README.md

metadata

title: Developmental Stuttering Detection
emoji: 🎤
colorFrom: green
colorTo: blue
sdk: gradio
sdk_version: 6.20.0
python_version: '3.11'
app_file: app.py
pinned: false

Developmental Stuttering Detection

Screens recorded speech for signs of stutter-like disfluencies using AI.

Part of the NeuroSage multi-modal neurodevelopmental assessment system.

Usage

1. Upload an audio file (WAV recommended, 4+ seconds)
2. Adjust the SLD threshold if needed
3. Click "Analyze" to get results

Features

• Analysis Tab: Quick SLD detection with JSON results
• Detailed Report Tab: Gemini-powered clinical-style analysis

Model

• Architecture: ChildFluencyNet (WavLM-Large + Linguistic Position Encoding)
• Training Data: UCLASS (3,547 windows, 20 child speakers)
• Performance: F1=0.94, MCC=0.40, AUC=0.80

Disclaimer

This is a screening aid only — not a clinical diagnosis. Always consult a Speech-Language Pathologist for formal evaluation.