (DNN) accompanied by deep metric\nlearning (DML) techniques boost the discriminative ability of the model in\nfacial expression recognition (FER) applications. DNN, equipped with only\nclassification loss functions such as Cross-Entropy cannot compact intra-class\nfeature variation or separate inter-class feature distance as well as when it\ngets fortified by a DML supporting loss item. The triplet center loss (TCL)\nfunction is applied on all dimensions of the sample's embedding in the\nembedding space. In our work, we developed three strategies: fully-synthesized,\nsemi-synthesized, and prediction-based negative sample selection strategies. To\nachieve better results, we introduce a selective attention module that provides\na combination of pixel-wise and element-wise attention coefficients using\nhigh-semantic deep features of input samples. We evaluated the proposed method\non the RAF-DB, a highly imbalanced dataset. The experimental results reveal\nsignificant improvements in comparison to the baseline for all three negative\nsample selection strategies.\n", "date": "2023-02-08", "categories": []}
{"id": "2302.02419", "title": "deep learning of segment-level feature representation for speech emotion\n  recognition in conversations", "abstract": "  Accurately detecting emotions in conversation is a necessary yet challenging\ntask due to the complexity of emotions and dynamics in dialogues. The emotional\nstate of a speaker can be influenced by many different factors, such as\ninterlocutor stimulus, dialogue scene, and topic. In this work, we propose a\nconversational speech emotion recognition method to deal with capturing\nattentive contextual dependency and speaker-sensitive interactions. First, we\nuse a pretrained VGGish model to extract segment-based audio representation in\nindividual utterances. Second, an attentive bi-directional gated recurrent unit\n(GRU) models contextual-sensitive information and explores intra- and\ninter-speaker dependencies jointly in a dynamic manner. The experiments\nconducted on the standard conversational dataset MELD demonstrate the\neffectiveness of the proposed method when compared against state-of the-art\nmethods.\n", "date": "2023-02-05", "categories": []}
{"id": "2302.02499", "title": "Handwriting and Drawing for Depression Detection: A Preliminary Study", "abstract": "  The events of the past 2 years related to the pandemic have shown that it is\nincreasingly important to find new tools to help mental health experts in\ndiagnosing mood disorders. Leaving aside the longcovid cognitive (e.g.,\ndifficulty in concentration) and bodily (e.g., loss of smell) effects, the\nshort-term covid effects on mental health were a significant increase in\nanxiety and depressive symptoms. The aim of this study is to use a new tool,\nthe online handwriting and drawing analysis, to discriminate between healthy\nindividuals and depressed patients. To this aim, patients with clinical\ndepression (n = 14), individuals with high sub-clinical (diagnosed by a test\nrather than a doctor) depressive traits (n = 15) and healthy individuals (n =\n20) were recruited and asked to perform four online drawing /handwriting tasks\nusing a digitizing tablet and a special writing device. From the raw collected\nonline data, seventeen drawing/writing features (categorized into five\ncategories) were extracted, and compared among the three groups of the involved\nparticipants, through ANOVA repeated measures analyses. Results shows that Time\nfeatures are more effective in discriminating between healthy and participants\nwith sub-clinical depressive characteristics. On the other hand, Ductus and\nPressure features are more effective in discriminating between clinical\ndepressed and healthy participants.\n", "date": "2023-02-05", "categories": []}
{"id": "2302.02821", "title": "A Valid Self-Report is Never Late, Nor is it Early: On Considering the\n  \"Right\" Temporal Distance for Assessing Emotional Experience", "abstract": "  Developing computational models for automatic affect prediction requires\nvalid self-reports about individuals' emotional interpretations of stimuli. In\nthis article, we highlight the important influence of the temporal distance\nbetween a stimulus event and the moment when its experience is reported on the\nprovided information's validity. This influence stems from the time-dependent\nand time-demanding nature of the involved cognitive processes. As such, reports\ncan be collected too late: forgetting is a widely acknowledged challenge for\naccurate descriptions of past experience. For this reason, methods striving for\nassessment as early as possible have become increasingly popular. However, here\nwe argue that collection may also occur too early: descriptions about very\nrecent stimuli might be collected before emotional processing has fully\nconverged. Based on these notions, we champion the existence of a temporal\ndistance for each type of stimulus that maximizes the validity of self-reports\n-- a \"right\" time. Consequently, we recommend future research to (1)\nconsciously consider the potential influence of temporal distance on affective\nself-reports when planning data collection, (2) document the temporal distance\nof affective self-reports wherever possible as part of corpora for\ncomputational modelling, and finally (3) and explore the effect of temporal\ndistance on self-reports across different types of stimuli.\n", "date": "2023-01-27", "categories": []}
{"id": "2302.01555", "title": "Bridging the Emotional Semantic Gap via Multimodal Relevance Estimation", "abstract": "  Human beings have rich ways of emotional expressions, including facial\naction, voice, and natural languages. Due to the diversity and complexity of\ndifferent individuals, the emotions expressed by various modalities may be\nsemantically irrelevant. Directly fusing information from different modalities\nmay inevitably make the model subject to the noise from semantically irrelevant\nmodalities. To tackle this problem, we propose a multimodal relevance\nestimation network to capture the relevant semantics among modalities in\nmultimodal emotions. Specifically, we take advantage of an attention mechanism\nto reflect the semantic relevance weights of each modality. Moreover, we\npropose a relevant semantic estimation loss to weakly supervise the semantics\nof each modality. Furthermore, we make use of contrastive learning to optimize\nthe similarity of category-level modality-relevant semantics across different\nmodalities in feature space, thereby bridging the semantic gap between\nheterogeneous modalities. In order to better reflect the emotional state in the\nreal interactive scenarios and perform the semantic relevance analysis, we\ncollect a single-label discrete multimodal emotion dataset named SDME, which\nenables researchers to conduct multimodal semantic relevance research with\nlarge category bias. Experiments on continuous and discrete emotion datasets\nshow that our model can effectively capture the relevant semantics, especially\nfor the large deviations in modal semantics. The code and SDME dataset will be\npublicly available.\n", "date": "2023-02-03", "categories": []}
{"id": "2302.00606", "title": "The RW3D: A multi-modal panel dataset to understand the psychological\n  impact of the pandemic", "abstract": "  Besides far-reaching public health consequences, the COVID-19 pandemic had a\nsignificant psychological impact on people around the world. To gain further\ninsight into this matter, we introduce the Real World Worry Waves Dataset\n(RW3D). The dataset combines rich open-ended free-text responses with survey\ndata on emotions, significant life events, and psychological stressors in a\nrepeated-measures design in the UK over three years (2020: n=2441, 2021: n=1716\nand 2022: n=1152). This paper provides background information on the data\ncollection procedure, the recorded variables, participants' demographics, and\nhigher-order psychological and text-based derived variables that emerged from\nthe data. The RW3D is a unique primary data resource that could inspire new\nresearch questions on the psychological impact of the pandemic, especially\nthose that connect modalities (here: text data, psychological survey variables\nand demographics) over time.\n", "date": "2023-02-01", "categories": []}
{"id": "2210.15824", "title": "Improving the Modality Representation with Multi-View Contrastive\n  Learning for Multimodal Sentiment Analysis", "abstract": "  Modality representation learning is an important problem for multimodal\nsentiment analysis (MSA), since the highly distinguishable representations can\ncontribute to improving the analysis effect. Previous works of MSA have usually\nfocused on multimodal fusion strategies, and the deep study of modal\nrepresentation learning was given less attention. Recently, contrastive\nlearning has been confirmed effective at endowing the learned representation\nwith stronger discriminate ability. Inspired by this, we explore the\nimprovement approaches of modality representation with contrastive learning in\nthis study. To this end, we devise a three-stages framework with multi-view\ncontrastive learning to refine representations for the specific objectives. At\nthe first stage, for the improvement of unimodal representations, we employ the\nsupervised contrastive learning to pull samples within the same class together\nwhile the other samples are pushed apart. At the second stage, a\nself-supervised contrastive learning is designed for the improvement of the\ndistilled unimodal representations after cross-modal interaction. At last, we\nleverage again the supervised contrastive learning to enhance the fused\nmultimodal representation. After all the contrast trainings, we next achieve\nthe classification task based on frozen representations. We conduct experiments\non three open datasets, and results show the advance of our model.\n", "date": "2022-10-28", "categories": []}
{"id": "2301.12051", "title": "Predicting Students' Exam Scores Using Physiological Signals", "abstract": "  While acute stress has been shown to have both positive and negative effects\non performance, not much is known about the impacts of stress on students\ngrades during examinations. To answer this question, we examined whether a\ncorrelation could be found between physiological stress signals and exam\nperformance. We conducted this study using multiple physiological signals of\nten undergraduate students over three different exams. The study focused on\nthree signals, i.e., skin temperature, heart rate, and electrodermal activity.\nWe extracted statistics as features and fed them into a variety of binary\nclassifiers to predict relatively higher or lower grades. Experimental results\nshowed up to 0.81 ROC-AUC with k-nearest neighbor algorithm among various\nmachine learning algorithms.\n", "date": "2023-01-28", "categories": []}
{"id": "2301.10939", "title": "Affective Faces for Goal-Driven Dyadic Communication", "abstract": "  We introduce a video framework for modeling the association between verbal\nand non-verbal communication during dyadic conversation. Given the input speech\nof a speaker, our approach retrieves a video of a listener, who has facial\nexpressions that would be socially appropriate given the context. Our approach\nfurther allows the listener to be conditioned on their own goals,\npersonalities, or backgrounds. Our approach models conversations through a\ncomposition of large language models and vision-language models, creating\ninternal representations that are interpretable and controllable. To study\nmultimodal communication, we propose a new video dataset of unscripted\nconversations covering diverse topics and demographics. Experiments and\nvisualizations show our approach is able to output listeners that are\nsignificantly more socially appropriate than baselines. However, many\nchallenges remain, and we release our dataset publicly to spur further\nprogress. See our website for video results, data, and code:\nhttps://realtalk.cs.columbia.edu.\n", "date": "2023-01-26", "categories": []}
{"id": "2301.07650", "title": "Facial Thermal and Blood Perfusion Patterns of Human Emotions:\n  Proof-of-Concept", "abstract": "  In this work, a preliminary study of proof-of-concept was conducted to\nevaluate the performance of the thermographic and blood perfusion data when\nemotions of positive and negative valence are applied, where the blood\nperfusion data are obtained from the thermographic data. The images were\nobtained for baseline, positive, and negative valence according to the protocol\nof the Geneva Affective Picture Database. Absolute and percentage differences\nof average values of the data between the valences and the baseline were\ncalculated for different regions of interest (forehead, periorbital eyes,\ncheeks, nose and upper lips). For negative valence, a decrease in temperature\nand blood perfusion was observed in the regions of interest, and the effect was\ngreater on the left side than on the right side. In positive valence, the\ntemperature and blood perfusion increased in some cases, showing a complex\npattern. The temperature and perfusion of the nose was reduced for both\nvalences, which is indicative of the arousal dimension. The blood perfusion\nimages were found to be greater contrast; the percentage differences in the\nblood perfusion images are greater than those obtained in thermographic images.\nMoreover, the blood perfusion images, and vasomotor answer are consistent,\ntherefore, they can be a better biomarker than thermographic analysis in\nidentifying emotions.\n", "date": "2023-01-18", "categories": []}
{"id": "2301.09174", "title": "MATT: Multimodal Attention Level Estimation for e-learning Platforms", "abstract": "  This work presents a new multimodal system for remote attention level\nestimation based on multimodal face analysis. Our multimodal approach uses\ndifferent parameters and signals obtained from the behavior and physiological\nprocesses that have been related to modeling cognitive load such as faces\ngestures (e.g., blink rate, facial actions units) and user actions (e.g., head\npose, distance to the camera). The multimodal system uses the following modules\nbased on Convolutional Neural Networks (CNNs): Eye blink detection, head pose\nestimation, facial landmark detection, and facial expression features. First,\nwe individually evaluate the proposed modules in the task of estimating the\nstudent's attention level captured during online e-learning sessions. For that\nwe trained binary classifiers (high or low attention) based on Support Vector\nMachines (SVM) for each module. Secondly, we find out to what extent multimodal\nscore level fusion improves the attention level estimation. The mEBAL database\nis used in the experimental framework, a public multi-modal database for\nattention level estimation obtained in an e-learning environment that contains\ndata from 38 users while conducting several e-learning tasks of variable\ndifficulty (creating changes in student cognitive loads).\n", "date": "2023-01-22", "categories": []}
{"id": "2301.06745", "title": "BERT-ERC: Fine-tuning BERT is Enough for Emotion Recognition in\n  Conversation", "abstract": "  Previous works on emotion recognition in conversation (ERC) follow a two-step\nparadigm, which can be summarized as first producing context-independent\nfeatures via fine-tuning pretrained language models (PLMs) and then analyzing\ncontextual information and dialogue structure information among the extracted\nfeatures. However, we discover that this paradigm has several limitations.\nAccordingly, we propose a novel paradigm, i.e., exploring contextual\ninformation and dialogue structure information in the fine-tuning step, and\nadapting the PLM to the ERC task in terms of input text, classification\nstructure, and training strategy. Furthermore, we develop our model BERT-ERC\naccording to the proposed paradigm, which improves ERC performance in three\naspects, namely suggestive text, fine-grained classification module, and\ntwo-stage training. Compared to existing methods, BERT-ERC achieves substantial\nimprovement on four datasets, indicating its effectiveness and generalization\ncapability. Besides, we also set up the limited resources scenario and the\nonline prediction scenario to approximate real-world scenarios. Extensive\nexperiments demonstrate that the proposed paradigm significantly outperforms\nthe previous one and can be adapted to various scenes.\n", "date": "2023-01-17", "categories": []}
{"id": "2301.06890", "title": "AI-Based Affective Music Generation Systems: A Review of Methods, and\n  Challenges", "abstract": "  Music is a powerful medium for altering the emotional state of the listener.\nIn recent years, with significant advancement in computing capabilities,\nartificial intelligence-based (AI-based) approaches have become popular for\ncreating affective music generation (AMG) systems that are empowered with the\nability to generate affective music. Entertainment, healthcare, and\nsensor-integrated interactive system design are a few of the areas in which\nAI-based affective music generation (AI-AMG) systems may have a significant\nimpact. Given the surge of interest in this topic, this article aims to provide\na comprehensive review of AI-AMG systems. The main building blocks of an AI-AMG\nsystem are discussed, and existing systems are formally categorized based on\nthe core algorithm used for music generation. In addition, this article\ndiscusses the main musical features employed to compose affective music, along\nwith the respective AI-based approaches used for tailoring them. Lastly, the\nmain challenges and open questions in this field, as well as their potential\nsolutions, are presented to guide future research. We hope that this review\nwill be useful for readers seeking to understand the state-of-the-art in AI-AMG\nsystems, and gain an overview of the methods used for developing them, thereby\nhelping them explore this field in the future.\n", "date": "2023-01-10", "categories": []}
{"id": "2208.11868", "title": "Interpretable Multimodal Emotion Recognition using Hybrid Fusion of\n  Speech and Image Data", "abstract": "  This paper proposes a multimodal emotion recognition system based on hybrid\nfusion that classifies the emotions depicted by speech utterances and\ncorresponding images into discrete classes. A new interpretability technique\nhas been developed to identify the important speech & image features leading to\nthe prediction of particular emotion classes. The proposed system's\narchitecture has been determined through intensive ablation studies. It fuses\nthe speech & image features and then combines speech, image, and intermediate\nfusion outputs. The proposed interpretability technique incorporates the divide\n& conquer approach to compute shapely values denoting each speech & image\nfeature's importance. We have also constructed a large-scale dataset (IIT-R\nSIER dataset), consisting of speech utterances, corresponding images, and class\nlabels, i.e., 'anger,' 'happy,' 'hate,' and 'sad.' The proposed system has\nachieved 83.29% accuracy for emotion recognition. The enhanced performance of\nthe proposed system advocates the importance of utilizing complementary\ninformation from multiple modalities for emotion recognition.\n", "date": "2022-08-25", "categories": []}
{"id": "2301.01982", "title": "Emotion-Cause Pair Extraction as Question Answering", "abstract": "  The task of Emotion-Cause Pair Extraction (ECPE) aims to extract all\npotential emotion-cause pairs of a document without any annotation of emotion\nor cause clauses. Previous approaches on ECPE have tried to improve\nconventional two-step processing schemes by using complex architectures for\nmodeling emotion-cause interaction. In this paper, we cast the ECPE task to the\nquestion answering (QA) problem and propose simple yet effective BERT-based\nsolutions to tackle it. Given a document, our Guided-QA model first predicts\nthe best emotion clause using a fixed question. Then the predicted emotion is\nused as a question to predict the most potential cause for the emotion. We\nevaluate our model on a standard ECPE corpus. The experimental results show\nthat despite its simplicity, our Guided-QA achieves promising results and is\neasy to reproduce. The code of Guided-QA is also provided.\n", "date": "2023-01-05", "categories": []}
{"id": "2211.13564", "title": "More comprehensive facial inversion for more effective expression\n  recognition", "abstract": "  Facial expression recognition (FER) plays a significant role in the\nubiquitous application of computer vision. We revisit this problem with a new\nperspective on whether it can acquire useful representations that improve FER\nperformance in the image generation process, and propose a novel generative\nmethod based on the image inversion mechanism for the FER task, termed\nInversion FER (IFER). Particularly, we devise a novel Adversarial Style\nInversion Transformer (ASIT) towards IFER to comprehensively extract features\nof generated facial images. In addition, ASIT is equipped with an image\ninversion discriminator that measures the cosine similarity of semantic\nfeatures between source and generated images, constrained by a distribution\nalignment loss. Finally, we introduce a feature modulation module to fuse the\nstructural code and latent codes from ASIT for the subsequent FER work. We\nextensively evaluate ASIT on facial datasets such as FFHQ and CelebA-HQ,\nshowing that our approach achieves state-of-the-art facial inversion\nperformance. IFER also achieves competitive results in facial expression\nrecognition datasets such as RAF-DB, SFEW and AffectNet. The code and models\nare available at https://github.com/Talented-Q/IFER-master.\n", "date": "2022-11-24", "categories": []}
{"id": "2301.00142", "title": "Computational Charisma -- A Brick by Brick Blueprint for Building\n  Charismatic Artificial Intelligence", "abstract": "  Charisma is considered as one's ability to attract and potentially also\ninfluence others. Clearly, there can be considerable interest from an\nartificial intelligence's (AI) perspective to provide it with such skill.\nBeyond, a plethora of use cases opens up for computational measurement of human\ncharisma, such as for tutoring humans in the acquisition of charisma, mediating\nhuman-to-human conversation, or identifying charismatic individuals in big\nsocial data. A number of models exist that base charisma on various dimensions,\noften following the idea that charisma is given if someone could and would help\nothers. Examples include influence (could help) and affability (would help) in\nscientific studies or power (could help), presence, and warmth (both would\nhelp) as a popular concept. Modelling high levels in these dimensions for\nhumanoid robots or virtual agents, seems accomplishable. Beyond, also automatic\nmeasurement appears quite feasible with the recent advances in the related\nfields of Affective Computing and Social Signal Processing. Here, we,\nthereforem present a blueprint for building machines that can appear\ncharismatic, but also analyse the charisma of others. To this end, we first\nprovide the psychological perspective including different models of charisma\nand behavioural cues of it. We then switch to conversational charisma in spoken\nlanguage as an exemplary modality that is essential for human-human and\nhuman-computer conversations. The computational perspective then deals with the\nrecognition and generation of charismatic behaviour by AI. This includes an\noverview of the state of play in the field and the aforementioned blueprint. We\nthen name exemplary use cases of computational charismatic skills before\nswitching to ethical aspects and concluding this overview and perspective on\nbuilding charisma-enabled AI.\n", "date": "2022-12-31", "categories": []}
{"id": "2301.00254", "title": "Depression Diagnosis and Analysis via Multimodal Multi-order Factor\n  Fusion", "abstract": "  Depression is a leading cause of death worldwide, and the diagnosis of\ndepression is nontrivial. Multimodal learning is a popular solution for\nautomatic diagnosis of depression, and the existing works suffer two main\ndrawbacks: 1) the high-order interactions between different modalities can not\nbe well exploited; and 2) interpretability of the models are weak. To remedy\nthese drawbacks, we propose a multimodal multi-order factor fusion (MMFF)\nmethod. Our method can well exploit the high-order interactions between\ndifferent modalities by extracting and assembling modality factors under the\nguide of a shared latent proxy. We conduct extensive experiments on two recent\nand popular datasets, E-DAIC-WOZ and CMDC, and the results show that our method\nachieve significantly better performance compared with other existing\napproaches. Besides, by analyzing the process of factor assembly, our model can\nintuitively show the contribution of each factor. This helps us understand the\nfusion mechanism.\n", "date": "2022-12-31", "categories": []}
{"id": "2212.13885", "title": "Emotion Recognition with Pre-Trained Transformers Using Multimodal\n  Signals", "abstract": "  In this paper, we address the problem of multimodal emotion recognition from\nmultiple physiological signals. We demonstrate that a Transformer-based\napproach is suitable for this task. In addition, we present how such models may\nbe pretrained in a multimodal scenario to improve emotion recognition\nperformances. We evaluate the benefits of using multimodal inputs and\npre-training with our approach on a state-ofthe-art dataset.\n", "date": "2022-12-22", "categories": []}
{"id": "2212.13917", "title": "Multimodal Emotion Recognition among Couples from Lab Settings to Daily\n  Life using Smartwatches", "abstract": "  Couples generally manage chronic diseases together and the management takes\nan emotional toll on both patients and their romantic partners. Consequently,\nrecognizing the emotions of each partner in daily life could provide an insight\ninto their emotional well-being in chronic disease management. The emotions of\npartners are currently inferred in the lab and daily life using self-reports\nwhich are not practical for continuous emotion assessment or observer reports\nwhich are manual, time-intensive, and costly. Currently, there exists no\ncomprehensive overview of works on emotion recognition among couples.\nFurthermore, approaches for emotion recognition among couples have (1) focused\non English-speaking couples in the U.S., (2) used data collected from the lab,\nand (3) performed recognition using observer ratings rather than partner's\nself-reported / subjective emotions. In this body of work contained in this\nthesis (8 papers - 5 published and 3 currently under review in various\njournals), we fill the current literature gap on couples' emotion recognition,\ndevelop emotion recognition systems using 161 hours of data from a total of\n1,051 individuals, and make contributions towards taking couples' emotion\nrecognition from the lab which is the status quo, to daily life. This thesis\ncontributes toward building automated emotion recognition systems that would\neventually enable partners to monitor their emotions in daily life and enable\nthe delivery of interventions to improve their emotional well-being.\n", "date": "2022-12-21", "categories": []}
{"id": "2212.11772", "title": "A Self-Adjusting Fusion Representation Learning Model for Unaligned\n  Text-Audio Sequences", "abstract": "  Inter-modal interaction plays an indispensable role in multimodal sentiment\nanalysis. Due to different modalities sequences are usually non-alignment, how\nto integrate relevant information of each modality to learn fusion\nrepresentations has been one of the central challenges in multimodal learning.\nIn this paper, a Self-Adjusting Fusion Representation Learning Model (SA-FRLM)\nis proposed to learn robust crossmodal fusion representations directly from the\nunaligned text and audio sequences. Different from previous works, our model\nnot only makes full use of the interaction between different modalities but\nalso maximizes the protection of the unimodal characteristics. Specifically, we\nfirst employ a crossmodal alignment module to project different modalities\nfeatures to the same dimension. The crossmodal collaboration attention is then\nadopted to model the inter-modal interaction between text and audio sequences\nand initialize the fusion representations. After that, as the core unit of the\nSA-FRLM, the crossmodal adjustment transformer is proposed to protect original\nunimodal characteristics. It can dynamically adapt the fusion representations\nby using single modal streams. We evaluate our approach on the public\nmultimodal sentiment analysis datasets CMU-MOSI and CMU-MOSEI. The experiment\nresults show that our model has significantly improved the performance of all\nthe metrics on the unaligned text-audio sequences.\n", "date": "2022-11-12", "categories": []}
{"id": "2212.10030", "title": "InterMulti:Multi-view Multimodal Interactions with Text-dominated\n  Hierarchical High-order Fusion for Emotion Analysis", "abstract": "  Humans are sophisticated at reading interlocutors' emotions from multimodal\nsignals, such as speech contents, voice tones and facial expressions. However,\nmachines might struggle to understand various emotions due to the difficulty of\neffectively decoding emotions from the complex interactions between multimodal\nsignals. In this paper, we propose a multimodal emotion analysis framework,\nInterMulti, to capture complex multimodal interactions from different views and\nidentify emotions from multimodal signals. Our proposed framework decomposes\nsignals of different modalities into three kinds of multimodal interaction\nrepresentations, including a modality-full interaction representation, a\nmodality-shared interaction representation, and three modality-specific\ninteraction representations. Additionally, to balance the contribution of\ndifferent modalities and learn a more informative latent interaction\nrepresentation, we developed a novel Text-dominated Hierarchical High-order\nFusion(THHF) module. THHF module reasonably integrates the above three kinds of\nrepresentations into a comprehensive multimodal interaction representation.\nExtensive experimental results on widely used datasets, (i.e.) MOSEI, MOSI and\nIEMOCAP, demonstrate that our method outperforms the state-of-the-art.\n", "date": "2022-12-20", "categories": []}
{"id": "2212.08661", "title": "EffMulti: Efficiently Modeling Complex Multimodal Interactions for\n  Emotion Analysis", "abstract": "  Humans are skilled in reading the interlocutor's emotion from multimodal\nsignals, including spoken words, simultaneous speech, and facial expressions.\nIt is still a challenge to effectively decode emotions from the complex\ninteractions of multimodal signals. In this paper, we design three kinds of\nmultimodal latent representations to refine the emotion analysis process and\ncapture complex multimodal interactions from different views, including a\nintact three-modal integrating representation, a modality-shared\nrepresentation, and three modality-individual representations. Then, a\nmodality-semantic hierarchical fusion is proposed to reasonably incorporate\nthese representations into a comprehensive interaction representation. The\nexperimental results demonstrate that our EffMulti outperforms the\nstate-of-the-art methods. The compelling performance benefits from its\nwell-designed framework with ease of implementation, lower computing\ncomplexity, and less trainable parameters.\n", "date": "2022-12-16", "categories": []}
{"id": "2212.09465", "title": "Improving the Generalizability of Text-Based Emotion Detection by\n  Leveraging Transformers with Psycholinguistic Features", "abstract": "  In recent years, there has been increased interest in building predictive\nmodels that harness natural language processing and machine learning techniques\nto detect emotions from various text sources, including social media posts,\nmicro-blogs or news articles. Yet, deployment of such models in real-world\nsentiment and emotion applications faces challenges, in particular poor\nout-of-domain generalizability. This is likely due to domain-specific\ndifferences (e.g., topics, communicative goals, and annotation schemes) that\nmake transfer between different models of emotion recognition difficult. In\nthis work we propose approaches for text-based emotion detection that leverage\ntransformer models (BERT and RoBERTa) in combination with Bidirectional Long\nShort-Term Memory (BiLSTM) networks trained on a comprehensive set of\npsycholinguistic features. First, we evaluate the performance of our models\nwithin-domain on two benchmark datasets: GoEmotion and ISEAR. Second, we\nconduct transfer learning experiments on six datasets from the Unified Emotion\nDataset to evaluate their out-of-domain robustness. We find that the proposed\nhybrid models improve the ability to generalize to out-of-distribution data\ncompared to a standard transformer-based approach. Moreover, we observe that\nthese models perform competitively on in-domain data.\n", "date": "2022-12-19", "categories": []}
{"id": "2212.07751", "title": "Combating Uncertainty and Class Imbalance in Facial Expression\n  Recognition", "abstract": "  Recognition of facial expression is a challenge when it comes to computer\nvision. The primary reasons are class imbalance due to data collection and\nuncertainty due to inherent noise such as fuzzy facial expressions and\ninconsistent labels. However, current research has focused either on the\nproblem of class imbalance or on the problem of uncertainty, ignoring the\nintersection of how to address these two problems. Therefore, in this paper, we\npropose a framework based on Resnet and Attention to solve the above problems.\nWe design weight for each class. Through the penalty mechanism, our model will\npay more attention to the learning of small samples during training, and the\nresulting decrease in model accuracy can be improved by a Convolutional Block\nAttention Module (CBAM). Meanwhile, our backbone network will also learn an\nuncertain feature for each sample. By mixing uncertain features between\nsamples, the model can better learn those features that can be used for\nclassification, thus suppressing uncertainty. Experiments show that our method\nsurpasses most basic methods in terms of accuracy on facial expression data\nsets (e.g., AffectNet, RAF-DB), and it also solves the problem of class\nimbalance well.\n", "date": "2022-12-15", "categories": []}
{"id": "2212.06400", "title": "Improving Depression estimation from facial videos with face alignment,\n  training optimization and scheduling", "abstract": "  Deep learning models have shown promising results in recognizing depressive\nstates using video-based facial expressions. While successful models typically\nleverage using 3D-CNNs or video distillation techniques, the different use of\npretraining, data augmentation, preprocessing, and optimization techniques\nacross experiments makes it difficult to make fair architectural comparisons.\nWe propose instead to enhance two simple models based on ResNet-50 that use\nonly static spatial information by using two specific face alignment methods\nand improved data augmentation, optimization, and scheduling techniques. Our\nextensive experiments on benchmark datasets obtain similar results to\nsophisticated spatio-temporal models for single streams, while the score-level\nfusion of two different streams outperforms state-of-the-art methods. Our\nfindings suggest that specific modifications in the preprocessing and training\nprocess result in noticeable differences in the performance of the models and\ncould hide the actual originally attributed to the use of different neural\nnetwork architectures.\n", "date": "2022-12-13", "categories": []}
{"id": "2208.03506", "title": "Multi-Task Transformer with uncertainty modelling for Face Based\n  Affective Computing", "abstract": "  Face based affective computing consists in detecting emotions from face\nimages. It is useful to unlock better automatic comprehension of human\nbehaviours and could pave the way toward improved human-machines interactions.\nHowever it comes with the challenging task of designing a computational\nrepresentation of emotions. So far, emotions have been represented either\ncontinuously in the 2D Valence/Arousal space or in a discrete manner with\nEkman's 7 basic emotions. Alternatively, Ekman's Facial Action Unit (AU) system\nhave also been used to caracterize emotions using a codebook of unitary\nmuscular activations. ABAW3 and ABAW4 Multi-Task Challenges are the first work\nto provide a large scale database annotated with those three types of labels.\nIn this paper we present a transformer based multi-task method for jointly\nlearning to predict valence arousal, action units and basic emotions. From an\narchitectural standpoint our method uses a taskwise token approach to\nefficiently model the similarities between the tasks. From a learning point of\nview we use an uncertainty weighted loss for modelling the difference of\nstochasticity between the three tasks annotations.\n", "date": "2022-08-06", "categories": []}
{"id": "2212.05463", "title": "Vision Transformer with Attentive Pooling for Robust Facial Expression\n  Recognition", "abstract": "  Facial Expression Recognition (FER) in the wild is an extremely challenging\ntask. Recently, some Vision Transformers (ViT) have been explored for FER, but\nmost of them perform inferiorly compared to Convolutional Neural Networks\n(CNN). This is mainly because the new proposed modules are difficult to\nconverge well from scratch due to lacking inductive bias and easy to focus on\nthe occlusion and noisy areas. TransFER, a representative transformer-based\nmethod for FER, alleviates this with multi-branch attention dropping but brings\nexcessive computations. On the contrary, we present two attentive pooling (AP)\nmodules to pool noisy features directly. The AP modules include Attentive Patch\nPooling (APP) and Attentive Token Pooling (ATP). They aim to guide the model to\nemphasize the most discriminative features while reducing the impacts of less\nrelevant features. The proposed APP is employed to select the most informative\npatches on CNN features, and ATP discards unimportant tokens in ViT. Being\nsimple to implement and without learnable parameters, the APP and ATP\nintuitively reduce the computational cost while boosting the performance by\nONLY pursuing the most discriminative features. Qualitative results demonstrate\nthe motivations and effectiveness of our attentive poolings. Besides,\nquantitative results on six in-the-wild datasets outperform other\nstate-of-the-art methods.\n", "date": "2022-12-11", "categories": []}
{"id": "2212.06009", "title": "An Approach for Improving Automatic Mouth Emotion Recognition", "abstract": "  The study proposes and tests a technique for automated emotion recognition\nthrough mouth detection via Convolutional Neural Networks (CNN), meant to be\napplied for supporting people with health disorders with communication skills\nissues (e.g. muscle wasting, stroke, autism, or, more simply, pain) in order to\nrecognize emotions and generate real-time feedback, or data feeding supporting\nsystems. The software system starts the computation identifying if a face is\npresent on the acquired image, then it looks for the mouth location and\nextracts the corresponding features. Both tasks are carried out using Haar\nFeature-based Classifiers, which guarantee fast execution and promising\nperformance. If our previous works focused on visual micro-expressions for\npersonalized training on a single user, this strategy aims to train the system\nalso on generalized faces data sets.\n", "date": "2022-12-12", "categories": []}
{"id": "2212.02907", "title": "Emotion Conditioned Creative Dialog Generation", "abstract": "  We present a DialGPT based model for generating creative dialog responses\nthat are conditioned based on one of the following emotions: anger, disgust,\nfear, happiness, pain, sadness and surprise. Our model is capable of producing\na contextually apt response given an input sentence and a desired emotion\nlabel. Our model is capable of expressing the desired emotion with an accuracy\nof 0.6. The best performing emotions are neutral, fear and disgust. When\nmeasuring the strength of the expressed emotion, we find that anger, fear and\ndisgust are expressed in the most strong fashion by the model.\n", "date": "2022-12-06", "categories": []}
{"id": "2212.02995", "title": "Knowledge-Bridged Causal Interaction Network for Causal Emotion\n  Entailment", "abstract": "  Causal Emotion Entailment aims to identify causal utterances that are\nresponsible for the target utterance with a non-neutral emotion in\nconversations. Previous works are limited in thorough understanding of the\nconversational context and accurate reasoning of the emotion cause. To this\nend, we propose Knowledge-Bridged Causal Interaction Network (KBCIN) with\ncommonsense knowledge (CSK) leveraged as three bridges. Specifically, we\nconstruct a conversational graph for each conversation and leverage the\nevent-centered CSK as the semantics-level bridge (S-bridge) to capture the deep\ninter-utterance dependencies in the conversational context via the CSK-Enhanced\nGraph Attention module. Moreover, social-interaction CSK serves as\nemotion-level bridge (E-bridge) and action-level bridge (A-bridge) to connect\ncandidate utterances with the target one, which provides explicit causal clues\nfor the Emotional Interaction module and Actional Interaction module to reason\nthe target emotion. Experimental results show that our model achieves better\nperformance over most baseline models. Our source code is publicly available at\nhttps://github.com/circle-hit/KBCIN.\n", "date": "2022-12-06", "categories": []}
{"id": "2212.03102", "title": "A comparative study of emotion recognition methods using facial\n  expressions", "abstract": "  Understanding the facial expressions of our interlocutor is important to\nenrich the communication and to give it a depth that goes beyond the explicitly\nexpressed. In fact, studying one's facial expression gives insight into their\nhidden emotion state. However, even as humans, and despite our empathy and\nfamiliarity with the human emotional experience, we are only able to guess what\nthe other might be feeling. In the fields of artificial intelligence and\ncomputer vision, Facial Emotion Recognition (FER) is a topic that is still in\nfull growth mostly with the advancement of deep learning approaches and the\nimprovement of data collection. The main purpose of this paper is to compare\nthe performance of three state-of-the-art networks, each having their own\napproach to improve on FER tasks, on three FER datasets. The first and second\nsections respectively describe the three datasets and the three studied network\narchitectures designed for an FER task. The experimental protocol, the results\nand their interpretation are outlined in the remaining sections.\n", "date": "2022-12-05", "categories": []}
{"id": "2210.17233", "title": "CorrLoss: Integrating Co-Occurrence Domain Knowledge for Affect\n  Recognition", "abstract": "  Neural networks are widely adopted, yet the integration of domain knowledge\nis still underutilized. We propose to integrate domain knowledge about\nco-occurring facial movements as a constraint in the loss function to enhance\nthe training of neural networks for affect recognition. As the co-ccurrence\npatterns tend to be similar across datasets, applying our method can lead to a\nhigher generalizability of models and a lower risk of overfitting. We\ndemonstrate this by showing performance increases in cross-dataset testing for\nvarious datasets. We also show the applicability of our method for calibrating\nneural networks to different facial expressions.\n", "date": "2022-10-31", "categories": []}
{"id": "2212.01844", "title": "Pair-Based Joint Encoding with Relational Graph Convolutional Networks\n  for Emotion-Cause Pair Extraction", "abstract": "  Emotion-cause pair extraction (ECPE) aims to extract emotion clauses and\ncorresponding cause clauses, which have recently received growing attention.\nPrevious methods sequentially encode features with a specified order. They\nfirst encode the emotion and cause features for clause extraction and then\ncombine them for pair extraction. This lead to an imbalance in inter-task\nfeature interaction where features extracted later have no direct contact with\nthe former. To address this issue, we propose a novel Pair-Based Joint Encoding\n(PBJE) network, which generates pairs and clauses features simultaneously in a\njoint feature encoding manner to model the causal relationship in clauses. PBJE\ncan balance the information flow among emotion clauses, cause clauses and\npairs. From a multi-relational perspective, we construct a heterogeneous\nundirected graph and apply the Relational Graph Convolutional Network (RGCN) to\ncapture the various relationship between clauses and the relationship between\npairs and clauses. Experimental results show that PBJE achieves\nstate-of-the-art performance on the Chinese benchmark corpus.\n", "date": "2022-12-04", "categories": []}
{"id": "2211.15425", "title": "FAF: A novel multimodal emotion recognition approach integrating face,\n  body and text", "abstract": "  Multimodal emotion analysis performed better in emotion recognition depending\non more comprehensive emotional clues and multimodal emotion dataset. In this\npaper, we developed a large multimodal emotion dataset, named \"HED\" dataset, to\nfacilitate the emotion recognition task, and accordingly propose a multimodal\nemotion recognition method. To promote recognition accuracy, \"Feature After\nFeature\" framework was used to explore crucial emotional information from the\naligned face, body and text samples. We employ various benchmarks to evaluate\nthe \"HED\" dataset and compare the performance with our method. The results show\nthat the five classification accuracy of the proposed multimodal fusion method\nis about 83.75%, and the performance is improved by 1.83%, 9.38%, and 21.62%\nrespectively compared with that of individual modalities. The complementarity\nbetween each channel is effectively used to improve the performance of emotion\nrecognition. We had also established a multimodal online emotion prediction\nplatform, aiming to provide free emotion prediction to more users.\n", "date": "2022-11-20", "categories": []}
{"id": "2210.10262", "title": "Free energy model of emotional valence in dual-process perceptions", "abstract": "  An appropriate level of arousal induces positive emotions, and a high arousal\npotential may provoke negative emotions. To explain the effect of arousal on\nemotional valence, we propose a novel mathematical framework of arousal\npotential variations in the dual process of human cognition: automatic and\ncontrolled. A suitable mathematical formulation to explain the emotions in the\ndual process is still absent. Our model associates free energy with arousal\npotential and its variations to explain emotional valence. Decreasing and\nincreasing free energy consequently induce positive and negative emotions,\nrespectively. We formalize a transition from the automatic to the controlled\nprocess in the dual process as a change of Bayesian prior. Further, we model\nemotional valence using free energy increase (FI) when one tries changing one's\nBayesian prior and its reduction (FR) when one succeeds in recognizing the same\nstimuli with a changed prior and define three emotions: \"interest,\"\n\"confusion,\" and \"boredom\" using the variations. The results of our\nmathematical analysis comparing various Gaussian model parameters reveals the\nfollowing: 1) prediction error (PR) increases FR (representing \"interest\") when\nthe first prior variance is greater than the second prior variance, 2) PR\ndecreases FR when the first prior variance is less than the second prior\nvariance, and 3) the distance between priors' means always increases FR. We\nalso discuss the association of the outcomes with emotions in the controlled\nprocess. The proposed mathematical model provides a general framework for\npredicting and controlling emotional valence in the dual process that varies\nwith viewpoint and stimuli, as well as for understanding the contradictions in\nthe effects of arousal on the valence.\n", "date": "2022-10-19", "categories": []}
{"id": "2211.10780", "title": "ArtELingo: A Million Emotion Annotations of WikiArt with Emphasis on\n  Diversity over Language and Culture", "abstract": "  This paper introduces ArtELingo, a new benchmark and dataset, designed to\nencourage work on diversity across languages and cultures. Following ArtEmis, a\ncollection of 80k artworks from WikiArt with 0.45M emotion labels and\nEnglish-only captions, ArtELingo adds another 0.79M annotations in Arabic and\nChinese, plus 4.8K in Spanish to evaluate \"cultural-transfer\" performance. More\nthan 51K artworks have 5 annotations or more in 3 languages. This diversity\nmakes it possible to study similarities and differences across languages and\ncultures. Further, we investigate captioning tasks, and find diversity improves\nthe performance of baseline models. ArtELingo is publicly available at\nhttps://www.artelingo.org/ with standard splits and baseline models. We hope\nour work will help ease future research on multilinguality and culturally-aware\nAI.\n", "date": "2022-11-19", "categories": []}
{"id": "2211.11256", "title": "UniMSE: Towards Unified Multimodal Sentiment Analysis and Emotion\n  Recognition", "abstract": "  Multimodal sentiment analysis (MSA) and emotion recognition in conversation\n(ERC) are key research topics for computers to understand human behaviors. From\na psychological perspective, emotions are the expression of affect or feelings\nduring a short period, while sentiments are formed and held for a longer\nperiod. However, most existing works study sentiment and emotion separately and\ndo not fully exploit the complementary knowledge behind the two. In this paper,\nwe propose a multimodal sentiment knowledge-sharing framework (UniMSE) that\nunifies MSA and ERC tasks from features, labels, and models. We perform\nmodality fusion at the syntactic and semantic levels and introduce contrastive\nlearning between modalities and samples to better capture the difference and\nconsistency between sentiments and emotions. Experiments on four public\nbenchmark datasets, MOSI, MOSEI, MELD, and IEMOCAP, demonstrate the\neffectiveness of the proposed method and achieve consistent improvements\ncompared with state-of-the-art methods.\n", "date": "2022-11-21", "categories": []}
{"id": "2211.09956", "title": "A Persian ASR-based SER: Modification of Sharif Emotional Speech\n  Database and Investigation of Persian Text Corpora", "abstract": "  Speech Emotion Recognition (SER) is one of the essential perceptual methods\nof humans in understanding the situation and how to interact with others,\ntherefore, in recent years, it has been tried to add the ability to recognize\nemotions to human-machine communication systems. Since the SER process relies\non labeled data, databases are essential for it. Incomplete, low-quality or\ndefective data may lead to inaccurate predictions. In this paper, we fixed the\ninconsistencies in Sharif Emotional Speech Database (ShEMO), as a Persian\ndatabase, by using an Automatic Speech Recognition (ASR) system and\ninvestigating the effect of Farsi language models obtained from accessible\nPersian text corpora. We also introduced a Persian/Farsi ASR-based SER system\nthat uses linguistic features of the ASR outputs and Deep Learning-based\nmodels.\n", "date": "2022-11-18", "categories": []}
{"id": "2211.08237", "title": "Multilingual Speech Emotion Recognition With Multi-Gating Mechanism and\n  Neural Architecture Search", "abstract": "  Speech emotion recognition (SER) classifies audio into emotion categories\nsuch as Happy, Angry, Fear, Disgust and Neutral. While Speech Emotion\nRecognition (SER) is a common application for popular languages, it continues\nto be a problem for low-resourced languages, i.e., languages with no pretrained\nspeech-to-text recognition models. This paper firstly proposes a\nlanguage-specific model that extract emotional information from multiple\npre-trained speech models, and then designs a multi-domain model that\nsimultaneously performs SER for various languages. Our multidomain model\nemploys a multi-gating mechanism to generate unique weighted feature\ncombination for each language, and also searches for specific neural network\nstructure for each language through a neural architecture search module. In\naddition, we introduce a contrastive auxiliary loss to build more separable\nrepresentations for audio data. Our experiments show that our model raises the\nstate-of-the-art accuracy by 3% for German and 14.3% for French.\n", "date": "2022-10-31", "categories": []}
{"id": "2211.07737", "title": "Describing emotions with acoustic property prompts for speech emotion\n  recognition", "abstract": "  Emotions lie on a broad continuum and treating emotions as a discrete number\nof classes limits the ability of a model to capture the nuances in the\ncontinuum. The challenge is how to describe the nuances of emotions and how to\nenable a model to learn the descriptions. In this work, we devise a method to\nautomatically create a description (or prompt) for a given audio by computing\nacoustic properties, such as pitch, loudness, speech rate, and articulation\nrate. We pair a prompt with its corresponding audio using 5 different emotion\ndatasets. We trained a neural network model using these audio-text pairs. Then,\nwe evaluate the model using one more dataset. We investigate how the model can\nlearn to associate the audio with the descriptions, resulting in performance\nimprovement of Speech Emotion Recognition and Speech Audio Retrieval. We expect\nour findings to motivate research describing the broad continuum of emotion\n", "date": "2022-11-14", "categories": []}
{"id": "2211.06609", "title": "AU-Aware Vision Transformers for Biased Facial Expression Recognition", "abstract": "  Studies have proven that domain bias and label bias exist in different Facial\nExpression Recognition (FER) datasets, making it hard to improve the\nperformance of a specific dataset by adding other datasets. For the FER bias\nissue, recent researches mainly focus on the cross-domain issue with advanced\ndomain adaption algorithms. This paper addresses another problem: how to boost\nFER performance by leveraging cross-domain datasets. Unlike the coarse and\nbiased expression label, the facial Action Unit (AU) is fine-grained and\nobjective suggested by psychological studies. Motivated by this, we resort to\nthe AU information of different FER datasets for performance boosting and make\ncontributions as follows. First, we experimentally show that the naive joint\ntraining of multiple FER datasets is harmful to the FER performance of\nindividual datasets. We further introduce expression-specific mean images and\nAU cosine distances to measure FER dataset bias. This novel measurement shows\nconsistent conclusions with experimental degradation of joint training. Second,\nwe propose a simple yet conceptually-new framework, AU-aware Vision Transformer\n(AU-ViT). It improves the performance of individual datasets by jointly\ntraining auxiliary datasets with AU or pseudo-AU labels. We also find that the\nAU-ViT is robust to real-world occlusions. Moreover, for the first time, we\nprove that a carefully-initialized ViT achieves comparable performance to\nadvanced deep convolutional networks. Our AU-ViT achieves state-of-the-art\nperformance on three popular datasets, namely 91.10% on RAF-DB, 65.59% on\nAffectNet, and 90.15% on FERPlus. The code and models will be released soon.\n", "date": "2022-11-12", "categories": []}
{"id": "2211.07307", "title": "Sentiment recognition of Italian elderly through domain adaptation on\n  cross-corpus speech dataset", "abstract": "  The aim of this work is to define a speech emotion recognition (SER) model\nable to recognize positive, neutral and negative emotions in natural\nconversations of Italian elderly people. Several datasets for SER are available\nin the literature. However most of them are in English or Chinese, have been\nrecorded while actors and actresses pronounce short phrases and thus are not\nrelated to natural conversation. Moreover only few speeches among all the\ndatabases are related to elderly people. Therefore, in this work, a\nmulti-language and multi-age corpus is considered merging a dataset in English,\nthat includes also elderly people, with a dataset in Italian. A general model,\ntrained on young and adult English actors and actresses is proposed, based on\nXGBoost. Then two strategies of domain adaptation are proposed to adapt the\nmodel either to elderly people and to Italian speakers. The results suggest\nthat this approach increases the classification performance, underlining also\nthat new datasets should be collected.\n", "date": "2022-11-14", "categories": []}
{"id": "2208.07216", "title": "Class-attention Video Transformer for Engagement Intensity Prediction", "abstract": "  In order to deal with variant-length long videos, prior works extract\nmulti-modal features and fuse them to predict students' engagement intensity.\nIn this paper, we present a new end-to-end method Class Attention in Video\nTransformer (CavT), which involves a single vector to process class embedding\nand to uniformly perform end-to-end learning on variant-length long videos and\nfixed-length short videos. Furthermore, to address the lack of sufficient\nsamples, we propose a binary-order representatives sampling method (BorS) to\nadd multiple video sequences of each video to augment the training set.\nBorS+CavT not only achieves the state-of-the-art MSE (0.0495) on the EmotiW-EP\ndataset, but also obtains the state-of-the-art MSE (0.0377) on the DAiSEE\ndataset. The code and models have been made publicly available at\nhttps://github.com/mountainai/cavt.\n", "date": "2022-08-12", "categories": []}
{"id": "2211.04769", "title": "Interpretable Explainability in Facial Emotion Recognition and\n  Gamification for Data Collection", "abstract": "  Training facial emotion recognition models requires large sets of data and\ncostly annotation processes. To alleviate this problem, we developed a gamified\nmethod of acquiring annotated facial emotion data without an explicit labeling\neffort by humans. The game, which we named Facegame, challenges the players to\nimitate a displayed image of a face that portrays a particular basic emotion.\nEvery round played by the player creates new data that consists of a set of\nfacial features and landmarks, already annotated with the emotion label of the\ntarget facial expression. Such an approach effectively creates a robust,\nsustainable, and continuous machine learning training process. We evaluated\nFacegame with an experiment that revealed several contributions to the field of\naffective computing. First, the gamified data collection approach allowed us to\naccess a rich variation of facial expressions of each basic emotion due to the\nnatural variations in the players' facial expressions and their expressive\nabilities. We report improved accuracy when the collected data were used to\nenrich well-known in-the-wild facial emotion datasets and consecutively used\nfor training facial emotion recognition models. Second, the natural language\nprescription method used by the Facegame constitutes a novel approach for\ninterpretable explainability that can be applied to any facial emotion\nrecognition model. Finally, we observed significant improvements in the facial\nemotion perception and expression skills of the players through repeated game\nplay.\n", "date": "2022-11-09", "categories": []}
{"id": "2211.05047", "title": "A Comparative Study of Data Augmentation Techniques for Deep Learning\n  Based Emotion Recognition", "abstract": "  Automated emotion recognition in speech is a long-standing problem. While\nearly work on emotion recognition relied on hand-crafted features and simple\nclassifiers, the field has now embraced end-to-end feature learning and\nclassification using deep neural networks. In parallel to these models,\nresearchers have proposed several data augmentation techniques to increase the\nsize and variability of existing labeled datasets. Despite many seminal\ncontributions in the field, we still have a poor understanding of the interplay\nbetween the network architecture and the choice of data augmentation. Moreover,\nonly a handful of studies demonstrate the generalizability of a particular\nmodel across multiple datasets, which is a prerequisite for robust real-world\nperformance. In this paper, we conduct a comprehensive evaluation of popular\ndeep learning approaches for emotion recognition. To eliminate bias, we fix the\nmodel architectures and optimization hyperparameters using the VESUS dataset\nand then use repeated 5-fold cross validation to evaluate the performance on\nthe IEMOCAP and CREMA-D datasets. Our results demonstrate that long-range\ndependencies in the speech signal are critical for emotion recognition and that\nspeed/rate augmentation offers the most robust performance gain across models.\n", "date": "2022-11-09", "categories": []}
{"id": "2210.15834", "title": "GM-TCNet: Gated Multi-scale Temporal Convolutional Network using Emotion\n  Causality for Speech Emotion Recognition", "abstract": "  In human-computer interaction, Speech Emotion Recognition (SER) plays an\nessential role in understanding the user's intent and improving the interactive\nexperience. While similar sentimental speeches own diverse speaker\ncharacteristics but share common antecedents and consequences, an essential\nchallenge for SER is how to produce robust and discriminative representations\nthrough causality between speech emotions. In this paper, we propose a Gated\nMulti-scale Temporal Convolutional Network (GM-TCNet) to construct a novel\nemotional causality representation learning component with a multi-scale\nreceptive field. GM-TCNet deploys a novel emotional causality representation\nlearning component to capture the dynamics of emotion across the time domain,\nconstructed with dilated causal convolution layer and gating mechanism.\nBesides, it utilizes skip connection fusing high-level features from different\ngated convolution blocks to capture abundant and subtle emotion changes in\nhuman speech. GM-TCNet first uses a single type of feature, mel-frequency\ncepstral coefficients, as inputs and then passes them through the gated\ntemporal convolutional module to generate the high-level features. Finally, the\nfeatures are fed to the emotion classifier to accomplish the SER task. The\nexperimental results show that our model maintains the highest performance in\nmost cases compared to state-of-the-art techniques.\n", "date": "2022-10-28", "categories": []}
{"id": "2211.02366", "title": "SPEAKER VGG CCT: Cross-corpus Speech Emotion Recognition with Speaker\n  Embedding and Vision Transformers", "abstract": "  In recent years, Speech Emotion Recognition (SER) has been investigated\nmainly transforming the speech signal into spectrograms that are then\nclassified using Convolutional Neural Networks pretrained on generic images and\nfine tuned with spectrograms. In this paper, we start from the general idea\nabove and develop a new learning solution for SER, which is based on Compact\nConvolutional Transformers (CCTs) combined with a speaker embedding. With CCTs,\nthe learning power of Vision Transformers (ViT) is combined with a diminished\nneed for large volume of data as made possible by the convolution. This is\nimportant in SER, where large corpora of data are usually not available. The\nspeaker embedding allows the network to extract an identity representation of\nthe speaker, which is then integrated by means of a self-attention mechanism\nwith the features that the CCT extracts from the spectrogram. Overall, the\nsolution is capable of operating in real-time showing promising results in a\ncross-corpus scenario, where training and test datasets are kept separate.\nExperiments have been performed on several benchmarks in a cross-corpus setting\nas rarely used in the literature, with results that are comparable or superior\nto those obtained with state-of-the-art network architectures. Our code is\navailable at https://github.com/JabuMlDev/Speaker-VGG-CCT.\n", "date": "2022-11-04", "categories": []}
{"id": "2211.02629", "title": "Multi-view Multi-label Fine-grained Emotion Decoding from Human Brain\n  Activity", "abstract": "  Decoding emotional states from human brain activity plays an important role\nin brain-computer interfaces. Existing emotion decoding methods still have two\nmain limitations: one is only decoding a single emotion category from a brain\nactivity pattern and the decoded emotion categories are coarse-grained, which\nis inconsistent with the complex emotional expression of human; the other is\nignoring the discrepancy of emotion expression between the left and right\nhemispheres of human brain. In this paper, we propose a novel multi-view\nmulti-label hybrid model for fine-grained emotion decoding (up to 80 emotion\ncategories) which can learn the expressive neural representations and\npredicting multiple emotional states simultaneously. Specifically, the\ngenerative component of our hybrid model is parametrized by a multi-view\nvariational auto-encoder, in which we regard the brain activity of left and\nright hemispheres and their difference as three distinct views, and use the\nproduct of expert mechanism in its inference network. The discriminative\ncomponent of our hybrid model is implemented by a multi-label classification\nnetwork with an asymmetric focal loss. For more accurate emotion decoding, we\nfirst adopt a label-aware module for emotion-specific neural representations\nlearning and then model the dependency of emotional states by a masked\nself-attention mechanism. Extensive experiments on two visually evoked\nemotional datasets show the superiority of our method.\n", "date": "2022-10-26", "categories": []}
{"id": "2211.02637", "title": "Emotion Recognition With Temporarily Localized 'Emotional Events' in\n  Naturalistic Context", "abstract": "  Emotion recognition using EEG signals is an emerging area of research due to\nits broad applicability in BCI. Emotional feelings are hard to stimulate in the\nlab. Emotions do not last long, yet they need enough context to be perceived\nand felt. However, most EEG-related emotion databases either suffer from\nemotionally irrelevant details (due to prolonged duration stimulus) or have\nminimal context doubting the feeling of any emotion using the stimulus. We\ntried to reduce the impact of this trade-off by designing an experiment in\nwhich participants are free to report their emotional feelings simultaneously\nwatching the emotional stimulus. We called these reported emotional feelings\n\"Emotional Events\" in our Dataset on Emotion with Naturalistic Stimuli (DENS).\nWe used EEG signals to classify emotional events on different combinations of\nValence(V) and Arousal(A) dimensions and compared the results with benchmark\ndatasets of DEAP and SEED. STFT is used for feature extraction and used in the\nclassification model consisting of CNN-LSTM hybrid layers. We achieved\nsignificantly higher accuracy with our data compared to DEEP and SEED data. We\nconclude that having precise information about emotional feelings improves the\nclassification accuracy compared to long-duration EEG signals which might be\ncontaminated by mind-wandering.\n", "date": "2022-10-25", "categories": []}
{"id": "2211.00794", "title": "Impact of annotation modality on label quality and model performance in\n  the automatic assessment of laughter in-the-wild", "abstract": "  Laughter is considered one of the most overt signals of joy. Laughter is\nwell-recognized as a multimodal phenomenon but is most commonly detected by\nsensing the sound of laughter. It is unclear how perception and annotation of\nlaughter differ when annotated from other modalities like video, via the body\nmovements of laughter. In this paper we take a first step in this direction by\nasking if and how well laughter can be annotated when only audio, only video\n(containing full body movement information) or audiovisual modalities are\navailable to annotators. We ask whether annotations of laughter are congruent\nacross modalities, and compare the effect that labeling modality has on machine\nlearning model performance. We compare annotations and models for laughter\ndetection, intensity estimation, and segmentation, three tasks common in\nprevious studies of laughter. Our analysis of more than 4000 annotations\nacquired from 48 annotators revealed evidence for incongruity in the perception\nof laughter, and its intensity between modalities. Further analysis of\nannotations against consolidated audiovisual reference annotations revealed\nthat recall was lower on average for video when compared to the audio\ncondition, but tended to increase with the intensity of the laughter samples.\nOur machine learning experiments compared the performance of state-of-the-art\nunimodal (audio-based, video-based and acceleration-based) and multi-modal\nmodels for different combinations of input modalities, training label modality,\nand testing label modality. Models with video and acceleration inputs had\nsimilar performance regardless of training label modality, suggesting that it\nmay be entirely appropriate to train models for laughter detection from body\nmovements using video-acquired labels, despite their lower inter-rater\nagreement.\n", "date": "2022-11-02", "categories": []}
{"id": "2211.00233", "title": "Detection of (Hidden) Emotions from Videos using Muscles Movements and\n  Face Manifold Embedding", "abstract": "  We provide a new non-invasive, easy-to-scale for large amounts of subjects\nand a remotely accessible method for (hidden) emotion detection from videos of\nhuman faces. Our approach combines face manifold detection for accurate\nlocation of the face in the video with local face manifold embedding to create\na common domain for the measurements of muscle micro-movements that is\ninvariant to the movement of the subject in the video. In the next step, we\nemploy the Digital Image Speckle Correlation (DISC) and the optical flow\nalgorithm to compute the pattern of micro-movements in the face. The\ncorresponding vector field is mapped back to the original space and\nsuperimposed on the original frames of the videos. Hence, the resulting videos\ninclude additional information about the direction of the movement of the\nmuscles in the face. We take the publicly available CK++ dataset of visible\nemotions and add to it videos of the same format but with hidden emotions. We\nprocess all the videos using our micro-movement detection and use the results\nto train a state-of-the-art network for emotions classification from videos --\nFrame Attention Network (FAN). Although the original FAN model achieves very\nhigh out-of-sample performance on the original CK++ videos, it does not perform\nso well on hidden emotions videos. The performance improves significantly when\nthe model is trained and tested on videos with the vector fields of muscle\nmovements. Intuitively, the corresponding arrows serve as edges in the image\nthat are easily captured by the convolutions filters in the FAN network.\n", "date": "2022-11-01", "categories": []}
{"id": "2208.00344", "title": "Towards Intercultural Affect Recognition: Audio-Visual Affect\n  Recognition in the Wild Across Six Cultures", "abstract": "  In our multicultural world, affect-aware AI systems that support humans need\nthe ability to perceive affect across variations in emotion expression patterns\nacross cultures. These systems must perform well in cultural contexts without\nannotated affect datasets available for training models. A standard assumption\nin affective computing is that affect recognition models trained and used\nwithin the same culture (intracultural) will perform better than models trained\non one culture and used on different cultures (intercultural). We test this\nassumption and present the first systematic study of intercultural affect\nrecognition models using videos of real-world dyadic interactions from six\ncultures. We develop an attention-based feature selection approach under\ntemporal causal discovery to identify behavioral cues that can be leveraged in\nintercultural affect recognition models. Across all six cultures, our findings\ndemonstrate that intercultural affect recognition models were as effective or\nmore effective than intracultural models. We identify and contribute useful\nbehavioral features for intercultural affect recognition; facial features from\nthe visual modality were more useful than the audio modality in this study's\ncontext. Our paper presents a proof-of-concept and motivation for the future\ndevelopment of intercultural affect recognition systems, especially those\ndeployed in low-resource situations without annotated data.\n", "date": "2022-07-31", "categories": []}
{"id": "2210.16642", "title": "Unifying the Discrete and Continuous Emotion labels for Speech Emotion\n  Recognition", "abstract": "  Traditionally, in paralinguistic analysis for emotion detection from speech,\nemotions have been identified with discrete or dimensional (continuous-valued)\nlabels. Accordingly, models that have been proposed for emotion detection use\none or the other of these label types. However, psychologists like Russell and\nPlutchik have proposed theories and models that unite these views, maintaining\nthat these representations have shared and complementary information. This\npaper is an attempt to validate these viewpoints computationally. To this end,\nwe propose a model to jointly predict continuous and discrete emotional\nattributes and show how the relationship between these can be utilized to\nimprove the robustness and performance of emotion recognition tasks. Our\napproach comprises multi-task and hierarchical multi-task learning frameworks\nthat jointly model the relationships between continuous-valued and discrete\nemotion labels. Experimental results on two widely used datasets (IEMOCAP and\nMSPPodcast) for speech-based emotion recognition show that our model results in\nstatistically significant improvements in performance over strong baselines\nwith non-unified approaches. We also demonstrate that using one type of label\n(discrete or continuous-valued) for training improves recognition performance\nin tasks that use the other type of label. Experimental results and reasoning\nfor this approach (called the mismatched training approach) are also presented.\n", "date": "2022-10-29", "categories": []}
{"id": "2210.16807", "title": "The Florence 4D Facial Expression Dataset", "abstract": "  Human facial expressions change dynamically, so their recognition / analysis\nshould be conducted by accounting for the temporal evolution of face\ndeformations either in 2D or 3D. While abundant 2D video data do exist, this is\nnot the case in 3D, where few 3D dynamic (4D) datasets were released for public\nuse. The negative consequence of this scarcity of data is amplified by current\ndeep learning based-methods for facial expression analysis that require large\nquantities of variegate samples to be effectively trained. With the aim of\nsmoothing such limitations, in this paper we propose a large dataset, named\nFlorence 4D, composed of dynamic sequences of 3D face models, where a\ncombination of synthetic and real identities exhibit an unprecedented variety\nof 4D facial expressions, with variations that include the classical\nneutral-apex transition, but generalize to expression-to-expression. All these\ncharacteristics are not exposed by any of the existing 4D datasets and they\ncannot even be obtained by combining more than one dataset. We strongly believe\nthat making such a data corpora publicly available to the community will allow\ndesigning and experimenting new applications that were not possible to\ninvestigate till now. To show at some extent the difficulty of our data in\nterms of different identities and varying expressions, we also report a\nbaseline experimentation on the proposed dataset that can be used as baseline.\n", "date": "2022-10-30", "categories": []}
{"id": "2210.17421", "title": "I am Only Happy When There is Light: The Impact of Environmental Changes\n  on Affective Facial Expressions Recognition", "abstract": "  Human-robot interaction (HRI) benefits greatly from advances in the machine\nlearning field as it allows researchers to employ high-performance models for\nperceptual tasks like detection and recognition. Especially deep learning\nmodels, either pre-trained for feature extraction or used for classification,\nare now established methods to characterize human behaviors in HRI scenarios\nand to have social robots that understand better those behaviors. As HRI\nexperiments are usually small-scale and constrained to particular lab\nenvironments, the questions are how well can deep learning models generalize to\nspecific interaction scenarios, and further, how good is their robustness\ntowards environmental changes? These questions are important to address if the\nHRI field wishes to put social robotic companions into real environments acting\nconsistently, i.e. changing lighting conditions or moving people should still\nproduce the same recognition results. In this paper, we study the impact of\ndifferent image conditions on the recognition of arousal and valence from human\nfacial expressions using the FaceChannel framework \\cite{Barro20}. Our results\nshow how the interpretation of human affective states can differ greatly in\neither the positive or negative direction even when changing only slightly the\nimage properties. We conclude the paper with important points to consider when\nemploying deep learning models to ensure sound interpretation of HRI\nexperiments.\n", "date": "2022-10-28", "categories": []}
{"id": "2206.14976", "title": "Semi-Supervised Generative Adversarial Network for Stress Detection\n  Using Partially Labeled Physiological Data", "abstract": "  Physiological measurements involves observing variables that attribute to the\nnormative functioning of human systems and subsystems directly or indirectly.\nThe measurements can be used to detect affective states of a person with aims\nsuch as improving human-computer interactions. There are several methods of\ncollecting physiological data, but wearable sensors are a common, non-invasive\ntool for accurate readings. However, valuable information is hard to extract\nfrom the raw physiological data, especially for affective state detection.\nMachine Learning techniques are used to detect the affective state of a person\nthrough labeled physiological data. A clear problem with using labeled data is\ncreating accurate labels. An expert is needed to analyze a form of recording of\nparticipants and mark sections with different states such as stress and calm.\nWhile expensive, this method delivers a complete dataset with labeled data that\ncan be used in any number of supervised algorithms. An interesting question\narises from the expensive labeling: how can we reduce the cost while\nmaintaining high accuracy? Semi-Supervised learning (SSL) is a potential\nsolution to this problem. These algorithms allow for machine learning models to\nbe trained with only a small subset of labeled data (unlike unsupervised which\nuse no labels). They provide a way of avoiding expensive labeling. This paper\ncompares a fully supervised algorithm to a SSL on the public WESAD (Wearable\nStress and Affect Detection) Dataset for stress detection. This paper shows\nthat Semi-Supervised algorithms are a viable method for inexpensive affective\nstate detection systems with accurate results.\n", "date": "2022-06-30", "categories": []}
{"id": "2207.03572", "title": "The ACII 2022 Affective Vocal Bursts Workshop & Competition:\n  Understanding a critically understudied modality of emotional expression", "abstract": "  The ACII Affective Vocal Bursts Workshop & Competition is focused on\nunderstanding multiple affective dimensions of vocal bursts: laughs, gasps,\ncries, screams, and many other non-linguistic vocalizations central to the\nexpression of emotion and to human communication more generally. This year's\ncompetition comprises four tracks using a large-scale and in-the-wild dataset\nof 59,299 vocalizations from 1,702 speakers. The first, the A-VB-High task,\nrequires competition participants to perform a multi-label regression on a\nnovel model for emotion, utilizing ten classes of richly annotated emotional\nexpression intensities, including; Awe, Fear, and Surprise. The second, the\nA-VB-Two task, utilizes the more conventional 2-dimensional model for emotion,\narousal, and valence. The third, the A-VB-Culture task, requires participants\nto explore the cultural aspects of the dataset, training native-country\ndependent models. Finally, for the fourth task, A-VB-Type, participants should\nrecognize the type of vocal burst (e.g., laughter, cry, grunt) as an 8-class\nclassification. This paper describes the four tracks and baseline systems,\nwhich use state-of-the-art machine learning methods. The baseline performance\nfor each track is obtained by utilizing an end-to-end deep learning model and\nis as follows: for A-VB-High, a mean (over the 10-dimensions) Concordance\nCorrelation Coefficient (CCC) of 0.5687 CCC is obtained; for A-VB-Two, a mean\n(over the 2-dimensions) CCC of 0.5084 is obtained; for A-VB-Culture, a mean CCC\nfrom the four cultures of 0.4401 is obtained; and for A-VB-Type, the baseline\nUnweighted Average Recall (UAR) from the 8-classes is 0.4172 UAR.\n", "date": "2022-07-07", "categories": []}
{"id": "2210.15937", "title": "On the Use of Modality-Specific Large-Scale Pre-Trained Encoders for\n  Multimodal Sentiment Analysis", "abstract": "  This paper investigates the effectiveness and implementation of\nmodality-specific large-scale pre-trained encoders for multimodal sentiment\nanalysis~(MSA). Although the effectiveness of pre-trained encoders in various\nfields has been reported, conventional MSA methods employ them for only\nlinguistic modality, and their application has not been investigated. This\npaper compares the features yielded by large-scale pre-trained encoders with\nconventional heuristic features. One each of the largest pre-trained encoders\npublicly available for each modality are used; CLIP-ViT, WavLM, and BERT for\nvisual, acoustic, and linguistic modalities, respectively. Experiments on two\ndatasets reveal that methods with domain-specific pre-trained encoders attain\nbetter performance than those with conventional features in both unimodal and\nmultimodal scenarios. We also find it better to use the outputs of the\nintermediate layers of the encoders than those of the output layer. The codes\nare available at https://github.com/ando-hub/MSA_Pretrain.\n", "date": "2022-10-28", "categories": []}
{"id": "2208.12812", "title": "Speech Emotion Recognition using Supervised Deep Recurrent System for\n  Mental Health Monitoring", "abstract": "  Understanding human behavior and monitoring mental health are essential to\nmaintaining the community and society's safety. As there has been an increase\nin mental health problems during the COVID-19 pandemic due to uncontrolled\nmental health, early detection of mental issues is crucial. Nowadays, the usage\nof Intelligent Virtual Personal Assistants (IVA) has increased worldwide.\nIndividuals use their voices to control these devices to fulfill requests and\nacquire different services. This paper proposes a novel deep learning model\nbased on the gated recurrent neural network and convolution neural network to\nunderstand human emotion from speech to improve their IVA services and monitor\ntheir mental health.\n", "date": "2022-08-26", "categories": []}
{"id": "2210.15247", "title": "A few-shot learning approach with domain adaptation for personalized\n  real-life stress detection in close relationships", "abstract": "  We design a metric learning approach that aims to address computational\nchallenges that yield from modeling human outcomes from ambulatory real-life\ndata. The proposed metric learning is based on a Siamese neural network (SNN)\nthat learns the relative difference between pairs of samples from a target user\nand non-target users, thus being able to address the scarcity of labelled data\nfrom the target. The SNN further minimizes the Wasserstein distance of the\nlearned embeddings between target and non-target users, thus mitigating the\ndistribution mismatch between the two. Finally, given the fact that the base\nrate of focal behaviors is different per user, the proposed method approximates\nthe focal base rate based on labelled samples that lay closest to the target,\nbased on which further minimizes the Wasserstein distance. Our method is\nexemplified for the purpose of hourly stress classification using real-life\nmultimodal data from 72 dating couples. Results in few-shot and one-shot\nlearning experiments indicate that proposed formulation benefits stress\nclassification and can help mitigate the aforementioned challenges.\n", "date": "2022-10-27", "categories": []}
{"id": "2210.15359", "title": "Exploiting modality-invariant feature for robust multimodal emotion\n  recognition with missing modalities", "abstract": "  Multimodal emotion recognition leverages complementary information across\nmodalities to gain performance. However, we cannot guarantee that the data of\nall modalities are always present in practice. In the studies to predict the\nmissing data across modalities, the inherent difference between heterogeneous\nmodalities, namely the modality gap, presents a challenge. To address this, we\npropose to use invariant features for a missing modality imagination network\n(IF-MMIN) which includes two novel mechanisms: 1) an invariant feature learning\nstrategy that is based on the central moment discrepancy (CMD) distance under\nthe full-modality scenario; 2) an invariant feature based imagination module\n(IF-IM) to alleviate the modality gap during the missing modalities prediction,\nthus improving the robustness of multimodal joint representation. Comprehensive\nexperiments on the benchmark dataset IEMOCAP demonstrate that the proposed\nmodel outperforms all baselines and invariantly improves the overall emotion\nrecognition performance under uncertain missing-modality conditions. We release\nthe code at: https://github.com/ZhuoYulang/IF-MMIN.\n", "date": "2022-10-27", "categories": []}
{"id": "2210.14419", "title": "Discourse-Aware Emotion Cause Extraction in Conversations", "abstract": "  Emotion Cause Extraction in Conversations (ECEC) aims to extract the\nutterances which contain the emotional cause in conversations. Most prior\nresearch focuses on modelling conversational contexts with sequential encoding,\nignoring the informative interactions between utterances and\nconversational-specific features for ECEC. In this paper, we investigate the\nimportance of discourse structures in handling utterance interactions and\nconversationspecific features for ECEC. To this end, we propose a\ndiscourse-aware model (DAM) for this task. Concretely, we jointly model ECEC\nwith discourse parsing using a multi-task learning (MTL) framework and\nexplicitly encode discourse structures via gated graph neural network (gated\nGNN), integrating rich utterance interaction information to our model. In\naddition, we use gated GNN to further enhance our ECEC model with\nconversation-specific features. Results on the benchmark corpus show that DAM\noutperform the state-of-theart (SOTA) systems in the literature. This suggests\nthat the discourse structure may contain a potential link between emotional\nutterances and their corresponding cause expressions. It also verifies the\neffectiveness of conversationalspecific features. The codes of this paper will\nbe available on GitHub.\n", "date": "2022-10-26", "categories": []}
{"id": "2210.13876", "title": "Towards emotion recognition for virtual environments: an evaluation of\n  EEG features on benchmark dataset", "abstract": "  One of the challenges in virtual environments is the difficulty users have in\ninteracting with these increasingly complex systems. Ultimately, endowing\nmachines with the ability to perceive users emotions will enable a more\nintuitive and reliable interaction. Consequently, using the\nelectroencephalogram as a bio-signal sensor, the affective state of a user can\nbe modelled and subsequently utilised in order to achieve a system that can\nrecognise and react to the user's emotions. This paper investigates features\nextracted from electroencephalogram signals for the purpose of affective state\nmodelling based on Russell's Circumplex Model. Investigations are presented\nthat aim to provide the foundation for future work in modelling user affect to\nenhance interaction experience in virtual environments. The DEAP dataset was\nused within this work, along with a Support Vector Machine and Random Forest,\nwhich yielded reasonable classification accuracies for Valence and Arousal\nusing feature vectors based on statistical measurements and band power from the\n\\'z, \\b{eta}, \\'z, and \\'z\\'z waves and High Order Crossing of the EEG signal.\n", "date": "2022-10-25", "categories": []}
{"id": "2210.12430", "title": "Speech Emotion Recognition via an Attentive Time-Frequency Neural\n  Network", "abstract": "  Spectrogram is commonly used as the input feature of deep neural networks to\nlearn the high(er)-level time-frequency pattern of speech signal for speech\nemotion recognition (SER). \\textcolor{black}{Generally, different emotions\ncorrespond to specific energy activations both within frequency bands and time\nframes on spectrogram, which indicates the frequency and time domains are both\nessential to represent the emotion for SER. However, recent spectrogram-based\nworks mainly focus on modeling the long-term dependency in time domain, leading\nto these methods encountering the following two issues: (1) neglecting to model\nthe emotion-related correlations within frequency domain during the\ntime-frequency joint learning; (2) ignoring to capture the specific frequency\nbands associated with emotions.} To cope with the issues, we propose an\nattentive time-frequency neural network (ATFNN) for SER, including a\ntime-frequency neural network (TFNN) and time-frequency attention.\nSpecifically, aiming at the first issue, we design a TFNN with a\nfrequency-domain encoder (F-Encoder) based on the Transformer encoder and a\ntime-domain encoder (T-Encoder) based on the Bidirectional Long Short-Term\nMemory (Bi-LSTM). The F-Encoder and T-Encoder model the correlations within\nfrequency bands and time frames, respectively, and they are embedded into a\ntime-frequency joint learning strategy to obtain the time-frequency patterns\nfor speech emotions. Moreover, to handle the second issue, we also adopt\ntime-frequency attention with a frequency-attention network (F-Attention) and a\ntime-attention network (T-Attention) to focus on the emotion-related frequency\nband ranges and time frame ranges, which can enhance the discriminability of\nspeech emotion features.\n", "date": "2022-10-22", "categories": []}
{"id": "2207.05691", "title": "The MuSe 2022 Multimodal Sentiment Analysis Challenge: Humor, Emotional\n  Reactions, and Stress", "abstract": "  The Multimodal Sentiment Analysis Challenge (MuSe) 2022 is dedicated to\nmultimodal sentiment and emotion recognition. For this year's challenge, we\nfeature three datasets: (i) the Passau Spontaneous Football Coach Humor\n(Passau-SFCH) dataset that contains audio-visual recordings of German football\ncoaches, labelled for the presence of humour; (ii) the Hume-Reaction dataset in\nwhich reactions of individuals to emotional stimuli have been annotated with\nrespect to seven emotional expression intensities, and (iii) the Ulm-Trier\nSocial Stress Test (Ulm-TSST) dataset comprising of audio-visual data labelled\nwith continuous emotion values (arousal and valence) of people in stressful\ndispositions. Using the introduced datasets, MuSe 2022 2022 addresses three\ncontemporary affective computing problems: in the Humor Detection Sub-Challenge\n(MuSe-Humor), spontaneous humour has to be recognised; in the Emotional\nReactions Sub-Challenge (MuSe-Reaction), seven fine-grained `in-the-wild'\nemotions have to be predicted; and in the Emotional Stress Sub-Challenge\n(MuSe-Stress), a continuous prediction of stressed emotion values is featured.\nThe challenge is designed to attract different research communities,\nencouraging a fusion of their disciplines. Mainly, MuSe 2022 targets the\ncommunities of audio-visual emotion recognition, health informatics, and\nsymbolic sentiment analysis. This baseline paper describes the datasets as well\nas the feature sets extracted from them. A recurrent neural network with LSTM\ncells is used to set competitive baseline results on the test partitions for\neach sub-challenge. We report an Area Under the Curve (AUC) of .8480 for\nMuSe-Humor; .2801 mean (from 7-classes) Pearson's Correlations Coefficient for\nMuSe-Reaction, as well as .4931 Concordance Correlation Coefficient (CCC) and\n.4761 for valence and arousal in MuSe-Stress, respectively.\n", "date": "2022-06-23", "categories": []}
{"id": "2207.09508", "title": "HSE-NN Team at the 4th ABAW Competition: Multi-task Emotion Recognition\n  and Learning from Synthetic Images", "abstract": "  In this paper, we present the results of the HSE-NN team in the 4th\ncompetition on Affective Behavior Analysis in-the-wild (ABAW). The novel\nmulti-task EfficientNet model is trained for simultaneous recognition of facial\nexpressions and prediction of valence and arousal on static photos. The\nresulting MT-EmotiEffNet extracts visual features that are fed into simple\nfeed-forward neural networks in the multi-task learning challenge. We obtain\nperformance measure 1.3 on the validation set, which is significantly greater\nwhen compared to either performance of baseline (0.3) or existing models that\nare trained only on the s-Aff-Wild2 database. In the learning from synthetic\ndata challenge, the quality of the original synthetic training set is increased\nby using the super-resolution techniques, such as Real-ESRGAN. Next, the\nMT-EmotiEffNet is fine-tuned on the new training set. The final prediction is a\nsimple blending ensemble of pre-trained and fine-tuned MT-EmotiEffNets. Our\naverage validation F1 score is 18% greater than the baseline convolutional\nneural network.\n", "date": "2022-07-19", "categories": []}
{"id": "2210.11341", "title": "SS-VAERR: Self-Supervised Apparent Emotional Reaction Recognition from\n  Video", "abstract": "  This work focuses on the apparent emotional reaction recognition (AERR) from\nthe video-only input, conducted in a self-supervised fashion. The network is\nfirst pre-trained on different self-supervised pretext tasks and later\nfine-tuned on the downstream target task. Self-supervised learning facilitates\nthe use of pre-trained architectures and larger datasets that might be deemed\nunfit for the target task and yet might be useful to learn informative\nrepresentations and hence provide useful initializations for further\nfine-tuning on smaller more suitable data. Our presented contribution is\ntwo-fold: (1) an analysis of different state-of-the-art (SOTA) pretext tasks\nfor the video-only apparent emotional reaction recognition architecture, and\n(2) an analysis of various combinations of the regression and classification\nlosses that are likely to improve the performance further. Together these two\ncontributions result in the current state-of-the-art performance for the\nvideo-only spontaneous apparent emotional reaction recognition with continuous\nannotations.\n", "date": "2022-10-20", "categories": []}
{"id": "2210.08713", "title": "Supervised Prototypical Contrastive Learning for Emotion Recognition in\n  Conversation", "abstract": "  Capturing emotions within a conversation plays an essential role in modern\ndialogue systems. However, the weak correlation between emotions and semantics\nbrings many challenges to emotion recognition in conversation (ERC). Even\nsemantically similar utterances, the emotion may vary drastically depending on\ncontexts or speakers. In this paper, we propose a Supervised Prototypical\nContrastive Learning (SPCL) loss for the ERC task. Leveraging the Prototypical\nNetwork, the SPCL targets at solving the imbalanced classification problem\nthrough contrastive learning and does not require a large batch size.\nMeanwhile, we design a difficulty measure function based on the distance\nbetween classes and introduce curriculum learning to alleviate the impact of\nextreme samples. We achieve state-of-the-art results on three widely used\nbenchmarks. Further, we conduct analytical experiments to demonstrate the\neffectiveness of our proposed SPCL and curriculum learning strategy. We release\nthe code at https://github.com/caskcsg/SPCL.\n", "date": "2022-10-17", "categories": []}
{"id": "2210.10130", "title": "PERI: Part Aware Emotion Recognition In The Wild", "abstract": "  Emotion recognition aims to interpret the emotional states of a person based\non various inputs including audio, visual, and textual cues. This paper focuses\non emotion recognition using visual features. To leverage the correlation\nbetween facial expression and the emotional state of a person, pioneering\nmethods rely primarily on facial features. However, facial features are often\nunreliable in natural unconstrained scenarios, such as in crowded scenes, as\nthe face lacks pixel resolution and contains artifacts due to occlusion and\nblur. To address this, in the wild emotion recognition exploits full-body\nperson crops as well as the surrounding scene context. In a bid to use body\npose for emotion recognition, such methods fail to realize the potential that\nfacial expressions, when available, offer. Thus, the aim of this paper is\ntwo-fold. First, we demonstrate our method, PERI, to leverage both body pose\nand facial landmarks. We create part aware spatial (PAS) images by extracting\nkey regions from the input image using a mask generated from both body pose and\nfacial landmarks. This allows us to exploit body pose in addition to facial\ncontext whenever available. Second, to reason from the PAS images, we introduce\ncontext infusion (Cont-In) blocks. These blocks attend to part-specific\ninformation, and pass them onto the intermediate features of an emotion\nrecognition network. Our approach is conceptually simple and can be applied to\nany existing emotion recognition method. We provide our results on the publicly\navailable in the wild EMOTIC dataset. Compared to existing methods, PERI\nachieves superior performance and leads to significant improvements in the mAP\nof emotion categories, while decreasing Valence, Arousal and Dominance errors.\nImportantly, we observe that our method improves performance in both images\nwith fully visible faces as well as in images with occluded or blurred faces.\n", "date": "2022-10-18", "categories": []}
{"id": "2206.12469", "title": "Burst2Vec: An Adversarial Multi-Task Approach for Predicting Emotion,\n  Age, and Origin from Vocal Bursts", "abstract": "  We present Burst2Vec, our multi-task learning approach to predict emotion,\nage, and origin (i.e., native country/language) from vocal bursts. Burst2Vec\nutilises pre-trained speech representations to capture acoustic information\nfrom raw waveforms and incorporates the concept of model debiasing via\nadversarial training. Our models achieve a relative 30 % performance gain over\nbaselines using pre-extracted features and score the highest amongst all\nparticipants in the ICML ExVo 2022 Multi-Task Challenge.\n", "date": "2022-06-24", "categories": []}
{"id": "2210.09396", "title": "Affective Idiosyncratic Responses to Music", "abstract": "  Affective responses to music are highly personal. Despite consensus that\nidiosyncratic factors play a key role in regulating how listeners emotionally\nrespond to music, precisely measuring the marginal effects of these variables\nhas proved challenging. To address this gap, we develop computational methods\nto measure affective responses to music from over 403M listener comments on a\nChinese social music platform. Building on studies from music psychology in\nsystematic and quasi-causal analyses, we test for musical, lyrical, contextual,\ndemographic, and mental health effects that drive listener affective responses.\nFinally, motivated by the social phenomenon known as w\\v{a}ng-y\\`i-y\\'un, we\nidentify influencing factors of platform user self-disclosures, the social\nsupport they receive, and notable differences in discloser user activity.\n", "date": "2022-10-17", "categories": []}
{"id": "2210.09434", "title": "Modelling Emotion Dynamics in Song Lyrics with State Space Models", "abstract": "  Most previous work in music emotion recognition assumes a single or a few\nsong-level labels for the whole song. While it is known that different emotions\ncan vary in intensity within a song, annotated data for this setup is scarce\nand difficult to obtain. In this work, we propose a method to predict emotion\ndynamics in song lyrics without song-level supervision. We frame each song as a\ntime series and employ a State Space Model (SSM), combining a sentence-level\nemotion predictor with an Expectation-Maximization (EM) procedure to generate\nthe full emotion dynamics. Our experiments show that applying our method\nconsistently improves the performance of sentence-level baselines without\nrequiring any annotated songs, making it ideal for limited training data\nscenarios. Further analysis through case studies shows the benefits of our\nmethod while also indicating the limitations and pointing to future directions.\n", "date": "2022-10-17", "categories": []}
{"id": "2210.10039", "title": "How Would The Viewer Feel? Estimating Wellbeing From Video Scenarios", "abstract": "  In recent years, deep neural networks have demonstrated increasingly strong\nabilities to recognize objects and activities in videos. However, as video\nunderstanding becomes widely used in real-world applications, a key\nconsideration is developing human-centric systems that understand not only the\ncontent of the video but also how it would affect the wellbeing and emotional\nstate of viewers. To facilitate research in this setting, we introduce two\nlarge-scale datasets with over 60,000 videos manually annotated for emotional\nresponse and subjective wellbeing. The Video Cognitive Empathy (VCE) dataset\ncontains annotations for distributions of fine-grained emotional responses,\nallowing models to gain a detailed understanding of affective states. The Video\nto Valence (V2V) dataset contains annotations of relative pleasantness between\nvideos, which enables predicting a continuous spectrum of wellbeing. In\nexperiments, we show how video models that are primarily trained to recognize\nactions and find contours of objects can be repurposed to understand human\npreferences and the emotional content of videos. Although there is room for\nimprovement, predicting wellbeing and emotional response is on the horizon for\nstate-of-the-art models. We hope our datasets can help foster further advances\nat the intersection of commonsense video understanding and human preference\nlearning.\n", "date": "2022-10-18", "categories": []}
{"id": "2210.07630", "title": "The Invariant Ground Truth of Affect", "abstract": "  Affective computing strives to unveil the unknown relationship between affect\nelicitation, manifestation of affect and affect annotations. The ground truth\nof affect, however, is predominately attributed to the affect labels which\ninadvertently include biases inherent to the subjective nature of emotion and\nits labeling. The response to such limitations is usually augmenting the\ndataset with more annotations per data point; however, this is not possible\nwhen we are interested in self-reports via first-person annotation. Moreover,\noutlier detection methods based on inter-annotator agreement only consider the\nannotations themselves and ignore the context and the corresponding affect\nmanifestation. This paper reframes the ways one may obtain a reliable ground\ntruth of affect by transferring aspects of causation theory to affective\ncomputing. In particular, we assume that the ground truth of affect can be\nfound in the causal relationships between elicitation, manifestation and\nannotation that remain \\emph{invariant} across tasks and participants. To test\nour assumption we employ causation inspired methods for detecting outliers in\naffective corpora and building affect models that are robust across\nparticipants and tasks. We validate our methodology within the domain of\ndigital games, with experimental results showing that it can successfully\ndetect outliers and boost the accuracy of affect models. To the best of our\nknowledge, this study presents the first attempt to integrate causation tools\nin affective computing, making a crucial and decisive step towards general\naffect modeling.\n", "date": "2022-10-14", "categories": []}
{"id": "2210.07642", "title": "Training speech emotion classifier without categorical annotations", "abstract": "  There are two paradigms of emotion representation, categorical labeling and\ndimensional description in continuous space. Therefore, the emotion recognition\ntask can be treated as a classification or regression. The main aim of this\nstudy is to investigate the relation between these two representations and\npropose a classification pipeline that uses only dimensional annotation. The\nproposed approach contains a regressor model which is trained to predict a\nvector of continuous values in dimensional representation for given speech\naudio. The output of this model can be interpreted as an emotional category\nusing a mapping algorithm. We investigated the performances of a combination of\nthree feature extractors, three neural network architectures, and three mapping\nalgorithms on two different corpora. Our study shows the advantages and\nlimitations of the classification via regression approach.\n", "date": "2022-10-14", "categories": []}
{"id": "2210.05790", "title": "Transfer Learning with Joint Fine-Tuning for Multimodal Sentiment\n  Analysis", "abstract": "  Most existing methods focus on sentiment analysis of textual data. However,\nrecently there has been a massive use of images and videos on social platforms,\nmotivating sentiment analysis from other modalities. Current studies show that\nexploring other modalities (e.g., images) increases sentiment analysis\nperformance. State-of-the-art multimodal models, such as CLIP and VisualBERT,\nare pre-trained on datasets with the text paired with images. Although the\nresults obtained by these models are promising, pre-training and sentiment\nanalysis fine-tuning tasks of these models are computationally expensive. This\npaper introduces a transfer learning approach using joint fine-tuning for\nsentiment analysis. Our proposal achieved competitive results using a more\nstraightforward alternative fine-tuning strategy that leverages different\npre-trained unimodal models and efficiently combines them in a multimodal\nspace. Moreover, our proposal allows flexibility when incorporating any\npre-trained model for texts and images during the joint fine-tuning stage,\nbeing especially interesting for sentiment classification in low-resource\nscenarios.\n", "date": "2022-10-11", "categories": []}
{"id": "2210.04935", "title": "Deep Insights of Learning based Micro Expression Recognition: A\n  Perspective on Promises, Challenges and Research Needs", "abstract": "  Micro expression recognition (MER) is a very challenging area of research due\nto its intrinsic nature and fine-grained changes. In the literature, the\nproblem of MER has been solved through handcrafted/descriptor-based techniques.\nHowever, in recent times, deep learning (DL) based techniques have been adopted\nto gain higher performance for MER. Also, rich survey articles on MER are\navailable by summarizing the datasets, experimental settings, conventional and\ndeep learning methods. In contrast, these studies lack the ability to convey\nthe impact of network design paradigms and experimental setting strategies for\nDL-based MER. Therefore, this paper aims to provide a deep insight into the\nDL-based MER frameworks with a perspective on promises in network model\ndesigning, experimental strategies, challenges, and research needs. Also, the\ndetailed categorization of available MER frameworks is presented in various\naspects of model design and technical characteristics. Moreover, an empirical\nanalysis of the experimental and validation protocols adopted by MER methods is\npresented. The challenges mentioned earlier and network design strategies may\nassist the affective computing research community in forging ahead in MER\nresearch. Finally, we point out the future directions, research needs, and draw\nour conclusions.\n", "date": "2022-10-10", "categories": []}
{"id": "2210.05246", "title": "Cluster-level pseudo-labelling for source-free cross-domain facial\n  expression recognition", "abstract": "  Automatically understanding emotions from visual data is a fundamental task\nfor human behaviour understanding. While models devised for Facial Expression\nRecognition (FER) have demonstrated excellent performances on many datasets,\nthey often suffer from severe performance degradation when trained and tested\non different datasets due to domain shift. In addition, as face images are\nconsidered highly sensitive data, the accessibility to large-scale datasets for\nmodel training is often denied. In this work, we tackle the above-mentioned\nproblems by proposing the first Source-Free Unsupervised Domain Adaptation\n(SFUDA) method for FER. Our method exploits self-supervised pretraining to\nlearn good feature representations from the target data and proposes a novel\nand robust cluster-level pseudo-labelling strategy that accounts for in-cluster\nstatistics. We validate the effectiveness of our method in four adaptation\nsetups, proving that it consistently outperforms existing SFUDA methods when\napplied to FER, and is on par with methods addressing FER in the UDA setting.\n", "date": "2022-10-11", "categories": []}
{"id": "2210.05419", "title": "Exploring Interactions and Regulations in Collaborative Learning: An\n  Interdisciplinary Multimodal Dataset", "abstract": "  Collaborative learning is an educational approach that enhances learning\nthrough shared goals and working together. Interaction and regulation are two\nessential factors related to the success of collaborative learning. Since the\ninformation from various modalities can reflect the quality of collaboration, a\nnew multimodal dataset with cognitive and emotional triggers is introduced in\nthis paper to explore how regulations affect interactions during the\ncollaborative process. Specifically, a learning task with intentional\ninterventions is designed and assigned to high school students aged 15 years\nold (N=81) in average. Multimodal signals, including video, Kinect, audio, and\nphysiological data, are collected and exploited to study regulations in\ncollaborative learning in terms of individual-participant-single-modality,\nindividual-participant-multiple-modality, and\nmultiple-participant-multiple-modality. Analysis of annotated emotions, body\ngestures, and their interactions indicates that our multimodal dataset with\ndesigned treatments could effectively examine moments of regulation in\ncollaborative learning. In addition, preliminary experiments based on baseline\nmodels suggest that the dataset provides a challenging in-the-wild scenario,\nwhich could further contribute to the fields of education and affective\ncomputing.\n", "date": "2022-10-11", "categories": []}
{"id": "2210.03853", "title": "Revisiting Self-Supervised Contrastive Learning for Facial Expression\n  Recognition", "abstract": "  The success of most advanced facial expression recognition works relies\nheavily on large-scale annotated datasets. However, it poses great challenges\nin acquiring clean and consistent annotations for facial expression datasets.\nOn the other hand, self-supervised contrastive learning has gained great\npopularity due to its simple yet effective instance discrimination training\nstrategy, which can potentially circumvent the annotation issue. Nevertheless,\nthere remain inherent disadvantages of instance-level discrimination, which are\neven more challenging when faced with complicated facial representations. In\nthis paper, we revisit the use of self-supervised contrastive learning and\nexplore three core strategies to enforce expression-specific representations\nand to minimize the interference from other facial attributes, such as identity\nand face styling. Experimental results show that our proposed method\noutperforms the current state-of-the-art self-supervised learning methods, in\nterms of both categorical and dimensional facial expression recognition tasks.\n", "date": "2022-10-08", "categories": []}
{"id": "2210.04240", "title": "Less is More: Facial Landmarks can Recognize a Spontaneous Smile", "abstract": "  Smile veracity classification is a task of interpreting social interactions.\nBroadly, it distinguishes between spontaneous and posed smiles. Previous\napproaches used hand-engineered features from facial landmarks or considered\nraw smile videos in an end-to-end manner to perform smile classification tasks.\nFeature-based methods require intervention from human experts on feature\nengineering and heavy pre-processing steps. On the contrary, raw smile video\ninputs fed into end-to-end models bring more automation to the process with the\ncost of considering many redundant facial features (beyond landmark locations)\nthat are mainly irrelevant to smile veracity classification. It remains unclear\nto establish discriminative features from landmarks in an end-to-end manner. We\npresent a MeshSmileNet framework, a transformer architecture, to address the\nabove limitations. To eliminate redundant facial features, our landmarks input\nis extracted from Attention Mesh, a pre-trained landmark detector. Again, to\ndiscover discriminative features, we consider the relativity and trajectory of\nthe landmarks. For the relativity, we aggregate facial landmark that\nconceptually formats a curve at each frame to establish local spatial features.\nFor the trajectory, we estimate the movements of landmark composed features\nacross time by self-attention mechanism, which captures pairwise dependency on\nthe trajectory of the same landmark. This idea allows us to achieve\nstate-of-the-art performances on UVA-NEMO, BBC, MMI Facial Expression, and SPOS\ndatasets.\n", "date": "2022-10-09", "categories": []}
{"id": "2206.05238", "title": "Dimensional Modeling of Emotions in Text with Appraisal Theories: Corpus\n  Creation, Annotation Reliability, and Prediction", "abstract": "  The most prominent tasks in emotion analysis are to assign emotions to texts\nand to understand how emotions manifest in language. An observation for NLP is\nthat emotions can be communicated implicitly by referring to events, appealing\nto an empathetic, intersubjective understanding of events, even without\nexplicitly mentioning an emotion name. In psychology, the class of emotion\ntheories known as appraisal theories aims at explaining the link between events\nand emotions. Appraisals can be formalized as variables that measure a\ncognitive evaluation by people living through an event that they consider\nrelevant. They include the assessment if an event is novel, if the person\nconsiders themselves to be responsible, if it is in line with the own goals,\nand many others. Such appraisals explain which emotions are developed based on\nan event, e.g., that a novel situation can induce surprise or one with\nuncertain consequences could evoke fear. We analyze the suitability of\nappraisal theories for emotion analysis in text with the goal of understanding\nif appraisal concepts can reliably be reconstructed by annotators, if they can\nbe predicted by text classifiers, and if appraisal concepts help to identify\nemotion categories. To achieve that, we compile a corpus by asking people to\ntextually describe events that triggered particular emotions and to disclose\ntheir appraisals. Then, we ask readers to reconstruct emotions and appraisals\nfrom the text. This setup allows us to measure if emotions and appraisals can\nbe recovered purely from text and provides a human baseline. Our comparison of\ntext classification methods to human annotators shows that both can reliably\ndetect emotions and appraisals with similar performance. Therefore, appraisals\nconstitute an alternative computational emotion analysis paradigm and further\nimprove the categorization of emotions in text with joint models.\n", "date": "2022-06-10", "categories": []}
{"id": "2210.03428", "title": "Missing Modality meets Meta Sampling (M3S): An Efficient Universal\n  Approach for Multimodal Sentiment Analysis with Missing Modality", "abstract": "  Multimodal sentiment analysis (MSA) is an important way of observing mental\nactivities with the help of data captured from multiple modalities. However,\ndue to the recording or transmission error, some modalities may include\nincomplete data. Most existing works that address missing modalities usually\nassume a particular modality is completely missing and seldom consider a\nmixture of missing across multiple modalities. In this paper, we propose a\nsimple yet effective meta-sampling approach for multimodal sentiment analysis\nwith missing modalities, namely Missing Modality-based Meta Sampling (M3S). To\nbe specific, M3S formulates a missing modality sampling strategy into the modal\nagnostic meta-learning (MAML) framework. M3S can be treated as an efficient\nadd-on training component on existing models and significantly improve their\nperformances on multimodal data with a mixture of missing modalities. We\nconduct experiments on IEMOCAP, SIMS and CMU-MOSI datasets, and superior\nperformance is achieved compared with recent state-of-the-art methods.\n", "date": "2022-10-07", "categories": []}
{"id": "2210.01946", "title": "Affection: Learning Affective Explanations for Real-World Visual Data", "abstract": "  In this work, we explore the emotional reactions that real-world images tend\nto induce by using natural language as the medium to express the rationale\nbehind an affective response to a given visual stimulus. To embark on this\njourney, we introduce and share with the research community a large-scale\ndataset that contains emotional reactions and free-form textual explanations\nfor 85,007 publicly available images, analyzed by 6,283 annotators who were\nasked to indicate and explain how and why they felt in a particular way when\nobserving a specific image, producing a total of 526,749 responses. Even though\nemotional reactions are subjective and sensitive to context (personal mood,\nsocial status, past experiences) - we show that there is significant common\nground to capture potentially plausible emotional responses with a large\nsupport in the subject population. In light of this crucial observation, we ask\nthe following questions: i) Can we develop multi-modal neural networks that\nprovide reasonable affective responses to real-world visual data, explained\nwith language? ii) Can we steer such methods towards producing explanations\nwith varying degrees of pragmatic language or justifying different emotional\nreactions while adapting to the underlying visual stimulus? Finally, iii) How\ncan we evaluate the performance of such methods for this novel task? With this\nwork, we take the first steps in addressing all of these questions, thus paving\nthe way for richer, more human-centric, and emotionally-aware image analysis\nsystems. Our introduced dataset and all developed methods are available on\nhttps://affective-explanations.org\n", "date": "2022-10-04", "categories": []}
{"id": "2210.00263", "title": "Fine-tuning Wav2vec for Vocal-burst Emotion Recognition", "abstract": "  The ACII Affective Vocal Bursts (A-VB) competition introduces a new topic in\naffective computing, which is understanding emotional expression using the\nnon-verbal sound of humans. We are familiar with emotion recognition via verbal\nvocal or facial expression. However, the vocal bursts such as laughs, cries,\nand signs, are not exploited even though they are very informative for behavior\nanalysis. The A-VB competition comprises four tasks that explore non-verbal\ninformation in different spaces. This technical report describes the method and\nthe result of SclabCNU Team for the tasks of the challenge. We achieved\npromising results compared to the baseline model provided by the organizers.\n", "date": "2022-10-01", "categories": []}
{"id": "2209.13914", "title": "An Efficient Multitask Learning Architecture for Affective Vocal Burst\n  Analysis", "abstract": "  Affective speech analysis is an ongoing topic of research. A relatively new\nproblem in this field is the analysis of vocal bursts, which are nonverbal\nvocalisations such as laughs or sighs. Current state-of-the-art approaches to\naddress affective vocal burst analysis are mostly based on wav2vec2 or HuBERT\nfeatures. In this paper, we investigate the use of the wav2vec successor\ndata2vec in combination with a multitask learning pipeline to tackle different\nanalysis problems at once. To assess the performance of our efficient multitask\nlearning architecture, we participate in the 2022 ACII Affective Vocal Burst\nChallenge, showing that our approach substantially outperforms the baseline\nestablished there in three different subtasks.\n", "date": "2022-09-28", "categories": []}
{"id": "2209.13999", "title": "CEFER: A Four Facets Framework based on Context and Emotion embedded\n  features for Implicit and Explicit Emotion Recognition", "abstract": "  People's conduct and reactions are driven by their emotions. Online social\nmedia is becoming a great instrument for expressing emotions in written form.\nPaying attention to the context and the entire sentence help us to detect\nemotion from texts. However, this perspective inhibits us from noticing some\nemotional words or phrases in the text, particularly when the words express an\nemotion implicitly rather than explicitly. On the other hand, focusing only on\nthe words and ignoring the context results in a distorted understanding of the\nsentence meaning and feeling. In this paper, we propose a framework that\nanalyses text at both the sentence and word levels. We name it CEFER (Context\nand Emotion embedded Framework for Emotion Recognition). Our four approach\nfacets are to extracting data by considering the entire sentence and each\nindividual word simultaneously, as well as implicit and explicit emotions. The\nknowledge gained from these data not only mitigates the impact of flaws in the\npreceding approaches but also it strengthens the feature vector. We evaluate\nseveral feature spaces using BERT family and design the CEFER based on them.\nCEFER combines the emotional vector of each word, including explicit and\nimplicit emotions, with the feature vector of each word based on context. CEFER\nperforms better than the BERT family. The experimental results demonstrate that\nidentifying implicit emotions are more challenging than detecting explicit\nemotions. CEFER, improves the accuracy of implicit emotion recognition.\nAccording to the results, CEFER perform 5% better than the BERT family in\nrecognizing explicit emotions and 3% in implicit.\n", "date": "2022-09-28", "categories": []}
{"id": "2209.14085", "title": "PTSD in the Wild: A Video Database for Studying Post-Traumatic Stress\n  Disorder Recognition in Unconstrained Environments", "abstract": "  POST-traumatic stress disorder (PTSD) is a chronic and debilitating mental\ncondition that is developed in response to catastrophic life events, such as\nmilitary combat, sexual assault, and natural disasters. PTSD is characterized\nby flashbacks of past traumatic events, intrusive thoughts, nightmares,\nhypervigilance, and sleep disturbance, all of which affect a person's life and\nlead to considerable social, occupational, and interpersonal dysfunction. The\ndiagnosis of PTSD is done by medical professionals using self-assessment\nquestionnaire of PTSD symptoms as defined in the Diagnostic and Statistical\nManual of Mental Disorders (DSM). In this paper, and for the first time, we\ncollected, annotated, and prepared for public distribution a new video database\nfor automatic PTSD diagnosis, called PTSD in the wild dataset. The database\nexhibits \"natural\" and big variability in acquisition conditions with different\npose, facial expression, lighting, focus, resolution, age, gender, race,\nocclusions and background. In addition to describing the details of the dataset\ncollection, we provide a benchmark for evaluating computer vision and machine\nlearning based approaches on PTSD in the wild dataset. In addition, we propose\nand we evaluate a deep learning based approach for PTSD detection in respect to\nthe given benchmark. The proposed approach shows very promising results.\nInterested researcher can download a copy of PTSD-in-the wild dataset from:\nhttp://www.lissi.fr/PTSD-Dataset/\n", "date": "2022-09-28", "categories": []}
{"id": "2209.07384", "title": "Self-Supervised Attention Networks and Uncertainty Loss Weighting for\n  Multi-Task Emotion Recognition on Vocal Bursts", "abstract": "  Vocal bursts play an important role in communicating affect, making them\nvaluable for improving speech emotion recognition. Here, we present our\napproach for classifying vocal bursts and predicting their emotional\nsignificance in the ACII Affective Vocal Burst Workshop & Challenge 2022\n(A-VB). We use a large self-supervised audio model as shared feature extractor\nand compare multiple architectures built on classifier chains and attention\nnetworks, combined with uncertainty loss weighting strategies. Our approach\nsurpasses the challenge baseline by a wide margin on all four tasks.\n", "date": "2022-09-15", "categories": []}
{"id": "2209.07629", "title": "Self-Relation Attention and Temporal Awareness for Emotion Recognition\n  via Vocal Burst", "abstract": "  The technical report presents our emotion recognition pipeline for\nhigh-dimensional emotion task (A-VB High) in The ACII Affective Vocal Bursts\n(A-VB) 2022 Workshop \\& Competition. Our proposed method contains three stages.\nFirstly, we extract the latent features from the raw audio signal and its\nMel-spectrogram by self-supervised learning methods. Then, the features from\nthe raw signal are fed to the self-relation attention and temporal awareness\n(SA-TA) module for learning the valuable information between these latent\nfeatures. Finally, we concatenate all the features and utilize a\nfully-connected layer to predict each emotion's score. By empirical\nexperiments, our proposed method achieves a mean concordance correlation\ncoefficient (CCC) of 0.7295 on the test set, compared to 0.5686 on the baseline\nmodel. The code of our method is available at\nhttps://github.com/linhtd812/A-VB2022.\n", "date": "2022-09-15", "categories": []}
{"id": "2209.12172", "title": "Optimal Transport-based Identity Matching for Identity-invariant Facial\n  Expression Recognition", "abstract": "  Identity-invariant facial expression recognition (FER) has been one of the\nchallenging computer vision tasks. Since conventional FER schemes do not\nexplicitly address the inter-identity variation of facial expressions, their\nneural network models still operate depending on facial identity. This paper\nproposes to quantify the inter-identity variation by utilizing pairs of similar\nexpressions explored through a specific matching process. We formulate the\nidentity matching process as an Optimal Transport (OT) problem. Specifically,\nto find pairs of similar expressions from different identities, we define the\ninter-feature similarity as a transportation cost. Then, optimal identity\nmatching to find the optimal flow with minimum transportation cost is performed\nby Sinkhorn-Knopp iteration. The proposed matching method is not only easy to\nplug in to other models, but also requires only acceptable computational\noverhead. Extensive simulations prove that the proposed FER method improves the\nPCC/CCC performance by up to 10\\% or more compared to the runner-up on wild\ndatasets. The source code and software demo are available at\nhttps://github.com/kdhht2334/ELIM_FER.\n", "date": "2022-09-25", "categories": []}
{"id": "2209.10170", "title": "FV2ES: A Fully End2End Multimodal System for Fast Yet Effective Video\n  Emotion Recognition Inference", "abstract": "  In the latest social networks, more and more people prefer to express their\nemotions in videos through text, speech, and rich facial expressions.\nMultimodal video emotion analysis techniques can help understand users' inner\nworld automatically based on human expressions and gestures in images, tones in\nvoices, and recognized natural language. However, in the existing research, the\nacoustic modality has long been in a marginal position as compared to visual\nand textual modalities. That is, it tends to be more difficult to improve the\ncontribution of the acoustic modality for the whole multimodal emotion\nrecognition task. Besides, although better performance can be obtained by\nintroducing common deep learning methods, the complex structures of these\ntraining models always result in low inference efficiency, especially when\nexposed to high-resolution and long-length videos. Moreover, the lack of a\nfully end-to-end multimodal video emotion recognition system hinders its\napplication. In this paper, we designed a fully multimodal video-to-emotion\nsystem (named FV2ES) for fast yet effective recognition inference, whose\nbenefits are threefold: (1) The adoption of the hierarchical attention method\nupon the sound spectra breaks through the limited contribution of the acoustic\nmodality and outperforms the existing models' performance on both IEMOCAP and\nCMU-MOSEI datasets; (2) the introduction of the idea of multi-scale for visual\nextraction while single-branch for inference brings higher efficiency and\nmaintains the prediction accuracy at the same time; (3) the further integration\nof data pre-processing into the aligned multimodal learning model allows the\nsignificant reduction of computational costs and storage space.\n", "date": "2022-09-21", "categories": []}
{"id": "2209.10223", "title": "Dynamic Time-Alignment of Dimensional Annotations of Emotion using\n  Recurrent Neural Networks", "abstract": "  Most automatic emotion recognition systems exploit time-continuous\nannotations of emotion to provide fine-grained descriptions of spontaneous\nexpressions as observed in real-life interactions. As emotion is rather\nsubjective, its annotation is usually performed by several annotators who\nprovide a trace for a given dimension, i.e. a time-continuous series describing\na dimension such as arousal or valence. However, annotations of the same\nexpression are rarely consistent between annotators, either in time or in\nvalue, which adds bias and delay in the trace that is used to learn predictive\nmodels of emotion. We therefore propose a method that can dynamically\ncompensate inconsistencies across annotations and synchronise the traces with\nthe corresponding acoustic features using Recurrent Neural Networks.\nExperimental evaluations were carried on several emotion data sets that include\nChinese, French, German, and Hungarian participants who interacted remotely in\neither noise-free conditions or in-the-wild. The results show that our method\ncan significantly increase inter-annotator agreement, as well as correlation\nbetween traces and audio features, for both arousal and valence. In addition,\nimprovements are obtained in the automatic prediction of these dimensions using\nsimple light-weight models, especially for valence in noise-free conditions,\nand arousal for recordings captured in-the-wild.\n", "date": "2022-09-21", "categories": []}
{"id": "2207.10299", "title": "Learn From All: Erasing Attention Consistency for Noisy Label Facial\n  Expression Recognition", "abstract": "  Noisy label Facial Expression Recognition (FER) is more challenging than\ntraditional noisy label classification tasks due to the inter-class similarity\nand the annotation ambiguity. Recent works mainly tackle this problem by\nfiltering out large-loss samples. In this paper, we explore dealing with noisy\nlabels from a new feature-learning perspective. We find that FER models\nremember noisy samples by focusing on a part of the features that can be\nconsidered related to the noisy labels instead of learning from the whole\nfeatures that lead to the latent truth. Inspired by that, we propose a novel\nErasing Attention Consistency (EAC) method to suppress the noisy samples during\nthe training process automatically. Specifically, we first utilize the flip\nsemantic consistency of facial images to design an imbalanced framework. We\nthen randomly erase input images and use flip attention consistency to prevent\nthe model from focusing on a part of the features. EAC significantly\noutperforms state-of-the-art noisy label FER methods and generalizes well to\nother tasks with a large number of classes like CIFAR100 and Tiny-ImageNet. The\ncode is available at\nhttps://github.com/zyh-uaiaaaa/Erasing-Attention-Consistency.\n", "date": "2022-07-21", "categories": []}
{"id": "2209.09666", "title": "Documenting use cases in the affective computing domain using Unified\n  Modeling Language", "abstract": "  The study of the ethical impact of AI and the design of trustworthy systems\nneeds the analysis of the scenarios where AI systems are used, which is related\nto the software engineering concept of \"use case\" and the \"intended purpose\"\nlegal term. However, there is no standard methodology for use case\ndocumentation covering the context of use, scope, functional requirements and\nrisks of an AI system. In this work, we propose a novel documentation\nmethodology for AI use cases, with a special focus on the affective computing\ndomain. Our approach builds upon an assessment of use case information needs\ndocumented in the research literature and the recently proposed European\nregulatory framework for AI. From this assessment, we adopt and adapt the\nUnified Modeling Language (UML), which has been used in the last two decades\nmostly by software engineers. Each use case is then represented by an UML\ndiagram and a structured table, and we provide a set of examples illustrating\nits application to several affective computing scenarios.\n", "date": "2022-09-19", "categories": []}
{"id": "2209.09768", "title": "An Efficient End-to-End Transformer with Progressive Tri-modal Attention\n  for Multi-modal Emotion Recognition", "abstract": "  Recent works on multi-modal emotion recognition move towards end-to-end\nmodels, which can extract the task-specific features supervised by the target\ntask compared with the two-phase pipeline. However, previous methods only model\nthe feature interactions between the textual and either acoustic and visual\nmodalities, ignoring capturing the feature interactions between the acoustic\nand visual modalities. In this paper, we propose the multi-modal end-to-end\ntransformer (ME2ET), which can effectively model the tri-modal features\ninteraction among the textual, acoustic, and visual modalities at the low-level\nand high-level. At the low-level, we propose the progressive tri-modal\nattention, which can model the tri-modal feature interactions by adopting a\ntwo-pass strategy and can further leverage such interactions to significantly\nreduce the computation and memory complexity through reducing the input token\nlength. At the high-level, we introduce the tri-modal feature fusion layer to\nexplicitly aggregate the semantic representations of three modalities. The\nexperimental results on the CMU-MOSEI and IEMOCAP datasets show that ME2ET\nachieves the state-of-the-art performance. The further in-depth analysis\ndemonstrates the effectiveness, efficiency, and interpretability of the\nproposed progressive tri-modal attention, which can help our model to achieve\nbetter performance while significantly reducing the computation and memory\ncost. Our code will be publicly available.\n", "date": "2022-09-20", "categories": []}
{"id": "2207.11679", "title": "Affective Behaviour Analysis Using Pretrained Model with Facial Priori", "abstract": "  Affective behaviour analysis has aroused researchers' attention due to its\nbroad applications. However, it is labor exhaustive to obtain accurate\nannotations for massive face images. Thus, we propose to utilize the prior\nfacial information via Masked Auto-Encoder (MAE) pretrained on unlabeled face\nimages. Furthermore, we combine MAE pretrained Vision Transformer (ViT) and\nAffectNet pretrained CNN to perform multi-task emotion recognition. We notice\nthat expression and action unit (AU) scores are pure and intact features for\nvalence-arousal (VA) regression. As a result, we utilize AffectNet pretrained\nCNN to extract expression scores concatenating with expression and AU scores\nfrom ViT to obtain the final VA features. Moreover, we also propose a\nco-training framework with two parallel MAE pretrained ViT for expression\nrecognition tasks. In order to make the two views independent, we random mask\nmost patches during the training process. Then, JS divergence is performed to\nmake the predictions of the two views as consistent as possible. The results on\nABAW4 show that our methods are effective.\n", "date": "2022-07-24", "categories": []}
{"id": "2209.08988", "title": "MSA-GCN:Multiscale Adaptive Graph Convolution Network for Gait Emotion\n  Recognition", "abstract": "  Gait emotion recognition plays a crucial role in the intelligent system. Most\nof the existing methods recognize emotions by focusing on local actions over\ntime. However, they ignore that the effective distances of different emotions\nin the time domain are different, and the local actions during walking are\nquite similar. Thus, emotions should be represented by global states instead of\nindirect local actions. To address these issues, a novel Multi Scale Adaptive\nGraph Convolution Network (MSA-GCN) is presented in this work through\nconstructing dynamic temporal receptive fields and designing multiscale\ninformation aggregation to recognize emotions. In our model, a adaptive\nselective spatial-temporal graph convolution is designed to select the\nconvolution kernel dynamically to obtain the soft spatio-temporal features of\ndifferent emotions. Moreover, a Cross-Scale mapping Fusion Mechanism (CSFM) is\ndesigned to construct an adaptive adjacency matrix to enhance information\ninteraction and reduce redundancy. Compared with previous state-of-the-art\nmethods, the proposed method achieves the best performance on two public\ndatasets, improving the mAP by 2\\%. We also conduct extensive ablations studies\nto show the effectiveness of different components in our methods.\n", "date": "2022-09-19", "categories": []}
{"id": "2209.09068", "title": "Audio-Visual Fusion for Emotion Recognition in the Valence-Arousal Space\n  Using Joint Cross-Attention", "abstract": "  Automatic emotion recognition (ER) has recently gained lot of interest due to\nits potential in many real-world applications. In this context, multimodal\napproaches have been shown to improve performance (over unimodal approaches) by\ncombining diverse and complementary sources of information, providing some\nrobustness to noisy and missing modalities. In this paper, we focus on\ndimensional ER based on the fusion of facial and vocal modalities extracted\nfrom videos, where complementary audio-visual (A-V) relationships are explored\nto predict an individual's emotional states in valence-arousal space. Most\nstate-of-the-art fusion techniques rely on recurrent networks or conventional\nattention mechanisms that do not effectively leverage the complementary nature\nof A-V modalities. To address this problem, we introduce a joint\ncross-attentional model for A-V fusion that extracts the salient features\nacross A-V modalities, that allows to effectively leverage the inter-modal\nrelationships, while retaining the intra-modal relationships. In particular, it\ncomputes the cross-attention weights based on correlation between the joint\nfeature representation and that of the individual modalities. By deploying the\njoint A-V feature representation into the cross-attention module, it helps to\nsimultaneously leverage both the intra and inter modal relationships, thereby\nsignificantly improving the performance of the system over the vanilla\ncross-attention module. The effectiveness of our proposed approach is validated\nexperimentally on challenging videos from the RECOLA and AffWild2 datasets.\nResults indicate that our joint cross-attentional A-V fusion model provides a\ncost-effective solution that can outperform state-of-the-art approaches, even\nwhen the modalities are noisy or absent.\n", "date": "2022-09-19", "categories": []}
{"id": "2209.07424", "title": "CMSBERT-CLR: Context-driven Modality Shifting BERT with Contrastive\n  Learning for linguistic, visual, acoustic Representations", "abstract": "  Multimodal sentiment analysis has become an increasingly popular research\narea as the demand for multimodal online content is growing. For multimodal\nsentiment analysis, words can have different meanings depending on the\nlinguistic context and non-verbal information, so it is crucial to understand\nthe meaning of the words accordingly. In addition, the word meanings should be\ninterpreted within the whole utterance context that includes nonverbal\ninformation. In this paper, we present a Context-driven Modality Shifting BERT\nwith Contrastive Learning for linguistic, visual, acoustic Representations\n(CMSBERT-CLR), which incorporates the whole context's non-verbal and verbal\ninformation and aligns modalities more effectively through contrastive\nlearning. First, we introduce a Context-driven Modality Shifting (CMS) to\nincorporate the non-verbal and verbal information within the whole context of\nthe sentence utterance. Then, for improving the alignment of different\nmodalities within a common embedding space, we apply contrastive learning.\nFurthermore, we use an exponential moving average parameter and label smoothing\nas optimization strategies, which can make the convergence of the network more\nstable and increase the flexibility of the alignment. In our experiments, we\ndemonstrate that our approach achieves state-of-the-art results.\n", "date": "2022-08-21", "categories": []}
{"id": "2207.14087", "title": "CubeMLP: An MLP-based Model for Multimodal Sentiment Analysis and\n  Depression Estimation", "abstract": "  Multimodal sentiment analysis and depression estimation are two important\nresearch topics that aim to predict human mental states using multimodal data.\nPrevious research has focused on developing effective fusion strategies for\nexchanging and integrating mind-related information from different modalities.\nSome MLP-based techniques have recently achieved considerable success in a\nvariety of computer vision tasks. Inspired by this, we explore multimodal\napproaches with a feature-mixing perspective in this study. To this end, we\nintroduce CubeMLP, a multimodal feature processing framework based entirely on\nMLP. CubeMLP consists of three independent MLP units, each of which has two\naffine transformations. CubeMLP accepts all relevant modality features as input\nand mixes them across three axes. After extracting the characteristics using\nCubeMLP, the mixed multimodal features are flattened for task predictions. Our\nexperiments are conducted on sentiment analysis datasets: CMU-MOSI and\nCMU-MOSEI, and depression estimation dataset: AVEC2019. The results show that\nCubeMLP can achieve state-of-the-art performance with a much lower computing\ncost.\n", "date": "2022-07-28", "categories": []}
{"id": "2209.04908", "title": "Automatic Detection of Sentimentality from Facial Expressions", "abstract": "  Emotion recognition has received considerable attention from the Computer\nVision community in the last 20 years. However, most of the research focused on\nanalyzing the six basic emotions (e.g. joy, anger, surprise), with a limited\nwork directed to other affective states. In this paper, we tackle\nsentimentality (strong feeling of heartwarming or nostalgia), a new emotional\nstate that has few works in the literature, and no guideline defining its\nfacial markers. To this end, we first collect a dataset of 4.9K videos of\nparticipants watching some sentimental and non-sentimental ads, and then we\nlabel the moments evoking sentimentality in the ads. Second, we use the\nad-level labels and the facial Action Units (AUs) activation across different\nframes for defining some weak frame-level sentimentality labels. Third, we\ntrain a Multilayer Perceptron (MLP) using the AUs activation for sentimentality\ndetection. Finally, we define two new ad-level metrics for evaluating our model\nperformance. Quantitative and qualitative results show promising results for\nsentimentality detection. To the best of our knowledge this is the first work\nto address the problem of sentimentality detection.\n", "date": "2022-09-11", "categories": []}
{"id": "2209.04112", "title": "Joint Alignment of Multi-Task Feature and Label Spaces for Emotion Cause\n  Pair Extraction", "abstract": "  Emotion cause pair extraction (ECPE), as one of the derived subtasks of\nemotion cause analysis (ECA), shares rich inter-related features with emotion\nextraction (EE) and cause extraction (CE). Therefore EE and CE are frequently\nutilized as auxiliary tasks for better feature learning, modeled via multi-task\nlearning (MTL) framework by prior works to achieve state-of-the-art (SoTA) ECPE\nresults. However, existing MTL-based methods either fail to simultaneously\nmodel the specific features and the interactive feature in between, or suffer\nfrom the inconsistency of label prediction. In this work, we consider\naddressing the above challenges for improving ECPE by performing two alignment\nmechanisms with a novel A^2Net model. We first propose a feature-task alignment\nto explicitly model the specific emotion-&cause-specific features and the\nshared interactive feature. Besides, an inter-task alignment is implemented, in\nwhich the label distance between the ECPE and the combinations of EE&CE are\nlearned to be narrowed for better label consistency. Evaluations of benchmarks\nshow that our methods outperform current best-performing systems on all ECA\nsubtasks. Further analysis proves the importance of our proposed alignment\nmechanisms for the task.\n", "date": "2022-09-09", "categories": []}
{"id": "2209.03901", "title": "Dyadic Interaction Assessment from Free-living Audio for Depression\n  Severity Assessment", "abstract": "  Psychomotor retardation in depression has been associated with speech timing\nchanges from dyadic clinical interviews. In this work, we investigate speech\ntiming features from free-living dyadic interactions. Apart from the\npossibility of continuous monitoring to complement clinical visits, a study in\nfree-living conditions would also allow inferring sociability features such as\ndyadic interaction frequency implicated in depression. We adapted a speaker\ncount estimator as a dyadic interaction detector with a specificity of 89.5%\nand a sensitivity of 86.1% in the DIHARD dataset. Using the detector, we\nobtained speech timing features from the detected dyadic interactions in\nmulti-day audio recordings of 32 participants comprised of 13 healthy\nindividuals, 11 individuals with depression, and 8 individuals with psychotic\ndisorders. The dyadic interaction frequency increased with depression severity\nin participants with no or mild depression, indicating a potential diagnostic\nmarker of depression onset. However, the dyadic interaction frequency decreased\nwith increasing depression severity for participants with moderate or severe\ndepression. In terms of speech timing features, the response time had a\nsignificant positive correlation with depression severity. Our work shows the\npotential of dyadic interaction analysis from audio recordings of free-living\nto obtain markers of depression severity.\n", "date": "2022-09-08", "categories": []}
{"id": "2209.02785", "title": "Read it to me: An emotionally aware Speech Narration Application", "abstract": "  In this work we try to perform emotional style transfer on audios. In\nparticular, MelGAN-VC architecture is explored for various emotion-pair\ntransfers. The generated audio is then classified using an LSTM-based emotion\nclassifier for audio. We find that \"sad\" audio is generated well as compared to\n\"happy\" or \"anger\" as people have similar expressions of sadness.\n", "date": "2022-09-06", "categories": []}
{"id": "2208.01369", "title": "The Face of Affective Disorders", "abstract": "  We study the statistical properties of facial behaviour altered by the\nregulation of brain arousal in the clinical domain of psychiatry. The\nunderlying mechanism is linked to the empirical interpretation of the vigilance\ncontinuum as behavioral surrogate measurement for certain states of mind.\nReferring to the classical scalp-based obtrusive measurements, we name the\npresented method Opto-Electronic Encephalography (OEG) which solely relies on\nmodern camera-based real-time signal processing and computer vision. Based upon\na stochastic representation as coherence of the face dynamics, reflecting the\nhemifacial asymmetry in emotion expressions, we demonstrate an almost flawless\ndistinction between patients and healthy controls as well as between the mental\ndisorders depression and schizophrenia and the symptom severity. In contrast to\nthe standard diagnostic process, which is time-consuming, subjective and does\nnot incorporate neurobiological data such as real-time face dynamics, the\nobjective stochastic modeling of the affective responsiveness only requires a\nfew minutes of video-based facial recordings. We also highlight the potential\nof the methodology as a causal inference model in transdiagnostic analysis to\npredict the outcome of pharmacological treatment. All results are obtained on a\nclinical longitudinal data collection with an amount of 99 patients and 43\ncontrols.\n", "date": "2022-08-02", "categories": []}
{"id": "2208.14244", "title": "Expressions Causing Differences in Emotion Recognition in Social\n  Networking Service Documents", "abstract": "  It is often difficult to correctly infer a writer's emotion from text\nexchanged online, and differences in recognition between writers and readers\ncan be problematic. In this paper, we propose a new framework for detecting\nsentences that create differences in emotion recognition between the writer and\nthe reader and for detecting the kinds of expressions that cause such\ndifferences. The proposed framework consists of a bidirectional encoder\nrepresentations from transformers (BERT)-based detector that detects sentences\ncausing differences in emotion recognition and an analysis that acquires\nexpressions that characteristically appear in such sentences. The detector,\nbased on a Japanese SNS-document dataset with emotion labels annotated by both\nthe writer and three readers of the social networking service (SNS) documents,\ndetected \"hidden-anger sentences\" with AUC = 0.772; these sentences gave rise\nto differences in the recognition of anger. Because SNS documents contain many\nsentences whose meaning is extremely difficult to interpret, by analyzing the\nsentences detected by this detector, we obtained several expressions that\nappear characteristically in hidden-anger sentences. The detected sentences and\nexpressions do not convey anger explicitly, and it is difficult to infer the\nwriter's anger, but if the implicit anger is pointed out, it becomes possible\nto guess why the writer is angry. Put into practical use, this framework would\nlikely have the ability to mitigate problems based on misunderstandings.\n", "date": "2022-08-30", "categories": []}
{"id": "2209.02604", "title": "Make Acoustic and Visual Cues Matter: CH-SIMS v2.0 Dataset and AV-Mixup\n  Consistent Module", "abstract": "  Multimodal sentiment analysis (MSA), which supposes to improve text-based\nsentiment analysis with associated acoustic and visual modalities, is an\nemerging research area due to its potential applications in Human-Computer\nInteraction (HCI). However, the existing researches observe that the acoustic\nand visual modalities contribute much less than the textual modality, termed as\ntext-predominant. Under such circumstances, in this work, we emphasize making\nnon-verbal cues matter for the MSA task. Firstly, from the resource\nperspective, we present the CH-SIMS v2.0 dataset, an extension and enhancement\nof the CH-SIMS. Compared with the original dataset, the CH-SIMS v2.0 doubles\nits size with another 2121 refined video segments with both unimodal and\nmultimodal annotations and collects 10161 unlabelled raw video segments with\nrich acoustic and visual emotion-bearing context to highlight non-verbal cues\nfor sentiment prediction. Secondly, from the model perspective, benefiting from\nthe unimodal annotations and the unsupervised data in the CH-SIMS v2.0, the\nAcoustic Visual Mixup Consistent (AV-MC) framework is proposed. The designed\nmodality mixup module can be regarded as an augmentation, which mixes the\nacoustic and visual modalities from different videos. Through drawing\nunobserved multimodal context along with the text, the model can learn to be\naware of different non-verbal contexts for sentiment prediction. Our\nevaluations demonstrate that both CH-SIMS v2.0 and AV-MC framework enables\nfurther research for discovering emotion-bearing acoustic and visual cues and\npaves the path to interpretable end-to-end HCI applications for real-world\nscenarios.\n", "date": "2022-08-22", "categories": []}
{"id": "2207.09373", "title": "Multi-Task Learning Framework for Emotion Recognition in-the-wild", "abstract": "  This paper presents our system for the Multi-Task Learning (MTL) Challenge in\nthe 4th Affective Behavior Analysis in-the-wild (ABAW) competition. We explore\nthe research problems of this challenge from three aspects: 1) For obtaining\nefficient and robust visual feature representations, we propose MAE-based\nunsupervised representation learning and IResNet/DenseNet-based supervised\nrepresentation learning methods; 2) Considering the importance of temporal\ninformation in videos, we explore three types of sequential encoders to capture\nthe temporal information, including the encoder based on transformer, the\nencoder based on LSTM, and the encoder based on GRU; 3) For modeling the\ncorrelation between these different tasks (i.e., valence, arousal, expression,\nand AU) for multi-task affective analysis, we first explore the dependency\nbetween these different tasks and propose three multi-task learning frameworks\nto model the correlations effectively. Our system achieves the performance of\n$1.7607$ on the validation dataset and $1.4361$ on the test dataset, ranking\nfirst in the MTL Challenge. The code is available at\nhttps://github.com/AIM3-RUC/ABAW4.\n", "date": "2022-07-19", "categories": []}
{"id": "2208.13549", "title": "Learning a General Clause-to-Clause Relationships for Enhancing\n  Emotion-Cause Pair Extraction", "abstract": "  Emotion-cause pair extraction (ECPE) is an emerging task aiming to extract\npotential pairs of emotions and corresponding causes from documents. Previous\napproaches have focused on modeling the pair-to-pair relationship and achieved\npromising results. However, the clause-to-clause relationship, which\nfundamentally symbolizes the underlying structure of a document, has still been\nin its research infancy. In this paper, we define a novel clause-to-clause\nrelationship. To learn it applicably, we propose a general clause-level\nencoding model named EA-GAT comprising E-GAT and Activation Sort. E-GAT is\ndesigned to aggregate information from different types of clauses; Activation\nSort leverages the individual emotion/cause prediction and the sort-based\nmapping to propel the clause to a more favorable representation. Since EA-GAT\nis a clause-level encoding model, it can be broadly integrated with any\nprevious approach. Experimental results show that our approach has a\nsignificant advantage over all current approaches on the Chinese and English\nbenchmark corpus, with an average of $2.1\\%$ and $1.03\\%$.\n", "date": "2022-08-29", "categories": []}
{"id": "2208.13954", "title": "Video-based Cross-modal Auxiliary Network for Multimodal Sentiment\n  Analysis", "abstract": "  Multimodal sentiment analysis has a wide range of applications due to its\ninformation complementarity in multimodal interactions. Previous works focus\nmore on investigating efficient joint representations, but they rarely consider\nthe insufficient unimodal features extraction and data redundancy of multimodal\nfusion. In this paper, a Video-based Cross-modal Auxiliary Network (VCAN) is\nproposed, which is comprised of an audio features map module and a cross-modal\nselection module. The first module is designed to substantially increase\nfeature diversity in audio feature extraction, aiming to improve classification\naccuracy by providing more comprehensive acoustic representations. To empower\nthe model to handle redundant visual features, the second module is addressed\nto efficiently filter the redundant visual frames during integrating\naudiovisual data. Moreover, a classifier group consisting of several image\nclassification networks is introduced to predict sentiment polarities and\nemotion categories. Extensive experimental results on RAVDESS, CMU-MOSI, and\nCMU-MOSEI benchmarks indicate that VCAN is significantly superior to the\nstate-of-the-art methods for improving the classification accuracy of\nmultimodal sentiment analysis.\n", "date": "2022-08-30", "categories": []}
{"id": "2208.13386", "title": "Affective Manifolds: Modeling Machine's Mind to Like, Dislike, Enjoy,\n  Suffer, Worry, Fear, and Feel Like A Human", "abstract": "  After the development of different machine learning and manifold learning\nalgorithms, it may be a good time to put them together to make a powerful mind\nfor machine. In this work, we propose affective manifolds as components of a\nmachine's mind. Every affective manifold models a characteristic group of mind\nand contains multiple states. We define the machine's mind as a set of\naffective manifolds. We use a learning model for mapping the input signals to\nthe embedding space of affective manifold. Using this mapping, a machine or a\nrobot takes an input signal and can react emotionally to it. We use deep metric\nlearning, with Siamese network, and propose a loss function for affective\nmanifold learning. We define margins between states based on the psychological\nand philosophical studies. Using triplets of instances, we train the network to\nminimize the variance of every state and have the desired distances between\nstates. We show that affective manifolds can have various applications for\nmachine-machine and human-machine interactions. Some simulations are also\nprovided for verification of the proposed method. It is possible to have as\nmany affective manifolds as required in machine's mind. More affective\nmanifolds in the machine's mind can make it more realistic and effective. This\npaper opens the door; we invite the researchers from various fields of science\nto propose more affective manifolds to be inserted in machine's mind.\n", "date": "2022-08-29", "categories": []}
{"id": "2208.12622", "title": "Play with Emotion: Affect-Driven Reinforcement Learning", "abstract": "  This paper introduces a paradigm shift by viewing the task of affect modeling\nas a reinforcement learning (RL) process. According to the proposed paradigm,\nRL agents learn a policy (i.e. affective interaction) by attempting to maximize\na set of rewards (i.e. behavioral and affective patterns) via their experience\nwith their environment (i.e. context). Our hypothesis is that RL is an\neffective paradigm for interweaving affect elicitation and manifestation with\nbehavioral and affective demonstrations. Importantly, our second\nhypothesis-building on Damasio's somatic marker hypothesis-is that emotion can\nbe the facilitator of decision-making. We test our hypotheses in a racing game\nby training Go-Blend agents to model human demonstrations of arousal and\nbehavior; Go-Blend is a modified version of the Go-Explore algorithm which has\nrecently showcased supreme performance in hard exploration tasks. We first vary\nthe arousal-based reward function and observe agents that can effectively\ndisplay a palette of affect and behavioral patterns according to the specified\nreward. Then we use arousal-based state selection mechanisms in order to bias\nthe strategies that Go-Blend explores. Our findings suggest that Go-Blend not\nonly is an efficient affect modeling paradigm but, more importantly,\naffect-driven RL improves exploration and yields higher performing agents,\nvalidating Damasio's hypothesis in the domain of games.\n", "date": "2022-08-26", "categories": []}
{"id": "2208.11893", "title": "Cross-Modality Gated Attention Fusion for Multimodal Sentiment Analysis", "abstract": "  Multimodal sentiment analysis is an important research task to predict the\nsentiment score based on the different modality data from a specific opinion\nvideo. Many previous pieces of research have proved the significance of\nutilizing the shared and unique information across different modalities.\nHowever, the high-order combined signals from multimodal data would also help\nextract satisfied representations. In this paper, we propose CMGA, a\nCross-Modality Gated Attention fusion model for MSA that tends to make adequate\ninteraction across different modality pairs. CMGA also adds a forget gate to\nfilter the noisy and redundant signals introduced in the interaction procedure.\nWe experiment on two benchmark datasets in MSA, MOSI, and MOSEI, illustrating\nthe performance of CMGA over several baseline models. We also conduct the\nablation study to demonstrate the function of different components inside CMGA.\n", "date": "2022-08-25", "categories": []}
{"id": "2208.12238", "title": "Supervised Contrastive Learning for Affect Modelling", "abstract": "  Affect modeling is viewed, traditionally, as the process of mapping\nmeasurable affect manifestations from multiple modalities of user input to\naffect labels. That mapping is usually inferred through end-to-end\n(manifestation-to-affect) machine learning processes. What if, instead, one\ntrains general, subject-invariant representations that consider affect\ninformation and then uses such representations to model affect? In this paper\nwe assume that affect labels form an integral part, and not just the training\nsignal, of an affect representation and we explore how the recent paradigm of\ncontrastive learning can be employed to discover general high-level\naffect-infused representations for the purpose of modeling affect. We introduce\nthree different supervised contrastive learning approaches for training\nrepresentations that consider affect information. In this initial study we test\nthe proposed methods for arousal prediction in the RECOLA dataset based on user\ninformation from multiple modalities. Results demonstrate the representation\ncapacity of contrastive learning and its efficiency in boosting the accuracy of\naffect models. Beyond their evidenced higher performance compared to end-to-end\narousal classification, the resulting representations are general-purpose and\nsubject-agnostic, as training is guided though general affect information\navailable in any multimodal corpus.\n", "date": "2022-08-25", "categories": []}
{"id": "2207.11875", "title": "Seeking Subjectivity in Visual Emotion Distribution Learning", "abstract": "  Visual Emotion Analysis (VEA), which aims to predict people's emotions\ntowards different visual stimuli, has become an attractive research topic\nrecently. Rather than a single label classification task, it is more rational\nto regard VEA as a Label Distribution Learning (LDL) problem by voting from\ndifferent individuals. Existing methods often predict visual emotion\ndistribution in a unified network, neglecting the inherent subjectivity in its\ncrowd voting process. In psychology, the \\textit{Object-Appraisal-Emotion}\nmodel has demonstrated that each individual's emotion is affected by his/her\nsubjective appraisal, which is further formed by the affective memory. Inspired\nby this, we propose a novel \\textit{Subjectivity Appraise-and-Match Network\n(SAMNet)} to investigate the subjectivity in visual emotion distribution. To\ndepict the diversity in crowd voting process, we first propose the\n\\textit{Subjectivity Appraising} with multiple branches, where each branch\nsimulates the emotion evocation process of a specific individual. Specifically,\nwe construct the affective memory with an attention-based mechanism to preserve\neach individual's unique emotional experience. A subjectivity loss is further\nproposed to guarantee the divergence between different individuals. Moreover,\nwe propose the \\textit{Subjectivity Matching} with a matching loss, aiming at\nassigning unordered emotion labels to ordered individual predictions in a\none-to-one correspondence with the Hungarian algorithm. Extensive experiments\nand comparisons are conducted on public visual emotion distribution datasets,\nand the results demonstrate that the proposed SAMNet consistently outperforms\nthe state-of-the-art methods. Ablation study verifies the effectiveness of our\nmethod and visualization proves its interpretability.\n", "date": "2022-07-25", "categories": []}
{"id": "2208.08909", "title": "\"Are you okay, honey?\": Recognizing Emotions among Couples Managing\n  Diabetes in Daily Life using Multimodal Real-World Smartwatch Data", "abstract": "  Couples generally manage chronic diseases together and the management takes\nan emotional toll on both patients and their romantic partners. Consequently,\nrecognizing the emotions of each partner in daily life could provide an insight\ninto their emotional well-being in chronic disease management. Currently, the\nprocess of assessing each partner's emotions is manual, time-intensive, and\ncostly. Despite the existence of works on emotion recognition among couples,\nnone of these works have used data collected from couples' interactions in\ndaily life. In this work, we collected 85 hours (1,021 5-minute samples) of\nreal-world multimodal smartwatch sensor data (speech, heart rate,\naccelerometer, and gyroscope) and self-reported emotion data (n=612) from 26\npartners (13 couples) managing diabetes mellitus type 2 in daily life. We\nextracted physiological, movement, acoustic, and linguistic features, and\ntrained machine learning models (support vector machine and random forest) to\nrecognize each partner's self-reported emotions (valence and arousal). Our\nresults from the best models (balanced accuracies of 63.8% and 78.1% for\narousal and valence respectively) are better than chance and our prior work\nthat also used data from German-speaking, Swiss-based couples, albeit, in the\nlab. This work contributes toward building automated emotion recognition\nsystems that would eventually enable partners to monitor their emotions in\ndaily life and enable the delivery of interventions to improve their emotional\nwell-being.\n", "date": "2022-08-16", "categories": []}
{"id": "2208.10491", "title": "Improving Speech Emotion Recognition Through Focus and Calibration\n  Attention Mechanisms", "abstract": "  Attention has become one of the most commonly used mechanisms in deep\nlearning approaches. The attention mechanism can help the system focus more on\nthe feature space's critical regions. For example, high amplitude regions can\nplay an important role for Speech Emotion Recognition (SER). In this paper, we\nidentify misalignments between the attention and the signal amplitude in the\nexisting multi-head self-attention. To improve the attention area, we propose\nto use a Focus-Attention (FA) mechanism and a novel Calibration-Attention (CA)\nmechanism in combination with the multi-head self-attention. Through the FA\nmechanism, the network can detect the largest amplitude part in the segment. By\nemploying the CA mechanism, the network can modulate the information flow by\nassigning different weights to each attention head and improve the utilization\nof surrounding contexts. To evaluate the proposed method, experiments are\nperformed with the IEMOCAP and RAVDESS datasets. Experimental results show that\nthe proposed framework significantly outperforms the state-of-the-art\napproaches on both datasets.\n", "date": "2022-08-21", "categories": []}
{"id": "2207.00907", "title": "Emotion Analysis using Multi-Layered Networks for Graphical\n  Representation of Tweets", "abstract": "  Anticipating audience reaction towards a certain piece of text is integral to\nseveral facets of society ranging from politics, research, and commercial\nindustries. Sentiment analysis (SA) is a useful natural language processing\n(NLP) technique that utilizes both lexical/statistical and deep learning\nmethods to determine whether different sized texts exhibit a positive,\nnegative, or neutral emotion. However, there is currently a lack of tools that\ncan be used to analyse groups of independent texts and extract the primary\nemotion from the whole set. Therefore, the current paper proposes a novel\nalgorithm referred to as the Multi-Layered Tweet Analyzer (MLTA) that\ngraphically models social media text using multi-layered networks (MLNs) in\norder to better encode relationships across independent sets of tweets. Graph\nstructures are capable of capturing meaningful relationships in complex\necosystems compared to other representation methods. State of the art Graph\nNeural Networks (GNNs) are used to extract information from the Tweet-MLN and\nmake predictions based on the extracted graph features. Results show that not\nonly does the MLTA predict from a larger set of possible emotions, delivering a\nmore accurate sentiment compared to the standard positive, negative or neutral,\nit also allows for accurate group-level predictions of Twitter data.\n", "date": "2022-07-02", "categories": []}
{"id": "2207.10003", "title": "BYEL : Bootstrap Your Emotion Latent", "abstract": "  With the improved performance of deep learning, the number of studies trying\nto apply deep learning to human emotion analysis is increasing rapidly. But\neven with this trend going on, it is still difficult to obtain high-quality\nimages and annotations. For this reason, the Learning from Synthetic Data (LSD)\nChallenge, which learns from synthetic images and infers from real images, is\none of the most interesting areas. In general, Domain Adaptation methods are\nwidely used to address LSD challenges, but there is a limitation that target\ndomains (real images) are still needed. Focusing on these limitations, we\npropose a framework Bootstrap Your Emotion Latent (BYEL), which uses only\nsynthetic images in training. BYEL is implemented by adding Emotion Classifiers\nand Emotion Vector Subtraction to the BYOL framework that performs well in\nSelf-Supervised Representation Learning. We train our framework using synthetic\nimages generated from the Aff-wild2 dataset and evaluate it using real images\nfrom the Aff-wild2 dataset. The result shows that our framework (0.3084)\nperforms 2.8% higher than the baseline (0.3) on the macro F1 score metric.\n", "date": "2022-07-20", "categories": []}
{"id": "2208.10335", "title": "Intensity-Aware Loss for Dynamic Facial Expression Recognition in the\n  Wild", "abstract": "  Compared with the image-based static facial expression recognition (SFER)\ntask, the dynamic facial expression recognition (DFER) task based on video\nsequences is closer to the natural expression recognition scene. However, DFER\nis often more challenging. One of the main reasons is that video sequences\noften contain frames with different expression intensities, especially for the\nfacial expressions in the real-world scenarios, while the images in SFER\nfrequently present uniform and high expression intensities. However, if the\nexpressions with different intensities are treated equally, the features\nlearned by the networks will have large intra-class and small inter-class\ndifferences, which is harmful to DFER. To tackle this problem, we propose the\nglobal convolution-attention block (GCA) to rescale the channels of the feature\nmaps. In addition, we introduce the intensity-aware loss (IAL) in the training\nprocess to help the network distinguish the samples with relatively low\nexpression intensities. Experiments on two in-the-wild dynamic facial\nexpression datasets (i.e., DFEW and FERV39k) indicate that our method\noutperforms the state-of-the-art DFER approaches. The source code will be made\npublicly available.\n", "date": "2022-08-19", "categories": []}
{"id": "2208.09269", "title": "Feature Selection Enhancement and Feature Space Visualization for\n  Speech-Based Emotion Recognition", "abstract": "  Robust speech emotion recognition relies on the quality of the speech\nfeatures. We present speech features enhancement strategy that improves speech\nemotion recognition. We used the INTERSPEECH 2010 challenge feature-set. We\nidentified subsets from the features set and applied Principle Component\nAnalysis to the subsets. Finally, the features are fused horizontally. The\nresulting feature set is analyzed using t-distributed neighbour embeddings\n(t-SNE) before the application of features for emotion recognition. The method\nis compared with the state-of-the-art methods used in the literature. The\nempirical evidence is drawn using two well-known datasets: Emotional Speech\nDataset (EMO-DB) and Ryerson Audio-Visual Database of Emotional Speech and Song\n(RAVDESS) for two languages, German and English, respectively. Our method\nachieved an average recognition gain of 11.5\\% for six out of seven emotions\nfor the EMO-DB dataset, and 13.8\\% for seven out of eight emotions for the\nRAVDESS dataset as compared to the baseline study.\n", "date": "2022-08-19", "categories": []}
{"id": "2207.06180", "title": "Multi-modal Depression Estimation based on Sub-attentional Fusion", "abstract": "  Failure to timely diagnose and effectively treat depression leads to over 280\nmillion people suffering from this psychological disorder worldwide. The\ninformation cues of depression can be harvested from diverse heterogeneous\nresources, e.g., audio, visual, and textual data, raising demand for new\neffective multi-modal fusion approaches for automatic estimation. In this work,\nwe tackle the task of automatically identifying depression from multi-modal\ndata and introduce a sub-attention mechanism for linking heterogeneous\ninformation while leveraging Convolutional Bidirectional LSTM as our backbone.\nTo validate this idea, we conduct extensive experiments on the public DAIC-WOZ\nbenchmark for depression assessment featuring different evaluation modes and\ntaking gender-specific biases into account. The proposed model yields effective\nresults with 0.89 precision and 0.70 F1-score in detecting major depression and\n4.92 MAE in estimating the severity. Our attention-based fusion module\nconsistently outperforms conventional late fusion approaches and achieves\ncompetitive performance compared to the previously published depression\nestimation frameworks, while learning to diagnose the disorder end-to-end and\nrelying on far fewer preprocessing steps.\n", "date": "2022-07-13", "categories": []}
{"id": "2208.08106", "title": "Disentangling Identity and Pose for Facial Expression Recognition", "abstract": "  Facial expression recognition (FER) is a challenging problem because the\nexpression component is always entangled with other irrelevant factors, such as\nidentity and head pose. In this work, we propose an identity and pose\ndisentangled facial expression recognition (IPD-FER) model to learn more\ndiscriminative feature representation. We regard the holistic facial\nrepresentation as the combination of identity, pose and expression. These three\ncomponents are encoded with different encoders. For identity encoder, a well\npre-trained face recognition model is utilized and fixed during training, which\nalleviates the restriction on specific expression training data in previous\nworks and makes the disentanglement practicable on in-the-wild datasets. At the\nsame time, the pose and expression encoder are optimized with corresponding\nlabels. Combining identity and pose feature, a neutral face of input individual\nshould be generated by the decoder. When expression feature is added, the input\nimage should be reconstructed. By comparing the difference between synthesized\nneutral and expressional images of the same individual, the expression\ncomponent is further disentangled from identity and pose. Experimental results\nverify the effectiveness of our method on both lab-controlled and in-the-wild\ndatabases and we achieve state-of-the-art recognition performance.\n", "date": "2022-08-17", "categories": []}
{"id": "2207.06958", "title": "Proceedings of the ICML 2022 Expressive Vocalizations Workshop and\n  Competition: Recognizing, Generating, and Personalizing Vocal Bursts", "abstract": "  This is the Proceedings of the ICML Expressive Vocalization (ExVo)\nCompetition. The ExVo competition focuses on understanding and generating vocal\nbursts: laughs, gasps, cries, and other non-verbal vocalizations that are\ncentral to emotional expression and communication. ExVo 2022, included three\ncompetition tracks using a large-scale dataset of 59,201 vocalizations from\n1,702 speakers. The first, ExVo-MultiTask, requires participants to train a\nmulti-task model to recognize expressed emotions and demographic traits from\nvocal bursts. The second, ExVo-Generate, requires participants to train a\ngenerative model that produces vocal bursts conveying ten different emotions.\nThe third, ExVo-FewShot, requires participants to leverage few-shot learning\nincorporating speaker identity to train a model for the recognition of 10\nemotions conveyed by vocal bursts.\n", "date": "2022-07-14", "categories": []}
{"id": "2207.05577", "title": "Learning from Label Relationships in Human Affect", "abstract": "  Human affect and mental state estimation in an automated manner, face a\nnumber of difficulties, including learning from labels with poor or no temporal\nresolution, learning from few datasets with little data (often due to\nconfidentiality constraints) and, (very) long, in-the-wild videos. For these\nreasons, deep learning methodologies tend to overfit, that is, arrive at latent\nrepresentations with poor generalisation performance on the final regression\ntask. To overcome this, in this work, we introduce two complementary\ncontributions. First, we introduce a novel relational loss for multilabel\nregression and ordinal problems that regularises learning and leads to better\ngeneralisation. The proposed loss uses label vector inter-relational\ninformation to learn better latent representations by aligning batch label\ndistances to the distances in the latent feature space. Second, we utilise a\ntwo-stage attention architecture that estimates a target for each clip by using\nfeatures from the neighbouring clips as temporal context. We evaluate the\nproposed methodology on both continuous affect and schizophrenia severity\nestimation problems, as there are methodological and contextual parallels\nbetween the two. Experimental results demonstrate that the proposed methodology\noutperforms all baselines. In the domain of schizophrenia, the proposed\nmethodology outperforms previous state-of-the-art by a large margin, achieving\na PCC of up to 78% performance close to that of human experts (85%) and much\nhigher than previous works (uplift of up to 40%). In the case of affect\nrecognition, we outperform previous vision-based methods in terms of CCC on\nboth the OMG and the AMIGOS datasets. Specifically for AMIGOS, we outperform\nprevious SoTA CCC for both arousal and valence by 9% and 13% respectively, and\nin the OMG dataset we outperform previous vision works by up to 5% for both\narousal and valence.\n", "date": "2022-07-12", "categories": []}
{"id": "2208.07178", "title": "Computational Empathy Counteracts the Negative Effects of Anger on\n  Creative Problem Solving", "abstract": "  How does empathy influence creative problem solving? We introduce a\ncomputational empathy intervention based on context-specific affective mimicry\nand perspective taking by a virtual agent appearing in the form of a\nwell-dressed polar bear. In an online experiment with 1,006 participants\nrandomly assigned to an emotion elicitation intervention (with a control\nelicitation condition and anger elicitation condition) and a computational\nempathy intervention (with a control virtual agent and an empathic virtual\nagent), we examine how anger and empathy influence participants' performance in\nsolving a word game based on Wordle. We find participants who are assigned to\nthe anger elicitation condition perform significantly worse on multiple\nperformance metrics than participants assigned to the control condition.\nHowever, we find the empathic virtual agent counteracts the drop in performance\ninduced by the anger condition such that participants assigned to both the\nempathic virtual agent and the anger condition perform no differently than\nparticipants in the control elicitation condition and significantly better than\nparticipants assigned to the control virtual agent and the anger elicitation\ncondition. While empathy reduces the negative effects of anger, we do not find\nevidence that the empathic virtual agent influences performance of participants\nwho are assigned to the control elicitation condition. By introducing a\nframework for computational empathy interventions and conducting a two-by-two\nfactorial design randomized experiment, we provide rigorous, empirical evidence\nthat computational empathy can counteract the negative effects of anger on\ncreative problem solving.\n", "date": "2022-08-15", "categories": []}
{"id": "2208.07221", "title": "CIAO! A Contrastive Adaptation Mechanism for Non-Universal Facial\n  Expression Recognition", "abstract": "  Current facial expression recognition systems demand an expensive re-training\nroutine when deployed to different scenarios than they were trained for.\nBiasing them towards learning specific facial characteristics, instead of\nperforming typical transfer learning methods, might help these systems to\nmaintain high performance in different tasks, but with a reduced training\neffort. In this paper, we propose Contrastive Inhibitory Adaptati On (CIAO), a\nmechanism that adapts the last layer of facial encoders to depict specific\naffective characteristics on different datasets. CIAO presents an improvement\nin facial expression recognition performance over six different datasets with\nvery unique affective representations, in particular when compared with\nstate-of-the-art models. In our discussions, we make an in-depth analysis of\nhow the learned high-level facial features are represented, and how they\ncontribute to each individual dataset's characteristics. We finalize our study\nby discussing how CIAO positions itself within the range of recent findings on\nnon-universal facial expressions perception, and its impact on facial\nexpression recognition research.\n", "date": "2022-08-10", "categories": []}
{"id": "2208.03051", "title": "Hybrid Multimodal Feature Extraction, Mining and Fusion for Sentiment\n  Analysis", "abstract": "  In this paper, we present our solutions for the Multimodal Sentiment Analysis\nChallenge (MuSe) 2022, which includes MuSe-Humor, MuSe-Reaction and MuSe-Stress\nSub-challenges. The MuSe 2022 focuses on humor detection, emotional reactions\nand multimodal emotional stress utilizing different modalities and data sets.\nIn our work, different kinds of multimodal features are extracted, including\nacoustic, visual, text and biological features. These features are fused by\nTEMMA and GRU with self-attention mechanism frameworks. In this paper, 1)\nseveral new audio features, facial expression features and paragraph-level text\nembeddings are extracted for accuracy improvement. 2) we substantially improve\nthe accuracy and reliability of multimodal sentiment prediction by mining and\nblending the multimodal features. 3) effective data augmentation strategies are\napplied in model training to alleviate the problem of sample imbalance and\nprevent the model from learning biased subject characters. For the MuSe-Humor\nsub-challenge, our model obtains the AUC score of 0.8932. For the MuSe-Reaction\nsub-challenge, the Pearson's Correlations Coefficient of our approach on the\ntest set is 0.3879, which outperforms all other participants. For the\nMuSe-Stress sub-challenge, our approach outperforms the baseline in both\narousal and valence on the test dataset, reaching a final combined result of\n0.5151.\n", "date": "2022-08-05", "categories": []}
{"id": "2208.03473", "title": "Learning Human Cognitive Appraisal Through Reinforcement Memory Unit", "abstract": "  We propose a novel memory-enhancing mechanism for recurrent neural networks\nthat exploits the effect of human cognitive appraisal in sequential assessment\ntasks. We conceptualize the memory-enhancing mechanism as Reinforcement Memory\nUnit (RMU) that contains an appraisal state together with two positive and\nnegative reinforcement memories. The two reinforcement memories are decayed or\nstrengthened by stronger stimulus. Thereafter the appraisal state is updated\nthrough the competition of positive and negative reinforcement memories.\nTherefore, RMU can learn the appraisal variation under violent changing of the\nstimuli for estimating human affective experience. As shown in the experiments\nof video quality assessment and video quality of experience tasks, the proposed\nreinforcement memory unit achieves superior performance among recurrent neural\nnetworks, that demonstrates the effectiveness of RMU for modeling human\ncognitive appraisal.\n", "date": "2022-08-06", "categories": []}
{"id": "2207.08002", "title": "EEG2Vec: Learning Affective EEG Representations via Variational\n  Autoencoders", "abstract": "  There is a growing need for sparse representational formats of human\naffective states that can be utilized in scenarios with limited computational\nmemory resources. We explore whether representing neural data, in response to\nemotional stimuli, in a latent vector space can serve to both predict emotional\nstates as well as generate synthetic EEG data that are participant- and/or\nemotion-specific. We propose a conditional variational autoencoder based\nframework, EEG2Vec, to learn generative-discriminative representations from EEG\ndata. Experimental results on affective EEG recording datasets demonstrate that\nour model is suitable for unsupervised EEG modeling, classification of three\ndistinct emotion categories (positive, neutral, negative) based on the latent\nrepresentation achieves a robust performance of 68.49%, and generated synthetic\nEEG sequences resemble real EEG data inputs to particularly reconstruct\nlow-frequency signal components. Our work advances areas where affective EEG\nrepresentations can be useful in e.g., generating artificial (labeled) training\ndata or alleviating manual feature extraction, and provide efficiency for\nmemory constrained edge computing applications.\n", "date": "2022-07-16", "categories": []}
{"id": "2207.09012", "title": "SS-MFAR : Semi-supervised Multi-task Facial Affect Recognition", "abstract": "  Automatic affect recognition has applications in many areas such as\neducation, gaming, software development, automotives, medical care, etc. but it\nis non trivial task to achieve appreciable performance on in-the-wild data\nsets. In-the-wild data sets though represent real-world scenarios better than\nsynthetic data sets, the former ones suffer from the problem of incomplete\nlabels. Inspired by semi-supervised learning, in this paper, we introduce our\nsubmission to the Multi-Task-Learning Challenge at the 4th Affective Behavior\nAnalysis in-the-wild (ABAW) 2022 Competition. The three tasks that are\nconsidered in this challenge are valence-arousal(VA) estimation, classification\nof expressions into 6 basic (anger, disgust, fear, happiness, sadness,\nsurprise), neutral, and the 'other' category and 12 action units(AU) numbered\nAU-{1,2,4,6,7,10,12,15,23,24,25,26}. Our method Semi-supervised Multi-task\nFacial Affect Recognition titled SS-MFAR uses a deep residual network with task\nspecific classifiers for each of the tasks along with adaptive thresholds for\neach expression class and semi-supervised learning for the incomplete labels.\nSource code is available at https://github.com/1980x/ABAW2022DMACS.\n", "date": "2022-07-19", "categories": []}
{"id": "2208.02953", "title": "A Novel Enhanced Convolution Neural Network with Extreme Learning\n  Machine: Facial Emotional Recognition in Psychology Practices", "abstract": "  Facial emotional recognition is one of the essential tools used by\nrecognition psychology to diagnose patients. Face and facial emotional\nrecognition are areas where machine learning is excelling. Facial Emotion\nRecognition in an unconstrained environment is an open challenge for digital\nimage processing due to different environments, such as lighting conditions,\npose variation, yaw motion, and occlusions. Deep learning approaches have shown\nsignificant improvements in image recognition. However, accuracy and time still\nneed improvements. This research aims to improve facial emotion recognition\naccuracy during the training session and reduce processing time using a\nmodified Convolution Neural Network Enhanced with Extreme Learning Machine\n(CNNEELM). The system entails (CNNEELM) improving the accuracy in image\nregistration during the training session. Furthermore, the system recognizes\nsix facial emotions happy, sad, disgust, fear, surprise, and neutral with the\nproposed CNNEELM model. The study shows that the overall facial emotion\nrecognition accuracy is improved by 2% than the state of art solutions with a\nmodified Stochastic Gradient Descent (SGD) technique. With the Extreme Learning\nMachine (ELM) classifier, the processing time is brought down to 65ms from\n113ms, which can smoothly classify each frame from a video clip at 20fps. With\nthe pre-trained InceptionV3 model, the proposed CNNEELM model is trained with\nJAFFE, CK+, and FER2013 expression datasets. The simulation results show\nsignificant improvements in accuracy and processing time, making the model\nsuitable for the video analysis process. Besides, the study solves the issue of\nthe large processing time required to process the facial images.\n", "date": "2022-08-05", "categories": []}
{"id": "2208.00877", "title": "Self-supervised Group Meiosis Contrastive Learning for EEG-Based Emotion\n  Recognition", "abstract": "  The progress of EEG-based emotion recognition has received widespread\nattention from the fields of human-machine interactions and cognitive science\nin recent years. However, how to recognize emotions with limited labels has\nbecome a new research and application bottleneck. To address the issue, this\npaper proposes a Self-supervised Group Meiosis Contrastive learning framework\n(SGMC) based on the stimuli consistent EEG signals in human being. In the SGMC,\na novel genetics-inspired data augmentation method, named Meiosis, is\ndeveloped. It takes advantage of the alignment of stimuli among the EEG samples\nin a group for generating augmented groups by pairing, cross exchanging, and\nseparating. And the model adopts a group projector to extract group-level\nfeature representations from group EEG samples triggered by the same emotion\nvideo stimuli. Then contrastive learning is employed to maximize the similarity\nof group-level representations of augmented groups with the same stimuli. The\nSGMC achieves the state-of-the-art emotion recognition results on the publicly\navailable DEAP dataset with an accuracy of 94.72% and 95.68% in valence and\narousal dimensions, and also reaches competitive performance on the public SEED\ndataset with an accuracy of 94.04%. It is worthy of noting that the SGMC shows\nsignificant performance even when using limited labels. Moreover, the results\nof feature visualization suggest that the model might have learned video-level\nemotion-related feature representations to improve emotion recognition. And the\neffects of group size are further evaluated in the hyper parametric analysis.\nFinally, a control experiment and ablation study are carried out to examine the\nrationality of architecture. The code is provided publicly online.\n", "date": "2022-07-12", "categories": []}
{"id": "2208.01041", "title": "Voice Analysis for Stress Detection and Application in Virtual Reality\n  to Improve Public Speaking in Real-time: A Review", "abstract": "  Stress during public speaking is common and adversely affects performance and\nself-confidence. Extensive research has been carried out to develop various\nmodels to recognize emotional states. However, minimal research has been\nconducted to detect stress during public speaking in real time using voice\nanalysis. In this context, the current review showed that the application of\nalgorithms was not properly explored and helped identify the main obstacles in\ncreating a suitable testing environment while accounting for current\ncomplexities and limitations. In this paper, we present our main idea and\npropose a stress detection computational algorithmic model that could be\nintegrated into a Virtual Reality (VR) application to create an intelligent\nvirtual audience for improving public speaking skills. The developed model,\nwhen integrated with VR, will be able to detect excessive stress in real time\nby analysing voice features correlated to physiological parameters indicative\nof stress and help users gradually control excessive stress and improve public\nspeaking performance\n", "date": "2022-08-01", "categories": []}
{"id": "2208.00544", "title": "Analysis of Semi-Supervised Methods for Facial Expression Recognition", "abstract": "  Training deep neural networks for image recognition often requires\nlarge-scale human annotated data. To reduce the reliance of deep neural\nsolutions on labeled data, state-of-the-art semi-supervised methods have been\nproposed in the literature. Nonetheless, the use of such semi-supervised\nmethods has been quite rare in the field of facial expression recognition\n(FER). In this paper, we present a comprehensive study on recently proposed\nstate-of-the-art semi-supervised learning methods in the context of FER. We\nconduct comparative study on eight semi-supervised learning methods, namely\nPi-Model, Pseudo-label, Mean-Teacher, VAT, MixMatch, ReMixMatch, UDA, and\nFixMatch, on three FER datasets (FER13, RAF-DB, and AffectNet), when various\namounts of labeled samples are used. We also compare the performance of these\nmethods against fully-supervised training. Our study shows that when training\nexisting semi-supervised methods on as little as 250 labeled samples per class\ncan yield comparable performances to that of fully-supervised methods trained\non the full labeled datasets. To facilitate further research in this area, we\nmake our code publicly available at: https://github.com/ShuvenduRoy/SSL_FER\n", "date": "2022-07-31", "categories": []}
{"id": "2208.00737", "title": "e-Genia3 An AgentSpeak extension for empathic agents", "abstract": "  In this paper, we present e-Genia3 an extension of AgentSpeak to provide\nsupport to the development of empathic agents. The new extension modifies the\nagent's reasoning processes to select plans according to the analyzed event and\nthe affective state and personality of the agent. In addition, our proposal\nallows a software agent to simulate the distinction between self and other\nagents through two different event appraisal processes: the empathic appraisal\nprocess, for eliciting emotions as a response to other agents emotions, and the\nregular affective appraisal process for other non-empathic affective events.\nThe empathic regulation process adapts the elicited empathic emotion based on\nintrapersonal factors (e.g., the agent's personality and affective memory) and\ninterpersonal characteristics of the agent (e.g., the affective link between\nthe agents). The use of a memory of past events and their corresponding\nelicited emotions allows the maintaining of an affective link to support\nlong-term empathic interaction between agents.\n", "date": "2022-08-01", "categories": []}
{"id": "2207.14640", "title": "EmoSens: Emotion Recognition based on Sensor data analysis using\n  LightGBM", "abstract": "  Smart wearables have played an integral part in our day to day life. From\nrecording ECG signals to analysing body fat composition, the smart wearables\ncan do it all. The smart devices encompass various sensors which can be\nemployed to derive meaningful information regarding the user's physical and\npsychological conditions. Our approach focuses on employing such sensors to\nidentify and obtain the variations in the mood of a user at a given instance\nthrough the use of supervised machine learning techniques. The study examines\nthe performance of various supervised learning models such as Decision Trees,\nRandom Forests, XGBoost, LightGBM on the dataset. With our proposed model, we\nobtained a high recognition rate of 92.5% using XGBoost and LightGBM for 9\ndifferent emotion classes. By utilizing this, we aim to improvise and suggest\nmethods to aid emotion recognition for better mental health analysis and mood\nmonitoring.\n", "date": "2022-07-12", "categories": []}
{"id": "2206.13101", "title": "SpeechEQ: Speech Emotion Recognition based on Multi-scale Unified\n  Datasets and Multitask Learning", "abstract": "  Speech emotion recognition (SER) has many challenges, but one of the main\nchallenges is that each framework does not have a unified standard. In this\npaper, we propose SpeechEQ, a framework for unifying SER tasks based on a\nmulti-scale unified metric. This metric can be trained by Multitask Learning\n(MTL), which includes two emotion recognition tasks of Emotion States Category\n(EIS) and Emotion Intensity Scale (EIS), and two auxiliary tasks of phoneme\nrecognition and gender recognition. For this framework, we build a Mandarin SER\ndataset - SpeechEQ Dataset (SEQD). We conducted experiments on the public CASIA\nand ESD datasets in Mandarin, which exhibit that our method outperforms\nbaseline methods by a relatively large margin, yielding 8.0% and 6.5%\nimprovement in accuracy respectively. Additional experiments on IEMOCAP with\nfour emotion categories (i.e., angry, happy, sad, and neutral) also show the\nproposed method achieves a state-of-the-art of both weighted accuracy (WA) of\n78.16% and unweighted accuracy (UA) of 77.47%.\n", "date": "2022-06-27", "categories": []}
{"id": "2207.13254", "title": "Contextual Information and Commonsense Based Prompt for Emotion\n  Recognition in Conversation", "abstract": "  Emotion recognition in conversation (ERC) aims to detect the emotion for each\nutterance in a given conversation. The newly proposed ERC models have leveraged\npre-trained language models (PLMs) with the paradigm of pre-training and\nfine-tuning to obtain good performance. However, these models seldom exploit\nPLMs' advantages thoroughly, and perform poorly for the conversations lacking\nexplicit emotional expressions. In order to fully leverage the latent knowledge\nrelated to the emotional expressions in utterances, we propose a novel ERC\nmodel CISPER with the new paradigm of prompt and language model (LM) tuning.\nSpecifically, CISPER is equipped with the prompt blending the contextual\ninformation and commonsense related to the interlocutor's utterances, to\nachieve ERC more effectively. Our extensive experiments demonstrate CISPER's\nsuperior performance over the state-of-the-art ERC models, and the\neffectiveness of leveraging these two kinds of significant prompt information\nfor performance gains. To reproduce our experimental results conveniently,\nCISPER's sourcecode and the datasets have been shared at\nhttps://github.com/DeqingYang/CISPER.\n", "date": "2022-07-27", "categories": []}
{"id": "2207.11389", "title": "Two-Aspect Information Fusion Model For ABAW4 Multi-task Challenge", "abstract": "  In this paper, we propose the solution to the Multi-Task Learning (MTL)\nChallenge of the 4th Affective Behavior Analysis in-the-wild (ABAW)\ncompetition. The task of ABAW is to predict frame-level emotion descriptors\nfrom videos: discrete emotional state; valence and arousal; and action units.\nAlthough researchers have proposed several approaches and achieved promising\nresults in ABAW, current works in this task rarely consider interactions\nbetween different emotion descriptors. To this end, we propose a novel end to\nend architecture to achieve full integration of different types of information.\nExperimental results demonstrate the effectiveness of our proposed solution.\n", "date": "2022-07-23", "categories": []}
{"id": "2207.11482", "title": "Multimodal Emotion Recognition with Modality-Pairwise Unsupervised\n  Contrastive Loss", "abstract": "  Emotion recognition is involved in several real-world applications. With an\nincrease in available modalities, automatic understanding of emotions is being\nperformed more accurately. The success in Multimodal Emotion Recognition (MER),\nprimarily relies on the supervised learning paradigm. However, data annotation\nis expensive, time-consuming, and as emotion expression and perception depends\non several factors (e.g., age, gender, culture) obtaining labels with a high\nreliability is hard. Motivated by these, we focus on unsupervised feature\nlearning for MER. We consider discrete emotions, and as modalities text, audio\nand vision are used. Our method, as being based on contrastive loss between\npairwise modalities, is the first attempt in MER literature. Our end-to-end\nfeature learning approach has several differences (and advantages) compared to\nexisting MER methods: i) it is unsupervised, so the learning is lack of data\nlabelling cost; ii) it does not require data spatial augmentation, modality\nalignment, large number of batch size or epochs; iii) it applies data fusion\nonly at inference; and iv) it does not require backbones pre-trained on emotion\nrecognition task. The experiments on benchmark datasets show that our method\noutperforms several baseline approaches and unsupervised learning methods\napplied in MER. Particularly, it even surpasses a few supervised MER\nstate-of-the-art.\n", "date": "2022-07-23", "categories": []}
{"id": "2207.11808", "title": "ArmanEmo: A Persian Dataset for Text-based Emotion Detection", "abstract": "  With the recent proliferation of open textual data on social media platforms,\nEmotion Detection (ED) from Text has received more attention over the past\nyears. It has many applications, especially for businesses and online service\nproviders, where emotion detection techniques can help them make informed\ncommercial decisions by analyzing customers/users' feelings towards their\nproducts and services. In this study, we introduce ArmanEmo, a human-labeled\nemotion dataset of more than 7000 Persian sentences labeled for seven\ncategories. The dataset has been collected from different resources, including\nTwitter, Instagram, and Digikala (an Iranian e-commerce company) comments.\nLabels are based on Ekman's six basic emotions (Anger, Fear, Happiness, Hatred,\nSadness, Wonder) and another category (Other) to consider any other emotion not\nincluded in Ekman's model. Along with the dataset, we have provided several\nbaseline models for emotion classification focusing on the state-of-the-art\ntransformer-based language models. Our best model achieves a macro-averaged F1\nscore of 75.39 percent across our test dataset. Moreover, we also conduct\ntransfer learning experiments to compare our proposed dataset's generalization\nagainst other Persian emotion datasets. Results of these experiments suggest\nthat our dataset has superior generalizability among the existing Persian\nemotion datasets. ArmanEmo is publicly available for non-commercial use at\nhttps://github.com/Arman-Rayan-Sharif/arman-text-emotion.\n", "date": "2022-07-24", "categories": []}
{"id": "2207.12135", "title": "Label Uncertainty Modeling and Prediction for Speech Emotion Recognition\n  using t-Distributions", "abstract": "  As different people perceive others' emotional expressions differently, their\nannotation in terms of arousal and valence are per se subjective. To address\nthis, these emotion annotations are typically collected by multiple annotators\nand averaged across annotators in order to obtain labels for arousal and\nvalence. However, besides the average, also the uncertainty of a label is of\ninterest, and should also be modeled and predicted for automatic emotion\nrecognition. In the literature, for simplicity, label uncertainty modeling is\ncommonly approached with a Gaussian assumption on the collected annotations.\nHowever, as the number of annotators is typically rather small due to resource\nconstraints, we argue that the Gaussian approach is a rather crude assumption.\nIn contrast, in this work we propose to model the label distribution using a\nStudent's t-distribution which allows us to account for the number of\nannotations available. With this model, we derive the corresponding\nKullback-Leibler divergence based loss function and use it to train an\nestimator for the distribution of emotion labels, from which the mean and\nuncertainty can be inferred. Through qualitative and quantitative analysis, we\nshow the benefits of the t-distribution over a Gaussian distribution. We\nvalidate our proposed method on the AVEC'16 dataset. Results reveal that our\nt-distribution based approach improves over the Gaussian approach with\nstate-of-the-art uncertainty modeling results in speech-based emotion\nrecognition, along with an optimal and even faster convergence.\n", "date": "2022-07-25", "categories": []}
{"id": "2207.10878", "title": "An Ensemble Approach for Multiple Emotion Descriptors Estimation Using\n  Multi-task Learning", "abstract": "  This paper illustrates our submission method to the fourth Affective Behavior\nAnalysis in-the-Wild (ABAW) Competition. The method is used for the Multi-Task\nLearning Challenge. Instead of using only face information, we employ full\ninformation from a provided dataset containing face and the context around the\nface. We utilized the InceptionNet V3 model to extract deep features then we\napplied the attention mechanism to refine the features. After that, we put\nthose features into the transformer block and multi-layer perceptron networks\nto get the final multiple kinds of emotion. Our model predicts arousal and\nvalence, classifies the emotional expression and estimates the action units\nsimultaneously. The proposed system achieves the performance of 0.917 on the\nMTL Challenge validation dataset.\n", "date": "2022-07-22", "categories": []}
{"id": "2207.11123", "title": "Adaptive Graph-Based Feature Normalization for Facial Expression\n  Recognition", "abstract": "  Facial Expression Recognition (FER) suffers from data uncertainties caused by\nambiguous facial images and annotators' subjectiveness, resulting in excursive\nsemantic and feature covariate shifting problem. Existing works usually correct\nmislabeled data by estimating noise distribution, or guide network training\nwith knowledge learned from clean data, neglecting the associative relations of\nexpressions. In this work, we propose an Adaptive Graph-based Feature\nNormalization (AGFN) method to protect FER models from data uncertainties by\nnormalizing feature distributions with the association of expressions.\nSpecifically, we propose a Poisson graph generator to adaptively construct\ntopological graphs for samples in each mini-batches via a sampling process, and\ncorrespondingly design a coordinate descent strategy to optimize proposed\nnetwork. Our method outperforms state-of-the-art works with accuracies of\n91.84% and 91.11% on the benchmark datasets FERPlus and RAF-DB, respectively,\nand when the percentage of mislabeled data increases (e.g., to 20%), our\nnetwork surpasses existing works significantly by 3.38% and 4.52%.\n", "date": "2022-07-22", "categories": []}
{"id": "2207.10201", "title": "Hybrid CNN-Transformer Model For Facial Affect Recognition In the ABAW4\n  Challenge", "abstract": "  This paper describes our submission to the fourth Affective Behavior Analysis\n(ABAW) competition. We proposed a hybrid CNN-Transformer model for the\nMulti-Task-Learning (MTL) and Learning from Synthetic Data (LSD) task.\nExperimental results on validation dataset shows that our method achieves\nbetter performance than baseline model, which verifies that the effectiveness\nof proposed network.\n", "date": "2022-07-20", "categories": []}
{"id": "2207.10644", "title": "CTL-MTNet: A Novel CapsNet and Transfer Learning-Based Mixed Task Net\n  for the Single-Corpus and Cross-Corpus Speech Emotion Recognition", "abstract": "  Speech Emotion Recognition (SER) has become a growing focus of research in\nhuman-computer interaction. An essential challenge in SER is to extract common\nattributes from different speakers or languages, especially when a specific\nsource corpus has to be trained to recognize the unknown data coming from\nanother speech corpus. To address this challenge, a Capsule Network (CapsNet)\nand Transfer Learning based Mixed Task Net (CTLMTNet) are proposed to deal with\nboth the singlecorpus and cross-corpus SER tasks simultaneously in this paper.\nFor the single-corpus task, the combination of Convolution-Pooling and\nAttention CapsNet module CPAC) is designed by embedding the self-attention\nmechanism to the CapsNet, guiding the module to focus on the important features\nthat can be fed into different capsules. The extracted high-level features by\nCPAC provide sufficient discriminative ability. Furthermore, to handle the\ncross-corpus task, CTL-MTNet employs a Corpus Adaptation Adversarial Module\n(CAAM) by combining CPAC with Margin Disparity Discrepancy (MDD), which can\nlearn the domain-invariant emotion representations through extracting the\nstrong emotion commonness. Experiments including ablation studies and\nvisualizations on both singleand cross-corpus tasks using four well-known SER\ndatasets in different languages are conducted for performance evaluation and\ncomparison. The results indicate that in both tasks the CTL-MTNet showed better\nperformance in all cases compared to a number of state-of-the-art methods. The\nsource code and the supplementary materials are available at:\nhttps://github.com/MLDMXM2017/CTLMTNet\n", "date": "2022-07-18", "categories": []}
{"id": "2207.10654", "title": "Emotion detection of social data: APIs comparative study", "abstract": "  The development of emotion detection technology has emerged as a highly\nvaluable possibility in the corporate sector due to the nearly limitless uses\nof this new discipline, particularly with the unceasing propagation of social\ndata. In recent years, the electronic marketplace has witnessed the\nestablishment of a large number of start-up businesses with an almost sole\nfocus on building new commercial and open-source tools and APIs for emotion\ndetection and recognition. Yet, these tools and APIs must be continuously\nreviewed and evaluated, and their performances should be reported and\ndiscussed. There is a lack of research to empirically compare current emotion\ndetection technologies in terms of the results obtained from each model using\nthe same textual dataset. Also, there is a lack of comparative studies that\napply benchmark comparison to social data. This study compares eight\ntechnologies; IBM Watson NLU, ParallelDots, Symanto-Ekman, Crystalfeel, Text to\nEmotion, Senpy, Textprobe, and NLP Cloud. The comparison was undertaken using\ntwo different datasets. The emotions from the chosen datasets were then derived\nusing the incorporated APIs. The performance of these APIs was assessed using\nthe aggregated scores that they delivered as well as the theoretically proven\nevaluation metrics such as the micro-average of accuracy, classification error,\nprecision, recall, and f1-score. Lastly, the assessment of these APIs\nincorporating the evaluation measures is reported and discussed.\n", "date": "2022-07-08", "categories": []}
{"id": "2207.09716", "title": "Multi-Task Learning for Emotion Descriptors Estimation at the fourth\n  ABAW Challenge", "abstract": "  Facial valence/arousal, expression and action unit are related tasks in\nfacial affective analysis. However, the tasks only have limited performance in\nthe wild due to the various collected conditions. The 4th competition on\naffective behavior analysis in the wild (ABAW) provided images with\nvalence/arousal, expression and action unit labels. In this paper, we introduce\nmulti-task learning framework to enhance the performance of three related tasks\nin the wild. Feature sharing and label fusion are used to utilize their\nrelations. We conduct experiments on the provided training and validating data.\n", "date": "2022-07-20", "categories": []}
{"id": "2207.08811", "title": "Fusion of Physiological and Behavioural Signals on SPD Manifolds with\n  Application to Stress and Pain Detection", "abstract": "  Existing multimodal stress/pain recognition approaches generally extract\nfeatures from different modalities independently and thus ignore cross-modality\ncorrelations. This paper proposes a novel geometric framework for multimodal\nstress/pain detection utilizing Symmetric Positive Definite (SPD) matrices as a\nrepresentation that incorporates the correlation relationship of physiological\nand behavioural signals from covariance and cross-covariance. Considering the\nnon-linearity of the Riemannian manifold of SPD matrices, well-known machine\nlearning techniques are not suited to classify these matrices. Therefore, a\ntangent space mapping method is adopted to map the derived SPD matrix sequences\nto the vector sequences in the tangent space where the LSTM-based network can\nbe applied for classification. The proposed framework has been evaluated on two\npublic multimodal datasets, achieving both the state-of-the-art results for\nstress and pain detection tasks.\n", "date": "2022-07-17", "categories": []}
{"id": "2207.07693", "title": "Towards Understanding Confusion and Affective States Under Communication\n  Failures in Voice-Based Human-Machine Interaction", "abstract": "  We present a series of two studies conducted to understand user's affective\nstates during voice-based human-machine interactions. Emphasis is placed on the\ncases of communication errors or failures. In particular, we are interested in\nunderstanding \"confusion\" in relation with other affective states. The studies\nconsist of two types of tasks: (1) related to communication with a voice-based\nvirtual agent: speaking to the machine and understanding what the machine says,\n(2) non-communication related, problem-solving tasks where the participants\nsolve puzzles and riddles but are asked to verbally explain the answers to the\nmachine. We collected audio-visual data and self-reports of affective states of\nthe participants. We report results of two studies and analysis of the\ncollected data. The first study was analyzed based on the annotator's\nobservation, and the second study was analyzed based on the self-report.\n", "date": "2022-07-15", "categories": []}
{"id": "2207.07973", "title": "Learn-to-Decompose: Cascaded Decomposition Network for Cross-Domain\n  Few-Shot Facial Expression Recognition", "abstract": "  Most existing compound facial expression recognition (FER) methods rely on\nlarge-scale labeled compound expression data for training. However, collecting\nsuch data is labor-intensive and time-consuming. In this paper, we address the\ncompound FER task in the cross-domain few-shot learning (FSL) setting, which\nrequires only a few samples of compound expressions in the target domain.\nSpecifically, we propose a novel cascaded decomposition network (CDNet), which\ncascades several learn-to-decompose modules with shared parameters based on a\nsequential decomposition mechanism, to obtain a transferable feature space. To\nalleviate the overfitting problem caused by limited base classes in our task, a\npartial regularization strategy is designed to effectively exploit the best of\nboth episodic training and batch training. By training across similar tasks on\nmultiple basic expression datasets, CDNet learns the ability of\nlearn-to-decompose that can be easily adapted to identify unseen compound\nexpressions. Extensive experiments on both in-the-lab and in-the-wild compound\nexpression datasets demonstrate the superiority of our proposed CDNet against\nseveral state-of-the-art FSL methods. Code is available at:\nhttps://github.com/zouxinyi0625/CDNet.\n", "date": "2022-07-16", "categories": []}
{"id": "2207.08104", "title": "A Multibias-mitigated and Sentiment Knowledge Enriched Transformer for\n  Debiasing in Multimodal Conversational Emotion Recognition", "abstract": "  Multimodal emotion recognition in conversations (mERC) is an active research\ntopic in natural language processing (NLP), which aims to predict human's\nemotional states in communications of multiple modalities, e,g., natural\nlanguage and facial gestures. Innumerable implicit prejudices and\npreconceptions fill human language and conversations, leading to the question\nof whether the current data-driven mERC approaches produce a biased error. For\nexample, such approaches may offer higher emotional scores on the utterances by\nfemales than males. In addition, the existing debias models mainly focus on\ngender or race, where multibias mitigation is still an unexplored task in mERC.\nIn this work, we take the first step to solve these issues by proposing a\nseries of approaches to mitigate five typical kinds of bias in textual\nutterances (i.e., gender, age, race, religion and LGBTQ+) and visual\nrepresentations (i.e, gender and age), followed by a Multibias-Mitigated and\nsentiment Knowledge Enriched bi-modal Transformer (MMKET). Comprehensive\nexperimental results show the effectiveness of the proposed model and prove\nthat the debias operation has a great impact on the classification performance\nfor mERC. We hope our study will benefit the development of bias mitigation in\nmERC and related emotion studies.\n", "date": "2022-07-17", "categories": []}
{"id": "2207.07238", "title": "Emotion Recognition in Conversation using Probabilistic Soft Logic", "abstract": "  Creating agents that can both appropriately respond to conversations and\nunderstand complex human linguistic tendencies and social cues has been a long\nstanding challenge in the NLP community. A recent pillar of research revolves\naround emotion recognition in conversation (ERC); a sub-field of emotion\nrecognition that focuses on conversations or dialogues that contain two or more\nutterances. In this work, we explore an approach to ERC that exploits the use\nof neural embeddings along with complex structures in dialogues. We implement\nour approach in a framework called Probabilistic Soft Logic (PSL), a\ndeclarative templating language that uses first-order like logical rules, that\nwhen combined with data, define a particular class of graphical model.\nAdditionally, PSL provides functionality for the incorporation of results from\nneural models into PSL models. This allows our model to take advantage of\nadvanced neural methods, such as sentence embeddings, and logical reasoning\nover the structure of a dialogue. We compare our method with state-of-the-art\npurely neural ERC systems, and see almost a 20% improvement. With these\nresults, we provide an extensive qualitative and quantitative analysis over the\nDailyDialog conversation dataset.\n", "date": "2022-07-14", "categories": []}
{"id": "2207.06410", "title": "MDEAW: A Multimodal Dataset for Emotion Analysis through EDA and PPG\n  signals from wireless wearable low-cost off-the-shelf Devices", "abstract": "  We present MDEAW, a multimodal database consisting of Electrodermal Activity\n(EDA) and Photoplethysmography (PPG) signals recorded during the exams for the\ncourse taught by the teacher at Eurecat Academy, Sabadell, Barcelona in order\nto elicit the emotional reactions to the students in a classroom scenario.\nSignals from 10 students were recorded along with the students' self-assessment\nof their affective state after each stimulus, in terms of 6 basic emotion\nstates. All the signals were captured using portable, wearable, wireless,\nlow-cost, and off-the-shelf equipment that has the potential to allow the use\nof affective computing methods in everyday applications. A baseline for\nstudent-wise affect recognition using EDA and PPG-based features, as well as\ntheir fusion, was established through ReMECS, Fed-ReMECS, and Fed-ReMECS-U.\nThese results indicate the prospects of using low-cost devices for affective\nstate recognition applications. The proposed database will be made publicly\navailable in order to allow researchers to achieve a more thorough evaluation\nof the suitability of these capturing devices for emotion state recognition\napplications.\n", "date": "2022-07-14", "categories": []}
{"id": "2207.04697", "title": "Multi-level Fusion of Wav2vec 2.0 and BERT for Multimodal Emotion\n  Recognition", "abstract": "  The research and applications of multimodal emotion recognition have become\nincreasingly popular recently. However, multimodal emotion recognition faces\nthe challenge of lack of data. To solve this problem, we propose to use\ntransfer learning which leverages state-of-the-art pre-trained models including\nwav2vec 2.0 and BERT for this task. Multi-level fusion approaches including\ncoattention-based early fusion and late fusion with the models trained on both\nembeddings are explored. Also, a multi-granularity framework which extracts not\nonly frame-level speech embeddings but also segment-level embeddings including\nphone, syllable and word-level speech embeddings is proposed to further boost\nthe performance. By combining our coattention-based early fusion model and late\nfusion model with the multi-granularity feature extraction framework, we obtain\nresult that outperforms best baseline approaches by 1.3% unweighted accuracy\n(UA) on the IEMOCAP dataset.\n", "date": "2022-07-11", "categories": []}
{"id": "2207.03334", "title": "Speech Emotion: Investigating Model Representations, Multi-Task Learning\n  and Knowledge Distillation", "abstract": "  Estimating dimensional emotions, such as activation, valence and dominance,\nfrom acoustic speech signals has been widely explored over the past few years.\nWhile accurate estimation of activation and dominance from speech seem to be\npossible, the same for valence remains challenging. Previous research has shown\nthat the use of lexical information can improve valence estimation performance.\nLexical information can be obtained from pre-trained acoustic models, where the\nlearned representations can improve valence estimation from speech. We\ninvestigate the use of pre-trained model representations to improve valence\nestimation from acoustic speech signal. We also explore fusion of\nrepresentations to improve emotion estimation across all three emotion\ndimensions: activation, valence and dominance. Additionally, we investigate if\nrepresentations from pre-trained models can be distilled into models trained\nwith low-level features, resulting in models with a less number of parameters.\nWe show that fusion of pre-trained model embeddings result in a 79% relative\nimprovement in concordance correlation coefficient CCC on valence estimation\ncompared to standard acoustic feature baseline (mel-filterbank energies), while\ndistillation from pre-trained model embeddings to lower-dimensional\nrepresentations yielded a relative 12% improvement. Such performance gains were\nobserved over two evaluation sets, indicating that our proposed architecture\ngeneralizes across those evaluation sets. We report new state-of-the-art\n\"text-free\" acoustic-only dimensional emotion estimation $CCC$ values on two\nMSP-Podcast evaluation sets.\n", "date": "2022-07-02", "categories": []}
{"id": "2207.01138", "title": "ABAW: Learning from Synthetic Data & Multi-Task Learning Challenges", "abstract": "  This paper describes the fourth Affective Behavior Analysis in-the-wild\n(ABAW) Competition, held in conjunction with European Conference on Computer\nVision (ECCV), 2022. The 4th ABAW Competition is a continuation of the\nCompetitions held at IEEE CVPR 2022, ICCV 2021, IEEE FG 2020 and IEEE CVPR 2017\nConferences, and aims at automatically analyzing affect. In the previous runs\nof this Competition, the Challenges targeted Valence-Arousal Estimation,\nExpression Classification and Action Unit Detection. This year the Competition\nencompasses two different Challenges: i) a Multi-Task-Learning one in which the\ngoal is to learn at the same time (i.e., in a multi-task learning setting) all\nthe three above mentioned tasks; and ii) a Learning from Synthetic Data one in\nwhich the goal is to learn to recognise the basic expressions from artificially\ngenerated data and generalise to real data. The Aff-Wild2 database is a large\nscale in-the-wild database and the first one that contains annotations for\nvalence and arousal, expressions and action units. This database is the basis\nfor the above Challenges. In more detail: i) s-Aff-Wild2 -- a static version of\nAff-Wild2 database -- has been constructed and utilized for the purposes of the\nMulti-Task-Learning Challenge; and ii) some specific frames-images from the\nAff-Wild2 database have been used in an expression manipulation manner for\ncreating the synthetic dataset, which is the basis for the Learning from\nSynthetic Data Challenge. In this paper, at first we present the two\nChallenges, along with the utilized corpora, then we outline the evaluation\nmetrics and finally present the baseline systems per Challenge, as well as\ntheir derived results. More information regarding the Competition can be found\nin the competition's website:\nhttps://ibug.doc.ic.ac.uk/resources/eccv-2023-4th-abaw/.\n", "date": "2022-07-03", "categories": []}
{"id": "2207.02104", "title": "A cross-corpus study on speech emotion recognition", "abstract": "  For speech emotion datasets, it has been difficult to acquire large\nquantities of reliable data and acted emotions may be over the top compared to\nless expressive emotions displayed in everyday life. Lately, larger datasets\nwith natural emotions have been created. Instead of ignoring smaller, acted\ndatasets, this study investigates whether information learnt from acted\nemotions is useful for detecting natural emotions. Cross-corpus research has\nmostly considered cross-lingual and even cross-age datasets, and difficulties\narise from different methods of annotating emotions causing a drop in\nperformance. To be consistent, four adult English datasets covering acted,\nelicited and natural emotions are considered. A state-of-the-art model is\nproposed to accurately investigate the degradation of performance. The system\ninvolves a bi-directional LSTM with an attention mechanism to classify emotions\nacross datasets. Experiments study the effects of training models in a\ncross-corpus and multi-domain fashion and results show the transfer of\ninformation is not successful. Out-of-domain models, followed by adapting to\nthe missing dataset, and domain adversarial training (DAT) are shown to be more\nsuitable to generalising to emotions across datasets. This shows positive\ninformation transfer from acted datasets to those with more natural emotions\nand the benefits from training on different corpora.\n", "date": "2022-07-05", "categories": []}
{"id": "2206.11049", "title": "Dynamic Restrained Uncertainty Weighting Loss for Multitask Learning of\n  Vocal Expression", "abstract": "  We propose a novel Dynamic Restrained Uncertainty Weighting Loss to\nexperimentally handle the problem of balancing the contributions of multiple\ntasks on the ICML ExVo 2022 Challenge. The multitask aims to recognize\nexpressed emotions and demographic traits from vocal bursts jointly. Our\nstrategy combines the advantages of Uncertainty Weight and Dynamic Weight\nAverage, by extending weights with a restraint term to make the learning\nprocess more explainable. We use a lightweight multi-exit CNN architecture to\nimplement our proposed loss approach. The experimental H-Mean score (0.394)\nshows a substantial improvement over the baseline H-Mean score (0.335).\n", "date": "2022-06-22", "categories": []}
{"id": "2206.05286", "title": "AHD ConvNet for Speech Emotion Classification", "abstract": "  Accomplishments in the field of artificial intelligence are utilized in the\nadvancement of computing and making of intelligent machines for facilitating\nmankind and improving user experience. Emotions are rudimentary for people,\naffecting thinking and ordinary exercises like correspondence, learning and\ndirection. Speech emotion recognition is domain of interest in this regard and\nin this work, we propose a novel mel spectrogram learning approach in which our\nmodel uses the datapoints to learn emotions from the given wav form voice notes\nin the popular CREMA-D dataset. Our model uses log mel-spectrogram as feature\nwith number of mels = 64. It took less training time compared to other\napproaches used to address the problem of emotion speech recognition.\n", "date": "2022-06-10", "categories": []}
{"id": "2206.07981", "title": "Multi-scale Cooperative Multimodal Transformers for Multimodal Sentiment\n  Analysis in Videos", "abstract": "  Multimodal sentiment analysis in videos is a key task in many real-world\napplications, which usually requires integrating multimodal streams including\nvisual, verbal and acoustic behaviors. To improve the robustness of multimodal\nfusion, some of the existing methods let different modalities communicate with\neach other and modal the crossmodal interaction via transformers. However,\nthese methods only use the single-scale representations during the interaction\nbut forget to exploit multi-scale representations that contain different levels\nof semantic information. As a result, the representations learned by\ntransformers could be biased especially for unaligned multimodal data. In this\npaper, we propose a multi-scale cooperative multimodal transformer (MCMulT)\narchitecture for multimodal sentiment analysis. On the whole, the \"multi-scale\"\nmechanism is capable of exploiting the different levels of semantic information\nof each modality which are used for fine-grained crossmodal interactions.\nMeanwhile, each modality learns its feature hierarchies via integrating the\ncrossmodal interactions from multiple level features of its source modality. In\nthis way, each pair of modalities progressively builds feature hierarchies\nrespectively in a cooperative manner. The empirical results illustrate that our\nMCMulT model not only outperforms existing approaches on unaligned multimodal\nsequences but also has strong performance on aligned multimodal sequences.\n", "date": "2022-06-16", "categories": []}
{"id": "2206.07359", "title": "The Emotion is Not One-hot Encoding: Learning with Grayscale Label for\n  Emotion Recognition in Conversation", "abstract": "  In emotion recognition in conversation (ERC), the emotion of the current\nutterance is predicted by considering the previous context, which can be\nutilized in many natural language processing tasks. Although multiple emotions\ncan coexist in a given sentence, most previous approaches take the perspective\nof a classification task to predict only a given label. However, it is\nexpensive and difficult to label the emotion of a sentence with confidence or\nmulti-label. In this paper, we automatically construct a grayscale label\nconsidering the correlation between emotions and use it for learning. That is,\ninstead of using a given label as a one-hot encoding, we construct a grayscale\nlabel by measuring scores for different emotions. We introduce several methods\nfor constructing grayscale labels and confirm that each method improves the\nemotion recognition performance. Our method is simple, effective, and\nuniversally applicable to previous systems. The experiments show a significant\nimprovement in the performance of baselines.\n", "date": "2022-06-15", "categories": []}
{"id": "2206.07229", "title": "Accurate Emotion Strength Assessment for Seen and Unseen Speech Based on\n  Data-Driven Deep Learning", "abstract": "  Emotion classification of speech and assessment of the emotion strength are\nrequired in applications such as emotional text-to-speech and voice conversion.\nThe emotion attribute ranking function based on Support Vector Machine (SVM)\nwas proposed to predict emotion strength for emotional speech corpus. However,\nthe trained ranking function doesn't generalize to new domains, which limits\nthe scope of applications, especially for out-of-domain or unseen speech. In\nthis paper, we propose a data-driven deep learning model, i.e. StrengthNet, to\nimprove the generalization of emotion strength assessment for seen and unseen\nspeech. This is achieved by the fusion of emotional data from various domains.\nWe follow a multi-task learning network architecture that includes an acoustic\nencoder, a strength predictor, and an auxiliary emotion predictor. Experiments\nshow that the predicted emotion strength of the proposed StrengthNet is highly\ncorrelated with ground truth scores for both seen and unseen speech. We release\nthe source codes at: https://github.com/ttslr/StrengthNet.\n", "date": "2022-06-15", "categories": []}
{"id": "2206.07593", "title": "HICEM: A High-Coverage Emotion Model for Artificial Emotional\n  Intelligence", "abstract": "  As social robots and other intelligent machines enter the home, artificial\nemotional intelligence (AEI) is taking center stage to address users' desire\nfor deeper, more meaningful human-machine interaction. To accomplish such\nefficacious interaction, the next-generation AEI need comprehensive human\nemotion models for training. Unlike theory of emotion, which has been the\nhistorical focus in psychology, emotion models are a descriptive tools. In\npractice, the strongest models need robust coverage, which means defining the\nsmallest core set of emotions from which all others can be derived. To achieve\nthe desired coverage, we turn to word embeddings from natural language\nprocessing. Using unsupervised clustering techniques, our experiments show that\nwith as few as 15 discrete emotion categories, we can provide maximum coverage\nacross six major languages--Arabic, Chinese, English, French, Spanish, and\nRussian. In support of our findings, we also examine annotations from two\nlarge-scale emotion recognition datasets to assess the validity of existing\nemotion models compared to human perception at scale. Because robust,\ncomprehensive emotion models are foundational for developing real-world\naffective computing applications, this work has broad implications in social\nrobotics, human-machine interaction, mental healthcare, and computational\npsychology.\n", "date": "2022-06-15", "categories": []}
{"id": "2206.06775", "title": "DeepEmotex: Classifying Emotion in Text Messages using Deep Transfer\n  Learning", "abstract": "  Transfer learning has been widely used in natural language processing through\ndeep pretrained language models, such as Bidirectional Encoder Representations\nfrom Transformers and Universal Sentence Encoder. Despite the great success,\nlanguage models get overfitted when applied to small datasets and are prone to\nforgetting when fine-tuned with a classifier. To remedy this problem of\nforgetting in transferring deep pretrained language models from one domain to\nanother domain, existing efforts explore fine-tuning methods to forget less. We\npropose DeepEmotex an effective sequential transfer learning method to detect\nemotion in text. To avoid forgetting problem, the fine-tuning step is\ninstrumented by a large amount of emotion-labeled data collected from Twitter.\nWe conduct an experimental study using both curated Twitter data sets and\nbenchmark data sets. DeepEmotex models achieve over 91% accuracy for\nmulti-class emotion classification on test dataset. We evaluate the performance\nof the fine-tuned DeepEmotex models in classifying emotion in EmoInt and\nStimulus benchmark datasets. The models correctly classify emotion in 73% of\nthe instances in the benchmark datasets. The proposed DeepEmotex-BERT model\noutperforms Bi-LSTM result on the benchmark datasets by 23%. We also study the\neffect of the size of the fine-tuning dataset on the accuracy of our models.\nOur evaluation results show that fine-tuning with a large set of\nemotion-labeled data improves both the robustness and effectiveness of the\nresulting target task model.\n", "date": "2022-06-12", "categories": []}
{"id": "2206.04975", "title": "NR-DFERNet: Noise-Robust Network for Dynamic Facial Expression\n  Recognition", "abstract": "  Dynamic facial expression recognition (DFER) in the wild is an extremely\nchallenging task, due to a large number of noisy frames in the video sequences.\nPrevious works focus on extracting more discriminative features, but ignore\ndistinguishing the key frames from the noisy frames. To tackle this problem, we\npropose a noise-robust dynamic facial expression recognition network\n(NR-DFERNet), which can effectively reduce the interference of noisy frames on\nthe DFER task. Specifically, at the spatial stage, we devise a dynamic-static\nfusion module (DSF) that introduces dynamic features to static features for\nlearning more discriminative spatial features. To suppress the impact of target\nirrelevant frames, we introduce a novel dynamic class token (DCT) for the\ntransformer at the temporal stage. Moreover, we design a snippet-based filter\n(SF) at the decision stage to reduce the effect of too many neutral frames on\nnon-neutral sequence classification. Extensive experimental results demonstrate\nthat our NR-DFERNet outperforms the state-of-the-art methods on both the DFEW\nand AFEW benchmarks.\n", "date": "2022-06-10", "categories": []}
{"id": "2206.04399", "title": "Depression Recognition using Remote Photoplethysmography from Facial\n  Videos", "abstract": "  Depression is a mental illness that may be harmful to an individual's health.\nThe detection of mental health disorders in the early stages and a precise\ndiagnosis are critical to avoid social, physiological, or psychological side\neffects. This work analyzes physiological signals to observe if different\ndepressive states have a noticeable impact on the blood volume pulse (BVP) and\nthe heart rate variability (HRV) response. Although typically, HRV features are\ncalculated from biosignals obtained with contact-based sensors such as\nwearables, we propose instead a novel scheme that directly extracts them from\nfacial videos, just based on visual information, removing the need for any\ncontact-based device. Our solution is based on a pipeline that is able to\nextract complete remote photoplethysmography signals (rPPG) in a fully\nunsupervised manner. We use these rPPG signals to calculate over 60\nstatistical, geometrical, and physiological features that are further used to\ntrain several machine learning regressors to recognize different levels of\ndepression. Experiments on two benchmark datasets indicate that this approach\noffers comparable results to other audiovisual modalities based on voice or\nfacial expression, potentially complementing them. In addition, the results\nachieved for the proposed method show promising and solid performance that\noutperforms hand-engineered methods and is comparable to deep learning-based\napproaches.\n", "date": "2022-06-09", "categories": []}
{"id": "2206.04625", "title": "AttX: Attentive Cross-Connections for Fusion of Wearable Signals in\n  Emotion Recognition", "abstract": "  We propose cross-modal attentive connections, a new dynamic and effective\ntechnique for multimodal representation learning from wearable data. Our\nsolution can be integrated into any stage of the pipeline, i.e., after any\nconvolutional layer or block, to create intermediate connections between\nindividual streams responsible for processing each modality. Additionally, our\nmethod benefits from two properties. First, it can share information\nuni-directionally (from one modality to the other) or bi-directionally. Second,\nit can be integrated into multiple stages at the same time to further allow\nnetwork gradients to be exchanged in several touch-points. We perform extensive\nexperiments on three public multimodal wearable datasets, WESAD, SWELL-KW, and\nCASE, and demonstrate that our method can effectively regulate and share\ninformation between different modalities to learn better representations. Our\nexperiments further demonstrate that once integrated into simple CNN-based\nmultimodal solutions (2, 3, or 4 modalities), our method can result in superior\nor competitive performance to state-of-the-art and outperform a variety of\nbaseline uni-modal and classical multimodal methods.\n", "date": "2022-06-09", "categories": []}
{"id": "2206.03173", "title": "Speaker-Guided Encoder-Decoder Framework for Emotion Recognition in\n  Conversation", "abstract": "  The emotion recognition in conversation (ERC) task aims to predict the\nemotion label of an utterance in a conversation. Since the dependencies between\nspeakers are complex and dynamic, which consist of intra- and inter-speaker\ndependencies, the modeling of speaker-specific information is a vital role in\nERC. Although existing researchers have proposed various methods of speaker\ninteraction modeling, they cannot explore dynamic intra- and inter-speaker\ndependencies jointly, leading to the insufficient comprehension of context and\nfurther hindering emotion prediction. To this end, we design a novel speaker\nmodeling scheme that explores intra- and inter-speaker dependencies jointly in\na dynamic manner. Besides, we propose a Speaker-Guided Encoder-Decoder (SGED)\nframework for ERC, which fully exploits speaker information for the decoding of\nemotion. We use different existing methods as the conversational context\nencoder of our framework, showing the high scalability and flexibility of the\nproposed framework. Experimental results demonstrate the superiority and\neffectiveness of SGED.\n", "date": "2022-06-07", "categories": []}
{"id": "2206.02119", "title": "A Multimodal Corpus for Emotion Recognition in Sarcasm", "abstract": "  While sentiment and emotion analysis have been studied extensively, the\nrelationship between sarcasm and emotion has largely remained unexplored. A\nsarcastic expression may have a variety of underlying emotions. For example, \"I\nlove being ignored\" belies sadness, while \"my mobile is fabulous with a battery\nbackup of only 15 minutes!\" expresses frustration. Detecting the emotion behind\na sarcastic expression is non-trivial yet an important task. We undertake the\ntask of detecting the emotion in a sarcastic statement, which to the best of\nour knowledge, is hitherto unexplored. We start with the recently released\nmultimodal sarcasm detection dataset (MUStARD) pre-annotated with 9 emotions.\nWe identify and correct 343 incorrect emotion labels (out of 690). We double\nthe size of the dataset, label it with emotions along with valence and arousal\nwhich are important indicators of emotional intensity. Finally, we label each\nsarcastic utterance with one of the four sarcasm types-Propositional, Embedded,\nLikeprefixed and Illocutionary, with the goal of advancing sarcasm detection\nresearch. Exhaustive experimentation with multimodal (text, audio, and video)\nfusion models establishes a benchmark for exact emotion recognition in sarcasm\nand outperforms the state-of-art sarcasm detection. We release the dataset\nenriched with various annotations and the code for research purposes:\nhttps://github.com/apoorva-nunna/MUStARD_Plus_Plus\n", "date": "2022-06-05", "categories": []}
{"id": "2206.02187", "title": "M2FNet: Multi-modal Fusion Network for Emotion Recognition in\n  Conversation", "abstract": "  Emotion Recognition in Conversations (ERC) is crucial in developing\nsympathetic human-machine interaction. In conversational videos, emotion can be\npresent in multiple modalities, i.e., audio, video, and transcript. However,\ndue to the inherent characteristics of these modalities, multi-modal ERC has\nalways been considered a challenging undertaking. Existing ERC research focuses\nmainly on using text information in a discussion, ignoring the other two\nmodalities. We anticipate that emotion recognition accuracy can be improved by\nemploying a multi-modal approach. Thus, in this study, we propose a Multi-modal\nFusion Network (M2FNet) that extracts emotion-relevant features from visual,\naudio, and text modality. It employs a multi-head attention-based fusion\nmechanism to combine emotion-rich latent representations of the input data. We\nintroduce a new feature extractor to extract latent features from the audio and\nvisual modality. The proposed feature extractor is trained with a novel\nadaptive margin-based triplet loss function to learn emotion-relevant features\nfrom the audio and visual data. In the domain of ERC, the existing methods\nperform well on one benchmark dataset but not on others. Our results show that\nthe proposed M2FNet architecture outperforms all other methods in terms of\nweighted average F1 score on well-known MELD and IEMOCAP datasets and sets a\nnew state-of-the-art performance in ERC.\n", "date": "2022-06-05", "categories": []}
{"id": "2606.21069", "title": "Quality and Agreement in Multilabel Emotion Annotation: A Case Study and Evaluation Framework", "abstract": "Emotion annotation is inherently subjective, yet most NLP pipelines still assume \"gold\" labels, typically produced by majority voting, and treat annotator variation as noise. In this paper, we present a multilabel emotion annotation case study and use it to examine how annotator behavior and aggregation choices affect both agreement estimates and downstream emotion classifiers. Rather than collapsing disagreement into a single label, we represent targets as soft vote-share labels (including an intensity-weighted variant) and evaluate models using both thresholded metrics (macro-/micro-F1) and probabilistic alignment (Bernoulli cross-entropy SoftBCE), alongside data-derived disagreement diagnostics. Across annotation regimes, we show that disagreement is structured and leaves measurable traces in model behavior: hard labels may maximize F1 metrics, while soft supervision yields predictions that better reflect empirical annotator variance and uncertainty. Our results provide practical guidance for designing, aggregating, and evaluating multilabel emotion datasets when multiple interpretations are plausible.", "date": "2026-06-19", "categories": []}
{"id": "2606.21930", "title": "MindTailor: Personalized Emotional Support via Post History-Grounded Case Formulation and Collaborative Refinement", "abstract": "As mental health concerns continue to rise globally, social media has emerged as a vital space where individuals seek emotional support. While prior work on personalized emotional support has leveraged seekers' emotional states, personas, and situational context, these approaches primarily capture the seeker's current state, overlooking the formative experiences that shape present concerns. In this work, we propose MindTailor, a framework that generates personalized emotional support responses by constructing a case formulation from the seeker's post history and iteratively refining responses through collaborative critique among counselor agents grounded in distinct counseling strategies. To enable research on this history-aware task, we construct ReddiSupp, a dataset of 798 Reddit posts paired with seekers' prior post histories. Through LLM-as-a-Judge evaluation, expert human evaluation, and a user study with seekers, we demonstrate that MindTailor outperforms baselines across these evaluations, improving empathy, personalization, understanding, and achieving the highest overall preference.", "date": "2026-06-20", "categories": []}
{"id": "2606.22072", "title": "A Controlled Study of CLIP-Based Body-Scene Fusion for Emotion Recognition in Context", "abstract": "Apparent emotion in natural images is often not visible from the face alone. The face may be small, hidden, or neutral, while posture and scene context carry much of the evidence. This work studies context-aware emotion recognition on EMOTIC with an image-only two-stream model. A ResNet-18 body stream encodes the target-person crop, and a CLIP ViT-B/16 scene stream encodes the full image. The fused feature predicts 26 categorical emotion labels and the continuous valence, arousal, and dominance values. This study examines whether small context-debiasing or rare-class training changes still help after adding a CLIP scene encoder. The clean two-stream model is compared with simplified CCIM-style intervention, CLEF-lite context-bias subtraction, ASL tuning, and class-balanced sampling under the same implementation pipeline. No tested variant improves over the clean two-stream model, which achieves 34.52% mAP on the EMOTIC test split. CLIP gives the model broad scene semantics, but the simplified causal, counterfactual, and rare-class changes do not automatically improve performance. Most remaining errors are in rare and subtle emotion categories, so the next step should focus on label relationships and finer subject-context interaction.", "date": "2026-06-20", "categories": []}
{"id": "2606.22725", "title": "Interpretable Uncertainty Routing Separating Emotion Ambiguity from Distribution Shift in Facial Expression Recognition", "abstract": "Facial expression recognition (FER) is inherently ambiguous: human annotators frequently disagree, and models deployed in real environments face distribution shift. Crucially, these two conditions demand different downstream actions, as ambiguous in-distribution faces should be reported with their ambiguity whereas out-of-distribution inputs should be rejected. However, a single uncertainty score conflates the two. In this study, uncertainty decomposition into aleatoric and epistemic components for FER is investigated, and Uncertainty-Aware Routing (UAR), an inference-time routing mechanism that exploits the separation, is introduced. Specifically, aleatoric and epistemic uncertainties are obtained from a Deep Ensemble of fully fine-tuned DINOv2 models and are each validated against an independent external signal: aleatoric against human annotator disagreement, and epistemic against distribution shift induced by image corruptions. The proposed dual-validation protocol reveals that aleatoric recovers annotator disagreement with Spearman correlation 0.66 (95% CI: 0.64-0.68), and epistemic detects corruption-induced shifts, achieving average AUROC of 0.699 at the highest corruption severity. UAR retains approximately 1.8 times more ambiguous in-distribution faces than single-uncertainty routing at a matched out-of-distribution rejection rate. A strong label-distribution-learning baseline achieves comparable disagreement recovery but cannot separate ambiguity from shift and therefore cannot route, establishing that the value of decomposition lies in the separation enabling interpretable and differentiated action selection.", "date": "2026-06-21", "categories": []}
{"id": "2606.24941", "title": "EmotionAI: A Privacy-Preserving Computational Intelligence Pipeline for Speech-Emotion-Grounded Conversational Analysis", "abstract": "Reviewing recorded interviews for affective cues such as composure, hesitation and agitation is slow and subjective, and cloud services that could automate it require sensitive audio to leave the device. EmotionAI is a fully local Computational Intelligence (CI) pipeline that couples Speech Emotion Recognition (SER) with generative reasoning. Speaker diarisation, Whisper Automatic Speech Recognition (ASR) and a wav2vec2 emotion classifier produce per-segment affective evidence, which is then passed to an adversarial three-model local Large Language Model (LLM) panel for timestamp-grounded and citation-constrained question answering. Zero-shot evaluation on the RAVDESS four-class English subset (n = 672) exposes cross-corpus fragility rather than classifier superiority: the deployed classifier scores 48.8% accuracy, above random (24.9%) and majority (28.6%) baselines but below an in-domain MFCC + logistic-regression comparator (71.0%). The complete pipeline runs in a mean 157 s on CPU (real-time factor approximately 1.33) with zero external calls. The contribution is not state-of-the-art SER but an auditable, privacy-preserving integration of imperfect affective evidence into grounded conversational analysis, together with an honest empirical account of where cross-corpus transfer and human-centred validation still fall short.", "date": "2026-06-22", "categories": []}
{"id": "2606.25325", "title": "Omni-Perception Policy Optimization for Multimodal Emotion Reasoning", "abstract": "We find that current emotion-oriented Omni-MLLMs still lack reliable omni-modal perception: they (i) underutilize multimodal cues in their reasoning trajectories and (ii) exhibit unfaithful behavior, often hallucinating modality-specific statements from other modalities. Building on these insights, we propose OPPO (Omni-Perception Policy Optimization), a reinforcement learning framework that explicitly optimizes multimodal perception. First, an Omni-Perception Reward decomposes ground-truth reasoning into fine-grained visual, acoustic, and emotion cues and rewards trajectories that semantically recover these cues. Second, an Omni-Perception Loss compares the policy under full and unimodally masked inputs, applying a KL penalty only to modality-specific evidence tokens to suppress cross-modal hallucination. We further introduce MEP-Bench, a diagnostic benchmark that quantifies utilization and faithfulness. Experiments show that OPPO achieves state-of-the-art performance on MER-UniBench and MME-Emotion, while substantially improving utilization and faithfulness scores on MEP-Bench, highlighting the importance of sufficient and faithful omni perception for multimodal emotion reasoning.", "date": "2026-06-24", "categories": []}
{"id": "2606.26083", "title": "Real-Time Voice AI Hears but Does Not Listen", "abstract": "Speech conveys information through both words and vocal delivery. We evaluate four leading production realtime voice systems-OpenAI's GPT Realtime 2, Google's Gemini 3.1 Flash Live, and Alibaba's Qwen3.5 Omni Plus and Omni Flash-on tasks where the words and the delivery patterns both convey meaningful information. Across three consequential scenarios, all four systems act on the words rather than the voice. They end calls with crying callers who insist nothing is wrong, approve wire transfers authorized in frightened voices, and enroll callers whose agreement is clearly sarcastic. Surprisingly, this is often not a failure of perception. When asked directly, three of the four systems reliably identify the distress, fear, or sarcasm they later ignore when making decisions. We observe a similar pattern when these realtime voice systems estimate accent and age, as their responses frequently follow the biases of the words rather than the acoustic properties of the speaker. We term this disconnect between perception and action the emotional intelligence gap of voice AI. Prompting systems to explicitly attend to vocal delivery improves performance only partially and inconsistently. Our findings show that current realtime voice AI systems often behave as if speech had been reduced to a transcript, suggesting that they should be used with caution in settings where the tone and emotion of delivery convey important information.", "date": "2026-06-24", "categories": []}
{"id": "2606.26723", "title": "State-Specific Respiratory Signatures for Affective and Stress Recognition: Interpretable Respiratory Markers, Autocorrelation Lags, and Compact CNN Models", "abstract": "Respiratory activity is a direct and interpretable physiological channel for wearable stress and affective-state recognition, yet many studies emphasize classification accuracy without identifying which respiratory properties separate different states. This work reframes RESP-based recognition as a joint predictive and explanatory problem. Using the chest respiratory channel of the WESAD dataset, we analyze 60 s windows under leave-one-subject-out validation and combine two complementary branches: compact raw-signal one-dimensional convolutional neural networks (1D-CNNs) and physically grouped handcrafted respiratory signatures. The primary application task is binary stress versus non-stress detection, while baseline, stress, amusement, and meditation are additionally analyzed in a one-vs-rest setting to reveal state-specific respiratory markers. The feature space is organized into respiratory timing, breath-to-breath variability, waveform statistics, spectral/time-frequency descriptors, and autocorrelation/nonlinear predictability descriptors, with the raw 60 s signal treated as a sixth representation for the CNN branch. We introduce autocorrelation transition lags (Zpm/Zmp) as interpretable markers of respiratory correlation scale and separately evaluate exploratory FEG-Pro/Lyapunov-like descriptors. In the final CNN refit setting, the raw-signal model achieved the strongest stress-vs-rest performance, with accuracy 96.72 percent, macro-F1 95.30 percent, and MCC 90.61 percent. In contrast, compact feature models were stronger for baseline, with MCC 65.34 percent, amusement, with MCC 35.69 percent, and especially meditation, with MCC 88.65 percent. These results show that CNNs are most useful for the practical stress detector, whereas interpretable respiratory signatures provide stronger and more physiologically transparent state-specific markers for several non-stress conditions.", "date": "2026-06-25", "categories": []}
{"id": "2606.27652", "title": "MER-R1: Multimodal Emotion Reasoning via Slow-Fast Thinking Synergy", "abstract": "We find that explicit reasoning does not necessarily translate into better multimodal emotion recognition (MER) accuracy, even though it makes predictions more interpretable. Specifically, for reasoning-based MLLMs, fast thinking by triggering direct answers often outperforms slow thinking after deliberative reasoning. Our empirical analyses show that fast thinking improves recall with broader and more confident predictions, whereas slow thinking favors precision through conservative filtering of incorrect categories. Building on these insights, we propose MER-R1, a reinforcement learning framework that turns slow-fast complementarity into explicit optimization. Dual-objective disentanglement separates recall and precision into two optimization signals, allowing them to be jointly optimized rather than traded off against each other. Slow-fast confidence calibration further aligns the final slow-thinking answer with fast-thinking intuition, strengthening correct emotions while suppressing incorrect ones. In this way, MER-R1 unifies the recall-oriented intuition of fast thinking with the precision-oriented selectivity of slow thinking. We further provide theoretical justification for this synergy, showing that it mitigates variance-induced interference during optimization. Extensive experiments on MER-UniBench and MME-Emotion show that MER-R1 achieves state-of-the-art performance and makes reasoning genuinely benefit emotion recognition.", "date": "2026-06-26", "categories": []}
{"id": "2606.28083", "title": "STAG: Spatio-temporal Evolving Structural Representation of Action Units for Micro-expression Recognition", "abstract": "Micro-expression recognition is challenging due to subtle and short-lived facial muscle movements. Existing methods rely heavily on apex-onset frames, overlook fine-grained inter-frame dynamics, and separately model spatial and temporal information, limiting generalization across datasets. To address these challenges, we propose STAG, a dynamic ROI-AU-coupled spatial-temporal network that jointly models motion flow and adaptive facial connectivity. The framework extracts optical flow from discriminative frames using magnitude-based selection and temporal attention. A dual-branch architecture combines an enhanced graph attention network for structured spatial reasoning with a transformer encoder for temporal modeling. A bidirectional cross-attention module enables mutual refinement of spatial and temporal features, while AU-guided dynamic connectivity adapts facial region interactions according to muscle activation patterns. The transformer captures subtle temporal dynamics beyond apex-based approaches, improving semantic consistency and interpretability for explainable micro-expression recognition. The fused representation is optimized using focal loss and evaluated on CASME II, 4DME, DFME, NaME, SAMM, and SMIC-HS. Extensive experiments demonstrate improved robustness, generalization, interpretability, and computational efficiency, confirming the effectiveness of adaptive relational reasoning, AU-guided dynamic connectivity, and deep spatial-temporal feature fusion for accurate cross-dataset micro-expression recognition.", "date": "2026-06-26", "categories": []}
{"id": "2606.28769", "title": "Generative Learning as a Tool to Improve Perception of Emotional Body Motion Expressions", "abstract": "Emotional body motion expressions are an essential element of non-verbal communication. Effectively conveying these expressions through technology is of utmost importance, for example, with virtual reality avatars and in social robotics. Recent advances in generative models have opened new opportunities for advancing research on emotional body motion learning. However, generating accurate emotional expression representations is challenging, given the subtlety of emotional cues, individual variability, and cultural differences. We investigate whether a generative model can implicitly learn emotional body motions directly from culturally grounded motion-capture data, without explicit emotion-motion guidance. Using a dataset of emotional performances by 49 Japanese actors, we trained a Transformer-based generative model to generate expressive motions conditioned on 13 discrete emotion labels. We evaluate the generated motions from two perspectives: (1) an LSTM-based classifier to assess recognizability by machine observers, achieving a recognition accuracy of 22.80%, and (2) a human perception study with Japanese raters to assess alignment with human affective interpretations, yielding a recognition accuracy of 24.91%. Beyond these, we evaluate the utility of generative modeling for three practical tasks: augmenting emotion recognition models, extracting representative emotion-specific motion patterns, and synthesizing smooth transitions between emotion intensities. Our findings highlight the potential of implicit, data-driven generative modeling to enhance affective computing applications and our understanding of emotion expressions.", "date": "2026-06-27", "categories": []}
{"id": "2606.29068", "title": "A Comparative Study on Affective Cues in Text Embeddings Across Psychological Emotion Theories", "abstract": "Text encoders are known for their utility in natural language processing, as they are able to efficiently compress inputs into dense vectors while preserving semantics. These models have been applied to affective computing, in particular to help with solving sentiment analysis and emotion recognition tasks. Nevertheless, it remains unclear to what extent the latent representations produced by modern text encoders capture well-defined psychological theories of affect. In this work, we investigate the affective capabilities of twelve recently released text encoders by probing their generated embeddings as input features for solving regression and classification tasks across three established emotion frameworks, using both word- and sentence-level data. Additionally, we apply a semantic data-leakage prevention technique to improve robustness in word-level evaluations. Our main findings show that the latent manifolds of the latest instruction-aware open-weight encoders enclose an equal or even a larger amount of affective information in comparison with proprietary counterparts when evaluated at word level. In contrast, embeddings of task-tuned and proprietary encoders reach the highest scores on sentence-level affective classification. Furthermore, a qualitative analysis of latent representations and their encoded affective cues is provided.", "date": "2026-06-27", "categories": []}
{"id": "2606.30543", "title": "TRACE: Temporal Relationship-Aware Conversational Entrainment Detection in Dyadic Speech", "abstract": "With the proliferation of speech AI agents, understanding emotional entrainment in conversational interaction has become increasingly important. Emotional entrainment is shaped by social relationships and conversational context, influencing affective coordination over time. We introduce DyadEE, a dataset for emotional entrainment detection in dyadic speech interactions, containing both emotionally entrained conversations and synthetic interactions where entrainment is disrupted through partner swapping and emotion resynthesis. We further propose TRACE, a window-level framework that models dyadic interaction as ordered sequences of acoustic embeddings derived from emotion fine-tuned Whisper representations, treating each sample as an interaction trace rather than pooled utterances. Experimental results on DyadEE show that incorporating conversational context and relationship information improves emotional entrainment detection, with TRACE achieving the best accuracy of 97.01%.", "date": "2026-06-29", "categories": []}
{"id": "2606.31249", "title": "Rethinking the Role of Feature Engineering and Learning Strategies in Few-Shot Hidden Emotion Recognition", "abstract": "In this paper, we present the solution developed by our team, XInsight Lab, which achieved first place in Track 3 of the 4th EI-MIGA-IJCAI Challenge with a test accuracy of 0.76923. To address the challenge of weak and sparse implicit emotion evidence in long videos, this paper extends the winning solution from the previous competition and proposes a compact multi-modal temporal modeling framework. The framework integrates and evaluates the effects of multi-source features, including 2D/3D skeletons, facial expression Blendshapes, DINOv2/v3 vision foundation models, X-CLIP video features, and Gemini semantic priors. Architecturally, we propose a cross-attention mechanism that utilizes static pose features, denoted as Base, as the Query and dynamic micro-motion differential features, denoted as Offset, as the Key and Value. By capturing local relative velocities, this mechanism eliminates static biases related to individual body shape and identity. Concurrently, an adaptive pooling method based on Multiple Instance Learning is employed to extract instantaneous emotions while suppressing background noise in long sequences. Finally, the paper reveals the representation collapse phenomenon of general vision foundation models in micro-dynamic tasks, and analyzes the underlying mechanisms where networks fall into public-leaderboard-driven pseudo-generalization due to shortcut learning and rote memorization.", "date": "2026-06-30", "categories": []}
{"id": "2606.31442", "title": "Who Determines the Meaning of an Emotion? Affective Sovereignty as an Epistemic Consequence of Measurement Limits", "abstract": "Emotion-sensing AI is rapidly becoming embedded in vehicles, home appliances, dialogue agents, and social infrastructure, giving rise to a sphere in which emotion is no longer confined to individual experience but is instead observed and computed at a societal scale, a domain we term the Affectosphere. Yet a central normative question in this domain has remained underexplored: who has the final authority to determine the meaning of one's own emotion? This study addresses the question from the epistemological side of measurement's structural limits. We define a meaning distribution as the distribution of labels assigned by annotators drawn from a population under a fixed annotation protocol, and decompose its uncertainty into reducible and irreducible components. We then demonstrate that, while emotion AI can assign high-confidence point labels and discriminate real differences at an aggregate level, the irreducible component of the meaning distribution for individual instances cannot be estimated with adequate coverage under realistic annotator counts, a systematic divergence we term the epistemic gap. The key finding is that high device confidence does not constitute evidence that irrecoverable meaning has been recovered. From this epistemic gap, together with an explicitly stated normative premise, namely that the output of a system which cannot recover a quantity in principle must not be treated as its authoritative determination, we derive the norm that the final interpretive authority over the meaning of one's emotion is procedurally reserved for the experiencing subject, the norm of affective sovereignty. These results suggest that the design, evaluation, and regulation of emotion AI should place explicit allocation of interpretive authority, rather than accuracy maximisation, at their core.", "date": "2026-06-30", "categories": []}