diff --git a/marked/P/T-REC-P.1010-200407-I_PDF-E/raw.md b/marked/P/T-REC-P.1010-200407-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..26973700e7ff324bc5c5f2ed9c42a3fb5623829c --- /dev/null +++ b/marked/P/T-REC-P.1010-200407-I_PDF-E/raw.md @@ -0,0 +1,339 @@ + + +![ITU logo: A globe with a red lightning bolt and the letters ITU.](2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg) + +ITU logo: A globe with a red lightning bolt and the letters ITU. + +INTERNATIONAL TELECOMMUNICATION UNION + +**ITU-T** + +**P.1010** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +(07/2004) + +SERIES P: TELEPHONE TRANSMISSION QUALITY, +TELEPHONE INSTALLATIONS, LOCAL LINE +NETWORKS + +Transmission performance and QoS aspects of IP end-points + +--- + +**Fundamental voice transmission objectives for +VoIP terminals and gateways** + +ITU-T Recommendation P.1010 + +--- + +# ITU-T P-SERIES RECOMMENDATIONS + +## TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS + +| | | | +|-----------------------------------------------------------------------------------------------|---------------|---------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Subscribers' lines and sets | Series | P.30
P.300 | +| Transmission standards | Series | P.40 | +| Objective measuring apparatus | Series | P.50
P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of quality | Series | P.80
P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | +| Transmission performance and QoS aspects of IP end-points | Series | P.1000 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +# **ITU-T Recommendation P.1010** + +# **Fundamental voice transmission objectives for VoIP terminals and gateways** + +## **Summary** + +This Recommendation provides 3.1 kHz telephony speech transmission performance requirements for the whole range of packet-based gateways and terminals, including wireless and softphones. Measurement methodologies are not covered by this Recommendation, however, work on this topic is under way in Study Group 12 and is planned to be incorporated in a future revision or, alternatively, in a separate new Recommendation. Also, requirements for wideband telephony may be added in a future version of this Recommendation. + +###### **Source** + +ITU-T Recommendation P.1010 was approved on 7 July 2004 by ITU-T Study Group 12 (2001-2004) under the ITU-T Recommendation A.8 procedure. + +# FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications. The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure e.g. interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementors are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database. + +© ITU 2004 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +# CONTENTS + +| | Page | +|------------------------------------------------------------------------|------| +| Scope ..... | 1 | +| 1 Introduction ..... | 1 | +| 2 References..... | 1 | +| 3 Terms and definitions ..... | 2 | +| 4 Abbreviations and acronyms ..... | 3 | +| 5 End-to-end considerations ..... | 3 | +| 6 Transmission characteristics..... | 3 | +| 6.1 Default coding algorithm..... | 4 | +| 6.2 Delay..... | 4 | +| 6.3 Send Loudness Rating ..... | 5 | +| 6.4 Receive Loudness Rating ..... | 5 | +| 6.5 Weighted Terminal Coupling Loss (TCLw) ..... | 6 | +| 6.6 Residual Echo Level..... | 6 | +| 6.7 Further parameters with respect to speech processing devices ..... | 6 | +| 7 Measurement considerations ..... | 6 | + + + +# Fundamental voice transmission objectives for VoIP terminals and gateways + +## Scope + +This Recommendation provides fundamental speech transmission performance requirements for 3.1 kHz handset telephony of packet-based terminals and gateways; it applies to both access gateways and trunking gateways. Requirements are given for handset operation, with the assumption that the corresponding requirements for headset and handsfree are obtained via appropriate conversions to take into account the different acoustic reference points. + +This Recommendation addresses the whole range of IP-based gateways and terminals, including wireless and softphones. + +Measurement methodologies are not covered by this Recommendation; however, work on this topic is under way in Study Group 12 and is planned to be incorporated in a future revision or alternatively in a separate new Recommendation. + +# 1 Introduction + +This Recommendation limits its scope to the fundamental speech transmission performance requirements for VoIP terminals and VoIP gateways with respect to the interface with the packet-based network. Thus, a number of traditional parameters have not been considered in this Recommendation. This is mainly because it is anticipated that the respective requirements from existing Recommendations, e.g., P.310 [11] or in the Q.55x series of Recommendations, can be applied to VoIP terminals and VoIP gateways in an appropriate way. Attention should also be paid to the end-to-end considerations provided in clause 5. + +# 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- [1] ITU-T Recommendation G.100 (2001), *Definitions used in Recommendations on general characteristics of international telephone connections and circuits*. +- [2] ITU-T Recommendation G.107 (2003), *The E-model, a computational model for use in transmission planning*. +- [3] ITU-T Recommendation G.108 (1999), *Application of the E-model: A planning guide, plus Amendment 1 (2003): New Appendix I – The relationship between and interaction of talker echo and absolute delay, and Amendment 2 (2004): New Appendix II – Planning examples regarding delay in packet-based networks*. +- [4] ITU-T Recommendation G.108.2 (2003), *Transmission planning aspects of echo cancellers*. +- [5] ITU-T Recommendation G.109 (1999), *Definition of categories of speech transmission quality*. + +- [6] ITU-T Recommendation G.168 (2004), *Digital network echo cancellers*. +- [7] ITU-T Recommendation G.711 (1988), *Pulse code modulation (PCM) of voice frequencies*. +- [8] ITU-T Recommendation G.799.1/Y.1451.1 (2004), *Functionality and interface specifications for GSTN transport network equipment for interconnecting GSTN and IP networks*. +- [9] ITU-T Recommendation G.1020 (2003), *Performance parameter definitions for quality of speech and other voiceband applications utilizing IP network*, plus Amendment 1. +- [10] ITU-T Recommendation P.10 (1998), *Vocabulary of terms on telephone transmission quality and telephone sets*, plus Amendment 1 (2003): *New Annex A – List of psychoacoustic parameters*. +- [11] ITU-T Recommendation P.310 (2003), *Transmission characteristics for telephone band (300-3400 Hz) digital telephones*. +- [12] ITU-T Recommendation P.330 (2003), *Speech processing devices for acoustic enhancement*. +- [13] ITU-T Recommendation P.340 (2000), *Transmission characteristics and speech quality parameters of hands-free terminals*. +- [14] ITU-T Recommendation P.380 (2003), *Electro-acoustic measurements on headsets*. +- [15] ITU-T Recommendation P.501 (2000), *Test signals for use in telephonometry*. +- [16] ITU-T Recommendation P.502 (2000), *Objective test methods for speech communication systems using complex test signals*. +- [17] ITU-T Recommendation Q.115.1 (2002), *Logic for the control of echo control devices and functions*. +- [18] ITU-T Recommendation Y.1541 (2002), *Network performance objectives for IP-based services*. +- [19] ANSI/TIA/EIA-810-A (2000), *Transmission Requirements for Narrowband Voice over IP and Voice over PCM Digital Wireline Telephones*. + +# 3 Terms and definitions + +Beside those in ITU-T Recs P.10 [10] and G.100 [1] this Recommendation defines the following terms: + +**3.1 send loudness rating (SLR):** $SLR = SLR(\text{set}) + \sum_{i=1}^n CLR_i$ + +**3.2 receive loudness rating (RLR):** $RLR = RLR(\text{set}) + \sum_{i=n+1}^N CLR_i$ + +**3.3 overall loudness rating (OLR):** $OLR = SLR + RLR$ + +**3.4 circuit loudness rating (CLR):** The loudness loss between two electrical interfaces in a connection or circuit. + +**3.5 softphone:** A softphone is an implementation of a VoIP terminal which is realized by a software program enabling a personal computer (PC) or personal digital assistant (PDA) equipped with an acoustic interface (microphone and earpiece/loudspeaker) to be used for two-way voice communications. + +# 4 Abbreviations and acronyms + +This Recommendation uses the following abbreviations: + +| | | +|----------|---------------------------------| +| CLR | Circuit Loudness Rating | +| IP | Internet Protocol | +| IPDV | IP Packet Delay Variation | +| OLR | Overall Loudness Rating | +| RLR | Receive Loudness Rating | +| RLR(set) | RLR of a telephone set | +| SCN | Switched-Circuit Network | +| SLR | Send Loudness Rating | +| SLR(set) | SLR of a telephone set | +| T | Mean One-Way Delay | +| TCLw | Weighted Terminal Coupling Loss | +| UNI | User-Network Interface | +| VAD | Voice Activity Detection | +| VoIP | Voice over Internet Protocol | + +# 5 End-to-end considerations + +In order to achieve a desired end-to-end speech transmission performance (mouth-to-ear) it is recommended that transmission planning tasks be carried out with the E-model of ITU-T Rec. G.107 [2] as described in ITU-T Rec. G.108 [3]; this includes the *a priori* determination of the desired category of speech transmission quality as defined in ITU-T Rec. G.109 [5]. + +While, in general, the transmission characteristics of single circuit-oriented network elements, such as switches or terminals, can be assumed to have a single input value for the planning tasks of ITU-T Rec. G.108 [3], this approach is not applicable in packet-based systems and, thus, there is a need for the transmission planner's specific attention. + +In particular, the decision as to which delay category given in this Recommendation should be chosen to represent the specific configuration is the responsibility of the individual transmission planner. + +ITU-T Rec. G.108 with its Amendments [3] provides further guidance on this important issue. + +# 6 Transmission characteristics + +The requirements specified in this clause apply to both VoIP terminals and VoIP gateways with respect to the interface with the packet-based network as illustrated in Figure 1. + +![Figure 1/P.1010 – Area of applicability of this Recommendation. The diagram shows two parallel horizontal signal paths. Top path: A VoIP terminal (represented by a box with a handset icon) connects to a box labeled 'path through IP network(s)'. Between them is a dot labeled 'UNI'. Another 'UNI' dot is at the output of the network path. A double-headed arrow spans the VoIP terminal. Bottom path: A box labeled 'SCN' connects to a 'VoIP gateway' box, which then connects to a 'path through IP network(s)' box. 'UNI' dots are placed between the gateway and network, and at the network output. A double-headed arrow spans the VoIP gateway. The label 'P1010_F1' is in the bottom right.](a5ee5c23b6dc52ec1d724b76d5a5f58f_img.jpg) + +Figure 1/P.1010 – Area of applicability of this Recommendation. The diagram shows two parallel horizontal signal paths. Top path: A VoIP terminal (represented by a box with a handset icon) connects to a box labeled 'path through IP network(s)'. Between them is a dot labeled 'UNI'. Another 'UNI' dot is at the output of the network path. A double-headed arrow spans the VoIP terminal. Bottom path: A box labeled 'SCN' connects to a 'VoIP gateway' box, which then connects to a 'path through IP network(s)' box. 'UNI' dots are placed between the gateway and network, and at the network output. A double-headed arrow spans the VoIP gateway. The label 'P1010\_F1' is in the bottom right. + +**Figure 1/P.1010 – Area of applicability of this Recommendation** + +## 6.1 Default coding algorithm + +VoIP terminals and VoIP gateways shall support the coding algorithm according to ITU-T Rec. G.711 [7] (both $\mu$ -law and A-law). VoIP terminals and VoIP gateways may support other coding algorithms. + +## 6.2 Delay + +While delay incurred from traditional terminals was significantly lower than the delay possibly incurred from the network, this is no longer true for VoIP terminals and for VoIP gateways. + +Hence, for traditional terminals, the appropriate Recommendations (e.g., ITU-T Rec. P.310) were able to specify one single requirement for delay. + +With the introduction of packet-based transmission, however, the increased delays introduced by the terminals and the gateways can have a tremendous impact on the end-to-end delay and, thus, need proper consideration in transmission planning. For example, depending on the delay variation accumulated by packets traversing an IP network, the design of the de-jitter buffer of an IP gateway or terminal is extremely critical and has to consider a trade-off between the permissible number of dropped packets and the maximum delay to be incurred (see Figure 3/G.1020 [9]). + +Accordingly, for the parameter of delay, IP terminals and gateways may be designed to operate in a number of modes, with the mode selection being based on end-to-end considerations, especially desired speech quality level and likely network configurations. + +For IP networks which comply with or which perform better than Class 0 of ITU-T Rec. Y.1541 [18], terminals or gateways that comply with any of the three delay categories given in this Recommendation will be a starting point to achieve acceptable speech quality. + +For other kinds of networks, for example, networks with larger delay variations than any class of Y.1541 limits, only delay categories A or B may be a valid choice. Furthermore, there are cases such as the interconnection of two local IP islands via an intercontinental path through the PSTN, where only delay category A will be a valid basis for achieving a reasonable end-to-end speech transmission performance (due to the long delay of the PSTN path). + +Irrespective of network configurations encountered, the deployment of some low bit rate codecs may further decrease the perceived speech quality and narrow the valid choice of delay categories. + +The following subclauses specify three delay categories for either send or receive directions. It is the common understanding that, in order to comply with a specific delay category, a VoIP terminal or a VoIP gateway has to comply with both send and receive requirements of this category. + +NOTE – For some VoIP terminals, e.g., softphones, it is well recognized that, under certain circumstances, technical and practical considerations may not permit any of the above categories to be met at this time. Therefore, this issue, as well as the introduction of a fourth delay category, are for further study. + +### 6.2.1 Send delay categories (provisional values) + +For a VoIP terminal, send delay is defined as the one-way delay from the acoustical input (mouthpiece) of this VoIP terminal to its interface to the packet-based network. VoIP gateway send delay is defined as the one-way delay from the electrical input (TDM or analogue) of this VoIP gateway to the interface to the packet-based network. In each case, the total send delay is the upper bound on the mean delay and takes into account the delay contributions of all of the elements shown in Figures 2 and A.1/G.1020 [9], respectively. + +- Category A: $T_s \leq 20$ ms; +- Category B: $T_s \leq 35$ ms; +- Category C: $T_s \leq 50$ ms. + +### 6.2.2 Receive delay categories (provisional values) + +For a VoIP terminal, receive delay is defined as the one-way delay from the interface to the packet-based network of this VoIP terminal to its acoustical output (earpiece). VoIP gateway receive delay is defined as the one-way delay from the interface to the packet-based network to the electrical output (TDM or analogue) of this VoIP gateway. In each case, the total receive delay is the upper bound on the mean delay and takes into account the delay contributions of all of the elements shown in Figures 3 and A.2/G.1020 [9], respectively. + +- Category A: $T_r \leq 30$ ms; +- Category B: $T_r \leq 65$ ms; +- Category C: $T_r \leq 100$ ms. + +#### 6.2.2.1 IP Packet delay variation (IPDV) + +VoIP terminals and gateways should accommodate network IPDV values of up to 50 ms; this is consistent with ITU-T Rec. Y.1541 Classes 0 and 1. It is left to the designer's discretion whether the VoIP terminal or gateway is capable of accommodation of IPDV values above this value, which may occur in other networks. + +## 6.3 Send Loudness Rating + +For VoIP terminals, the recommended nominal value of Send Loudness Rating (SLR) should be: + +- $SLR = SLR(set) = 8$ dB. + +For VoIP gateways, the recommended nominal value of Send Loudness Rating (SLR) should be: + +- $SLR = SLR(set) + CLR = 8$ dB. + +For the gateway at the ingress to the IP network, the nominal SLR has to be achieved solely by means of transmission planning; this requires a proper dimensioning of the CLR. + +## 6.4 Receive Loudness Rating + +For VoIP terminals, the recommended nominal value of Receive Loudness Rating (RLR) should be: + +- $RLR = RLR(set) = 2$ dB. + +For VoIP gateways, the recommended nominal value of Receive Loudness Rating (RLR) should be: + +- $RLR = RLR(set) + CLR = 2$ dB. + +For the gateway at the egress from the IP network, the nominal RLR has to be achieved solely by means of transmission planning; this requires a proper dimensioning of the CLR. + +## 6.5 Weighted Terminal Coupling Loss (TCLw) + +Due to the increased delay values introduced by VoIP terminals compared to traditional digital terminals as per ITU-T Rec. P.310 [11], the TCLw requirement needs to be increased (see also ANSI/TIA/EIA-810-A [19]). + +For VoIP terminals the recommended nominal value of weighted Terminal Coupling Loss (TCLw) should be: + +- $TCLw \geq 55$ dB + +For further details see clause 10/P.310 [11]; in addition, the TCLw has to be maintained under all operational conditions. Relaxation of the TCLw value by not more than 10 dB is only permissible during double-talk periods. Note that this value is provisional and for further study. + +NOTE – TCLw is strongly influenced by the amount of acoustic coupling in the acoustic interface. In the absence of any standardized requirements for this interface, the recommended value of TCLw may not always be met. Nevertheless, in the interest of achieving satisfactory speech quality, designers of VoIP terminals should strive to meet this value as closely as possible, taking into account the characteristics of the acoustic interfaces to be used. + +## 6.6 Residual Echo Level + +For VoIP gateways, an echo canceller according to ITU-T Rec. G.168 [6], is recommended for cancelling echoes that may originate in the PSTN. For details regarding control of Echo Cancellers see Annex A/Q.115.1 [17]. Details on the residual echo level that may be achieved and on transmission planning aspects for echo cancelling devices, are discussed in ITU-T Rec. G.108.2 [4]. Details on functional aspects of trunking gateways can be found in ITU-T Rec. G.799.1/Y.1451.1 [8]. + +## 6.7 Further parameters with respect to speech processing devices + +For VoIP terminals and gateways that contain non-linear speech processing devices, the following parameters require additional attention and are for further study in the context of this Recommendation: + +- objective evaluation of speech quality for VoIP gateways and VoIP terminals; +- double-talk capability; +- switching behaviour; +- partial echo effects; +- occurrence of artefacts; +- time-variant impairments. + +As is the case for traditional telephone sets, reduced performance for handsfree and headset mode operation is anticipated for VoIP terminals. Further guidance in this area is provided by ITU-T Recs P.340 [13] and P.380 [14]. + +# 7 Measurement considerations + +Measurement methodologies are not covered by this Recommendation; however, work on this topic is under way in Study Group 12 and is planned to be included in a future revision or, alternatively, in a separate new Recommendation. + +Nevertheless, it can be assumed that VoIP terminals and gateways contain non-linear and time-variant functions (e.g., automatic gain control, adaptive jitter buffer, VAD). This needs to be considered when approaching the measurement of transmission characteristics. ITU-T Recs P.330 [12], P.501 [15] and P.502 [16] provide further guidance in this respect. + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|--------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series B | Means of expression: definitions, symbols, classification | +| Series C | General telecommunication statistics | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Construction, installation and protection of cables and other elements of outside plant | +| Series M | TMN and network maintenance: international transmission systems, telephone circuits, telegraphy, facsimile and leased circuits | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks and open system communications | +| Series Y | Global information infrastructure, Internet protocol aspects and Next Generation Networks | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.1201-201210-I_PDF-E/raw.md b/marked/P/T-REC-P.1201-201210-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..0a28d0cb3189c60b0f23e1e1f989b83f0611f14f --- /dev/null +++ b/marked/P/T-REC-P.1201-201210-I_PDF-E/raw.md @@ -0,0 +1,799 @@ + + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.1201** + +(10/2012) + +SERIES P: TERMINALS AND SUBJECTIVE AND +OBJECTIVE ASSESSMENT METHODS + +Models and tools for quality assessment of streamed +media + +# --- **Parametric non-intrusive assessment of audiovisual media streaming quality** + +Recommendation ITU-T P.1201 + +![ITU logo](84a1d09fb489061482111515543b60dc_img.jpg) + +The logo of the International Telecommunication Union (ITU) features a globe with a red lightning bolt striking across it. To the right of the globe, the text "International Telecommunication Union" is written in a blue, sans-serif font, with "ITU" in a larger, bold font above it. + +ITU logo + +## ITU-T P-SERIES RECOMMENDATIONS **TERMINALS AND SUBJECTIVE AND OBJECTIVE ASSESSMENT METHODS** + +| | | | +|-----------------------------------------------------------------------------------------------|---------------|---------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Voice terminal characteristics | Series | P.30 | +| | | P.300 | +| Reference systems | Series | P.40 | +| Objective measuring apparatus | Series | P.50 | +| | | P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of speech quality | Series | P.80 | +| | | P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | +| Transmission performance and QoS aspects of IP end-points | Series | P.1000 | +| Communications involving vehicles | Series | P.1100 | +| Models and tools for quality assessment of streamed media | Series | P.1200 | +| Telemeeting assessment | Series | P.1300 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | Series | P.1400 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +## Recommendation ITU-T P.1201 + +# Parametric non-intrusive assessment of audiovisual media streaming quality + +## Summary + +Recommendation ITU-T P.1201 provides an overview of algorithmic models for non-intrusive monitoring of the audio, video and audiovisual quality of IP-based video services based on packet-header information. The ITU-T P.1201-series of Recommendations addresses two application areas: + +- ITU-T P.1201.1 specifies the model algorithm for the lower resolution (LR) application area, including services such as mobile TV. +- ITU-T P.1201.2 specifies the model algorithm for the higher resolution (HR) application area, including services such as IPTV. + +The two ITU-T P.1201 model algorithms are no-reference (i.e., non-intrusive) models which operate by analysing packet header information as available from respective packet trace data, provided to the model algorithms in the packet capture format (PCAP). Further input information on more general aspects of the stream, such as the video resolution, which may not be available from packet header information, is provided to the model algorithm out-of band, for example in the form of stream-specific side information. + +As output, the model algorithms provide individual estimates of audio, video and audiovisual quality in terms of the five-point absolute category rating (ACR) mean opinion score (MOS) scale. Further, diagnostic information on causes of quality degradations can also be made available. + +Complementary to the ITU-T P.1201 models, there are further models described in the ITU-T P.1202-series of Recommendations. These so-called ITU-T P.1202 models are bitstream-based video quality models. The main differences with ITU-T P.1201 can be summarized as follows: + +- The ITU-T P.1201 models provide audio, video and audiovisual quality estimates, while the ITU-T P.1202-only models provide video quality estimates. +- The ITU-T P.1201 models use packet header information, while the ITU-T P.1202 models exploit further bitstream information, such as coding-related information. As a consequence, the ITU-T P.1202-models can be more accurate in their quality predictions. In turn, they require non-encrypted streams, to enable the access to payload information. Since the ITU-T P.1202 models are more complex, they also require more computational power to estimate the video quality. + +## History + +| Edition | Recommendation | Approval | Study Group | +|---------|----------------------------|------------|-------------| +| 1.0 | ITU-T P.1201 | 2012-10-14 | 12 | +| 1.1 | ITU-T P.1201 (2012) Amd. 1 | 2013-03-28 | 12 | + +## Keywords + +Audio, audiovisual, IPTV, mean opinion score (MOS), mobile TV, monitoring, multimedia, QoE, video. + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database at . + +© ITU 2013 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## Table of Contents + +| | Page | +|------------------------------------------------------------------------------------------------------|------| +| 1 Scope ..... | 1 | +| 2 References..... | 4 | +| 3 Definitions ..... | 5 | +| 3.1 Terms defined elsewhere ..... | 5 | +| 3.2 Terms defined in this Recommendation..... | 5 | +| 4 Abbreviations and acronyms ..... | 5 | +| 5 Conventions ..... | 7 | +| 6 Areas of application..... | 7 | +| 6.1 Application range for the models ..... | 7 | +| 6.2 Modes of operation..... | 11 | +| 7 Model input interfaces ..... | 14 | +| 8 Model output information and performance details ..... | 19 | +| 9 Description of the ITU-T P.1201 model algorithms..... | 20 | +| Appendix I – Detailed performance figures for the ITU-T P.1201.1 and ITU-T P.1201.2 algorithms ..... | 21 | +| Appendix II – Performance figures for the ITU-T P.1201.2 algorithm in the case of MPEG2-TS/UDP ..... | 23 | + + + +# Parametric non-intrusive assessment of audiovisual media streaming quality + +## 1 Scope + +This Recommendation provides an overview of two recommended objective parametric quality assessment models that predict the impact of observed IP network impairments on quality experienced by the end user in multimedia mobile streaming and IPTV applications over transport formats such as: RTP (over UDP), MPEG2-TS (over UDP or RTP/UDP), 3GPP-PSS (over RTP). + +As an umbrella Recommendation, ITU-T P.1201 provides the framework for the actual two algorithmic models described in [ITU-T P.1201.1] and [ITU-T P.1201.2]. These algorithms are aimed at monitoring the audio, video and audiovisual quality of IP-based video services based on packet-header information, with two application areas: + +- [ITU-T P.1201.1] specifies the model algorithm for the lower resolution application area, including services such as mobile TV. +- [ITU-T P.1201.2] specifies the model algorithm for the higher resolution application area, including services such as IPTV. + +These models are restricted to input information contained in packet headers, prior and static knowledge about the media stream and dynamic buffering information from the client. Information from decoding the bit-stream or parsing the packet payload is not used. + +These models predict mean opinion scores (MOS) on a five-point ACR scale (see [ITU-T P.910]) for audio and video parts of the stream, as well as a global audiovisual MOS score (as defined in [ITU-T P.911], for instance). + +The primary application for these models is the monitoring of transmission quality during in-service operation or for maintenance purposes. The ITU-T P.1201 model may be deployed both in end-point locations and at mid-network monitoring points. The location of the model and the location of the measurement probe together determine the mode of operation, as described in more detail in clause 6.1. + +The primary quality prediction made by such models is not based on the information in the audio and video coding layer (payload) of the stream being analysed, but on information on the payload as it is available from packet headers and additional side-information. + +This Recommendation cannot provide a comprehensive evaluation of audiovisual quality as perceived by a particular end user because its scores reflect the impairments due to encoding and the subsequent IP network being assessed, which may only be one part of the end-to-end connection. An explicit inclusion of processing steps, such as content contribution from, for example, satellite networks, display properties, etc., are not considered. Hence, also the effects of audio level, noise, delay (and corresponding similar video factors) and other impairments related to the media payload are not reflected in the scores computed by such a model. Further, the quality impact due to a specific audio or video encoder implementation or a specific decoder-side packet loss concealment implementation is not explicitly addressed. Instead, the models have been developed for a set of dedicated service implementations, which are assumed to be meaningful representations of today's IP-based streaming video services. As a consequence, however, in case of significant deviations of a given service being assessed from the service configurations used for developing this standard, it is possible to obtain high-quality scores with this Recommendation, but yet to have a poor quality of the stream as it is perceived by actual users, or vice versa. Moreover, the scores predicted by a parametric model (i.e., without access to payload information) necessarily reflect a somewhat simplified representation of the perceptual impairment of the considered stream. + +However, with only using packet-header information, the models still enable estimation of payload-related information, and thus allow to provide valid and, in most cases, accurate predictions, presuming that they are applied in an appropriate manner, following this Recommendation. + +As a consequence, this Recommendation can be used for applications such as: + +- in-service quality monitoring for specific IP-based audiovisual services, as specified in more detail below; +- benchmarking of different service implementations. However, it cannot be used for direct benchmarking of different encoder implementations, but only the effect of different encoding bitrates and transmission errors for a given decoder-based packet loss concealment. The implementations that can be assessed with ITU-T P.1201 include the audio and video encoding bitrates, the employed video GOP-structure, frame rate, resolution, potential packet loss and the audio codec type. + +The audio model algorithm is considered as a part of ITU-T P.1201 that may be employed in combination with other than the ITU-T P.1201 video model algorithm, for example, with the bitstream-based ITU-T P.1202 model algorithms (ITU-T P.1202-series of Recommendations), or be used as a stand-alone audio model. Consequently, care is taken that the application range for the audio model algorithm is designed to be slightly larger than necessary for a component of an audiovisual model applied to services as considered for the ITU-T P.1201 activity. + +The application areas of the ITU-T P.1201 model algorithms are summarized in Tables 1, 2, and 3 below: + +**Table 1 – Application areas, test factors and coding technologies where [ITU-T P.1201.1] and [ITU-T P.1201.2] have been verified and are known to produce reliable results. For details about the settings, see clause 6** + +| [ITU-T P.1201.1] – Lower resolution (LR) | [ITU-T P.1201.2] – Higher resolution (HR) | +|----------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------| +| Applications the models are intended for | | +| In-service monitoring of audiovisual, video, and audio UDP-based streaming | (same for both models, see entry on the left) | +| Performance and quality assessment of live networks (including codecs) including the effect due to encoding bitrate, and transmission errors | (same for both models, see entry on the left) | +| Test factors the models have been validated for | | +| Encoding (compression) degradation of audio and video with a variety of bitrates
Video: 40-6000 kbit/s
Audio: 4.75-576 kbit/s | Encoding (compression) degradation of audio and video with a variety of bitrates
Video: 0.5-30 Mbit/s
Audio: 16-384 kbit/s | +| Packet loss degradation of audio and video (both random and bursty packet loss patterns) | Packet loss degradation of audio and video (both random and bursty packet loss patterns) | +| Re-buffering degradation (audio-only re-buffering not validated) | – | +| Video contents of different spatio-temporal complexity | (same for both models, see entry on the left) | + +**Table 1 – Application areas, test factors and coding technologies where [ITU-T P.1201.1] and [ITU-T P.1201.2] have been verified and are known to produce reliable results. For details about the settings, see clause 6** + +| [ITU-T P.1201.1] – Lower resolution (LR) | [ITU-T P.1201.2] – Higher resolution (HR) | +|--------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| Different video keyframe and frame rates
Frame rates: 5-30 Hz
GOP lengths (1/keyframe rate): 2-10 s | Different video group of picture (GOP) structures and video frame rates
Supporting variable GOP structures with all frame types | +| Different video resolutions: HVGA, QVGA, QCIF | Different video resolutions: SD (PAL/NTSC), HD (720p50, 720p60, 1080i50, 1080p25, 1080i60, 1080p30) | +| Different decoder-side packet loss concealment strategies (freezing with skipping, one slice per RTP packet/frame) | Different decoder-side packet loss concealment strategies (freezing with skipping, one/multiple slices per frame – based on the slicing settings chosen in the encoder) | +| – | Interlaced and progressive scan | +| Coding technologies the models have been trained on | | +| Video: MPEG4 Part 2, ITU-T H.264 (MPEG4 Part 10) | Video: ITU-T H.264 (MPEG4 Part 10) | +| Audio: AMR-NB/WB+, AAC-LC, HE-AACv1/v2 | Audio: AAC-LC, HE-AACv2, AC3, MPEG-LII | + +**Table 2 – Application areas, test factors and coding technologies for which further investigation of ITU-T P.1201.1 and ITU-T P.1201.2 models is needed** + +| [ITU-T P.1201.1] – Lower resolution (LR) | [ITU-T P.1201.2] – Higher resolution (HR) | +|----------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| Applications where the models can be used, but the results may not be reliable | | +| In-service monitoring of live network audiovisual, video, and audio TCP-based streaming (assuming that parameter extraction from TCP-based streaming is implemented) | In-service monitoring of live network audiovisual, video, and audio TCP-based streaming without re-buffering (assuming that parameter extraction from TCP-based streaming is implemented) | +| Test factors where the models can be used but the results may not be reliable | | +| – | – | +| Coding technologies where the models can be used but the results may not be reliable | | +| Specific implementations of video encoders and decoders other than the codecs used in the development, see clause 6 | (same for both models, see entry on the left) | + +**Table 3 – Application areas, test factors, and coding technologies for which [ITU-T P.1201.1] and [ITU-T P.1201.2] are not intended to be used** + +| [ITU-T P.1201.1] – Lower resolution (LR) | [ITU-T P.1201.2] – Higher resolution (HR) | +|------------------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------| +| Applications the models are not intended for | | +| Direct comparison/benchmarking of encoder and decoder implementations, and thus of services that employ different encoder or decoder implementations | (same for both models, see entry on the left) | +| Evaluation of audiovisual quality including display/device properties | (same for both models, see entry on the left) | +| Test factors the models are not intended for | | +| Audiovisual streaming with significant rate adaptation (such as used in dynamic adaptive streaming over HTTP (DASH) streaming) | (same for both models, see entry on the left) | +| Transcoding situations | (same for both models, see entry on the left) | +| The effects of audio level, noise, delay (and corresponding similar video factors) | (same for both models, see entry on the left) | +| – | Re-buffering degradation of audio, video and audiovisual | +| Coding technologies the models are not intended for | | +| ITU-T H.261, MPEG-2, ITU-T H.263, ITU-T H.265, etc. (Note) | | +| NOTE – For the exact set of codecs for which the models have been validated, see clause 6. | | + +## 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- [ITU-T H.264] Recommendation ITU-T H.264 (2012), *Advanced video coding for generic audiovisual services*. +- [ITU-T P.800.1] Recommendation ITU-T P.800.1 (2006), *Mean Opinion Score (MOS) terminology*. +- [ITU-T P.910] Recommendation ITU-T P.910 (2008), *Subjective video quality assessment methods for multimedia applications*. +- [ITU-T P.911] Recommendation ITU-T P.911 (1998), *Subjective audiovisual quality assessment methods for multimedia applications*. +- [ITU-T P.1201.1] Recommendation ITU-T P.1201.1 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – Lower resolution application area*. + +- [ITU-T P.1201.2] Recommendation ITU-T P.1201.2 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – Higher resolution application area.* +- [ITU-T P.1202] Recommendation ITU-T P.1202 (2012), *Parametric non-intrusive bitstream assessment of video media streaming quality.* +- [ITU-T P.1202.1] Recommendation ITU-T P.1202.1 (2012), *Parametric non-intrusive bitstream assessment of video media streaming quality – Lower resolution application area.* +- [ITU-T P.1401] Recommendation ITU-T P.1401 (2012), *Methods, metrics and procedures for statistical evaluation, qualification and comparison of objective quality prediction models.* + +# 3 Definitions + +## 3.1 Terms defined elsewhere + +This Recommendation uses the following terms defined elsewhere: + +#### 3.1.1 mean opinion score (MOS): [ITU-T P.800.1]. + +## 3.2 Terms defined in this Recommendation + +This Recommendation defines the following terms: + +**3.2.1 model, model algorithm:** An algorithm with the purpose of estimating the subjective (perceived) quality of a media sequence. + +**3.2.2 sequence:** A short decoded audio, video or audiovisual portion of a stream, typically shorter than 30 seconds. + +**3.2.3 bitstream:** The part of an IP-based transmission where the actual audiovisual, video, or audio content is available in encoded and packetized form. + +**3.2.4 compression artefacts:** Artefacts introduced due to lossy compression of the encoding process. + +**3.2.5 slicing artefacts:** Artefacts introduced when packet losses are concealed using a packet loss concealment (PLC) scheme trying to repair erroneous frames. + +**3.2.6 freezing artefacts:** Artefacts introduced when the PLC scheme of the receiver replaces the erroneous frames (either due to packet loss or error propagation) with the previous error free frame until a decoded picture without errors has been received. Since the erroneous frames are not displayed, this type of artefact is also referred to as freezing with skipping. + +**3.2.7 rebuffering artefacts:** Artefacts coming from rebuffering events at the client side, which could be a result of video data arriving late. Usually, rebuffering events are indicated to the viewer, e.g., in the form of a spinning wheel. This is also referred to as freezing without skipping. + +## 4 Abbreviations and acronyms + +This Recommendation uses the following abbreviations and acronyms: + +- | | | +|--------|----------------------------------------| +| AAC | Advanced Audio Coding | +| AAC-LC | Advanced Audio Coding – Low Complexity | +| AC3 | Audio Coding 3 | +| ACR | Absolute Category Rating | + +| | | +|--------|----------------------------------------| +| AMR-NB | Adaptive Multi-Rate – Narrowband | +| AMR-WB | Adaptive Multi-Rate – Wideband | +| API | Application Programming Interface | +| ARQ | Automatic Repeat Request | +| DASH | Dynamic Adaptive Streaming over HTTP | +| FB | Fullband | +| FEC | Forward Error Correction | +| GOP | Group Of Pictures | +| HD | High Definition (television) | +| HE-AAC | High-Efficiency Advanced Audio Coding | +| HR | Higher Resolution | +| HRC | Hypothetical Reference Circuit | +| HTTP | Hypertext Transfer Protocol | +| HVGA | Half Video Graphics Array | +| IDR- | Instantaneous Decoder Refresh (frame) | +| I- | Inline-(frame) | +| LR | Lower Resolution | +| MBMS | Multimedia Broadcast/Multicast Service | +| MOS | Mean Opinion Score | +| MPEG | Motion Pictures Expert Group | +| NB | Narrowband | +| NTSC | National Television Standard Committee | +| PAL | Phase Alternating Line | +| PCAP | Packet Capture Format | +| PCC | Pearson Correlation Coefficient | +| PES | Packetized Elementary Stream | +| PLC | Packet Loss Concealment | +| PSS | Packet-switched Streaming Service | +| PVS | Processed Video Sequence | +| QCIF | Quarter Common Intermediate Format | +| QoE | Quality of Experience | +| QVGA | Quarter Video Graphics Array | +| RMSE | Root Mean Square Error | +| RTP | Real-time Transport Protocol | +| RTSP | Real Time Streaming Protocol | +| SD | Standard Definition | +| SRC | Source Reference Channel or Circuit | + +| | | +|-----|------------------------| +| SWB | Superwideband | +| TS | Transport Stream | +| UDP | User Datagram Protocol | +| VSP | Visual Simple Profile | +| WB | Wideband | + +## 5 Conventions + +None. + +# 6 Areas of application + +The two application areas for ITU-T P.1201 are: + +- ITU-T P.1201.1 (lower resolution (LR) mode): +QCIF-QVGA-HVGA, mostly for mobile TV and streaming with the sub-application areas: + - Linear mobile TV over RTP (includes mobile TV over a 3G mobile network with MBMS and with unicast transport over RTP/UDP/IP) + - Multimedia streaming (includes 3GPP PSS with transport over RTP/UDP/IP). +- ITU-T P.1201.2 (higher resolution (HR) mode): SD and HD television, mostly for IPTV with the sub-application areas: + - Linear broadcast TV (includes transmission over MPEG2-TS/RTP/UDP/IP, MPEG2-TS/UDP/IP (see [ITU-T P.1201.2]), and is assumed to be applicable to RTP/UDP/IP transport with similar, but so far unverified, accuracy as compared to MPEG2-TS/RTP/UDP/IP and MPEG2-TS/UDP/IP) + - Video on-demand (includes transmission over MPEG2-TS/RTP/UDP/IP, MPEG2-TS/UDP/IP (see [ITU-T P.1201.2]), and is assumed to be applicable to RTP/UDP/IP transport with similar, but so far unverified, accuracy as compared to MPEG2-TS/RTP/UDP/IP and MPEG2-TS/UDP/IP). + +## 6.1 Application range for the models + +Table 4 below shows the application range of the models based on what the models have actually been developed for. Note that all cases represent the CC mode of operation, see clause 6.2 for more details about the modes. + +**Table 4 – Factors and application ranges of the ITU-T P.1201.1 and ITU-T P.1201.2 model algorithms** + +| | ITU-T P.1201.1 – Lower resolution (LR) | ITU-T P.1201.2 – Higher resolution (HR) | +|--------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| Application information | Value range, unit | | +| Sequence duration (Ts) | The model has been validated on source sequence lengths of:
10 s: no rebuffering
16 s: rebuffering
No rebuffering: PVS length = SRC length
Rebuffering: PVS length = SRC length + rebuffering length
(no rebuffering at end and start)
It is expected that the model will give reliable prediction results for sequence durations within the range 8-24 seconds | The model has been validated on source sequence lengths of:
10 s: no rebuffering
–
PVS length = SRC length
It is expected that the model will give reliable prediction results for sequence durations of approximately 8-16 seconds | +| Packetization | 3GPP MBMS, PSS or using RTSP directly (all three over RTP/UDP/IP) | MPEG2-TS/RTP/UDP/IP
RTP/UDP/IP (Note 3)
MPEG2-TS/UDP/IP (Note 3) | +| Video codec | MPEG4 visual simple profile (VSP)
ITU-T H.264 baseline profile | ITU-T H.264 main profile, ITU-T H.264 high profile | +| Video size | QCIF, QVGA,
HVGA | SD: PAL, NTSC
HD: 720p, 1080p, 1080i
(High profile: 1080; main profile: 720, SD) | +| Audio codec | AMR-NB, AMR-WB+, AAC-LC,
HE-AAC (v1, v2) | MPEG-4 AAC-LC
MPEG-4 HE-AAC (V1 and V2 = 3GPP enhanced AAC+)
MPEG-1 Layer 2
AC3
MPEG-2 audio | +| Coded video bitrate | [ITU-T H.264]
QCIF: 32-1000 kbit/s
QVGA: 80-3000 kbit/s
HVGA: 192-6000 kbit/s

MPEG4
QCIF: 40-1500 kbit/s
QVGA: 90-3500 kbit/s
HVGA: 192-6000 kbit/s | HD
[ITU-T H.264]:
0.5 up to 30 Mbit/s
SD
[ITU-T H.264]:
0.5 up to 9 Mbit/s | + +**Table 4 – Factors and application ranges of the ITU-T P.1201.1 and ITU-T P.1201.2 model algorithms** + +| | ITU-T P.1201.1 – Lower resolution (LR) | ITU-T P.1201.2 – Higher resolution (HR) | +|-------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| Application information | Value range, unit | | +| Coded audio bitrate | AMR-NB: 4.75-12.2 kbit/s
AMR-WB+: 10.4-48 kbit/s
AAC-LC: 16-128+ kbit/s
HE-AAC (v1, v2): 32-128 kbit/s | AAC-LC:
32-576 kbit/s
HE AACv2:
16-96 kbit/s
MPEG-1 Layer 2:
64-384 kbit/s
AC3:
64-384 kbit/s | +| Video decoder packet loss concealment | Two types of assumed decoder behaviour are covered:
1) freezing with skipping;
2) slicing with:
MPEG4: 1 slices/frame
[ITU-T H.264]: 1 slice/packet
Both MPEG4 and [ITU-T H.264]:
Fixed PLC (using fixed decoder, details and settings) | Types of decoder behaviour: two dimensions: slicing, PLC
1) freezing with skipping
(duration(source)=duration(processed sequence));
2) slicing with 1 slice per frame;
3) slicing with 1 slice per macroblock row, PLC with zero-motion copy (temporal from same region of previous good frame). | +| Audio decoder packet loss concealment | Decoder default modes: Codec-implementation specific loss concealment | (same for both models, see entry on the left) | +| Retransmission mechanisms (ARQ); forward error correction (FEC); client jitter buffer behaviour | Rebuffering handling, particular to LR-case: without skipping of length 0 to 8 seconds
Developed models represent CC-mode (see clause 6.1), hence applied as if dejitter buffer, ARQ and FEC mechanisms have already corrected the stream.
For other modes of operation, see clause 6.1, appropriate methods to correct the streams in ways reflecting the expected FEC, ARQ and dejitter buffer behaviour are for further study. | (No rebuffering)

Developed models represent CC-mode (see clause 6.1), hence applied as if dejitter buffer, ARQ and FEC mechanisms have already corrected the stream.
For other modes of operation, see clause 6.1. Appropriate methods to correct the streams in ways reflecting the expected FEC, ARQ and dejitter buffer behaviour are for further study. | +| Encoder implementation | The model has been developed using the following encoders (Note 1):
• MPEG4 Part 2: ffmpeg
• ITU-T H.264 (MPEG4 Part 10): x264
• AMR-NB/WB+: According to standard
• AAC: Nero | The model has been developed using the following encoders (Note 1):
Video:
• ITU-T H.264 (MPEG-4 Part 10): x264
Audio:
• AAC-LC, HE-AAC v2: Nero
• MPEG1-LII and AC3: ffmpeg | + +**Table 4 – Factors and application ranges of the ITU-T P.1201.1 and ITU-T P.1201.2 model algorithms** + +| | ITU-T P.1201.1 – Lower resolution (LR) | ITU-T P.1201.2 – Higher resolution (HR) | +|------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| Application information | Value range, unit | | +| Decoder implementation | Reference decoder was a proprietary decoder provided by one proponent, which also performed de-packetization and audio-video-demultiplexing. The ITU-T H.264-decoding is standard-conformant, with the PLC as described above (Note 2). | (same for both models, see entry on the left) | +| Group of pictures (GOP) | GOP structure is estimated from the stream.
Typical GOP structure for which the model has been trained:
M = 1, N = 40 (typically no B frames for mobile case)
Length: fixed, variable, adaptive
Structure (e.g., IPPP..PPPI) | Supporting default modes for typical GOP structures
E.g., M = 3, N = 15
Length: fixed, variable, adaptive
Structure (e.g., IBBPBB..PBBI)
NOTE – GOP structure is explicitly estimated from stream. | +| Frame rate | 5, 8.33, 12.5, 15, 20, 25, 30 fps | SD: 50i (PAL), 59.94i (NTSC)
HD: 50p, 59.94p, 60p, 50i, 59.94i, 60i, 25p, 29.97p, 30p | +| Marker bit in RTP header | Model can handle that marker bit means end of frame | For TS, see TS-related information below | +| Audio channel number | 1 (diotic mono), 2 (stereo) | 2 (stereo) | +| # of Audio frames per RTP packet | 1 to 5 audio frames | Bitrate-specific (depending on both the audio and video bitrate) | +| Audio-video multiplexed? | Default: No, at RTP-level; no audio-video asynchrony | In MPEG-2 TS/RTP/UDP and MPEG-2 TS/UDP: Supported | +| Encrypted payload | Only payload level | Transport streams: Both PES- and TS-level encryption, i.e., including and excluding PES-headers, respectively | +| Usage of: payload-unit-start indicator in TS headers | Not applicable | Information on whether payload-unit-start indicator in TS header means the start of frame is provided to the model as side information (Boolean) | +| Usage of: marker bit in RTP header | "End of frame" (True/false) | Not applicable | +| Usage of: random-access indicator in TS header | Not applicable | Information on whether random-access indicator in TS header means the start of I-frame or of IDR-frame is provided to the model as side information (Boolean) | + +**Table 4 – Factors and application ranges of the ITU-T P.1201.1 and ITU-T P.1201.2 model algorithms** + +| | ITU-T P.1201.1 – Lower resolution (LR) | ITU-T P.1201.2 – Higher resolution (HR) | +|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| Application information | Value range, unit | | +| Usage of: elementary-stream-propriety indicator in TS header | Not applicable | Information on whether elementary-stream-propriety indicator in TS header means the start of I-frame or of IDR-frame is provided to the model as side information (Boolean) | +| Packet loss degradation, video | Uniform loss:
0-10%
Burst loss:
0-10% (4-state Markov model) | Uniform loss:
0-2%
Burst loss:
0-2%
(4-state Markov model) | +| Packet loss degradation, audio | Uniform loss:
0-10%
Burst loss:
0-10% (4-state Markov model) | Uniform loss:
0-6%
Burst loss:
0-6%
(4-state Markov model) | +| Symmetrical versus asymmetrical handling of audio and video in audiovisual case | Model application: Symmetrical, but can handle asymmetric cases due to specific model development process | (same for both models, see entry on the left) | +| NOTE 1 – It is assumed that the model can be used for estimating quality when other encoder implementations for the given codec have been used. However, if the encoder performance is significantly worse or better than for the encoder used, the model prediction accuracy will be reduced. | | | +| NOTE 2 – One aspect not covered by decoder packet loss concealment is postfiltering. Guidance on how to adjust internal model parameters for specific other decoders, including set-top boxes, is for further study. | | | +| NOTE 3 – The ITU-T P.1201.2 model has been trained on MPEG-2 TS/RTP/UDP. However, due to the design of the ITU-T P.1201.2 algorithm, it is also applicable to MPEG2-TS/UDP/IP. Further, to the model's design, it is assumed to also work for RTP/UDP/IP transport with similar, but so far unverified, accuracy as compared to MPEG2-TS/RTP/UDP/IP. | | | + +## 6.2 Modes of operation + +The four modes of operation are described in Table 5 and Figures 1-a to 1-e below. Note that the models as described in [ITU-T P.1201.1] and [ITU-T P.1201.2] only support one of the four possible modes (the so-called CC mode). Additional adaptation is required to use the ITU-T P.1201 for the other modes. + +**Table 5 – Modes of operations of ITU-T P.1201** + +| Class | Name | Mode abbreviation* | Description | +|------------------------|------------------------------------|--------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| Mid-point or end-point | Static operation | NN | The model uses information from the local transport layer, prior knowledge about coding and prior knowledge about the end point | +| Mid-point | Non-embedded dynamic operation | BN | The model uses information from the local transport layer, prior knowledge about coding and information about the end point collected through measurement reporting protocols | +| Mid-point | Non-embedded distributed operation | CN | The model, located inside the network, uses information from the transport layer measured at an end point and collected through signalling protocols, prior knowledge about coding and information about the end point collected through signalling protocols | +| End-point | Embedded operation | CC | The model uses information from the local transport layer, information from the end point, and prior knowledge about coding | + +\*Mode abbreviation naming scheme: +XY, where +X corresponds to place of measurement (N: Network, C: Client, B: Both network and client) +Y corresponds to place of model (N: Network, C: Client). + +In Figures 1-a to 1-e below, the following arrow style is used: + +- Media stream +- - - - → Signalling protocol +- . - . → Static information +- . . . → Buffering information + +![Diagram of Static operation mode (NN) inside the network. A 'Send point' box has a solid blue arrow labeled 'Media stream' pointing to an 'End point' box. From the 'End point', a solid blue arrow labeled 'Media signal' points right. A 'Buffering estimation' box is connected to the 'Media stream' line by a solid blue arrow pointing down to it. From the 'Buffering estimation' box, a dashed blue arrow points down to a 'Model' box. A cylinder labeled 'Static media stream and decoder behaviour information' has a dash-dot blue arrow pointing to the 'Model' box. The 'Model' box has a solid blue arrow labeled 'MOS' pointing right.](9b9d2abd741ed4bafe7f78f89961c663_img.jpg) + +Diagram of Static operation mode (NN) inside the network. A 'Send point' box has a solid blue arrow labeled 'Media stream' pointing to an 'End point' box. From the 'End point', a solid blue arrow labeled 'Media signal' points right. A 'Buffering estimation' box is connected to the 'Media stream' line by a solid blue arrow pointing down to it. From the 'Buffering estimation' box, a dashed blue arrow points down to a 'Model' box. A cylinder labeled 'Static media stream and decoder behaviour information' has a dash-dot blue arrow pointing to the 'Model' box. The 'Model' box has a solid blue arrow labeled 'MOS' pointing right. + +**Figure 1-a – Static operation mode (NN) inside the network** + +![Figure 1-b: Static operation mode (NN) inside a terminal. A 'Send point' sends a 'Media stream' to an 'End point'. Inside the 'End point', the stream goes to a 'Media buffer' and a 'Buffering estimation' block. The 'Buffering estimation' block feeds into a 'Model' block, which outputs 'MOS'. The 'Media buffer' feeds into a 'Decoder and PLC' block, which outputs 'Media signal'. A cylinder labeled 'Static media stream and decoder behaviour information' also feeds into the 'Model' block.](9c6461e1e94afae4dec455e69a2ce152_img.jpg) + +``` + +graph LR + SendPoint[Send point] -- "Media stream" --> EndPoint + subgraph EndPoint [End point] + direction TB + Cyl[Static media stream and decoder behaviour information] --> Model + Buffering[Buffering estimation] --> Model + MediaBuffer[Media buffer] --> Decoder[Decoder and PLC] + end + SendPoint --> Buffering + SendPoint --> MediaBuffer + Model --> MOS[MOS] + Decoder --> MediaSignal[Media signal] + +``` + +Figure 1-b: Static operation mode (NN) inside a terminal. A 'Send point' sends a 'Media stream' to an 'End point'. Inside the 'End point', the stream goes to a 'Media buffer' and a 'Buffering estimation' block. The 'Buffering estimation' block feeds into a 'Model' block, which outputs 'MOS'. The 'Media buffer' feeds into a 'Decoder and PLC' block, which outputs 'Media signal'. A cylinder labeled 'Static media stream and decoder behaviour information' also feeds into the 'Model' block. + +**Figure 1-b – Static operation mode (NN) inside a terminal** + +![Figure 1-c: Non-embedded dynamic operation mode (BN). A 'Send point' sends a 'Media stream' to an 'End point', which outputs 'Media signal'. A branch of the 'Media stream' goes to a 'Model' block, which outputs 'MOS'. A cylinder labeled 'Static media stream and decoder behaviour information' also feeds into the 'Model' block. A dashed line connects the 'End point' to the 'Model' block.](2b3a967f6ce4f23649be995a353e39f8_img.jpg) + +``` + +graph LR + SendPoint[Send point] -- "Media stream" --> EndPoint[End point] + EndPoint --> MediaSignal[Media signal] + SendPoint -- "Media stream" --> Model + Cyl[Static media stream and decoder behaviour information] -.-> Model + Model --> MOS[MOS] + EndPoint -.-> Model + +``` + +Figure 1-c: Non-embedded dynamic operation mode (BN). A 'Send point' sends a 'Media stream' to an 'End point', which outputs 'Media signal'. A branch of the 'Media stream' goes to a 'Model' block, which outputs 'MOS'. A cylinder labeled 'Static media stream and decoder behaviour information' also feeds into the 'Model' block. A dashed line connects the 'End point' to the 'Model' block. + +**Figure 1-c – Non-embedded dynamic operation mode (BN)** + +![Figure 1-d: Non-embedded distributed operation mode (CN). A 'Send point' sends a 'Media stream' to an 'End point', which outputs 'Media signal'. A branch of the 'Media stream' goes to a 'Model' block, which outputs 'MOS'. A cylinder labeled 'Static media stream and decoder behaviour information' also feeds into the 'Model' block. A dashed line connects the 'End point' to the 'Model' block.](10781f43062bf3e9601a1e086710556c_img.jpg) + +``` + +graph LR + SendPoint[Send point] -- "Media stream" --> EndPoint[End point] + EndPoint --> MediaSignal[Media signal] + Cyl[Static media stream and decoder behaviour information] -.-> Model + Model --> MOS[MOS] + EndPoint -.-> Model + +``` + +Figure 1-d: Non-embedded distributed operation mode (CN). A 'Send point' sends a 'Media stream' to an 'End point', which outputs 'Media signal'. A branch of the 'Media stream' goes to a 'Model' block, which outputs 'MOS'. A cylinder labeled 'Static media stream and decoder behaviour information' also feeds into the 'Model' block. A dashed line connects the 'End point' to the 'Model' block. + +**Figure 1-d – Non-embedded distributed operation mode (CN)** + +![Figure 1-e – Embedded operation mode (CC)](b6671cfafda3820aafe9a24fa7a4d8c7_img.jpg) + +This diagram illustrates the embedded operation mode (CC). A 'Send point' block on the left sends a 'Media stream' to a 'Media buffer' block. The 'Media buffer' is part of an 'End point' (indicated by a dashed box). Inside the 'End point', the 'Media buffer' sends data to a 'Decoder and PLC' block, which outputs a 'Media signal'. The 'Media buffer' also sends data to a 'Model' block. The 'Model' block receives 'Static media stream and decoder behaviour information' from a cylinder-shaped input and outputs a 'MOS' score. There is also a dashed arrow from the 'Model' block back to the 'Media buffer'. + +Figure 1-e – Embedded operation mode (CC) + +**Figure 1-e – Embedded operation mode (CC)** + +## 7 Model input interfaces + +The ITU-T P.1201 models will receive dynamic transport layer information and static side information. For the models as they are described in [ITU-T P.1201.1] and [ITU-T P.1201.2], the dynamic transport layer input information is expected to be provided in PCAP file format with transport header information. However, in practical implementations, realizations of the transport layer information extraction other than PCAP-based realizations can be envisaged. For the models as they are described here, the PCAP file could be created based on packets being captured at a network interface. The static side information is information about the media stream and the decoder behaviour. The overview information per application area and mode is described in Table 6 and Figure 2 below: + +![Figure 2 – Overview of ITU-T P.1201 model interfaces](523ab7b925beb555f88b2e1e1336974f_img.jpg) + +This diagram shows the overview of ITU-T P.1201 model interfaces. A dashed box labeled 'ITU-T P.1201 model' contains three main components: 'Parameter extraction', 'MOS estimation (model core)', and 'Buffering information'. 'Packet header information (dynamic)' enters from the left through interface 'I.2' into the 'Parameter extraction' block. The 'Parameter extraction' block outputs through interface 'I.3' to the 'MOS estimation (model core)' block. The 'MOS estimation (model core)' block outputs a 'MOS' score to the right. Below the model box, 'Static side information' (cylinder) enters through interface 'I.1' into the 'MOS estimation (model core)' block. 'Buffering information' (rectangle) enters through interface 'I.4' into the 'MOS estimation (model core)' block. + +Figure 2 – Overview of ITU-T P.1201 model interfaces + +**Figure 2 – Overview of ITU-T P.1201 model interfaces** + +The I.3 interface in the CN mode conveys the signalling information from the parameter extraction module located in the end point. + +The ITU-T P.1201 model has three main inputs: + +- **Packet-header information:** This input can be taken from a PCAP file, or streamed from a network interface or from a parameter extraction module located in the end point in a CN mode implementation. +- **Buffering information:** This input can be taken from the media buffer in the client, or estimated by a buffering estimation module using packet information. +- **Static media- and decoder information:** This input is obtained from packet information or from a player API. + +**Table 6 – Overview of input to the ITU-T P.1201 model for the modes of operation** + +| | Static operation (NN) | Non-embedded dynamic operation (BN) | Non-embedded distributed operation (CN) | Embedded operation (CC) | +|-------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------| +| Interface 1 (I.1) | Static information about the media stream (codec, usage of flags, etc.), and static information about decoder. Detailed description in Table 7. | Static information about the media stream (codec, usage of flags, etc.), and static information about decoder. Detailed description in Table 7.
NOTE – Static decoder behaviour information might be provided via signalling and included in interface I.3 (see Figure 2) | Static information about the media stream (codec, usage of flags, etc.), and static information about decoder. Detailed description in Table 7.
NOTE – Static decoder behaviour information might be provided via signalling and included in interface I.3 (see Figure 2) | Static information about the media stream (codec, usage of flags, etc.), and static information about decoder. Detailed description in Table 7. | +| Interface 2 (I.2) | PCAP file (payload not required) | PCAP file (payload not required) | Not available in this mode | PCAP file (payload not required) | +| Interface 3 (I.3) | Parameters extracted from the PCAP file | Output from a parameter extraction module for transport header information located in the end point and transferred to the model with a protocol. | Output from a parameter extraction module for transport header information located in the end point and transferred to the model with a protocol. | Parameters extracted from the PCAP file | +| Interface 4 (I.4)
Only available for the ITU-T P.1201 LR model | Rebuffering information (estimated) | Rebuffering parameters measured/extracted in the end point and transferred to the model with a protocol. | Rebuffering parameters measured/extracted in the end point and transferred to the model with a protocol. | Rebuffering information from the media buffer | + +The following Tables 7 and 8 give more detailed examples of information that is provided at the input interfaces I.1 (see Figure 2 above). + +**Table 7 – Input to the ITU-T P.1201.1 model algorithm (LR)** + +| Input | Typical values | +|-------------------------------|------------------------------------------------------| +| Dynamic input | | +| Media stream | PCAP file or other capture format | +| Rebuffering information | Text file containing rebuffering information | +| Static information | | +| Audio destination port | 1234 | +| Video destination port | 5678 | +| Video codec | ITU-T H.264, MPEG4 | +| Video codec profile | Simple, baseline | +| Video resolution | QCIF, QVGA, HVGA | +| Video scanning type | Progressive | +| Video frame rate | 5, 8.33, 10, 12.5, 15, 25, 30 | +| Video packet loss concealment | Slicing, freezing | +| Audio codec | AMRNB; AMRWB+; MPEG1L2;
AAC-LC; HE-AAC v1,v2; AC3 | +| Audio channel number | 1, 2 | +| Audio bandwidth | NB, WB, SWB, FB | +| Audio frame length | 20 to 42.67 | +| RTP marker bit usage | "Last packet of frame" (TRUE/FALSE) | + +**Table 8 – Input to the ITU-T P.1201.2 model algorithm (HR)** + +| Input | Typical values | +|-------------------------------|-----------------------------------| +| Dynamic input | | +| Media stream | PCAP file or other capture format | +| Static information | | +| Audio destination port | 1234 | +| Video destination port | 5678 | +| Video codec | ITU-T H.264 | +| Video codec profile | Main, High | +| Video resolution | PAL, NTSC, HD720, HD1080 | +| Video scanning type | Progressive, Interlaced | +| Video frame rate | 24, 25, 29.97, 30, 50, 60 | +| Video packet loss concealment | Slicing, freezing | +| Number of slices per frame | 1, 30, 34, 68 | +| Audio codec | MPEG1L2, AAC-LC, HE-AAC v2, AC3 | + +**Table 8 – Input to the ITU-T P.1201.2 model algorithm (HR)** + +| Input | Typical values | +|------------------------------------------|-------------------------------------------------| +| Audio channel number | 2 | +| Audio bandwidth | FB | +| Audio frame length | 21.33 to 42.67 ms | +| Transport stream random access indicator | "First TS-packet of I-frame", "Not applicable" | +| Elementary stream priority | "First TS-packet of I-frame", "Not applicable " | +| Payload unit start indicator | "First TS-packet of a frame", "Not applicable" | + +The models described in [ITU-T P.1201.1] and [ITU-T P.1201.2] have been validated assuming that the available information already reflects the impact of any error resilience methods, such as forward error correction (FEC), or packet re-transmission mechanisms such as automatic repeat request (ARQ), and of the impact due to the dejitter buffer. This is equivalent to the parameter extraction module being located behind these processing steps, that is, implementing the CC mode of operation. In case of the NN mode, the measurement point is located prior to the actual FEC, ARQ and dejitter buffer mechanisms. Since these mechanisms may have a very strong impact on factors such as the packet loss seen by the decoder, the NN-mode requires an explicit handling of these mechanisms, to reflect a CN- or BN-type of behaviour. Two principal approaches are conceivable to capture this case, which are both based on prior knowledge: + +- a) The packet stream S is converted into a stream that reflects an assumed behaviour of FEC, ARQ, and/or jitter de-buffering, reflecting the input format to be provided to the parameter extraction module (PCAP). This step results in a converted stream S', see Figure 3-c. +- b) The input parameters to the model are converted into values that reflect the application of FEC, ARQ and/or dejitter buffering (module "loss-related parameter mapping" in Figure 3-b). + +Figure 3 shows how the case of error resilience methods such as forward error correction (FEC) and automatic repeat request (ARQ) could be handled in case of the NN/CN/BN modes. Figure 3-a shows the model architecture when FEC/ARQ is not used. In Figure 3-b the extracted internal model parameters are mapped to reflect the properties of a corrected stream. Note that the "Loss-related parameter mapping" module is considered transparent for all non loss-related parameters. In Figure 3-c the media stream is first corrected using a FEC/ARQ dejitter buffer and then the parameters are extracted in exactly the same way as in the case without FEC/ARQ. + +![Figure 3-a: Model structure when error resilience methods (FEC/ARQ) are not used. The diagram shows a flow from 'Media stream without FEC/ARQ' to 'Parameter extraction', which outputs 'Internal parameters' to 'Model core', which in turn outputs 'MOS'. A database labeled 'Static media stream and decoder behaviour information' is connected to the 'Model core' via a dashed arrow.](c5452f95f3b28f1bfe29e84fbc2e1267_img.jpg) + +``` + +graph LR + A[Media stream without FEC/ARQ] --> B[Parameter extraction] + B -- Internal parameters --> C[Model core] + C --> D[MOS] + E[(Static media stream and decoder behaviour information)] -.-> C + +``` + +Figure 3-a: Model structure when error resilience methods (FEC/ARQ) are not used. The diagram shows a flow from 'Media stream without FEC/ARQ' to 'Parameter extraction', which outputs 'Internal parameters' to 'Model core', which in turn outputs 'MOS'. A database labeled 'Static media stream and decoder behaviour information' is connected to the 'Model core' via a dashed arrow. + +**Figure 3-a – Model structure when error resilience methods (FEC/ARQ) are not used** + +![Figure 3-b: Model structure when FEC/ARQ are in use and extracted parameters are mapped. The diagram shows a flow from 'Media stream using FEC/ARQ' to 'Parameter extraction', which outputs 'Internal parameters' to 'Loss-related parameter mapping'. This block outputs 'Mapped/adjusted internal parameters' to 'Model core', which outputs 'MOS'. Two databases are connected to the 'Model core' via dashed arrows: 'FEC/ARQ/de-jitter buffer information' and 'Static media stream and decoder behaviour information'.](366a77fdefb0097b3289b4a011911390_img.jpg) + +``` + +graph LR + A[Media stream using FEC/ARQ] --> B[Parameter extraction] + B -- Internal parameters --> C[Loss-related parameter mapping] + C -- Mapped/adjusted internal parameters --> D[Model core] + D --> E[MOS] + F[(FEC/ARQ/de-jitter buffer information)] -.-> D + G[(Static media stream and decoder behaviour information)] -.-> D + +``` + +Figure 3-b: Model structure when FEC/ARQ are in use and extracted parameters are mapped. The diagram shows a flow from 'Media stream using FEC/ARQ' to 'Parameter extraction', which outputs 'Internal parameters' to 'Loss-related parameter mapping'. This block outputs 'Mapped/adjusted internal parameters' to 'Model core', which outputs 'MOS'. Two databases are connected to the 'Model core' via dashed arrows: 'FEC/ARQ/de-jitter buffer information' and 'Static media stream and decoder behaviour information'. + +**Figure 3-b – Model structure when FEC/ARQ are in use and extracted parameters are mapped** + +![Figure 3-c: Model structure when FEC/ARQ are used and the stream is corrected before parameter extraction. The flowchart shows: Media stream using FEC/ARQ -> FEC/ARQ de-jitter buffer -> Corrected media stream -> Parameter extraction -> Internal parameters -> Model core -> MOS. A cylinder labeled 'Static media stream and decoder behaviour information' points to the Model core.](7133ccf78043568ca62ecbcd43628a4a_img.jpg) + +``` + +graph LR + A[Media stream using FEC/ARQ] --> B[FEC/ARQ de-jitter buffer] + B --> C[Corrected media stream] + C --> D[Parameter extraction] + D --> E[Internal parameters] + E --> F[Model core] + F --> G[MOS] + H[(Static media stream and decoder behaviour information)] -.-> F + +``` + +Figure 3-c: Model structure when FEC/ARQ are used and the stream is corrected before parameter extraction. The flowchart shows: Media stream using FEC/ARQ -> FEC/ARQ de-jitter buffer -> Corrected media stream -> Parameter extraction -> Internal parameters -> Model core -> MOS. A cylinder labeled 'Static media stream and decoder behaviour information' points to the Model core. + +**Figure 3-c – Model structure when FEC/ARQ are used and the stream is corrected before parameter extraction** + +## 8 Model output information and performance details + +The ITU-T P.1201.1 and ITU-T P.1201.2 models have three output parameters: + +- 1) Estimated audiovisual MOS on the 1 to 5 scale, which is an estimation of the perceived audiovisual quality. +- 2) Estimated video MOS on the 1 to 5 scale, which is an estimation of the perceived video quality (without audio present). Note that the model is able to give a video score for a stream without audio, and a stream including audio. +- 3) Estimated audio MOS on the 1 to 5 scale, which is an estimation of the perceived audio quality (without video present). Note that the model is able to give an audio score for a stream without video, and a stream including video. + +The performance information for the ITU-T P.1201 models can be found in Table 9 and in Appendix I. The statistical metrics root mean square error (RMSE) and Pearson correlation are used to describe the performance, see [ITU-T P.1401]. Note that for these performance figures, the subjective ratings have been mapped to the model scores using a linear, i.e., 1st-order mapping function, at a per-database-level. This has been done in order to avoid misalignment due to bias in the different subjective tests, e.g., as a result of different test settings. + +**Table 9 – Performance information for ITU-T P.1201.1 (LR)** + +| | RMSE | Pearson correlation | +|-------------|-------------------------------|-------------------------------| +| Audiovisual | 0.470 (based on 1166 samples) | 0.852 (based on 1166 samples) | +| Video | 0.535 (based on 1430 samples) | 0.830 (based on 1430 samples) | +| Audio | 0.351 (based on 690 samples) | 0.941 (based on 690 samples) | + +**Table 10 – Performance information for ITU-T P.1201.2 (HR)** + +| | RMSE | Pearson correlation | +|-------------|-----------------------------------------------------|-----------------------------------------------------| +| Audiovisual | 0.435 (based on 3190 samples, PES and TS encrypted) | 0.911 (based on 3190 samples, PES and TS encrypted) | +| Video | 0.461 (based on 6138 samples, PES and TS encrypted) | 0.902 (based on 6138 samples, PES and TS encrypted) | +| Audio | 0.336 (based on 1360 samples, PES and TS encrypted) | 0.949 (based on 1360 samples, PES and TS encrypted) | + +# **9 Description of the ITU-T P.1201 model algorithms** + +The ITU-T P.1201 lower resolution model is described in [ITU-T P.1201.1], and the description of the ITU-T P.1201 higher resolution model algorithm can be found in [ITU-T P.1201.2]. + +## Appendix I + +### Detailed performance figures for the ITU-T P.1201.1 and ITU-T P.1201.2 algorithms + +(This appendix does not form an integral part of this Recommendation.) + +**Table I.1 – Detailed performance figures for the ITU-T P.1201.1 algorithm** + +| Media | Video codec | Audio codec | RMSE | Pearson correlation | # files | +|-------------|--------------|-------------|-------|---------------------|---------| +| All | Overall | | 0.478 | 0.867 | 3286 | +| Audio | Overall | | 0.351 | 0.941 | 690 | +| | – | AAC-LC | 0.300 | 0.955 | 207 | +| | – | AAC-HEv1 | 0.354 | 0.960 | 24 | +| | – | AAC-HEv2 | 0.429 | 0.808 | 16 | +| | – | AMRNB | 0.263 | 0.755 | 161 | +| | – | AMRWBP | 0.419 | 0.853 | 282 | +| Video | Overall | | 0.535 | 0.830 | 1430 | +| | H264 (QCIF) | – | 0.531 | 0.848 | 207 | +| | MPEG4 (QCIF) | – | 0.461 | 0.829 | 184 | +| | H264 (QVGA) | – | 0.529 | 0.851 | 375 | +| | MPEG4 (QVGA) | – | 0.523 | 0.708 | 264 | +| | H264 (HVGA) | – | 0.584 | 0.832 | 400 | +| Audiovisual | Overall | | 0.470 | 0.852 | 1166 | +| | H264 (QCIF) | AAC-LC | 0.360 | 0.951 | 87 | +| | H264 (QCIF) | AMRNB | 0.480 | 0.701 | 64 | +| | H264 (QCIF) | AMRWBP | 0.499 | 0.882 | 80 | +| | MPEG4 (QCIF) | AAC-LC | 0.420 | 0.754 | 48 | +| | MPEG4 (QCIF) | AMRNB | 0.379 | 0.917 | 48 | +| | MPEG4 (QCIF) | AMRWBP | 0.474 | 0.823 | 56 | +| | H264 (QVGA) | AAC-LC | 0.463 | 0.888 | 153 | +| | H264 (QVGA) | AMRNB | 0.443 | 0.811 | 71 | +| | H264 (QVGA) | AMRWBP | 0.639 | 0.729 | 100 | +| | MPEG4 (QVGA) | AAC-LC | 0.645 | 0.724 | 42 | +| | MPEG4 (QVGA) | AMRNB | 0.495 | 0.706 | 73 | +| | MPEG4 (QVGA) | AMRWBP | 0.395 | 0.749 | 104 | +| | H264 (HVGA) | AAC-LC | 0.422 | 0.927 | 80 | + +**Table I.1 – Detailed performance figures for the ITU-T P.1201.1 algorithm** + +| Media | Video codec | Audio codec | RMSE | Pearson correlation | # files | +|-------|-------------|-------------|-------|---------------------|---------| +| | H264 (HVGA) | AMRNB | 0.465 | 0.784 | 72 | +| | H264 (HVGA) | AMRWBP | 0.455 | 0.868 | 88 | + +**Table I.2a – Detailed performance figures for the ITU-T P.1201.2 algorithm** + +| Media | Video codec | Audio codec | RMSE | Pearson correlation | # files | +|-------------|---------------|-------------|-------|---------------------|---------| +| All | Overall | | 0.439 | 0.911 | 10688 | +| Audio | Overall | | 0.336 | 0.949 | 1360 | +| | – | AAC-HEv2 | 0.463 | 0.861 | 188 | +| | – | AAC-LC | 0.328 | 0.950 | 514 | +| | – | MPEG1L2 | 0.317 | 0.941 | 450 | +| | – | AC3 | 0.258 | 0.975 | 208 | +| Video | Overall | | 0.461 | 0.902 | 6138 | +| | H264 (SD) | – | 0.465 | 0.898 | 1396 | +| | H264 (HD720) | – | 0.465 | 0.897 | 1438 | +| | H264 (HD1080) | – | 0.458 | 0.905 | 3304 | +| Audiovisual | Overall | | 0.435 | 0.911 | 3190 | +| | H264 (SD) | AAC-LC | 0.408 | 0.921 | 502 | +| | H264 (SD) | AC3 | 0.519 | 0.857 | 316 | +| | H264 (HD720) | AAC-HEv2 | 0.318 | 0.943 | 48 | +| | H264 (HD720) | AAC-LC | 0.427 | 0.932 | 256 | +| | H264 (HD720) | MPEG1L2 | 0.454 | 0.881 | 482 | +| | H264 (HD720) | AC3 | 0.411 | 0.884 | 160 | +| | H264 (HD1080) | AAC-LC | 0.444 | 0.920 | 744 | +| | H264 (HD1080) | MPEG1L2 | 0.406 | 0.920 | 682 | + +**Table I.2b – Detailed performance figures for the ITU-T P.1201.2 algorithm** + +| Media | Encryption level | RMSE | Pearson correlation | # files | +|-------------|------------------|-------|---------------------|---------| +| Audio | PES | 0.337 | 0.949 | 680 | +| | M2TS | 0.336 | 0.949 | 680 | +| Video | PES | 0.459 | 0.903 | 3069 | +| | M2TS | 0.463 | 0.901 | 3069 | +| Audiovisual | PES | 0.433 | 0.913 | 1595 | +| | M2TS | 0.438 | 0.91 | 1595 | + +## Appendix II + +### Performance figures for the ITU-T P.1201.2 algorithm in the case of MPEG2-TS/UDP + +(This appendix does not form an integral part of this Recommendation.) + +In order to validate the ITU-T P.1201.2 algorithm in the case of MPEG2-TS/UDP, two sets of PCAP files were used: + +- A) MPEG2-TS/RTP/UDP case: The PES-encrypted ITU-T P.1201.1 and ITU-T P.1201.2 training databases for higher resolutions (HRs) with MPEG2-TS/RTP/UDP as packetization scheme. +- B) MPEG2-TS/UDP case: The same PCAP files as A). However, this time, a MPEG2-TS/UDP packetization scheme was simulated by: + - not reading the RTP-header (to ensure a true simulation none of the parameters of the RTP-header were actually read at all); + - setting a CTS-flag, which implies a replacement of the RTP-timestamp by the video-DTS, as described in clause 4.2.3 of [ITU-T P.1201.2]; + - not using the RTP-sequence number for the calculation of the number of lost packets, as described in clause 4.2.2 of [ITU-T P.1201.2]. Instead, a different (new) function was used that utilized the CC of the TS-headers in order to estimate the number of lost packets. + +Both sets A) and B), that were used for the validation, contain the following number of databases and PCAP files: + +- One audio database: 240 files +- Five video-only databases: 1200 files +- Three audiovisual databases: 714 files + +As performance indicators, the Pearson correlation coefficient (PCC) and the root-mean-square-error (RMSE) were computed between the MOS score predictions on sets A) and B). Performance results are reported in Table II.1. The percentages of PCAP files with a MOS estimation difference smaller than 0.05 between the MPEG2-TS/RTP/UDP and MPEG2-TS/UDP PCAP files are also provided. + +**Table II.1 – Performance results for the ITU-T P.1201.2 algorithm in the case of MPEG2-TS/UDP** + +| Media | PCC | RMSE | % files with MOS diff < 0.05 | +|---------------------------------|------|------|------------------------------| +| Audio for audio database | 0.99 | 0.01 | 98.75% | +| Audio for audiovisual databases | 0.99 | 0.01 | 99.86% | +| Video for video databases | 0.99 | 0.06 | 90.42% | +| Video for audiovisual databases | 0.99 | 0.05 | 93.56% | +| Audiovideo | 0.99 | 0.03 | 95.10 | + + + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|---------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Terminals and subjective and objective assessment methods | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects and next-generation networks | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.1203.1-201901-I_PDF-E/raw.md b/marked/P/T-REC-P.1203.1-201901-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..2da6129ac452235a9420a22e577496b3bd8a921c --- /dev/null +++ b/marked/P/T-REC-P.1203.1-201901-I_PDF-E/raw.md @@ -0,0 +1,952 @@ + + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.1203.1** + +(01/2019) + +SERIES P: TELEPHONE TRANSMISSION QUALITY, +TELEPHONE INSTALLATIONS, LOCAL LINE +NETWORKS + +Models and tools for quality assessment of streamed +media + +--- + +**Parametric bitstream-based quality assessment +of progressive download and adaptive +audiovisual streaming services over reliable +transport – Video quality estimation module** + +Recommendation ITU-T P.1203.1 + +![ITU logo](84a1d09fb489061482111515543b60dc_img.jpg) + +The logo of the International Telecommunication Union (ITU) features a stylized globe with a red lightning bolt striking through it, symbolizing global communication and technology. + +ITU logo + +International +Telecommunication +Union + +# ITU-T P-SERIES RECOMMENDATIONS + +## TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS + +| | | | +|----------------------------------------------------------------------------------------------------|---------------|---------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Voice terminal characteristics | Series | P.30 | +| | | P.300 | +| Reference systems | Series | P.40 | +| Objective measuring apparatus | Series | P.50 | +| | | P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of speech quality | Series | P.80 | +| Methods for objective and subjective assessment of speech and video quality | Series | P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | +| Transmission performance and QoS aspects of IP end-points | Series | P.1000 | +| Communications involving vehicles | Series | P.1100 | +| Models and tools for quality assessment of streamed media | Series | P.1200 | +| Telemeeting assessment | Series | P.1300 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | Series | P.1400 | +| Methods for objective and subjective assessment of quality of services other than speech and video | Series | P.1500 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +# Recommendation ITU-T P.1203.1 + +## Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Video quality estimation module + +## Summary + +Recommendation ITU-T P.1203.1 specifies the short-term video representation quality estimation modules for ITU-T P.1203 (Pv module). The ITU-T P.1203-series of Recommendations specifies modules for a set of model algorithms for monitoring the integral media session quality for transport control protocol (TCP) type video streaming. The models comprise modules for short-term video-quality (described in this part of the Recommendation family) and audio-quality estimation. The per-one-second outputs of these short-term modules are integrated into estimates of audio-visual quality and together with information about initial loading delay and media playout stalling events, they are further integrated into the final model output, to provide an estimate of integral quality. The respective ITU-T work item has formerly been referred to as "Parametric non-intrusive assessment of TCP-based multimedia streaming quality", or "P.NATS". The ITU-T P.1203.1 part of ITU-T P.1203 provides details for the modules for bitstream-based, short-term video quality estimation. + +Four different modes can be used for the Pv module specified in this Recommendation. These modes – referred to as mode 0 to 3 – use input information of differing complexity and amount and represent four model algorithms each with a different level of complexity. The Pv modules comprise components reflecting the effects due to video compression, up-scaling of content and the effect due to low frame rates. The four different modes use the same overall model architecture and individual coefficients and all have the same components for up-scaling and framerate. The only Pv module component that differs between modes is the Pv module component related to video compression. + +This revision to Recommendation ITU-T P.1203.1 (2017) replaces the scale factor defined in clause 8. + +## History + +| Edition | Recommendation | Approval | Study Group | Unique ID* | +|---------|----------------|------------|-------------|---------------------------------------------------------------------------| +| 1.0 | ITU-T P.1203.1 | 2016-12-22 | 12 | 11.1002/1000/13159 | +| 2.0 | ITU-T P.1203.1 | 2017-10-29 | 12 | 11.1002/1000/13400 | +| 3.0 | ITU-T P.1203.1 | 2019-01-13 | 12 | 11.1002/1000/13845 | + +## Keywords + +Adaptive streaming, audio, audiovisual, IPTV, mean opinion score (MOS), mobile TV, mobile video, monitoring, multimedia, progressive download, QoE, TV, video. + +--- + +\* To access the Recommendation, type the URL in the address field of your web browser, followed by the Recommendation's unique ID. For example, . + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database at . + +© ITU 2019 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## Table of Contents + +###### Page + +| | | | +|-----|-------------------------------------------------------------------|----| +| 1 | Scope..... | 1 | +| 2 | References..... | 3 | +| 3 | Definitions ..... | 4 | +| 3.1 | Terms defined elsewhere ..... | 4 | +| 3.2 | Terms defined in this Recommendation..... | 4 | +| 4 | Abbreviations and acronyms ..... | 4 | +| 5 | Conventions ..... | 4 | +| 6 | Pv module in ITU-T P.1203 context..... | 4 | +| 6.1 | Pv module modes ..... | 5 | +| 7 | Model input..... | 5 | +| 7.1 | I.13 input specification ..... | 6 | +| 8 | Model algorithm and output ..... | 6 | +| 8.1 | Core model ..... | 7 | +| | Annex A – Mode 0 Pv module: quant description..... | 10 | +| A.1 | Performance analysis..... | 11 | +| | Annex B – Mode 1 Pv module: Dq description ..... | 12 | +| B.1 | quant calculation ..... | 12 | +| B.2 | Dq calculation for mode 1 ..... | 13 | +| B.3 | Performance analysis..... | 14 | +| | Annex C – Mode 2 Pv module: quant description..... | 15 | +| C.1 | Consideration on parameter extraction..... | 15 | +| C.2 | quant calculation ..... | 17 | +| C.3 | Performance analysis..... | 17 | +| | Annex D – Mode 3 Pv module: quant description..... | 18 | +| D.1 | Performance analysis..... | 19 | +| | Annex E – MOSfromR and RfromMOS and functions ..... | 20 | +| E.1 | MOSfromR definition ..... | 20 | +| E.2 | RfromMOS definition ..... | 20 | +| | Annex F – Byte-counting algorithm for Mode 2 ..... | 22 | + + + +## **Recommendation ITU-T P.1203.1** + +## **Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Video quality estimation module** + +# **1 Scope** + +This Recommendation describes the video module as an integral part of the ITU-T P.1203-series Recommendations. The ITU-T P.1203 Recommendations describe a set of objective parametric quality assessment modules. Combined, these modules can be used to predict the impact of audio and video media encodings as well as Internet protocol (IP) network impairments on the quality experienced by an end-user of multi-media streaming applications. + +The addressed streaming techniques comprise progressive download as well as adaptive streaming, for both mobile and fixed network streaming applications over transport control protocol (TCP) or other TCP like protocols which are not affected by transmission errors. + +The model described is restricted to information provided to it by an appropriate packet- or bitstream-analysis module. The overall ITU-T P.1203 model is applicable for the effects due to audio- and video-coding as well as initial loading delay or stalling (which are both caused by rebuffering at the client) as the typical degradations associated with progressive download. As final output, the ITU-T P.1203-series models target integral audiovisual media quality scores. + +This Recommendation describes four different quality modules, one for each mode of ITU-T P.1203, that is, modes 0, 1, 2 and 3. The video quality module predicts mean opinion scores (MOS) on a 5-point ACR scale (see [ITU-T P.910]) as a per-one-second MOS score. The underlying measurement window is described in [ITU-T P.1203]. If used stand-alone, the video module can provide estimates of short-term video quality at per-one-second intervals. + +This model cannot provide a comprehensive evaluation of video transmission quality as perceived by an individual end user because its scores reflect the impairments due to video scaling, framerate and coding only. Furthermore, the scores predicted by a parametric model necessarily reflect an average perceptual impairment. + +Note also that the model was developed and validated for one specific encoder and decoder implementation. If a different encoder and decoder pair is used in a monitoring situation the scores may not reflect that. + +Effects such as flicker due to low source bitrate or other, not coding or transmission caused impairments related to the payload are not reflected in the scores computed by this model. Moreover, the scores predicted by a parametric model (especially in case of no access to payload or pixel information) necessarily reflect a somewhat simplified representation of the perceptual impairment of the considered stream. + +However, the model still enables estimation of some coding quality related information and thus valid and in most cases accurate predictions, presuming that it is applied in an appropriate manner, following this Recommendation. + +Table 1-1 shows application areas, test factors and coding technologies where ITU-T P.1203.1 for adaptive streaming and progressive download has been verified and is known to produce reliable results. + +**Table 1-1 – Application areas, test factors and coding technologies where ITU-T P.1203.1 for adaptive streaming and progressive download has been verified and is known to produce reliable results** + +| Applications for which the model is intended | | +|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| In-service monitoring of TCP- video. Both so called over-the-top (OTT) services (for example YouTube) and operator managed video services (over TCP), using the protocols HTTP/TCP/IP and RTMP/TCP/IP. Note that this model is agnostic to the type of container format (Flash (FLV), MP4, WebM and 3GP). | | +| Performance and quality assessment of live networks (including codecs) considering the effect due to encoding bitrate, encoding resolution, and encoding framerate. | | +| Video test factors for which the model has been validated | | +| Video content | Movie trailers, sports videos, documentaries, freely available HD content, time lapse videos, etc. | +| Input video length | Maximum 20 seconds. The video model produces a per-second score considering input data from a measurement window of max. 20 s length. | +| Bitstream container | Elementary stream contained in MPEG-2 transport stream (TS) segments | +| Encoder/decoder implementation |

The model has been trained using the following video encoder/decoder:

  • – ITU-T H.264/MPEG-4 AVC High profile: x264 (ffmpeg)
  • – A common framework was developed based on the above codec, all the test data was generated using the common framework.

It is assumed that the model can be used for estimating quality when other encoder implementations for the given codec have been used. However, model performance cannot be guaranteed in this case.

| +| Slice size | 1 slice per video frame | +| Scene-cut detection | Off | +| x264 Preset | Medium | +| Video resolution/bitrate |

240p: 75-150 kbit/s

360p: 220-450 kbit/s

480p: 375-750 kbit/s

720p: 1 050-2 100 kbit/s

1080p: 1 875-12 500 kbit/s

Note that aspect ratio of 16:9 is maintained for all quality levels.

| +| Group of pictures (GOP) | 1 second length, IBBBP only | +| Segment length |

1-9 seconds

NOTE – The segment length determines how often the quality can be adapted.

| + +# 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- [ITU-T G.107] Recommendation ITU-T G.107 (2015), *The E-model: a computational model for use in transmission planning*. +- [ITU-T H.264] Recommendation ITU-T H.264 (2016), *Advanced video coding for generic audiovisual services*. +- [ITU-T P.800.1] Recommendation ITU-T P.800.1 (2006), *Mean opinion score (MOS) terminology*. +- [ITU-T P.910] Recommendation ITU-T P.910 (2008), *Subjective video quality assessment methods for multimedia applications*. +- [ITU-T P.911] Recommendation ITU-T P.911 (1998), *Subjective audiovisual quality assessment methods for multimedia applications*. +- [ITU-T P.1201.1] Recommendation ITU-T P.1201.1 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – Lower resolution application area*. +- [ITU-T P.1201.2] Recommendation ITU-T P.1201.2 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – Higher resolution application area*. +- [ITU-T P.1202] Recommendation ITU-T P.1202 (2012), *Parametric non-intrusive bitstream assessment of video media streaming quality*. +- [ITU-T P.1202.1] Recommendation ITU-T P.1202.1 (2012), *Parametric non-intrusive bitstream assessment of video media streaming quality – Lower resolution application area*. +- [ITU-T P.1203] Recommendation ITU-T P.1203 (2016), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport*. +- [ITU-T P.1203.2] Recommendation ITU-T P.1203.2 (2016), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Audio quality estimation module*. +- [ITU-T P.1203.3] Recommendation ITU-T P.1203.3 (2016), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Quality integration module*. +- [ITU-T P.1401] Recommendation ITU-T P.1401 (2012), *Methods, metrics and procedures for statistical evaluation, qualification and comparison of objective quality prediction models*. + +# **3 Definitions** + +## **3.1 Terms defined elsewhere** + +This Recommendation uses the following term defined elsewhere: + +### **3.1.1 mean opinion score (MOS): [ITU-T P.800.1]** + +Further terms are defined in [ITU-T P.1203]. + +## **3.2 Terms defined in this Recommendation** + +None. + +# **4 Abbreviations and acronyms** + +This Recommendation uses the following abbreviations and acronyms: + +| | | +|------|--------------------------------------| +| AAC | Advanced Audio Coding | +| DASH | Dynamic Adaptive Streaming over HTTP | +| FEC | Forward Error Correction | +| GOP | Group Of Pictures | +| HD | High Definition (television) | +| IP | Internet Protocol | +| MOS | Mean Opinion Score | +| MPEG | Moving Pictures Expert Group | +| OTT | Over The Top | +| RBSP | Raw Byte Sequence Payload | +| SD | Standard Definition | +| TS | Transport Stream | +| TCP | Transport Control Protocol | +| UDP | User Datagram Protocol | + +# **5 Conventions** + +None. + +# **6 Pv module in ITU-T P.1203 context** + +The overall model structure is shown in Figure 6-1, highlighting the position of the Pv module. More details on the general structure can be found in [ITU-T P.1203]. + +![Figure 6-1: Pv module in context of building blocks of the ITU-T P.1203 model. The diagram shows a flow from 'Stream I.01' through 'Media parameter extraction' (I.11) and 'Buffer parameter extraction' (I.14) to 'Input information'. 'I. GEN: Device info available to all modules' feeds into the 'Pv: Video quality estimation module (ITU-T P.1203.1)' (I.13) and 'Pa: Audio quality estimation module (ITU-T P.1203.2)'. Both modules feed into 'Diagnostic information', which includes 'Pq: Quality integration module (ITU-T P.1203.3)', 'Pav: A/V integration/temporal', and 'Pb: quality impact due to buffering'. The final output is 'Integral MOS' (0.46).](cfef993dcc8fb513de79eb1f93cf26ae_img.jpg) + +The diagram illustrates the architecture of the ITU-T P.1203 model. A stream labeled 'Stream I.01' enters from the left. It is processed by two extraction blocks: 'Media parameter extraction' (output I.11) and 'Buffer parameter extraction' (output I.14). These outputs are grouped under 'Input information'. A box labeled 'I. GEN: Device info available to all modules' provides input to both the 'Pv: Video quality estimation module (ITU-T P.1203.1)' (receiving I.13) and the 'Pa: Audio quality estimation module (ITU-T P.1203.2)' (receiving I.11). Both modules output to a 'Diagnostic information' block. This block contains the 'Pq: Quality integration module (ITU-T P.1203.3)', which further contains 'Pav: A/V integration/temporal' and 'Pb: quality impact due to buffering'. The outputs from these sub-modules (0.34, 0.35, and 0.23) are combined into the final 'Integral MOS' of 0.46. A reference label 'P.1203.1(17)\_F6-1' is at the bottom right. + +Figure 6-1: Pv module in context of building blocks of the ITU-T P.1203 model. The diagram shows a flow from 'Stream I.01' through 'Media parameter extraction' (I.11) and 'Buffer parameter extraction' (I.14) to 'Input information'. 'I. GEN: Device info available to all modules' feeds into the 'Pv: Video quality estimation module (ITU-T P.1203.1)' (I.13) and 'Pa: Audio quality estimation module (ITU-T P.1203.2)'. Both modules feed into 'Diagnostic information', which includes 'Pq: Quality integration module (ITU-T P.1203.3)', 'Pav: A/V integration/temporal', and 'Pb: quality impact due to buffering'. The final output is 'Integral MOS' (0.46). + +**Figure 6-1 – Pv module in context of building blocks of the ITU-T P.1203 model** + +## 6.1 Pv module modes + +The modes of operation for ITU-T P.1203.1 are defined in Table 6-1. Detailed information on exactly which inputs are available for each mode is provided in Table 7-1. Individual models are specified per mode and are described in Annexes A, B, C, D. + +**Table 6-1 – ITU-T P.1203.1 modes of operation** + +| Mode | Encryption | Input | Complexity | +|------|-------------------------------------------------|-----------------------------------------------------|------------| +| 0 | Encrypted media payload and media frame headers | Meta-data | Low | +| 1 | Encrypted media payload | Meta-data and frame header information | Low | +| 2 | No encryption | Meta-data and up-to 2% of the media stream | Medium | +| 3 | No encryption | Meta-data and any information from the video stream | Unlimited | + +# 7 Model input + +The model receives media information and prior knowledge about the media stream. In various modes of operation, inputs may be extracted or estimated from network transmissions in different ways, which are outside the scope of this Recommendation but which may be added in future annexes. The core video quality model receives the following input signals regardless of the mode of operation, following the measurement window-based procedure as specified in [ITU-T P.1203], clause 7.4: + +I.13: video coding information, as specified in [ITU-T P.1203], clause 7.1. + +Note that fault correction techniques, such as automatic repeat request (ARQ) and forward error correction (FEC) used for user datagram protocol (UDP) based streaming are not applicable in this case, where the streaming is TCP based. In TCP-based transport all retransmissions and packet loss information is typically handled transparently by the transport layer and while it can be available to + +the models described in this Recommendation it is not needed. The only information provided to the model that may implicitly include effects such as packet loss and respective retransmission is the initial loading delay and stalling information provided to the quality integration module (see [ITU-T P.1203.3]). + +## 7.1 I.13 input specification + +Table 7-1 specifies the inputs to the module. Note that the actual information available to the module at a specific output sample timestamp is restricted by the measurement window as defined in [ITU-T P.1203], clause 7.4. + +**Table 7-1 – Description of I.13** + +| ID | Description | Values | Frequency | Modes | +|-------------|------------------------------|--------------------------------------------------------------------------|-------------------|-------| +| I.13 | | | | | +| 12 | Target video bit-rate | Bit-rate in kbit/s | Per media segment | All | +| 13 | Video frame-rate | Frame rate in frames per second | Per media segment | All | +| 14 | Segment duration | Duration in seconds | Per media segment | All | +| 15 | Video encoding resolution | Number of pixels (WxH) in transmitted video | Per media segment | All | +| 16 | Video codec and profile | One of: H264-baseline, H264-high, H264-main | Per media segment | All | +| 17 | Video frame number | Integer, starting at 1, denoting frame sequence number in encoding order | Per video frame | 1,2,3 | +| 18 | Video frame duration | Duration of the frame in seconds | Per video frame | 1,2,3 | +| 19 | Frame presentation timestamp | The frame presentation timestamp | Per video frame | 1,2,3 | +| 20 | Frame decoding timestamp | The frame decoding timestamp | Per video frame | 1,2,3 | +| 21 | Video frame size | The size of the encoded video frame in bytes | Per video frame | 1,2,3 | +| 22 | Type of each picture | "I" or "Non-I" for mode 1 | Per video frame | 1 | +| 23 | Video bitstream | Encoded video bytes for the frame | Per video frame | 2,3 | + +# 8 Model algorithm and output + +The video module defined in this Recommendation has one output, O.22. It provides output values on the 5-point ACR scale ("MOS") per output sampling interval. + +There is a core model algorithm, which is described hereafter. It comprises four different modes, which all are based on the same basic function. For all four modes, the degradation due to spatial up-scaling of content that is streamed at lower resolution as well as the handling of temporal/jerkiness-related degradations due to low video framerates are the same. Only a third degradation term addressing video compression artefacts, referred to as $D_q$ in clause 8.1.1, is mode-dependent. + +The core model as well as the terms that are mode-independent are described in this clause. The model component addressing coding-related degradations, the "quantization-related component" that depends on the respective mode, is given in Annexes A (mode 0) through D (mode 3). + +The following parameters are used in the description of the model: + +- *quant*, *quant* $\in [0,1]$ : a parameter capturing the quantization degradation; its computation depends on the variant of the quantization degradation module. +- *scaleFactor*, *scaleFactor* $\in [1, \infty]$ : a parameter capturing the upscaling degradation. +- *framerate*: the video frame rate in frames per second. +- *disRes*: the video display resolution in number of pixels. For instance, high definition (HD) 1080p screens have a video display resolution of $disRes = 1920 \times 1080 = 2073600$ pixels. +- *codRes*: the video encoding resolution in pixels. The video resolution used to encode the video, for instance, $codRes = 854 \times 480 = 409920$ pixels. +- *displaySize*: the video display diagonal length, in cm or inches. +- *deviceType*: a parameter describing how the device is used; if the video display is handheld/mobile or stationary, such as a PC or TV. This parameter is preferably set based on knowledge about the device but can also be estimated from the *displaySize* parameter if it is unknown. In the case of unknown device, *deviceType* must be set to "handheld" if *displaySize* is smaller than 33 cm (13"). If in the same scenario it is equal to or larger than 33 cm, *deviceType* is "TV". + +## 8.1 Core model + +Figure 8-1 shows the Pv module in the context of building blocks of the ITU-T P.1203 model. + +### 8.1.1 Quantization degradation + +$$Dq = 100 - R_{fromMOS}(\widehat{MOSq}) \quad (1)$$ + +where *RfromMOS* is described in Annex E (as in [ITU-T G.107]), and + +$$Dq = \max(\min(Dq, 100), 0)$$ + +and where: + +$$\widehat{MOSq} = q_1 + q_2 \cdot \exp(q_3 \cdot quant) \quad (2)$$ + +and + +$$\widehat{MOSq} = \max(\min(\widehat{MOSq}, 5), 1)$$ + +with $q_1 = 4.66$ , $q_2 = -0.07$ and $q_3 = 4.06$ . + +![Diagram of the Pv module showing three input blocks (Quantization, Temporal, Upscaling) feeding into an Integration module. The Quantization module takes 'quant' as input and outputs 'Dq'. The Temporal module takes 'framerate' as input and outputs 'Dt'. The Upscaling module takes 'disRes' and 'codRes' as inputs and outputs 'Du'. All three outputs (Dq, Dt, Du) are fed into the Integration module (Block 400), which outputs the final 'MOS' value. The diagram is labeled P.1203.1(17)_F8-1.](5dfc130b129ace4df375839020a5700d_img.jpg) + +The diagram illustrates the Pv module's architecture. It consists of three parallel processing blocks on the left that feed into a single integration block on the right. + + +- Block 100 (Quantization module):** Takes the parameter *quant* as input and produces the degradation metric *Dq*. +- Block 300 (Temporal module):** Takes the *framerate* parameter as input and produces the degradation metric *Dt*. +- Block 200 (Upscaling module):** Takes the parameters *disRes* and *codRes* as inputs and produces the degradation metric *Du*. + + The metrics *Dq*, *Dt*, and *Du* are then fed into **Block 400 (Integration module)**, which outputs the final Mean Opinion Score (*MOS*). The diagram is labeled P.1203.1(17)\_F8-1. + +Diagram of the Pv module showing three input blocks (Quantization, Temporal, Upscaling) feeding into an Integration module. The Quantization module takes 'quant' as input and outputs 'Dq'. The Temporal module takes 'framerate' as input and outputs 'Dt'. The Upscaling module takes 'disRes' and 'codRes' as inputs and outputs 'Du'. All three outputs (Dq, Dt, Du) are fed into the Integration module (Block 400), which outputs the final 'MOS' value. The diagram is labeled P.1203.1(17)\_F8-1. + +Figure 8-1 – Pv module in the context of building blocks of the ITU-T P.1203 model + +### 8.1.2 Upscaling degradation + +$$Du = u_1 \cdot \log_{10}(u_2 \cdot (\text{scaleFactor} - 1) + 1) \quad (3)$$ + +with + +$$Du = \max(\min(Du, 100), 0)$$ + +and + +$$\text{scaleFactor} = \max\left(\frac{\text{disRes}}{\text{codRes}}, 1\right) \quad (4)$$ + +with $u_1 = 72.61$ and $u_2 = 0.32$ . + +### 8.1.3 Temporal degradation + +$$D_t = \begin{cases} D_{t1} - D_{t2} - D_{t3}, & \text{framerate} < 24 \\ 0, & \text{framerate} \geq 24 \end{cases} \quad (5)$$ + +Pure temporal degradation: + +$$D_{t1} = \frac{100 \cdot (t_1 - t_2 \cdot \text{framerate})}{t_3 + \text{framerate}} \quad (6)$$ + +Compensation variable 1, relative to coding impact: + +$$D_{t2} = \frac{D_q \cdot (t_1 - t_2 \cdot \text{framerate})}{t_3 + \text{framerate}} \quad (7)$$ + +Compensation variable 2, relative to spatial scaling impact: + +$$D_{t3} = \frac{Du \cdot (t_1 - t_2 \cdot \text{framerate})}{t_3 + \text{framerate}} \quad (8)$$ + +$$D_t = \max(\min(D_t, 100), 0)$$ + +where $t_1 = 30.98$ , $t_2 = 1.29$ and $t_3 = 64.65$ . + +### 8.1.4 Integration + +Degradation ( $D$ ) is calculated based on the quantization ( $D_q$ ), upscaling ( $D_u$ ) and temporal ( $D_t$ ) degradations as follows: + +$$D = \max(\min(D_q + D_u + D_t, 100), 0) \quad (9)$$ + +Maximum quality according to the pure upscaling and temporal degradations is calculated as follows: + +$$\widehat{Qmax} = \begin{cases} 100 - D_u - D_{t1}, & \text{framerate} < 24 \\ 100 - D_u, & \text{framerate} \geq 24 \end{cases} \quad (10)$$ + +Then, quality ( $Q$ ) is calculated as follows: + +$$\begin{aligned} \widehat{Q} &= 100 - D = \\ &\begin{cases} 100 - \max\left(\min\left((100 - \widehat{Qmax}) + D_q - D_{t2} - D_{t3}, 100\right), 0\right), & \text{framerate} < 24 \\ 100 - \max\left(\min\left((100 - \widehat{Qmax}) + D_q, 100\right), 0\right), & \text{framerate} \geq 24 \end{cases} \end{aligned} \quad (11)$$ + +Finally, + +$$\widehat{MOS} = MOSfromR(\widehat{Q}) \quad (12)$$ + +where $\widehat{Q}$ is the estimated video encoding, with $\widehat{Q} \in [0, 100]$ . + +*MOSfromR* is described in Annex E. + +$D$ is the overall degradation due to video representation (spatial and temporal sampling, video compression). + +If deviceType is "handheld", the following adjustment must be made on the final score: + +$$\begin{aligned}\widehat{MOS}_{handheld} &= htv1 + htv2 \cdot \widehat{MOS} + htv3 \cdot \widehat{MOS}^2 + htv4 \cdot \widehat{MOS}^3 \\ \widehat{MOS} &= \max(\min(\widehat{MOS}_{handheld}, 5), 1)\end{aligned}\quad (13)$$ + +The four coefficients in the TV/handheld-adjustment are: + +| Coefficient | Value | +|-------------|----------| +| htv1 | -0.60293 | +| htv2 | 2.12382 | +| htv3 | -0.36936 | +| htv4 | 0.03409 | + +It is noted that higher *D*, *Dq*, *Du* and *Dt* values mean higher degradations and therefore lower quality. Higher $\widehat{Q}$ and $\widehat{MOS}$ values mean higher quality. + +In the following Annexes A through D, different variants of the quant parameter are described, one for each of the four modes. + +The initial output delivered addresses the case of presentation on a PC screen of 1920x1080 pixel resolution. To estimate the O.22 output delivered by the video quality estimation module Pv for the case of viewing on a mobile device, the following adjustment of the output score is proposed. + +# Annex A + +## Mode 0 Pv module: *quant* description + +(This annex forms an integral part of this Recommendation.) + +In the case of mode 0, *quant* is calculated as described in Equation (A.2): + +$$quant = a_1 + a_2 \cdot \ln(a_3 + \ln(br) + \ln(br \cdot bpp + a_4)) \quad (A.1)$$ + +and *bpp* is the number of kbits per pixel: + +$$bpp = \frac{br}{codRes \cdot fr} \quad (A.2)$$ + +where: + +*fr* : framerate in frames per second + +*codRes* : coding resolution in number of pixels of the frame + +*br* : video stream bitrate in kbit/s. + +In case of TS-segments, the bitrate can be estimated using information about the chunk size, setting + +$$br = brChunkSize \quad (A.3)$$ + +*brChunkSize* : estimated encoding video bitrate in kbit/s, computed from the chunk size as follows: + +$$brChunkSize = \frac{chunkSize \cdot 8 - audioSize - tsHeader - pesHeader}{videoDur \cdot 1000} \quad (A.4)$$ + +where: + +*chunkSize* : chunk (audio + video) size in Bytes + +*videoDur* : duration of the video chunk in seconds + +and where: + +$$tsHeader = 4 \cdot 8 \cdot \frac{chunkSize}{188} \quad (A.5)$$ + +and + +$$pesHeader = 17 \cdot 8 \cdot (numVideoFrames + numAudioFrames) \quad (A.6)$$ + +where: + +*numVideoFrames* : number of video frames in the chunk + +*numAudioFrames* : number of video and audio frames in the chunk. + +*numVideoFrames* and *numAudioFrames* are computed as follows: + +$$numVideoFrames = int(ceil(videoDur \cdot framerate)) \quad (A.7)$$ + +and + +$$numAudioFrames = int(ceil(audioDur \cdot \frac{audioSampleRate}{samplesPerFrame})) \quad (A.8)$$ + +where: + +*audioSampleRate* : audio sample rate in Hz, e.g., 48 000 + +*samplesPerFrame* : number of samples per audio frame, e.g., 1024 for AAC + +*audioDur* : audio duration within the chunk in seconds. + +Further: + +$$audioSize = audioBrTarget \cdot audioDur \cdot 1000 \quad (\text{A.9})$$ + +where: + +*audioBrTarget*: targeted encoding audio bitrate in kbit/s + +If other means of determining the video stream bitrate are available, these may be used instead of the above calculation approach, which uses the chunk sizes. + +Table A.1 shows coefficient sets for *quant* parameter calculation for mode 0. + +**Table A.1 – Coefficient sets for *quant* parameter calculation for Mode 0** + +| Coefficient | Value | +|-------------|----------| +| $a_1$ | 11.99835 | +| $a_2$ | -2.99992 | +| $a_3$ | 41.24751 | +| $a_4$ | 0.13183 | + +## A.1 Performance analysis + +See [ITU-T P.1203] for details on the performance of the complete model with the mode 0 Pv module. + +# Annex B + +## Mode 1 Pv module: *Dq* description + +(This annex forms an integral part of this Recommendation.) + +For mode 1, the *Dq* estimation for mode 0, that is, the estimation of the degradation due to coding, is amended by a frame-information-dependent term that addresses the complexity of the video content. This way, the prediction is improved over mode 0, by considering how contents with different degrees of complexity lead to a modified encoding effectiveness. + +As shown in the equations below, the modification is carried out outside of the mode 0 *Dq*-calculation and not made part of the *quant* parameter. As a consequence, Annex B describes the calculation of the complete *Dq* for this case, instead of only the calculation of the *quant* parameter. This is in contrast to the modes 0, 2 and 3, where all effects covered by the respective module instantiation are captured by *quant* directly. + +### B.1 *quant* calculation + +In the case of mode 1, *quant* is calculated as described in Equation (B.4) for mode 2, but with different coefficient values than in the case of mode 0 (see Annex A): + +$$quant = a_1 + a_2 \cdot \ln(a_3 + \ln(br) + \ln(br \cdot bpp)) \quad (B.1)$$ + +and *bpp* is the number of kbits per pixel: + +$$bpp = \frac{brFrameSize}{codRes \cdot fr} \quad (B.2)$$ + +where: + +*fr* : framerate in frames per second +*br* *FrameSize*. + +Here, *bpp* is calculated based on the parameter *brFrameSize*, the video bitrate in kbit/s computed from the video frame sizes. + +*brFramesize* is computed as: + +$$brFramesize = \frac{\sum_{index=1}^{numVideoFrames} frameSizes(index) \cdot 8}{frameDuration \cdot numVideoFrames \cdot 1000} \quad (B.3)$$ + +Definitions: + +- *frameSizes*: Vector containing the frame size in bytes for each frame in the video chunk. +- *frameDuration*: Video frame duration in seconds, which is the same value for all frames in the video chunk. +- *numVideoFrames*: Number of video frames in the chunk. + +Table B.1 shows coefficient sets for *quant* parameter calculation for mode 1. + +**Table B.1 – Coefficient sets for *quant* parameter calculation for mode 1** + +| Coefficient | Value | +|-----------------------|----------| +| a 1 | 5.00012 | +| a 2 | −1.19631 | +| a 3 | 41.35850 | + +### B.2 $Dq$ calculation for mode 1 + +$Dq$ is calculated in the same manner as indicated in clause 8.1.1. For mode 1, however, there is an external correction factor not handled in the *quant* parameter, as outlined above, to include the effect of content-complexity. + +$$Dq = 100 - R_{fromMOS}(\widehat{MOSq}) \quad (B.4)$$ + +(compare with clause 8.1.1 Equation (1)), where $R_{fromMOS}$ is described in Annex E (same as in [ITU-T G.107]) and: + +$$Dq = \max(\min(Dq, 100), 0)$$ + +and where: + +$$\widehat{MOSq}_1 = q_1 + q_2 \cdot \exp(q_3 \cdot quant) \quad (B.5)$$ + +with the coefficients shown in Table B.2: + +**Table B.2 – Coefficients and values** + +| Coefficient | Value | +|-------------|-------| +| $q_1$ | 4.66 | +| $q_2$ | -0.07 | +| $q_3$ | 4.06 | + +In addition, we set: + +$$iFrameRatio = \frac{\bar{I}}{\bar{I}_n} \quad (B.6)$$ + +with + +$\bar{I}$ : Average size of I-frames in chunk + +$\bar{I}_n$ : Average size of non-I-frames in chunk + +$$\widehat{MOSq} = \widehat{MOSq}_1 + \text{sigmoid}(c0, c1, c2, c3, iFrameRatio) \quad (B.7)$$ + +where $\text{sigmoid}(k0, k1, k2, iFrameRatio)$ is defined with a number of input parameters as follows: + +$$scalex = \frac{10}{k2 - k1} \quad (B.8)$$ + +$$midx = \frac{k1 + k2}{2} \quad (B.9)$$ + +$$\text{sigmoid}(k0, k1, k2, iFrameRatio) = k0 - \frac{k0}{1 + e^{-scalex \cdot (iFrameRatio - midx)}} \quad (B.10)$$ + +The modelling coefficients in the sigmoid-function are set as shown in Table B.3: + +**Table B.3 – Settings of the modelling coefficients in the sigmoid-function** + +| Coefficient | Value | +|-------------|-------------| +| $k0$ | -0.91562479 | +| $k1$ | -3.28579526 | +| $k2$ | 20.4098663 | + +The final score is calculated using the bounding function: + +$$\widehat{MOS}_q = \max(\min(\widehat{MOS}_q, 5), 1) \quad (\text{B.11})$$ + +### **B.3 Performance analysis** + +See [ITU-T P.1203] for details on the performance of the complete model with the mode 1 Pv module. + +# Annex C + +## Mode 2 Pv module: *quant* description + +(This annex forms an integral part of this Recommendation.) + +## C.1 Consideration on parameter extraction + +In the case of mode 2, only 2% of bitstream data is allowed to be used. This has some implications for the way in which bitstream information is being parsed. Because the video module is run separately and the audio module used for all modes corresponds to the light-weight mode 0, the byte limit is enforced on video streams only. These considerations are described in the following: + +#### Payload not considered in the video byte count + +- Container metadata (e.g., TS headers, MP4 non-MDAT boxes, etc.) +- [ITU-T H.264] Annex B start codes +- Sequence parameter sets and picture parameter sets +- The `nal_unit_type`, the `forbidden_zero_bit` and the `nal_ref_idc` indicator, i.e., the complete NAL unit header. + +In other words, the bytes counting towards the percentage limit correspond to the raw byte sequence payload (RBSP) as defined in [ITU-T H.264], including emulation prevention bytes in the NAL payload. + +#### C.1.1 Implications of 2%-limitation for frame-specific information + +It is noted that it may happen that some frames cannot be decoded, e.g., due to corrupted frames, bugs in the decoding software or if 2% of the frame is not enough information in order to extract information from a single frame. To compensate for such cases, the model includes a so called "frame compensation", which will allow the feature *quant* to be estimated more reliably. The frame compensation shall be run directly after bitstream parsing and before the estimation of the feature *quant*. + +To do this compensation, for every output sampling interval, first all frames within a chunk will be analysed whether there are bitstream statistics available or not, using the targeted 2% of bytes per frame. During this analysis, a vector *statsVec* of length $N$ , where $N$ is the number of frames in the chunk, is created consisting of a list of $N$ tuples, where each tuple contains: + +- a boolean value and +- the frame type + +per position/frame within this chunk. The *statsVec*'s frame types component can adopt the following values: + +- "I", "P", "B", used when there are bitstream statistics available +- "I", "Non-I" where when those statistics are not present. + +Figure C.1 shows the overall pipeline of the compensation procedure: + +![Flowchart showing the relationship between frame information and workaround by going back to mode 1. The process starts with 'statsVec' pointing to a 'Bitstream statistics' box containing 'some frames', 'all frames', and 'no frames'. 'some frames' leads to 'compensate', which then leads to 'mode 2/3'. 'all frames' also leads to 'mode 2/3'. 'no frames' leads to 'mode 1'. Both 'mode 2/3' and 'mode 1' lead to 'score'. A small label 'P.1203.1(16)_FC.1' is at the bottom right.](e69b9188aa2c14ec6b21c83f711fef65_img.jpg) + +``` + +graph TD + statsVec[statsVec] --> BitstreamStatistics[Bitstream statistics] + subgraph BitstreamStatistics + someFrames[some frames] + allFrames[all frames] + noFrames[no frames] + end + someFrames --> compensate[compensate] + allFrames --> mode23[mode 2/3] + noFrames --> mode1[mode 1] + compensate --> mode23 + mode23 --> score[score] + mode1 --> score + P1203116_FC1[P.1203.1(16)_FC.1] + +``` + +Flowchart showing the relationship between frame information and workaround by going back to mode 1. The process starts with 'statsVec' pointing to a 'Bitstream statistics' box containing 'some frames', 'all frames', and 'no frames'. 'some frames' leads to 'compensate', which then leads to 'mode 2/3'. 'all frames' also leads to 'mode 2/3'. 'no frames' leads to 'mode 1'. Both 'mode 2/3' and 'mode 1' lead to 'score'. A small label 'P.1203.1(16)\_FC.1' is at the bottom right. + +**Figure C.1 – Relation between frame information and workaround by going back to mode 1 in the case of "no frames have stats" for a given chunk / model window** + +The previously created *statsVec* will now be analysed and classified dependent on the number of bitstream statistics available in the current chunk. There are three scenarios: + +- 1 No frame has bitstream statistics +- 2 All frames have bitstream statistics +- 3 Some frames have bitstream statistics. + +##### No frames have bitstream statistics + +In this case the model uses the algorithm specified in mode 1 for the calculation of the feature *quant*. + +##### All frames have bitstream statistics + +Since all frames can be decoded, there is no need for compensation. Therefore, the algorithm for the calculation of *quant* in mode 2/3 is applied directly. + +##### Some frames have bitstream statistics + +Since some frames have bitstream statistics it is possible to compensate for frames where no bitstream statistics are available in order to avoid the drawbacks of falling back to the mode 1 algorithm. This compensation is done in four steps: + +- 1 Iterate over all the frames in the chunk and generate *statsVec* (as described above). +- 2 Iterate over *statsVec* starting at position *pos* = 0 where 0 is the index of the first frame in the chunk. +- 3 If the frame at *pos* = 0 has statistics: + - a) Continue iterating *pos* until a frame without bitstream statistics is reached, store *statsPosLastFound* = *pos* - 1. + - b) Continue iterating *pos* until a frame with bitstream statistics is reached. The area between *pos* and *statsPosLastFound* + 1 is now called a "hole", i.e., a contiguous region of frames without bitstream statistics. + - c) For every frame within the hole at position *pos* with type *type*, call the function *findNearest(pos, type)* (as described below) in order to fill the bitstream statistics. +- 4 If the frame at *pos* = 0 has no bitstream statistics: + - a) Set *statsPosLastFound* to -1 + +b) Continue with step 3.b. + +##### **findNearest(pos, type)** + +This function searches for the nearest frame (starting at position *pos*) having bitstream statistics, which has the type *type*. If no frame matching these criteria can be found, the search is performed again, this time allowing any frame type to be matched. + +### **C.2 quant calculation** + +The quant parameter is calculated using the following formula: + +$$quant = \frac{\overline{QP_{PB}}}{51} \quad (C.1)$$ + +Where $\overline{QP_{PB}}$ is the average over all values in the vector $QP_{PB}$ , where $QP_{PB}$ is computed as described in the pseudocode below. + +Definitions: + +- *initialQP*: vector containing the initial quantization parameter (QP, with $QP \in [0, 51]$ ) value for each frame in the video chunk; +- *twoPercentQP*, a vector containing, for each frame in the video chunk, the average of the QP values for each complete macroblock (MB) contained in the first 2% (in bytes) of the current frame's payload. If the frame contains black borders on the top and bottom of the frame (so-called "letterboxing"), the QP values of the MBs contained in these black borders are ignored; +- *frameTypes*, a vector containing the frame type for each video frame in the chunk. + +Here, video frames are parsed in their order of arrival and are therefore captured in decoding order, not in presentation/display order. + +Note: Array indices start at 0. + +``` +frameTypes ← [I,P,B,...] +QPPB ← [] +p ← 2 +index ← 0 + +FOR frameType IN frameTypes: + IF (frameType = P) OR (frameType = B): + IF (twoPercentQP[index] = 0) OR + (twoPercentQP[index] could not be computed) OR + (p% for the frame contains only MBs in black border region): + QPPB[index] ← initialQP[index] + ELSE: + QPPB[index] ← twoPercentQP[index] + index ← index + 1 +``` + +### **C.3 Performance analysis** + +See [ITU-T P.1203] for details on the performance of the complete model with the mode 2 Pv module. + +# Annex D + +## Mode 3 Pv module: *quant* description + +(This annex forms an integral part of this Recommendation.) + +$$quant = \frac{\overline{QP_{PB}}}{51} \quad (D.1)$$ + +where $\overline{QP_{PB}}$ is the average over all values in the vector $QP_{PB}$ . For mode 3, $QP_{PB}$ is computed as described in the pseudocode below. + +Definitions: + +- *numMBdec*: Vector containing the number of decoded MBs in the current frame for each video frame in the chunk. +- *numMBskip*: Vector containing the number of skipped MBs in the current frame for each video frame in the chunk. +- *averageQP*: Vector containing, for each video frame in the chunk, the average over the current frame of the QP values per MB, ignoring MBs in the black border areas, if the frame contains black borders. +- *frameTypes*: Vector containing the frame types for each video frame in the chunk. + +Video frames are parsed from the bitstream and are therefore captured in decoding order. + +Note: Array indices start at 0. + +``` +frameTypes ← [I,P,B,...] +QPP ← [] +QPB ← [] +QPPB ← [] +index ← 0 + +FOR frameType IN frameTypes: + + CASE frameType = P: + + IF (length(QPP) = 0) OR + (numMBskip[index] / numMBdec[index] < 0.99): + QPP.append(averageQP[index]) + + CASE frameType = B: + QPB.append(averageQP[index]) + + CASE (frameType = I) AND (length(QPP) > 0): + + IF length(QPP) > 1: + QPP[length(QPP)-1] ← QPP[length(QPP)-2] + ELSE: + QPP ← [] + + index ← index + 1 + +QPPB ← QPP +QPPB.append(QPB) +``` + +### **D.1 Performance analysis** + +See [ITU-T P.1203] for details on the performance of the complete model with the mode 3 Pv module. + +# Annex E + +## *MOSfromR* and *RfromMOS* and functions + +(This annex forms an integral part of this Recommendation.) + +## E.1 *MOSfromR* definition + +*MOSfromR* is expressed as follows: + +$$MOSfromR: \mathbb{R} \mapsto \mathbb{R}$$ + +$$Q \mapsto MOS := MOSfromR(Q)$$ + +$$MOS = MOS_{MIN} + (MOS_{MAX} - MOS_{MIN}) * \frac{Q}{100} + Q * (Q - 60) * (100 - Q) * 0.000007 \quad (E.1)$$ + +$$MOS = \min(MOS_{MAX}, \max(MOS, MOS_{MIN})) \quad (E.2)$$ + +where $MOS_{MAX} = 4.9$ and $MOS_{MIN} = 1.05$ . + +### E.2 *RfromMOS* definition + +*RfromMOS* is expressed as follows: + +$$RfromMOS: \mathbb{R} \mapsto \mathbb{R}$$ + +$$MOS \mapsto Q := RfromMOS(MOS)$$ + +It is the inverse function of *MOSfromR*. + +Since *RfromMOS* is not explicitly defined, the resulting value $Q$ for a given $MOS$ value shall be calculated using a lookup table based on *MOSfromR*, using linear interpolation. + +For the implementation of the inverse function, the absolute error of performing a double-conversion must be less than 0.01 in the $MOS$ scale for all $MOS$ values within the valid range $[MOS_{MIN}, MOS_{MAX}]$ . In other words, it must be ensured that: + +$$|MOS - MOSfromR(RfromMOS(MOS))| < 0.01 \quad MOS \in [MOS_{MIN}, MOS_{MAX}] \quad (E.3)$$ + +#### E.2.1 Example for lookup table + +The following algorithm gives an example of how to construct such a lookup table: + +Define a table $h$ consisting of a unique set of keys $K$ and a set of values $V$ , where $h(k)$ is the respective value $v$ for the given key $k$ . The lookup table must have only one value $v$ for any given key $k$ . + +Let $h(MOSfromR(v)) = v$ for every $v \in \{0, 0.25, 0.5, \dots, 100\}$ and let $h(MOS_{MIN}) = 0$ . + +To obtain the value of $Q$ for a given value of $MOS$ , it must first be clipped to the allowed range: + +$$MOS = \min(MOS_{MAX}, \max(MOS, MOS_{MIN})) \quad (E.4)$$ + +Then, $Q$ is calculated based on a linear interpolation: + +$$Q = \frac{V_p(K_q - MOS) + V_q(MOS - K_p)}{K_q - K_p} \quad (\text{E. 5})$$ + +where $K_p, K_q$ are the nearest interpolation points available in $h$ , and $V_p, V_q$ the respective values at these points. + +# Annex F + +## Byte-counting algorithm for Mode 2 + +(This annex forms an integral part of this Recommendation.) + +### Byte-counting algorithm + +Below is pseudocode that illustrates the expected behaviour of a conformant byte-counting implementation: + +``` +pv = PVModule(metaData) +perFrameBytesConsumed = [] +bytesConsumed = 0 +totalBytes = 0 +for h264File in h264FileList: + byteStream = open(h264File) + while not eof: + consume start code at start of byteStream + nalUnitType = getNalUnitType(byteStream[0]) + nextStartCodeOffset = findNextStartCode(byteStream) + nalUnit = byteStream[:nextStartCodeOffset] + if nalUnitType == SPS: + pv.consumeSps(nalUnit) + elif nalUnitType == PPS: + pv.consumePps(nalUnit) + elif nalUnitType in (IDR_SLICE, SLICE): + totalBytes += len(nalUnit) - 1 + frameBytesConsumed = pv.consumeSliceNalUnit(nalUnit) + perFrameBytesConsumed.append(frameBytesConsumed) + bytesConsumed += frameBytesConsumed + byteStream = byteStream[nextStartCodeOffset:] +pvScores = pv.getScores() +assert bytesConsumed <= (totalBytes * 0.02), "Byte count exceeded" +return (pvScores, perFrameBytesConsumed) +``` + + + +# SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | Tariff and accounting principles and international telecommunication/ICT economic and policy issues | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound-programme and television-transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling, and associated measurements and tests | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects, next-generation networks, Internet of Things and smart cities | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.1204.1-202510-I_PDF-E/7f17c430b9598e4d748a8041457810b3_img.jpg b/marked/P/T-REC-P.1204.1-202510-I_PDF-E/7f17c430b9598e4d748a8041457810b3_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dddcee612263f835c39fd5b321742eb69c46729a --- /dev/null +++ b/marked/P/T-REC-P.1204.1-202510-I_PDF-E/7f17c430b9598e4d748a8041457810b3_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:cda41b8892708c3b5cf5679403e5e875e6f34acf72ec311794e985489a6074af +size 45382 diff --git a/marked/P/T-REC-P.1204.1-202510-I_PDF-E/84a1d09fb489061482111515543b60dc_img.jpg b/marked/P/T-REC-P.1204.1-202510-I_PDF-E/84a1d09fb489061482111515543b60dc_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..ac295b1b0c7f14b1c5a33ef6682b1b017c812477 --- /dev/null +++ b/marked/P/T-REC-P.1204.1-202510-I_PDF-E/84a1d09fb489061482111515543b60dc_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:37aed4fc114a39843438175326e0b7e1d09a9d6075723d82fea32040fd560ee4 +size 7192 diff --git a/marked/P/T-REC-P.1204.4-202207-I_PDF-E/raw.md b/marked/P/T-REC-P.1204.4-202207-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..c8197429acaa648e94a9178515af2ceae3e6ce20 --- /dev/null +++ b/marked/P/T-REC-P.1204.4-202207-I_PDF-E/raw.md @@ -0,0 +1,1325 @@ + + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.1204.4** + +(07/2022) + +SERIES P: TELEPHONE TRANSMISSION QUALITY, +TELEPHONE INSTALLATIONS, LOCAL LINE +NETWORKS + +Models and tools for quality assessment of streamed +media + +--- + +**Video quality assessment of streaming services +over reliable transport for resolutions up to 4K +with access to full and reduced reference pixel +information** + +Recommendation ITU-T P.1204.4 + +# ITU-T P-SERIES RECOMMENDATIONS **TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS** + +| | | +|----------------------------------------------------------------------------------------------------|----------------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | P.10–P.19 | +| Voice terminal characteristics | P.30–P.39 | +| Reference systems | P.40–P.49 | +| Objective measuring apparatus | P.50–P.59 | +| Objective electro-acoustical measurements | P.60–P.69 | +| Measurements related to speech loudness | P.70–P.79 | +| Methods for objective and subjective assessment of speech quality | P.80–P.89 | +| Voice terminal characteristics | P.300–P.399 | +| Objective measuring apparatus | P.500–P.599 | +| Measurements related to speech loudness | P.700–P.709 | +| Methods for objective and subjective assessment of speech and video quality | P.800–P.899 | +| Audiovisual quality in multimedia services | P.900–P.999 | +| Transmission performance and QoS aspects of IP end-points | P.1000–P.1099 | +| Communications involving vehicles | P.1100–P.1199 | +| Models and tools for quality assessment of streamed media | P.1200–P.1299 | +| Telemeeting assessment | P.1300–P.1399 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | P.1400–P.1499 | +| Methods for objective and subjective assessment of quality of services other than speech and video | P.1500–P.1599 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +# Recommendation ITU-T P.1204.4 + +# Video quality assessment of streaming services over reliable transport for resolutions up to 4K with access to full and reduced reference pixel information + +## Summary + +Recommendation ITU-T P.1204.4 describes the reduced-reference and full-reference video quality estimation model for Recommendation ITU-T P.1204 for monitoring the video quality for streaming using reliable transport (e.g., hypertext transfer protocol- (HTTP-) based adaptive streaming (HAS) over the transmission control protocol (TCP), and quick user datagram protocol Internet connections (QUIC)). The estimate is validated for videos encoded with H.264, H.265, video payload type 9 (VP9), or AOMedia Video 1 (AV1) codecs at any resolution up to 4K/ultra-high definition (UHD) ( $3\,840 \times 2\,160$ ) resolution for personal computer (PC) monitors and television (TV) and up to $2\,560 \times 1\,440$ for smartphone and tablet displays. + +The ITU-T P.1204 series of Recommendations provides sequence-related (between 5 s and 10 s) and per-1-second video-quality estimation. In principle, the per-one-second outputs of these video-quality models can be used together with an audio model for integration into audiovisual quality and, together with information about initial loading delay and media playout stalling events, further into a final per-session model output, an estimate of integral per-session quality (see e.g., Recommendations ITU-T P.1203, ITU-T P.1203.2, ITU-T P.1203.3). + +Recommendation ITU-T P.1204.4 was developed in collaboration with the Video Quality Experts Group (VQEG). + +This ITU-T P.1204-series of Recommendations addresses three application areas: + +- large-screen presentation as with fixed-network video streaming; +- mobile streaming on handheld devices such as smartphones; +- presentation on tablet-type devices. + +## History + +| Edition | Recommendation | Approval | Study Group | Unique ID* | +|---------|------------------------------|------------|-------------|---------------------------------------------------------------------------| +| 1.0 | ITU-T P.1204.4 | 2020-01-13 | 12 | 11.1002/1000/14157 | +| 1.1 | ITU-T P.1204.4 (2020) Amd. 1 | 2021-01-07 | 12 | 11.1002/1000/14592 | +| 2.0 | ITU-T P.1204.4 | 2022-07-29 | 12 | 11.1002/1000/15007 | + +## Keywords + +Adaptive streaming, IPTV, mean opinion score (MOS), mobile video, mobile TV, monitoring, multimedia, OTT, progressive download, QoE, TV, video. + +--- + +\* To access the Recommendation, type the URL in the address field of your web browser, followed by the Recommendation's unique ID. For example, . + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had received notice of intellectual property, protected by patents/software copyrights, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the appropriate ITU-T databases available via the ITU-T website at . + +© ITU 2022 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## Table of Contents + +| | Page | +|--------------------------------------------------------------------------|------| +| 1 Scope ..... | 1 | +| 2 References..... | 2 | +| 3 Definitions ..... | 3 | +| 3.1 Terms defined elsewhere ..... | 3 | +| 3.2 Terms defined in this Recommendation ..... | 3 | +| 4 Abbreviations and acronyms ..... | 3 | +| 5 Conventions ..... | 4 | +| 6 Areas of application..... | 5 | +| 6.1 Application range for the models ..... | 5 | +| 7 Video module in the ITU-T P.1204 context ..... | 7 | +| 8 Model input..... | 7 | +| 9 Model algorithm and output ..... | 8 | +| 10 Model description ..... | 8 | +| 10.1 Notation ..... | 8 | +| 10.2 Multi-resolution frame pyramid ..... | 9 | +| 10.3 Frame edge representation..... | 9 | +| 10.4 Local edge statistic ..... | 11 | +| 10.5 Sharpness..... | 12 | +| 10.6 Features of reference (side information) ..... | 13 | +| 10.7 Features of test video..... | 13 | +| 10.8 Additional features ..... | 14 | +| 10.9 Step functions ..... | 16 | +| 10.10 Transformations..... | 17 | +| 10.11 Core model ..... | 17 | +| 10.12 Viewing distance ..... | 19 | +| 10.13 Parameters ..... | 19 | +| 10.14 Device-dependent model parameters ..... | 20 | +| 10.15 S -transformation parameters..... | 20 | +| Appendix I – Performance figures ..... | 22 | +| Appendix II – Long term integration module (Pq) for ITU-T P.1204.4 ..... | 23 | +| II.1 Model input ..... | 23 | +| II.2 Model output ..... | 23 | +| II.3 Model description..... | 24 | +| II.4 Performance figures..... | 26 | +| Bibliography..... | 28 | + + + +# Recommendation ITU-T P.1204.4 + +## Video quality assessment of streaming services over reliable transport for resolutions up to 4K with access to full and reduced reference pixel information + +## 1 Scope + +This Recommendation describes a full-reference/reduced-reference video quality assessment module that together with audio and integration modules can be used to form a complete model to predict the impact of audio and video media encodings and observed Internet protocol (IP) network impairments on quality experienced by the end-user in multimedia streaming applications. The streaming techniques addressed comprise progressive download and adaptive streaming, for both mobile and fixed network streaming applications. The video quality module can also be used stand-alone as a video quality prediction model. + +The model described here is a pixel-based reduced-reference model, which can be seen as a special form of a full-reference model. It can be used from monitoring to accurate measurement of client-side quality of experience (QoE) for network or service optimization or benchmarking purposes. The model developed may be deployed both in end-point locations and at mid-network monitoring points, as long as the reference information can be made available at that location. The model in this Recommendation can also be used for laboratory testing of video systems. + +The model described here is applicable to progressive download and adaptive streaming or other streaming applications with reliable transport, where the quality experienced by the end user is affected by video degradations due to coding, spatial re-scaling or variations in video frame rates. Quality assessment of adaptive streaming includes aspects of media adaptation that may be handled in integration modules such as [ITU-T P.1203.3], and not in the video modules in this Recommendation. This Recommendation is able to handle various video codecs (i.e., H.264, H.265/high-efficiency video coding (HEVC), video payload type 9 (VP9), and AV1, resolutions up to 4K/ ultra-high definition-1 (UHD-1) and frame rates up to 60 frames/s. The video-quality module Pv of [b-ITU-T P.1203], i.e., [ITU-T P.1203.1], only addresses H.264 and full high definition (HD) with up to 30 frames/s. + +The model predicts a mean opinion score (MOS) on a five-point absolute category rating (ACR) scale (see [ITU-T P.910]) as an overall video quality MOS (5 s to 10 s). In addition to the overall quality score, this video quality model produces a per-one-second quality score, suitable for diagnostics or integration into an integral quality score for longer sessions (see, for example, [ITU-T P.1203.3] for 1 min to 5 min duration sessions). + +The model associated with this Recommendation cannot provide a comprehensive evaluation of the video quality as perceived by an *individual end-user* because the scores reflect the perceived impairments due to coded video media data being transmitted over an IP connection with a certain performance and do not include specific terminal devices or user-specific information. The scores predicted by such a general quality model necessarily reflect *average perceptual quality*. + +Effects due to source generations, such as signal noise, video shake, certain colour properties (and other similar video factors) and other impairments related to the payload, are not reflected in the scores computed by this model. + +As a consequence, this Recommendation can be used for applications such as: + +- in-service quality monitoring for specific IP-based audiovisual services, as specified in more detail in clause 6.1; +- performance and quality assessment of live networks (including video encoding) considering the effect due to encoding bitrate, encoding resolution and encoding frame rate; + +- laboratory testing of video systems; +- benchmarking of different service implementations; +- benchmarking of different encoder implementations; +- evaluation of transcoding solutions. + +In particular, targeted applications are progressive download streaming and adaptive streaming (using reliable transport), which includes the following. + +- Over-the-top (OTT) services, as well as operator-managed video services (over the transmission control protocol (TCP)). +- Video over both mobile and fixed connections. +- The streaming protocols HTTP live streaming (HLS) or dynamic adaptive streaming over HTTP (DASH) used with the hypertext transfer protocol (HTTP) or HTTP2 over TCP/IP or quick user datagram protocol Internet connections (QUIC), or real-time messaging protocol (RTMP) over TCP/IP. Note that the model is agnostic to the specific application or transport layer protocol, with the exception that it assumes reliable delivery of video packets. +- Video services typically using container formats based on the ISO/IEC base media file format such as Moving Picture Experts Group-4 (MPEG-4) Part 14 (MP4), or other container formats such as audio video interleave (AVI), Matroska video (MKV), WebM, Third Generation Partnership (3GP), and MPEG-2 transport stream (MPEG2-TS). Note that the model is agnostic to the type of container format. + +## 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- [ITU-T H.264] Recommendation ITU-T H.264 (2021), *Advanced video coding for generic audiovisual services*. +- [ITU-T H.265] Recommendation ITU-T H.265 (2021), *High efficiency video coding*. +- [ITU-T P.910] Recommendation ITU-T P.910 (2022), *Subjective video quality assessment methods for multimedia applications*. +- [ITU-T P.1203.1] Recommendation ITU-T P.1203.1 (2019), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Video quality estimation module*. +- [ITU-T P.1203.3] Recommendation ITU-T P.1203.3 (2019), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Quality integration module*. + +## 3 Definitions + +### 3.1 Terms defined elsewhere + +This Recommendation uses the following term defined elsewhere: + +**3.1.1 bitstream** [ITU-T H.264]: A sequence of bits that forms the representation of coded pictures and associated data forming one or more coded video sequences. Bitstream is a collective term used to refer either to a NAL unit stream or a byte stream. + +**3.1.2 integral quality** [b-ITU-T P.1203]: The quality as perceived by a subject in a subjective test, which corresponds to the scope of this Recommendation. Artefacts presented in the subjective tests typically include a combination of audio compression, video compression, and stalling effects. + +**3.1.3 mean opinion score (MOS)** [b-ITU-T P.1204]: The mean of opinion scores, which are values on a predefined scale that subjects assign to their opinion of the performance of the telephone transmission system used either for conversation or for listening to spoken material. + +NOTE – Paraphrased from clause 7 of [b-ITU-T P.800.1]. + +**3.1.4 media adaptation** [b-ITU-T P.1203]: Events where the player switches video playback between a known set of media quality levels while adapting to network conditions, by downloading and decoding individual segments in sequence. + +**3.1.5 media quality level** [b-ITU-T P.1203]: A particular encoding setting applied to a video or audio stream. + +**3.1.6 model, model algorithm** [b-ITU-T P.1203]: An algorithm with the purpose of estimating the subjective (perceived) quality of a media sequence. + +**3.1.7 sequence** [b-ITU-T P.1203]: An audiovisual stream composed of multiple non-overlapping segments. + +**3.1.8 video chunk** [b-ITU-T G.1022]: A contiguous set of samples for one track of a video. + +### 3.2 Terms defined in this Recommendation + +None. + +## 4 Abbreviations and acronyms + +This Recommendation uses the following abbreviations and acronyms: + +| | | +|------|--------------------------------------| +| 3GP | Third Generation Partnership | +| AAC | Advanced Audio Coding | +| ACR | Absolute Category Rating | +| AV1 | AOMedia Video 1 | +| AVC | Advanced Video Coding | +| AVI | Audio Video Interleave | +| DASH | Dynamic Adaptive Streaming over HTTP | +| GoP | Group of Pictures | +| HAS | HTTP-based adaptive streaming | +| HD | High Definition | +| HEVC | High-Efficiency Video Coding | +| HLS | HTTP live streaming | + +| | | +|----------|---------------------------------------------------| +| HTTP | Hypertext Transfer Protocol | +| IP | Internet Protocol | +| MKV | Matroska Video | +| MO | Mobile | +| MOS | Mean Opinion Score | +| MP4 | MPEG-4 Part 14 | +| MPEG | Moving Pictures Expert Group | +| MPEG2-TS | MPEG-2 Transport Stream | +| NAL | Network Abstraction Layer | +| PC | Personal Computer | +| QoE | Quality of Experience | +| QUIC | Quick User datagram protocol Internet Connections | +| RMSE | Root Mean Square Error | +| RTMP | Real-Time Messaging Protocol | +| RTP | Real-time Transport Protocol | +| TA | Tablet | +| TCP | Transmission Control Protocol | +| TS | Transport Stream | +| UDP | User Datagram Protocol | +| UHD | Ultra-High Definition | +| VVC | Versatile Video Coding | + +## 5 Conventions + +This Recommendation uses the following conventions: + +- 4K: Video resolution of $4\,096 \times 2\,160$ or $3\,840 \times 2\,160$ ; +- Pv designates the video quality estimation module (as specified in this Recommendation for the case of full and reduced reference pixel-based prediction, see [b-ITU-T P.1204] for alternative implementations such as bitstream based and hybrid); +- Reliable transport: Reliable delivery with protocols guaranteeing no loss of information. + +## 6 Areas of application + +### 6.1 Application range for the models + +Table 1 shows the application range of the model in this Recommendation based on what the model has actually been developed for. Table 2 lists areas where it is not applicable. Table 3 lists test factors and coding technologies for which this Recommendation has been validated for H.264, H.265, VP9, and correspondingly Table 4 lists test factors for the case of AV1 encoding. + +**Table 1 – Application areas for which this Recommendation is applicable** + +| Areas for which the model is applicable | +|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| In-service monitoring of video sent over reliable transport. Both OTT services and operator-managed video services, using reliable delivery with protocols such as HTTP or HTTP2 over TCP/IP or QUIC, or RTMP over TCP/IP. Note that this model is agnostic to the type of container format. | +| Performance and quality assessment of live networks (including video encoding) considering impairments due to encoding bitrate, encoding resolution, and encoding frame rate. | +| Laboratory testing of video systems. | +| Benchmarking of different service implementations. | +| Benchmarking of different encoder implementations. Note that only the full and reduced reference pixel-based model type can be used for direct benchmarking of this type. | +| Evaluation of transcoding solutions. | + +**Table 2 – Application areas for which this Recommendation is not applicable** + +| Areas for which the model is not applicable | +|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| In-service monitoring of video streaming using unreliable transport (e.g., real-time transport protocol/user datagram protocol (RTP/UDP), where packet loss introduces visible quality degradations. | +| Evaluation of visual quality of display/device properties. | +| Evaluation of audio/video sync distortions. | +| Evaluation of video codecs for which the model is not validated (MPEG-I Part 3 [versatile video coding (VVC)], etc.). | +| Evaluation of the effects of noise, delay, colour correctness or other content-production-related aspects. | + +**Table 3 – Test factors, and coding technologies for which this Recommendation has been validated** + +| Video test factors for which the model has been validated | | +|------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| Video content | Movies and movie trailers, sports videos, documentaries, computer-generated graphics/games, etc. | +| Input video length | The video modules were trained and validated to produce one overall video-quality score for a chunk of ~7-9 s and also provide the per-second scores. Optimal performance for ~8 s. Models are assumed to provide valid overall video-quality estimations for 5-10 s long sequences. | +| Bitstream container | AVI, MP4, MKV, WebM | +| Encoder types (and implementation), see Note 1 | H.264/advanced video coding (AVC) (libx264), H.265/HEVC (libx265), VP9 (libvpx-vp9), AV1 (libaom-av1) | + +**Table 3 – Test factors, and coding technologies for which this Recommendation has been validated** + +| Video test factors for which the model has been validated | | | | | +|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------|---------------------------------------------|------------------------------| +| Encoder profiles | H.264 (MPEG-4 Part 10): Constrained baseline, Main, Hi, Hi10, Hi422.
H.265: Main, Main10, Rext.
VP9: 0, 1, 2, 3. | | | | +| Video resolution and bitrate | Resolution definition | Video height range | Personal computer/television (PC/TV) | Mobile/tablet (MO/TA) | +| | Below SD | 180-70 | – | 90 Kbps – 1 Mbps | +| | SD | 360-540 | 150 Kbps – 4 Mbps | 150 Kbps – 4 Mbps | +| | HD | 720-1 080 | 500 Kbps – 15 Mbps | 500 Kbps – 15 Mbps | +| | Above HD | 1 440-2 160 | 1.5 Mbps – 45 Mbps | 1.5 Mbps – 20 Mbps | +| Video aspect ratio | 16:9, see Note 2 | | | | +| Group of pictures (GoP) | Variable. Average GoP length can be between 0.5 s and chunk duration | | | | +| Bit-depth | 8 bits or 10 bits | | | | +| Chroma subsampling | YUV 4:2:0 and YUV 4:2:2 | | | | +| OTTs | Online providers that offer video on demand and video encoding as a service. It should be noted that the models are applicable for similar OTTs. | | | | +| Display resolution and frame rate | PC/TV: 2 160p, up to 60 frames/s.
Mobile/Tablet: 1 440p, up to 60 frames/s. | | | | +| Viewing distances | PC/TV: $1.5H$ to $3H$ ( $H$ : Screen height), see Note 3
MO/TA: $4H$ to $6H$ | | | | +| NOTE 1 – During training and validation, FFmpeg 3.2.2 was used with x264 snapshot 20170202-2245, x265 v2.2, libvpx 1.6.1.
NOTE 2 – For original content with a larger aspect ratio, letterboxing of up to 30% was allowed, that is 1512 pixels height for video coded at 2160 pixels height. Video content with 1.89:1 aspect ratio (e.g., cinema 4K) may also be used.
NOTE 3 – It is noted that for PC/MO, the model output is conservative and should be interpreted to correspond to a viewing distance of $1.5H$ to $1.6H$ . | | | | | + +**Table 4 – Additional test factors for which this Recommendation has been validated for the case of AV1 encoding** + +| | | | | | +|---------------------------------------------------------------------------------------------------------|------------------------------|---------------------------|---------------------------------------------|------------------------------| +| Encoder profile | Main | | | | +| Video resolution and bitrate | Resolution definition | Video height range | Personal computer/television (PC/TV) | Mobile/tablet (MO/TA) | +| | Below SD | 180-70 | 180 Kbps – 8 Mbps | 35 Kbps – 4.5 Mbps | +| | SD | 360-540 | 200 Kbps – 8 Mbps | 100 Kbps – 6 Mbps | +| | HD | 720-1 080 | 500 Kbps – 30 Mbps | 300 Kbps – 30 Mbps | +| | Above HD | 1 440-2 160 | 600 Kbps – 45 Mpbs | 300 Kbps – 45 Mbps | +| NOTE – AV1 encoding was performed using FFmpeg 4.2.2 and libaom-av1 (1.0.0-errata1-avif-73-g3aca30199). | | | | | + +## 7 Video module in the ITU-T P.1204 context + +The building blocks of the ITU-T P.1204 model used stand-alone are depicted in Figure 1. + +![Figure 1: Building blocks of the ITU-T P.1204 video quality model used stand-alone. The diagram shows a 'Stream I.01' entering a 'Pixel information extraction (decoder)' block, which outputs 'I.15'. This 'I.15' signal and 'Input information' from 'I.16 (Reference)' are fed into a 'Pixel-based [ITU-T P.1204.4]' block. A 'Media parameter extraction' block also receives 'I.15' and provides input to the 'Pixel-based' block. The 'Pixel-based' block outputs 'Pv ITU-T P.1204'. Below the 'Pixel-based' block is an 'I.GEN: Device info available to all modules' block. The final outputs are 'O.27 (Video quality MOS, 5-10 s)' and 'O.22 (5-point per-1-second video quality)'. A label 'P.1204.4(20)_F01' is at the bottom right.](eefe19c5e14dc4d6c316b7f7fbb7d7d7_img.jpg) + +Figure 1: Building blocks of the ITU-T P.1204 video quality model used stand-alone. The diagram shows a 'Stream I.01' entering a 'Pixel information extraction (decoder)' block, which outputs 'I.15'. This 'I.15' signal and 'Input information' from 'I.16 (Reference)' are fed into a 'Pixel-based [ITU-T P.1204.4]' block. A 'Media parameter extraction' block also receives 'I.15' and provides input to the 'Pixel-based' block. The 'Pixel-based' block outputs 'Pv ITU-T P.1204'. Below the 'Pixel-based' block is an 'I.GEN: Device info available to all modules' block. The final outputs are 'O.27 (Video quality MOS, 5-10 s)' and 'O.22 (5-point per-1-second video quality)'. A label 'P.1204.4(20)\_F01' is at the bottom right. + +**Figure 1 – Building blocks of the ITU-T P.1204 video quality model used stand-alone in the case of this Recommendation** + +## 8 Model input + +In a full-reference model, quality $q$ of a test video $v$ – called a degraded video – is estimated by a function $G$ depending on the degraded video $v$ and on the reference video $v_{\text{ref}}$ , + +$$q = G(v, v_{\text{ref}})$$ + +In the reduced-reference case, the function $G$ depends on the reference through features $f_{\text{ref}}$ of the reference $v_{\text{ref}}$ only. The features are extracted by the reference feature extraction function $\phi$ , + +$$f_{\text{ref}} = \phi(v_{\text{ref}})$$ + +and there is a restriction on the size of the features. The quality of the degraded video is estimated by function $G'$ by + +$$q = G'(v, f_{\text{ref}}).$$ + +Thus, a reduced-reference model is a special form of a full-reference model. The reference features $f_{\text{ref}}$ are sometimes called the *side information*, as in an operational setup this information can be transmitted over a side-channel to the measurement device. + +The model receives information **I.GEN** about the viewing condition per media session: *display resolution, device type, display size, and relative viewing distance*. + +The video quality model receives the following input signals: the reference video for extraction of the reference features, and the degraded video signal, decoded and possibly upscaled. It is possible to process the video in chunks, respecting the sampling of the output signal. See Table 5 for an overview of model inputs. + +**Table 5 – Input description of I.GEN, I.15, I.16** + +| ID | Description | Values | Frequency | +|--------------|---------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-----------------| +| I.GEN | | | | +| 0 | The resolution of the image displayed to the user | Number of pixels ( $W \times H$ ) in displayed video | Per media chunk | +| 1 | The device type on which the media is played | "PC", "TV", "MO", "TA" | Per media chunk | +| 2 | Device display size | Diagonal size (diagonal in inches) | Per media chunk | +| 3 | Relative viewing distance in multiple of display height | Relative viewing distance | Per media chunk | +| I.15 | | | | +| 4 | Degraded video | The raw pixels (YUV file including metadata required for parsing; width, height, frame rate, and pixel format) of the processed video, i.e., the video decoded and upscaled to display resolution without buffering or stalling. The frame information in I.16 and I.15 is synchronized, i.e., no frame misalignments are present. | Per media chunk | +| I.16 | | | | +| 5 | Reference video information | The reference-side information extraction module takes as input the reference video and outputs the side information file. The reference model side channel bandwidth limit is 256 kbit/s. Thus, the side information of the reference model for a video sequence $v$ is stored in a file with size at most $256/8 * t_v$ kB, where $t_v$ is the duration of video $v$ in seconds. | Per media chunk | + +## 9 Model algorithm and output + +The video module defined in this Recommendation has two outputs, O.22 and O.27. It provides output values on the five-point ACR scale (MOS). + +## 10 Model description + +This clause describes the video model in detail. + +### 10.1 Notation + +Variable names are single or multiple characters of the form *frame\_height* or using subscripts like $n_{\text{resolution}}$ , to shorten formulas. For an array, $x$ , the notations $(x)$ and $(x_{ij})_{i=0..m-1, j=0..n-1}$ are used. The elements are denoted $x_{ij}$ or $x[i, j]$ , where array locators start at 0. The shorthand $x[i]$ is used for $x[i, \cdot]$ . + +A video sequence is given by a triple $(Y, Cb, Cr)$ of sequences of planes of $YCbCr$ pixel values, like in e.g., YUV420p, together with a sequence of display times $(dt)$ for each frame. The variables *frame\_height*, *frame\_width* refer to the *height*, *width* of the $Y$ plane, and $n_{\text{frame}}$ denotes the number of frames of the video sequence. + +Model constants and parameters are given in the lists at the end of the descriptions. Most clauses describe a set of computations, with input and output variables given at the start. In addition, the dimensions of the variables are given in the rightmost column. Some of these dimensions are given by model constants, e.g., $n_{\text{resolution}}$ , their values can be found in the list at the end. + +### 10.2 Multi-resolution frame pyramid + +INPUT: + +| | | +|-------------|------------------------------------------------------------------| +| Name | Description/dimension | +| $Y, Cb, Cr$ | Sequence of $Y, Cb, Cr$ planes, pixel values of the video signal | +| $dt$ | number of frames, $n_{\text{frame}}$ | + +OUTPUT: + +| | | +|-------------|-------------------------------------------------------------------------------------------------------| +| Name | Dimension | +| $(Y_{l,r})$ | $n_{\text{frame}} \times n_{\text{resolution}} \times \sum_i (frame\_height_i \times frame\_width_i)$ | + +All video frames are read and subsampled to a multi-resolution pyramid as follows: The $Y$ planes of frames are read and re-scaled by bicubic interpolation to a fixed resolution of $(f\_height\_init, f\_width\_init)$ and scaled to the range $[0, y\_range\_max]$ . As a reference ffmpeg version 3.2.2 can serve, see [b-FFmpeg]. To each frame a display time $dt$ is associated, e.g., for a video at a constant frame rate of 25 frames/s, the display time ( $dt$ ) is the constant vector of 40 ms for each frame. + +Frames are iteratively down-scaled by smoothing the frame data by convolution with filter kernel $f = [\frac{1}{4}, \frac{1}{2}, \frac{1}{4}]$ , keeping the same size of the matrix, computing the convolution by extending the border with constant values, and subsampling by 2 in either spatial direction, resulting for each frame $F_l$ in a frame multi-resolution pyramid of the $Y$ data given by + +$$(Y_{l,0}, Y_{l,1}, \dots, Y_{l,N-1}), \hfill (1)$$ + +with $N = n_{\text{resolution}}$ . The dimension of $Y_{l,r}$ is + +$$\dim(Y_{l,r}) = \frac{f\_height\_init}{2^{N-1-i}} \times \frac{f\_width\_init}{2^{N-1-i}} \hfill (2)$$ + +### 10.3 Frame edge representation + +INPUT: + +| | | +|-------------|-------------------------------------------------------------------------------------------------------| +| Name | Dimension | +| $(Y_{l,r})$ | $n_{\text{frame}} \times n_{\text{resolution}} \times \sum_i (frame\_height_i \times frame\_width_i)$ | +| $dt$ | $n_{\text{frame}}$ | + +OUTPUT: + +| | | +|------|-------------------------------------------------------------------------------------------------------| +| Name | Dimension | +| $Z$ | $n_{\text{frame}} \times n_{\text{resolution}} \times \sum_i (frame\_height_i \times frame\_width_i)$ | + +$$\phi \quad n_{\text{frame}} \times n_{\text{resolution}} \times \sum_i (\text{frame\_height}_i \times \text{frame\_width}_i)$$ + +Features based on edge orientation and strength will be computed for all $Y_{l,r}$ . To simplify notation in the following paragraph, the subscripts are omitted, and $Y$ signifies a $Y_{l,r}$ for some values of $l, r$ . Start by computing edge strength $R$ and edge orientation $\phi$ in the following way: the horizontal edges $H$ are computed as + +$$H[i, j] = \frac{2}{\pi} \cdot \arctan\left(\frac{1}{y_{\text{rescale}}}(Y[i, j] - Y[i - 1, j])\right) \quad (3)$$ + +and accordingly + +$$V[i, j] = \frac{2}{\pi} \cdot \arctan\left(\frac{1}{y_{\text{rescale}}}(Y[i, j] - Y[i, j - 1])\right) \quad (4)$$ + +where the locators $i, j$ run from 0 to $\dim(Y) - 1$ . A polar representation is computed composed of edge strength $R$ , + +$$R[i, j] = \sqrt{H[i, j]^2 + V[i, j]^2} \quad (5)$$ + +and angle $\phi$ as + +$$\phi[i, j] = \text{angle}(H[i, j], V[i, j]) \quad (6)$$ + +where the angle is measured in the counterclockwise direction, in the range $[0, 2\pi]$ . Further, a normalized edge strength with lateral inhibition is computed. + +Compute $c$ as the average of the constant $c_{\text{lat}}$ and the average of $R$ , + +$$c = \frac{1}{2} \cdot (c_{\text{lat}} + E(R)), \quad (7)$$ + +here $E$ is the average computed over all spatial position locators $i, j$ . + +For a position locator $(i, j)$ , a pair of positions $(i_0, j_0)$ and $(i_1, j_1)$ perpendicular to the edge are defined by + +$$i_0 = \frac{\Delta_{\text{lat}} \cdot H[i, j]}{\max(d_{\text{lat}}, R[i, j])} \quad (8)$$ + +$$j_0 = \frac{\Delta_{\text{lat}} \cdot V[i, j]}{\max(d_{\text{lat}}, R[i, j])} \quad (9)$$ + +and + +$$i_1 = -i_0, \quad j_1 = -j_0. \quad (10)$$ + +A normalization array $S$ by lateral inhibition is given by the edge strength at positions perpendicular to the edge + +$$S[i, j] = 0.5 * (Z[i + r(i_0), j + r(j_0)] + Z[i + r(i_1), j + r(j_1)]) \quad (11)$$ + +where the function $r$ rounds to the next integer value. The normalized edge strength $Z$ is computed as + +$$Z[i, j] = \frac{\max(0, R[i, j] - S[i, j])}{c + R[i, j] + S[i, j]} \quad (12)$$ + +### 10.4 Local edge statistic + +INPUT: + +Name Dimension + +$$Z \quad n_{\text{frame}} \times n_{\text{resolution}} \times \sum_i (\text{frame\_height}_i \times \text{frame\_width}_i)$$ + +$$\phi \quad n_{\text{frame}} \times n_{\text{resolution}} \times \sum_i (\text{frame\_height}_i \times \text{frame\_width}_i)$$ + +OUTPUT: + +Name Dimension + +$$s \quad n_{\text{frame}} \times n_{\text{resolution}} \times n_{\text{orient}} \times n_{s_h} \times n_{s_w}$$ + +A partition of unity is a family of positive continuous $[0,1]$ -valued functions $(\theta_k)_{k=0,\dots,L-1}$ for some integer $L$ having + +$$\sum_{k=0}^{L-1} \theta_k = 1. \quad (13)$$ + +A partition of unity on the unit circle is given by the functions $(\theta_k)_{k=0,\dots,L_o-1}$ mapping an angle $\phi$ in $[0,2\pi]$ to + +$$\theta_k(\phi) = 1 \quad \text{if} \quad \Delta(\phi, \alpha_k) < \beta \quad (14)$$ + +further + +$$\theta_k(\phi) = \frac{2\beta - \Delta(\phi, \alpha_k)}{\beta} \quad \text{if} \quad \beta < \Delta(\phi, \alpha_k) < 2\beta \quad (15)$$ + +and + +$$\theta_k(\phi) = 0 \quad \text{if} \quad \Delta(\phi, \alpha_k) \geq 2\beta \quad (16)$$ + +with $\Delta(\phi, \alpha_k) = \min(|\phi - \alpha_k|, 2\pi - |\phi - \alpha_k|)$ . Here $\alpha_k = k \cdot 2\pi/L_o$ and $\beta = 2\pi/(3L_o)$ . + +Let $v_\delta$ denote the continuous partial linear function given by + +$$v_\delta(x) = \frac{1}{2} \cdot \frac{x}{\delta} \quad 0 \leq x < \delta \quad (17)$$ + +$$v_\delta(x) = \frac{1}{2} \quad \delta \leq x < 2\delta \quad (18)$$ + +$$v_\delta(x) = \frac{1}{2} \left(3 - \frac{x}{\delta}\right) \quad 2\delta \leq x < 3\delta \quad (19)$$ + +$$v_\delta(x) = 0 \quad \text{otherwise} \quad (20)$$ + +A partition of unity $\Psi_{mn}$ in the spatial domain of the frame data $Y$ is defined by + +$$\Psi_{mn}[i, j] = v_\delta(i - m) * v_\epsilon(j - n) \quad (21)$$ + +with $\delta = \text{height}(Y)/(n_h + 2)$ and $\epsilon = \text{width}(Y)/(n_w + 2)$ . + +A family of patches $(P_{mnk})$ is computed using the partition of unity $\Psi_{mn}$ in the spatial domain and the partition $(\theta_k)$ in orientation. For orientation subscript $k$ , and locators $i, j$ a local patch $P$ is computed by + +$$P_{mnk}[i, j] = \Psi_{mn}[i, j] \cdot Z_{mn}[i, j] \cdot \theta_k(\phi_{mn}[i, j]) \quad (22)$$ + +The local patch $P_{mnk}$ is used to compute a patch statistic: Let $(z_{mnk}[i])_{i=0,\dots,N-1}$ denote the entries of the local patch $P_{mnk}$ sorted in ascending order, then a statistic $s$ is defined as the average over all values of $z_{mnk}$ above the quantile $q$ , + +$$s_{mnk} = s(z_{mnk}) = \frac{1}{N - q} \sum_{i=q}^{N-1} z_{mnk}[i] \quad (23)$$ + +with $q = \text{floor}((\text{width}(Y) - q_{\text{pos}})/\text{width}(Y))$ . The patch statistic $s_{mnk}$ computed based on the $l$ th frame $Y_{l,r}$ at resolution $r$ is denoted by $s_{mnk}^{l,r}$ if the reference to the frame number and resolution index is needed, otherwise, if clear from the context, the simpler notation $s_{mnk}$ is used, as in Formula (23). + +For each subsampled frame and each resolution, the local patch statistic $s = (s_{mnk})$ is computed. To reduce the number of local statistics, a border of 2 is skipped, and the subscripts $m, n$ are sampled by 2, e.g., sampled at positions 2,4,6,..., thus resulting in dimension of $\text{dim}(s) = \text{number}_{\text{frame}} \times \text{number}_{\text{resolution}} \times n_{\text{orient}} \times n_{s_h} \times n_{s_w}$ . + +To speed up computation, the local statistics are computed for a subset only. The patch statistic $s_{mnk}^{l,r}$ for frame $l$ at resolution $r$ is computed only if $r < \text{num}_{h\_res}$ or the following two conditions are fulfilled, otherwise it is set to zero. The first is that the following frame sampling condition is met, $(l \bmod \text{fps}) \bmod m = 0$ , where $\text{fps}$ is the video frame rate and $m$ is the largest integer less than or equal to $\text{fps}/\text{num}_{\text{high\_res\_per\_chunk}}$ . The second condition is that the local patch statistic at lower resolution is above the average overall patch statistics at lower resolution, + +$$s_{mnk}^{l,r-1} > E(s^{l,r-1}), \quad (24)$$ + +where the average $E$ is computed over all local patches in space, $m = 0, \dots, n_{s_h} - 1$ , $n = 0, \dots, n_{s_w} - 1$ , and orientation, $k = 0, \dots, n_{\text{orient}} - 1$ . + +### 10.5 Sharpness + +INPUT: + +| | | +|------|-----------| +| Name | Dimension | +|------|-----------| + +| | | +|-----|------------------------------------------------------------------------------------------------------------| +| $s$ | $n_{s_{\text{frame}}} \times n_{\text{resolution}} \times n_{\text{orient}} \times n_{s_h} \times n_{s_w}$ | +|-----|------------------------------------------------------------------------------------------------------------| + +OUTPUT: + +| | | +|------|-----------| +| Name | Dimension | +|------|-----------| + +| | | +|----------------|------------------------| +| $\text{sharp}$ | $n_{s_{\text{frame}}}$ | +|----------------|------------------------| + +At the highest resolution $r = \text{number\_resolution} - 1$ of the pyramid and for each frame $l$ with local edge statistic $(s_{mnk}^{l,r})_{mnk}$ , let $(h_i)_{i=0,\dots,N}$ denote the values of $(s_{mnk}^{l,r})_{mnk}$ in ascending order. Then sharpness $\text{sharp}$ is computed as the mean over the fraction of the largest $\text{sharp\_frac}$ values of $h$ , + +$$\text{sharp} = \frac{1}{N - i_0 + w_0} \cdot \left( w_0 \cdot h[i_0] + \sum_{i=i_0+1}^{N-1} h[i] \right) \quad (25)$$ + +Here $i_0 = \text{floor}(N \cdot \text{sharp\_frac})$ and $w_0 = 1 - (N \cdot \text{sharp\_frac} - i_0)$ . + +The computed sharpness value is scaled by multiplication with $\text{sharp\_scale\_fac}$ and stored as a 16-bit float. + +### 10.6 Features of reference (side information) + +INPUT: + +| | | +|------|-----------| +| Name | Dimension | +|------|-----------| + +| | | +|-------------|------------------------------------------------------------------| +| $Y, Cb, Cr$ | Sequence of $Y, Cb, Cr$ planes, pixel values of the video signal | +|-------------|------------------------------------------------------------------| + +| | | +|------|-------------------------| +| $dt$ | $n_{\text{ref\_frame}}$ | +|------|-------------------------| + +OUTPUT: + +| | | +|------|-----------| +| Name | Dimension | +|------|-----------| + +| | | +|------------------|--------------------------------------------------------------------------------------------| +| $s_{\text{ref}}$ | $n_{\text{ref\_frame}} \times n_{\text{orient}} \times n_{\text{sh}} \times n_{\text{sw}}$ | +|------------------|--------------------------------------------------------------------------------------------| + +| | | +|----------------|-------------------------| +| $\text{sharp}$ | $n_{\text{ref\_frame}}$ | +|----------------|-------------------------| + +Starting from the reference video sequence, the multi-resolution frame pyramid ( $Y_{l,r}$ ), the frame edge representation, the local edge statistic $s^{l,r}$ and sharpness $\text{sharp}$ are computed. + +Set $fs_{\text{step}} = 1$ , but for frame rates higher than 30 frames/s, set $fs_{\text{step}} = 2$ . The values of $s^{l,r}$ will be sampled by $fs_{\text{step}}$ , resulting in $n_{\text{ref\_frame}} = \text{floor}(n_{\text{ref\_frame}}/fs_{\text{step}})$ many frames with corresponding local statistics, in addition, the statistics will be kept at one resolution, only. In more detail, for each frame, the local orientation edge statistics $s$ at resolution $\text{dissim\_res}$ is sampled by $fs_{\text{step}}$ , to reduce the number of statistics to keep, thus + +$$s_{\text{ref}}[l, m, n, k] = s[l \cdot fs_{\text{step}}, \text{dissim\_res}, m, n, k] \quad (26)$$ + +with $i = 0, \dots, n_{\text{sh}} - 1, j = 0, \dots, n_{\text{sw}} - 1, k = 0, \dots, n_{\text{orient}} - 1$ . Note that for clarity the subscript $\text{ref}$ is used to denote the statistic of the reference video. + +The extracted reference features are composed of the local edge statistics $s_{\text{ref}}$ , compressed by multiplying the values by $\text{stat\_scale\_fac}$ , rounded to integer, clipped to $[0,255]$ and stored as a uint8, together with display\_time $dt$ and sharpness $\text{sharp}$ , where the latter two are stored as float16 values. + +### 10.7 Features of test video + +INPUT: + +| | | +|------|-----------| +| Name | Dimension | +|------|-----------| + +| | | +|-------------|------------------------------------------------------------------| +| $Y, Cb, Cr$ | Sequence of $Y, Cb, Cr$ planes, pixel values of the video signal | +|-------------|------------------------------------------------------------------| + +| | | +|------|--------------------| +| $dt$ | $n_{\text{frame}}$ | +|------|--------------------| + +OUTPUT: + +| | | +|------|-----------| +| Name | Dimension | +|------|-----------| + +| | | +|-----|--------------------------------------------------------------------------------------------------------------------| +| $s$ | $n_{\text{frame}} \times n_{\text{resolution}} \times n_{\text{orient}} \times n_{\text{sh}} \times n_{\text{sw}}$ | +|-----|--------------------------------------------------------------------------------------------------------------------| + +| | | +|---------------------|--------------------| +| $\text{rep\_frame}$ | $n_{\text{frame}}$ | +|---------------------|--------------------| + +| | | +|---------------|-------------------------------------------------------------------------| +| $y\_low\_res$ | $n_{\text{frame}} \times y\_low\_res\_height \times y\_low\_res\_width$ | +|---------------|-------------------------------------------------------------------------| + +| | | +|----------------|--------------------| +| $\text{sharp}$ | $n_{\text{frame}}$ | +|----------------|--------------------| + +Starting from the test video sequence, the multi-resolution frame pyramid ( $Y_{l,r}$ ), the frame edge representation, the local edge statistic $s_{mnk}^{l,r}$ and sharpness $\text{sharp}$ are computed. To speed up computation, the values of $s_{mnk}^{l,r}$ are computed for $l = 0, d, 2d, \dots$ for a frame sampling step $d$ defined by the test video frame rate as given in the following list: + +Step $d$ Video frames per second (fps) + +4 $\text{fps} > 30$ + +2 $20 < \text{fps} \leq 30$ + +1 $\leq 20$ + +Hence, $n_{s_{\text{frame}}} = \text{floor}(n_{\text{frame}}/d)$ . + +To match the features of the reference, the same compression is applied to the values of $s_{mnk}^{l,r}$ to be stored as uint8 values, and sharpness *sharp* as float16. + +A vector *rep\_frame* of indicators for repeated frames is computed as follows. If frame $i$ is a repetition of frame $i - 1$ , then $\text{rep\_frame}[i] = 1$ , otherwise $\text{rep\_frame}[i] = 0$ . + +For a sequence of matrices $(x_l)_{n=0,\dots,L-1}$ , each of dimension $M \times N$ , the spatial local mean $(z_l)$ is a sequence of dimension $L \times M_0 \times N_0$ given by + +$$z_l[p, q] = E(x_l[i, j]) \quad (27)$$ + +where the average $E$ is computed over all $i = \text{floor}(p * M/M_0), \dots, \text{floor}((p + 1) * M/M_0)$ , and $j = \text{floor}(q * N/N_0), \dots, \text{floor}((q + 1) * N/N_0)$ . Accordingly, starting from the $Y$ values at the lowest resolution in the pyramid, $Y_{l,0}$ , an array of local average values $y\_low\_res$ is computed using + +$$y\_low\_res_l[p, q] = E(Y_{l,0}), \quad (28)$$ + +with $p = 0, \dots, y\_low\_res\_height - 1$ , and $q = 0, \dots, y\_low\_res\_width - 1$ . + +### 10.8 Additional features + +INPUT: + +| Name | Dimension | +|----------------------------|----------------------------------------------------------------------------------------------------------------| +| $s$ | $n_{s_{\text{frame}}} \times n_{s_{\text{resolution}}} \times n_{\text{orient}} \times n_{s_h} \times n_{s_w}$ | +| $s_{\text{ref}}$ | $n_{s_{\text{ref\_frame}}} \times n_{\text{orient}} \times n_{s_h} \times n_{s_w}$ | +| $dt$ | $n_{s_{\text{frame}}}$ | +| $dt_{\text{ref}}$ | $n_{s_{\text{frame}}}$ | +| rep_frame | $n_{s_{\text{frame}}}$ | +| sharp | $n_{s_{\text{frame}}}$ | +| sharpref | $n_{s_{\text{ref\_frame}}}$ | + +OUTPUT: + +| Name | Dimension | +|-----------------------------|------------------------------------------------------| +| dissim | $n_{s_{\text{frame}}} \times n_{s_h} \times n_{s_w}$ | +| dissiminc | $n_{s_{\text{frame}}} \times n_{s_h} \times n_{s_w}$ | +| motionavg | $n_{s_{\text{frame}}}$ | +| fps | $n_{s_{\text{frame}}}$ | +| sharp | $n_{s_{\text{frame}}}$ | +| sharpref | $n_{s_{\text{frame}}}$ | + +First, a vector of indices $I_{ref} = (I_{ref,i})_i$ is determined, such that frame $i$ of the test video corresponds to frame $I_{ref,i}$ of the reference video. The procedure is done to possibly support frame rate reductions of the test video with respect to the reference. First, a matrix $A$ of root mean square error (RMSE) values is computed, + +$$A_{ij} = \text{RMSE}(s[i, \text{dissim\_res}], s_{ref}[j]). \quad (29)$$ + +Then, the index is determined as the best match close to a regression line estimated by robust regression. The details are given by the following pseudo code, the resulting index is $I_{ref} = \text{ref\_ind\_reg}(A)$ . + +``` +function arg_min_constraint(v,x,delta=2): + # Determine the position of the minimum within [x-delta,x+delta]. + # INPUT: v -- numpy 1-d array, x -- float, delta -- int. + # OUTPUT: i -- int, position of minimum under constraint 'close to x'. + i0 = clip( int(ceil(x-delta)), 0, len(v)-1) + i1 = clip( int(floor(x+delta)), 1, len(v)) + return i0 + argmin(v[i0:i1]) + +function ref_ind_reg(A): + # INPUT: A -- numpy array (2D) with dim m x n, of dissimilarity values. + # OUTPUT: ref_ind -- numpy array (1-dim) of length m. + + # determine a starting point + for i in [0,1,..,m-1]: + x[i] = arg_min_constraint(A[i,:], floor(i*n/m),delta=6) + r = ref_ind(A) + + # limit outliers using a Huber regression + TA = sklearn.linear_model.HuberRegressor() + TA.fit(x,r) + r_estim = TA.predict(x) + # take as ref ind values close by 'r_estim' + for i in [0,..,m-1]: + r_c[i] = arg_min_constraint(A[i,:],r_estim[i]) + + return r_c +``` + +For more details of the Huber robust regression, see [b-Scikit-learn], [b-Huber]. + +Let the function $D$ compute for two arrays $(a[k, i, j])$ , $(b[k, i, j])$ the average positive part of the difference, + +$$D(a, b) = \sum_k \max(0, a - b) / \sum_k \text{sign}(\max(0, a - b)) \quad (30)$$ + +where the max and sign functions are applied element-wise. Then, two arrays of dissimilarities $\text{dissim}$ , $\text{dissim}_{inc}$ are computed by + +$$\text{dissim}[i] = D(s[i, \text{dissim\_res}], s_{ref}[i]) \quad (31)$$ + +and + +$$\text{dissim}_{inc}[i] = D(s_{ref}[i], s[i, \text{dissim\_res}]) \quad (32)$$ + +An average motion feature is computed by + +$$\text{motion}_{avg}[i] = E(\text{abs}(s[i, \text{dissim\_res}] - s[\max(0, i - 1), \text{dissim\_res}])) \quad (33)$$ + +where abs denotes the absolute values and $E$ denotes the mean computed over all $n_{orient} \times n_{S_h} \times n_{S_w}$ array entries of $s$ . + +A windowed frame rate is computed as follows. The vector of frame display times is split into chunks of *fps\_chunk\_dur* seconds duration with chunk start and end given by $(start_i, end_i)$ , where the last chunk is allowed to be longer, if the video duration is not a multiple of *fps\_chunk\_dur*. A frame rate is computed for each chunk given by $(start_i, end_i)$ by computing + +$$fps[i] = \frac{1000}{E(dt_{no\_rep})} \quad i \text{ in } [start_i, \dots, end_i] \quad (34)$$ + +where the function *E* computes the average, $dt_{no\_rep}$ is the display time in milliseconds of each frame in the chunk ignoring repeated frames, i.e., + +$$dt_{no\_rep}[j] = dt[j_0] + \dots + dt[j_m] \quad (35)$$ + +for $j_0$ being the *j*th non-repeated frame, and there are *m* consecutive repeated frames following frame $j_0$ . A non-repeated frame is a frame *l* having *rep\_frame*[*l*] = 0. The frame rate vector is resampled to match the size of *dissim* using the step function resampling approach described in clause 10.9. + +The sharpness vector *sharp* and *sharp\_ref* contain zeros reflecting the skipped computation of the corresponding local edge statistic. These missing values are estimated by applying the function *estimate\_missing\_by\_local\_average* given by the following pseudo-code. + +``` +function estimate_missing_by_local_average(x): + # Input a 1-dim array of positive values, with some zero values + # corresponding to skipped computations + # Return a 1-dim array with zero entries replaced by local averages. + + m = num_a + i_non_z = non_zero(x) # return indices on non-zero elements of x + n = int(ceil(len(x)/len(i_non_z))) + z = zeros(len(x)) + for i in range(len(z)): + j0 = max(0, i - m*n) + j1 = min(j0+2*m*n, len(x)) + j0 = max(0, j1-2*m*n) + x_local = x[j0:j1] + j_non_z = non_missing(x_local)[0] + z[i] = np.mean(x_local[j_non_z]) + if len(z) < num_s: + z[i] = ones(len(x))*mean(z) + return z +``` + +In addition, the sharpness vector *sharp\_ref* of the reference video is resampled to the size *ns\_frame* of vector *sharp* using the step function resampling approach described in clause 10.9. + +### 10.9 Step functions + +For a vector $(v_i)_{i=0, \dots, N-1}$ and a monotonically increasing vector $(t_i)_{i=0, \dots, N}$ let $S_{v,t}: \mathbb{R} \rightarrow \mathbb{R}$ denote a step function, a real-valued function taking the constant value $v_i$ over the intervall $[t_i, t_{i+1}]$ and zero otherwise. Let $Avg(S_{v,t}, t_0, t_1)$ denote the average value of the step function $S_{v,t}$ over the interval $[t_0, t_1]$ . + +Given an additional monotonically increasing vector $(t'_i)_{i=0, \dots, N'}$ , the step function $S_{v,t}$ can be resampled to the interval limits of $(t'_i)$ to the step function $S_{v',t'}$ in a natural way by integrating the function $S_{v,t}$ over the intervals defined by $(t'_i)$ , explicitly + +$$v'_i = \frac{1}{t'_{i+1} - t'_i} \int_{t'_i}^{t'_{i+1}} S_{v,t}(x) dx = Avg(S_{v,t}, t'_i, t'_{i+1}) \quad (36)$$ + +### 10.10 Transformations + +Let $S: \mathbb{R} \rightarrow [0,1]$ be a non-linear transformation composed of two parts, and parameterized by $p_x$ the joining position of the two parts and its mapped-value $S(p_x) = p_y$ and slope $p_q$ at $p_x$ . The first part is given by + +$$S(x) = a \cdot x^b \quad \text{if } x \leq p_x \quad (37)$$ + +and the second part by an exponential saturation of the form + +$$S(x) = 2 \cdot d \cdot \left( \frac{1}{1 + \exp(-c \cdot (x[I] - p_x))} - \frac{1}{2} \right) + p_y, \quad x > p_x \quad (38)$$ + +with + +$$a = p_y/p_x^b, \quad b = p_x * p_q/p_y \quad c = 2 * p_q/d \quad d = 1 - p_y. \quad (39)$$ + +For most of the parameters that are used in the following, the transformation has an S-shape, and will therefore be called an S-shaped transformation. The list in clause 10.11 summarizes the transformations that will be used together with their parameters. + +### 10.11 Core model + +INPUT: + +| Name | Dimension | +|-----------------------|----------------------------------------------------------------------------------------------------| +| $s$ | $nS_{\text{frame}} \times nS_{\text{resolution}} \times n_{\text{orient}} \times nS_h \times nS_w$ | +| $dissim$ | $nS_{\text{frame}} \times nS_h \times nS_w$ | +| $dissim_{\text{inc}}$ | $nS_{\text{frame}} \times nS_h \times nS_w$ | +| $motion_{\text{avg}}$ | $nS_{\text{frame}}$ | +| $sharp$ | $nS_{\text{frame}}$ | +| $sharp_{\text{ref}}$ | $nS_{\text{frame}}$ | +| $fps$ | $nS_{\text{frame}}$ | + +OUTPUT: + +| Name | Dimension | +|--------------------|------------------------------| +| $q_{\text{frame}}$ | $nS_{\text{frame}}$ | +| $q_{\text{sec}}$ | video_duration_in_sec | +| $q$ | 1 | +| 0.22 | 1 | +| 0.27 | 1 | + +A weighting factor is computed by first computing $s_{\text{max}}$ as the maximum over all orientations of $s$ , + +$$s_{\text{max}}[n, i, j] = \max_k s[n, \text{dissim}_{\text{res}}, k, i, j], \quad (40)$$ + +and then computing the weighting factor $w_s$ as + +$$w_s[n, i, j] = \left( \frac{1}{\max(0, c_{\text{lim}} - c_{\text{scale}} \cdot s_{\text{max}}[n, i, j])} \right)^{c_{\text{exp}}} \quad (41)$$ + +and an additional border weighting is applied using + +$$w[n, i, j] = \frac{d_{border}[i, j] + 1}{w_{max\_border\_dist}} \cdot w_s[n, i, j] \quad (42)$$ + +where $d_{border}$ is the distance to the border, capped at $w_{max\_border\_dist} - 1$ , i.e., + +$$d_{border}[i, j] = \min(i, j, (ns_h - 1 - i), (ns_w - 1 - j), (w_{max\_border\_dist} - 1)). \quad (43)$$ + +The weighting is applied to the dissimilarity values by + +$$dissim_w[n] = dissim[n] \cdot \frac{w[n]}{E(w[n])} \quad (44)$$ + +$$dissim_{inc\_w}[n] = dissim_{inc}[n] \cdot \frac{w[n]}{E(w[n])}, \quad (45)$$ + +where the function $E$ computes the average of $w[n, i, j]$ over all locators $i = 0..ns_h - 1, j = 0..ns_w - 1$ . + +A frame rate degradation is computed by + +$$d_{fps} = (1 - S_{fps}(fps)) \cdot (1 - \exp(E(motion\_avg)/par_{motion\_fps})) \quad (46)$$ + +A motion weighting is computed by + +$$mo_{avg\_weight} = 1 - par_{motion\_c} \cdot S_{mo}(motion\_avg) \quad (47)$$ + +Further, an additional correction is computed by + +$$dissim_{cor} = L(dissim_w) \quad (48)$$ + +$$dissim_{inc\_cor} = L(dissim_{inc}), \quad (49)$$ + +where + +$$L(s)[n, i, j] = s[n, i, j] \cdot (1 + par_{lum\_fac} * (1 + y_{lowres}[n, p, q])^{par_{lum\_exp}}) \quad (50)$$ + +with $p = floor(i * y_{lowres\_height}/ns_h)$ and $q = floor(j * y_{lowres\_width}/ns_w)$ . + +Further, the values are mapped to range [0,1] by + +$$d_{dis}[n, i, j] = mo_{avg\_weight}[n] \cdot S_{dis}(dissim_{corr}[n, i, j]) \quad (51)$$ + +$$d_{dis\_inc}[n, i, j] = mo_{avg\_weight}[n] \cdot S_{dis\_inc}(dissim_{inc\_corr}[n, i, j]) \quad (52)$$ + +A degradation due to loss in sharpness $d_{sharp}$ is computed by + +$$d_{sharp} = 1 - S_{rel\_sharp} \left( \min(1.0, \frac{sharp + c_{sharp}}{sharp\_ref + c_{sharp}}) \right) \quad (53)$$ + +using the S-transformation $S_{rel\_sharp}$ , on the other hand, a degradation related to an increase in sharpness is computed by + +$$d_{sharp\_inc} = S_{sharp\_inc}(\max(0.0, sharp - sharp\_ref)) \quad (54)$$ + +using the S-transformation $S_{sharp\_inc}$ . For a degradation factor, higher values are worse. For a quality factor, higher values are good. Degradation factors in the range [0,1] can be inverted to quality factors by subtraction from one. Hence, a quality estimate is computed by a product of inverted degradations, + +$$q_{frame_l} = (1 - d_{sharp}) \cdot (1 - d_{sharp\_inc}) \cdot (1 - d_{fps}) \cdot E_{spat}((1 - d_{dis}) \cdot (1 - d_{dis\_inc})), \quad (55)$$ + +where $E_{spat}$ computes the spatial mean, in more detail, for the sequence of matrices the mean for each sequence element is computed. To model temporal aspects of degradations the following is computed: + +$$q_{frame} = 1 - \text{deg\_fade\_out}((1 - q_{frame_i}), dt) \quad (56)$$ + +where the function $\text{deg\_fade\_out}$ is described in the next paragraph using pseudo code. + +For a vector $(v_i)_{i=0,\dots,N-1}$ and a monotonically increasing vector $(t_i)_{i=0,\dots,N}$ let $S_{v,t}$ denote a step function, a real-valued function taking the constant value $v_i$ over the interval $[t_i, t_{i+1}]$ . Let $\text{Avg}(S_{v,t}, t_0, t_1)$ denote the average value of the step function $S_{v,t}$ over the interval $[t_0, t_1]$ . Then the function $\text{deg\_fade\_out}$ is defined by the following pseudo code: + +``` +function deg_fade_out(v,t) + +w = [0,0,...,0] +for i in [1,2,...]: + a = exp(-par_fade_dt) + v_avg = Avg(S_{v,t}, t[i+1]-par_fade_smooth, t[i+1]) + w[i] = max(v_avg, a * w[i-1] + (1-a) * v_avg) + +return w +``` + +The values of the parameters $\text{par\_fade\_dt}$ , $\text{par\_fade\_smooth}$ are specified in clauses 10.14 and 10.13, respectively.. + +The temporal average quality $q$ for the entire sequence is given by + +$$q = E(q_{frame}) \quad (57)$$ + +Furthermore, a per-second score is computed using the step function resampling approach in clause 10.9, resulting in $q_{sec}$ . These $[0,1]$ -valued scores $q$ , $q_{frame}$ , $q_{sec}$ are mapped to the range $[1,5]$ by the rescale function $r: x \mapsto 4x + 1$ resulting in + +$$0.27 = r(q), \quad 0.22 = r(q_{sec}) \quad (58)$$ + +Higher values of $q$ , $0.22$ , $0.27$ mean higher quality. + +### 10.12 Viewing distance + +For a relative viewing distance above $4H$ , the parameters for the mobile case are used. For a relative viewing distance below $2H$ , the parameters for the PC/TV case are used. For intermediate viewing distances, the parameters are linearly interpolated between the two. + +### 10.13 Parameters + +| Name | Value | +|-------------------|-------| +| $f\_width\_init$ | 1 920 | +| $f\_height\_init$ | 1 080 | +| $y\_range\_max$ | 255 | +| $y\_rescale$ | 20 | +| $n_{resolution}$ | 4 | +| $n_h$ | 18 | +| $n_w$ | 32 | +| $num_{h\_res}$ | 2 | +| $c_{lat}$ | 0.3 | + +$d_{lat}$ 0.001 + $\Delta_{lat}$ 2 + $L_o$ 8 + $q_{pos}$ 2 +sharp\_scale\_fac 10.0 +stat\_scale\_fac 4\*255 +y\_low\_res\_width 5 +y\_low\_res\_height 3 +c\_sharp 0.05 +w\_max\_border\_dist 3 +dissim\_res 1 +fps\_chunk\_dur 2 + $n_{orient}$ 8 + $n_{S_h}$ 7 + $n_{S_w}$ 14 +sharp\_frac 0.05 +num\_a 3 +num\_s 200 +par\_weight\_scale 100 +par\_fade\_smooth 0.5 + +### 10.14 Device-dependent model parameters + +| Name | Value PC/TV | Value MO/TA | +|---------------------|---------------------|---------------------| +| par_weight_lim | 5.593268792046344 | 4.656208421713784 | +| par_weight_exp | 0.9985031497295792 | 0.9999821534030532 | +| $par_{motion\_fps}$ | 0.10338749688116727 | 0.1000006225291463 | +| $par_{motion\_c}$ | 0.9683245820065315 | 0.7604347879732595 | +| $par_{lum\_fac}$ | 0.5573475746950503 | 0.5574799921101337 | +| $par_{lum\_exp}$ | 0.10014977581205474 | 0.10412368985745854 | +| $par\_fade\_dt$ | 0.1616170238997139 | 0.1871980057940932 | + +### 10.15 S-transformation parameters + +| Transform PC/TV | x-Position | y-Position | Slope | +|------------------|---------------------|--------------------|---------------------| +| $S_{mo}$ | 1.0464757777038356 | 0.5 | 0.47124514999456596 | +| $S_{dis}$ | 0.5450173005392799 | 0.7980273056330967 | 2.048041212706822 | +| $S_{dis\_inc}$ | 0.36420555146972666 | 0.6165825542863502 | 2.235668875917247 | +| $S_{rel\_sharp}$ | 0.6745913663781392 | 0.5 | 2.177200231342128 | + +| | | | | +|------------------|---------------------|--------------------|----------------------| +| $S_{sharp\_inc}$ | 0.289504984526356 | 0.5 | 2.028729717455461 | +| $S_{fps}$ | 15.0 | 0.7500024932923486 | 0.01805843377341594 | +| Transform Mobile | x -Position | y -Position | Slope | +| $S_{mo}$ | 1.2972708989704074 | 0.5 | 0.1882251589297096 | +| $S_{dis}$ | 0.7211019847289146 | 0.6830850971844077 | 2.3914975476194362 | +| $S_{dis\_inc}$ | 0.4041098766701082 | 0.5404927853257431 | 1.3109987046856608 | +| $S_{rel\_sharp}$ | 0.28071248315138375 | 0.5 | 0.9889249368712523 | +| $S_{sharp\_inc}$ | 0.6740897012131203 | 0.5 | 2.9946362074534 | +| $S_{fps}$ | 15.0 | 0.7665500949169916 | 0.021999942089236887 | + +## Appendix I + +## Performance figures + +(This appendix does not form an integral part of this Recommendation.) + +In this appendix, RMSE is reported, for the model used stand-alone. Note that the numbers are reported after a final per-database mapping between the model output and the subjective scores of a database. This linear mapping is used to account for scale and bias variations between different databases. + +**Table I.1 – Validation performance of model used stand-alone: The *submitted model* is the model trained on the exchanged training databases and frozen before creation of validation data. The model was retrained using a fivefold cross-validation approach, with the validation performance listed to show stability of the performance indicating no over-fitting** + +| | | | | | | | +|-----------------------|-----------------------------------|-------|-------|-------|-------|-------| +| Pixel-based reference | Submitted model | 0.444 | | | | | +| | Five-fold cross-validation | 0.418 | 0.418 | 0.425 | 0.442 | 0.429 | + +The re-training of the submitted model was performed on five different splits. The splits were defined on the database level. The following is the procedure that was followed to determine the splits. + +- All training and validation databases were merged to obtain in total 26 different short databases (18 PC/TV and 8 MO/TA). +- A level of difficulty of prediction for each database was determined based on average prediction error over all models. +- A 50-50 training-validation split was determined randomly but respecting the level of difficulty. In total, five different splits were defined. Each split had a balanced distribution of databases based on difficulty in both the training and validation. +- The 50-50 split was separately performed for PC/TV and MO/TA cases. +- The final model coefficients correspond to the best performing split. + +**Table I.2 – Validation performance of model used stand-alone: Performance of P.1204.4 on the validation databases containing AV1 encoded videos, only. Listed in the table are RMSE over all 6 databases, per-database minimum and maximum RMSE** + +| Model | Average RMSE | Min RMSE | Max RMSE | +|----------|--------------|----------|----------| +| P.1204.4 | 0.362 | 0.269 | 0.425 | + +## Appendix II + +## Long term integration module (Pq) for ITU-T P.1204.4 + +(This appendix does not form an integral part of this Recommendation.) + +The current approach for long-term integration described in this appendix is to be considered as an intermediate solution. It has been developed based on a total of 6 databases (2 for training and 4 for validation), while the short-term video quality model (Pv) described in the normative part of this Recommendation was developed using 26 databases (13 for training and 13 for validation). + +As a consequence of the relatively low number of long-sequence databases used, the integration module presented in this appendix does not form an integral part of the standard and is considered for information. It is planned to be superseded in the future by a more comprehensive integration module trained and validated on a higher number of databases, which is currently under development. + +The streaming parameters ranges used in the training and validation tests of the long-term integration model presented in this appendix are summarized in Table II.1. + +**Table II.1 – Parameters range used in the tests for the development ITU-T P.1204.4 long-term integration model** + +| | | +|----------------------------------------|------------------------| +| Video sequence duration | 60 seconds – 5 minutes | +| Initial loading delay | 0-30 seconds | +| Total stalling duration | 0-26 seconds | +| Number of stalling events | 0-5 | +| Total number of quality level switches | 0-39 | + +The ranges of parameter settings of other test factors related to each of the segments in a video session are summarized in Table 3 of this Recommendation. It should be noted that audio quality was not varied in any of the 6 tests used to train and validate the long-term integration model presented in this appendix. For the PC/TV case, the audio from the source video was encoded using 16-bit PCM with 2 channels and 48 kHz sampling frequency. For the case of mobile/tablet, an advanced audio coding (AAC) codec was used for audio with a bit rate of 512 kbit/s. + +### II.1 Model input + +The model must receive the following input signals regardless of the mode of operation: + +- O.21: audio quality per output sampling interval, as specified in [b-ITU-T P.1203.2] +- O.22: video quality per output sampling interval, as specified in this Recommendation – see clause 9. +- I.14: stalling events, as described in clause 7.1 of [b-ITU-T P.1204]. +- I.GEN: device type (either of "PC", "TV", "Mobile" or "Tablet"), as specified in [b-ITU-T P.1204]. + +### II.2 Model output + +The Pq model outputs the following information: + +- O.23: perceptual buffering indication +- O.34: audiovisual segment coding quality per output sampling interval +- O.35: final audiovisual coding quality score + +– 0.46: final media session quality score + +The values 0.23, 0.35 and 0.46 will be output once per session. + +The value 0.34 will be output once per output sampling interval. + +### II.3 Model description + +#### II.3.1 Notation + +All the variables related to time are expressed in seconds. + +"Per output sampling interval" scores are expressed as vectors: + +$$x = [x_0, x_1, \dots, x_N]$$ + +$L+1$ is the total duration of the video sequence. + +Model constants and parameters are given in the list at the end of the descriptions. + +#### II.3.2 Audio-video segment quality + +First, the inputs audio (O.21) and video (O.22) coding quality vectors mapped to the $[0, 1]$ -scale by the rescale function $r': x \mapsto (x - 1)/4$ , resulting in a per-second video quality $v_j$ and audio quality, $a_j$ . The audio-video quality $q_{avj}$ per each second $j$ is given by the minimum of video quality $v_j$ and audio quality, $a_j$ , + +$$q_{avj} = \min(v_j, a_j) \quad (\text{II.1})$$ + +#### II.3.3 Integrated audio-video coding quality + +The integrated audio-video coding quality is expressed as: + +$$q_{avg} = \sum_{j=0}^L q_{avj} \cdot \frac{av_{recj}}{W} \quad (\text{II.2})$$ + +where the weighted average is computed using the recency weight $av_{recj}$ at the time $j$ and the normalization constant is given by $W = \sum_{j=0}^L av_{recj}$ . + +#### II.3.4 Recency weight + +The recency weight at sampling time $j$ is given by: + +$$av_{rec}(j) = e^{-\frac{L-j}{k_{rec}L}} \quad (\text{II.3})$$ + +for $j = 0, 1, \dots, L-1$ . + +The recency weight for stalling is defined as: + +$$b_{rec}(j) = \begin{cases} rec\_lin(j) & \text{for } j \in [1, k_{lin}[ \\ av_{rec}(j) & \text{for } j \in [k_{lin}, L] \end{cases} \quad (\text{II.4})$$ + +with: + +$$rec\_lin(j) = \frac{av_{rec}(k_{lin}) - 0.1}{k_{lin}} \cdot j + 0.1 \quad (\text{II.5})$$ + +#### II.3.5 Final quality score + +The final quality score $q$ for a video sequence is determined by two factors: the integrated audio-video coding quality $q_{avg}$ and a factor determined by the stalling events impact $w_b$ . + +$$q = q_{avg} \cdot (1 - w_b) \quad (\text{II.6})$$ + +The impact $w_b$ is the total impact of stalling events, defined as the summation of the impact of each stalling event: + +$$w_b = \sum_{i=0}^{N_{buff}-1} w_{b_i} \quad (\text{II.7})$$ + +with $N_{buff}$ the number of stalling events and the individual stalling impacts $w_{b_i}$ are described in the next paragraph. + +#### II.3.6 Stalling impact + +The $i$ -th stalling event $st_i$ is given by a tuple $(s_i, l_i)$ where $s_i$ and $l_i$ are respectively the start time relative to the start of the original video clip and the duration of the stalling event. + +Given two consecutive stalling events $st_i$ and $st_{i+1}$ , the temporal distance between the two stalling events is defined as $\Delta_{i+1} = s_{i+1} - s_i$ and $\Delta_0 = 0$ . + +The impact of stalling events depends on their temporal distribution. First, a pair of stalling events $st_i$ and $st_{i+1}$ with $\Delta_{i+1} \leq \Delta_{min}$ is replaced by one stalling event $st_i = (s_i, l_i + l_{i+1})$ . + +Further, the stalling impact is given by a sum of two weighted relative stalling durations, + +$$w_{b_i} = \left( k_0 \cdot \frac{l_i}{L} + f(\Delta_i) \frac{\min(l_i, l_{max})}{L} \right) b_{rec}(s_i) \quad (\text{II.8})$$ + +where the weights depend on the recency weight, see equation (II.4), a constant $k_0$ and the function $f(\Delta_i)$ given by + +$$f(\Delta_i) = \frac{x^{\alpha-1}}{\Gamma(\theta) \cdot \theta^\alpha} \cdot \exp\left(-\frac{x}{\theta}\right) \quad (\text{II.9})$$ + +with $x = \Delta_i - \Delta_{min}$ for $\Delta_i \in [\Delta_{min}, \Delta_{max}]$ and zero otherwise. + +#### II.3.7 Multiple short stalling events + +In case of multiple stallings, and where some of the stalling durations are very short, the following modification needs to be applied: + +If $l_i < l_{short}$ in equation (II.8) the value $l_i$ is replace by $l'_i$ given by + +$$l'_i = \min(l_i + g(i) \cdot w(l_i), l_{short}) \quad (\text{II.10})$$ + +with + +$$g(i) = \begin{cases} 0 & \text{for } i = 0 \\ \frac{l_{short}}{2} & \text{for } i = 1 \\ \frac{l_{short}}{i} & \text{else} \end{cases} \quad (\text{II.11})$$ + +$$w(x) = \begin{cases} \frac{x}{2} & \text{for } x \leq \frac{l_{short}}{10} \\ -\frac{2x-10}{3} & \text{for } x \geq \frac{7}{10} l_{short} \\ 1 & \text{else} \end{cases} \quad (\text{II.12})$$ + +#### II.3.8 Very long stalling + +In case of a very long stalling, the following modification needs to be applied: + +If $l_i > \frac{1}{3}L$ the stalling's weight is + +$$w_{b_i} = (k_0 + f(\Delta_i)) \cdot \frac{k_{vts} \cdot l_i}{L} \cdot b_{rec}(s_i) \quad (\text{II.13})$$ + +#### II.3.9 Output + +For the final output the $[0,1]$ -valued scores $q, q_{av}, q_{avg}$ are mapped to the range $[1,5]$ by the rescale function $r: x \mapsto 4x + 1$ resulting in + +$$0.34 = r(q_{av}) \quad (\text{II.14})$$ + +$$0.35 = r(q_{avg}) \quad (\text{II.15})$$ + +$$0.46 = r(q) \quad (\text{II.16})$$ + +$$0.23 = 0.35 - 0.46 \quad (\text{II.17})$$ + +#### II.3.10 Parameters + +Table II.2 lists the model parameters: + +**Table II.2 – Model parameters** + +| | | +|----------------|----------| +| $k_0$ | 1.51293 | +| $k_{rec}$ | 0.731715 | +| $\alpha$ | 4.41 | +| $\theta$ | 2.21 | +| $\Delta_{min}$ | 2 s | +| $\Delta_{max}$ | 30 s | +| $l_{max}$ | 5 s | +| $l_{short}$ | 5 s | +| $k_{lin}$ | 5 s | +| $k_{vls}$ | 1.3 | + +### II.4 Performance figures + +In this clause, the aggregated RMSE of the model is presented. Aggregated RMSE is defined as: + +$$RMSE = \frac{1}{W} \sum_{k=1}^M w_k \cdot RMSE_k$$ + +Where $M$ represents the total number of (training and validation) databases, $w_k$ and $RMSE_k$ the weight and root mean square error for database $k$ respectively, and $W = \sum_{k=1}^M w_k$ . Training and validation databases have different weights: + +$$w_{training} = 0.1$$ + +$$w_{validation} = 0.9$$ + +Note that the numbers are reported after a final per-database mapping between the model output and the subjective scores of a database. This linear mapping is used to account for scale and bias variations between different databases. + +Tables II.3 and II.4 show the performance of the Pq, based respectively on the submitted version and standardized versions of ITU-T P.1204.4. + +**Table II.3 – Performance of the Pq described above, based on the submitted version of ITU-T P.1204.4** + +| RMSE | Training | | Validation | | +|-------|-------------------------|-----------------------|-------------------------|-----------------------| +| | Nr. of databases | Nr. of Samples | Nr. of databases | Nr. of Samples | +| 0.480 | 2 | 82 | 4 | 134 | + +**Table II.4 – Performance of the Pq described above, based on the standardized ITU-T P.1204.4** + +| RMSE | Training | | Validation | | +|-------|-------------------------|-----------------------|-------------------------|-----------------------| +| | Nr. of databases | Nr. of Samples | Nr. of databases | Nr. of Samples | +| 0.469 | 2 | 82 | 4 | 134 | + +## Bibliography + +- [b-ITU-T G.1022] Recommendation ITU-T G.1022 (2016), *Buffer models for media streams on TCP transport*. +- [b-ITU-T P.800.1] Recommendation ITU-T P.800.1 (2016), *Mean opinion score (MOS) terminology*. +- [b-ITU-T P.911] Recommendation ITU-T P.911 (1998), *Subjective audiovisual quality assessment methods for multimedia applications*. +- [b-ITU-T P.1201.1] Recommendation ITU-T P.1201.1 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – Lower resolution application area*. +- [b-ITU-T P.1201.2] Recommendation ITU-T P.1201.2 (2012), *Parametric non-intrusive assessment of audiovisual media streaming quality – Higher resolution application area*. +- [b-ITU-T P.1202] Recommendation ITU-T P.1202 (2012), *Parametric non-intrusive bitstream assessment of video media streaming quality*. +- [b-ITU-T P.1202.1] Recommendation ITU-T P.1202.1 (2012), *Parametric non-intrusive bitstream assessment of video media streaming quality – Lower resolution application area*. +- [b-ITU-T P.1203] Recommendation ITU-T P.1203 (2017), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport*. +- [b-ITU-T P.1203.2] Recommendation ITU-T P.1203.2 (2017), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Audio quality estimation module*. +- [b-ITU-T P.1204] Recommendation ITU-T P.1204 (2020), *Video quality assessment of streaming services over reliable transport for resolutions up to 4K*. +- [b-ITU-T P.1204.3] Recommendation ITU-T P.1204.3 (2020), *Video quality assessment of streaming services over reliable transport for resolutions up to 4K with access to full bitstream information*. +- [b-ITU-T P.1204.5] Recommendation ITU-T P.1204.5 (2020), *Video quality assessment of streaming services over reliable transport for resolutions up to 4K with access to transport and received pixel information*. +- [b-ITU-T P.1401] Recommendation ITU-T P.1401 (2020), *Methods, metrics and procedures for statistical evaluation, qualification and comparison of objective quality prediction models*. +- [b-FFmpeg] FFmpeg Developers (Internet). *FFmpeg: A complete, cross-platform solution to record, convert and stream audio and video*, version 4.2, "Ada" [Software]. Available [viewed 2020-02-26] from: +- [b-Huber] Scikit-learn developers (2007-19). Huber regression. In: *Scikit-learn*, version 0.22.1 [Software]. Available [viewed 2020-02-26] from: [https://scikit-learn.org/stable/modules/linear\\_model.html#huber-regression](https://scikit-learn.org/stable/modules/linear_model.html#huber-regression) +- [b-Scikit-learn] Scikit-learn developers (2007-19). *Scikit-learn*, version 0.22.1 [Software]. Available [viewed 2020-02-26] from: [https://scikit-learn.org/stable/whats\\_new/v0.19.html#version-0-19-1](https://scikit-learn.org/stable/whats_new/v0.19.html#version-0-19-1) + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | Tariff and accounting principles and international telecommunication/ICT economic and policy issues | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling, and associated measurements and tests | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects, next-generation networks, Internet of Things and smart cities | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.1211-202310-I_PDF-E/raw.md b/marked/P/T-REC-P.1211-202310-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..1b554d0c8ba5956630ff7e6639d2f6f5a9201461 --- /dev/null +++ b/marked/P/T-REC-P.1211-202310-I_PDF-E/raw.md @@ -0,0 +1,742 @@ + + +# Recommendation + +# **ITU-T P.1211 (10/2023)** + +SERIES P: Telephone transmission quality, telephone installations, local line networks + +Models and tools for quality assessment of streamed media + +--- + +# **Derivation procedure for contribution values to quality degradation of adaptive audiovisual streaming services** + +![ITU logo](84a1d09fb489061482111515543b60dc_img.jpg) + +The logo of the International Telecommunication Union (ITU) is located in the bottom right corner. It features a blue globe with white lines representing latitude and longitude, and the letters 'ITU' in a bold, blue, sans-serif font superimposed on the globe. + +ITU logo + +## ITU-T P-SERIES RECOMMENDATIONS + +### **Telephone transmission quality, telephone installations, local line networks** + +| | | +|----------------------------------------------------------------------------------------------------|----------------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | P.10-P.19 | +| Voice terminal characteristics | P.30-P.39 | +| Reference systems | P.40-P.49 | +| Objective measuring apparatus | P.50-P.59 | +| Objective electro-acoustical measurements | P.60-P.69 | +| Measurements related to speech loudness | P.70-P.79 | +| Methods for objective and subjective assessment of speech quality | P.80-P.89 | +| Voice terminal characteristics | P.300-P.399 | +| Objective measuring apparatus | P.500-P.599 | +| Measurements related to speech loudness | P.700-P.709 | +| Methods for objective and subjective assessment of speech and video quality | P.800-P.899 | +| Audiovisual quality in multimedia services | P.900-P.999 | +| Transmission performance and QoS aspects of IP end-points | P.1000-P.1099 | +| Communications involving vehicles | P.1100-P.1199 | +| Models and tools for quality assessment of streamed media | P.1200-P.1299 | +| Telemeeting assessment | P.1300-P.1399 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | P.1400-P.1499 | +| Methods for objective and subjective assessment of quality of services other than speech and video | P.1500-P.1599 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +# Recommendation ITU-T P.1211 + +## Derivation procedure for contribution values to quality degradation of adaptive audiovisual streaming services + +## Summary + +Recommendation ITU-T P.1211 describes a derivation procedure for contribution values that provide information about a relationship between quality-influencing factors (e.g., media quality levels and stalling) and the final media session quality score of adaptive audiovisual streaming services. More precisely, contribution values show how much each quality-influencing factor affects (i.e., decreases) the final media session quality score provided by a quality estimation model (e.g., that of ITU-T P.1203 or ITU-T P.1204) using Shapley theory, which is a key concept in cooperative game theory for distributing profits in a game among participants. This derivation procedure can be used to monitor the streaming service since operators can ascertain which quality-influencing factors decrease the final media session quality score. + +## History \* + +| Edition | Recommendation | Approval | Study Group | Unique ID | +|---------|----------------|------------|-------------|--------------------| +| 1.0 | ITU-T P.1211 | 2023-10-29 | 12 | 11.1002/1000/15695 | + +## Keywords + +Adaptive streaming, audiovisual, audiovisual quality estimation, contribution value, mean opinion score (MOS), monitoring, multimedia, QoE, Shapley theory. + +--- + +\* To access the Recommendation, type the URL in the address field of your web browser, followed by the Recommendation's unique ID. + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had received notice of intellectual property, protected by patents/software copyrights, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the appropriate ITU-T databases available via the ITU-T website at . + +© ITU 2023 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## Table of Contents + +| | Page | +|---------------------------------------------------------------------------------------------------------------------------------------|------| +| 1 Scope ..... | 1 | +| 2 References..... | 2 | +| 3 Definitions ..... | 2 | +| 3.1 Terms defined elsewhere ..... | 2 | +| 3.2 Terms defined in this Recommendation..... | 3 | +| 4 Abbreviations and acronyms ..... | 3 | +| 5 Conventions ..... | 3 | +| 6 Field of application ..... | 4 | +| 7 Building blocks ..... | 4 | +| 7.1 Input data ..... | 5 | +| 8 Derivation procedure ..... | 6 | +| Appendix I – Example of the deviation procedure with ITU-T P.1203 mode 0 (metadata-based model) as a quality estimation model ..... | 8 | +| Appendix II – Example of the deviation procedure with ITU-T P.1203 mode 3 (bitstream-based model) as a quality estimation model ..... | 14 | +| Appendix III – Example of the deviation procedure with [ITU-T P.1204.4] (pixel-based model) as a quality estimation model ..... | 17 | +| Bibliography..... | 20 | + + + +## Derivation procedure for contribution values to quality degradation of adaptive audiovisual streaming services + +## 1 Scope + +This Recommendation describes a derivation procedure for contribution values that provides information about a relationship between a quality-influencing factor (i.e., media quality levels and stalling) and media session quality of adaptive audiovisual streaming services. + +In streaming services, the quality experienced by the end user can be assessed by using quality estimation models such as: + +- A metadata-based model, which uses audio and video bitrates, framerate, resolution ITU-T P.1203.1 mode 0, [ITU-T P.1203.2], [ITU-T P.1203.3]; +- A bitstream-based model that uses meta-data and any information from the video bitstream ITU-T P.1203.1 mode 3, [ITU-T P.1204.3], [ITU-T P.1203.2], [ITU-T P.1203.3]; +- A pixel-based model that uses information derived from reference video [ITU-T P.1204.4], [ITU-T P.1203.2], [ITU-T P.1203.3]; +- A hybrid metadata-based and pixel-based model [ITU-T P.1204.5], [ITU-T P.1203.2], [ITU-T P.1203.3]. + +These models estimate a mean opinion score (MOS) on a five-point absolute category rating scale [ITU-T P.910] as a global multi-media MOS score [ITU-T P.911] by taking into account quality-influencing factors such as media quality level (i.e., set of audio-coding or video degradations due to coding, spatial re-scaling or variations in video frame rates) and stalling (i.e., delivery degradations due to initial loading delay and rebuffering at the client) [ITU-T P.1203]. In this Recommendation, the output of these quality estimation models (i.e., global multi-media MOS score) is referred to as the final media session quality score as specified in [ITU-T P.1203]. + +The derivation procedure described provides contribution values that show how much each media quality level and stalling affects (i.e., decreases) the final media session quality score provided by quality estimation models such as [ITU-T P.1203] as previously described for each media session. This contribution value is calculated as the degree of quality degradation due to each media quality level and stalling by comparing with the maximum final media session quality score (i.e., the highest audiovisual quality and no stalling during the media session) based on Shapley theory, which is a key solution concept in cooperative game theory for distributing profits in a game among participants [b-Shapley]. Here, for example, different media quality levels selected or stalling occurrences during the media session result in different contribution values of each media quality level and stalling. Also, if the final media session quality score of a certain media session is different because of the use of a different quality estimation model, the contribution values of each media quality level and stalling are also different. + +The derivation procedure described is agnostic to the specified quality estimation models such as the ITU-T P.1203 type of model. This means that, although a metadata-based model (i.e., ITU-T P.1203 mode 0), bitstream-based model (i.e., ITU-T P.1203 mode 3), pixel-based model (i.e., [ITU-T P.1204.4]) and hybrid model (i.e., [ITU-T P.1204.5]), and their respective appendices for long term integration, generally exist to estimate the final media session quality score, this Recommendation can be used with all types of quality estimation model. As examples, the procedures when ITU-T P.1203 mode 0 (metadata-based model), ITU-T P.1203 mode 3 (bitstream-based model) and [ITU-T P.1204.4] (pixel-based model) are used as quality estimation model are described in Appendices I, II and III, respectively. + +This Recommendation primarily applies to the analysis of the quality aspect of progressive download or adaptive audiovisual streaming so that operators can ascertain degradation factors by deploying the derivation procedure described in end-point locations, mid-network monitoring points or a streaming server. + +## 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- [ITU-T P.910] Recommendation ITU-T P.910 (2023), *Subjective video quality assessment methods for multimedia applications.* +- [ITU-T P.911] Recommendation ITU-T P.911 (1998), *Subjective audiovisual quality assessment methods for multimedia applications.* +- [ITU-T P.1203] Recommendation ITU-T P.1203 (2017), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport.* +- [ITU-T P.1203.1] Recommendation ITU-T P.1203.1 (2019), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Video quality estimation module.* +- [ITU-T P.1203.2] Recommendation ITU-T P.1203.2 (2017), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Audio quality estimation module.* +- [ITU-T P.1203.3] Recommendation ITU-T P.1203.3 (2019), *Parametric bitstream-based quality assessment of progressive download and adaptive audiovisual streaming services over reliable transport – Quality integration module.* +- [ITU-T P.1204] Recommendation ITU-T P.1204 (2020), *Video quality assessment of streaming services over reliable transport for resolutions up to 4K.* +- [ITU-T P.1204.3] Recommendation ITU-T P.1204.3 (2020), *Video quality assessment of streaming services over reliable transport for resolutions up to 4K with access to full bitstream information.* +- [ITU-T P.1204.4] Recommendation ITU-T P.1204.4 (2022), *Video quality assessment of streaming services over reliable transport for resolutions up to 4K with access to full and reduced reference pixel information.* +- [ITU-T P.1204.5] Recommendation ITU-T P.1204.5 (2023), *Video quality assessment of streaming services over reliable transport for resolutions up to 4K with access to transport and received pixel information.* + +## 3 Definitions + +### 3.1 Terms defined elsewhere + +This Recommendation uses the following terms defined elsewhere: + +**3.1.1 adaptation set** [ITU-T P.1203]: A set of distinct media quality levels to be used for HTTP adaptive streaming, between which the player can perform media adaptation. + +**3.1.2 mean opinion score (MOS)** [b-ITU-T P.10]: The mean of opinion scores. + +**3.1.3 media quality level** [ITU-T P.1203]: A particular encoding setting applied to a video or audio stream. + +**3.1.4 segment** [ITU-T P.1203]: An audiovisual file belonging to one particular media quality level. + +**3.1.5 sequence** [ITU-T P.1203]: An audiovisual stream composed of multiple non-overlapping segments. + +### **3.2 Terms defined in this Recommendation** + +This Recommendation defines the following terms: + +**3.2.1 maximum media session quality score**: The final media session quality score for a certain sequence in which only the highest media quality level is selected, and no stalling occurs. + +**3.2.2 modified sequence**: A sequence in which specified selected media quality levels during a media session are replaced by the highest media quality level in an adaptation set or the number of stalling occurrences is replaced by zero. + +**3.2.3 stalling**: Stalling is caused by initial loading delay and rebuffering events at the client side, which could be a result of video data arriving late. Usually, stalling events are indicated to the viewer, e.g., in the form of a spinning wheel, and result in stalling of the media playout. + +## **4 Abbreviations and acronyms** + +This Recommendation uses the following abbreviations and acronyms: + +AAC Advanced Audio Coding + +AC3 Audio Compression 3 + +FQ Final Quality + +HE High Efficiency + +HRC Hypothetical Reference Circuit + +HTTP Hyper Text Transfer Protocol + +LC Low Complexity + +MO Mobile + +MOS Mean Opinion Score + +PC Personal Computer + +PVS Processed Video Sequence + +TA Tablet + +TV Television + +VP9 Video Payload type 9 + +## **5 Conventions** + +None. + +## 6 Field of application + +This Recommendation analyses the quality aspect of progressive download streaming or adaptive bitrate streaming. Limitations regarding the transport protocol, video length or encoding protocol and other attributes related to streaming services are the same as those of the quality estimation model used since the procedure described requires the final media session quality score from it. + +The procedure described is used on a server providing streaming service, as shown in Figure 1. After finishing a video playback on a client device, stalling information is sent to the streaming server. The contribution values are then calculated by a derivation procedure described in this Recommendation using a quality estimation model and selected quality level during media sessions that are on the streaming server. + +The contribution values are generally calculated on the streaming server so that operators can ascertain degradation factors. This is because the contribution calculation needs bitstream or pixel information of the highest media quality level at each segment, which is stored on a streaming server and not on a client device, if the bitstream-based (i.e., ITU-T P.1203 mode 3) or pixel-based (i.e., [ITU-T P.1204.4]) quality estimation model is used. However, if the metadata-based quality estimation model (i.e., ITU-T P.1203 mode 0) is used, the contribution values can be calculated on the client device since the metadata-based quality estimation model can be executed there and the metadata of the highest media quality level is also found on that site. + +![Sequence diagram showing the interaction between a Streaming server and a Client. The sequence is: 1. Send media information (server to client), 2. Video playback (client), 3. Send stalling information (client to server), 4. Calculate contribution values (server).](daa4a6fa7e2ba1954258f86b4928eb32_img.jpg) + +``` +sequenceDiagram + participant Streaming server + participant Client + Note right of Client: P.1211(23) + Streaming server->>Client: 1. Send media information + Note right of Client: 2. Video playback + Client->>Streaming server: 3. Send stalling information + Note left of Streaming server: 4. Calculate contribution values +``` + +Sequence diagram showing the interaction between a Streaming server and a Client. The sequence is: 1. Send media information (server to client), 2. Video playback (client), 3. Send stalling information (client to server), 4. Calculate contribution values (server). + +**Figure 1 – Sequence diagram** + +## 7 Building blocks + +The building blocks are depicted in Figure 2. This Recommendation uses the selected quality level sequence and stalling information extracted from the sequence and final media session quality score output from a quality estimation module. In addition to that, if the bitstream-based or pixel-based quality estimation model is used, bitstream information or pixel information is needed as shown in Figure 2. The details are described in clause 7.1. + +During the calculation of contribution values, since the derivation procedure requires its final media session quality score, the modified sequence is sent to the quality estimation module. Here, in a modified sequence, specified selected media quality levels during media session are replaced by the highest media quality level in an adaptation set or the number of stalling occurrences is replaced by zero. The details of the modified sequence are explained in clause 8. The derivation procedure described also needs adaptation set information, which is prepared at the streaming server. + +![Figure 2 – Building blocks. A block diagram showing the flow of data from a 'Sequence' input through three parallel extraction modules: 'Media parameter extraction', 'Pixel information extraction', and 'Stalling extraction'. These modules feed into a central 'Audiovisual quality estimation module (e.g., ITU-T P.1203)'. This module also receives 'Reference video pixel information' and an 'Adaptation set and bitstream and pixel information of highest media quality level'. The output of the quality estimation module is a 'Final media session quality score (e.g., 0.46)' and a 'Modified sequence'. These outputs, along with the extracted information, feed into a 'Contribution value estimation module'. The final output is 'Contribution values'. The diagram is labeled P.1211(23) in the bottom right corner.](cfef993dcc8fb513de79eb1f93cf26ae_img.jpg) + +Figure 2 – Building blocks. A block diagram showing the flow of data from a 'Sequence' input through three parallel extraction modules: 'Media parameter extraction', 'Pixel information extraction', and 'Stalling extraction'. These modules feed into a central 'Audiovisual quality estimation module (e.g., ITU-T P.1203)'. This module also receives 'Reference video pixel information' and an 'Adaptation set and bitstream and pixel information of highest media quality level'. The output of the quality estimation module is a 'Final media session quality score (e.g., 0.46)' and a 'Modified sequence'. These outputs, along with the extracted information, feed into a 'Contribution value estimation module'. The final output is 'Contribution values'. The diagram is labeled P.1211(23) in the bottom right corner. + +**Figure 2 – Building blocks** + +### 7.1 Input data + +This Recommendation uses the adaptation set information at a streaming server [ITU-T P.1203], which consists of multiple media quality levels (i.e., multiple levels of tuple of audio bitrate, video bitrate, video resolution, and video frame rate), a sequence that is a history of selected media quality levels during a media session, stalling information tuple (i.e., start time and duration of stalling [ITU-T P.1203.3]) during a media session, and the final media session quality score. In addition, if the bitstream-based quality estimation model is used, bitstream information of the selected quality levels during the media session is extracted from the sequence and the bitstream information of the highest media quality level at each segment, is needed as shown in Figure 2. When a pixel-based quality estimation model is used, video pixel information of the selected quality levels is extracted from the sequence and pixel information of the highest media quality level, and video reference pixel information obtained as input, are needed. Appendices II and III describe how to use the information. + +Let $N = \{r_1, \dots, r_R, s\}$ be a set of an adaptation set (i.e., prepared media quality levels on a streaming server) and stalling information, where $r_1$ is the lowest media quality level, $r_R$ is the highest media quality level, $R$ is the number of media quality levels, and $s$ represents all stalling events. Here, media quality level is the tuple of audio bitrate, video bitrate, video resolution, and video frame rate. Let $x_j \in N$ be the $j$ th element in $N$ , $j = 1, \dots, R + 1$ . $H$ denotes the sequence, which consists of selected media quality levels $r_i$ and all stalling events $s$ during video playback. The notation used in this Recommendation is listed in Table 1. + +Note that, since the quality estimation model is used during the contribution value calculation, the input data in this Recommendation is specified by the input data of the quality estimation model. As examples, the specific input data is described if the ITU-T P.1203 mode 0, ITU-T P.1203 mode 3 or ITU-T P.1204 is used in Appendices I, II and III. + +**Table 1 – Notation used in this Recommendation** + +| Symbol | Description | +|------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| $N = \{r_1, \dots, r_R, s\}$ | A set of adaptation set and stalling information. $r_1$ is the lowest media quality level, $r_R$ is the highest media quality level, $R$ is the number of media quality levels, and $s$ represents all stalling events. Here, media quality level is the tuple of audio bitrate, video bitrate, video resolution, and video frame rate | +| $x_j$ | $j$ th element in $N, j = 1, \dots, R + 1$ . | +| $Z$ | All of subsets of $N$ without $x_j$ . | +| $z_i$ | One of the elements of $Z. z_i \in Z, i = 1, \dots, Z $ | +| $H$ | A sequence, that consists of selected media quality level $r_i$ and all stalling events $s$ during video playback. | +| $H'$ (or $H''$ ) | A modified sequence of an input sequence $H$ and $z_i$ , in which a part of the selected media quality level during a media session is replaced with the highest media quality level in an adaptation set. | +| $f$ | A function that takes a given $H$ and subset $z_i$ (or $z_i \cup x_j$ ) as input and outputs a modified sequence $H'$ (or $H''$ ). | +| $v$ | A quality estimation model such as that of [ITU-T P.1203] or [ITU-T P.1204], which estimates the quality of an input sequence, $H$ . | +| $c_j$ | Contribution value of $x_j$ . | + +## 8 Derivation procedure + +The derivation procedure for contribution values consists of three steps, which are summarized in Table 2. + +**Table 2 – Calculation steps of contribution values** + +| Step | Description | +|------|----------------------------------------------------------------------------------------------------------------------| +| 1 | Pick up one element $x_j$ in $N$ and iterate steps 2 and 3 for over all elements $x_j$ in $N, j = 1, \dots, R + 1$ . | +| 2 | Determine the set $Z$ consisting of all subsets of $N$ without $x_j$ . | +| 3 | Calculate the contribution value, $c_j$ , of $x_j$ by adapting the Shapley value calculation given in Equation 1. | + +The first step is to pick up one element $x_j$ and iterate steps 2 and 3 for over all elements $x_j$ in $N$ . Here, the order of picking up element in $N$ does not affect the calculation results. For simplicity, the case $x_j$ is explained in the following. However, the procedure described is executed over all elements in $N$ to calculate all contribution values, $c_j, j = 1, \dots, R, R + 1$ . The second step is to determine the composition of all subsets $Z$ of $N$ without $x_j$ . Here, $Z$ is all possible patterns of subsets induced from $N$ without $x_j$ , and the cardinality of $Z$ is $2^{|N|-1}$ . The third step is to calculate the contribution value, $c_j$ , of $x_j$ by adapting the Shapley value calculation. + +The details of the third step, in which the contribution value, $c_j$ , of $x_j$ is calculated, are as follows. Let $f$ be a function that takes a given sequence $H$ and subset $z_i \in Z, i = 1, \dots, |Z|$ (or $z_i \cup x_j$ ) as input and outputs a modified sequence, $H'$ (or $H''$ ). Let $v$ be a quality estimation model such as that of [ITU-T P.1203] or [ITU-T P.1204], which estimates the quality of an input sequence, $H$ . From here, + +although the contribution value calculation algorithm is explained only for a given subset, $z_i$ , it is also applicable to other subsets, $(z_i \cup x_j)$ . The modification procedure is described in the following paragraphs. The contribution value, $c_j$ , of $x_j$ is calculated by adopting the Shapley value calculation algorithm as follows. + +$$c_j(v, f, H) = \sum_{i=1}^{|Z|} \frac{|z_i|!(|N|-|z_i|-1)!}{|N|!} \left( v(f(z_i, H)) - v(f(z_i \cup x_j, H)) \right) \quad (1)$$ + +where $|Z|$ and $|N|$ denote the cardinality of $Z$ and $N$ , respectively, and the sum extends over all subsets, $z_i \in Z$ , which are the subsets of $N$ that do not include $x_j$ . + +The function $f$ maps a subset $z_i$ and a sequence of quality levels $H = (h_t)$ to a new sequence of quality levels $H' = (h'_t)$ by replacing all occurrences of quality levels in $z_i$ by the highest quality level $r_R$ , + +$$h'_t = \begin{cases} r_R & \text{if } h_t \in z_i \\ h_t & \text{otherwise} \end{cases} \quad (2)$$ + +For the stalling, if $s$ is in $z_i$ , $f$ generates $H'$ in which the stalling events are removed, while if $s$ is not in $z_i$ , $f$ generates $H'$ in which stalling information is the same as the original $H$ . + +For the case, $z_i \cup x_j$ , $f$ outputs $H''$ similarly to modifying $H$ to $H'$ . By generating sequences $H'$ and $H''$ on the basis of a given subset $z_i$ and calculating the quality decrease over all possible subsets using Equation (1), the contribution value, $c_j$ , of $x_j$ is calculated. + +## Appendix I + +### Example of the deviation procedure with ITU-T P.1203 mode 0 (metadata-based model) as a quality estimation model + +(This appendix does not form an integral part of this Recommendation.) + +This appendix describes the example of the deviation procedure when ITU-T P.1203 mode 0 is used as a quality estimation model. Especially, prepared processed video sequences (PVSs), input data specification, and contribution values are described. + +#### Prepared PVSs + +In this example, four PVSs listed under this heading are used to show the derivation procedure and the contribution values for various kinds of hypothetical reference circuit (HRC). These PVSs are generated on the basis of the HRC of an ITU-T P.1203 open dataset [b-Robitza] by using a prepared source video, which is 1 min long. The characteristics of these PVSs follow and the HRCs used are shown in Figures I.1 and I.2. Table I.1 shows the media quality level of the HRC. + +PVS01: HRC80 in TR04 [b-Robitza] in which only one media quality level is selected during a media session with no stalling. + +PVS02: HRC03 TR04 [b-Robitza] in which multiple media quality levels are selected during a media session with no stalling. + +PVS03: HRC85 in TR04 [b-Robitza] in which multiple media quality levels are selected and a stalling occurs during a media session. + +PVS04: HRC02 in TR04 [b-Robitza] in which multiple media quality levels are selected and stalling occurs multiple times during a media session. + +![Figure I.1: Bitrate (kbit/s) vs Viewing time (s) for HRC80 and HRC03 in TR04 in ITU-T P.1203 open dataset [b-Robitza]. The left graph shows a constant bitrate of 500 kbit/s for HRC80. The right graph shows a variable bitrate for HRC03, with values of 2500, 500, 250, 500, 2500, 500, 250, 500, 2500, 500, 250, 500 kbit/s over 60 seconds.](70f5f44cb855bbb0c203c00187b2113e_img.jpg) + +Figure I.1 consists of two line graphs. The left graph, representing HRC80, shows a constant bitrate of 500 kbit/s over a 60-second viewing time. The right graph, representing HRC03, shows a variable bitrate over the same 60-second period. The HRC03 graph is a step function with the following bitrates: 2500 kbit/s (0-5s), 500 kbit/s (5-10s), 250 kbit/s (10-15s), 500 kbit/s (15-20s), 2500 kbit/s (20-25s), 500 kbit/s (25-30s), 250 kbit/s (30-35s), 500 kbit/s (35-40s), 2500 kbit/s (40-45s), 500 kbit/s (45-50s), 250 kbit/s (50-55s), and 500 kbit/s (55-60s). + +| Viewing time (s) | HRC80 Bitrate (kbit/s) | HRC03 Bitrate (kbit/s) | +|------------------|------------------------|------------------------| +| 0 - 5 | 500 | 2500 | +| 5 - 10 | 500 | 500 | +| 10 - 15 | 500 | 250 | +| 15 - 20 | 500 | 500 | +| 20 - 25 | 500 | 2500 | +| 25 - 30 | 500 | 500 | +| 30 - 35 | 500 | 250 | +| 35 - 40 | 500 | 500 | +| 40 - 45 | 500 | 2500 | +| 45 - 50 | 500 | 500 | +| 50 - 55 | 500 | 250 | +| 55 - 60 | 500 | 500 | + +Figure I.1: Bitrate (kbit/s) vs Viewing time (s) for HRC80 and HRC03 in TR04 in ITU-T P.1203 open dataset [b-Robitza]. The left graph shows a constant bitrate of 500 kbit/s for HRC80. The right graph shows a variable bitrate for HRC03, with values of 2500, 500, 250, 500, 2500, 500, 250, 500, 2500, 500, 250, 500 kbit/s over 60 seconds. + +Figure I.1 – HRC80 and HRC03 in TR04 in ITU-T P.1203 open dataset [b-Robitza] + +![Two line graphs showing Bitrate (kbit/s) vs Viewing time (s) for HRC85 and HRC02. The left graph for HRC85 shows a sawtooth pattern between 500 and 2500 kbit/s over 60 seconds. The right graph for HRC02 shows a step-like pattern between 250 and 2500 kbit/s over 80 seconds. A label 'P.1211(23)' is in the bottom right corner.](71ab4df17511d75261da8d462d643b1a_img.jpg) + +Two line graphs showing Bitrate (kbit/s) vs Viewing time (s) for HRC85 and HRC02. The left graph for HRC85 shows a sawtooth pattern between 500 and 2500 kbit/s over 60 seconds. The right graph for HRC02 shows a step-like pattern between 250 and 2500 kbit/s over 80 seconds. A label 'P.1211(23)' is in the bottom right corner. + +**Figure I.2 – HRC85 and HRC02 in TR04 in ITU-T P.1203 open dataset [b-Robitza]** + +**Table I.1 – Media quality level information of PVSs** + +NOTE – Quality levels QL1, QL3, and QL5 are not defined in this dataset [b-Robitza]. + +| Quality level | Video bitrate (kbit/s) | Resolution (pixel) | +|---------------|------------------------|--------------------| +| QL7 | 10 000 | 1 080 | +| QL6 | 2 500 | 1 080 | +| QL4 | 500 | 480 | +| QL2 | 150 | 240 | + +#### Input data specifications + +The input data of this Recommendation is the adaptation set information, a sequence that is a history of selected media quality levels during a media session, stalling information during a media session and the final media session quality score. The input data are specified in Table I.2, where the listed input data is based on ITU-T P.1203.1 mode 0, [ITU-T P.1203.2] and ITU-T P.1203.3]. + +**Table I.2 – Input specifications** + +| Description | Values | Frequency | +|---------------------------------------------------------|------------------------------------------------------|-------------------| +| The resolution of the image displayed to the user | Number of pixels ( $W \times H$ ) in displayed video | Per media chunk | +| The device type on which the media is played | "PC", "TV", "MO", "TA" | Per media chunk | +| Device display size | Display size (diagonal in inches) | Per media chunk | +| Relative viewing distance in multiple of display height | Relative viewing distance | Per media chunk | +| Target Audio bitrate | Bitrate in kilobits per second | Per media segment | +| Segment duration | Duration in seconds | Per media segment | +| Audio codec | One of: AAC-LC, AAC-HEv1, AAC-HEv2, AC3 | Per media segment | +| Audio sampling frequency | In hertz | Per media segment | +| Number of audio channels | 2 | Per media segment | + +**Table I.2 – Input specifications** + +| Description | Values | Frequency | +|-----------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------| +| Video target bitrate | Bitrate in kilobits per second | Per media chunk | +| Video frame rate | Frame rate in frames per second. | Per media chunk | +| Segment duration | Duration in seconds | Per media chunk | +| Video encoding resolution | Number of pixels ( $W \times H$ ) in transmitted video | Per media chunk | +| Video codec and profile | ITU-T H.264-high | Per media chunk | +| Stalling or initial loading event start | The start time of the stalling event in seconds relative to the start of the original video clip, expressed in media time (not wall clock time)
NOTE – This is 0 for initial loading delay. | Per stalling event | +| Event duration | The duration of the stalling event in seconds | Per stalling event | +| Final media session quality score | Scores provided per session and on a quality scale of 1 to 5 | Per session | + +Legend – AAC: advanced audio coding; AC3: audio compression 3; HE: high efficiency; LC: low complexity; MO: mobile; MPEG: Moving Picture Experts Group; PC: personal computer; TA: tablet; TV: television; VP9: video payload type 9. + +#### **Example of calculation steps** + +An example calculation step is described. Assume a sequence and adaptation set as follows. + +$$H = [QL4, QL6, QL2, QL2, QL7], N = \{QL7, QL6, QL4, QL2, s\},$$ + +where, $H$ has no stalling. The first step of the calculation is to select $x_j$ from $N$ and determine all of subset $Z$ . Assume that the selected $x_j$ in the iteration of step 1 is $QL2$ , + +$$x_j = QL2,$$ + +and $Z$ will be determined, according to Table 2, as all the patterns of subsets of $N$ without $x_j$ , which is + +$$\begin{aligned} Z = \{ & z_1 = \{QL4\}, z_2 = \{QL6\}, z_3 = \{QL7\}, z_4 = \{s\}, z_5 = \{QL4, QL6\}, z_6 = \{QL4, QL7\}, z_7 \\ & = \{QL4, s\}, z_8 = \{QL6, QL7\}, z_9 = \{QL6, s\}, z_{10} = \{QL7, s\}, z_{11} \\ & = \{QL4, QL6, QL7\}, z_{12} = \{QL4, QL6, s\}, z_{13} = \{QL4, QL7, s\}, z_{14} \\ & = \{QL6, QL7, s\}, z_{15} = \{QL4, QL6, QL7, s\}, z_{16} = \{\} \}. \end{aligned}$$ + +Then, $z_i$ is selected from $Z$ to modify the sequence $H$ . For instance, if $z_1$ is selected, + +$$z_1, \{z_1 \cup x_j\} = \{QL2, QL4\}.$$ + +Then, by using Equation (2), the modified sequence is generated. In this case, since $QL2$ and $QL4$ are in subsets $z_1$ and $\{z_1 \cup x_j\}$ , $H'$ and $H''$ become as follows. + +$$H' = [\mathbf{QL7}, QL6, QL2, QL2, QL7], v(H') = 2.880$$ + +$$H'' = [\mathbf{QL7}, QL6, \mathbf{QL7}, \mathbf{QL7}, QL7], v(H'') = 4.885$$ + +The bold segments are replaced with the highest media quality level in $N$ and the final media session quality score of both sequence $H'$ and $H''$ is calculated using ITU-T P.1203 mode 0. Then, the part of the contribution values is calculated as follows. + +$$\frac{|z_1|! (|N| - |z_1| - 1)!}{|N|!} (v(f(z_1, H)) - v(f(z_1 \cup x_j, H))) = \frac{1! 3!}{5!} (2.880 - 4.885) = -0.100.$$ + +Note that, the outputs of $v$ is rounded to the fourth decimal place. + +This process is executed over all elements in $Z$ as follows. From here onwards, the term + +$$\frac{|z_i|! (|N| - |z_i| - 1)!}{|N|!} (v(f(z_i, H)) - v(f(z_i \cup x_j, H)))$$ + +is denoted $E$ . + +- For $z_2$ + - $z_2, \{z_2 \cup x_j\} = \{QL2, QL6\}.$ + - $H' = [QL4, \mathbf{QL7}, QL2, QL2, QL7], v(H') = 2.822,$ + - $H'' = [QL4, \mathbf{QL7}, \mathbf{QL7}, \mathbf{QL7}, QL7], v(H'') = 4.425$ + - $E = \frac{1!3!}{5!} (2.822 - 4.425) = -0.080$ +- For $z_3$ + - $z_3, \{z_3 \cup x_j\} = \{QL2, QL7\}.$ + - $H' = [QL4, QL6, QL2, QL2, QL7], v(H') = 2.822,$ + - $H'' = [QL4, QL6, \mathbf{QL7}, \mathbf{QL7}, QL7], v(H'') = 4.423$ + - $E = \frac{1!3!}{5!} (2.822 - 4.423) = -0.080$ +- For $z_4$ + - $z_4, \{z_4 \cup x_j\} = \{QL2, s\}.$ + - $H' = [QL4, QL6, QL2, QL2, QL7], v(H') = 2.822,$ + - $H'' = [QL4, QL6, \mathbf{QL7}, \mathbf{QL7}, QL7], v(H'') = 4.423$ + - $E = \frac{1!3!}{5!} (2.822 - 4.423) = -0.080$ +- For $z_5$ + - $z_5, \{z_5 \cup x_j\} = \{QL2, QL4, QL6\}.$ + - $H' = [\mathbf{QL7}, \mathbf{QL7}, QL2, QL2, QL7], v(H') = 2.880,$ + - $H'' = [\mathbf{QL7}, \mathbf{QL7}, \mathbf{QL7}, \mathbf{QL7}, QL7], v(H'') = 4.896$ + - $E = \frac{2!2!}{5!} (2.880 - 4.896) = -0.067$ +- For $z_6$ + - $z_6, \{z_6 \cup x_j\} = \{QL2, QL4, QL7\}.$ + - $H' = [\mathbf{QL7}, QL6, QL2, QL2, QL7], v(H') = 2.880,$ + - $H'' = [\mathbf{QL7}, QL6, \mathbf{QL7}, \mathbf{QL7}, QL7], v(H'') = 4.885$ + - $E = \frac{2!2!}{5!} (2.880 - 4.885) = -0.067$ +- For $z_7$ + - $z_7, \{z_7 \cup x_j\} = \{QL2, QL4, s\}.$ + +- $H' = [QL7, QL6, QL2, QL2, QL7], v(H') = 2.880,$ +- $H'' = [QL7, QL6, QL7, QL7, QL7], v(H'') = 4.885$ +- $E = \frac{2!2!}{5!} (2.880 - 4.885) = -0.067$ + +– For $z_8$ + +- $z_8, \{z_8 \cup x_j\} = \{QL2, QL6, QL7\}.$ +- $H' = [QL4, QL7, QL2, QL2, QL7], v(H') = 2.822,$ +- $H'' = [QL4, QL7, QL7, QL7, QL7], v(H'') = 4.425$ +- $E = \frac{2!2!}{5!} (2.822 - 4.425) = -0.053$ + +– For $z_9$ + +- $z_9, \{z_9 \cup x_j\} = \{QL2, QL6, s\}.$ +- $H' = [QL4, QL7, QL2, QL2, QL7], v(H') = 2.822,$ +- $H'' = [QL4, QL7, QL7, QL7, QL7], v(H'') = 4.425$ +- $E = \frac{2!2!}{5!} (2.822 - 4.425) = -0.053$ + +– For $z_{10}$ + +- $z_{10}, \{z_{10} \cup x_j\} = \{QL2, QL7, s\}.$ +- $H' = [QL4, QL6, QL2, QL2, QL7], v(H') = 2.822,$ +- $H'' = [QL4, QL6, QL7, QL7, QL7], v(H'') = 4.423$ +- $E = \frac{2!2!}{5!} (2.822 - 4.423) = -0.053$ + +– For $z_{11}$ + +- $z_{11}, \{z_{11} \cup x_j\} = \{QL2, QL4, QL6, QL7\}.$ +- $H' = [QL7, QL7, QL2, QL2, QL7], v(H') = 2.880,$ +- $H'' = [QL7, QL7, QL7, QL7, QL7], v(H'') = 4.896$ +- $E = \frac{3!1!}{5!} (2.880 - 4.896) = -0.100$ + +– For $z_{12}$ + +- $z_{12}, \{z_{12} \cup x_j\} = \{QL2, QL4, QL6, s\}.$ +- $H' = [QL7, QL7, QL2, QL2, QL7], v(H') = 2.880,$ +- $H'' = [QL7, QL7, QL7, QL7, QL7], v(H'') = 4.896$ +- $E = \frac{3!1!}{5!} (2.880 - 4.896) = -0.100$ + +– For $z_{13}$ + +- $z_{13}, \{z_{13} \cup x_j\} = \{QL2, QL4, QL7, s\}.$ +- $H' = [QL7, QL6, QL2, QL2, QL7], v(H') = 2.880,$ +- $H'' = [QL7, QL6, QL7, QL7, QL7], v(H'') = 4.885$ +- $E = \frac{3!1!}{5!} (2.880 - 4.885) = -0.100$ + +– For $z_{14}$ + +- $z_{14}, \{z_{14} \cup x_j\} = \{QL2, QL6, QL7, s\}.$ +- $H' = [QL4, QL7, QL2, QL2, QL7], v(H') = 2.822$ +- $H'' = [QL4, QL7, QL7, QL7, QL7], v(H'') = 4.425$ +- $E = \frac{3!1!}{5!} (2.822 - 4.425) = -0.080$ + +– For $z_{15}$ + +- $z_{15}, \{z_{15} \cup x_j\} = \{QL2, QL4, QL6, QL7, s\}.$ +- $H' = [QL7, QL7, QL2, QL2, QL7], v(H') = 2.880,$ +- $H'' = [QL7, QL7, QL7, QL7, QL7], v(H'') = 4.896$ +- $E = \frac{4!0!}{5!} (2.880 - 4.896) = -0.403$ + +– For $z_{16}$ + +- $z_{16}, \{z_{16} \cup x_j\} = \{QL2\}.$ +- $H' = [QL4, QL6, QL2, QL2, QL7], v(H') = 2.822,$ +- $H'' = [QL4, QL6, QL7, QL7, QL7], v(H'') = 4.423$ +- $E = \frac{0!4!}{5!} (2.822 - 4.423) = -0.320$ + +Finally, the contribution value $c_j$ of $x_j$ (i.e., QL2) can be obtained by summing $E$ from $z_1$ to $z_{16}$ . In this example case, $c_j$ is $-1.807$ . To calculate contribution values for QL4, QL6, QL7, and stalling, the preceding process is repeated. + +Note that as described in clause 6, since all information to estimate the quality of modified sequence $H'$ and $H''$ is from the metadata-based model, the contribution calculation can be executed on the client device. + +#### Calculation results + +Table I.3 shows the calculated contribution values. In Table I.3, final quality (FQ) means the final media session quality score of corresponding PVS calculated by ITU-T P.1203 mode 0. QL7, QL6, QL4, QL2, and stalling columns mean the contribution value of each media quality level. The column total means the total value of contribution value. + +In PVS01, since QL7, QL6 and QL2 are not selected and do not induce any quality degradation, the contribution value equals 0. Also, since stalling does not occur or induce any quality degradation, the contribution value equals 0. On the other hand, since QL4 was selected during a media session, it has $-1.375$ points. This means that QL4 reduces the final media session quality score by approximately 1.375 compared with the maximum media session quality score. In PVS02, QL6, QL4 and QL2 were selected. Thus, each media quality level reduced the FQ, and QL2 reduced the quality the most. In PVS03, since stalling occurred, the contribution value of the stalling was $-0.375$ . In PVS04, since several quality levels were selected, each quality level reduces the quality. Specifically, since stalling occurred twice and media quality level QL2 was selected many times in this PVS, the stalling and QL2 reduced the final media session quality score. These contribution values of each PVS also indicate that if the PVS differs, the contribution values of each PVS also differ since the selected media quality level or stalling occurrence differs between PVSs. + +**Table I.3 – Calculated results of contribution values for PVS01, PVS02, PVS03, and PVS04** + +| | FQ | QL7 | QL6 | QL4 | QL2 | Stalling | Total | +|-------|-------|-----|--------|--------|--------|----------|--------| +| PVS01 | 3.531 | 0 | 0 | -1.357 | 0 | 0 | -1.357 | +| PVS02 | 2.749 | 0 | -0.008 | -0.974 | -1.157 | 0 | -2.139 | +| PVS03 | 3.881 | 0 | -0.036 | -0.595 | 0 | -0.375 | -1.007 | +| PVS04 | 1.599 | 0 | -0.003 | -0.203 | -2.320 | -0.763 | -3.288 | + +## Appendix II + +### Example of the deviation procedure with ITU-T P.1203 mode 3 (bitstream-based model) as a quality estimation model + +(This appendix does not form an integral part of this Recommendation.) + +This appendix describes an example of the deviation procedure when ITU-T P.1203 mode 3 is used as a quality estimation model. Especially, PVSs used, input data specification and contribution values are described. + +#### Prepared PVSs + +In this example, the same PVSs as those of Appendix I are used. Thus, detailed information about the PVSs is omitted. + +#### Input data specifications + +The input data of this Recommendation is the adaptation set information, a sequence that is a history of selected media quality level during a media session, stalling information during a media session, and the final media session quality score. In addition to that, since ITU-T P.1203 mode 3 is a bitstream-based model, the bitstream information at each segment is used. The specifications of input data are in Table II.1, where the listed input data is based on ITU-T P.1203 mode 3, [ITU-T P.1203.2], [ITU-T P.1203.3]. + +**Table II.1 – Input specifications** + +| Description | Values | Frequency | +|---------------------------------------------------------|------------------------------------------------------|-------------------| +| The resolution of the image displayed to the user | Number of pixels ( $W \times H$ ) in displayed video | Per media chunk | +| The device type on which the media is played | "PC", "TV", "MO", "TA" | Per media chunk | +| Device display size | Display size (diagonal in inches) | Per media chunk | +| Relative viewing distance in multiple of display height | Relative viewing distance | Per media chunk | +| Target audio bitrate | Bitrate in kilobits per second | Per media segment | +| Segment duration | Duration in seconds | Per media segment | +| Audio frame number | Integer, starting with 1 | Per media segment | +| Audio frame size | Size of the frame in bytes | Per audio frame | +| Audio frame duration | Duration in seconds | Per audio frame | +| Audio codec | One of: AAC-LC, AAC-HEv1, AAC-HEv2, AC3 | Per media segment | +| Audio sampling frequency | In hertz | Per media segment | +| Number of audio channels | 2 | Per media segment | +| Audio bitstream | Encoded audio bytes for the frame | Per audio frame | +| Video bitrate | Bitrate in kilobits per second | Per media chunk | + +**Table II.1 – Input specifications** + +| Description | Values | Frequency | +|-----------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------| +| Video frame rate | Frame rate in frames per second | Per media chunk | +| Segment duration | Duration in seconds | Per media chunk | +| Video encoding resolution | Number of pixels ( $W \times H$ ) in transmitted video | Per media chunk | +| Video codec and profile | ITU-T H.264-high | Per media chunk | +| Video frame number | Integer, starting at 1, denoting the frame sequence number in encoding order | Per video frame | +| Video frame duration | Duration of the frame in seconds | Per video frame | +| Video frame size | The size of the encoded video frame in bytes | Per video frame | +| Type of each picture | "I"/"P"/"B" | Per video frame | +| Video bitstream | Encoded video bytes for the frame | Per video frame | +| Stalling or initial loading event start | The start time of the stalling event in seconds relative to the start of the original video clip, expressed in media time (not wall clock time)
NOTE – This is 0 for initial loading delay. | Per stalling event | +| Event duration | The duration of the stalling event in seconds | Per stalling event | +| Final media session quality score | Scores provided per session and on a quality scale of 1 to 5 | Per session | + +#### **Example of calculation steps** + +Most parts of the calculation step are the same as in Appendix I. Since the bitstream information is needed in ITU-T P.1203 mode 3, generation of the modified sequence is slightly different. Assume that the same $N, x_j, S$ as in Appendix I and a sequence $H$ as follows. + +$$H = [(QL4, b_{1,ql4}), (QL6, b_{2,ql6}), (QL2, b_{3,ql2}), (QL2, b_{4,ql2}), (QL7, b_{5,ql7})],$$ + +where $b_{j,qlk}$ represents a tuple of video frame size, type of each picture and quantization parameter values of media quality level $k$ at segment $j$ that are parsed from video bitstream. Then, the modified sequences become as follows. + +$$H' = [(QL7, b_{1,ql7}), (QL6, b_{2,ql6}), (QL2, b_{3,ql2}), (QL2, b_{4,ql2}), (QL7, b_{5,ql7})],$$ + +$$H'' = [(QL7, b_{1,ql7}), (QL6, b_{2,ql6}), (QL7, b_{3,ql7}), (QL7, b_{4,ql7}), (QL7, b_{5,ql7})].$$ + +Here, $b_{j,ql7}$ is obtained as an input. Different from the metadata-based model, ITU-T P.1203 mode 0 in Appendix I, the media quality level and bitstream information at bold segments are replaced with the highest media quality level and bitstream information of the highest media quality level at those segments. Here, since the bitstream information of the highest quality level at each segment is placed at the streaming server, the contribution calculation is executed at the streaming server. The other steps are the same as those in Appendix I. + +#### **Calculation results** + +The calculated contribution values are listed in Table II.2, in which FQ means the final media session quality score of the corresponding PVS calculated by ITU-T P.1203 mode 3. QL7, QL6, QL4, QL2, and stalling columns mean the contribution value of each media quality level. The column total means the total value of the contribution value. + +Most calculated contribution values tend to be the same as those of ITU-T P.1203 mode 0 in Appendix I. However, since the quality estimation model and final quality media session quality score differ, the contribution values also differ from those of ITU-T P.1203 mode 0. + +In PVS01, since QL7, QL6, and QL2 are not selected and do not induce any quality degradation, the contribution value equals 0. Also, since stalling does not occur or induce any quality degradation, the contribution value equals 0. On the other hand, since QL4 was selected during the media session, it was $-2.197$ points. In PVS02, QL6, QL4, and QL2 were selected. Thus, each media quality level reduced the FQ. In PVS03, since a stalling occurred, the contribution value of the stalling was $-0.337$ . In PVS04, since several quality levels were selected, each quality level reduces the quality. Specifically, since stalling occurred twice and media quality level QL2 was selected many times in this PVS, the stalling and QL2 reduced the final media session quality score. + +**Table II.2 – Calculated results of contribution values for PVS01, PVS02, PVS03 and PVS04** + +| | FQ | QL7 | QL6 | QL4 | QL2 | Stalling | Total | +|-------|-------|-----|--------|--------|--------|----------|--------| +| PVS01 | 2.631 | 0 | 0 | -2.197 | 0 | 0 | -2.197 | +| PVS02 | 2.310 | 0 | -0.146 | -1.383 | -0.990 | 0 | -2.518 | +| PVS03 | 3.192 | 0 | -0.200 | -1.099 | 0 | -0.337 | -1.636 | +| PVS04 | 1.569 | 0 | -0.008 | -0.275 | -2.256 | -0.721 | -3.259 | + +## Appendix III + +### Example of the deviation procedure with [ITU-T P.1204.4] (pixel-based model) as a quality estimation model + +(This appendix does not form an integral part of this Recommendation.) + +This appendix describes an example of the deviation procedure when [ITU-T P.1204.4] is used as a quality estimation model. Especially, the PVSs used, input data specification and contribution values are described. + +#### Prepared PVSs + +In this example, the same PVSs as those in Appendix I are used. Thus, detailed information about PVSs is omitted. + +#### Input data specifications + +The input data of this Recommendation is the adaptation set information, a sequence that is a history of selected media quality level during a media session, stalling information during a media session, and the final media session quality score. In addition to that, since [ITU-T P.1204.4] is a pixel-based model, the video pixel information at each segment is used. The input data are specified in Table III.1, where the listed input data is based on [ITU-T P.1204.4], [ITU-T P.1203.2], [ITU-T P.1203.3]. + +**Table III.1 – Input specifications** + +| Description | Values | Frequency | +|----------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------| +| The resolution of the image displayed to the user | Number of pixels ( $W \times H$ ) in displayed video | Per media chunk | +| The device type on which the media is played | "PC", "TV", "MO", "TA" | Per media chunk | +| Device display size | Display size (diagonal in inches) | Per media chunk | +| Relative viewing distance in multiples of display height | Relative viewing distance | Per media chunk | +| Target audio bit-rate | Bitrate in kilobits per second | Per media segment | +| Segment duration | Duration in seconds | Per media segment | +| Audio codec | One of: AAC-LC, AAC-HEv1, AAC-HEv2, AC3 | Per media segment | +| Audio sampling frequency | In hertz | Per media segment | +| Number of audio channels | 2 | Per media segment | +| Stalling or initial loading event start | The start time of the stalling event in seconds relative to the start of the original video clip, expressed in media time (not wall clock time)
NOTE – This is 0 for initial loading delay. | Per stalling event | + +**Table III.1 – Input specifications** + +| Description | Values | Frequency | +|-----------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------| +| Event duration | The duration of the stalling event in seconds | Per stalling event | +| Degraded video | The raw pixels (YUV file including metadata required for parsing; width, height, frame rate and pixel format) of the processed video, i.e., the video decoded and upscaled to display resolution without buffering or stalling. The frame information in degraded video and reference video is synchronized, i.e., no frame misalignments are present | Per media chunk | +| Reference video information | The reference-side information extraction module takes as input the reference video and outputs the side information file.
The reference model side channel bandwidth limit is 256 kbit/s. Thus, the side information of the reference model for a video sequence $v$ is stored in a file with size at most $256/8 * t_v$ kB, where $t_v$ is the duration of video $v$ in seconds. | Per media chunk | +| Final media session quality score | Scores provided per session and on a quality scale of 1 to 5 | Per session | + +#### **Example of calculation steps** + +Most parts of the calculation step are the same as those in Appendix I. Since the pixel information is needed in [ITU-T P.1204.4], a sequence and modified sequences are slightly different. Assume the same $N, x_j, S, H$ as in Appendix I. In addition to that in the pixel-based model, the pixel information of that sequence $H$ , + +$$H = [(QL4, p_{1,q14}), (QL6, p_{2,q16}), (QL2, p_{3,q12}), (QL2, p_{4,q12}), (QL7, p_{5,q17})],$$ + +where $p_{j,q1k}$ represents the degraded video pixel information of media quality level $k$ at segment $j$ . Then, the modified sequences become as follows. + +$$H' = [(QL7, p_{1,q17}), (QL6, p_2), (QL2, p_3), (QL2, p_4), (QL7, p_5)],$$ + +$$H'' = [(QL7, p_{1,q17}), (QL6, p_2), (QL7, p_{3,q17}), (QL7, p_{4,q17}), (QL7, p_5)].$$ + +Also, when the final media session quality score of $H, H'$ and $H''$ is calculated using [ITU-T P.1204.4], reference video pixel information is used. + +Here, since the pixel information of the highest media quality level at each segment is placed at the streaming server, the contribution calculation is executed at the streaming server. The other steps are the same as those in Appendix I. + +#### **Calculation results** + +Table III.2 shows the calculated contribution values. In Table III.2, FQ means the final media session quality score of the corresponding PVS calculated by [ITU-T P.1204.4]. QL7, QL6, QL4, QL2, and stalling columns mean the contribution value of each media quality level. The column total means the total value of the contribution value. + +Most calculated contribution values tend to be the same as those of ITU-T P.1203 mode 0 in Appendix I. However, since the quality estimation model and final quality media session quality score differ, the contribution values also differ from those of ITU-T P.1203 mode 0. + +In PVS01, since QL7, QL6, and QL2 are not selected and do not induce any quality degradation, the contribution value equals 0. Also, since stalling does not occur or induce any quality degradation, the + +contribution value equals 0. On the other hand, since QL4 was selected during a media session, it was the $-0.965$ points. In PVS02, QL6, QL4 and QL2 were selected. Thus, each media quality level reduced the FQ. In PVS03, since stalling occurred, its contribution value was $-0.051$ . In PVS04, since several quality levels were selected, each quality level reduced the quality. Specifically, since stalling occurred twice and media quality level QL2 was selected many times in this PVS, the stalling and QL2 reduced the final media session quality score. + +**Table III.2 – Calculated results of contribution values for PVS01, PVS02, PVS03, and PVS04** + +| | FQ | QL7 | QL6 | QL4 | QL2 | Stalling | Total | +|-------|-------|-----|----------|----------|----------|----------|----------| +| PVS01 | 3.442 | 0 | 0 | $-0.965$ | 0 | 0 | $-0.965$ | +| PVS02 | 3.243 | 0 | $-0.064$ | $-0.511$ | $-0.590$ | 0 | $-1.164$ | +| PVS03 | 3.775 | 0 | $-0.225$ | $-0.375$ | 0 | $-0.051$ | $-0.632$ | +| PVS04 | 1.871 | 0 | $-0.002$ | $-0.040$ | $-1.675$ | $-0.818$ | $-2.536$ | + +## Bibliography + +- [b-ITU-T H.264] Recommendation ITU-T H.264 (2021), *Advanced video coding for generic audiovisual services*. +- [b-ITU-T H.265] Recommendation ITU-T H.265 (2023), *High efficiency video coding*. +- [b-ITU-T P.10] Recommendation ITU-T P.10/G.100 (2017), *Vocabulary for performance, quality of service and quality of experience*. +- [b-Robitza] Robitza, W., Göring, S., Raake, A., Lindegren, D., Heikkilä, G., Gustafsson, J., List, P., Feiten, B., Wüstenhagen, U., Garcia, M.-N., Yamagishi, K., Broom, S. (2018). HTTP adaptive streaming QoE estimation with ITU-T Rec. P.1203 – Open databases and software. In: *9th ACM Multimedia Systems Conference*, Amsterdam, 2018-06-12/15, pp. 466-471. New York, NY: Association for Computing Machinery. +- [b-Shapley] Shapley, L.S. (1953). A value for $n$ -person games. In: Kuhn, H., Tucker, A., editors. *Contributions to the theory of games*, vol. 2, pp. 307–317. Princeton, NJ: Princeton University Press. + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | Tariff and accounting principles and international telecommunication/ICT economic and policy issues | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling, and associated measurements and tests | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects, next-generation networks, Internet of Things and smart cities | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.1302-201410-I_PDF-E/raw.md b/marked/P/T-REC-P.1302-201410-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..5d0a45eaae23475be6360ab561655cd655cf39b3 --- /dev/null +++ b/marked/P/T-REC-P.1302-201410-I_PDF-E/raw.md @@ -0,0 +1,2007 @@ + + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.1302** + +(10/2014) + +SERIES P: TERMINALS AND SUBJECTIVE AND +OBJECTIVE ASSESSMENT METHODS + +Telemeeting assessment + +--- + +**Subjective method for simulated conversation +tests addressing speech and audio-visual call +quality** + +Recommendation ITU-T P.1302 + +# ITU-T P-SERIES RECOMMENDATIONS + +## TERMINALS AND SUBJECTIVE AND OBJECTIVE ASSESSMENT METHODS + +| | | | +|--------------------------------------------------------------------------------------------------|---------------|---------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Voice terminal characteristics | Series | P.30 | +| | | P.300 | +| Reference systems | Series | P.40 | +| Objective measuring apparatus | Series | P.50 | +| | | P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of speech quality | Series | P.80 | +| | | P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | +| Transmission performance and QoS aspects of IP end-points | Series | P.1000 | +| Communications involving vehicles | Series | P.1100 | +| Models and tools for quality assessment of streamed media | Series | P.1200 | +| Telemeeting assessment | Series | P.1300 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | Series | P.1400 | +| Methods for objective and subjective assessment of quality of services other than voice services | Series | P.1500 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +## Recommendation ITU-T P.1302 + +# Subjective method for simulated conversation tests addressing speech and audio-visual call quality + +## Summary + +Recommendation ITU-T P.1302 describes a subjective method for assessing the quality of simulated speech or audio-visual telephony calls with time-varying transmission conditions. Simulated calls consist of several stretches of speech or audio-visual material which are ordered in a logical sequence, according to a storyline defined by a scenario. After each stretch, test participants have to answer a content-related question, and they have to rate the overall quality of the call at the end of the entire sequence. The method is meant as a complement to standard listening-only or conversational tests, provoking a fixed structure in the material to be judged, while maintain a more-or-less conversational attention focus of the test participants. + +## History + +| Edition | Recommendation | Approval | Study Group | Unique ID* | +|---------|----------------|------------|-------------|---------------------------------------------------------------------------| +| 1.0 | ITU-T P.1302 | 2014-10-29 | 12 | 11.1002/1000/12325 | + +## Keywords + +Audio-visual quality, call quality, conversation quality, speech quality, subjective methodology. + +--- + +\* To access the Recommendation, type the URL in the address field of your web browser, followed by the Recommendation's unique ID. For example, . + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database at . + +© ITU 2015 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## Table of Contents + +| | Page | +|-------------------------------------------------------------------------------------------------------------------------|------| +| 1 Scope..... | 1 | +| 2 References..... | 2 | +| 3 Definitions ..... | 2 | +| 3.1 Terms defined elsewhere ..... | 2 | +| 3.2 Terms defined in this Recommendation..... | 2 | +| 4 Abbreviations and acronyms ..... | 2 | +| 5 Conventions ..... | 3 | +| 6 Subjective test methodology..... | 3 | +| 6.1 Simulated conversation test addressing speech quality..... | 3 | +| 6.2 Simulated conversation tests addressing audio-visual quality ..... | 4 | +| 6.3 Features of the text material ..... | 4 | +| 6.4 Pre-processing of the short segments ..... | 4 | +| 6.5 Questions and post-screening of participants ..... | 4 | +| Appendix I – Exemplary scenarios for simulated conversations addressing speech quality
and audio-visual quality..... | 6 | +| Appendix II – Exemplary questions and answers to the scenarios from Appendix I ..... | 11 | +| Appendix III..... | 41 | +| Bibliography..... | 42 | + + + +# Subjective method for simulated conversation tests addressing speech and audio-visual call quality + +## 1 Scope + +In this Recommendation, a subjective test procedure is described which is able to assess the quality of speech or audio-visual telephone calls, the so-called "call quality". In contrast to standard listening-only tests as described in [ITU-T P.800], call quality refers to the experience of a user during a typically 1-2 min long call. Such experiences are commonly addressed in conversation tests as described in [ITU-T P.800] and [ITU-T P.805], however real conversation tests commonly lack analytical power, because naïve test participants have to concentrate mainly on the flow of the conversation, and are unable to rate instances of good or bad quality. In turn, methods for the continuous assessment of quality as described in [ITU-T P.880] do not put the listener in a situation which is comparable to a conversation. + +In order to circumvent this problem, ETSI has developed a method for "simulated conversation tests" in [b-TR 102 506], which has been adopted in this Recommendation and extended to audio-visual telephony. Simulated conversations consist of a number of short (8-12 s) stretches of speech, typically five, which are presented to the listening or listening-and-viewing test participants, and which follow a logical storyline. After each stretch of speech, the test participant has to answer a short content-related question, e.g., in terms of an answer to a multiple-choice question. Following the answer, the next stretch of speech or audio-visual material is presented, followed by the next question, and so forth until the end of the logical sequence is reached. Upon the end of the sequence, the test participant has to judge the perceived quality of the entire call on a standard (typically five-point ACR according to [ITU-T P.800] and [ITU-T P.911]) quality scale. Those multiple-choice questions may refer to information transmitted via the audio, video or audio-visual modality. + +Quality of short segments can either be predicted with subjective methods, e.g., [ITU-T P.911] for audiovisual segments or [ITU-T P.800] for speech segments, or can be assessed with instrumental models like [ITU-T P.863] or [ITU-T J.341], in order to analyse the relationship between the quality of individual samples and the quality of the entire call as assessed with this Recommendation. + +The method is not intended to replace listening-only tests according to [ITU-T P.800] or listening-and-viewing tests according to [ITU-T P.911], or speech conversation tests according to [ITU-T P.800] or audio-visual conversation tests according to [ITU-T P.920]. Rather, it is recommended to make use of the appropriate method as follows: + +- In case that the listening-only or listening-and-viewing quality is of interest and that the quality is considered to be constant throughout the duration of a speech-only or audio-visual call, the method according to [ITU-T P.800] or [ITU-T P.911] should be used. [ITU-T P.800] is, however, not recommended for longer durations.1 +- In case that the listening-and-viewing quality is of interest and that the quality is considered not to be constant throughout the duration of a speech-only or audio-visual call, the present method should be used; in contrast to the task of rating overall quality in [ITU-T P.880] it addresses also audio-visual material and aims at establishing a more valid situation by introducing active participation of the participants. + +--- + +1 A validation within ITU-T SG12 has revealed no significant differences between the results of a [ITU-T P.911] test and a test according to the present simulated conversation test method results, for a limited number of constant audio, video or audio-visual quality degradations. + +- In case that conversational quality which may be impacted by conversation-specific degradations like echo or delay is of interest, the method according to [ITU-T P.800], [ITU-T P.805] or [ITU-T P.920] should be used. + +## 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +[ITU-T J.341] Recommendation ITU-T J.341 (2011), *Objective perceptual multimedia video quality measurement of HDTV for digital cable television in the presence of a full reference*. + +[ITU-T P.10] Recommendation ITU-T P.10/G.100 (2006), *Vocabulary for performance and quality of service*. + +[ITU-T P.800] Recommendation ITU-T P.800 (1996), *Methods for subjective determination of transmission quality*. + +[ITU-T P.805] Recommendation ITU-T P.805 (2007), *Subjective evaluation of conversational quality*. + +[ITU-T P.863] Recommendation ITU-T P.863 (2011), *Perceptual objective listening quality assessment*. + +[ITU-T P.880] Recommendation ITU-T P.880 (2004), *Continuous evaluation of time-varying speech quality*. + +[ITU-T P.911] Recommendation ITU-T P.911 (1998), *Subjective audiovisual quality assessment methods for multimedia applications*. + +[ITU-T P.920] Recommendation ITU-T P.920 (2000), *Interactive test methods for audiovisual communications*. + +# 3 Definitions + +### 3.1 Terms defined elsewhere + +This Recommendation uses the following terms defined elsewhere: + +**3.1.1 conversational quality** [ITU-T P.10]: Quality with which a bi- or multidirectional conversation is perceived by a communication partner. + +**3.1.2 conversational speech quality** [ITU-T P.10]: Speech quality as experienced in a bi- or multidirectional conversation. + +### 3.2 Terms defined in this Recommendation + +This Recommendation defines the following term: + +**3.2.1 call quality**: Listening quality as perceived by a user at the end of a conversational call. + +## 4 Abbreviations and acronyms + +None. + +## 5 Conventions + +None. + +## 6 Subjective test methodology + +The characteristics of this methodology are as follows: + +- By orally answering the questions related to each stretch of speech, test participants should be made to be involved in a conversation; nevertheless, the situation is a pure listening or listening-and viewing one, allowing the test participants to concentrate more on the form of the media than it would be possible in a real conversation. +- The conversation is highly structured; in fact, it consists of short samples as they are used in standard listening-only or watching and listening-only tests according to [ITU-T P.800] and [ITU-T P.911]. + +Details of the methodology are described in the following paragraphs; the method is thus particularly tailored to situations where the quality varies over the time of a call. + +### 6.1 Simulated conversation test addressing speech quality + +A typical telephony situation is a dialogue between two persons in which both interlocutors have a certain amount of speaking and listening activity. Thus, the test situation is divided into parts with listening-only activity and parts with speaking activity. It is supposed that both participants in a conversation are interested in the conversation topic; in the test, this is realized by the test participant being asked for certain aspects of the content, thus to direct the focus of attention to the content of the presented speech stimuli, as it would be in a natural conversation situation. + +In order to ensure the ecological validity of the test, typical contents of telephone conversations are chosen (e.g., request for a rental car, discussing the furniture in a room). The realization of such a simulated conversation consists of a series of shorter "utterances", which have a pause between them for interaction but are connected logically with regard to the content of the presentation. Instead of the own speech activity as a reply, an artificial content-directed task is to be done by the test participants. For example, test participants have to answer a question related to the content of the past stimulus, which can be done by keyword spotting. + +The speech samples used for the test should address topics like in a natural telephone situation. A simulated conversation consists of five samples of approximately 8-12 seconds and is constructed in such a way that four breaks are possible. After the fifth test stimulus, the participant is asked for an assessment for the overall call quality of the simulated call, instead of the content-related question. The questions asked to the participants after each sample of the simulated conversations and the corresponding answer options are presented on paper or on a computer screen after each sample. For each sample, a set of different questions is recommended in order to prevent test participants from memorizing the answers. The position of the right answer should also be randomly changed in order to avoid participants answering mechanically. The number of multiple choice answer options may vary between three and five. Simulated calls should be constructed for at least four speakers balanced in gender and comprising different scene backgrounds and conversation topics. + +![Schematic representation of the simulated conversation structure](99bae07626f60f9ede10e2e387ef7051_img.jpg) + +The diagram illustrates the temporal structure of a simulated conversation. It consists of a horizontal bar divided into nine alternating segments: five dark grey segments and four white segments. Above the first dark grey segment is the label "Pre-recorded processed segment" with a downward arrow. Above the first white segment is the label "Pause with content-oriented task" with a downward arrow. Above the final dark grey segment is the label "Quality assessment" with a rightward arrow. + +Schematic representation of the simulated conversation structure + +**Figure 1 – Schematic representation of the simulated conversation structure** + +Figure 1 shows an exemplary structure of such a simulated conversation, using samples of 12 second length. The used speech material consists of five speech parts (samples), which correspond to the utterances of one party. After a 12 seconds speech sample there is a 12 seconds pause during which the test person had to perform a content regarding task. At the end of the sequence a score for the Call Quality is solicited from the test participants using an absolute category rating scale as described in [ITU-T P.800]. Thus, the overall length of the simulated conversation is approx. 120 seconds, including the answering and quality-rating phases. + +### 6.2 Simulated conversation tests addressing audio-visual quality + +Similar to the methodology for speech-only call quality, a simulated videotelephony conversation is constructed as a set of five consecutive samples of short video clips. The samples are linked by the storyline. Participants are asked to watch each clip and then to verbally answer a question about the audiovisual content of the clip. The question could be related to the auditory content only, to the video content only, or to the audiovisual one, that is to say, a question that could be answered by using clues from both visual and auditory channels. The questions asked to the participants after each sample of the simulated conversations and the corresponding answer options are presented on a computer screen after each sample. The number of multiple choice answer options may vary between three and five. For each sample, a set of different questions is recommended in order to prevent test participants from memorizing the answers. The position of the right answer should also be randomly changed in order to avoid participants answering mechanically. + +Audiovisual clips should be composed of at least four speakers balanced in gender, comprising different scene backgrounds and conversation topics. The clips should involve the video channel by the means of visual cues (i.e., showing objects to the camera, pointing dates on a wall calendar, body gestures), so that the test participants have to pay attention to the video channel. The video clips and the time period allowed to answer each question should have approximately the same length, leading to a total duration of 10 times the stimulus length or 8-12 seconds. At the end of each simulated conversation, test participants have to rate the audiovisual quality of the entire dialog using a standard overall quality scale, as recommended in [ITU-T P.911]. + +### 6.3 Features of the text material + +When creating text material for recording the short stimuli, a whole scenario should be decided upon first, so each stimulus (utterance) represents an authentic sequel of the previous stimuli. Validated so far have been scenarios using a narrative form, i.e., without implying interlocutor's verbal action. In that way, listeners/raters are not confused when just answering questions instead of reacting on the stimuli, and two short utterances can be combined to one longer stimulus, if desired. + +Also, each utterance should provide more than one important piece of information. That allows for more questions, but also prohibits a too easy listening task. + +For the audio-visual tests, information should be spread on the two modalities, and accordingly, questions should be drafted explicitly addressing either audio or video transmitted information. See Appendices I and II for examples. + +### 6.4 Pre-processing of the short segments + +The original samples should be processed to obtain transmission quality as needed. As this method aims at time-varying transmission characteristics, changes in transmission characteristics can be induced between or within the short segments. Example profiles already used are illustrated in Appendix III. + +### 6.5 Questions and post-screening of participants + +During the development of this method, the participant's answers have been recorded. But it has been shown that noting errors during a session was easily manageable by the instructor to automatically + +calculate the error rates. The purpose of this post-evaluation screening is to find possible outliers, i.e., participants who obviously did not concentrate sufficiently on the content or all modalities provided in order to give valid ratings of call quality. As with strongly degraded transmission, errors may occur independent of concentration, it is suggested to define the exclusion criterion depending on the performance of subjects throughout the test. The alternative answers should be displayed on a screen without alpha-numerical labels to prevent participants just referring to these labels, but uttering (at least parts of) the answer. + +Furthermore, while questions concerning audio, video or audio-visual content can be used, without observing major difference in the perceived quality of the samples, a combination of audio and video content questions is preferred in order to prevent priming towards one modality. + +## Appendix I + +### Exemplary scenarios for simulated conversations addressing speech quality and audio-visual quality + +(This appendix does not form an integral part of this Recommendation.) + +This Recommendation is based on experiments, which were originally partially carried out in German. Because of this, the original German scenario descriptions are added here for completeness. In principle, experiments following this procedure should be carried out in the language under consideration, using scenario description in that language. A re-translation of the English scenario descriptions into German might provide slightly different results when compared to the original experiments. + +The following scenarios are designed to combine sentences (e.g., two) to one single stimulus. + +### Scenario 1: Scheduling medical treatment + +- 1) Hello, this is Markus Lehmann. I would like to make an appointment concerning the treatment of my back complaint as discussed. [point to your back] +- 2) We had agreed on six weeks with two radiation sessions each. I would like to start the treatment now. [go to the wall calendar] +- 3) In the morning I am always busy. But on Tuesdays and on Thursdays would be fine. [point to Tuesday and Thursday on the calendar] +- 4) There is no way to come on Wednesdays, as I attend a language course at these days. I would like best to have regular appointments. [point to a huge blocker on Wednesday] +- 5) Besides, do you think that additional gymnastic exercises could decrease my afflictions in the long run? [do some exercises] +- 6) I was thinking about an introduction under the tutelage of a physiotherapist. Maybe, that could be a good complement to the radiation. [do some more exercises] +- 7) I hope the treatment costs will be fully paid by the health insurance. For this treatment, however, I do not know for certain. +- 8) Yesterday, I read in a newspaper once more about the approval of new austerity plan, mainly concerning the amount of assumption of costs. [show a newspaper] +- 9) Additionally, I would like to make an appointment with your partner in the group practice, the dentist Dr Kollwitz. +- 10) The date can even be next month. I would actually prefer it next month, as I am very busy this one. [point to the end of the month on a calendar] + +(German) + +- 1) Guten Tag, hier ist Marcus Lehmann. Ich hätte gerne einen Termin bei Ihnen wegen der besprochenen Behandlung meiner Rückenbeschwerden. [Zeigt auf seinen Rücken] +- 2) Wir hatten abgestimmt, dass ich sechs Wochen lang je zwei Bestrahlungen bekommen soll. Ich würde die Behandlung nun gerne beginnen. [Geht zum Wandkalender] +- 3) Auf keinen Fall kann ich vormittags. Dienstags und donnerstags wäre es mir sehr recht. [zeigt am Wandkalender auf Dienstag und Donnerstag] +- 4) Mittwoch ist bei mir unmöglich, da ich an diesem Tag einen Sprachkurs besuche. Am liebsten wären mir regelmäßige Termine. [Zeigt am Wandkalender auf Mittwoch, wo ein großer Blocker steht] + +- 5) Außerdem möchte ich Ihre Meinung dazu hören, ob zusätzliche Gymnastik meine Beschwerden vielleicht langfristig verringern könnte. [Macht etwas Gymnastik in die Kamera] +- 6) Ich denke an eine Einführung unter Anleitung einer Krankengymnastin. Vielleicht ist das eine gute Ergänzung zu den Bestrahlungen. [Macht weiterhin Gymnastik] +- 7) Ich hoffe, die Kosten der Behandlung werden komplett von der Krankenkasse übernommen. Allerdings weiß ich nicht, ob es auch meine Behandlung betrifft. +- 8) Gestern habe ich wieder in der Zeitung gelesen, dass neue Sparpläne verabschiedet worden sind, die hauptsächlich die Höhe der Eigenbeteiligung betreffen. [Hält eine Zeitung in die Kamera] +- 9) Zusätzlich möchte ich noch einen Gesprächstermin mit Ihrer Praxispartnerin, der Zahnärztin Dr. Kollwitz, vereinbaren. +- 10) Der Termin kann ruhig erst im nächsten Monat liegen. Das ist mir sogar lieber, da ich diesen Monat sehr beschäftigt bin. [Zeigt am Wandkalender das Ende des Monats] + +### **Scenario 2: Car Rental** + +- 1) Hello, this is Maria Eibig and I would like to rent a car. [grab a catalogue and show it] +- 2) A bigger estate car would be sufficient. But if a van is cheaper I would also take one of those. +- 3) I would prefer midweek, that is Wednesday. At that day I would need the car close to half past 8 pm. [show a picture with an estate car] +- 4) But if you offer a special rate for the weekend, I would do my transport then. [point to a wall calendar behind you] +- 5) For me it is important to take out an assurance for the car. I would also take the small trolley as per your offer. [point to the weekend on the calendar] +- 6) Additionally, I would like to have a mileage allowance for more than 500 kilometres and blankets as packaging material. +- 7) I would like to pay with my credit card. My customer number is 24 12. [show your credit card] +- 8) Last time I rented a Golf. That was beginning of last month. [point at the begin of the month on the calendar] +- 9) That time I returned the car fully fuelled. This time I will not manage to do so, as I will come back not until the evening. [shake your head] +- 10) I will then bring the car directly to your car park and return the following morning to do the billing. [point to a certain position on a map] + +(German) + +- 1) Guten Tag, hier ist Maria Eibig und ich möchte bei Ihnen gern einen Wagen reservieren. [Hole die Broschüre eines Autoverleihers heraus und halte sie in die Kamera] +- 2) Ein größerer Kombi reicht mir dafür aus, wenn ein Kleintransporter preiswerter ist, würde ich auch diesen nehmen. [Zeige ein Bild mit einem Kombi] +- 3) Mir wäre Mitte der Woche, also Mittwoch, am liebsten. Da würde ich den Wagen gegen zwanzig Uhr dreißig benötigen. [Zeige dabei auf einen Wandkalender hinter Dir] +- 4) Wenn sie einen Spezialtarif am Wochenende anbieten, würde ich den Transport auch am Wochenende machen. [Zeige am Wandkalender auf das Wochenende] +- 5) Wichtig ist mir noch der Abschluss einer Versicherung für das Fahrzeug. Den kleinen Rollwagen aus Ihrem Programm nehme ich ebenfalls mit. + +- 6) Außerdem brauche ich eine Kilometerpauschale von über 500 Kilometern und Decken als Verpackungsmaterial. +- 7) Ich habe eine Kreditkarte mit der ich den Wagen bezahlen möchte. Meine Kundennummer ist 24 12. [Zeige die Kreditkarte in die Kamera] +- 8) Zuletzt hatte ich mir einen Golf geliehen. Das war am Anfang des letzten Monats. [Zeige im Kalender auf den Monatsbeginn] +- 9) Das letzte Mal hatte ich den Wagen vollgetankt zurückgegeben. Diesmal werde ich das wohl nicht schaffen, da ich erst abends zurückkomme. [Kopf schütteln] +- 10) Ich fahre den Wagen dann gleich auf Ihren Firmenparkplatz und komme dann am darauffolgenden Morgen zu Ihnen zur Abrechnung. [Zeige auf einen Punkt auf einer Karte] + +### **Scenario 3: Kitchen purchase order** + +- 1) Hello. I would like to order a built-in kitchen. Would it be possible to talk to Mrs Müller? [show a catalogue] +- 2) She consulted me in your salesrooms and we loosely agreed upon calling her until Wednesday. +- 3) I decided on the version in light grey. Giving a thought to it, a white kitchen appeared to be too cold to me. [show a picture of a light grey kitchen] +- 4) Furthermore, I would like to have the doorknobs with the lozenge shape, as the round shape is a little bit too small in my opinion. [show a detailed picture of the doorknob] +- 5) Concerning the sink I consider taking a single unit. Thereby there is still space for a dishwasher. [show a picture of a sink] +- 6) I decided on the small dishwasher from the brand "Futura". Do you have one of these on offer? [show a picture of a small washer] +- 7) I suppose this washer provides aqua-stop. If not, I would rather take the next bigger model. [point to a "Aqua-Stop" label; alternatively change to "energy label"] +- 8) I am not decided yet on the cooker. I would like best to have one with glass-ceramic top and sensor circuitry. [point to top in the catalogue] +- 9) I am fine with the table. I would like to have one in the size of 100x160 cm. [show roughly the size with your hands in the air] +- 10) If I understood the catalogue correctly, it can be extended to the length of 200 cm. [show the extended size] + +(German) + +- 1) Hallo, ich möchte bei Ihnen gerne eine Einbauküche bestellen. Ist es möglich, dass ich mit Frau Müller spreche? [Zeigt eine Küchenbroschüre in die Kamera] +- 2) Sie hat mich in Ihren Verkaufsräumen beraten und ich hatte grob mit Ihr abgesprochen, dass ich mich bis Mittwoch bei Ihr melden werde. +- 3) Ich habe mich für die Variante in Hellgrau entschieden. Eine weiße Küche kam mir bei längerer Überlegung doch zu kalt vor. [Zeigt das Bild einer hellgrauen Küche in die Kamera] +- 4) Außerdem möchte ich alle Türknäufe in Rautenform, die Runden sind mir doch ein wenig zu klein. [Zeigt das Detailbild von Türknäufen] +- 5) Für die Spüle habe ich mir überlegt, doch nur ein Becken zu nehmen. Dadurch bleibt Platz für eine Spülmaschine. [Zeigt das Bild eines Spülbeckens] +- 6) Ich habe mich für die schmale Maschine der Firma Futura entschieden. Haben Sie so ein Gerät im Angebot? [Zeige eine schmale Maschine im Katalog] + +- 7) Ich nehme an, dass diese Maschine über einen Aqua-Stop verfügt. Falls nicht möchte ich lieber das nächst größere Modell. [Zeige Aqua-Stop Label im Katalog; alternativ ändern in Energiesparlabel] +- 8) Bei dem Herd bin ich mir noch nicht sicher. Am liebsten hätte ich ein Gerät mit Glaskeramik Kochfeldern und Sensorschaltung. [zeige Kochfeld im Katalog] +- 9) Der Tisch gefällt mir gut. Ich hätte ihn gerne in der Größe 100 x 100 cm. [Zeigt die gewünschte Größe in der Luft] +- 10) Wenn ich den Prospekt richtig verstehe lässt er sich auf eine Länge von 150 cm vergrößern. [Zeigt eine Vergrößerung in die Luft] + +#### **Alternative Scenario 4: Birthday party** + +- 1) Hi Jan. Last Saturday I went to Peter's birthday party. It was really great! [please wave and make an inviting facial expression] +- 2) There I met an old acquaintance. Here is a photo of our old friends with me. This is Stefan, Linda, and Markus. [show the photo and point to Stefan, Linda, and Markus each at a time] +- 3) After dinner a band arrived at the party and played music. I believe the guests really enjoyed it. Spirits were running high. Here is a shot. [show photo of people dancing in front of a band] +- 4) Do you remember Julia from our time at school? The two of us danced in the garden and she even taught me a break-dance move. Have a look. [show a photo of Julia and do a dance move] +- 5) At midnight we all gathered around a big cake to sing "happy birthday". The cake was made by Peter's mother. It was marvellous. +- 6) Before having the cake Peter received his presents. He got a drum set jointly from his parents and friends. Here is a photo. [show a picture from a drum set] +- 7) I talked with some of our old friends about the good old days at school and told each other funny story from back then. We laughed quite a bit. [smile and laugh] +- 8) When Peter's brother had to go to sleep, we sang a children's song, so he could fall asleep. [sing a children's song] +- 9) Then, Peter's parents told me details about their journey to Asia. They also told me how to use eating sticks. Like this. [show using eating sticks] +- 10) I would like to meet you this week eventually and talk a while. I wish you well. Goodbye. [wave goodbye and smile] + +(German) + +- 1) Hallo Jan! Ich war letzten Samstag auf der Geburtstagsparty von Peter. Da war es wirklich super! [bitte winke und mache ein einladendes Gesicht] +- 2) Ich habe dort viele alte Bekannte getroffen. Da ist ein Foto von mir und unseren alten Freunden. Da ist Stefan, Linda und Markus. [zeige das Foto und zeige jeweils auf Stefan, Linda und Markus] +- 3) Nach dem Essen kam noch eine Band auf die Party und hat Musik gemacht. Ich glaube, die Gäste fanden das richtig gut. Die Stimmung war ziemlich ausgelassen. Da ist ein Foto. [zeige ein Bild auf dem viele Leute vor einer Band tanzen] +- 4) Erinnerst Du Dich noch an Julia aus der Schulzeit? Wie haben zusammen im Garten getanzt und sie hat mir sogar ein Break-Dance Move gezeigt. Schau mal! [zeige das Bild von Julia und mache eine Tanzbewegung] +- 5) Um Mitternacht haben wir uns dann alle um einen großen Kuchen versammelt, um Happy-Birthday zu singen. Den Kuchen hat Peters Mutter gemacht, der war fantastisch. [zeige ein Bild vom Kuchen.] + +- 6) Vor dem Kuchenessen haben wir dann alle Peter seine Geschenke gegeben. Er hat ein Drumset gemeinsam von seinen Eltern und seinen Freunden bekommen. Hier ist das Foto. [zeige ein Bild vom Schlagzeug] +- 7) Einige alte Freunde und ich haben uns über die gute alte Zeit an der Schule gesprochen und uns lustige Geschichten von damals erzählt. Wir haben ziemlich viel gelacht. [Lächeln und lachen] +- 8) Als Peters Bruder ins Bett musste, haben wir ein Kinderlied gesungen, damit er einschläft. [singe ein Kinderlied] +- 9) Peters Eltern haben mir noch Genaueres von ihrer Reise nach Asien erzählt und mir gezeigt, wie man mit Stäbchen isst! So ungefähr. [zeige die Stäbchen] +- 10) Ich hoffe, wir können uns noch in dieser Woche mal treffen und etwas länger reden. Mach es gut, Tschüss! [winke zum Abschied und zeige ein natürliches Lächeln] + +## Appendix II + +## Exemplary questions and answers to the scenarios from Appendix I + +(This appendix does not form an integral part of this Recommendation.) + +Questions are separated in those concerning information from the audio (A), video (V) or both channels (AV). These should be mixed for the AV condition. The order of the answers should be randomized. Correct answers are bold. Subject making errors frequently should be excluded from analysis. Gender of the speaker has to be adjusted. If two subsequent sentences are presented a one stimulus, the short answers can be extended up to whole sentences. + +### Scenario 1: Scheduling medical treatment + +#### 1.1A Why does the patient call? + +- a) he wants to speak to an employee +- b) he misdialled +- c) **he wants to get an appointment for a treatment** +- d) he needs a prescription +- e) he wants a vaccination + +#### 1.2AV What medical condition does he have? + +- a) toothache +- b) **backache** +- c) twisted ankle +- d) he complaints about the treatment +- e) gnawing headache + +#### 1.3V Which body part does the patient show? + +- a) **back** +- b) shoulder +- c) wrist +- d) elbow +- e) neck + +#### 2.1A What does the patient want to do with respect to the treatment? + +- a) **start it** +- b) pay for it +- c) stop it +- d) interrupt it +- e) change it + +#### 2.2A How long should the treatment be? + +- a) 1 week +- b) 1 year +- c) **6 weeks** +- d) 2 weeks +- e) 5 days + +**2.3A What kind of treatment is she supposed to get?** + +- a) pills +- b) radiation** +- c) infusions + +**2.4V Which date is shown on the calendar? (correct answer depends of material preparation)** + +- a) Monday, 15th +- b) Tuesday, 16th +- c) Wednesday, 17th +- d) Thursday, 18th +- e) Friday, 19th + +**3.1AV Which days does the patient prefer?** + +- a) Wednesdays only +- b) work days +- c) Tuesdays and Thursdays** +- d) 13th and 14th of February +- e) weekends + +**3.2A When can he not manage to come?** + +- a) weekends +- b) mornings** +- c) next week +- d) on foul weather +- e) in the evening + +**3.3V Which days are already marked in the calendar?** + +- a) Monday and Tuesday +- b) Tuesday and Thursday +- c) Thursday and Saturday +- d) Wednesday** + +**4.1A Why does he not want to come on Wednesdays?** + +- a) appointment with a dentist +- b) soccer training +- c) language course** +- d) taking care of his grandmother +- e) working as newspaper carrier + +**4.2A What else is important?** + +- a) short duration of treatment +- b) an electronic bill +- c) regular appointments** +- d) a short avenue + +#### **5.1A What else does he request?** + +- a) a bid +- b) an opinion on gymnastic exercises** +- c) sports medical examination +- d) admission to a hospital +- e) attestation to be incapable of work + +#### **5.2A What does he wants to be informed about?** + +- a) additional exercises** +- b) additional acupuncture +- c) taking a cure +- d) new surgery method +- e) new laser treatment + +#### **5.3V Which movement does he make?** + +- a) arm turn +- b) neck turn +- c) pelvis turn +- d) back turn** + +#### **6.1A Gymnastic exercises are ... ?** + +- a) under tutelage of a physiotherapist** +- b) conducted in a group +- c) explained by the doctor +- d) for free +- e) in the neighbourhood + +#### **6.2A What does he expects from the exercises?** + +- a) get to know like-minded patients +- b) allaying of pain +- c) good complement to the treatment** +- d) rapid improved performance +- e) an amusing evening + +#### **6.3V Which movement does he make?** + +- a) arm turn +- b) squat +- c) pelvis turn +- d) back turn** + +#### **7.1A The patent is concerned about ...?** + +- a) potential close-down of the doctor's practice +- b) the treatment being fully paid by insurance** +- c) savings for the family insurances +- d) the meaningfulness of his treatment + +- e) increase of medication prices + +#### **8.1A Which costs is the insurance supposed to borne?** + +- a) **complete costs** +- b) just necessary medicine +- c) lease costs for the practice +- d) parking fee during treatment +- e) travel expenses + +#### **8.2A Who is to bear the expenses?** + +- a) occupation cooperative +- b) **health insurance** +- c) insurance for accident costs +- d) employer +- e) his wife + +#### **8.3v What does he show to the camera?** + +- a) **newspaper** +- b) magazine +- c) catalogue +- d) straps +- e) crutches + +#### **9.1A What does he want to arrange?** + +- a) appointment with a psychotherapist +- b) discussion about cost sharing +- c) **appointment with a dentist** +- d) attendance at a scientific conference +- e) appointment for a heating repair + +#### **9.2A Who is Dr Kollwitz?** + +- a) **partner in the group practice** +- b) wife of the doctor +- c) wife of the patient +- d) another patient +- e) sister of the patient + +#### **10.1A Why does he want an appointment later?** + +- a) fear of visiting a dentist +- b) missing the appropriate funds currently +- c) **he is very busy currently** +- d) dentist does not have appointments on short notice +- e) he has to contact his insurance company first + +#### **10.2A Why does he want to visit the dentist apparently?** + +- a) root canal therapy +- b) routine checkup** +- c) refund of costs +- d) complaint about wrong diagnosis +- e) complaint about poor quality + +#### **10.3V Which days are marked in the calendar?** + +- a) Monday to Wednesday** +- b) Tuesday to Thursday +- c) Wednesday to Friday + +(German) + +#### **1.1A Weshalb ruft der Patient an?** + +- a) er möchte einen Mitarbeiter sprechen +- b) er hat sich verwählt +- c) er möchte einen Termin zur Behandlung** +- d) er braucht ein Rezept +- e) er braucht eine Impfung + +#### **1.2AV Welche Beschwerden hat die Patientin?** + +- a) Zahnschmerzen +- b) Rückenbeschwerden** +- c) einen verstauchten Fuß +- d) er beschwert sich über die Behandlung +- e) bohrenden Kopfschmerz + +#### **1.3V Welches Körperteil zeigt die Patientin?** + +- a) Rücken** +- b) Schulter +- c) Handgelenk +- d) Ellbogen +- e) Nacken + +#### **2.1A Möchte die Patientin die Therapie ...** + +- a) ... beginnen?** +- b) ... bezahlen? +- c) ... beenden? +- d) ... unterbrechen? +- e) ... ändern? + +#### **2.2A Wie lange soll die Behandlung dauern?** + +- a) 1 Woche +- b) 1 Jahr +- c) 6 Wochen** + +- d) 2 Wochen +- e) 5 Tage + +#### **2.3A Was für eine Behandlung soll sie bekommen?** + +- a) Tabletten +- b) Bestrahlungen** +- c) Infusionen + +#### **2.4V Welches Datum ist in dem Kalender zu sehen?** + +- a) Montag, der 15. +- b) Dienstag, der 16. +- c) Mittwoch, der 17. +- d) Donnerstag, der 18. +- e) Freitag, der 19. + +#### **3.1AV Welche Tage bevorzugt die Patientin?** + +- a) nur mittwochs +- b) werktags +- c) dienstags und donnerstags** +- d) den 13. und 14. Februar +- e) am Wochenende + +#### **3.2A Wann kann die Patientin nicht?** + +- a) am Wochenende +- b) vormittags** +- c) in der nächsten Woche +- d) bei schlechter Witterung +- e) am Abend + +#### **3.3V Welche Tage sind im Kalender markiert?** + +- a) Montag und Dienstag +- b) Dienstag und Donnerstag +- c) Donnerstag und Samstag +- d) Mittwoch** + +#### **4.1A Warum kann der Anrufer mittwochs nicht?** + +- a) Termin beim Zahnarzt +- b) Fußballtraining +- c) Sprachkurs** +- d) Betreuung der Großmutter +- e) Arbeit als Zeitungsausträger + +#### **4.2A Was ist ihm noch wichtig?** + +- a) eine kurze Behandlungsdauer +- b) eine elektronische Rechnung +- c) regelmäßige Termine** + +- d) eine kurze Anreise + +#### **5.1A Was möchte die Patientin außerdem vom Arzt?** + +- a) einen Kostenvoranschlag +- b) **seine Meinung zu zusätzlicher Gymnastik** +- c) ein sportmedizinisches Gutachten +- d) eine Einweisung ins Krankenhaus +- e) eine Arbeitsunfähigkeitsbescheinigung + +#### **5.2A Über welchen Nutzen will sich der Anrufer informieren?** + +- a) **von zusätzlicher Gymnastik** +- b) von zusätzlicher Akupunktur +- c) eines Kuraufenthalts +- d) einer neuen Operationsmethode +- e) einer Laserbehandlung + +#### **5.3V Welche Bewegung macht die Patientin?** + +- a) Armdrehung +- b) Nackendrehung +- c) Beckendrehung +- d) **Rückendrehung** + +#### **6.1A Wie sollen die Gymnastikkurse sein?** + +- a) **mit Anleitung durch eine Krankengymnastin** +- b) Übung in einer größeren Gruppe +- c) Einweisung vom Arzt selbst +- d) kostenfrei +- e) nahe seines Wohnortes + +#### **6.2A Was erhofft sich die Patientin von den Kursen?** + +- a) gleichgesinnte Patienten kennenzulernen +- b) Schmerzminderung +- c) **gute Ergänzung der Behandlung** +- d) kurzfristige Leistungssteigerung +- e) einen unterhaltsamen Abend + +#### **6.3V Welche Bewegung macht die Patientin?** + +- a) Armdrehung +- b) Kniebeuge +- c) Beckendrehung +- d) **Rückendrehung** + +#### **7.1A Worüber ist die Patientin verunsichert?** + +- a) mögliche Schließung der Arztpraxis +- b) **mögliche Erhöhung der Eigenbeteiligung** +- c) Einsparungen bei der Familienversicherung + +- d) ob seine Behandlung überhaupt sinnvoll ist +- e) Erhöhung der Arzneimittelpreise + +#### **8.1A Welche Kosten soll die Krankenkasse übernehmen?** + +- a) **kompletten Behandlungskosten** +- b) nur notwendige Medikamente +- c) Mietkosten der Arztpraxis +- d) Parkgebühren während der Behandlungszeiten +- e) Fahrtkosten zur Behandlung + +#### **8.2A Wer soll die Kosten der Behandlung übernehmen?** + +- a) die Berufsgenossenschaft +- b) **die Krankenkasse** +- c) die Unfallkasse +- d) der Arbeitgeber +- e) seine Frau + +#### **8.3v Was zeigt die Patient vor der Kamera?** + +- a) **Zeitung** +- b) Magazine +- c) Broschüre +- d) Bänder +- e) Krücken + +#### **9.1A Was möchte die Patientin vereinbaren?** + +- a) Termin mit dem Psychotherapeuten +- b) Gespräch wegen der Kostenbeteiligung +- c) **Termin mit der Zahnärztin** +- d) Teilnahme an einem wissenschaftlichen Kongress +- e) Termin zur Heizungsreparatur + +#### **9.2A Wer ist Frau Dr. Kollwitz?** + +- a) **Praxispartnerin des Arztes** +- b) Ehefrau des Arztes +- c) Ehefrau des Patienten +- d) Patientin +- e) Schwester des Patienten + +#### **10.1A Warum wünscht die Patientin einen späteren Termin?** + +- a) Angst vor dem Zahnarztbesuch +- b) im Moment fehlen die finanziellen Mittel +- c) **sie ist sehr beschäftigt** +- d) Zahnärztin vergibt keine kurzfristigen Termine +- e) sie muss erst bei der Krankenkasse anfragen + +#### **10.2A Warum will sie anscheinend zum Zahnarzt?** + +- a) Wurzelbehandlung +- b) Routineuntersuchung** +- c) Kostenrückerstattung +- d) Beschwerde über eine falsche Diagnose +- e) Beschwerde über mangelhafte Leistungen + +#### **10.3V Welche Tage sind in dem Kalender markiert?** + +- a) Montag bis Mittwoch** +- b) Dienstag bis Donnerstag +- c) Mittwoch bis Freitag + +### **Scenario 2: Car Rental** + +#### **1.1A What does the customer need?** + +- a) a car** +- b) information +- c) a lift +- d) tickets for the opera +- e) an electrician + +#### **1.2A The customer wants to ...** + +- a) sell a car +- b) buy a car +- c) rent a car** +- d) repair a car +- e) return a car + +#### **1.3V What is the name of the magazine?** + +- a) Autonet** +- b) Automobile Mag +- c) Cars +- d) Automobile News +- e) Cars Week + +#### **2.1AV What does the customer need?** + +- a) sweets +- b) a bigger estate car** +- c) a camper +- d) a motorcycle +- e) a cabriolet + +#### **2.2AV What alternative type of car would be OK?** + +- a) small bus +- b) small van** +- c) compact car + +- d) large limousine +- e) pickup + +#### **2.3V What colour was the car?** + +- a) grey +- b) **black** +- c) green +- d) red +- e) white + +#### **3.1AV On which day does she need it?** + +- a) yesterday +- b) next Sunday +- c) **Wednesday** +- d) today +- e) Saturday + +#### **3.2AV When does she want to rent it?** + +- a) in the morning +- b) **8.30 p.m.** +- c) in the morning or in the evening +- d) at noon +- e) in the morning or at midday + +#### **4.1AV On what other day could she take the car?** + +- a) **Weekend** +- b) Wednesday +- c) Tuesday +- d) Thursday +- e) Friday + +#### **4.2A In what case could she even agree with the weekend?** + +- a) traffic jam +- b) **special rate** +- c) bad weather +- d) complementary insurance +- e) never + +#### **4.3AV Which days are marked in the calendar? (depends on the calendar)** + +- a) Monday and Tuesday +- b) Monday and Wednesday +- c) Tuesday and Thursday +- d) Tuesday and Friday +- e) None + +#### **5.1A What is important to the caller?** + +- a) price +- b) **insurance** +- c) driving comfort +- d) colour +- e) year of construction + +#### **5.2A What else does she want to rent?** + +- a) elevating truck +- b) carrying strap +- c) **trolley** +- d) tail pulley +- e) belt + +#### **6.1A What kind of packaging material does she want?** + +- a) packaging foil +- b) boxes +- c) **blankets** +- d) paper +- e) foam material + +#### **6.2A How much mileage allowance does she want?** + +- a) 10 km +- b) 100 km +- c) **500 km** +- d) 5000 km +- e) 50 km + +#### **7.1AV What does the customer show?** + +- a) customer number +- b) **credit card** +- c) membership number for automobile association +- d) telephone number +- e) address + +#### **7.2AV What paying facility does she choose?** + +- a) cash on delivery +- b) **credit card** +- c) advanced payment + +#### **7.3V Which card was used?** + +- a) Maestro +- b) **Visa** +- c) Master Card +- d) Cirrus + +e) American Express + +#### **8.1A What car did she rent last time?** + +- a) BMW +- b) **GOLF** +- c) FORD +- d) POLO +- e) PORSCHE + +#### **8.2AV When did she rent a car last time?** + +- a) never +- b) last summer +- c) **beginning of the month** +- d) last week +- e) 1999 + +#### **8.3V Which day is marked in the calendar? (depends of the calendar)** + +- a) Monday +- b) Tuesday +- c) Wednesday +- d) Thursday +- e) Friday + +#### **9.1AV When does she want to be billed?** + +- a) not at all, because of debit payment +- b) Friday +- c) **the next morning** +- d) directly when returning the car +- e) in the next afternoon + +#### **9.2A In which condition was the car returned last time?** + +- a) washed +- b) **fully fuelled** +- c) without the boxes +- d) damaged +- e) too late + +#### **9.3V Which day is marked? (depends of the calendar)** + +- a) Monday +- b) Tuesday +- c) Wednesday +- d) Thursday +- e) Friday + +#### **10.1A When will she return the car?** + +- a) **at the evening** +- b) in the morning +- c) never +- d) midday +- e) the next day + +#### **10.2A Where does she return the car to?** + +- a) petrol station +- b) home +- c) **car park** +- d) basement garage +- e) companies entrance + +#### **10.3V What does she show?** + +- a) **a card** +- b) a bill +- c) car keys +- d) a calendar +- e) a credit card + +(German) + +#### **1.1A Was möchte der Kunde?** + +- a) **einen Wagen** +- b) eine Auskunft +- c) eine Mitfahrgelegenheit +- d) Opernkarten +- e) einen Elektriker + +#### **1.2A Der Kunde will einen Wagen ...** + +- a) ... verkaufen +- b) ... kaufen +- c) **... reservieren** +- d) ... reparieren lassen +- e) ... zurückgeben + +#### **1.3V Wie heißt die Zeitschrift Magazin?** + +- a) **... Autonet** +- b) ... Automobil mag +- c) ... Autobild +- d) ... Autoexpress +- e) ... Autoweek + +#### **2.1AV Was braucht der Kunde?** + +- a) Gummibärchen +- b) größeren Kombi** +- c) Campingwagen +- d) Motorrad +- e) Cabrio + +#### **2.2AV Welche Fahrzeuggröße benötigt er mindestens?** + +- a) kleinerer Bus +- b) größerer Kombi** +- c) Kompaktwagen +- d) größere Limousine +- e) Pritschenwagen + +#### **2.3V Was ist die Autofarbe?** + +- a) grau +- b) schwarz** +- c) grün +- d) rot +- e) weiß + +#### **3.1AV An welchem Tag möchte er vorbeikommen?** + +- a) am vorhergehenden Tag +- b) nächsten Sonntag +- c) kommenden Mittwoch / Mittwoch** +- d) heute +- e) Samstag + +#### **3.2AV Wann würde er den Wagen abholen?** + +- a) früh morgens +- b) 20:00h** +- c) morgens oder abends +- d) 12 h +- e) morgens oder mittags + +#### **4.1AV An welchen alternativen Termin könnte der Wagen auch abgeholt werden?** + +- a) Wochenende** +- b) Mittwoch +- c) Dienstag +- d) Donnerstag +- e) Freitag + +**4.2A In welchem Fällen könnte er den Wagen sogar am Wochenende abholen?** + +- a) bei Stau +- b) bei einem Spezialtarif** +- c) bei schlechtem Wetter +- d) bei einer Zusatzversicherung +- e) nie + +**4.3AV Welche Tage sind am Kalender markiert?** + +- a) Montag und Dienstag +- b) Montag und Mittwoch +- c) Dienstag und Donnerstag +- d) Dienstag und Freitag +- e) Keiner + +**5.1A Was ist wichtig für den Kunden?** + +- a) Preis +- b) Versicherung** +- c) Fahrkomfort +- d) Farbe +- e) Baujahr + +**5.2A Was soll zusätzlich gemietet werden?** + +- a) Hubwagen +- b) Tragegurt +- c) Rollwagen** +- d) Flaschenzug +- e) Befestigungsgurte + +**6.1A Was für Verpackungsmaterial benötigt der Kunde?** + +- a) Verpackungsfolie +- b) Kartons +- c) Decken** +- d) Papier +- e) Schaumstoff + +**6.2A Wie hoch soll die Kilometerpauschale mindestens sein?** + +- a) 10 km +- b) 100 km +- c) 500 km** +- d) 5000 km +- e) 50 km + +#### **7.1AV Was zeigt der Kunde vor?** + +- a) Kundennummer +- b) Kreditkarte** +- c) ADAC Mitgliedsnummer +- d) Telefonnummer +- e) Adresse + +#### **7.2AV Wie möchte der Kunde bezahlen?** + +- a) Nachnahme +- b) Kreditkarte** +- c) Vorkasse + +#### **7.3V Welche Karte würde verwendet?** + +- a) Maestro +- b) Visa** +- c) Mastercard +- d) Cirrus +- e) American Express + +#### **8.1A Was hat der Kunden schon gemietet?** + +- a) BMW +- b) GOLF** +- c) FORD +- d) POLO +- e) PORSCHE + +#### **8.2AV Wann hatte er das letzte mal einen Wagen gemietet?** + +- a) noch nie +- b) letzten Sommer +- c) Anfang des Monats** +- d) letzte Woche +- e) 1999 + +#### **8.3V Welcher Tag ist am Kalender angemerkt?** + +- a) Montag +- b) Dienstag +- c) Mittwoch +- d) Donnerstag +- e) Freitag + +#### **9.1AV Wann kommt er zur Abrechnung?** + +- a) gar nicht, wegen Lastschrifteinzug +- b) am Freitag +- c) am darauffolgenden Morgen** +- d) gleich bei der Abgabe des Fahrzeuges + +e) am darauffolgenden Nachmittag + +#### **9.2A Wie hat er den Wagen beim letzten Mal zurückgegeben?** + +- a) gewaschen +- b) **vollgetankt** +- c) ohne Umzugskisten +- d) beschädigt +- e) zu spät + +#### **9.3V Welcher Tag ist am Kalender markiert?** + +- a) Montag +- b) Dienstag +- c) Mittwoch +- d) Donnerstag +- e) Freitag + +#### **10.1A Wann kommt den Kunden zurück?** + +- a) **am Abend** +- b) am Morgen +- c) nie +- d) am Mittag +- e) noch einen Tag später + +#### **10.2A Wohin fährt er den Wagen bei der Rückgabe?** + +- a) Tankstelle +- b) nach Hause +- c) **Firmenparkplatz** +- d) Tiefgarage +- e) Firmeneinfahrt + +#### **10.3V Was zeigt der Kunde vor der Kamera?** + +- a) **Eine Karte** +- b) Eine Rechnung +- c) Die Autoschlüssel +- d) Ein Kalender +- e) Eine Kreditkarte + +### **Scenario 3: Kitchen purchase order** + +#### **1.1A Why does the customer call?** + +- a) He misdialled +- b) wants to rent a car +- c) **wants to order a kitchen** +- d) wants to order a taxi +- e) order opera tickets + +#### **1.2A What does he ask for?** + +- a) to speak Mrs Müller** +- b) to speak the boss +- c) to get a discount +- d) to obtain a telephone number +- e) to speaker Mr Braun + +#### **1.3V What does the customer show?** + +- a) his business card +- b) a novel +- c) a catalogue with a kitchen** +- d) a microphone +- e) a mobile phone + +#### **2.1A What is Mrs Müller's business?** + +- a) sales advisory service of consumer electronics +- b) sales advisory service in the salesroom** +- c) sales of car insurances +- d) warehousing +- e) training + +#### **2.2A What did both agree upon?** + +- a) Mrs Müller will contact him +- b) He will wait for a call +- d) He will contact her not before Thursday +- d) He will call her until Wednesday** +- e) He will pass by + +#### **3.1AV On which colour did he decide?** + +- a) white +- b) red +- c) light grey** +- d) green +- e) blue + +#### **3.2A Why did he not choose white?** + +- a) stains are easily visible +- b) it does not fit with the cooker +- c) out of stock +- d) difficult to maintain +- e) appears too cold** + +#### **3.3A How fast did he decide against white?** + +- a) after giving a thought** +- b) immediately + +- c) very fast + +#### **4.1AV Which door knobs does he like best?** + +- a) round shape +- b) none +- c) shape of pyramid +- d) conus shape +- e) **lozenge shape** + +#### **4.2A What does he not like with the round shape?** + +- a) too ugly +- b) edges too sharp +- c) **too small** +- d) too big +- e) the colour + +#### **4.3AV Which colour were the door knobs?** + +- a) **grey** +- b) black +- c) gold +- d) wood like +- e) white + +#### **5.1AV What is chosen to be small in favour of the dishwasher?** + +- a) the wall cupboard +- b) the fridge +- c) the rubbish bin +- d) **the sink** +- e) the cooker + +#### **5.2A Why did he decide on a single sink?** + +- a) **to save space** +- b) to save money +- c) because he only needs one +- d) to save time +- e) he was told to + +#### **5.3V How many sinks are shown?** + +- a) one +- b) zero +- c) **two** + +#### **6.1A What does he ask about the dishwasher?** + +- a) if it is on offer +- b) **if it provides aqua-stop** +- c) if it is also available in grey + +#### **6.2A What does the washer have to provide?** + +- a) sound absorber +- b) special washing programs +- c) transportable weights +- d) high usability +- e) **aqua-stop** + +#### **6.3A If the model does not meet the requirements ...** + +- b) **he orders the next bigger model** +- a) he rather takes none at all +- c) he wants to complain about it +- d) he will return it +- e) he wants a special offer + +#### **7.1A Which cooker did he choose?** + +- a) the largest +- b) AEG +- c) **none so far** +- d) MIELE +- e) BAUKNECHT + +#### **7.2AV What kind of top does he wants for his cooker?** + +- a) **glass-ceramic** +- b) a gas-stove +- c) steel +- d) cast iron +- e) enamel + +#### **7.3V How much does the cooker cost? (depends on the catalogue)** + +- a) 2999 +- b) 299 +- c) 499 +- d) **1999** +- e) 699 + +#### **8.A What else should the cooker provide?** + +- a) ceramic top +- b) fridge +- c) sink +- d) freezer +- e) **sensor-circuitry** + +#### **9.A How big is the table supposed to be?** + +- a) 100 x 100 +- b) **100 x 160** + +- d) 60 x 60 + +#### **10.1A How much should the table be extendable?** + +- a) only in its height +- b) 200 cm +- c) **150 cm** +- d) 80 cm + +(German) + +#### **1.1A Aus welchem Grund ruft der Kunde an?** + +- a) verwählt +- b) er möchte ein Auto bestellen +- c) **er möchte eine Küche bestellen** +- d) bestellen eines Taxis +- e) bestellen von Kinokarten + +#### **1.2A Wonach fragt der Kunde?** + +- a) **nach Frau Müller** +- b) nach dem Chef +- c) nach einem Rabatt +- d) nach einer Telefonnummer +- e) nach Herrn Braun + +#### **1.3V Was zeigt der Kunde?** + +- a) eine Visitenkarte +- b) einen Roman +- c) **eine Broschüre einer Küche** +- d) ein Mikrofon +- e) ein Handy + +#### **2.1A Was macht Frau Müller?** + +- a) Beratung für Unterhaltungselektronik +- b) **Beratung in den Verkaufsräumen** +- c) Beratung für Autoversicherungen +- d) Lagerhaltung +- e) Fortbildungen + +#### **2.2A Was hatte der Kunde vereinbart?** + +- a) Frau Müller meldet sich bei ihm +- b) er wartet auf einen Anruf +- d) er meldet sich ab Donnerstag +- d) **er meldet sich bis Mittwoch** +- e) er kommt vorbei + +#### **3.1AV Für welche Farbe hat sich der Kunde entschieden?** + +- a) weiß +- b) rot +- c) **hellgrau** +- d) grün +- e) blau + +#### **3.2A Warum wurde nicht "weiß" gewählt?** + +- a) man sieht Flecken so schnell +- b) passt nicht zum Herd +- c) nicht vorrätig +- d) schlecht zu pflegen +- e) **wirkt zu kalt** + +#### **3.3A Wie schnell hat sie sich gegen "weiß" entschieden?** + +- a) **nach längerer Überlegung** +- b) sofort +- c) sehr schnell + +#### **4.1AV Welche Türknäufe mag er am liebsten?** + +- a) runde Form +- b) gar keine +- c) Pyramidenform +- d) Kegelform +- e) **Rautenform** + +#### **4.2A Was missfällt ihm an den runden Türknäufen?** + +- a) zu hässlich +- b) zu scharfkantig +- c) **zu klein** +- d) zu groß +- e) die Farbe + +#### **4.3AV Welche Farbe haben die Türknäufe?** + +- a) **grau** +- b) schwarz +- c) gold +- d) holzfarben +- e) weiß + +#### **5.1AV Was soll für die Spülmaschine kleiner werden?** + +- a) der Hängeschrank +- b) der Kühlschränk +- c) der Müllcontainer +- d) **das Spülbecken** + +- e) der Herd + +#### **5.2A Warum will der Kunde nur ein Becken bestellen?** + +- a) **um Platz zu sparen** +- b) um Geld zu sparen +- c) weil er nichts anderes braucht +- d) um Zeit zu sparen +- e) weil es ihm empfohlen wurde + +#### **5.3V Wie viel Becken werden gezeigt?** + +- a) 1 +- b) 0 +- c) **2** + +#### **6.1A Was fragt der Kunde zu der Spülmaschine?** + +- a) ob sie angeboten wird +- b) ob Aqua-stop verfügbar ist +- c) ob es sie in hellgrau gibt + +#### **6.2A Über was soll die Maschine verfügen?** + +- a) Geräuschdämpfer +- b) spezielle Waschprogramme +- c) herausnehmbare Gewichte +- d) einfache Bedienung +- e) **Aqua-Stop** + +#### **6.3A Wenn die Maschine nicht ihren Wünschen entspricht ...** + +- b) **... nimmt er die nächst größere** +- a) ... nimmt sie lieber keine +- c) ... will sie sich beschweren +- d) ... gibt es sie zurück +- e) ... will sie einen Rabatt + +#### **7.1A Welchen Herd hat er sich ausgesucht?** + +- a) den Größten +- b) AEG +- c) **noch keinen** +- d) MIELE +- e) BAUKNECHT + +#### **7.2AV Was für ein Kochfeld soll der Herd haben?** + +- a) **Glaskeramik** +- b) Gas +- c) Edelstahl +- d) Gusseisen +- e) Emaille + +#### **7.3V Wie viel kostet das Kochfeld?** + +- a) 2999 +- b) 299 +- c) 499 +- d) 1999** +- e) 699 + +#### **8.A Was soll integriert sein?** + +- a) Kochfeld +- b) Kühlschrank +- c) Becken +- d) Tiefkühlschrank +- e) Sensorsteuerung** + +#### **9.A Wie groß soll die Mikrowelle sein?** + +- a) 100 x 100** +- b) 100 x 160 +- d) 60 x 60 + +#### **10.1A Auf welche Länge will er den Tisch vergrößern?** + +- a) er soll nur in der Höhe verstellbar sein +- b) 100 cm +- c) 150 cm** +- d) 80 cm + +### **Scenario 4: Bithday party** + +#### **1.1A Who is celebrating his birthday?** + +- a) Peter** +- b) Jan +- c) Jens +- d) Niklas + +#### **1.2A The birthday party was ...** + +- b) last Monday +- c) last Thursday +- a) last Saturday** +- d) last Wednesday + +#### **1.3V What is the caller doing with her hands?** + +- a) waving** +- b) folding +- c) grabbing + +#### **2.1A Who does she met there?** + +- b) Peter's friends +- a) old friends** + +- c) her professor +- d) Jan's girl-friend + +#### **2.2A What are the names of the people from the photo?** + +- b) Stefan, Linda and Peter +- c) Stefan, Jens and Markus +- a) Stefan, Linda and Markus** + +#### **2.3AV What does she present to the camera?** + +- b) a photo of Peter +- c) a photo of her parents +- a) a photo of an old friend** +- d) a photo of her at the party + +#### **3.1AV What happened after the meal?** + +- a) a small concert** +- b) a theatre play +- c) a dance show + +#### **3.2A How was the atmosphere?** + +- b) tensed situation +- a) spirits were running high** +- c) everyone felt giddy +- d) sad + +#### **4.1A Whom did she meet there?** + +- a) Julia from school** +- b) her brother +- c) her kindergarten teacher +- d) Peter's father + +#### **4.2V What is Julia doing according to the photo?** + +- b) she is reading +- c) she is eating +- d) she sleeps +- a) she is dancing** + +#### **4.3V What kind of movement is she doing?** + +- a) a break-dance move** +- b) a gymnastics exercise +- c) she turns over + +#### **5.1A Who baked the cake?** + +- a) Jens +- b) Peter +- c) Peter's mother** +- d) Julia + +**5.2A The guest gathered at midnight in order to ...** + +- b) go out dancing +- c) hand over the presents +- d) bake a cake +- a) sing happy birthday** + +**5.3V Which colour was the cake?** + +- a) white and red** +- b) green +- c) brown +- d) white + +**6.1AV What was Peter's birthday present?** + +- b) a guitar +- c) a piano +- d) a violin +- a) a drum set** + +**6.2A When did Peter receive his present?** + +- a) before eating the cake** +- b) after eating the cake +- c) at dawn +- d) the next day + +**7.1A What did Peter and his friends laughed about?** + +- b) Peter's jokes +- c) the party +- a) the past** +- d) at Peter's brother + +**8.1A Which song did she sing?** + +- b) a drinking song +- a) a children's song** +- c) a national anthem +- d) a birthday song + +**8.2A Why did she sing?** + +- b) for Peter's birthday +- c) without any reason +- d) to accompany the band +- a) Peter's brother had to go to sleep** + +**9.1A What did she learn?** + +- a) eating with chopsticks** +- b) a dance move +- c) a fight move + +- d) baking a cake + +#### **9.2AV What was she playing with?** + +- b) a pen +- a) Chinese chopsticks** +- c) a spoon +- d) her mobile phone + +#### **9.3A Who taught her to use chopsticks?** + +- a) Peter's parents** +- b) Peter +- c) Peter's brother +- d) Lina + +#### **10.1A About Jan, she want to ...** + +- b) play soccer together +- c) have dinner +- a) meet each other** +- d) discuss issues + +#### **10.2V What is she doing at the end?** + +- a) waving** +- b) juggling +- c) clapping + +(German) + +#### **1.1A Wer feiert seinen Geburtstag?** + +- a) Peter** +- b) Jan +- c) Jens +- d) Niklas + +#### **1.B Die Geburtstagfeier war...** + +- b) letzten Montag +- c) letzten Donnerstag +- a) letzten Samstag** +- d) letzten Mittwoch + +#### **1.3B Was macht sie mit ihren Händen vor der Kamera?** + +- a) winken** +- b) falten +- c) greifen + +#### **2.1A Wen hat sie bei der Party getroffen?** + +- b) Peters Freunde +- a) alte Bekannte** +- c) ihren Professor + +d) Jans Freundin + +#### **2.2A Wie heißen die Leute, die auf dem Foto sind?** + +b) Stefan, Linda und Peter + +c) Stefan, Jens und Markus + +**a) Stefan, Linda und Markus** + +#### **2.3AV Was zeigt sie vor der Kamera?** + +b) ein Foto von Peter + +c) ein Foto von seinen Eltern + +**a) ein Foto von alten Freunden** + +d) ein Foto von ihr bei der Party + +#### **3.1AV Welche Veranstaltung findet nach dem Essen statt?** + +**a) ein kleines Konzert** + +b) ein Theaterstück + +c) eine Tanzshow + +#### **3.2A Wie war die Stimmung?** + +b) gespannt + +**a) ausgelassen** + +c) euphorisch + +d) traurig + +#### **4.1A Wen hat sie bei der Party plötzlich getroffen?** + +**a) ihre alte Schulfreundin Julia** + +b) ihren Bruder + +c) ihren Kindergartenlehrer + +d) Peters Vater + +#### **4.2V Was macht Julia auf dem Foto?** + +b) sie liest + +c) sie isst + +d) sie schläft + +**a) sie tanzt** + +#### **4.3V Was macht sie für eine Bewegung?** + +**a) ein Breakdance-Bewegung** + +b) eine Gymnastik-Bewegung + +c) sie dreht sich um + +#### **5.1A Wer hat den Kuchen gebacken?** + +a) Jens + +b) Peter + +**c) Peters Mutter** + +d) Julia + +**5.2A Die Gäste treffen sich gegen Mitternacht um ...** + +- b) ... tanzen zu gehen +- c) ... Geschenke zu verschenken +- d) ... einen Kuchen zu backen +- a) ... Happy Birthday zu singen** + +**5.3V Welche Farbe hat der Kuchen?** + +- a) weiß und rot** +- b) grün +- c) braun +- d) weiß + +**6.1AV Was bekommt Peter als Geburtstagsgeschenk?** + +- b) eine Gitarre +- c) ein Klavier +- d) eine Geige +- a) ein Schlagzeug** + +**6.2A Wann hat Peter sein Geburtstagsgeschenk bekommen?** + +- a) vor dem Kuchen** +- b) nach dem Kuchen +- c) am Ende der Nacht +- d) am folgenden Tag + +**7.1A Worüber haben sie und ihre Freunde gelacht?** + +- b) Peters Witze +- c) die Party +- a) die Vergangenheit** +- d) Peters Bruder + +**8.1A Welches Lied hat sie gesungen?** + +- b) ein Trinklied +- a) ein Kinderlied** +- c) die Nationalhymne +- d) ein Geburtstagslied + +**8.2A Warum hat sie gesungen?** + +- b) um Peter ein Geburtstagslied zu singen +- c) ohne Grund +- d) um bei der Band mitzusingen +- a) Peters Bruder musste ins Bett gehen** + +**9.1A Was hat sie gelernt?** + +- a) essen mit Essstäbchen** +- b) eine Tanzbewegung +- c) eine Kampfbewegung + +- d) einen Kuchen zu backen + +#### **9.2AV Mit welchem Objekt spielt sie?** + +- b) Einem Stift +- a) Chinesischen Essstäbchen** +- c) einer Gabel +- d) seinem Handy + +#### **9.3A Wer hat ihr beigebracht, mit Essstäbchen zu essen?** + +- a) Peters Eltern** +- b) Peter +- c) Peters Bruder +- d) Lina + +#### **10.2A Mit Jan wünscht sie sich ...** + +- b) zusammen Fußball zu spielen +- c) zusammen essen zu gehen +- a) sich bald wiedersehen** +- d) wichtige Dinge zu diskutieren + +#### **10.3V Was macht sie zum Abschluss?** + +- a) winken** +- b) jonglieren +- c) klatschen + +# Appendix III + +(This appendix does not form an integral part of this Recommendation.) + +The purpose of this appendix is to show examples on how to build test material profiles for time-varying transmission conditions. For building up prediction models, extreme profiles are interesting to find psychophysical effects. For applied goals, profiles can be adjusted to the apparent network conditions, which may be less extreme. The same profiles can be used on audio and video data, but stimuli can also be created applying separate profiles for both channels. + +![Figure III.1: Schematic illustration of the profiles used in [b-Belmudez et al., 2013]. The figure displays 15 profiles (profile 1 to profile 15) arranged in two columns. Each profile is a bar chart showing 'AV Quality' (or 'Video Quality' and 'Audio Quality') on the y-axis against five categories (1, 2, 3, 4, 5) on the x-axis. The profiles show varying levels of quality across the categories, with some profiles showing high quality across all categories (e.g., profile 1) and others showing more varied or lower quality (e.g., profile 15).](77a781dfb114c3e2b399f876f1808cfd_img.jpg) + +Figure III.1 displays 15 profiles (profile 1 to profile 15) arranged in two columns. Each profile is a bar chart showing 'AV Quality' (or 'Video Quality' and 'Audio Quality') on the y-axis against five categories (1, 2, 3, 4, 5) on the x-axis. The profiles show varying levels of quality across the categories, with some profiles showing high quality across all categories (e.g., profile 1) and others showing more varied or lower quality (e.g., profile 15). + +**Left Column Profiles:** + +- profile 1: AV Quality is high across all categories (1-5). +- profile 2: AV Quality is high across all categories (1-5). +- profile 3: AV Quality is low across all categories (1-5). +- profile 4: AV Quality is low across all categories (1-5). +- profile 5: AV Quality is high across all categories (1-5). +- profile 6: AV Quality is high across all categories (1-5). +- profile 7: AV Quality is high across all categories (1-5). +- profile 8: AV Quality is high across all categories (1-5). +- profile 9: AV Quality is high across all categories (1-5). +- profile 10: AV Quality is high across all categories (1-5). +- profile 11: AV Quality is high across all categories (1-5). + +**Right Column Profiles:** + +- profile 12: Video Quality is high across all categories (1-5); Audio Quality is low across all categories (1-5). +- profile 13: Video Quality is high across all categories (1-5); Audio Quality is high across all categories (1-5). +- profile 14: Video Quality is high across all categories (1-5); Audio Quality is low across all categories (1-5). +- profile 15: Video Quality is low across all categories (1-5); Audio Quality is high across all categories (1-5). + +Figure III.1: Schematic illustration of the profiles used in [b-Belmudez et al., 2013]. The figure displays 15 profiles (profile 1 to profile 15) arranged in two columns. Each profile is a bar chart showing 'AV Quality' (or 'Video Quality' and 'Audio Quality') on the y-axis against five categories (1, 2, 3, 4, 5) on the x-axis. The profiles show varying levels of quality across the categories, with some profiles showing high quality across all categories (e.g., profile 1) and others showing more varied or lower quality (e.g., profile 15). + +Figure III.1 – Schematic illustration of the profiles used in [b-Belmudez et al., 2013] + +## Bibliography + +- [b-ETSI TR 102 506] ETSI TR 102 506 (2011), *Speech Processing, Transmission and Quality Aspects (STQ); "Estimating Speech Quality per Call"*, European Telecommunications Standards Institute, Sophia Antipolis. +- [b-Belmudez et al., 2013] Belmudez, B., Lewcio, B., Möller, S. (2013), *Call Quality Prediction for Audiovisual Time-Varying Impairments Using Simulated Conversational Structures*, submitted to Acta Acustica united with Acustica 99(5), 792-805. + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|---------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Terminals and subjective and objective assessment methods | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects and next-generation networks | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.1305-201607-I_PDF-E/raw.md b/marked/P/T-REC-P.1305-201607-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..eec0b47679e4938983eade897de9f2bccd1bf51f --- /dev/null +++ b/marked/P/T-REC-P.1305-201607-I_PDF-E/raw.md @@ -0,0 +1,823 @@ + + +International Telecommunication Union + +# **ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +# **P.1305** + +(07/2016) + +SERIES P: TERMINALS AND SUBJECTIVE AND +OBJECTIVE ASSESSMENT METHODS + +Telemeeting assessment + +--- + +## **Effect of delays on telemeeting quality** + +Recommendation ITU-T P.1305 + +**ITU-T** + +![ITU logo: A globe with a red lightning bolt and the text 'ITU International Telecommunication Union'.](1d7527f4316cfe2d342b08d1653d1592_img.jpg) + +The logo of the International Telecommunication Union (ITU) is located in the bottom right corner. It features a blue globe with a red lightning bolt striking across it. To the right of the globe, the text "ITU" is written in a large, bold, blue font, and below it, "International Telecommunication Union" is written in a smaller, blue font. + +ITU logo: A globe with a red lightning bolt and the text 'ITU International Telecommunication Union'. + +# ITU-T P-SERIES RECOMMENDATIONS + +## TERMINALS AND SUBJECTIVE AND OBJECTIVE ASSESSMENT METHODS + +| | | | +|----------------------------------------------------------------------------------------------------|---------------|---------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Voice terminal characteristics | Series | P.30 | +| | | P.300 | +| Reference systems | Series | P.40 | +| Objective measuring apparatus | Series | P.50 | +| | | P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of speech quality | Series | P.80 | +| Methods for objective and subjective assessment of speech and video quality | Series | P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | +| Transmission performance and QoS aspects of IP end-points | Series | P.1000 | +| Communications involving vehicles | Series | P.1100 | +| Models and tools for quality assessment of streamed media | Series | P.1200 | +| Telemeeting assessment | Series | P.1300 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | Series | P.1400 | +| Methods for objective and subjective assessment of quality of services other than speech and video | Series | P.1500 | + +For further details, please refer to the list of ITU-T Recommendations. + +## Recommendation ITU-T P.1305 + +# Effect of delays on telemeeting quality + +## Summary + +Recommendation ITU-T P.1305 has the primary purpose of describing the impact of echo-free transmission delays on telemeeting quality of experience (QoE). Delay has a major impact on the interaction performance of conference participants, but is not always explicitly noticed by them; its impact being commonly attributed to other participants behaviour rather than being a system feature. + +Before discussing the impact of the delays, it is useful to first consider how people behave when conversing in an ideal environment. Since this has been well studied in the linguistic discipline of conversation analysis (CA), a short discussion of conversations in that context and a discussion on the impact of delay from a CA perspective is included. Some analysis based on the alternative technique of conversation surface structures is also included. + +A secondary purpose of this Recommendation is to provide guidance on appropriate testing methods for evaluating the effect of delay in telemeetings. + +## History + +| Edition | Recommendation | Approval | Study Group | Unique ID* | +|---------|----------------|------------|-------------|---------------------------------------------------------------------------| +| 1.0 | ITU-T P.1305 | 2016-07-29 | 12 | 11.1002/1000/12974 | + +## Keywords + +Conversation, conversation analysis, delay, subjective test, telemeeting. + +--- + +\* To access the Recommendation, type the URL in the address field of your web browser, followed by the Recommendation's unique ID. For example, . + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database at . + +© ITU 2016 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## Table of Contents + +| | | Page | +|------|----------------------------------------------------------------------------------------------------------------|------| +| 1 | Scope..... | 1 | +| 2 | References..... | 1 | +| 3 | Definitions ..... | 2 | +| 3.1 | Terms defined elsewhere ..... | 2 | +| 3.2 | Terms defined in this Recommendation..... | 3 | +| 4 | Abbreviations and acronyms ..... | 3 | +| 5 | Conventions ..... | 3 | +| 6 | Existing Recommendations concerning the subjective quality evaluation in case of pure transmission delay ..... | 3 | +| 6.1 | Audio delay ..... | 3 | +| 6.2 | Video delay..... | 3 | +| 6.3 | Synchronization between audio and video ..... | 4 | +| 7 | Conversation features ..... | 4 | +| 7.1 | Two-party and multi-party conversations ..... | 4 | +| 7.2 | Two analysis techniques..... | 4 | +| 7.3 | Conversation analysis and turn-taking ..... | 5 | +| 7.4 | Conversation structure and styles..... | 7 | +| 7.5 | Backchannels and other behaviours during conversations..... | 7 | +| 7.6 | Conversation surface structure analysis ..... | 7 | +| 8 | The impact of delay ..... | 9 | +| 8.1 | Turn-taking with delay ..... | 9 | +| 8.2 | Conversation surface structure and delays ..... | 10 | +| 8.3 | Multiparty and two-party conversations with delay..... | 11 | +| 8.4 | Other side-impacts of delay..... | 11 | +| 9 | Factors influencing the impact of pure delay ..... | 11 | +| 9.1 | Audiovisual telemeetings ..... | 11 | +| 9.2 | User expectations..... | 12 | +| 9.3 | Asymmetric delay between participants and co-location..... | 12 | +| 9.4 | Interpersonal influences..... | 12 | +| 10 | Testing methods..... | 12 | +| 10.1 | General comments ..... | 12 | +| 10.2 | Conversational-opinion tests ..... | 13 | +| 10.3 | Group interaction tests..... | 14 | +| 10.4 | Communication system effectiveness tests ..... | 16 | +| 10.5 | Micro-feature tests..... | 16 | +| 11 | Areas for further research regarding the impact of delay ..... | 17 | +| | Annex A – Suggestions for free-conversation tasks ..... | 18 | + +| | Page | +|------------------------------------------------------|------| +| A.1 'Holiday' task ..... | 18 | +| A.2 Role-playing games ..... | 18 | +| A.3 "Who am I?" Celebrity Name-Guessing Task..... | 18 | +| A.4 Navigation tasks ..... | 19 | +| A.5 Story with missing parts ..... | 19 | +| A.6 Building blocks task – variation ..... | 19 | +| A.7 Survival task – variations ..... | 20 | +| Annex B – Suggestions for delay-critical tasks ..... | 21 | +| B.1 Information exchange tasks ..... | 21 | +| B.2 Random number verification task timed ..... | 21 | +| Bibliography..... | 22 | + +# Effect of delays on telemeeting quality + +# 1 Scope + +This Recommendation has the primary purpose of describing the impact of echo-free transmission delays on telemeeting quality of experience (QoE). Delay is special amongst communication impairments in that it can have a major impact on how well the users interact, but very often without them being explicitly aware of its presence. Typically users comment on its impact (e.g., "we kept on interrupting each other") or assume a delayed response is due to the other participants taking a long time to respond. Consequently its impact is not always captured in conversational-opinion tests. Crucially the impact has been shown to be highly dependent on the task or activity being undertaken. + +The focus in this Recommendation is therefore on the human factor – its purpose is not to define specific quantitative limits or values, discuss the sources of delay or to discuss methods of delay measurement or methods to reduce its impact. + +Before discussing the impact of the delays, it is useful to first consider how people behave when conversing in an ideal environment. Since this has been well studied in the linguistic discipline of conversation analysis (CA), a short discussion of conversations in that context and a discussion on the impact of delay from a CA perspective is included. Some analysis based on the alternative technique of conversation surface structures is also included. + +A secondary purpose of this Recommendation is to provide guidance on appropriate testing methods. A number of these have already been described in Annex D of [ITU-T P.1301] and will not be repeated here. However, there have been a number of new developments since that recommendation came into force and it is appropriate to expand on that knowledge. + +# 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- [ITU-T G.114] Recommendation ITU-T G.114 (2003), *One-way transmission time*. +- [ITU-T G.131] Recommendation ITU-T G.131 (2003), *Talker echo and its control*. +- [ITU-T P.800] Recommendation ITU-T P.800 (1996), *Methods for subjective determination of transmission quality*. +- [ITU-T P.805] Recommendation ITU-T P.805 (2007), *Subjective evaluation of conversational quality*. +- [ITU-T P.920] Recommendation ITU-T P.920 (2000), *Interactive test methods for audiovisual communications*. +- [ITU-T P.1301] Recommendation ITU-T P.1301 (2012), *Subjective quality evaluation of audio and audiovisual multiparty telemeetings*. + +[ITU-T P.1312] Recommendation ITU-T P.1312 (2016), *Method for the measurement of the communication effectiveness of multiparty telemeetings using task performance*. + +[ITU-R BT.1359-1] Recommendation ITU-R BT.1359-1 (1998), *Relative timing of sound and vision for broadcasting*. + +# 3 Definitions + +## 3.1 Terms defined elsewhere + +This Recommendation uses the following terms defined elsewhere: + +**3.1.1 telemeeting** [ITU-T P.1301]: A meeting in which participants are located at least two locations and the communication takes place via a telecommunication system. The term telemeeting is used to emphasize that a meeting is often more flexible and interactive than a conventional business teleconference and could also be a private meeting. The telemeeting could be audio-only, audiovisual, text-based, or a mix of these modes. + +**3.1.2 conversational quality** [ITU-T P.1301]: The perceived quality when two or more test participants have a conversation. + +**3.1.3 two-party** [ITU-T P.1301]: Two persons. Example: Two persons are participating in a telemeeting, having a conversation, performing a test task together, etc. If not explicitly stated differently, two-party implicates that the persons are at two locations. + +**3.1.4 multiparty** [ITU-T P.1301]: More than two persons. Example: More than two persons are participating in a telemeeting, having a conversation, performing a test task together, etc. The term multiparty does not specify if the persons are distributed across two or more locations. If not explicitly stated differently, multiparty implicates that the persons are at two or more than two locations. When further specification is necessary, additional terms will be used (see point-to-point and multi-point) or the number of locations will be explicitly stated. + +**3.1.5 point-to-point** [ITU-T P.1301]: Two locations. Example: A multiparty point-to-point telemeeting means that more than two interlocutors are taking part, and the interlocutors are at exactly two locations. That means that in one location more than one interlocutors are present, as there are more than two persons. + +**3.1.6 multi-point** [ITU-T P.1301]: More than two locations. Example: A multiparty multi-point telemeeting means that more than two interlocutors are taking part, and the interlocutors are located across more than two locations. Multi-point does not specify if one or more than one interlocutor may be present at each location. In the special case that only one person is present at each location, the term one-per-site will be used. + +**3.1.7 one-per-site** [ITU-T P.1301]: One person per connected location. Example: In a multiparty one-per-site telemeeting more than two sites are connected with only one person present at each site. + +**3.1.8 conversation analysis (CA)** [b-Sacks, 1974]: The study of social interaction, in particular focusing on conversations. + +**3.1.9 turn construction unit (TCU)** [b-Sacks, 1974]: A conversation analysis term describing the fundamental segment of speech in a conversation – essentially a piece of speech that constitutes an entire 'turn'. + +**3.1.10 transition relevance place (TRP)** [b-Sacks, 1974]: A conversation analysis term which indicates where a turn or floor exchange can take place between speakers. + +**3.1.11 system effectiveness** [ITU-T P.1312]: The ratio, expressed as a percentage, between the performance score achieved on a given system and the performance score obtained in face-to-face communication. + +## **3.2 Terms defined in this Recommendation** + +This Recommendation defines the following terms: + +**3.2.1 choral behaviour:** Many people doing the same thing at the same time. + +**3.2.2 phatic expression:** An expression whose function is part of social interaction rather than to convey information. + +# **4 Abbreviations and acronyms** + +This Recommendation uses the following abbreviations and acronyms: + +CA Conversation Analysis + +MOS Mean Opinion Score + +QoE Quality of Experience + +QoS Quality of Service + +TCU Turn Constructional Unit + +TRP Transition Relevance Place + +# **5 Conventions** + +None. + +# **6 Existing Recommendations concerning the subjective quality evaluation in case of pure transmission delay** + +In the following existing Recommendations on audio delay, video delay and audio-video synchronization the current understanding is outlined. These Recommendations have been developed for two-party interaction only. The multi-party case needs further study to ensure their applicability. + +## **6.1 Audio delay** + +Recommendation [ITU-T G.114] states that, + +"Regardless of the type of application, it is recommended to not exceed a one-way delay of 400 ms for general network planning..." + +"Although a few applications may be slightly affected by end-to-end (i.e., "mouth-to-ear" in the case of speech) delays of less than 150 ms, if delays can be kept below this figure, most applications, both speech and non-speech, will experience essentially transparent interactivity." + +The delay values in [ITU-T G.114] are based on the effect of pure audio delay only, i.e., in the complete absence of any echo. It is well-known that if echo is present, longer audio delays become easier detectable and much more disturbing. Effects of talker echo at different one-way echo-path delays are further described in [ITU-T G.131]. + +## **6.2 Video delay** + +Most investigations regarding quality impact of delay have been performed regarding audio delays, but in an earlier version of [ITU-T G.114] (02/96), in clause B.2.3, it is mentioned that: + +"Tests were performed to assess the interaction between delay and user applications. In these tests a comparison of telephone conversations with videophone were made and it was shown that there is little difference between both types of connection." + +This result was obtained for synchronized audio and video. Some research has been performed looking at pure audio-visual delay in an interactive setting, but there is a need for more research to clearly confirm the recommendations of [ITU-T G.114] for this case, especially regarding multi-party communication. + +## 6.3 Synchronization between audio and video + +The limits within which audio and video are perceived to be synchronized are well examined, at least for TV screens. According to [ITU-R BT.1359-1], + +"subjective evaluations show that detectability thresholds are about +45 ms to -125 ms and acceptability thresholds are about +90 ms to -185 ms on the average, a positive value indicates that sound is advanced with respect to vision". + +As video encoding usually requires more time than audio coding, audio should generally be delayed to obtain audio-video synchronization [b-Berndtsson, 2012]. It is likely that it is easier to identify audio-video asynchrony as a technical issue than synchronous delay. Following from this, audio-video asynchrony will most likely be reflected more critically in mean opinion score (MOS) ratings. + +Low video frame rates and video frame rate jitter can influence the impression of audio-video synchronization. + +# 7 Conversation features + +## 7.1 Two-party and multi-party conversations + +There are differences between two-way conversations and multi-party conversations. For instance in multi-party conversations: + +- participants are likely to spend a greater proportion of the time listening rather than talking. Usually this will vary as the discussion proceeds, with talkers being more active or passive at different times; +- telemeeting systems do not always attempt to reproduce the exact signals from the sending end at the receiving end; system designers are usually more interested in creating an environment where participants can interact effectively, and this can involve multiple technologies such as spatial audio, level compensation, noise reduction etc.; +- participant interaction is more complicated: in a two-party conversation the role of the speaker and the listener simply alternate, whereas in a multi-party conversation, the next speaker could be any one of several current listeners; +- passive participants are not only listening to what the active participants are saying, they are listening to the active participants interacting. + +## 7.2 Two analysis techniques + +In this Recommendation two different techniques for analysing a conversation are discussed. The first is based on the linguistic discipline of conversation analysis (CA) [b-Sacks, 1974] and focuses on analysing how the participants interact. For evaluating the impact of delay, the two most relevant analysis concepts from CA are turn-taking (see clause 7.3) and backchannels (see clause 7.6). + +The second technique analyses patterns of the temporal occurrence of utterances without labelling their purpose. This approach does not rely on the content of the utterances and can therefore be + +built on voice activity detection algorithms. This approach is referred to as conversational surface structure analysis (clause 7.4) and was pioneered by Brady in the 1960s. [b-Brady, 1965]. + +## 7.3 Conversation analysis and turn-taking + +In a free conversation the organization of the conversation, in terms of who speaks when, is referred to as 'turn-taking'. This is implicitly negotiated by a multitude of verbal cues within the conversation and also by nonverbal cues such as physical motion and eye contact. Turn-taking behaviour is more-or-less universal, even across languages of very different structures. This behaviour has been extensively studied in the discipline of conversation analysis, for example, [b-Sacks, 1974]. + +Some useful concepts from CA are: + +- The turn constructional unit (TCU), which is the fundamental segment of speech in a conversation – essentially a piece of speech that constitutes an entire 'turn'. TCUs are primarily determined as being pragmatically or grammatically complete utterances. That is, have they served a specific purpose, or are they syntactically complete. Typically they are separated by short pauses and a falling tone. +- The transition relevance place (TRP), which indicates where a turn or floor exchange can take place between speakers. TCUs are separated by TRPs. +- Turn allocation component, which relates to who should speak next. This is done either explicitly when the current talker selects the next talker – directly or subtlety; or the next talker 'self-selects' by simply interrupting at the end of the TCU. + +These processes enable the basic turn-taking process to take place, as described in [b-Sacks, 1974] and shown in a slightly modified form in Figure 1. As a TCU comes to an end the first decision to take place is, has the current talker selected the next talker? If so, the designated new talker usually feels obliged to talk. Otherwise, any participant including the current talker will self-select; a process which generally works by the first person to speak gaining the right to the next turn. The result of this process is a well ordered conversation as illustrated in Figure 2. + +![Flowchart of the conversation turn-taking process. It starts with '1. TCU', followed by '2. Current talker selects the next talker?'. If 'Yes', it loops back to '1. TCU'. If 'No', it goes to '3. Next talker self-selects?'. If 'Yes', it loops back to '1. TCU'. If 'No', it goes to '4. Current talker continues?'. If 'Yes', it loops back to '1. TCU'. If 'No', it loops back to the entry point of step 3. The diagram is labeled 'P.1305(16)_F01'.](d4af765160d04ecef538e5066006dc77_img.jpg) + +``` + +graph TD + 1[1. TCU] --> 2{2. Current talker selects the next talker?} + 2 -- Yes --> 1 + 2 -- No --> 3{3. Next talker self-selects?} + 3 -- Yes --> 1 + 3 -- No --> 4{4. Current talker continues?} + 4 -- Yes --> 1 + 4 -- No --> 3 + +``` + +Flowchart of the conversation turn-taking process. It starts with '1. TCU', followed by '2. Current talker selects the next talker?'. If 'Yes', it loops back to '1. TCU'. If 'No', it goes to '3. Next talker self-selects?'. If 'Yes', it loops back to '1. TCU'. If 'No', it goes to '4. Current talker continues?'. If 'Yes', it loops back to '1. TCU'. If 'No', it loops back to the entry point of step 3. The diagram is labeled 'P.1305(16)\_F01'. + +**Figure 1 – The conversation turn-taking process** + +![Timeline diagram showing turn-taking between three talkers (A, B, and C). Talker A has four TCU blocks, with TRP blocks above the first three. Talker B has four TCU blocks, with TRP blocks above the first two and last two. Talker C has one TCU block, with a TRP block above it. The diagram is labeled 'P.1305(16)_F02'.](042733dc5e8e7f5f30b60adba3266cde_img.jpg) + +Timeline diagram showing turn-taking between three talkers (A, B, and C). Talker A has four TCU blocks, with TRP blocks above the first three. Talker B has four TCU blocks, with TRP blocks above the first two and last two. Talker C has one TCU block, with a TRP block above it. The diagram is labeled 'P.1305(16)\_F02'. + +**Figure 2 – Example of turn-taking in a well-ordered conversation** + +Note of course that most of the turn-taking process is carried out subconsciously in line with what the participants consider to be appropriate behaviour. Knowing when it is acceptable to take your turn in a conversation is usually learned from an early age and becomes an important social skill. + +Depending on the type of conversation, there may also be a significant amount of talking taking place outside the underlying turn-taking process – in some cases obscuring it completely. This can include backchannels (clause 7.5), brief side-conversations, corrections, confirmations and choral laughter. + +Note that this description of the turn-taking process excludes interruptive behaviour. From a CA perspective, a participant starting a turn at a time other than a TRP is making a 'violative interruption'. This can often be interpreted by other participants as ill-mannered or rude behaviour. + +This turn-taking decision process leads to the following meeting characteristics: + +- Overwhelmingly, only one participant talks at a time [b-Sacks, 1974]. +- Occurrences of more than one talker at a time are common, but brief [b-Sacks, 1974]. +- Transitions from one turn to the next, with no gap or overlap are common. Together with transitions characterized by either a slight gap or slight overlap, they make up the vast majority of transitions [b-Sacks, 1974]. +- The most frequent gaps between talkers are in the region of 200 ms. [b-Gisladottir, 2015]. Gaps of more than 1 second are rare but do occur. [b-Jefferson, 1989] + +## 7.4 Conversation structure and styles + +Conversations can vary dramatically in their style of structure. At one extreme relaxed free conversation can occur amongst friends in social situations, where all turn taking is organized as described above. There may also be a significant amount of talking taking place outside the underlying turn-taking process, in some cases obscuring it completely. This additional content can include backchannels (clause 7.5), brief side-conversations, corrections, confirmations and choral laughter. + +At the other extreme some conversations are highly structured where most turn-taking is explicitly directed by a chairman or moderator. In this case a participant is directly asked by the participant currently holding the turn to speak. + +A conversation can change style dynamically, e.g., following an agenda but breaking out to free conversations in the discussion of single items. + +## 7.5 Backchannels and other behaviours during conversations + +A further aspect to consider is that there is often a lot of talking taking place outside the formal turn-taking process as described in CA. A conversation feature of particular interest is the use of continuers or backchannels [b-OConnail, 1993], also referred to as 'continuers' or 'listener responses' [b-Knapp, 2010]. These are listener responses in a primarily one-way communication, commonly used by participants in a conversation as a sign of acknowledgement and also to signal to the talker to continue with their turn. These can be both verbal and non-verbal in nature (e.g., "yeah", "mm", "uh-huh", "mm-hmm") and are frequently phatic expressions. They are often made between sentences and especially at transition points. Backchannels can also be visual in the form of nods or other facial expressions. + +The term "backchannel" was designed to imply that there are two channels of communication operating simultaneously during a conversation; the predominant channel is that of the speaker who directs primary speech flow. + +Other behaviours include: requests for confirmation or clarification; choral behaviour, especially laughter; and side conversations between subgroups. In many situations, particularly informal ones, it is not unusual to have several conversations taking place simultaneously with participants both talking and listening at the same time. Thus the ability of a telemeeting system to support doubletalk is a very important feature for natural human communications. + +## 7.6 Conversation surface structure analysis + +Conversation surface structure analysis has been introduced by Brady [b-Brady, 1965], [b-Brady, 1968], [b-Brady, 1971]. It analyses the temporal occurrence of speech activity, that is, on-off patterns of speech and uses this to define various conversation states such as single-talk, multiple-talk. These are then analysed to provide metrics such as: + +- State sojourn time: the mean time of staying in that state, [b-Brady1971] +- Speaker alteration rate: the mean rate at which speaker changes occurred [b-Hammer, 2004]) + +- Utterance Rhythm: mean time from one utterance of a speaker to its next utterance, [b-Schoenenberg, 2014]. + +Conversation surface structure analysis consists of the following procedure: + +1. A voice activity algorithm can be used to extract those on-off patterns of each single person recording. See for example Figure 3, top panel. +2. Then for each site, the single person on-off patterns need to be combined as they were experienced by the persons at the different ends which means the respective other persons need to be delayed if delay was present. See for example Figure 3, bottom panel. +3. In the next step, states need to be assigned to every sample of the combined on-off pattern, according to Figure 4. In this case 'S' and 'M' are used to indicate single and multiple talker states, and sub-scripts 'n,X' to indicate how many talkers there are and which one is talking. Thus for two participants. $S_{2,0}$ : silence, $S_{2,A}$ : single talk of person A, $S_{2,B}$ : single talk of person B, $M_{2,AB}$ : multi/double talk of persons A and B. State transitions are indicated by arrows. +4. From those states, it is possible to compute the measures such as state sojourn time, speaker alteration rate and utterance rhythm that characterize the conversation surface structure. + +An example of how to do this is given in clause 10. + +![Figure 3: Two horizontal timelines showing speech activity for Person A and Person B. The top panel shows a situation without delay, where Person A's speech (black blocks) and Person B's speech (grey blocks) overlap directly. The bottom panel shows a situation with delay, where Person B's speech is shifted to the right relative to Person A's. Below the timelines, state labels are placed in intervals: S_{2,A}, M_{2,AB}, S_{2,B}, S_{2,0}, S_{2,A}, S_{2,0}, S_{2,B}. The label 'P.1305(16)_F03' is in the bottom right corner.](411fa16c3211377525ba37c57784fee0_img.jpg) + +Figure 3: Two horizontal timelines showing speech activity for Person A and Person B. The top panel shows a situation without delay, where Person A's speech (black blocks) and Person B's speech (grey blocks) overlap directly. The bottom panel shows a situation with delay, where Person B's speech is shifted to the right relative to Person A's. Below the timelines, state labels are placed in intervals: S\_{2,A}, M\_{2,AB}, S\_{2,B}, S\_{2,0}, S\_{2,A}, S\_{2,0}, S\_{2,B}. The label 'P.1305(16)\_F03' is in the bottom right corner. + +**Figure 3 – Example of single and multi talk states in a two-party conversation situation.** +**Top panel: Situation without delay. Bottom panel: Situation from the perspective of Person A when delay is present, i.e., person B is delayed** + +![Figure 4: A state transition diagram for a two-party conversation. It consists of four circular nodes arranged in a diamond shape: S_{2,0} at the top, S_{2,A} on the left, S_{2,B} on the right, and M_{2,AB} at the bottom. Arrows indicate transitions: S_{2,0} to S_{2,A}, S_{2,0} to S_{2,B}, S_{2,A} to M_{2,AB}, S_{2,B} to M_{2,AB}, M_{2,AB} to S_{2,A}, and M_{2,AB} to S_{2,B}. A vertical double-headed arrow connects S_{2,0} and M_{2,AB}. The label 'P.1305(16)_F04' is in the bottom right corner.](1b5a812c8aa20fd5cba28e97001d32de_img.jpg) + +Figure 4: A state transition diagram for a two-party conversation. It consists of four circular nodes arranged in a diamond shape: S\_{2,0} at the top, S\_{2,A} on the left, S\_{2,B} on the right, and M\_{2,AB} at the bottom. Arrows indicate transitions: S\_{2,0} to S\_{2,A}, S\_{2,0} to S\_{2,B}, S\_{2,A} to M\_{2,AB}, S\_{2,B} to M\_{2,AB}, M\_{2,AB} to S\_{2,A}, and M\_{2,AB} to S\_{2,B}. A vertical double-headed arrow connects S\_{2,0} and M\_{2,AB}. The label 'P.1305(16)\_F04' is in the bottom right corner. + +**Figure 4 – Possible states (sets and subsets) for a two-party conversation situation** + +# 8 The impact of delay + +## 8.1 Turn-taking with delay + +From a conversation analysis perspective, the impact of delay in a conversation is primarily to disrupt the turn-taking process because each participant has a different understanding of when the TRP occurs. A new talker may believe they are starting to talk at the TRP, but the other participants will hear them sometime later when they themselves may have already started talking. This breaks the self-selection process described in clause 7.3 because the participants cannot be aware of who started talking first. Essentially, none of the participants is able to interrupt at the TRP experienced by the other participants, and if the delay is long both talkers interpret the others action as a violative interruption. + +Note that the participants may not realize that the confusion they are experiencing is due to transmission delay, often attributing it to aberrant behaviour of the other participants, for example, slow responding or even rudeness. + +A good example of this delay induced confusion is the 'false start'. These are delay induced conversation features that occur when two participants start talking simultaneously after a period of silence or in response to a stimulus from another participant. For example, in a multi-party conversation this can occur in response to a question from a third participant as shown in Figure 5. Once they realize somebody else is also talking, one of three things can happen: + +1. Both participants stop talking. This could be interpreted as two simultaneous unintended successful interrupts. More false starts can then follow in an attempt to correct the false-start, possibly involving other participants, leading to a short period of confusion as to who actually has the conference 'floor'. From a CA point of view this has a major impact since the conversation stops unexpectedly. +2. One participant stops talking and the other continues. Whilst this is probably the least disruptive on conversation flow, it can still be a significant disruption of the information exchange; the participant who stopped talking may have been making a more useful contribution than the participant who continued. It has also provided a significant distraction from the matter under discussion. +3. Both participants keep talking resulting in a longer period of multi-talk. Whilst the conversation still continues, the impact on intelligibility is severe and from a social point of view the behaviour of the participants can be perceived as rude or inappropriate. + +False starts are common in multi-party telemeetings where delay is present. However normal conversation turn-taking makes them less likely in two-way conversations. Of course two participants can start talking simultaneously even if there is no delay. However, delay appears to make the problem worse because it takes longer for the situation to become apparent to the participant. + +![Figure 5: False starts in 3-way conversation. A timeline diagram showing talk and listen states for Person A, B, and C. It illustrates three cases of false starts: Case 1 (B and C stop, leading to confusion), Case 2 (B continues, C stops, least disruptive), and Case 3 (both continue, resulting in multi-talk and reduced intelligibility).](ad29805cd4f64ad2828e14feb66de664_img.jpg) + +Figure 5 illustrates false starts in a 3-way conversation between Person A, B, and C. The diagram shows talk and listen states over time. Dashed lines connect talk periods of one person to listen periods of another. The diagram is divided into three cases: + +- Case 1:** Both B and C stop, leading to period of confusion. +- Case 2:** B continues, C stops. This is the least disruptive case. +- Case 3:** Both B and C continue, resulting in a period of multi-talk and reduced intelligibility. + +P.1305(16)\_F05 + +Figure 5: False starts in 3-way conversation. A timeline diagram showing talk and listen states for Person A, B, and C. It illustrates three cases of false starts: Case 1 (B and C stop, leading to confusion), Case 2 (B continues, C stops, least disruptive), and Case 3 (both continue, resulting in multi-talk and reduced intelligibility). + +**Figure 5 – False starts in 3-way conversation** + +## 8.2 Conversation surface structure and delays + +It has been shown in many studies e.g., [b-Brady, 1971], [b-Egger, 2012], that the conversation flow is disrupted by delays. In particular, it has been described in different works [b-Brady, 1971], [b-Schoenenberg, 2014] that the quality of the conversational experience compared to a face-to-face conversation diverges in terms of conversational realities of each telemeeting participant, and that the degree of divergence is dependent on the length of the delay [b-Schoenenberg, 2014]. The higher the degree of divergence, the more difficult it will be for participants to hold a fluent interaction without misunderstandings or breakdowns of the interactional flow. + +An example of a conversation part with high divergence can be found in Figure 6. + +![Figure 6: Example of conversation part with high degree of divergence between the two realities due to latency. A timeline diagram showing states (SI, BR, AS, NI) for Person A and Person B. It shows how latency leads to different perceived states for each participant.](fc857414626a8d94d132e12d9afe52a4_img.jpg) + +Figure 6 shows a conversation part with high divergence due to latency. The diagram illustrates states (SI, BR, AS, NI) for Person A and Person B. The states are shown for both participants' own and other realities. The diagram shows how latency leads to different perceived states for each participant. + +P.1305(16)\_F06 + +Figure 6: Example of conversation part with high degree of divergence between the two realities due to latency. A timeline diagram showing states (SI, BR, AS, NI) for Person A and Person B. It shows how latency leads to different perceived states for each participant. + +**Figure 6 – Example of conversation part with high degree of divergence between the two realities due to latency** + +Here it can be seen that owing to the delay the number of states has increased from the four states defined in clause 7.6, to include breaks (BR), alternating silence (AS), successful interruption (SI) + +non-successful interruption (NI). An explanation of how the degree of divergence figure can be calculated is given in clause 10.3. + +## 8.3 Multiparty and two-party conversations with delay + +So far there is only little research on multiparty conversations and delay; most research conducted has focused on the two-party case. However, it is likely the impact of delays will be different in multi-party conversations for the following reasons: + +- Participants are likely to spend a greater proportion of the time listening rather than talking. This will vary as the discussion proceeds, with talkers being more active or passive. +- Active participants, those actively contributing to the discussion, have shown to be more sensitive to delay compared to their non-active fellow participants [b-Schmitt, 2014a], [b-Schoenenberg, 2011]. +- The turn-taking process is more complicated. For example, in a two-party conversation the role of the speaker and the addressee alternate. In multi-party conversations, the speaker can have multiple addressees and listeners and thus more responders. A greater likelihood of false starts or passive interruptions can thus be expected. + +## 8.4 Other side-impacts of delay + +Delay can also impact the performance of the system, for example, it can affect jitter buffers and result in packet loss which in turn can lead to audible artifacts. + +# 9 Factors influencing the impact of pure delay + +In clause 7, it was noted that conversations can differ greatly in their structure, ranging from formal meetings with strict control, up to very informal social gatherings, and that the impact of delay will differ in each case. In this section other factors that influence the impact of delay are considered. + +## 9.1 Audiovisual telemeetings + +The previous clauses have primarily discussed the audio-only meetings. In the case of synchronized audiovisual telemeetings there are the following differences: + +- Delay is more obvious to non-naive participants since visual reactions are seen. +- Visual cues (e.g., gestures, nods, gaze) are available to support turn-taking and back-channel communications. +- The video provides additional information that helps the conversation organization. For example, participants can make visual signals to each other: e.g., raising hand to interrupt, collective voting. + +Subjective tests have shown that it is more important to synchronize audio and video than to keep the one-way audio transmission delay as low as possible [b-Berndtsson, 2012]. Since video is usually more delayed than audio, the audio should generally be delayed to obtain audio-video synchronization, at least for transmission delays below 600 ms. The audio should not be more than 90 ms ahead of the video, and also less than 185 ms after the video. This reflects the natural experience of light travelling faster than sound [ITU-R BT.1359-1]. + +A study [b-Tam, 2012] comparing two-party conversations with audio and audiovisual channels, found that perceived conversation naturalness was significantly impacted at a delay of 400 ms for audio-only connections and at 500 ms for audio-visual connections. Similar insights were drawn from studies comparing face-to-face, audio-only and audiovisual meeting [b-Tang, 1992]. In audiovisual meetings longer silences were observed but the subjective scores from the participants did not reflect this. + +## 9.2 User expectations + +Whilst user expectations will have little impact on the effect of delay in a telemeeting, users are likely to be more accepting of delays under some conditions. For example they will have lower expectations from lower cost services or when making calls on more challenging situations. + +The physical distance also adds to the delay and users are usually aware of this. For example, signals propagate through physical media at approximately 2/3 of the speed of light, so it takes about 50 ms to travel 10,000 km. Therefore expectations might be lower when talking to someone on the other side of the world than when talking to someone in the office next door. + +## 9.3 Asymmetric delay between participants and co-location + +Usually, there will be different delays in the connections to each site. In this situation, test participants experiencing longer delays will have more difficulties taking part in the conversation. This can have the knock-on effect of reducing the quality of the whole group experience. It can, for example, cause a disruption of the conversation flow, or as it has been shown [b-Schmitt, 2014b], a single delayed person can cause a worse experience for all participants in a five party telemeeting. + +A similar situation will occur if many participants are co-located and only a few are joining over a delayed connection. + +## 9.4 Interpersonal influences + +The sensitivity for delay can be influenced by the participant's expectations of the other participants. It can be difficult for participants to determine whether an apparently slow response is due to the system or to a genuine slow response. To a certain extent the degree of familiarity with the other participant can influence this based on previous interactions with them. + +However, it has been shown that conversation partners that know each other can be perceived to be unusually slow if there are long delays [b-Berndtsson, 2012]. Furthermore, persons that know each other well often have more spontaneous conversations, which can also make delays more detectable. Note that even though interruption may be a part of the normal conversations in this case, they do not occur with awkward timing or too often. + +The social role of the interlocutors and their corresponding interaction behaviour can also have an impact. For instance, a dominant person, the boss or mentor, may speak more and may ask more questions within a call. If this is the case, response times will appear to be long for the asking person but short for the responding person, which leads to different perceptions for the different sites. + +# 10 Testing methods + +## 10.1 General comments + +It is recommended to carry out the subjective quality evaluation of multiparty telemeetings as much as possible according to existing test methods recommended by ITU-T. + +Since the primary impact of delay is on group interaction, it will be necessary to use some form of conversation in the test. This could include multi-party conversation tests or third party listening tests where subjects listen and comment on the quality of a conversation between others. + +Tests on group inter-action can be divided into several types: + +- conversational-opinion tests; +- group inter-action tests based on conversation analysis and conversation surface structure analysis; + +- tests that measure group performance when undertaking tasks; +- micro-feature tests such as the task efficiency tests described in [ITU-T P.1312] which focus on a specific aspect of talker interaction such as double-talk or turn-taking. + +It is possible to design tests that include elements of more than one of these types. + +## 10.2 Conversational-opinion tests + +General procedures for these are laid down in Annex D of [ITU-T P.1301]. The test must allow the subjects to experience the problem and report its impact in terms of conversation difficulty. + +### 10.2.1 Training session + +In most cases, the test procedure will include a short training phase to familiarize the subjects with the system and the test task. This is important if the subjects do not know each other well in order to encourage more interactive conversations. This could be incorporated into an introduction session if appropriate and consist of a simple interactive game or task to learn each other names and to generally 'break the ice'. + +### 10.2.2 Conversational test tasks + +The test tasks should ideally be delay sensitive, but support natural conversation. Some suggestions for appropriate tasks are given in Annex D of [ITU-T P.1301] and in Annex A. More specifically: + +- Free conversation tasks are recommended for studying group behaviour since they allow natural subject interaction and subjects are able to adapt their behaviour to make the best use of the system under test. +- Short conversation tests (SCTs) as described in [ITU-T P.805] are not delay sensitive because subjects have to read from a script which masks the delay [b-Berndtsson, 2012]. +- Highly interactive tasks such as random number exchanges are sensitive to delay, but do not reflect normal conversation behaviour. +- Delays can be more easily detected as a technical issue in a highly interactive conversation where the motivation to interact fast is high [b-Schoenenberg, 2014]. As an example, some kind of competition might motivate subjects to interact more efficiently, and delays may be more apparent if they are hindering the interaction. + +### 10.2.3 Test questions + +Care should be used in the choice of questions in order to enable the test subjects to appropriately comment on their experience. For example, questions based on how the subjects perceived the interactivity are likely to be more revealing than questions about the overall quality. + +For example: + +- Did you experience difficulty in communicating? +- How would you judge the effort required to interrupt the other participants? +- How easy did you find it to communicate using the system? +- How attentive were your conversation partners during the call? + +A number of appropriate examples are given in [ITU-T P.805] and Annex D of [ITU-T P.1301], along with accompanying scales. + +Because delay induced interrupts are sometimes perceived as rude it would be good if these perceptions could also be captured. This can be difficult since subjects may object to their individual personalities being scrutinised in this manner. Perhaps the best approach, if the subjects knew each other well, is to ask questions such as, "Did the behaviour of any of the other participants seem out of character at any time?" + +## 10.3 Group interaction tests + +The principle behind group interaction tests lies in a comparison between the behaviour of test subjects in a reference condition, with their behaviour when using the system under test. + +Of the two approaches discussed earlier, conversation analysis, and conversation surface structure analysis, the latter is more mature and its focus on simply who is talking at any given time, lends itself more to an algorithmic approach [b-Schoenenberg, 2014]. Because conversation analysis assumes an understanding of the actual conversation content, detailed analysis is likely to require higher level resources such as speech recognition and natural language processing. Consequently at this stage no measurement techniques directly based on it have been published. However, analysis based on simple turn-taking could be considered. + +In group interaction tests, in addition to normal good practice in subject selection [ITU-T P.1301], [ITU-T P.800], careful consideration should be given to matching test subject profiles. This is important for free conversations since as far as possible we need to ensure consistent behaviour across groups. + +The number of variables in free conversations that influence conversational interaction can be very large. The interactivity in the conversation might be higher or more consistent if the test subjects have: + +- The same first language. +- A similar level of experience in using telemeeting systems. +- Similar levels of familiarity with each other. +- Similar cultural backgrounds. + +It may also be appropriate to consider matching ages, but of course, note that excessive restrictions placed on subject selection will reduce the generality of the results. + +### 10.3.1 Conversation surface structure measurements + +It can be seen from clause 8.2 that the quality of the conversational experience compared to a face to face conversation diverges in terms of conversational realities of each telemeeting participant, and that the degree of divergence is dependent on the length of the delay [b-Schoenenberg, 2014]. A measure of this divergence can be calculated using the following procedure, and with reference to Figure 6: + +1. First of all, the most used meaningful state walks needs to be defined, that are: +BR: Break – After a single talk period of one person, a silent period follows, after which single talk of the same person continues. +AS: Alternating silence – After a single talk period of one person, a silence follows, after which single talk of a different person continues. +NI: Non-successful interruption – After a single talk period of one person, double talk or multi talk (e.g., triple talk) follows, after which single talk of the same person continues. +SI: Successful interruption – After a single talk period of one person, double talk or multi talk (e.g., triple talk) follows, after which single talk of a different person who has started to talk in the double/multi talk period continues. +2. Compute for each person a speaker state pattern using the conversation surface structure analysis paradigm (clause 7.4) as it was experienced by the persons at the different ends which means the respective other persons need to be delayed if delay was present. See Figure 6. +3. Based on the state pattern, state walks can be detected. For each walk detected in i.e., the conversational reality of person A it is then looked into the same walk at the reality of person B to see how person B experience this walk. The origin for each walk is always a + +single talk period. When looking at person A's reality e.g. single talk of person A, followed by a silence, followed by single talk of person A, this will be noted as a break by person A in A's reality. In case of two interlocutors, eight different kinds of walks can occur, the four types listed above, starting with either person A or person B in single talk: + +| State walk | Description | States transitions (see Figure 4) | Speaker turn | +|------------|---------------------------------------------------------|----------------------------------------------------|--------------| +| BR.AA | Break, starting and ending with A | $S_{2,A} \Rightarrow S_{2,0} \Rightarrow S_{2,A}$ | No | +| BR.BB | Break, starting and ending with B | $S_{2,B} \Rightarrow S_{2,0} \Rightarrow S_{2,B}$ | No | +| AS.AB | Alternating silence, starting with A, ending with B | $S_{2,A} \Rightarrow S_{2,0} \Rightarrow S_{2,B}$ | Yes | +| AS.BA | Alternating silence, starting with B, ending with A | $S_{2,B} \Rightarrow S_{2,0} \Rightarrow S_{2,A}$ | Yes | +| NI.AA | Non-successful interruption, starting and ending with A | $S_{2,A} \Rightarrow M_{2,AB} \Rightarrow S_{2,A}$ | No | +| NI.BB | Non-successful interruption, starting and ending with B | $S_{2,B} \Rightarrow M_{2,AB} \Rightarrow S_{2,B}$ | No | +| SI.AB | Successful interruption, starting with A, ending with B | $S_{2,A} \Rightarrow M_{2,AB} \Rightarrow S_{2,B}$ | Yes | +| SI.BA | Successful interruption, starting with B, ending with A | $S_{2,B} \Rightarrow M_{2,AB} \Rightarrow S_{2,A}$ | Yes | + +Similarly, such state walks can be computed for multiparty scenarios. Here examples for a three-party call for state walks starting with person A: + +| State walk | Description | States transitions (see Figure 4) | Speaker turn | +|------------|-----------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------| +| BR.AA | Break | $S_{3,A} \Rightarrow S_{3,0} \Rightarrow S_{3,A}$ | No | +| AS.AB | Alternating silence | $S_{3,A} \Rightarrow S_{3,0} \Rightarrow S_{3,B}$ | Yes | +| AS.AC | Alternating silence | $S_{3,A} \Rightarrow S_{3,0} \Rightarrow S_{3,C}$ | Yes | +| NI.AA | Non-successful interruption | $S_{3,A} \Rightarrow M_{3,AB} (\Leftrightarrow M_{3,ABC} \Leftrightarrow) \Rightarrow S_{3,A}$
or: $S_{3,A} \Rightarrow M_{3,AC} (\Leftrightarrow M_{3,ABC} \Leftrightarrow) \Rightarrow S_{3,A}$ | No | +| SI.AB | Successful interruption | $S_{3,A} \Rightarrow M_{3,AB} (\Leftrightarrow M_{3,ABC} \Leftrightarrow) \Rightarrow S_{3,B}$
or: $S_{3,A} \Rightarrow M_{3,AC} (\Leftrightarrow M_{3,ABC} \Leftrightarrow) \Rightarrow S_{3,B}$ | Yes | +| SI.AC | Successful interruption | $S_{3,A} \Rightarrow M_{3,AB} (\Leftrightarrow M_{3,ABC} \Leftrightarrow) \Rightarrow S_{3,C}$
or: $S_{3,A} \Rightarrow M_{3,AC} (\Leftrightarrow M_{3,ABC} \Leftrightarrow) \Rightarrow S_{3,C}$ | Yes | + +In the next step the first period of single talk of a detected walk in one person's reality is searched in the other persons' reality and the walk that is found there is noted. It can be the same kind of walk or a different one. In this way, the number of matching walks and mismatching walks can be counted. A convenient approach for this is to use a confusion matrix as shown below: the rows represent the walks found in the "own" reality of a person, e.g., person A; the columns represent the walks found in the reality of the other person, e.g., person B; and the matrix elements contain the counts. In such a matrix, the diagonal represents the matching walks (indicated with $\bullet$ ), all other matrix elements represent the non-matching walks (indicated with $\times$ ). + +| | | Reality of person B | | | | | | | | +|---------------------|-------|---------------------|-------|-------|-------|-------|-------|-------|-------| +| | | BR.AA | BR.BB | AS.AB | AS.BA | NI.AA | NI.BB | SI.AB | SI.BA | +| Reality of person A | BR.AA | ● | × | × | × | × | × | × | × | +| | BR.BB | × | ● | × | × | × | × | × | × | +| | AS.AB | × | × | ● | × | × | × | × | × | +| | AS.BA | × | × | × | ● | × | × | × | × | +| | NI.AA | × | × | × | × | ● | × | × | × | +| | NI.BB | × | × | × | × | × | ● | × | × | +| | SI.AB | × | × | × | × | × | × | ● | × | +| | SI.BA | × | × | × | × | × | × | × | ● | + +In case of a multiparty scenario, this confusion matrix is extended with the additional state walks and it is computed for each pair of the reality of two persons. + +Then, from each such confusion matrix between two persons X and Y, the total number of matching and non-matching walks can be computed: $\text{numMatch\_XY} = \text{sum of counts in the diagonal elements}$ $\text{numMismatch\_XY} = \text{sum of counts in all non-diagonal elements}$ + +Finally, the number of mismatching walks needs to be related to the number of matching walks for building the parameter *divergence* [b-Schoenenberg, 2014], which is computed for each person using the corresponding confusion matrices. + +For the two-party scenario this leads to $\text{Divergence\_A} = \text{Divergence\_B} = \text{numMismatch\_AB} / \text{numMatch\_AB}$ + +For the three-party scenario this leads to: + +$\text{Divergence\_A} = (\text{numMismatch\_AB} + \text{numMismatch\_AC}) / (\text{numMatch\_AB} + \text{numMatch\_AC})$ + +$\text{Divergence\_B} = (\text{numMismatch\_AB} + \text{numMismatch\_BC}) / (\text{numMatch\_AB} + \text{numMatch\_BC})$ + +$\text{Divergence\_C} = (\text{numMismatch\_AC} + \text{numMismatch\_BC}) / (\text{numMatch\_AC} + \text{numMatch\_BC})$ + +The divergence is small if conversational realities are similar and high if they are rather different. + +## 10.4 Communication system effectiveness tests + +System effectiveness refers to the time that is needed to achieve the goals of a conversation using a communication system, relative to the time taken in a face to face meeting. Transmission delays will reduce the system effectiveness, increasing the meeting time not just because of the increased delay, but also because of the impact of the delay on the conversation turn-taking described above. + +## 10.5 Micro-feature tests + +An example of a micro-efficiency test is the task efficiency test [ITU-T P.1312] which is specifically designed to measure how well a system supports discussions with high levels of multi-talk. Whilst not designed to measure delay performance, it is delay sensitive since good performance in multi-talk relies on good interactivity behaviour which is in turn influenced by delay. This is shown in Appendix 3 of [ITU-T P.1312]. + +Other tests designed to focus on conversation features such as false starts or turn-taking can also be envisaged. + +# **11 Areas for further research regarding the impact of delay** + +There are many aspects of the impact of delay not yet fully understood and warrant further study. These include: + +- 1) The effects of asymmetric delay conditions between different participants in a telemeeting. +- 2) The effects of varying delays in a telemeeting. +- 3) Co-dependencies to other impairments such as echo, low frame rate, frame size. +- 4) The relationship between the impact of delay and the features of different types of audio rendering. +- 5) The impact of delay in relation to factors like the types of clients (screen size) and the number of participants using each client. +- 6) The relationship between delay and spatial audio rendering. +- 7) Analysis algorithms based on conversation analysis and/or turn-taking. + +## Annex A + +## Suggestions for free-conversation tasks + +(This annex forms an integral part of this Recommendation.) + +In [b-Berndtsson, 2012] it was noted that the best type of task for evaluating the impact of delay was a free conversation, where the subjects are free to express themselves in a natural manner, such as they would in normal life. However, it is also important to stimulate conversation of an appropriate style for the test. Typical conversation styles might include: + +- a. Relaxed, informal conversation. +- b. Formal, 'business' style conversation. +- c. Focused, problem solving discussion. + +A range of test tasks for audio conversational tests are described in [ITU-T P.800] and [ITU-T P.805], with further test tasks for audiovisual conversational tests are described in [ITU-T P.920]. However, these are primarily tasks suitable for two-way conversations and are not all suitable for extension to the multi-party situation. A range of tasks more appropriate for multi-party tests is given in Appendices III and V of [ITU-T P.1301]. + +### A.1 'Holiday' task + +In this task the participants are asked to: + +- a. Create a list of features that their ideal holiday would include (5 minutes). +- b. List them in order of importance (5 minutes). + +This task approach can be extended to include features of an ideal job, pastime, film, etc. + +Good points: A very good task for getting people involved in an informal type meeting – people like talking about experiences they enjoy. + +Poor points: It does not guarantee a balanced conversation – some participants could dominate or be too passive. + +### A.2 Role-playing games + +In this the test subjects take part in a role in a fictitious scenario. This is not a scripted test such as SCT, but relies on the test subjects taking on a role in a scenario. A good example of this is the "You be the Judge" game, () where the test subjects take on roles in a courtroom such as judge, council for the prosecution, defence, defendant, etc., and role play a series of scenarios in order to reach a solution. + +Good points: Stylistically this tends to replicate business type formal meetings. + +Poor points: Relies on subjects acting abilities and willingness to play a sufficiently active part. + +### A.3 "Who am I?" Celebrity Name-Guessing Task + +This task has been proposed in [b-Schoenenberg, 2014b] as a task for audiovisual communication with high delay sensitivity. In this task, participants are supposed to be a celebrity, and the task is to guess their own name. + +The task should not be confused with the *Name-Guessing Task* described in [ITU-T P.920]. The *Celebrity Name-Guessing Task* described here has strong social component due to the aspect that a participant is associated with a celebrity that he or she has to guess. In addition, the game character of this task let people find themselves in a joyful state, have fun and interact naturally. + +The game rules are described in [b-Schoenenberg, 2016] as follows: + +*Each participant receives a note with initials of a celebrity person that he or she has to guess. This note is clipped to the shirt of the person so that they and their interlocutor are reminded of it. The respective other knows the full name of the unknown celebrity person due to a hidden note with the full name. One participant begins by asking questions that can only be answered with "yes" or "no", for example, "Am I female?" or "Am I a movie star?". As long as the response of the other one is "yes" the participant is allowed to keep asking. If the answer is "no" the roles of asking and answering person are reversed. In this way, the roles are switched several times until one person correctly guesses the name. If one participant wins a game he or she gets a point. After the entire session (nine games) the participant who won most games wins the entire session and with this an additional prize (a voucher).* + +Advantage(s): Let people use both audio and video channel to communicate. In particular, the primarily social interaction in this task facilitates visual communication in terms of nonverbal signals such as smiles (and generally facial expressions), gestures or the change of body posture and gaze. + +Disadvantage(s): Works best for participants who know each other. For participants who do not know each other, difficulties in guessing the other's name may lead to frustration, especially when the reactions of the conversation partner cannot be properly interpreted due to the lack of familiarity. + +### **A.4 Navigation tasks** + +Each subject has a partially complete map of the same area, and take turns to explain to the other participants a route through the map. This will include filling in the incomplete sections of the map. Ideally the subjects will need to exchange and confirm a lot of relatively simple information. + +The time taken to complete the task may also be used as an indication of the performance. + +Good points: Forces all subjects to get involved. + +Poor points: Poor for subjective opinion, since subjects tend to rate the complexity of the task. + +### **A.5 Story with missing parts** + +In this task the test subjects each get a paper with the same story but with various small sections missing. One participant starts to read the story and the other participants reads the text silently, and interrupts when the reader missed a part of the text. Then the person that noticed that there was a part missing reads that part and continues to read the text until he is interrupted in a similar manner. As the reader continues to read until he hears the other person speak, this conversational task leads to several double talk situations. It is recommended that each conversation lasts between 3 and 5 minutes. + +Good points: Extendable to many test subjects. + +Poor points: Not suitable for audiovisual tests as the test subjects need to read the stories and are not free to look at their conversation partners. + +### **A.6 Building blocks task – variation** + +This is an extension of the small building blocks task described in [ITU-T P.920] for small groups. In the [ITU-T P.920] version one participant has an assembled version of a model and the other a disassembled version. The goal is that both participants have an assembled version at the end. To extend this to a multi-party scenario, all participants have a disassembled version of the model. The instruction manual to build the final model is however in parts distributed to the participant. The goal is again that each participant has an assembled version of the model in the end. Also the + +[ITU-T P.920] recommendation of a minimum size of $3\text{ cm} \times 3\text{ cm} \times 2\text{ cm}$ should be adjusted to video quality and resolution of the system under test. + +Good points: Requires audiovisual interaction. + +Poor points: Does not represent a typical conversation. + +### **A.7 Survival task – variations** + +This is a task similar to the survival task in [ITU-T P.1301]. Also in this task, the participants are asked to imagine themselves in a dangerous outdoor situation. Contrary to the "rank items"-style of the [ITU-T P.1301] survival task, this task has a quiz style "questions and select answer" interaction. This style makes it easier to use it multiple times with the same participants for a repeated measure design assessment. It shares many properties with the [ITU-T P.1301] survival task: it is a team building exercise in a cooperative consensus seeking task for small groups, thus suitable for participants which are not familiar with each other. It is further also a "knowledge free" task, meaning that it is assumed that participants are not experts in outdoor survival. The task was found to stimulate discussion and was commented as engaging by most participants. In the task a question about a particular outdoor situation is given with three answer options. The correct answer to the question is based on expert opinions and are accompanied with a specific reasoning. Questions and answers can be found under [b-Biech, 2007]. Every participant should answer the questions first alone before the group discussion begins. This is meant to help to assess how well this group of people works as a team: for a working team the participants should achieve at least the amount of points the highest scoring participant had. + +Good points: Good for assessing goal oriented ad-hoc group discussions. + +Poor points: Participation might be very unequal based on knowledge about the topic. + +## Annex B + +## Suggestions for delay-critical tasks + +(This annex forms an integral part of this Recommendation.) + +### B.1 Information exchange tasks + +A number of tasks to evaluate the effects of speech delay on communication quality are provided in Appendix I of [ITU-T P.920]. The first 5 of these are simple information exchange tasks and as such very delay sensitive. + +- 1) take turns in counting; +- 2) take turns reading random numbers aloud as quickly as possible; +- 3) take turns verifying random numbers aloud as quickly as possible; +- 4) words with missing letters are completed with letters supplied by the other talker; +- 5) take turns verifying city names as quickly as possible. + +### B.2 Random number verification task timed + +This task is a modified version of the number verification task, described in Appendix VIII in [ITU-T P.805]. The essential difference is that test participants are explicitly encouraged to solve the task as quickly as possible, for instance by informing them, that the fastest group of participants in the whole experiment will win a price. This modification has been found to increase the sensitivity [b-Schoenenberg, 2014]. + +An extension to three parties has been developed for the German language as well [b-Schoenenberg, 2014]. + +Advantage(s): High sensitivity to delay. Triggers fast interaction between participants. + +Disadvantage(s): Lack of naturalness of the task. Participants may try short-cuts by violating the rules of taking turns when reading and confirming the numbers. Thus, test supervisors should listen to the conversations and remind participants not to use short-cuts. + +# Bibliography + +- [b-Berndtsson, 2012] Berndtsson, G., Folkesson, M. and Kulyk, V. (2012), *Subjective quality assessment of video conferences and telemeetings*, Packet video workshop 2012, München, Germany. +- [b-Biech, 2007] Biech, E. (2007), *The Pfeiffer book of successful team-building tools: Best of the annuals*. Pfeiffer. +- [b-Brady, 1965] Brady, P.T. (1965), *A technique for investigating on-off patterns of speech*, Bell Syst. Tech. J., Vol. 44, No. 1, pp. 1-22. +- [b-Brady, 1968] Brady, P.T. (1968), *A statistical analysis of on-off patterns in 16 conversations*, Bell Syst. Tech. J., Vol. 47, No. 1, pp. 73-99. +- [b-Brady, 1971] Brady, P.T. (1971), *Effects of transmission delay on conversational behavior on echo-free telephone circuits*, Bell Syst. Tech. J., Vol. 50, No. 1, pp. 115-134. +- [b-Clift] Clift, R., *Conversation Analysis*, (Pre-publication draft), pp. 56-83. +- [b-Egger, 2012] Egger, S., Schatz, R., Schoenenberg, K., Raake, A. and Kubin, G. (2012), *Same but Different? – Using Speech Signal Features for Comparing Conversational VoIP Quality Studies*, IEEE International Conference on Communications (ICC 2012), Ottawa, Canada. +- [b-Gisladottir, 2015] Gisladottir, R., Chwillia, D.J., Levinson, S.C. (2015), *Conversation Electrified: ERP Correlates of Speech Act Recognition in Underspecified Utterances*, PLOS ONE | DOI:10.1371/journal.pone.0120068 March 20. +- [b-Guégain, 2006] Guégain, M., Le Bouquin-Jeannès, R., Faucon, G. and Barriac, V. (2006), *Towards an objective model of the conversational speech quality*, in Int. Conf. on Acoustics, Speech and Signal Processing (ICASP), IEEE. +- [b-Hammer, 2004] Hammer, Florian, Reichl, Peter, and Raake, Alexander (2004), *Elements of interactivity in telephone conversations*. In: Proceedings of the 8th International Conference on Spoken Language Processing (Interspeech, 2004). Jeju Island, Korea, 2004, pp. 1741-1744. +- [b-ITU Handbook] *ITU Handbook: Practical procedures for subjective testing* (2011). +- [b-Jefferson, 1989] Jefferson, G. (1989), *Preliminary notes on a possible metric which provides for a 'standard maximum' silence of approximately one second in conversation*. Conversation: An interdisciplinary perspective. Intercommunication series, 3, (pp. 166-196). Roger, Derek (Ed); Bull, Peter (Ed). +- [b-Kitawaki, 1991] Kitawaki, N. and K. Itoh, K. (1991), *Pure delay effects on speech quality in telecommunications*, in IEEE Journal on Selected Areas in Communications, Vol. 9, No. 4, pp. 586-593. +- [b-Knapp, 2010] Knapp, M.L. and Hall, J.A. (2010), *Nonverbal communication in human interaction*, 7th edition, Wadsworth, Cenage Learning, Boston, USA. + +- [b-OConnail, 1993] O'Connail, B., Whittaker, S. and Wilbur, J.A. (1993), *Conversations over video conferences: An evaluation of the spoken aspects of video-mediated communication*, Human-Computer Interaction, Vol. 8, pp. 389-428. +- [b-Raake, 2013] Raake, A., Schoenenberg, K., Skowronek, J. & Egger, S. (2013), *Predicting speech quality based on interactivity and delay*. In: Proc. Interspeech, Lyon, France, August. +- [b-Sacks,1974] Sacks, H., Schegloff, E.A., Jefferson, G. (1974), *A simplest systematics for the organisation of turn-taking in conversation*. [Journal] // Language 50. pp. 696-735. +- [b-Schmitt, 2014a] Schmitt, M., Gunkel, S., Cesar, P., and Bulterman, D. (2014), *The influence of interactivity patterns on the Quality of Experience in multi-party video-mediated conversations under symmetric delay conditions* in Proceedings of the 3rd International Workshop on Socially-aware Multimedia, New York, NY, USA. +- [b-Schmitt, 2014b] Schmitt, M., Gunkel, S., Pablo, C., and Bulterman, D. (2014), *Asymmetric Delay in Video-Mediated Group Discussions* in 6th International Workshop on Quality of Multimedia Experience (QoMEX). +- [b-Schoenenberg, 2011] Schoenenberg, K., Raake, A. & Skowronek, J. (2011), *A conversation analytic approach to the prediction of leadership in two to six-party audio conferences*. Third International Workshop on Quality of Multimedia Experience (QoMex), pp. 119-124. +- [b-Schoenenberg, 2014] Schoenenberg, K., Raake, A., Egger, S., & Schatz, R. (2014), *On interaction behaviour in telephone conversations under transmission delay*. Speech Communication, 63-64, pp. 1-14. +- [b-Schoenenberg, 2014b] Schoenenberg, K., Raake, A., Lebreton, P. (2014), *Conversational Quality and Visual Interaction of Video-Telephony under Synchronous and Asynchronous Transmission Delay*, Sixth International Workshop on Quality of Multimedia Experience (QoMex). +- [b-Schoenenberg, 2016] Schoenenberg, K. (2016), *The Quality of Mediated-Conversations under Transmission Delay*. PhD thesis, TU Berlin. +- [b-Tam, 2012] Tam, J., Carter, E., Kiesler, S. and Hodgins, J. (2012), *Video increases the perception of naturalness during remote interactions with latency*, in Proc. of CHI'12, New York, NY, USA, pp. 2045-2050. +- [b-Tang, 1992] Tang, J.C. (1992), *Why Do Users Like Video? Studies of Multimedia-Supported Collaboration*, Sun Microsystems, Inc., Mountain View, CA, USA. + + + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Terminals and subjective and objective assessment methods | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects and next-generation networks, Internet of Things and smart cities | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.16-198811-I_PDF-E/raw.md b/marked/P/T-REC-P.16-198811-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..12ca490665ed47ced85b1afed04d175b11943a87 --- /dev/null +++ b/marked/P/T-REC-P.16-198811-I_PDF-E/raw.md @@ -0,0 +1,241 @@ + + +![ITU logo](2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg) + +The logo of the International Telecommunication Union (ITU) features the letters 'ITU' in a bold, sans-serif font, superimposed on a stylized globe with latitude and longitude lines. A lightning bolt symbol is positioned to the right of the globe. + +ITU logo + +INTERNATIONAL TELECOMMUNICATION UNION + +**ITU-T** + +**P.16** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +# **TELEPHONE TRANSMISSION QUALITY** + +## **VOCABULARY AND EFFECTS OF TRANSMISSION PARAMETERS ON CUSTOMER OPINION OF TRANSMISSION QUALITY AND THEIR ASSESSMENT** --- + +## **SUBJECTIVE EFFECTS OF DIRECT CROSSTALK; THRESHOLDS OF AUDIBILITY AND INTELLIGIBILITY** + +**ITU-T Recommendation P.16** + +(Extract from the *Blue Book*) + +--- + +## NOTES + +1 ITU-T Recommendation P.16 was published in Volume V of the *Blue Book*. This file is an extract from the *Blue Book*. While the presentation and layout of the text might be slightly different from the *Blue Book* version, the contents of the file are identical to the *Blue Book* version and copyright conditions remain unchanged (see below). + +2 In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +### **SUBJECTIVE EFFECTS OF DIRECT CROSSTALK; THRESHOLDS OF AUDIBILITY AND INTELLIGIBILITY** + +*(Geneva, 1972; amended at Geneva, 1976, 1980; +Malaga-Torremolinos, 1984; Melbourne, 1988)* + +#### **1 Factors which affect the crosstalk threshold** + +The degree of audibility and intelligibility of a crosstalk signal depends on a large number of factors. + +The main factors influencing the intelligibility of the vocal crosstalk signal are listed below. + +##### **1.1 *Quality of transmission of telephone apparatus*** + +The sending and receiving loudness ratings are decisive factors. The same is true of the sidetone rating when room noise is present. The use of modern telephone apparatus with smooth frequency response curves is assumed. + +##### **1.2 *Circuit noise*** + +The circuit noise on the connection of the disturbed call must be taken into account. The noise level is measured by a psophometer equipped with a weighting network for telephone circuits, as described in Recommendation O.41. + +##### **1.3 *Room noise*** + +Room noise affects the ear directly through earcap leakage between the ear and the receiver and indirectly by sidetone. Sidetone also depends on operating conditions. Unlike circuit noise, the effect of room noise can be reduced to some extent by the user of the telephone. For this reason, and to allow for unfavourable cases, measurements on the audibility of crosstalk have been made with slight room noise as well as with negligible room noise. Because the audibility threshold is very sensitive to masking effects, "negligible" room noise means a noise level well below 10 dBA. The relatively low noise level of 40 dBA has a very marked masking effect and may therefore serve as an example of "slight" room noise. + +##### **1.4 *Telephone set noise*** + +In addition to the masking effects on crosstalk by circuit noise and room noise, the internal noise of the telephone set in the disturbed connection has to be considered. In modern telephone sets this noise is generated in the electronic circuitry (amplifiers, etc.) while in older sets the origin is noise from the carbon microphone. The internal noise can be expressed and treated as an equivalent circuit noise. + +##### **1.5 *Conversation on the disturbed connection*** + +During active speech on the disturbed connection, practical levels of crosstalk are inaudible. However, before the conversation starts or during long pauses in the conversation, it is possible for crosstalk to be heard and perhaps understood. In general, it would be unwise to plan on the basis that the disturbed connection is always active; accordingly, the information given in this Recommendation assumes no conversation on the disturbed connection. + +##### **1.6 *Crosstalk coupling*** + +The intelligibility of a vocal crosstalk signal depends largely on the nature of the crosstalk coupling, which is generally a function of frequency. + +The loudness rating of the crosstalk transmission path – from the speech signal present on the disturbing line to the subscriber's set subject to the disturbance – can be divided into the loudness loss of the crosstalk path from the disturbing to the disturbed line and the receive loudness rating of the disturbed subscriber's set. Figure 1/P.16 illustrates this subdivision. + +![Diagram illustrating the subdivision of crosstalk transmission path. It shows two parallel horizontal lines representing telephone lines. The top line has a sending terminal 'e' connected to a voice source 'Vc' and a receiving terminal 'r' connected to a Psophometer 'mV'. The bottom line has terminals 'B' at both ends. A dashed curved line labeled 'Crosstalk transmission path' connects terminal 'e' on the top line to terminal 'B' on the bottom line. A solid curved line labeled 'd' (loudness loss) connects terminal 'B' on the bottom line to terminal 'r' on the top line. An arrow labeled 'Direction of the voice signal' points to the right on the top line. The diagram is labeled 'T1202020-88'.](2763901b7a1fd1b5d704cdc450d12ed0_img.jpg) + +Diagram illustrating the subdivision of crosstalk transmission path. It shows two parallel horizontal lines representing telephone lines. The top line has a sending terminal 'e' connected to a voice source 'Vc' and a receiving terminal 'r' connected to a Psophometer 'mV'. The bottom line has terminals 'B' at both ends. A dashed curved line labeled 'Crosstalk transmission path' connects terminal 'e' on the top line to terminal 'B' on the bottom line. A solid curved line labeled 'd' (loudness loss) connects terminal 'B' on the bottom line to terminal 'r' on the top line. An arrow labeled 'Direction of the voice signal' points to the right on the top line. The diagram is labeled 'T1202020-88'. + +- e sending LR of the disturbing subscriber's set +- r receiving LR of the disturbed subscriber's set +- d crosstalk path loudness loss +- B terminals of subscriber's set +- $V_c$ conversational speech level + +Disturbing and disturbed subscriber's sets at the same end: near-end crosstalk. + +Disturbing and disturbed subscriber's sets at opposite ends: far-end crosstalk. + +FIGURE 1/P.16 + +#### Subdivision of crosstalk transmission path + +For a given speech level $V_c$ , the intelligibility of the crosstalk signal depends on the loudness rating $d + r$ . In Recommendation G.111, § A.4.4.4, the crosstalk receive loudness rating is defined as: + +$$XRLR = RLR(set) + L_x$$ + +where $RLR(set)$ refers to the disturbed telephone set. + +The crosstalk loudness $L_x$ is computed as a loudness rating but with the exponent $m = 1$ , which is valid near the audibility threshold. + +In the absence of further information, the value of $L_x$ may be approximately taken as the attenuation measured or calculated at a frequency of 1020 Hz. + +#### 2 Median listener threshold of the audibility and intelligibility of vocal crosstalk + +The curves in Figure 2/P.16 represent the crosstalk receive loudness rating corresponding to the threshold of audibility and intelligibility ( $XRLR_t$ ) as a function of circuit noise. For planning purposes, it is recommended that room noise be regarded as negligible, which represents the most unfavourable condition. + +![Figure 2/P.16: Threshold value of crosstalk receive loudness rating as a function of circuit noise. The graph shows two curves on a grid. The y-axis is 'Crosstalk receive loudness rating (XRLR_t) dB' from 40 to 80. The x-axis is 'Circuit noise dBmp' from -90 to -50. The top curve is labeled 'Audibility threshold' and the bottom curve is labeled 'Intelligibility threshold'. Both curves decrease as circuit noise increases. A label '(Without sidetone)' is at the top. A small code 'T1202030-88' is at the bottom right.](de6e8b740c69dac308cce9edfec3eff4_img.jpg) + +| Circuit noise (dBmp) | Audibility threshold (XRLR t dB) | Intelligibility threshold (XRLR t dB) | +|----------------------|---------------------------------------------|--------------------------------------------------| +| -90 | 80 | 67 | +| -80 | 75 | 62 | +| -70 | 68 | 55 | +| -60 | 60 | 48 | +| -50 | 50 | 40 | + +Figure 2/P.16: Threshold value of crosstalk receive loudness rating as a function of circuit noise. The graph shows two curves on a grid. The y-axis is 'Crosstalk receive loudness rating (XRLR\_t) dB' from 40 to 80. The x-axis is 'Circuit noise dBmp' from -90 to -50. The top curve is labeled 'Audibility threshold' and the bottom curve is labeled 'Intelligibility threshold'. Both curves decrease as circuit noise increases. A label '(Without sidetone)' is at the top. A small code 'T1202030-88' is at the bottom right. + +*Note 1* – The circuit noise is referred to the terminals of the subscriber's set. Room noise is assumed to be negligible. + +*Note 2* – The vocal level on the disturbing line is assumed to be $-18$ dBV active speech level. + +FIGURE 2/P.16 + +Threshold value of crosstalk receive loudness rating as a function of circuit noise + +The criterion for the threshold of audibility is that the presence of a speech signal is only just detectable but that no part of the speech can be understood. The criterion for the threshold of intelligibility is that single words or phrases can sometimes be understood while listening to a conversation. + +The threshold curves represent median values for the two criteria such that in each case 50% of subscriber's opinions are respectively above and below the particular curve. The standard deviation for listeners has been observed to lie in the range 4 to 6 dB and a value of 5 dB is recommended for planning purposes. Typical response curves for a large sample of listeners for the threshold criteria are shown in Figure 3/P.16 (no circuit noise). The difference in XRLR between the two curves is about 12 dB. + +The results of the original experiments (from which the curves in Figure 2/P.16 were drawn) were expressed in terms of speech level (e.g. in Volume Units (VU)) and on that basis showed a satisfactory degree of coherence. + +However, earlier versions of Recommendation P.16 were based on the assumption that there is a fixed relationship between the sending loudness rating and the speech level on the line. This assumption required a correction in the range of 11 dB and is therefore not justified. Furthermore, speech levels expressed in Volume Units appear to differ systematically as measured in different countries on identical speech samples. Therefore, a fixed speech level on the disturbing line is assumed, independent of the send loudness rating (SLR) of that circuit. + +The thresholds given in Figure 2/P.16 are based on the assumption that the speech level $V_c$ under normal conversational conditions is $-18$ dBV active speech level (measured according to Recommendation P.56) at the terminal of the disturbing telephone set. This value is the estimated average of the conversational level in many countries at the send end of a connection with fairly high overall loudness rating [between the optimum and the maximum permitted (OLR)]. + +![Figure 3/P.16: A graph showing the probability (0 to 100%) versus Crosstalk receive loudness rating (XRLR) in dB (50 to 100). Two curves are plotted: 'Audibility' and 'Intelligibility'. The 'Audibility' curve starts at 100% probability at 50 dB and decreases to 0% at approximately 95 dB. The 'Intelligibility' curve starts at 100% probability at 50 dB and decreases to 0% at approximately 85 dB. The graph is labeled 'TI202040-88' in the bottom right corner.](431b8889a0e7f676f0eef40859590349_img.jpg) + +| Crosstalk receive loudness rating (XRLR) [dB] | Audibility Probability [%] | Intelligibility Probability [%] | +|-----------------------------------------------|----------------------------|---------------------------------| +| 50 | 100 | 100 | +| 60 | 95 | 85 | +| 70 | 85 | 40 | +| 80 | 50 | 5 | +| 90 | 10 | 0 | +| 100 | 0 | 0 | + +Figure 3/P.16: A graph showing the probability (0 to 100%) versus Crosstalk receive loudness rating (XRLR) in dB (50 to 100). Two curves are plotted: 'Audibility' and 'Intelligibility'. The 'Audibility' curve starts at 100% probability at 50 dB and decreases to 0% at approximately 95 dB. The 'Intelligibility' curve starts at 100% probability at 50 dB and decreases to 0% at approximately 85 dB. The graph is labeled 'TI202040-88' in the bottom right corner. + +FIGURE 3/P.16 +Listener distribution of crosstalk receive loudness rating +required for different threshold definitions + +The standard deviation of talking levels is fairly high. For calculation purposes a value of $\sigma = 5$ dB should be used. + +To calculate the threshold value for a speech level different from $-18$ dBV, the $XRLR_t$ value should be corrected by the amount of the difference, with its sigh (higher levels require higher $XRLR$ values, and vice versa). + +The value $XRLR_t$ is the sum of the crosstalk path loudness loss and the receiving loudness loss on the disturbed line. In order to obtain the loudness loss of the crosstalk path, $L_x$ , for a particular threshold value, the $RLR(set)$ value has to be subtracted. + +In general, for any speech level and receiving loudness rating, $L_x$ is obtained from Figure 2/P.16 as: + +$$L_x = XRLR_t - RLR(set) + (18 + V_c)$$ + +#### 3 Effects of room noise + +Room noise reaches the listener's ear both by leakage under the earcap of the telephone handset and by the sidetone path. For a given sidetone the room noise can be converted to an equivalent circuit noise by means of a transmission model such as described in Supplement No. 3. A family of conversion curves with sidetone loss as parameter is found in Figure 2 of this Supplement. + +As an example, with a fairly high sidetone loss (the same as used in the previous version of Recommendation P.16) a level of 40 dBA room noise is equivalent to a circuit noise level of $-85$ dBmp. This noise level reduces the threshold $XRLR$ value by about 8 dB. An additional reduction will in most cases be caused by earcap leakage. + +However, the importance of this effect cannot be generally predicted, since it depends both on the shape of the earcap and on user habits. + +#### 4 Crosstalk probability + +While the curves in Figure 2/P.16 present the median values for various noise conditions, the curves in Figure 3/P.16 represent the probability of audible or intelligible crosstalk, in percent, for the negligible noise condition. Similar probability curves can be derived from the median values for any circuit noise condition by the use of cumulative normal distributions with a standard deviation of 6 dB. + +In a more general case, the talker variance should also be added. The mean speech level used in the calculations may be chosen to be lower than the relatively high level assumed in Figure 2/P.16, e.g. $-20$ dBV, which is closer to the average level in the network. An example of such an overall probability calculation is given in Annex A. + +The threshold values of crosstalk loudness rating given in this Recommendation can be used in different ways. One possible interpretation is to require all normal telephone connections (i.e. faulty connections excluded) to have crosstalk conditions between the two threshold criteria. This means that, on the one hand, there is no point in requiring a higher crosstalk attenuation than the one corresponding to the audibility threshold and, on the other hand, that the intelligibility threshold should not be exceeded. + +Another interpretation is to set the requirement so that there is a given small probability (e.g. 5%) that intelligible crosstalk can be encountered with negligible room noise and with the lowest circuit noise level found in the network. In practice, noise conditions are more favourable in the sense that crosstalk quite often is masked by room and circuit noise to the extent of becoming inaudible. For the average of all connections the risk of intelligible crosstalk will therefore be much smaller than the given percentage for the most unfavourable condition. + +Crosstalk requirements may not necessarily be the same for all parts of the network. Although the maintenance of telephone secrecy is primordial, the subscriber is more likely to make a severe judgement on crosstalk in a local call taking place in his immediate environment and in which indiscretion due to crosstalk may have unfortunate social consequences. The problem of "social crosstalk" is dealt with in [1]. + +In practice, simultaneity of speaking on the disturbing line and listening on the disturbed line (during conversation pauses) is not present in all cases. Information concerning this topic and showing how to calculate the probabilities concerned will be found in [2]. + +As guidelines, the probabilities of subscribers encountering potentially intelligible crosstalk should not be worse (i.e. higher) than the following: + +- own exchange calls: 1 in 1000, +- other calls: 1 in 100. + +*Note* – The fundamentals of calculating crosstalk probability in general are considered in Recommendation G.105. + +## ANNEX A + +(to Recommendation P.16) + +### Example of probability calculation + +The probability of understanding single words of a conversation overheard by crosstalk may be calculated for a listener chosen at random from a population of subscribers. The result of such a calculation can be used as a basis for establishing rules for, inter alia, the minimum required crosstalk attenuation between subscriber lines in a national network. + +In order to demonstrate the method of using the information given in this Recommendation to calculate the probability of encountering (intelligible) crosstalk, the following assumptions may be made: + +Mean speech level $V_c = -20$ dBV; + +Receive loudness rating of telephone sets $RLR(set) = -6$ dB; + +No room or circuit noise; + +Standard deviation of talking levels $\sigma_T = 5$ dB; + +Standard deviation of listener response distribution $\sigma_L = 6$ dB; + +Standard deviation of $RLR(set)$ $\sigma_s = 1$ dB. + +The threshold value for crosstalk intelligibility without noise, taken from Figure 2/P.16 is $XRLR_t = 67$ dB. + +According to the formula at the end of § 2 and with the given assumptions, the required median crosstalk path loudness loss becomes: + +$$L_x = 67 + 6 - 2 = 71 \text{ dB.}$$ + +The total standard deviation of the probability function is: + +$$\sqrt{\sigma_T^2 + \sigma_L^2 + \sigma_s^2} = 7.87 \text{ dB}$$ + +With these values of $L_x$ and $\sigma$ , a cumulative normal distribution function as in Figure A-1/P.16 can be drawn. The function indicates the probability that a listener can understand single words if crosstalk for a specific value of the crosstalk path loudness loss. For example, for $L_x = 75$ the probability is 30%. On the other hand, to obtain only 5% probability a crosstalk path loudness loss of 84 dB would be necessary. For 1% probability, 89 dB would be required, as well as 95 dB for 0.1% probability. + +This calculation was based on some typical values of speech level and receiving sensitivity under noise-free conditions. Similar calculations can easily be made with other data, also including the effects of noise. For a realistic estimation of the probability of intelligible crosstalk for subscribers in general, some statistical distribution of circuit noise (and possibly of room noise at the subscriber's locations) will have to be assumed. + +![Figure A-1/P.16: A graph showing the probability of understanding single words of an overheard conversation as a function of the (weighted) crosstalk path loudness loss L_x. The y-axis is labeled 'Probability' and uses a percentage scale from 0.01 to 99.9. The x-axis is labeled 'Crosstalk path loudness loss (L_x)' and ranges from 50 to 100 dB. A straight line is plotted on the graph, indicating a normal distribution. The line passes through approximately (50, 99.9), (60, 95), (70, 50), (80, 10), (90, 1), and (100, 0.01).](643d86ebba41e16a88461bfcb3741de6_img.jpg) + +Figure A-1/P.16: A graph showing the probability of understanding single words of an overheard conversation as a function of the (weighted) crosstalk path loudness loss L\_x. The y-axis is labeled 'Probability' and uses a percentage scale from 0.01 to 99.9. The x-axis is labeled 'Crosstalk path loudness loss (L\_x)' and ranges from 50 to 100 dB. A straight line is plotted on the graph, indicating a normal distribution. The line passes through approximately (50, 99.9), (60, 95), (70, 50), (80, 10), (90, 1), and (100, 0.01). + +$V_c$ mean = $-20$ dBV; $RLR(set) = -6$ dB; standard deviation = $7.9$ dB + +FIGURE A-1/P.16 + +Probability of understanding single words of an overheard conversation as a function of the (weighted) crosstalk path loudness loss $L_x$ + +## References + +- [1] WILLIAMS (H.), SILOCOCK (W. W.), SIBBALD (D.): Social crosstalk in the local area network, *El. Comm.*, Vol. 49, No. 4, London, 1974. +- [2] LAPSA (P. M.): Calculation of multidisturber crosstalk probabilities, *BSTJ*, Vol. 55, No. 7, New York, 1976. \ No newline at end of file diff --git a/marked/P/T-REC-P.300-200111-I_PDF-E/raw.md b/marked/P/T-REC-P.300-200111-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..c595b5738677285c7bab47b3b214fd16437c07a0 --- /dev/null +++ b/marked/P/T-REC-P.300-200111-I_PDF-E/raw.md @@ -0,0 +1,563 @@ + + +![ITU logo: A globe with a lightning bolt and the letters ITU.](2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg) + +ITU logo: A globe with a lightning bolt and the letters ITU. + +INTERNATIONAL TELECOMMUNICATION UNION + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.300** + +(11/2001) + +SERIES P: TELEPHONE TRANSMISSION QUALITY, +TELEPHONE INSTALLATIONS, LOCAL LINE +NETWORKS + +Subscribers' lines and sets + +--- + +**Transmission performance of group audio +terminals (GATs)** + +ITU-T Recommendation P.300 + +--- + +# ITU-T P-SERIES RECOMMENDATIONS + +## TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS + +| | | | +|-----------------------------------------------------------------------------------------------|---------------|--------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Subscribers' lines and sets | Series | P.30 | +| | | P.300 | +| Transmission standards | Series | P.40 | +| Objective measuring apparatus | Series | P.50 | +| | | P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of quality | Series | P.80 | +| | | P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +# **Transmission performance of group audio terminals (GATs)** + +## **Summary** + +This Recommendation provides the speech transmission requirements for Group Audio Terminals (GATs). The associated test methods for verifying the specified performances are contained in Annex A. An *in-situ* procedure for checking and aligning GAT terminals in the field is provided in Appendix I. + +Reference is made, as far as possible, to other Recommendations applicable to speech terminals for analogue and digital networks. Only terminal configurations not already covered by these Recommendations are addressed in detail. + +## **Source** + +ITU-T Recommendation P.300 is a revision of ITU-T Rec. P.30 (1998). It was revised by ITU-T Study Group 12 (2001-2004) and approved under the WTSA Resolution 1 procedure on 29 November 2001. + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications. The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementors are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database. + +© ITU 2002 + +All rights reserved. No part of this publication may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying and microfilm, without permission in writing from ITU. + +## CONTENTS + +###### Page + +| | | | +|-------|---------------------------------------------------------------------------------------------------------|----| +| 1 | Scope..... | 1 | +| 2 | Normative references ..... | 1 | +| 3 | Definitions and abbreviations ..... | 2 | +| 4 | Group Audio Terminals types..... | 2 | +| 5 | Speech transmission characteristics..... | 3 | +| 5.1 | Handset mode ..... | 3 | +| 5.2 | Headset mode..... | 3 | +| 5.3 | Hands-free mode (single-user facilities)..... | 3 | +| 5.4 | Hands-free mode (multiple-users facility)..... | 3 | +| 5.4.1 | Receiving volume control and sensitivity adjustments ..... | 4 | +| 5.4.2 | Sensitivity-frequency responses ..... | 4 | +| 5.4.3 | Loudness Rating ..... | 4 | +| 5.4.4 | Echo path loss characteristics ..... | 5 | +| 5.4.5 | Distortion ..... | 5 | +| 5.4.6 | Out-of-band requirements..... | 5 | +| 5.4.7 | Noise ..... | 5 | +| 5.4.8 | Delay..... | 5 | +| | Annex A – Objective measurement methods for hands-free, multiple users GATs ..... | 6 | +| A.1 | Introduction..... | 6 | +| A.1.1 | Testing arrangements..... | 6 | +| A.1.2 | Test signals levels ..... | 7 | +| A.2 | Testing of transmission requirements ..... | 8 | +| A.2.1 | Sensitivity-frequency response..... | 8 | +| A.2.2 | Loudness rating..... | 9 | +| A.2.3 | Terminal Coupling Loss ..... | 9 | +| | Appendix I – Practical in-situ audio alignment procedure for multiple-users hands-free GATs..... | 10 | +| I.1 | Introduction..... | 10 | +| I.2 | Audio alignment and sensitivity-frequency characteristics ..... | 10 | +| I.2.1 | Sending ..... | 11 | +| I.2.2 | Receiving ..... | 11 | + + + +# **Transmission performance of group audio terminals (GATs)** + +# **1 Scope** + +This Recommendation provides the audio performance requirements and test methods for Group Audio Terminals (GATs), i.e. terminals which are specifically designed for being used by several users at the same time. + +GATs cover a wide range of products, ranging from the hands-free telephone set, when used by several users at the same time, to the teleconference audio terminals incorporating sophisticated echo control mechanisms. GATs can be designed to operate both on the analog POTS networks and on digital ISDN links. While the former can only offer telephone bandwidth performances, the latter can be designed to offer either narrow-band or wideband audio communication facilities. + +This Recommendation addresses this whole range of devices by identifying, as far as possible, the already existing Recommendations applicable to each terminal type. + +# **2 Normative references** + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision: all users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. + +- [1] ITU-T Recommendation P.310 (2000), *Transmission characteristics for telephone-band (300-3400 Hz) digital telephones.* +- [2] ITU-T Recommendation P.311 (1998), *Transmission characteristics for wideband (150-7000 Hz) digital handset telephones.* +- [3] ITU-T Recommendation P.57 (1996), *Artificial ears.* +- [4] ITU-T Recommendation P.10 (1998), *Vocabulary of terms on telephone transmission quality and telephone sets.* +- [5] ITU-T Recommendation P.340 (2000), *Transmission characteristics and speech quality parameters of hands-free telephones.* +- [6] ITU-T Recommendation P.342 (2000), *Transmission characteristics for telephone band (300-3400 Hz) digital loudspeaking and hands-free telephony terminals.* +- [7] ITU-T Recommendation P.341 (1998), *Transmission characteristics for wideband (150-7000 Hz) digital hands-free telephony terminals.* +- [8] ITU-T P-Series Recommendations – Supplement 16 (1988), *Guidelines for placement of microphones and loudspeakers in telephone conference rooms [1] and for Group Audio Terminals (GATs).* +- [9] ITU-T Recommendation P.51 (1996), *Artificial mouth.* +- [10] ITU-T Recommendation P.79 (1999), *Calculation of loudness ratings for telephone sets.* +- [11] ISO 3:1973, *Preferred numbers – Series of preferred numbers.* + +- [12] ITU-T Recommendation G.122 (1993), *Influence of national systems on stability and talker echo in international connections*. +- [13] IEC 60651 (2001), *Sound level meters*. +- [14] ITU-T Recommendation P.50 (1999), *Artificial voices*. + +# 3 Definitions and abbreviations + +Relevant definitions and abbreviations in ITU-T Rec. P.10 [4] apply. + +For the purposes of this Recommendation, the following terms are additionally defined: + +**3.1 Group Audio Terminal (GAT):** Terminal designed for being used by several users at the same time and specifically intended for accessing point-to-point and multipoint teleconference sessions. + +**3.2 reference codec:** Practical implementation of an ideal codec. Used for carrying out speech transmission measurements on digital terminal equipment by means of analogue instrumentation. The Reference Codec shall provide a 0 dBr analogue interface to the testing instrumentation and implement the relevant coding algorithms (e.g. ITU-T Rec. G.711 for telephone band measurements and ITU-T Rec. G.722 for wide band applications). + +**3.3 reference sphere:** One meter radius sphere inside an echo-free environment, where anechoic conditions exist within the tolerance limits specified in ITU-T Rec. P.341 [7] (Table A.1/P.341). Three orthogonal axes are defined through the center of the Reference Sphere: x and y (horizontal) and z (vertical). + +The following abbreviations are used: + +- $d_r$ Loudspeaker-listener nominal operating distance of hands-free multiple users GATs +- $d_s$ Microphone-speaker operating distance of hands-free multiple users GATs +- $F_r$ Correction factor for receiving hands-free measurements of hands-free multiple users GATs +- $F_s$ Correction factor for sending hands-free measurements of hands-free multiple users GATs +- $F_{tcl}$ Correction factor for terminal coupling loss measurements of hands-free multiple users GATs +- ISDN Integrated Services Digital Network +- POTS Plain Old Telephone Service +- RLR Receiving Loudness Rating +- SLR Sending Loudness Rating + +# 4 Group Audio Terminals types + +The GAT terminals are intended for supporting communications between groups of users, either point-to-point or multipoint. + +The following types of audio facilities can be provided by GAT terminals, either separately or in combination: + +- handsets; +- headsets; +- hands-free (single-user facilities); +- hands-free (multiple-users facility). + +GATs can be designed to provide telephone-band and/or wideband audio facilities. + +# 5 Speech transmission characteristics + +The GAT terminals shall be able to interwork (end-to-end) with analogue (POTS) and digital (ISDN) narrow-band telephone sets, as well as with ISDN videophones and wideband telephones. As a consequence, the audio standards applicable to these devices also apply to GAT terminals intended for the same network access, as appropriate and as specified below. + +## 5.1 Handset mode + +For GATs designed for interconnecting with the analogue network access, the same relevant regional requirements applicable to POTS telephone sets are also applicable. + +For GATs designed for interconnecting with the ISDN network access, the requirements and measurement methods of ITU-T Recs. P.310 [1] and P.311 [2] apply, respectively for telephone-band and wideband operations. + +## 5.2 Headset mode + +In principle, the requirements for headset operations shall be based on the requirements for handset operations, with the following allowances: + +- the suitable positioning of the microphone with respect to the Mouth Reference Point (MRP) position shall be stated by the manufacturer; +- an appropriate Artificial Ear shall be used and the measurement results shall be referred to the Ear Reference Point (ERP) position as specified in ITU-T Rec. P.57 [3]. + +## 5.3 Hands-free mode (single-user facilities) + +The following Recommendations apply to GATs terminals designed for providing hands-free facilities to multiple users by combining more single-user facilities. + +ITU-T Rec. P.340 [5] applies to GATs designed for the analogue network access. + +The requirements and measurement methods of ITU-T Recs. P.342 [6] and P.341 [7] respectively apply for GATs intended for connection to the ISDN and implementing telephone-band (ITU-T Rec. G.711) or wide band (ITU-T Rec. G.722) speech coding. + +## 5.4 Hands-free mode (multiple-users facility) + +Multiple users GAT terminals are typically designed for guaranteeing optimum hands-free audio performance when installed in "typical" conference rooms (i.e. in "live" environments with acoustic reverberation phenomena and background noise). To this purpose, suitable signal processing techniques are frequently used for eliminating the reverberation effects in the transduced signals and/or suppressing the room noise. Performance under real operating conditions can then only be evaluated in the actual installation environments, whose characteristics can not be specified in a general way. + +This Recommendation is not then intended to address all the installation conditions of GAT terminals, but specifies their audio performances under reference echo-free conditions. In case this testing environment is not compatible with the technical characteristics of specific terminals, their manufacturers are given the option to state alternative, more suitable, acoustic testing environments for carrying out their characterisation. + +An *in-situ* testing procedure for checking the correct audio alignment of installed terminals is provided in Appendix I. + +Guidelines for the placement of microphones and loudspeakers in conference rooms can be found in Supplement 16 of the P-series Recommendations [8]. + +### **5.4.1 Receiving volume control and sensitivity adjustments** + +Due to the wide range of operating conditions, both the sending and receiving sensitivity of multiple-user handsfree terminals are generally adjusted at their installation on the basis of the room characteristics and of the distance between the transducers and the users. These adjustment controls shall not be accessible to normal users. An additional user-accessible receiving volume control (referred to in the following as "receiving volume control") can be provided in order to allow the user to establish the optimum listening conditions on the basis of the actual room noise, the number of participants and of other variable factors. + +#### **5.4.1.1 Sending sensitivity adjustment** + +The sending sensitivity adjustment is intended to allow for varying the positioning of the microphone(s) with respect to the user(s), according to the installation environment constraints. The manufacturer shall state the regulation range of this control and the microphone-to-speaker nominal operating distance ( $d_s$ ) for which the terminal submitted to the tests has been adjusted. + +#### **5.4.1.2 Receiving sensitivity adjustment** + +The receiving sensitivity adjustment is intended to allow for varying the positioning of the loudspeaker(s) with respect to the user(s) according to the installation environment constraints. The manufacturer shall state the regulation range of this control and the loudspeaker-to-listener nominal operating distance ( $d_r$ ) for which the terminal submitted to the tests has been regulated. + +#### **5.4.1.3 Receiving volume control** + +Unless stated otherwise, the compliance requirements here provided refer to the maximum position (maximum sensitivity) of the receiving volume control (when manually adjustable). + +The dynamic range of the receiving volume control (when manually adjustable) shall be of at least 15 dB. + +#### **5.4.1.4 Adaptive gain control** + +An adaptive gain control, sensible to the background noise level, may be implemented into the GAT. Its action should result into complementary and symmetrical gain variations, in the receiving and in the sending paths, as a function of the ambient noise level. + +### **5.4.2 Sensitivity-frequency responses** + +ITU-T Rec. P.340 [5] applies to GATs designed for the analogue network access. + +The requirements of ITU-T Recs. P.342 [6] and P.341 [7] respectively apply for GATs intended for connection to the ISDN and implementing telephone-band (ITU-T Rec.G.711) or wideband (ITU-T Rec. G.722) speech coding. + +### **5.4.3 Loudness Rating** + +ITU-T Rec. P.340 [5] applies to GATs designed for the analogue network access. + +For GATs intended for connection to the ISDN the following applies. + +#### **5.4.3.1 Sending** + +The nominal value of Sending Loudness Rating (SLR) shall be: + +$$\text{SLR} = (+13 - F_s) \text{ dB}$$ + +where $F_s$ is as defined in A.1.1.1. + +#### **5.4.3.2 Receiving** + +The nominal value of Receiving Loudness Rating (RLR) shall be: + +$$\text{RLR} = (+ 2 - F_r) \text{ dB}$$ + +for GATs implementing telephone-band (ITU-T Rec. G.711) speech coding and: + +$$\text{RLR} = (+ 5 - F_r) \text{ dB}$$ + +for GATs implementing wide band (ITU-T Rec. G.722) speech coding. + +In the above formulae $F_r$ is as defined in A.1.1.2. + +The volume control range should span the value of the receiving loudness rating equal to that of the corresponding handset telephone, as well as an RLR value about 10 dB lower. + +### **5.4.4 Echo path loss characteristics** + +The requirements of ITU-T Recs. P.342 [6] and P.341 [7] respectively apply for the telephone band and wideband operations of GATs intended for connection to the ISDN. + +### **5.4.5 Distortion** + +The requirements of ITU-T Recs. P.342 [6] and P.341 [7] respectively apply for GATs intended for connection to the ISDN and implementing telephone-band (ITU-T Rec. G.711) or wideband (ITU-T Rec. G.722) speech coding. + +### **5.4.6 Out-of-band requirements** + +The requirements of ITU-T Recs. P.342 [6] and P.341 [7] respectively apply for GATs intended for connection to the ISDN and implementing telephone-band (ITU-T Rec. G.711) or wideband (ITU-T Rec. G.722) speech coding. + +### **5.4.7 Noise** + +#### **5.4.7.1 Sending** + +The requirements of ITU-T Recs. P.342 [6] and P.341 [7] respectively apply for GATs intended for connection to the ISDN and implementing telephone-band (ITU-T Rec. G.711) or wideband (ITU-T Rec. G.722) speech coding. + +#### **5.4.7.2 Receiving** + +Irrespectively from the speech coding implemented, the A-weighted noise produced in the receiving direction at the measurement point by GATs intended for connection to the ISDN shall not exceed $-49 \text{ dBPa(A)} + F_r$ . + +### **5.4.8 Delay** + +The requirements of ITU-T Recs. P.342 [6] and P.341 [7] respectively apply for GATs intended for connection to the ISDN and implementing telephone-band (ITU-T Rec. G.711) or wideband (ITU-T Rec. G.722) speech coding. + +# ANNEX A + +## Objective measurement methods for hands-free, multiple users GATs + +## A.1 Introduction + +This annex describes methods which may be used to measure the speech transmission performances of hands-free, multiple users, Group Audio Terminals intended for ISDN connection. + +Unless stated otherwise, the same test conditions apply as specified in applicable ITU-T Recs. as appropriate (i.e. ITU-T Recs. P.342 [6] and P.341 [7] respectively for GATs implementing telephone-band (ITU-T Rec. G.711) or wideband (ITU-T Rec. G.722) speech coding). + +### A.1.1 Testing arrangements + +The acoustic reference conditions for measurement purposes are free-field within the Reference Sphere. However, for specific designs, it is up to the manufacturer to specify an alternative (reverberant) testing environment. In this case, the specified measurement conditions shall be stated in the test report. + +#### A.1.1.1 Sending + +Two different testing conditions shall be used, according to the type of microphone(s) equipping the GAT terminal: + +- microphones placed on the table surface in front of one or more speakers; +- microphone(s) placed elsewhere in the room. + +##### Microphones intended for being placed on the table + +- The arrangement specified in Figure A.1, placing each microphone in turn at point B. The other microphones shall be removed from the acoustic field of the Artificial Mouth. + +##### Microphones intended for being placed elsewhere in the room + +- The acoustic pick-up port of the microphone shall be placed at the centre of the Reference Sphere in an anechoic environment and the Artificial Mouth shall be placed with the lip plane tangential to the Sphere. The reference axis of the Mouth shall coincide with the x-axis of the Sphere, while the microphone shall be oriented as specified by the manufacturer. In case of microphones intended to pick up the sounds coming from multiple directions (either omni-directional or steered devices), the extreme incidence angles on the horizontal plane as stated by the manufacturer shall be tested, as well as an intermediate incidence direction. The testing angles actually used for measurements shall be stated in the test report. + +In order to take into account the difference between the reference test positioning and the actual microphone-speaker operating distance ( $d_s$ ), for which the terminal is adjusted, the following correction factor $F_s$ is defined: + +$$F_s \text{ dB} = 20 \text{ Log } \frac{d_s}{0.5} \quad (d_s \text{ in meters})$$ + +For sending noise measurements the sound source shall be muted and all microphones shall be placed in the testing environment. + +NOTE – The specified methodology is not applicable for testing hands-free terminals equipped with lapel-mounted (lavalier) microphones, the testing method for which is not yet defined. + +![Diagram of the measurement set-up for table placed microphones. It shows a side view of a microphone on a horizontal surface. A dashed vertical line from the microphone to a point 'C' is labeled '30 cm'. A dashed horizontal line from the microphone to the base of the vertical line is labeled '40 cm'. A dashed hypotenuse line from point 'C' to the microphone is labeled '50 cm'. Above point 'C' is a 'Lip ring' and a dashed line labeled 'C' pointing to the lip ring.](0f2a1e4a7b12fe5b8749882ecd636f5c_img.jpg) + +T1213030-02 + +Diagram of the measurement set-up for table placed microphones. It shows a side view of a microphone on a horizontal surface. A dashed vertical line from the microphone to a point 'C' is labeled '30 cm'. A dashed horizontal line from the microphone to the base of the vertical line is labeled '40 cm'. A dashed hypotenuse line from point 'C' to the microphone is labeled '50 cm'. Above point 'C' is a 'Lip ring' and a dashed line labeled 'C' pointing to the lip ring. + +**Figure A.1/P.300 – Measurement set-up for table placed microphones** + +#### A.1.1.2 Receiving + +Each loudspeaker shall be placed in turn at the centre of the Reference Sphere, with its main axis coincident with the x-axis and its outer surface containing the z-axis. The measurement microphone shall be placed on the x-axis, its acoustic centre being at the intersection with the Reference Sphere. + +In order to take into account the difference between the reference test positioning and the actual loudspeaker-listener operating distance ( $d_r$ ) for which the terminal is adjusted, the following correction factor $F_r$ is defined: + +$$F_r \text{ dB} = 20 \text{ Log } \frac{d_r}{0.5} \quad (d_r \text{ in metres})$$ + +#### A.1.1.3 Terminal coupling loss (TCL) + +For TCL and stability measurements the more efficient loudspeaker shall be placed at the centre of the Reference Sphere, with its main axis coincident with the x-axis and its outer surface containing the z-axis. The more efficient microphone shall be placed on the x-axis, its acoustic pick-up port being at the intersection with the Reference Sphere and its maximum efficiency pick-up direction parallel to the x-axis. For surface effect microphones, a suitable horizontal positioning plane shall be used with a minimum surface still guaranteeing the proper operation of the microphone. In any case, the positioning surface shall not enter the reference sphere by more than 100 mm. + +In order to take into account the difference between the reference test positioning and the actual loudspeaker-to-microphone operating distance for which the terminal is adjusted, the following correction factor $F_{tcl}$ is defined: + +$$F_{tcl} \text{ dB} = 20 \text{ Log } \frac{d_{min}}{0.5}$$ + +$d_{min}$ (in metres) is the minimum loudspeaker-to-microphone distance specified by the manufacturer. If no specification is provided, a default $d_{min} = 2 \text{ m}$ value shall be assumed and stated in the test report. + +### A.1.2 Test signals levels + +#### A.1.2.1 Sending + +The test signal level shall be $(-28.7 - F_s) \text{ dBPa}$ at the Hands-Free Reference Point (HFRP). The characteristics of the Artificial Mouth shall be according to ITU-T Rec. P.51 [9]. Two different methods can be alternatively used for calibrating the Artificial Mouth. + +##### **Method 1:** at the HFRP + +The input signal from the Artificial Mouth is equalized under free field conditions at the HFRP, such that the spectrum is as specified in A.1 and the total level in the frequency range corresponding to the third octave bands from 100 Hz to 8 kHz is $(-28.7 - F_s)$ dBPa. For calculating SLR and response characteristics the reference point is MRP. The sound pressure level at the MRP shall be calculated (by definition) by adding 24 dB to the sound pressure level at the HFRP. + +##### **Method 2:** at the MRP + +The signal generated by the Artificial Mouth is equalized at the MRP under free field conditions to obtain the spectrum specified in A.1, at a level of $-4.7$ dBPa in the frequency range corresponding to the third octave bands from 100 Hz to 8 kHz. The spectrum at the MRP is then recorded and the level is adjusted to obtain $(-28.7 - F_s)$ dBPa at the HFRP. The spectrum recorded at the MRP is used as a reference for calculating SLR and response characteristics. + +## **A.2 Testing of transmission requirements** + +### **A.2.1 Sensitivity-frequency response** + +#### **A.2.1.1 Sending** + +The measurement shall be carried out for each microphone, which shall be mounted as specified in A.1.1.1. In case of microphones intended to pick up the sounds coming from multiple directions (either omni-directional or steered devices), the extreme incidence angles on the horizontal plane (as stated by the manufacturer) shall be tested, as well as an intermediate incidence direction. If stated by the manufacturer, the nominal direction shall be used as intermediate. The testing angles actually used for measurements shall be stated in the test report. + +The noise signal shall be generated by the Artificial Mouth at the level specified in A.1.2.1. + +The spectrum of the output signal shall be measured at the output interface of the reference codec. + +The sending sensitivity shall be calculated as follows, according to the calibration method used (see A.1.2.1): + +##### **Method 1** + +The sending sensitivity is given by the difference between the electric spectrum and the acoustic spectrum at the MRP: + +$$S_{mj} = 20 \log V_s - 20 \log P_m, \quad P_m = P_{HFRP} + 24 \text{ dB}$$ + +where: + +$20 \log V_s$ : Electric spectrum; + +$20 \log P_m$ : Acoustic spectrum at MRP. + +##### **Method 2** + +The sending sensitivity $S_{mj}$ is given by the following relationship: + +$$S_{mj} = 20 \log V_s - 20 \log P_m - \text{Corr} - 24 \text{ dB}$$ + +where: + +$20 \log V_s$ : Electric spectrum; + +$20 \log P_m$ : Acoustic spectrum recorded at MRP; + +$\text{Corr}$ : Correction factor ( $20 \log P_{MRP}/P_{HFRP}$ ) of the Artificial Mouth. + +#### A.2.1.2 Receiving + +Each loudspeaker shall be placed in turn in the test room as specified in A.1.1.2. + +The noise signal generator shall be connected to the input of the reference codec. + +The sensitivity at each 1/3 octave band shall be calculated by subtracting the spectrum of the electric signal from the acoustic spectrum at the measurement point. + +The measurement shall be repeated at the minimum and maximum position of the (manual) volume control, changing the input level accordingly. In case of devices not provided with manual volume control, the measurement shall be repeated for excitation levels of $-30 \text{ dBm}_0$ and $-15 \text{ dBm}_0$ . + +### A.2.2 Loudness rating + +#### A.2.2.1 Sending Loudness Rating + +The sending sensitivity shall be measured for each of the fourteen 1/3rd octave bands given in Table 1/P.79 [10], bands 4 to 17 (200 Hz-4 kHz). The sensitivity is expressed in terms of $\text{dB(V/Pa)}$ and the SLR shall be calculated according to Formula A-23b/P.79, over bands 4 to 17 and using the sending weighting factors from Table A.2/P.79, adjusted by subtracting 0.3 dB from each value, using the sending sensitivity response of A.2.1.1. + +The measurement shall be carried out for each microphone, which shall be mounted as specified in A.1.1.1. In case of microphones intended to pick up the sounds coming from multiple directions (either omni-directional or steered devices), the extreme incidence angles on the horizontal plane (as stated by the manufacturer) shall be tested, as well as an intermediate incidence direction. If stated by the manufacturer, the nominal direction shall be used as intermediate. The testing angles actually used for measurements shall be stated in the test report. + +#### A.2.2.2 Receiving Loudness Rating + +The receiving sensitivity shall be measured for each of the fourteen 1/3rd octave bands given in Table 1/P.79 [10], bands 4 to 17 (200 Hz-4 kHz). The sensitivity is expressed in terms of $\text{dB(Pa/V)}$ and the RLR shall be calculated according to Formula A-23c/P.79, over bands 4 to 17 and using the receiving weighting factors from Table A.2/P.79, adjusted by subtracting 0.3 dB from each value, using the receiving sensitivity response of A.2.1.2. + +The receiving sensitivity shall not be corrected by the $L_e$ factor. The calculated RLR shall be corrected by subtracting 14 dB, according to ITU-T Rec. P.340 [5]. + +### A.2.3 Terminal Coupling Loss + +The measurements shall be carried out by coupling, as specified in A.1.1.3, the most efficient loudspeaker with the most efficient microphone. + +#### A.2.3.1 Weighted Terminal Coupling Loss + +The test signal shall be a pink noise, with a level of $(-15 - F_{\text{tcl}}) \text{ dBm}_0$ . + +The attenuation from digital input to digital output shall be measured at the 1/3rd octave frequencies given by the R10 series of preferred numbers in ISO 3 [11] for frequencies from 100 Hz to 8000 Hz. + +The $\text{TCL}_w$ shall be calculated according to the method in B.4/G.122 [12] (trapezoidal rule), on the frequency band from 100 Hz to 8 kHz. The value calculated according to ITU-T Rec. G.122 [12] ( $\text{TCL}_{w\text{calculated}}$ ) shall be corrected by adding the correction factor $F_{\text{tcl}}$ and by considering the total sound power produced by all the loudspeakers: + +$$\text{TCL}_w = \text{TCL}_{w\text{calculated}} + F_{\text{tcl}} - 10 \log \frac{n-1}{2}$$ + +where $n$ is the total number of loudspeakers equipping the terminal under test. + +NOTE – The formula above assumes that only one microphone is active at each time and that the sound pressure levels of signals picked up from the more distant loudspeakers is 3 dB less than the signal level generated by the better coupled loudspeaker. It is also assumed that a power summation law applies to the signals picked up by the different loudspeakers. This is not absolutely true in principle, but some uncorrelation of the sound signals generated by the different loudspeakers can be assumed to result from the different transmission path lengths and from the reverberation effects occurring in actual use. + +#### A.2.3.2 Stability loss + +The test signal shall be sinusoidal, with a level of $(-15 - F_{\text{tel}})$ dBm0. + +The attenuation from digital input to digital output shall be measured at 1/24th octave intervals for frequencies from 100 Hz to 8 kHz. The actually measured values shall be corrected by adding $F_{\text{tel}}$ and by subtracting $10 \log [(n+1)/2]$ (as for the TCLw measurement). + +# APPENDIX I + +## Practical *in-situ* audio alignment procedure for multiple-users hands-free GATs + +## I.1 Introduction + +The hands-free function for a group of users is normally provided with audio arrangements which widely interact with the installation environment and a guide is provided here for *in-situ* aligning the electroacoustic sensitivities and checking the correct transduction across the whole frequency band. + +## I.2 Audio alignment and sensitivity-frequency characteristics + +Audio alignment is an operation carried out when the terminal is installed, and any time the acoustic configuration of GATs is changed. It is intended to guarantee the proper setting of sending and receiving level adjustments, in order to prevent the channel overloading while still guaranteeing adequate signal to noise and a comfortable receiving audio level. + +The following equipment should be used for performing the audio alignment of GAT terminals: + +- Artificial Mouth (complying with ITU-T Rec. P.51 [9]); +- signal generator (digital or analogue); +- sound level meter (complying with IEC 60651 [13]); +- electric level meter. + +The acoustic test signal generated by the Artificial Mouth should consist of speech-shaped noise (ITU-T Rec. P.50 [14]), at a level of $-4.7$ dBPa at the MRP. + +The signal generator is used for performing the receiving alignment and can either generate an encoded digital path at the digital interface of the GAT, or an analogue signal to be applied at a suitable analogue test point. In the latter case, a reference codec could be necessary for digital GATs if no specific interface for alignment purposes is available within the terminal. The signal for receiving alignment should be a speech-shaped noise (ITU-T Rec. P.50 [14]) at a level of $-20$ dBm0. If necessary, the noise can be modulated ON-OFF. In this case, the signal level is defined as the level generated during the ON periods of the signal. + +NOTE – Terminal manufacturers are encouraged to implement an analogue 0 dBr interface for alignment purposes in their equipment. + +If no specific alignment interface is provided, the electric level meter for sending alignment should either directly process the digital pattern generated by the terminal, or be used in conjunction with a Reference Coder. + +### **I.2.1 Sending** + +#### **I.2.1.1 Sensitivity adjustment(s)** + +The Artificial Mouth should be placed over the border of the conference table, as shown in Figure A.1, in correspondence of each conferee position. The conference microphones should be positioned on the conference table (or in the conference room) as in normal use. + +If a sensitivity adjustment is provided for each microphone, it should be regulated in turn in order to achieve an output level of $-20 \text{ dBm}0$ when the Mouth is placed in correspondence with the associated microphone. Otherwise, if only one sensitivity control is provided, it should be regulated in order that the average reading, obtained by placing the Mouth in correspondence of each conferee position in turn, is equal to $-20 \text{ dBm}0$ . + +#### **I.2.1.2 *In-situ* frequency response** + +The *in-situ* measurement of the sending frequency response is defined as the difference between the octave spectra of the output signal and of the acoustic excitation signal at the MRP. Considering that the measurements are *in-situ*, the octave band analysis shall be restricted to the bandwidth from 125 Hz to 4 kHz (centre frequencies). It is recommended that the sum of the absolute differences between the measured values and their average is less than 8 dB. + +### **I.2.2 Receiving** + +#### **I.2.2.1 Sensitivity adjustment** + +The speech-shaped noise is fed to the input of the terminal at a level of $-20 \text{ dBm}0$ . With the manual volume control (if any) at its maximum position, the receiving gain should be adjusted in order to achieve a sound pressure level of at least 54 dBA at each conferee position. The measurements should be made by placing the measurement microphone at point C (Figure A.1) corresponding to each conferee position, and in the absence of the Artificial Mouth. + +NOTE – Practical experience has shown that higher listening levels, up to 65 dBA, may be desirable. The maximum settable gain being constrained by the terminal stability limit. + +#### **I.2.2.2 *In-situ* frequency response** + +The *in-situ* measurement of the receiving frequency response is defined as the difference between the octave spectra of the output signal, and of the input excitation signal at the terminal interface. Considering that measurements are *in-situ*, the octave band analysis shall be restricted to the bandwidth from 125 Hz to 4 kHz (centre frequencies). It is recommended that the sum of the absolute differences between the measured values and their average is less than 12 dB. + + + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|--------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series B | Means of expression: definitions, symbols, classification | +| Series C | General telecommunication statistics | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Construction, installation and protection of cables and other elements of outside plant | +| Series M | TMN and network maintenance: international transmission systems, telephone circuits, telegraphy, facsimile and leased circuits | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks and open system communications | +| Series Y | Global information infrastructure and Internet protocol aspects | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.310-200906-I_PDF-E/raw.md b/marked/P/T-REC-P.310-200906-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..5884a77e4847c2ddee033eee3deb83ee1825e4ee --- /dev/null +++ b/marked/P/T-REC-P.310-200906-I_PDF-E/raw.md @@ -0,0 +1,1177 @@ + + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.310** + +(06/2009) + +SERIES P: TERMINALS AND SUBJECTIVE AND +OBJECTIVE ASSESSMENT METHODS + +Voice terminal characteristics + +--- + +**Transmission characteristics for narrow-band +digital handset and headset telephones** + +Recommendation ITU-T P.310 + +# ITU-T P-SERIES RECOMMENDATIONS **TERMINALS AND SUBJECTIVE AND OBJECTIVE ASSESSMENT METHODS** + +| | | | +|-----------------------------------------------------------------------------------------------|---------------|--------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Voice terminal characteristics | Series | P.30 | +| | | P.300 | +| Reference systems | Series | P.40 | +| Objective measuring apparatus | Series | P.50 | +| | | P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of speech quality | Series | P.80 | +| | | P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | +| Transmission performance and QoS aspects of IP end-points | Series | P.1000 | +| Communications involving vehicles | Series | P.1100 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +## **Recommendation ITU-T P.310** + +# **Transmission characteristics for narrow-band digital handset and headset telephones** + +## **Summary** + +Recommendation ITU-T P.310 provides audio performance requirements and associated testing methods for narrow-band (300-3400 Hz) digital handset and headset telephones. + +Requirements and test methods are specified for the major audio transmission parameters including sending and receiving frequency response, loudness rating, noise, distortion, out-of-band signals, linearity, sidetone, echo and delay. + +This Recommendation is only applicable to digital telephones using encoding conforming to Recommendations ITU-T G.711 (64 kbit/s, PCM) and ITU-T G.726 (32 kbit/s, ADPCM). IP terminals and wireless headsets are not covered in this recommendation. + +## **Source** + +Recommendation ITU-T P.310 was approved on 22 June 2009 by ITU-T Study Group 12 (2009-2012) under Recommendation ITU-T A.8 procedures. + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database at . + +© ITU 2010 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## CONTENTS + +| | Page | +|----------------------------------------------------------------|------| +| 1 Scope ..... | 1 | +| 2 References..... | 1 | +| 3 Definitions and abbreviations ..... | 2 | +| 3.1 Definitions ..... | 2 | +| 3.2 Abbreviations ..... | 3 | +| 4 Test configuration..... | 4 | +| 4.1 Measurement approaches for testing digital telephones..... | 4 | +| 4.2 Test equipment ..... | 5 | +| 4.3 Measurement set-up..... | 6 | +| 4.4 Test environment ..... | 7 | +| 4.5 Test signal and level ..... | 7 | +| 4.6 Accuracy of test equipment ..... | 7 | +| 5 Handset technical requirements ..... | 8 | +| 5.1 Sending characteristics ..... | 8 | +| 5.2 Receiving characteristics ..... | 14 | +| 5.3 Sidetone characteristics ..... | 20 | +| 5.4 Echo path loss characteristics ..... | 22 | +| 5.5 Delay..... | 25 | +| 6 Headset technical requirements ..... | 25 | +| 6.1 Sending characteristics ..... | 25 | +| 6.2 Receiving characteristics ..... | 26 | +| 6.3 Echo path loss characteristics ..... | 26 | +| Annex A – Distortion allowances ..... | 28 | +| Bibliography..... | 29 | + + + +## Recommendation ITU-T P.310 + +## Transmission characteristics for narrow-band digital handset and headset telephones + +# 1 Scope + +This Recommendation deals with sending and receiving frequency response, loudness rating, noise, distortion, out-of-band signals, linearity, sidetone, echo and delay of narrow-band (300-3400 Hz) digital handset and headset telephones using "Waveform" encoding according to [ITU-T G.711] (PCM at both 64 and 56 kbit/s) and [ITU-T G.726] (ADPCM, 32 kbit/s). IP terminals and wireless headsets are not covered in this Recommendation. + +The use of digital telephones using [ITU-T G.728] (LD-CELP, 16 kbit/s) and mobile/cordless telephones are under study. + +Requirements applicable to low acoustic impedance transducers and digital telephone sets using non-linear techniques are under study. + +The requirements listed in this Recommendation should also be used as the basis of requirements for other "waveform" encoding schemes. + +The values given in this Recommendation should be used for developing specifications which will include assigning tolerances, etc. + +# 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- [ITU-T G.122] Recommendation ITU-T G.122 (1993), *Influence of national systems on stability and talker echo in international connections.* +- [ITU-T G.131] Recommendation ITU-T G.131 (2003), *Talker echo and its control.* +- [ITU-T G.711] Recommendation ITU-T G.711 (1988), *Pulse code modulation (PCM) of voice frequencies.* +- [ITU-T G.712] Recommendation ITU-T G.712 (2001), *Transmission performance characteristics of pulse code modulation channels.* +- [ITU-T G.726] Recommendation ITU-T G.726 (1990), *40, 32, 24, 16 kbit/s Adaptive Differential Pulse Code Modulation (ADPCM).* +- [ITU-T G.728] Recommendation ITU-T G.728 (1992), *Coding of speech at 16 kbit/s using low-delay code excited linear prediction.* +- [ITU-T I.412] Recommendation ITU-T I.412 (1988), *ISDN user-network interfaces – Interface structures and access capabilities.* +- [ITU-T I.430] Recommendation ITU-T I.430 (1995), *Basic user-network interface – Layer 1 specification.* + +- [ITU-T O.41] Recommendation ITU-T O.41 (1994), *Psophometer for use on telephone-type circuits.* +- [ITU-T O.131] Recommendation ITU-T O.131 (1988), *Quantizing distortion measuring equipment using a pseudo-random noise test signal.* +- [ITU-T O.133] Recommendation ITU-T O.133 (1993), *Equipment for measuring the performance of PCM encoders and decoders.* +- [ITU-T P.10] Recommendation ITU-T P.10/G.100 (2006), *Vocabulary for performance and quality of service.* +- [ITU-T P.50] Recommendation ITU-T P.50 (1999), *Artificial voices.* +- [ITU-T P.51] Recommendation ITU-T P.51 (1996), *Artificial mouth.* +- [ITU-T P.57] Recommendation ITU-T P.57 (2009), *Artificial ears.* +- [ITU-T P.58] Recommendation ITU-T P.58 (1996), *Head and torso simulator for telephonometry.* +- [ITU-T P.64] Recommendation ITU-T P.64 (2007), *Determination of sensitivity/frequency characteristics of local telephone systems.* +- [ITU-T P.79] Recommendation ITU-T P.79 (2007), *Calculation of loudness ratings for telephone sets.* +- [ITU-T P.380] Recommendation ITU-T P.380 (2003), *Electro-acoustic measurements on headsets.* +- [ITU-T P.501] Recommendation ITU-T P.501 (2007), *Test signals for use in telephonometry.* +- [ITU-T P.502] Recommendation ITU-T P.502 (2000), *Objective test methods for speech communication systems using complex test signals.* +- [ITU-T P.581] Recommendation ITU-T P.581 (2000), *Use of head and torso simulator (HATS) for hands-free terminal testing* +- [ISO 3] ISO 3:1973, *Preferred numbers – Series of preferred numbers.* +- [ISO 1996-1] ISO 1996-1:2003, *Acoustics – Description, measurement and assessment of environmental noise – Part 1: Basic quantities and assessment procedures.* + +# 3 Definitions and abbreviations + +## 3.1 Definitions + +This Recommendation defines the following terms: + +**3.1.1 0 dB reference point:** To preserve compatibility with existing codecs already in use in local digital switches, which are defined as a 0 dBr point, the codec (A- or $\mu$ -law) should be defined as follows: + +- *D/A converter* – A digital test sequence (DTS) representing the PCM equivalent of an analogue sinusoidal signal whose r.m.s. value is 3.14 dB (A-law) or 3.17 dB ( $\mu$ -law) below the maximum full-loaded capacity of the codec will generate 0 dBm across a 600-ohm load. Here DTS is defined as a periodic sequence of character signals as given in [ITU-T G.711]. +- *A/D converter* – A 0 dBm signal generated from a 600-ohm source will give the digital test sequence representing the PCM equivalent of an analogue sinusoidal signal whose r.m.s. value is 3.14 dB (A-law) or 3.17 dB ( $\mu$ -law) below the maximum full load capacity of the codec. + +**3.1.2 acoustic reference level (ARL):** Defined as the acoustic level at MRP which results in a –10 dBm0 output at the digital interface. + +**3.1.3 digital telephone set (DT):** A digital telephone set is one in which the A/D and D/A converters are built in and the connection to the network is via a digital bitstream. + +**3.1.4 HATS position:** The HATS (head and torso simulator) position (see Annexes D and E of [ITU-T P.64]) is the correct handset position for measuring sensitivity and frequency response characteristics. The HATS position has been shown to be essentially identical to the LRGP (loudness rating guard-ring position), except for the mouth simulator direction, which has been corrected with a 19° downwards rotation to more closely match real talkers. For handsets with omnidirectional microphones, measurements on the two heads may differ slightly, typically less than 1 dB. For handsets with directional or noise-cancelling microphones, the differences will be larger, and the HATS position will give the more realistic results. + +## **3.2 Abbreviations** + +Relevant abbreviations in [ITU-T P.10] will apply. + +This Recommendation uses the following abbreviations: + +| | | +|-------------------|-----------------------------------------| +| A/D | Analogue-to-Digital | +| ARL | Acoustic Reference Level | +| D/A | Digital-to-Analogue | +| DRP | earDrum Reference Point | +| DTS | Digital Test Sequence | +| ERP | Ear Reference Point | +| ERUP | Estimated Real Use Position | +| HATS | Head And Torso Simulator | +| ISDN | Integrated Services Digital Network | +| LRGP | Loudness Rating Guard-ring Position | +| LSTR | Listener SideTone Rating | +| MRP | Mouth Reference Point | +| PABX | Private Automatic Branch Exchange | +| PCM | Pulse Code Modulation | +| RLR | Receiving Loudness Rating | +| RTP | Recommended Test Position | +| RWP | Recommended Wearing Position | +| $S_{je}$ | Receiving Sensitivity (Artificial Ear) | +| $S_{JE}$ | Receiving Sensitivity (Real Ear) | +| SLR | Sending Loudness Rating | +| $S_{mj}$ | Sending Sensitivity (Artificial Mouth) | +| $S_{MJ}$ | Sending Sensitivity (Real Mouth) | +| $S_{si}$ (diff) | Sending Sensitivities for diffuse sound | +| $S_{si}$ (direct) | Sending Sensitivities for direct sound | + +| | | +|--------------|---------------------------------------------------------------------------| +| STMR | SideTone Masking Rating | +| TCL | Terminal Coupling Loss | +| TCLw | Weighted Terminal Coupling Loss | +| $\Delta S_m$ | Difference between the sending sensitivities for diffuse and direct sound | + +# 4 Test configuration + +## 4.1 Measurement approaches for testing digital telephones + +In general, there are two approaches for evaluating the transmission performance of a digital telephone, the direct approach and the codec approach. The direct approach is, in principle, the most accurate although the use of the codec approach may sometimes be advantageous. + +ITU-T recommends both methods to evaluate the voice transmission performance of a digital handset or headset telephone using "Waveform" encoding conforming to [ITU-T G.711] (PCM at 64 kbit/s and 56 kbit/s) and [ITU-T G.726] (ADPCM, 32 kbit/s). + +### 4.1.1 Direct digital processing approach + +In this approach, shown in Figure 4-1, the companded digital input/output bitstream of the telephone set is operated upon directly. The advantage is that most of the test signals, if sampled at 8 kHz, can be generated and analysed without the need for resampling and A/D or D/A conversion. + +![Figure 4-1: Digital telephone test arrangement (direct digital processing approach).](c2fc2621e8206d24427b56bcb2398fc0_img.jpg) + +The diagram shows a 'Digital set' enclosed in a dashed box. Inside, an input signal $S_M$ enters a microphone symbol, followed by an A/D converter. An output signal $S_E$ exits from a D/A converter. Both converters are connected to a 'Multi-plexer'. The 'Multi-plexer' is connected to an 'Interface' block outside the dashed box. The 'Interface' block receives a 'Control' signal from above and a 'PCM' signal from below. It is also connected to a 'Direct digital processing unit'. + +Figure 4-1: Digital telephone test arrangement (direct digital processing approach). + +Figure 4-1 – Digital telephone test arrangement +(direct digital processing approach) + +### 4.1.2 Codec approach + +In this approach, shown in Figure 4-2, a codec is used to convert the companded digital input/output bitstream of the telephone set to the equivalent analogue values, so that existing test procedures and equipment can be used. This codec should be a high-quality codec whose characteristics are as close as possible to ideal (see clause 4.2.2). + +![Figure 4-2: Digital telephone test arrangement (codec approach).](b10763be31553f31cbb795653b731a43_img.jpg) + +The diagram shows a 'Digital set' enclosed in a dashed box. Inside, an input signal $S_M$ enters a microphone symbol, followed by an A/D converter. An output signal $S_E$ exits from a D/A converter. Both converters are connected to a 'Multi-plexer'. The 'Multi-plexer' is connected to an 'interface' block outside the dashed box. The 'interface' block receives a 'Control' signal from above. It is connected to a '"Reference codec" test bed' (another dashed box). Inside this test bed, the signal from the interface goes to a D/A converter, which is followed by a $600\ \Omega$ resistor. The signal from the D/A converter also goes to an A/D converter, which is preceded by a $600\ \Omega$ resistor. Both A/D converters are connected to 'Analogue instruments'. + +Figure 4-2: Digital telephone test arrangement (codec approach). + +Figure 4-2 – Digital telephone test arrangement +(codec approach) + +NOTE – The digital telephone test equipment will, in general, be connected to the telephone under test through an interface. Such an interface should be able to provide all the signalling and supervisory sequences necessary for the telephone set to be working in all test modes. The interface must be capable of converting the digital output stream from the tested set (which may be in various formats, depending on the specific type of telephone set, e.g., conforming to [ITU-T I.412] for ISDN sets), to a form compatible with the test equipment. Interfaces can be applied for sending and receiving separately, taking into account telephone sets which are connected to various types of exchanges. + +## 4.2 Test equipment + +### 4.2.1 Electro-acoustic equipment + +The preferred acoustical access to terminals is the most realistic simulation of the "average" subscriber. This is made by using HATS (head and torso simulator) with appropriate ear simulator and appropriate means to fix handset or headset in a realistic and reproducible way. HATS, including the artificial mouth and the artificial ear, shall comply with [ITU-T P.58]. The recommended ear simulators are type 3.3 and type 3.4 artificial ear, described in [ITU-T P.57]. + +The horizontal positioning of the HATS reference plane shall be guaranteed with $\pm 2^\circ$ . The exact calibration and equalization of HATS can be found in [ITU-T P.581]. + +### 4.2.2 Codecs + +#### 4.2.2.1 Ideal codec + +The ideal codec consists of an independent encoder and decoder whose characteristics are hypothetical and comply with [ITU-T G.711]. The ideal encoder is a perfect analogue-to-digital converter preceded by an ideal low-pass filter (assumed to have no attenuation/frequency distortion and no envelope-delay distortion), and may be simulated by a digital processor. The ideal decoder is a perfect digital-to-analogue converter followed by an ideal low-pass filter (assumed to have no attenuation/frequency distortion and no envelope-delay distortion), and which may be simulated by a digital processor1. + +For the measurement of the sending side of a telephone set, the output digital signal is converted by the decoder to an analogue signal. The electrical characteristics of this output signal are measured using conventional analogue instruments. For the measurement of the receiving side of a telephone set, the analogue output from a signal source is converted to a digital signal by the ideal encoder and fed to the receiving input of the digital telephone set. + +NOTE – For codecs conforming to [ITU-T G.726], a G.711/G.726 conversion will be applied. + +#### 4.2.2.2 Reference codec + +A practical implementation of an ideal codec may be called a reference codec (see [ITU-T O.133]). + +For the reference codec, characteristics such as attenuation/frequency distortion, idle channel noise, quantizing distortion, etc. should be better than the requirements specified in [ITU-T G.712], so as not to mask the corresponding parameters of the set under test. A suitable reference codec may be realized by using: + +- 1) at least 14-bit linear A/D and D/A converters of high quality, and transcoding the output signal to the A- or $\mu$ -law PCM format; +- 2) a filter response that meets the requirements of Figure 4-3. + +--- + +1 This characteristic can be realized, for example, using over-sampling techniques and digital filters. + +![Figure 4-3: Attenuation/frequency distortion of the sending or receiving sides of the reference codec. The graph shows Loss (dB) on the y-axis (ranging from -0.25 to 40) versus Frequency (Hz) on the x-axis (logarithmic scale from 80 to 4000). The loss is 0.0 dB from 80 Hz to 100 Hz, then drops to -0.25 dB at 100 Hz. It remains at -0.25 dB until 3000 Hz, where it steps up to 0.25 dB. At 3400 Hz, it steps up to 0.9 dB. At 3600 Hz, it drops back to -0.25 dB. At 4000 Hz, it steps up to 40 dB. Shaded regions indicate the allowed loss limits.](36ac3e730a00d3f42d3400f5709f641a_img.jpg) + +Figure 4-3: Attenuation/frequency distortion of the sending or receiving sides of the reference codec. The graph shows Loss (dB) on the y-axis (ranging from -0.25 to 40) versus Frequency (Hz) on the x-axis (logarithmic scale from 80 to 4000). The loss is 0.0 dB from 80 Hz to 100 Hz, then drops to -0.25 dB at 100 Hz. It remains at -0.25 dB until 3000 Hz, where it steps up to 0.25 dB. At 3400 Hz, it steps up to 0.9 dB. At 3600 Hz, it drops back to -0.25 dB. At 4000 Hz, it steps up to 40 dB. Shaded regions indicate the allowed loss limits. + +**Figure 4-3 – Attenuation/frequency distortion of the sending or receiving sides of the reference codec** + +##### 4.2.2.2.1 Analogue interface + +The output and input impedances return loss and longitudinal conversion losses of the analogue interface of the reference codec should be in accordance with [ITU-T O.133]. + +##### 4.2.2.2.2 Digital interface + +The fundamental requirements for the reference codec digital interface are given in the appropriate ITU-T Recommendations (e.g., [ITU-T I.430] for ISDN telephone sets). + +## 4.3 Measurement set-up + +All measurements are performed using a HATS. The type of ear simulator and application force shall be indicated in a test report. For handsets, if not stated otherwise, 8N application force shall be used to apply the handset against the artificial ear. + +Proper positioning of handsets on HATS can be found in Annexes D and E of [ITU-T P.64]. + +Suggestions for positioning headsets on HATS are given in [ITU-T P.380]. The manufacturer should provide the recommended wearing position (RWP) in the user's guide, which describes a precise way the device should be placed on the user's head. From RWP, the recommended test position (RTP) information can be derived, as close as possible to RWP. Also the RTP description can be provided by the manufacturer which includes the information how the receiving part of the headset should be placed against or inside the ear simulator, and further describes the positioning and orientation of the microphone. If neither RTP nor RWP are provided by the manufacturer, the headset shall be tested in the estimated real use position (ERUP). The test lab shall define an ERUP that closely approximates real use. Natural headband pressure, or other positioning techniques normally used by a real user, shall be used. Because the headset position may affect test results, the test shall be repeated at least 5 times by completely repositioning the headset, following the rules described in [ITU-T P.380]. + +Use of the codec test approach means that test procedures for digital telephone sets in general follow those for analogue sets (see [ITU-T P.64]). + +The reference codec should meet the requirements of clause 4.2.2.2. An important difference, however, concerns the test circuits themselves (see Figures 5-2 and 5-4 to 5-7). + +The set is connected to the interface and is placed in the active call state. + +Handsets or headsets fitted with a volume control on receiving shall be set as close as possible to the nominal RLR and any residential difference from the nominal value will be corrected by the normalization process. + +## **4.4 Test environment** + +The test laboratory shall ensure that all measurements performed in the test room give identical results to those obtained in a free-field environment (e.g., an anechoic room). + +The environmental conditions for testing handset and headset terminals are specified as follows: + +- The test room shall be practically free-field down to a lowest frequency of 275 Hz, the handset or headset coupled with the HATS shall lie totally within this free-field volume. This is met if deviations from the ideal free-field conditions are less than $\pm 1$ dB. +- The ambient noise level shall be less than $-64$ dBPa(A) for all the tests, except for tests requesting the generation of ambient noise, e.g., D-Factor measurement. +- The test environment shall be free of mechanical disturbances. + +## **4.5 Test signal and level** + +In general, the test signals as described in this Recommendation should be applied. The use of the proposed test signals requires a linear and time invariant operation of the equipment under test. This cannot be ensured in all cases. For devices where the transmission properties are level and signal-dependent, alternative test signals should be chosen. In this case a more speech-like test signal such as described in [ITU-T P.50] and [ITU-T P.501] should be applied. Test house and manufacturer should ensure that the appropriate type of test signal is chosen. + +Unless stated otherwise, the test signal shall be set to $-4.7$ dBPa at MRP for the sending direction and $-16$ dBm0 for the receiving direction. The spectrum of acoustic signal produced by the artificial mouth is calibrated under free-field conditions at MRP. + +For a correct activation of the system, an artificial voice according to [ITU-T P.50] or a speech like test signal as described in [ITU-T P.501] shall be used for activation. The level of this activation signal shall be $-4.7$ dBPa at MRP. + +The use of test signals and levels should be stated in the test report. + +NOTE 1 – The use of sine signals may not be appropriate when speech processing and coding systems are implemented in the terminal. For example, in distortion measurement, if a sine wave is not usable, an alternative test signal could be a band-limited noise signal centred on the test frequencies. + +NOTE 2 – It should be ensured that the test signal is treated by speech processing algorithms as a speech-like signal, and not a noise-like signal. + +NOTE 3 – When measuring digital telephone sets, it is advisable to avoid measuring at sub-multiples of the sampling frequency. There is a tolerance on the frequencies of $\pm 2\%$ which may be used to avoid this problem, except for 4 kHz where only the $-2\%$ tolerance may be used. + +## **4.6 Accuracy of test equipment** + +Unless specified otherwise, the accuracy of the measurements made by the test equipment shall be better than those given in Table 1. + +**Table 1 – Minimum measurement accuracy** + +| Quantity | Accuracy | +|------------------------------|------------------------------------------------------------------------------------| +| Electrical signal power | ±0.2 dB for levels ≥ -50 dBm
±0.4 dB for levels < -50 dBm | +| Sound pressure | ±0.7 dB | +| Time | ±5% | +| Frequency | ±0.2% | +| Sound pressure level at MRP | ±3 dB for 100 Hz to 200 Hz
±1 dB for 200 Hz to 4 Hz
±3 dB for 4 kHz to 8 kHz | +| Electrical excitation levels | ±0.4 dB (see Note 1) | +| Frequency generation | ±0.2% (see Note 2) | + +NOTE 1 – Across the whole frequency range. +NOTE 2 – When measuring sampled systems, it is advisable to avoid measuring at sub-multiples frequency. There is a tolerance of ±2% on the generated frequencies, which may be used to avoid this problem, except for 4 kHz where only the -2% tolerance may be used. + +# 5 Handset technical requirements + +## 5.1 Sending characteristics + +### 5.1.1 Sending frequency response + +In view of the following consideration: + +- Unlike other standards, this Recommendation does not use ERP as the reference point for receiving, but uses the diffuse-field instead. With the concept of diffuse-field based receive measurements, a rising slope for the overall frequency response is achieved by a flat target frequency response in sending and a diffuse-field based receiving frequency response; the sensitivity mask of Table 2 is recommended. + +**Table 2 – Sending frequency response** + +| Frequency (Hz) | Upper limit (dB) | Lower limit (dB) | +|----------------|------------------|------------------| +| 100 | -18 | | +| 300 | 5 | -5 | +| 3400 | 5 | -5 | +| 4000 | 5 | | + +NOTE – The limits for intermediate frequencies lie on a straight line drawn between the given values on a linear (dB) – logarithmic (Hz) scale. + +![Figure 5-1: Sending mask graph showing dB vs Frequency (Hz).](71ab4df17511d75261da8d462d643b1a_img.jpg) + +| Frequency (Hz) | Upper Limit (dB) | Lower Limit (dB) | +|----------------|------------------|------------------| +| 100 | -19 | - | +| 200 | 5 | -19 | +| 300 | 5 | -5 | +| 400 | 5 | 0 | +| 3400 | 5 | 0 | +| 4000 | 5 | -5 | +| 4000+ | - | -20 | + +Figure 5-1: Sending mask graph showing dB vs Frequency (Hz). + +Figure 5-1 – Sending mask + +#### 5.1.1.1 Measurement method + +The sending frequency characteristic is measured according to [ITU-T P.64] using the measurement set-up shown in Figure 5-1. + +A sinewave or an artificial speech (e.g., as specified in [ITU-T P.50] and [ITU-T P.501]) generator and a spectrum analyser should be used. Additional test methods may be found in [ITU-T P.502]. The test signal level shall be $-4.7$ dBPa at MRP. The test signal level is averaged over the complete test signal sequence. The sensitivity is expressed in terms of dBV/Pa. + +![Figure 5-2: Block diagram of the measurement setup for sending frequency characteristic.](9b6b5924b48bf2fd5f347f88f06f45b3_img.jpg) + +``` + +graph LR + SG[Signal generator] --> AM[Artificial mouth] + subgraph QR[Quiet room] + AM --> DT[Digital telephone] + end + DT --> IF[Interface] + C[Control] --> IF + IF --> TB[Test bed] + subgraph TB[Test bed] + DA[D/A] + end + DA --> R[600 Ω] + DA --> MA[Measuring amplifier] + MA --> LRD[Level recorder or distortion meter] + +``` + +Figure 5-2: Block diagram of the measurement setup for sending frequency characteristic. + +Figure 5-2 – Measurements of sending frequency characteristic, HATS position + +### 5.1.2 Sending loudness rating (SLR) + +– The nominal value of SLR shall be $\pm 8$ dB, with a tolerance of $\pm 3$ dB. + +#### 5.1.2.1 Measurement method + +This should be calculated from the sensitivity/frequency characteristic determined in clause 5.1.1.1 by means of [ITU-T P.79]. + +NOTE – Other methods for calculating loudness ratings used by some Administrations for their own internal planning purposes can be found in [b-Teleph]. + +### 5.1.3 Sending noise + +In view of the following considerations: + +- compatibility with coder and decoder requirements according to [ITU-T G.712]; +- certain additions of noise must be allowed for in the electrical and acoustical parts (see clause 4.6); +- compatibility with existing analogue telephones, + +the following limits are recommended. + +- sending noise level: maximum –64 dBm0p; +- no peaks in any 1/3-octave band, with the level of 10 dB higher than the average noise spectrum in the frequency domain, shall occur. + +NOTE – The noise levels are related to the long-term objective for SLR and RLR. + +#### 5.1.3.1 Measurement method + +The sending noise shall be measured in a quiet environment as defined in clause 4.4. + +After a correct activation, the noise level at the digital output is measured in the frequency range from 100 Hz to 4000 Hz with apparatus including psophometric weighting described in [ITU-T O.41]. + +NOTE – The ambient noise criterion will be met if the ambient noise does not exceed NR20 [ISO 1996-1]. + +### 5.1.4 Sending distortion + +In view of the following considerations: + +- compatibility with coder and decoder requirements according to [ITU-T G.712]; +- certain additions of distortion must be allowed for in the electrical and acoustical parts (see clause 4.6); +- compatibility with existing analogue telephones, + +two different sets of values are recommended relating to two different measuring methods (the "Sinewave" method and the "Noise" method; see [ITU-T G.712]). Either is acceptable. + +NOTE – ETSI have found it desirable to use both the noise method and the sinewave method for the following reasons: + +- The "Sinewave" method is effective for the measurement of the coding distortion and overload distortion. +- The "Noise" method, being more speech-like and of lower frequency content, is more likely to indicate imperfections, including inter-modulation distortion, in the transducers as well as the coding. + +#### 5.1.4.1 Sending distortion using "Noise" method + +The "Noise" method is used routinely for A-law codecs. + +The ratio of signal-to-total distortion (harmonic and quantizing) power of the digitally encoded signal output by the terminal equipment measured with the proper noise weighting (see [ITU-T O.41]) shall be above the limits given in Tables 3 and 4 for [ITU-T G.711] (64 kbit/s) and [ITU-T G.726] (32 kbit/s), respectively. + +Limits for intermediate levels are found by drawing straight lines between the breaking points in the table on a linear (dB signal level) – linear (dB ratio) scale. + +**Table 3 – Limits for signal-to-total distortion ratio +([ITU-T G.711], 64 kbit/s) for "Noise" method** + +| Sending level dBPa
at MRP | Sending ratio
(dB) | +|--------------------------------------|-------------------------------| +| –45 | 5.0 | +| –30 | 20.0 | +| –24 | 25.5 | +| –17 | 30.2 | +| –10 | 32.4 | +| 0 | 33.0 | +| +4 | 33.0 | + +**Table 4 – Limits for signal-to-total distortion ratio +([ITU-T G.726], 32 kbit/s) for "Noise" method** + +| Sending level dBPa
at MRP | Sending ratio
(dB) | +|--------------------------------------|-------------------------------| +| –45 | 5.0 | +| –30 | 20.0 | +| –24 | 25.3 | +| –17 | 29.7 | +| –10 | 31.6 | +| 0 | 32.1 | +| +4 | 32.1 | + +##### **5.1.4.1.1 Measurement method** + +The input at MRP is a band-limited noise signal corresponding to [ITU-T O.131]. The test signal is then applied at –45, –40, –35, –30, –24, –20, –17, –10, –5, 0, 4 dBPa. + +The ratio of the signal-to-total distortion power of the digital signal output with a psophometric noise weighting according to [ITU-T O.41] is measured (see [ITU-T O.131]). + +##### **5.1.4.1.2 Sending distortion using "Sinewave" method** + +The ratio of signal-to-total distortion power measured with the proper noise weighting (see [ITU-T O.41]) shall be above the limits given in Tables 5, 6 and 7 for [ITU-T G.711] (64 kbit/s), [ITU-T G.711] (56 kbit/s) and [ITU-T G.726] (32 kbit/s) respectively. The input sound pressure level is limited at +10 dBPa for this measurement. + +Limits for intermediate levels are found by drawing straight lines between the breaking points in the table on a linear (dB signal level) – linear (dB ratio) scale. + +**Table 5 – Limits for signal-to-total distortion ratio +([ITU-T G.711], 64 kbit/s) for "Sinewave" method** + +| Frequency
(Hz) | Sending level
(dBPa at MRP) | Sending ratio
(dB) | +|---------------------------|----------------------------------------|-------------------------------| +| 315 | -4.7 | 26 | +| 510 | -4.7 | 30.5 | +| 1020 | -35 | 17.5 | +| | -30 | 22.5 | +| | -20 | 30.7 | +| | -10 | 33.3 | +| | 0 | 33.7 | +| | +7 | 31.7 | +| | +10 | 25.5 | + +**Table 6 – Limits for signal-to-total distortion ratio +([ITU-T G.711], 56 kbit/s) for "Sinewave" method** + +| Sending level
(dBPa at MRP) | Sending ratio
(dB) | +|----------------------------------------|-------------------------------| +| -35 | 15.3 | +| -30 | 20.3 | +| -20 | 27.5 | +| -10 | 28.5 | +| 0 | 28.6 | +| +7 | 27.9 | +| +10 | 24.2 | + +**Table 7 – Limits for signal-to-total distortion ratio +([ITU-T G.726], 32 kbit/s) for "Sinewave" method** + +| Sending level
(dBPa at MRP) | Sending ratio
(dB) | +|----------------------------------------|-------------------------------| +| -35 | 17.3 | +| -30 | 22.3 | +| -20 | 29.3 | +| -10 | 31.1 | +| 0 | 31.3 | +| +7 | 30.0 | +| +10 | 25.0 | + +###### 5.1.4.2.1 Measurement method + +After a correct activation of the system, a sinewave signal at frequencies of 315, 510 and 1020 Hz is applied at MRP respectively. The signal level shall be calibrated to $-4.7$ dBPa at MRP for all frequencies, except for the sinewave signal with a frequency of 1020 Hz that shall be applied at MRP at the following levels: $-35, -30, -25, -20, -15, -10, -5, 0, 7, 10$ dBPa. The input sound pressure level is limited at $+10$ dBPa for this measurement. + +The ratio of the signal-to-total distortion power of the digital signal output is measured with a psophometric noise weighting according to [ITU-T O.41]. + +NOTE – In cases where the sound pressure exceeds $+6$ dBPa, the linearity of the artificial mouth should be checked as it exceeds the ITU-T P.51 limits. For good performance, in this case, it is recommended to use a suitable individual precalibration of the artificial mouth for compensation of the deviation of the measured data by taking into account the calibration results. + +### 5.1.5 Sending out-of-band signals + +In view of the following considerations: + +- compatibility with coder and decoder requirements according to [ITU-T G.712]; +- compatibility with existing practices in the mixed analogue-digital network in use today, + +the following limits are recommended: + +With any sinewave signal above 4.6 kHz and up to 8 kHz applied at MRP at a level of $-4.7$ dBPa, the level of any image frequency produced at the digital interface shall be below a reference level obtained at 1020 Hz ( $-4.7$ dBPa at MRP) by at least the amount (in dB) specified in Table 8. + +**Table 8 – Discrimination levels – Sending out-of-band signals** + +| Applied sinewave frequency
(kHz) | Limit (minimum)
(dB) (Note) | +|-------------------------------------|--------------------------------| +| 4.6 | 30 | +| 8.0 | 40 | + +NOTE – The limit at intermediate frequencies lies on a straight line drawn between the given values on a log (frequency) – linear (dB) scale. + +#### 5.1.5.1 Measurement method + +The reference signal used is an activation signal followed by a sine wave at 1020 Hz. The reference signal is applied at MRP ( $-4.7$ dBPa at MRP). The level is measured at the digital interface. + +The actual test signals used are an activation signal followed by a sine wave with the frequencies of 4.65 kHz, 5 kHz, 6 kHz, 6.5 kHz, 7 kHz and 7.5 kHz ( $-4.7$ dBPa at MRP) respectively. The level of any image frequencies at the digital interface is measured. + +### 5.1.6 Sending linearity + +Considering that some non-linear techniques may be used in digital telephones, for example: + +- automatic gain control; +- compressor/expander techniques, + +these devices may be deliberately designed non-linear over the input level range specified, and may have dynamic characteristics (e.g., attack and hangover time). + +For digital telephones that are intended to have linear input-versus-output characteristics, the linearity relative to the gain for ARL should remain within the limits given in Table 9. For intermediate levels, the same limits for gain variation apply. + +**Table 9 – Sending linearity** + +| Sending dB relative to ARL | Upper limit (dB) | Lower limit (dB) | +|-----------------------------------|-------------------------|-------------------------| +| 13 | 0.5 | –0.5 | +| 0 | 0.5 | –0.5 | +| –30 | 0.5 | –0.5 | +| –30 | 1 | – $\infty$ | +| –40 | 1 | – $\infty$ | +| <–40 | 2 | – $\infty$ | + +NOTE 1 – In cases where the sound pressure at MRP exceeds +6 dBPa, the linearity of the artificial mouth should be checked, as it exceeds the ITU-T P.51 limits. For good performance, in this case, it is recommended to use a suitable individual pre-calibration of the artificial mouth for compensation of the deviation of the measured data by taking into account the calibration results. + +NOTE 2 – If a digital telephone has specifically designed non-linear characteristics, alternative test methods based on more complex test signals may be needed for these tests. The principles are described in [ITU-T P.501] and [ITU-T P.502]. + +#### **5.1.6.1 Measurement method** + +After a correct activation of the system, a sinewave signal with a frequency in the range 1004 Hz to 1025 Hz is applied at MRP. The level of this signal is adjusted until the output of the terminal is –10 dBm0. The level of the signal at MRP is then the ARL. + +The test signal shall be applied at the following levels: + +–45, –40, –35, –30, –25, –20, –15, –10, –5, 0, 5, 10, 13 dB relative to ARL. + +The variation of gain relative to the gain for ARL is measured. + +NOTE 1 – If the sinewave techniques cannot be used (which is likely if the terminal has a noise reduction device), another appropriate technique should be applied. For example, an artificial speech generator (e.g., as specified in [ITU-T P.50] and [ITU-T P.501]) and a spectrum analyser can be used. The test signal used shall be specified in the test report. + +NOTE 2 – Selective measurements may be used to avoid the effects of ambient noise. + +## **5.2 Receiving characteristics** + +### **5.2.1 Receiving frequency response** + +In view of the following consideration: + +- Unlike other standards, this Recommendation does not use ERP as the reference point for receiving, but uses the diffuse-field instead. With the concept of diffuse-field based receive measurements, a rising slope for the overall frequency response is achieved by a flat target frequency response in sending and a diffuse-field based receiving frequency response; the sensitivity mask of Table 10 is recommended. + +**Table 10 – Receiving frequency response** + +| Frequency (Hz) | Upper limit (dB) | Lower limit (dB) | +|----------------|------------------|------------------| +| 100 | 4 | | +| 300 | 4 | -4 | +| 3400 | 4 | -4 | +| 4000 | 4 | | + +NOTE – The limit curves shall be determined by straight lines joining successive coordinates given in the table, where frequency response is plotted on a linear dB scale against frequency on a logarithmic scale. It is a floating or 'best fit' mask. + +![Graph showing the receiving mask with HATS, diffuse-field equalized. The Y-axis is dB from -20 to 10. The X-axis is Frequency in Hz on a logarithmic scale from 100 to 10,000. A red curve shows the mask boundary, and a black line shows the measured response. The graph is labeled P.310(09)_F5.3.](ebce355620876e10f907f8b71926c112_img.jpg) + +The figure is a line graph on a semi-logarithmic scale. The vertical axis (Y-axis) represents gain in dB, ranging from -20 to 10 in increments of 5. The horizontal axis (X-axis) represents frequency in Hz on a logarithmic scale, with major markers at 100, 1000, and 10,000 Hz. A red curve represents the 'receiving mask' boundary. It starts at -20 dB at 100 Hz, rises steeply to 0 dB at approximately 200 Hz, remains flat at 0 dB until about 3000 Hz, and then drops sharply back to -20 dB at 4000 Hz. A black line represents the 'measured response'. It starts at 4 dB at 100 Hz, remains flat until 300 Hz, drops to -3 dB at 300 Hz, remains flat until 3400 Hz, and then drops sharply to -20 dB at 4000 Hz. Small black diamonds mark the corner points of the response at (100, 4), (300, -3), (3400, -3), and (4000, -20). The text 'P.310(09)\_F5.3' is located in the bottom right corner of the plot area. + +Graph showing the receiving mask with HATS, diffuse-field equalized. The Y-axis is dB from -20 to 10. The X-axis is Frequency in Hz on a logarithmic scale from 100 to 10,000. A red curve shows the mask boundary, and a black line shows the measured response. The graph is labeled P.310(09)\_F5.3. + +**Figure 5-3 – Receiving mask with HATS, diffuse-field equalized** + +#### **5.2.1.1 Measurement method** + +The receiving frequency characteristic is measured according to [ITU-T P.64] using the measurement set-up shown in Figure 5-4. + +A sinewave or an artificial speech generator (e.g., as specified in [ITU-T P.50] and [ITU-T P.501]), a spectrum analyser and a diffuse-field equalized HATS are used. Additional test methods may be found in [ITU-T P.502]. The test signal level shall be -16.0 dBm0, measured at the digital reference point or the equivalent analogue point. The test signal level is averaged over the complete test signal sequence. The sensitivity is expressed in terms of dBPa/V, referred to the HATS Reference Point (HRP). Information about HATS diffuse-field frequency response is available in [ITU-T P.58]. + +![Figure 5-4: Measurement of receiving frequency characteristic, HATS position. The diagram shows a 'Quiet room' (dashed box) containing an 'Artificial ear', a 'Digital telephone', and a 'Measuring amplifier'. The 'Measuring amplifier' is connected to a 'Level recorder or distortion meter'. The 'Digital telephone' is connected to an 'Interface', which is also connected to a 'Control'. The 'Interface' is connected to a 'Test bed' containing an 'A/D' converter. The 'A/D' converter is connected to a 'Signal generator' and a '600 Ω' resistor.](e69b9188aa2c14ec6b21c83f711fef65_img.jpg) + +``` + +graph LR + subgraph Quiet_room [Quiet room] + AE[Artificial ear] --- DT[Digital telephone] + MA[Measuring amplifier] --- AE + end + DT --- Interface + Control --- Interface + Interface --- Test_bed[Test bed] + subgraph Test_bed_sub [Test bed] + AD[A/D] + end + AD --- SG[Signal generator] + SG --- R[600 Ω] + R --- AD + MA --- LDM[Level recorder or distortion meter] + +``` + +Figure 5-4: Measurement of receiving frequency characteristic, HATS position. The diagram shows a 'Quiet room' (dashed box) containing an 'Artificial ear', a 'Digital telephone', and a 'Measuring amplifier'. The 'Measuring amplifier' is connected to a 'Level recorder or distortion meter'. The 'Digital telephone' is connected to an 'Interface', which is also connected to a 'Control'. The 'Interface' is connected to a 'Test bed' containing an 'A/D' converter. The 'A/D' converter is connected to a 'Signal generator' and a '600 Ω' resistor. + +P.310(09)\_F5.4 + +**Figure 5-4 – Measurement of receiving frequency characteristic, HATS position** + +### 5.2.2 Receiving loudness rating (RLR) + +- The nominal value of RLR shall be +2 dB, with a tolerance of $\pm 3$ dB + +#### 5.2.2.1 Measurement method + +This should be calculated from the sensitivity/frequency characteristic determined in clause 5.2.1.1 by means of [ITU-T P.79]. The DRP-ERP correction as defined in [ITU-T P.57] is applied. + +NOTE – Other methods for calculating loudness rating used by some Administrations for their own internal planning purposes can be found in [b-Teleph]. + +### 5.2.3 Receiving noise + +When driven by a signal corresponding to a decoder value quiet code, the maximum acoustic noise level at ERP shall be as follows: + +- if no user-controlled volume control is provided or when the volume control is set to nominal RLR value, the measured receiving noise at ERP shall not be greater than $-57$ dBPa(A); +- if a volume control is provided, the measured receiving noise at ERP shall not be greater than $-54$ dBPa(A) at the maximum setting of the volume control; +- no peaks in any 1/3-octave band with the level of 10 dB higher than the average noise spectrum in the frequency domain shall occur. + +#### 5.2.3.1 Measurement method + +The receiving noise shall be measured in a quiet environment as defined in clause 4.4. + +The receiving noise level at ERP is measured with A-weighting. The DRP-ERP correction as defined in [ITU-T P.57] is applied. + +### 5.2.4 Receiving distortion + +#### 5.2.4.1 Receiving distortion using "Noise" method + +The ratio of signal-to-total distortion (harmonic and quantizing) power of the signal measured at ERP with A-weighting shall be above the limits given in Tables 11 and 12 for [ITU-T G.711] (64 kbit/s) and [ITU-T G.726] (32 kbit/s), respectively, unless the signal in the artificial ear exceeds $+5$ dBPa or is less than $-50$ dBPa. + +Limits for intermediate levels are found by drawing straight lines between the breaking points in the table on a linear (dB signal level) – linear (dB ratio) scale. + +**Table 11 – Limits for signal-to-total distortion ratio +([ITU-T G.711], 64 kbit/s) for "Noise" method** + +| Receiving level
at the digital interface
(dBm0) | Receiving ratio
(dB) | +|----------------------------------------------------------------|---------------------------------| +| –55 | 5.0 | +| –40 | 20.0 | +| –34 | 25.0 | +| –27 | 30.6 | +| –20 | 33.0 | +| –10 | 33.7 | +| –6 | 33.8 | +| –3 | 24.0 | + +**Table 12 – Limits for signal-to-total distortion ratio +([ITU-T G.726], 32 kbit/s) for "Noise" method** + +| Receiving level
at the digital interface
(dBm0) | Receiving ratio
(dB) | +|----------------------------------------------------------------|---------------------------------| +| –55 | 5.0 | +| –40 | 20.0 | +| –34 | 24.8 | +| –27 | 30.1 | +| –20 | 32.3 | +| –10 | 32.9 | +| –6 | 32.9 | +| –3 | 23.4 | + +##### **5.2.4.1.1 Measurement method** + +A digitally simulated band-limited noise signal corresponding to [ITU-T O.131] is applied to the digital interface at the following levels: –55, –50, –45, –40, –34, –30, –27, –20, –15, –10, –6, –3 dBm0. + +The total distortion power shall be measured at ERP with A-weighting. The ratio of the signal-to-total distortion power is calculated (see [ITU-T O.131]). The DRP-ERP correction defined in [ITU-T P.57] is applied. + +#### **5.2.4.2 Receiving distortion using the "Sinewave" method** + +The ratio of signal-to-total distortion power measured at ERP with A-weighting shall be above the limits given in Tables 13, 14 and 15 for [ITU-T G.711] (64 kbit/s), [ITU-T G.711] (56 kbit/s) and [ITU-T G.726] (32 kbit/s) respectively, unless the signal in the artificial ear exceeds +10 dBPa or is less than –50 dBPa. + +Limits for intermediate levels are found by drawing straight lines between the breaking points in the table on a linear (dB signal level) – linear (dB ratio) scale. + +**Table 13 – Limits for signal-to-total distortion ratio +([ITU-T G.711], 64 kbit/s) for "Sinewave" method** + +| Frequency
(Hz) | Receiving level
at the digital interface
(dBm0) | Receiving ratio
(dB) | +|---------------------------|----------------------------------------------------------------|---------------------------------| +| 315 | -16 | 20 | +| 510 | -16 | 30.5 | +| 1020 | -45 | 17.5 | +| | -40 | 22.5 | +| | -30 | 30.5 | +| | -20 | 33.0 | +| | -10 | 33.5 | +| | -3 | 31.2 | +| | 0 | 25.5 | + +**Table 14 – Limits for signal-to-total distortion ratio +([ITU-T G.711], 56 kbit/s) for "Sinewave" method** + +| Receiving level at
the digital interface
(dBm0) | Receiving ratio
(dB) | +|----------------------------------------------------------------|---------------------------------| +| -45 | 15.3 | +| -40 | 20.3 | +| -30 | 27.4 | +| -20 | 28.4 | +| -10 | 28.6 | +| -3 | 27.7 | +| 0 | 24.2 | + +**Table 15 – Limits for signal-to-total distortion ratio +([ITU-T G.726], 32 kbit/s) for "Sinewave" method** + +| Receiving level at
the digital interface
(dBm0) | Receiving ratio
(dB) | +|----------------------------------------------------------------|---------------------------------| +| -45 | 17.3 | +| -40 | 22.3 | +| -30 | 29.2 | +| -20 | 30.9 | +| -10 | 31.2 | +| -3 | 29.7 | +| 0 | 25.0 | + +##### 5.2.4.2.1 Measurement method + +After a correct activation of the system, a digitally simulated sinewave signal at frequencies of 315, 510 and 1020 Hz is applied to the digital interface respectively. The signal level is $-16 \text{ dBm}0$ , except for the sinewave signal with a frequency of 1020 Hz that shall be applied to the digital interface at the following levels: $-45, -40, -35, -30, -25, -20, -15, -10, -3, 0 \text{ dBm}0$ . + +The total distortion power shall be measured at ERP with A-weighting. The DRP-ERP correction as defined in [ITU-T P.57] is applied. The ratio of the signal-to-total distortion power is calculated. + +### 5.2.5 Receiving out-of-band signals + +In view of the following considerations: + +- compatibility with coder and decoder requirements according to [ITU-T G.712]; +- compatibility with existing practices in the mixed analogue-digital network in use today, + +the following limits are recommended: + +Any spurious out-of-band image signal above 4.6 kHz and up to 8 kHz measured selectively shall be lower than the in-band acoustic level measured with a reference signal (1000 Hz, sinewave). The minimum level difference between the reference signal level and the out-of-band image signal level shall be as given in Table 16: + +**Table 16 – Discrimination levels – Receiving out-of-band signals** + +| Image signal frequency | Minimum level difference | +|------------------------|--------------------------| +| 4.6 kHz | 35 dB | +| 8.0 kHz | 45 dB | + +NOTE – The limit at intermediate frequencies lies on a straight line drawn between the given values on a log (frequency) – linear (dB) scale. + +#### 5.2.5.1 Measurement method + +The actual test signals used are a digitally activation signal followed by a digitally simulated sine wave with the frequencies of 500 Hz, 1000 Hz, 2000 Hz and 3150 Hz ( $-16 \text{ dBm}0$ , at the digital reference point or the equivalent analogue point) respectively. The level of spurious out-of-band image signals at frequencies up to 8 kHz are measured selectively at the ear simulator. + +The measured spurious out-of-band signals shall be compared to the measured in-band 1000 Hz signal. + +NOTE – Depending on the type of codec, the test signal used may need to be adapted. + +### 5.2.6 Receiving linearity + +Considering that some non-linear techniques may be used in digital telephones, for example: + +- automatic gain control; +- compressor/expander techniques, + +these devices may be deliberately designed non-linear over the input level range specified and may have dynamic characteristics (e.g., attack and hangover time). Unless a digital telephone has specifically designed non-linear characteristics, the following limits are recommended. + +NOTE – If a digital telephone has specifically designed non-linear characteristics, alternative test methods based on more complex test signals may be needed for these tests. The principles are described in [ITU-T P.501] and [ITU-T P.502]. + +For digital telephones that are intended to have linear input versus output characteristics, the gain variation relative to the gain at an input level of $-10$ dBm0, should be within the limits given in Table 17. For intermediate levels, the same limits for gain variation apply. + +**Table 17 – Receiving linearity** + +| Receiving level
at the digital interface
(dBm0) | Upper limit
(dB) | Lower limit
(dB) | +|-------------------------------------------------------|---------------------|---------------------| +| +3 | 0.5 | -0.5 | +| -40 | 0.5 | -0.5 | +| -40 | 1 | -1 | +| -50 | 1 | -1 | +| <-50 | 2 | -2 | + +#### 5.2.6.1 Measurement method + +After a correct activation of the system, a digitally simulated sinewave signal with a frequency in the range 1004 Hz to 1025 Hz shall be applied at the digital interface at the following levels: + +$-55, -50, -45, -40, -35, -30, -25, -20, -15, -10, -5, 0, 3$ dBm0. + +The variation of gain relative to the gain at an input level of $-10$ dBm0 shall be measured in the artificial ear. The measured sound pressure level should be transferred from DRP to ERP according to [ITU-T P.57]. + +NOTE 1 – If the sinewave techniques cannot be used (which is likely if the terminal has a noise reduction device), another appropriate technique should be applied. For example, an artificial speech generator (e.g., as specified in [ITU-T P.50] and [ITU-T P.501]) and a spectrum analyser can be used. + +NOTE 2 – Selective measurements may be used to avoid the effects of ambient noise. + +## 5.3 Sidetone characteristics + +### 5.3.1 Sidetone masking rating (STMR) + +In view of the following considerations: + +- the optimum STMR for conditions free from echo; +- the masking effect of sidetone on short delay talker echo, + +the value of STMR shall be 10 dB to 20 dB when corrected to nominal values of SLR (8 dB) and RLR (2 dB). + +NOTE – For this correction, the formula $STMR = STMR_m - (SLR_m - 8 + RLR_m - 2)$ should be used. Index m indicates measured values. + +Where a user-controlled receiving volume control is provided, the STMR shall be 10 dB to 20 dB for all volume control settings. It is preferable to have a constant STMR independent of the volume control setting. + +#### 5.3.1.1 Measurement method + +The test signal to be used for the measurements shall be the artificial voice according to [ITU-T P.50]. The spectrum of the acoustic signal produced by the artificial mouth is calibrated under free-field conditions at MRP. The test signal level shall be $-4.7$ dBPa at MRP. The test signal level is averaged over the complete test signal sequence. + +The talker sidetone sensitivity/frequency characteristic is measured according to [ITU-T P.64] using the measurement set-up of Figure 5-5. The reference codec is not used in this measurement but may remain in the test circuit, with no external coupling path. + +Sidetone masking rating (STMR) should be calculated from the sensitivity/frequency characteristic determined from above by means of [ITU-T P.79]. The measured sound pressure should be transferred from DRP to ERP. + +![Figure 5-5: Measurement of talker sidetone sensitivity/frequency characteristic, HATS position. The diagram shows a measurement setup where a Digital telephone is connected to an Interface, which is connected to a Test bed containing a D/A converter and a 600 Ω load. The Digital telephone is also connected to a Measuring amplifier and a Level recorder. The Measuring amplifier is connected to an Artificial ear, which is connected to a Signal generator. The Artificial ear and Signal generator are connected to a Quiet room. The Digital telephone is connected to the Artificial ear via a Control line. The Test bed is connected to the Interface via a Control line.](a149b400127a3e3e50b3c98d27c5935c_img.jpg) + +Figure 5-5: Measurement of talker sidetone sensitivity/frequency characteristic, HATS position. The diagram shows a measurement setup where a Digital telephone is connected to an Interface, which is connected to a Test bed containing a D/A converter and a 600 Ω load. The Digital telephone is also connected to a Measuring amplifier and a Level recorder. The Measuring amplifier is connected to an Artificial ear, which is connected to a Signal generator. The Artificial ear and Signal generator are connected to a Quiet room. The Digital telephone is connected to the Artificial ear via a Control line. The Test bed is connected to the Interface via a Control line. + +P.310(09)\_F5.5 + +**Figure 5-5 – Measurement of talker sidetone sensitivity/frequency characteristic, HATS position** + +### 5.3.2 D-factor + +In view of the following considerations: + +- the difficulties of high ambient noise conditions; +- what subscribers are used to having with present analogue sets, + +the value of D-factor shall not be less than 0 dB. As the long-term objective, the value of +3 dB is recommended. + +NOTE 1 – The key parameter for the handset performance in noisy conditions is the D-factor. + +NOTE 2 – When a telephone set is connected into the telecom network, there is a firm relation between STMR and LSTR: $D = LSTR - STMR$ , here LSTR is the listener sidetone loudness rating. + +#### 5.3.2.1 Measurement method + +The measurement method is defined according to Annex E of [ITU-T P.79]. + +For linear microphones and circuitry, D-factor is computed directly from measurements of the difference $\Delta S_m$ between the sending sensitivities for diffuse and direct sound, $S_{si}$ (diff) and $S_{si}$ (direct), respectively. + +$$\Delta S_m = S_{si} (\text{diff}) - S_{si} (\text{direct})$$ + +Both $S_{si}$ (diff) and $S_{si}$ (direct) can be measured according to clause 5.1.1, using the measurement set-up of Figure 5-6 except that diffuse sound is used as the input signal when measuring $S_{si}$ (diff). The sound pressure level of the diffuse sound field shall be adjusted in the range of –54 dBPa(A) to –29 dBPa(A) at MRP in the absence of artificial mouth. The actual level and type of noise should always be stated in quoting test results. + +$D$ is computed as a weighted average of $\Delta S_m$ , + +$$D = -\sum_{i=1}^N K_i \cdot \Delta S_m$$ + +The coefficients $K_i$ are given in Table E.1 of [ITU-T P.79]. + +NOTE – For sets with non-linear microphones and/or non-linear circuitry, $D$ depends on the room noise level. + +![Figure 5-6: Measurement of listener sidetone sensitivity/frequency characteristic, HATS position. The diagram shows a measurement setup. A 'Digital telephone' is connected to an 'Artificial mouth' and an 'Artificial ear'. The 'Artificial ear' is connected to a 'Measuring amplifier', which is connected to 'Channel 2' of a 'Spectrum analyser'. The 'Digital telephone' is connected to an 'Interface', which is connected to a 'Computer' and a 'Plotter'. The 'Computer' is also connected to 'Channel 1' of the 'Spectrum analyser'. The 'Interface' is connected to a 'Test bed' containing a 'D/A' converter, which is connected to a '600 Ω' load. A 'Control' signal is also connected to the 'Interface'. The 'Artificial mouth', 'Artificial ear', and 'Digital telephone' are enclosed in a dashed box labeled 'Diffuse sound field'. The diagram is labeled 'P.310(09)_F5.6'.](b235edb1dbe659e2782c9a0e47775ca4_img.jpg) + +Figure 5-6: Measurement of listener sidetone sensitivity/frequency characteristic, HATS position. The diagram shows a measurement setup. A 'Digital telephone' is connected to an 'Artificial mouth' and an 'Artificial ear'. The 'Artificial ear' is connected to a 'Measuring amplifier', which is connected to 'Channel 2' of a 'Spectrum analyser'. The 'Digital telephone' is connected to an 'Interface', which is connected to a 'Computer' and a 'Plotter'. The 'Computer' is also connected to 'Channel 1' of the 'Spectrum analyser'. The 'Interface' is connected to a 'Test bed' containing a 'D/A' converter, which is connected to a '600 Ω' load. A 'Control' signal is also connected to the 'Interface'. The 'Artificial mouth', 'Artificial ear', and 'Digital telephone' are enclosed in a dashed box labeled 'Diffuse sound field'. The diagram is labeled 'P.310(09)\_F5.6'. + +**Figure 5-6 – Measurement of listener sidetone sensitivity/frequency characteristic, HATS position** + +## 5.4 Echo path loss characteristics + +### 5.4.1 Weighted terminal coupling loss (TCLw) + +In view of the following considerations: + +- the aim to achieve as high an acoustic coupling loss as possible to minimize degradation caused by echo; +- that the far-end talker echo should be controlled under all volume control settings, and for the range of transducers sensitivities as long as the handset is properly used; +- the far-end terminal may be connected via a mobile or an IP network which introduces long talker echo path delay; +- what is practically obtainable in real use where the customer himself chooses the way to hold his handset, + +the following limit is recommended: + +In order to meet the ITU-T G.131 talker echo objective requirements, the weighted terminal coupling loss (TCLw) should be greater than 45 dB when measured under free-field conditions and with SLR normalized to SLR = +8 dB and RLR = +2 dB. For example, if the measured TCLw is 48 dB, the measured SLR is +9 dB and the measured RLR is +3 dB, then the normalized value of TCLw = 48 dB + (8 – 9) dB + (2 – 3) dB = 46 dB. + +For handsets fitted with a volume control, the TCLw shall be not less than 40 dB for the higher gain settings above the nominal setting of the volume control. Permanently increasing SLR as the RLR is decreased when increasing the receive volume level is not recommended. + +NOTE – In consideration of the increasing delays introduced by modern networks, a higher TCLw value than that specified here may be necessary for proper operation with these networks. + +#### 5.4.1.1 Measurement method + +Terminal coupling loss (TCL) is measured in such a way that the inherent mechanical coupling of the handset is not affected. + +When performing tests, the test space acoustics must not have a dominating influence. The test should be performed in the environment as defined in clause 4.4. + +The test is performed with the handset mounted at HATS position according to [ITU-T P.64]. + +The test signal may be a composite source signal (CSS) as defined in [ITU-T P.501] with a length of 4096 points (for the 48 kHz sampling rate) and a crest factor of 6 dB. The low crest factor is achieved by random alternation of the phase between $-180^\circ$ and $180^\circ$ . The length of the complete test signal composed of at least four sequences of CSS shall be at least one second. The test signal level is $-3$ dBm0. The calibration shall be determined during the ON portions of the signal. The test signal shall be band-limited to 100 Hz $\sim$ 4000 Hz. A logarithmically distributed multi-sinewave is equally well applicable with G.711 codecs. + +For non-linear and/or time variant systems, it must be ensured that the equipment under test is under "steady state conditions". Depending on the task, e.g., echo cancellers should be fully converged. This can be achieved by invoking the test signal for at least 2 seconds before the actual measurement occurs. + +The attenuation from digital input to digital output is measured at one-twelfth octave frequencies as given by the R.40-series of preferred numbers in [ISO 3] for frequencies from 300 to 3350 Hz, using the measurement arrangement shown in Figure 5-7. + +The TCLw is calculated according to clause B.4 of [ITU-T G.122] (trapezoidal rule). + +NOTE 1 – The echo impairment perceived by the person at the opposite end of the connection from a telephone set is a function of the magnitude of the talker echo signal as well as the talker echo path delay. The echo signal becomes more disturbing as the talker echo path delay increases. Thus, a telephone set with adequate TCLw performance on low delay connections may provide satisfactory performance while the same may not be true for connections that have a long delay. + +NOTE 2 – There might be problems measuring 46 dB TCL in the case where sophisticated coding with limited dynamic range is used. In such cases, typically speech or speech-like test signals need to be used that themselves have crest factors in the range of 15 dB, which reduces the measurement dynamic by the same amount. In such cases, the signal measured in the sending direction should be evaluated more carefully in order to find whether an echo signal is present, or whether this signal is completely masked by the noise signal introduced by the codec. If the signal measured in the sending direction is completely masked by the noise, the requirement can be considered to be fulfilled. If this is not the case, more sophisticated measurement procedures such as time averaging (in order to improve the signal-to-noise ratio) need to be applied in order to achieve reliable measurement results. + +![Figure 5-7: Measurement of terminal coupling loss, HATS position. The diagram shows a 'Digital telephone set' connected to an 'Interface'. A 'Control' signal is input to the Interface. The Interface is connected to a 'Test bed' (dashed box). Inside the Test bed, there are two parallel paths: one with a 'D/A' converter followed by a '600 Ω' resistor, and another with an 'A/D' converter followed by a '600 Ω' resistor. The output of the D/A path is labeled 'V_out' and the output of the A/D path is labeled 'V_in'. A microphone icon is shown next to the Digital telephone set. The label 'P.310(09)_F5.7' is at the bottom left.](f0a97d0d3818a253c1d2a009966081b1_img.jpg) + +Figure 5-7: Measurement of terminal coupling loss, HATS position. The diagram shows a 'Digital telephone set' connected to an 'Interface'. A 'Control' signal is input to the Interface. The Interface is connected to a 'Test bed' (dashed box). Inside the Test bed, there are two parallel paths: one with a 'D/A' converter followed by a '600 Ω' resistor, and another with an 'A/D' converter followed by a '600 Ω' resistor. The output of the D/A path is labeled 'V\_out' and the output of the A/D path is labeled 'V\_in'. A microphone icon is shown next to the Digital telephone set. The label 'P.310(09)\_F5.7' is at the bottom left. + +**Figure 5-7 – Measurement of terminal coupling loss, HATS position** + +### 5.4.2 Stability loss + +In view of the following considerations: + +- the aim to achieve a good stability; +- what is practically obtainable with normal type of handsets and transducers, + +the following limit is recommended: + +With the handset lying on and the transducers facing a hard surface, the attenuation from the digital input to the digital output shall be preferably at least 10 dB but not less than 6 dB at all frequencies in the range of 200 Hz to 4 kHz with SLR + RLR normalized to OLR = +10 dB. + +NOTE – Those handsets that are fitted with a volume control should maintain stability throughout the volume control range. + +#### 5.4.2.1 Measurement method + +The signal used in stability measurement is the same with TCLw, with a different input signal level of 0 dBm0. + +The measurement is made at one-twelfth octave intervals for frequencies from 200 Hz to 4000 Hz. With the handset and the transmission circuit fully active, the attenuation from digital input to digital output is measured under one of the following conditions. Stability is the least value of the attenuation in the band 200 to 4000 Hz. + +##### Method 1 + +- a) The handset shall be positioned on one inside surface at the corner formed by the intersection of three perpendicular, smooth, hard surfaces. Each surface shall extend 0.5 m from the vertex of the corner. One surface shall be marked with a diagonal line extending from the corner and a reference position 250 mm from the corner formed by the three surfaces, as shown in Figure 5-8. +- b) The handset, with the transmission circuit fully active, shall be positioned on the defined surface as follows: + - i) the mouthpiece and ear-cap shall face towards the surface; + - ii) the handset shall be placed centrally on the diagonal line with the ear-cap nearest to the apex of the corner; + - iii) the extremity of the handset shall coincide with the normal to the reference point, as shown in Figure 5-8. + +![Figure 5-8: Reference corner and reference position. The diagram shows a 3D corner formed by three perpendicular surfaces. A diagonal line is drawn on the bottom surface from the corner vertex. A reference point is marked on this diagonal line at a distance of 250 mm from the corner. Below the 3D diagram, a 2D side view shows a handset positioned on the diagonal line, with its ear-cap facing the corner and its mouthpiece facing away. The distance from the corner to the reference point is labeled 250. The text 'P.310(09)_F5.8' is visible in the 3D diagram.](27a0ba79d8b4a7d9d9380ca6a3229f1a_img.jpg) + +The diagram illustrates the setup for Method 1. It features a 3D perspective view of a corner where three surfaces meet. A diagonal line is marked on the bottom surface, starting from the corner vertex. A reference point is indicated on this line at a distance of 250 mm from the corner. Below this, a 2D side view shows a handset placed on the diagonal line. The handset's ear-cap is positioned closest to the corner, and its mouthpiece is further away. A dimension line indicates the 250 mm distance from the corner to the reference point. The label 'P.310(09)\_F5.8' is present in the 3D diagram. + +Figure 5-8: Reference corner and reference position. The diagram shows a 3D corner formed by three perpendicular surfaces. A diagonal line is drawn on the bottom surface from the corner vertex. A reference point is marked on this diagonal line at a distance of 250 mm from the corner. Below the 3D diagram, a 2D side view shows a handset positioned on the diagonal line, with its ear-cap facing the corner and its mouthpiece facing away. The distance from the corner to the reference point is labeled 250. The text 'P.310(09)\_F5.8' is visible in the 3D diagram. + +Figure 5-8 – Reference corner and reference position + +##### Method 2 + +The handset, with the transmission circuit fully active, is placed with the ear-cap and mouthpiece facing a hard, smooth surface free of any other object for 0.5 m. + +## 5.5 Delay + +Delay is a complex end-to-end issue. Certain aspects of delay can be optimized in digital terminals, such as the internal hardware/firmware delay. Other aspects are also important sources of delay, but they are a function of the selected codec, so they cannot be optimized in digital terminals. + +Considering the delay, referred to in this Recommendation, + +- the coding, decoding and filtering according to [ITU-T G.712] and [ITU-T G.726], as appropriate; +- the air paths and transducers involved, + +the following is recommended: + +The sum of the delays from MRP to the digital interface (sending direction), and from the digital interface to the artificial ear (receiving direction), shall not exceed 25 ms for digital telephones using G.711 encoding, and 30 ms for G.726 encoding. + +### 5.5.1 Measurement method + +#### 5.5.1.1 Sending delay measurement + +The test signal to be used for the measurements shall be a composite source signal (CSS) as described in [ITU-T P.501]. The handset terminal is set-up as described in clause 4.3. The delay is calculated using the cross-correlation function between the signal at the electrical test point and the signal at MRP. The measurement is corrected by the delay introduced by the test equipment. The delay is expressed in ms, determined from the maximum of the cross-correlation function. + +#### 5.5.1.2 Receiving delay measurement + +The test signal to be used for the measurements shall be a composite source signal (CSS) as described in [ITU-T P.501]. The handset terminal is set-up as described in clause 4.3. The delay is calculated using the cross-correlation function between the signal at the electrical test point and the signal arriving at the artificial ear. The measurement is corrected by the delay introduced by the test equipment. The delay is expressed in ms, determined from the maximum of the cross-correlation function. + +# 6 Headset technical requirements + +Most headset technical requirements are the same as those of the handset terminal. To condense contents and avoid redundancy, this clause only covers those which are particular to headsets. + +The headset should be mounted at the test position using a HATS as specified in [ITU-T P.380]. For binaural headsets, the following applies: + +- For headsets that do not have left and right ear wearing indication, they can be tested on a HATS that is equipped with one functional ear simulator. The primary receiver (where the microphone system is located) should be positioned on the functional ear simulator for all tests. The secondary receiver shall be moved into position for evaluating frequency response, RLR and distortion measurements. +- For headsets that are designed to be worn in only one manner or fashion, a HATS with two functional ear simulators shall be used. The two receivers shall be tested separately. The secondary receiver need only be tested for frequency response, RLR and distortion. + +## 6.1 Sending characteristics + +Refer to clause 4.3 for headset measurement set-up. Refer to clause 5 for sending requirements and measurements. + +## 6.2 Receiving characteristics + +### 6.2.1 Receiving loudness rating (RLR) + +- the nominal value of RLR shall be +2 dB, with a tolerance of $\pm 3$ dB; +- for binaural earphones, the nominal value of RLR shall be 8 dB for each earphone, with a tolerance of $\pm 3$ dB. + +#### 6.2.1.1 Measurement method + +This should be calculated from the sensitivity/frequency characteristic determined in clause 5.1.1.1 by means of [ITU-T P.79]. The DRP-ERP correction, as defined in [ITU-T P.57], is applied. + +NOTE – Other methods for calculating loudness rating used by some Administrations for their own internal planning purposes can be found in [b-Teleph]. + +### 6.2.2 Receiving noise + +When driven by a signal corresponding to a decoder value quiet code, the maximum acoustic noise level at ERP shall be as follows: + +- if no user-controlled volume control is provided or when the volume control is set to nominal RLR value, the measured receiving noise at ERP shall not be greater than $-57$ dBPa(A); + +For binaural headsets, each earphone shall not exceed $-60$ dBPa(A). + +- if a volume control is provided, the measured receiving noise at ERP shall not be greater than $-54$ dBPa(A) at the maximum setting of the volume control; + +For binaural headsets, each earphone shall not exceed $-57$ dBPa(A). + +- no peaks in any 1/3-octave band with the level of 10 dB higher than the average noise spectrum in the frequency domain shall occur. + +#### 6.2.2.1 Measurement method + +The receiving noise shall be measured in a quiet environment as defined in clause 4.4. + +The receiving noise level at ERP is measured with A-weighting. The DRP-ERP correction as defined in [ITU-T P.57] is applied. + +## 6.3 Echo path loss characteristics + +### 6.3.1 Stability loss + +With the headset positioned on the defined surface as specified in clause 5.4.2.1, the attenuation from the digital input to the digital output shall be preferably at least 10 dB but not less than 6 dB at all frequencies in the range of 200 Hz to 4 kHz with SLR + RLR normalized to OLR = +10 dB. + +NOTE – Those handsets that are fitted with a volume control should maintain stability throughout the volume control range. + +#### 6.3.1.1 Measurement method + +The signal used in stability measurement is the same with TCLw, with a different input signal level of 0 dBm0. + +The measurement is made at one-twelfth octave intervals for frequencies from 200 Hz to 4000 Hz. With the headset and the transmission circuit fully active, the attenuation from digital input to digital output is measured under one of the following conditions. Stability is the least value of the attenuation in the band 200 to 4000 Hz. + +The headset, with the transmission circuit fully active, shall be positioned on the defined surface as follows: + +- the microphone and the receiver shall face towards the surface; +- the headset receiver shall be placed centrally at the reference point as shown in Figure 5-8; +- the headset microphone is positioned as close as possible to the receiver. + +# Annex A + +## Distortion allowances + +(This annex forms an integral part of this Recommendation) + +In producing the distortion characteristics at sending and receiving allowances for non-linear products, the following ways have been accounted for: + +- The transducers (microphone and earphone) have a 1% distortion allowance for most input levels. The exceptions are the highest and lowest input levels which have been allowed 5% and the second lowest to have 2%. +- Noise levels for sending and receiving are equivalent to $-64$ dBm0p. + +The total contribution of these factors is calculated using power summation. + +NOTE 1 – This is of particular interest for developing specifications for other types of "Waveform" codecs not covered in this Recommendation. + +NOTE 2 – It may be prudent to allow 0.2-0.4 dB to the final calculation to account for other sources of non-linearity, e.g., artificial mouth, amplifiers. + +NOTE 3 – Room noise at $\leq -64$ dBPa(A) has no significant effect. + +# Bibliography + +- [b-Teleph] ITU-T Handbook on Telephonometry (1992), *Measurement Method: Telephonometry*. + + + + + +# **SERIES OF ITU-T RECOMMENDATIONS** + +| | | +|-----------------|---------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Terminals and subjective and objective assessment methods | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects and next-generation networks | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.32-198811-I_PDF-E/2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg b/marked/P/T-REC-P.32-198811-I_PDF-E/2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..64672153e39c851d9e16fdbc702e526c7ba7e8a7 --- /dev/null +++ b/marked/P/T-REC-P.32-198811-I_PDF-E/2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:454a9958ffe168868cb7d38a0eb24418dafe31a7a4245c992089b2316ac37d3e +size 7392 diff --git a/marked/P/T-REC-P.32-198811-I_PDF-E/4086a572c080354982c11f1de4d6921d_img.jpg b/marked/P/T-REC-P.32-198811-I_PDF-E/4086a572c080354982c11f1de4d6921d_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..8deeadbafdadddaf25efa5189f27cd317c03ace5 --- /dev/null +++ b/marked/P/T-REC-P.32-198811-I_PDF-E/4086a572c080354982c11f1de4d6921d_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3f351f333533042901d8a3694dad720d98f8e3829448487bba1320d57361d14e +size 40934 diff --git a/marked/P/T-REC-P.32-198811-I_PDF-E/431b8889a0e7f676f0eef40859590349_img.jpg b/marked/P/T-REC-P.32-198811-I_PDF-E/431b8889a0e7f676f0eef40859590349_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..eeb5eae50a3792635b9ad94dc157370569b01495 --- /dev/null +++ b/marked/P/T-REC-P.32-198811-I_PDF-E/431b8889a0e7f676f0eef40859590349_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2340f32533616304fdd67b5bdf26c5bb50f1558401cc92475ace5437b7d69f84 +size 44011 diff --git a/marked/P/T-REC-P.32-198811-I_PDF-E/c3c305cefbac2e7b13be34ab87054d1e_img.jpg b/marked/P/T-REC-P.32-198811-I_PDF-E/c3c305cefbac2e7b13be34ab87054d1e_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e6ad987dc7984b3db4ab5614eb041e89c31dd421 --- /dev/null +++ b/marked/P/T-REC-P.32-198811-I_PDF-E/c3c305cefbac2e7b13be34ab87054d1e_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:51ed14b8e9efa54cb50eb83107c5b2b8758e6790a12d13b2dfec7ede443954f7 +size 34581 diff --git a/marked/P/T-REC-P.340-200005-I_PDF-E/raw.md b/marked/P/T-REC-P.340-200005-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..8fff20d434e5bbc9e9b555a829a526b83c1b895b --- /dev/null +++ b/marked/P/T-REC-P.340-200005-I_PDF-E/raw.md @@ -0,0 +1,1383 @@ + + +![ITU logo](2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg) + +The logo of the International Telecommunication Union (ITU) features the letters 'ITU' in a bold, sans-serif font, superimposed on a stylized globe with latitude and longitude lines. A lightning bolt symbol is positioned to the right of the globe. + +ITU logo + +INTERNATIONAL TELECOMMUNICATION UNION + +**ITU-T** + +**P.340** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +(05/2000) + +SERIES P: TELEPHONE TRANSMISSION QUALITY, +TELEPHONE INSTALLATIONS, LOCAL LINE +NETWORKS + +Subscribers' lines and sets + +--- + +**Transmission characteristics and speech +quality parameters of hands-free terminals** + +ITU-T Recommendation P.340 + +(Formerly CCITT Recommendation) + +--- + +# ITU-T P-SERIES RECOMMENDATIONS + +# TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS + +| | | | +|-----------------------------------------------------------------------------------------------|---------------|--------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Subscribers' lines and sets | Series | P.30 | +| | | P.300 | +| Transmission standards | Series | P.40 | +| Objective measuring apparatus | Series | P.50 | +| | | P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of quality | Series | P.80 | +| | | P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +# **Transmission characteristics and speech quality parameters of hands-free terminals** + +## **Summary** + +This Recommendation defines the generic transmission characteristics of hands-free terminals, whatever the applications or technologies integrated. In addition to the parameters that ensure performance levels equivalent to these of handset terminals (sensitivity values, response curves, etc.), it defines the parameters which have a bearing on conversation quality (gain switching characteristics, double-talk performances, etc.). It also defines categories of hands-free terminals based on the terminal's double-talk capability. + +This Recommendation deals with the characteristics of the signal-processing equipment (e.g. acoustic echo cancellation (AEC)) that can be incorporated in hands-free terminals. + +## **Source** + +ITU-T Recommendation P.340 was revised by ITU-T Study Group 12 (1997-2000) and approved under the WTSC Resolution 1 procedure on 18 May 2000. + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications. The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Conference (WTSC), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSC Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementors are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database. + +© ITU 2001 + +All rights reserved. No part of this publication may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying and microfilm, without permission in writing from ITU. + +# CONTENTS + +###### Page + +| | | | +|-------|-----------------------------------------------------------------|----| +| 1 | Introduction..... | 1 | +| 2 | Normative references ..... | 1 | +| 3 | Definitions and abbreviations ..... | 2 | +| 4 | Transmission characteristics ..... | 4 | +| 4.1 | Sending sensitivity ..... | 4 | +| 4.2 | Receiving sensitivity ..... | 4 | +| 4.3 | Frequency response curves ..... | 4 | +| 4.3.1 | Sending ..... | 4 | +| 4.3.2 | Receiving ..... | 5 | +| 4.4 | Subjective evaluations of loudness ratings ..... | 5 | +| 4.4.1 | Sending ..... | 5 | +| 4.4.2 | Receiving ..... | 5 | +| 4.5 | Objective evaluations of sensitivity and loudness ratings ..... | 6 | +| 4.5.1 | Sensitivity measurement ..... | 6 | +| 4.5.2 | Measure and computation of loudness ratings ..... | 7 | +| 4.6 | Switching parameters ..... | 7 | +| 5 | Testing conditions ..... | 9 | +| 5.1 | Test table ..... | 9 | +| 5.2 | Test arrangement ..... | 9 | +| 5.3 | Calibration of the artificial mouth ..... | 10 | +| 5.4 | Test environment ..... | 11 | +| 5.5 | Test signals ..... | 13 | +| 5.6 | Test signal levels ..... | 13 | +| 6 | Conversational quality ..... | 13 | +| 7 | Parameters measured instrumentally ..... | 14 | +| 7.1 | Attenuation range $a_H$ ..... | 14 | +| 7.2 | Attenuation distribution at idle mode ..... | 14 | +| 7.3 | Hang-over time ..... | 15 | +| 7.4 | Dynamic compression ..... | 15 | +| 7.5 | Reverberance ..... | 15 | +| 7.6 | Time-dependent transfer function ..... | 16 | +| 7.7 | Duplex behaviour ..... | 16 | +| 7.8 | Control characteristics of echo cancellation ..... | 17 | +| 7.9 | Speech detection ..... | 17 | + +| | Page | +|-------------------------------------------------------------------------------------------------------------------------------------------------|-------------| +| 7.10 Speech detection with ambient noise..... | 17 | +| 7.11 Background noise transmission ..... | 18 | +| 8 Behaviour of hands-free terminals in double talk situation ..... | 19 | +| 8.1 Categorization ..... | 19 | +| 8.2 Terminal coupling loss in double talk situation..... | 19 | +| 9 Parameters to be evaluated for each type of HFT ..... | 20 | +| 10 Acoustic echo controllers and speech enhancement devices ..... | 21 | +| 10.1 Functional units..... | 21 | +| 10.2 Delay ..... | 22 | +| 10.2.1 Processing delay ..... | 22 | +| 10.2.2 Round trip echo path delay (EPDn) – network interface..... | 22 | +| 10.3 Acoustic echo control specifications ..... | 22 | +| 10.3.1 Acoustic echo path..... | 23 | +| 10.3.2 Parameters and recommended limits ..... | 23 | +| Annex A – Test results ..... | 25 | +| A.1 Categorization of hands-free terminals based on double talk performance..... | 25 | +| A.1.1 Results of the listening only tests with variable attenuations in sending path ..... | 25 | +| A.1.2 Results of the listening only tests with variable attenuations in receiving path ..... | 26 | +| A.1.3 Results of listening only tests on echo disturbance ..... | 27 | +| A.2 Relationship between Listening only tests, Double talk tests and Conversation tests: comparison of six different hands-free terminals ..... | 29 | +| Appendix I – Preliminary values..... | 33 | +| I.1 Double talk performance..... | 33 | +| I.2 Build-up time characteristics ..... | 34 | +| I.3 Level adjustments by companding or AGC ..... | 34 | +| Appendix II – Examples of processing delays for mobile hands-free terminal ..... | 35 | +| II.1 Example 1 ..... | 35 | +| Tadd_proc (additional time for processing)..... | 35 | +| II.2 Example 2 ..... | 35 | +| Appendix III – Recommended test bed..... | 36 | +| III.1 Example 1: Implemented test hybrid ..... | 36 | +| III.2 Example 2: Adaptive Echo Canceller ..... | 36 | +| Appendix IV – Bibliography..... | 38 | + +# Introduction + +The object of this Recommendation is to obtain for hands-free terminals transmission performance equivalent with handset telephones, at least with respect to send and receive loudness. + +Other important features contributing to the quality of telephone calls made from hands-free terminals, as switching characteristics, duplex capability, are defined in this Recommendation. + +This Recommendation covers generic characteristics and requirements that are applicable to both analogue and digital hands-free terminals. Additional requirements that are applicable strictly to digital terminals can be found in ITU-T P.342 [3] and P.341 [4]. + +For loudspeaking telephones (see ITU-T P.10 [14]) which do not provide full hands-free operation, the relevant parts of this Recommendation may be referred to. + +Test methods appropriate for parameters defined in this Recommendation may be found in ITU-T P.502 [12]. + +For the use of HATS for testing, ITU-T P.581 [21] applies. + + + +# **Transmission characteristics and speech quality parameters of hands-free terminals** + +# **1 Introduction** + +The object of this Recommendation is to obtain for hands-free terminals transmission performance equivalent with handset telephones, at least with respect to send and receive loudness. + +Other important features contributing to the quality of telephone calls made from hands-free terminals, as switching characteristics, duplex capability, are defined in this Recommendation. + +This Recommendation covers generic characteristics and requirements that are applicable to both analogue and digital hands-free terminals. Additional requirements that are applicable strictly to digital terminals can be found in ITU-T P.342 [3] and P.341 [4]. + +For loudspeaking telephones (see ITU-T P.10 [14]) which do not provide full hands-free operation, the relevant parts of this Recommendation may be referred to. + +Test methods appropriate for parameters defined in this Recommendation may be found in ITU-T P.502 [12]. + +For the use of HATS for testing, ITU-T P.581 [21] applies. + +# **2 Normative references** + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision: all users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. + +- [1] ITU-T G.121 (1993), *Loudness Ratings (LRs) of national systems.* +- [2] CCITT G.223 (1988), *Assumptions for the calculation of noise on hypothetical reference circuits for telephony.* +- [3] ITU-T P.342 (2000), *Transmission characteristics for telephone band (300-3400 Hz) digital loudspeaking and hands-free telephony terminals.* +- [4] ITU-T P.341 (1998), *Transmission characteristics for wideband (150-7000 Hz) digital hands-free telephony terminals.* +- [5] ITU-T P.78 (1996), *Subjective testing method for determination of loudness ratings in accordance with Recommendation P.76.* +- [6] ITU-T P.79 (1999), *Calculation of loudness ratings for telephone sets.* +- [7] ITU-T P.50 (1999), *Artificial voices.* +- [8] ITU-T P.51 (1996), *Artificial mouth.* +- [9] ITU-T P.501 (2000), *Test signals for use in telephonometry.* +- [10] ITU-T G.167 (1993), *Acoustic echo controllers.* +- [11] ITU-T Handbook on Telephonometry, 1993. +- [12] ITU-T P.502 (2000), *Objective test methods for speech communication systems using complex test signals.* + +- [13] ITU-T P.832 (2000), *Subjective performance evaluation of hands-free terminals.* +- [14] ITU-T P.10 (1998), *Vocabulary of terms on telephone transmission quality and telephone sets.* +- [15] ITU-T G.114 (2000), *One-way transmission time.* +- [16] ITU-T G.131 (1996), *Control of talker echo.* +- [17] ITU-T G.174 (1994), *Transmission performance objectives for terrestrial digital wireless systems using portable terminals to access the PSTN.* +- [18] ITU-T G.122 (1993), *Influence of national systems on stability and talker echo in international connections.* +- [19] ITU-T P.800 (1996), *Methods for subjective determination of transmission quality.* +- [20] ITU-T G.168 (2000), *Digital network echo cancellers.* +- [21] ITU-T P.581 (2000), *Use of head and torso simulator (HATS) for hands-free terminal testing.* + +# 3 Definitions and abbreviations + +The relevant definitions given in ITU-T P.10 apply along with the following: + +**3.1 hands-free reference point (HFRP):** A point located on the axis of the artificial mouth, at 50 cm from the lip ring, where the level calibration is made, in free field. It corresponds to the measurement point 11, as defined in ITU-T P.51 [8]. + +**3.2 hands-free (telephone) set (HFT):** A telephone set using a loudspeaker associated with an amplifier as a telephone receiver and which may be used without a handset. + +**3.3 loudspeaking (telephone) set (LST):** A handset telephone using a loudspeaker associated with an amplifier as a telephone receiver. + +**3.4 acoustic echo canceller (AEC):** A device which reduces the acoustic echo level with negligible effects on the local and distant users' speech. It is generally implemented by adaptive identification of the acoustic echo path response. + +**3.5 loss controller:** A device which reduces the acoustic echo level by inserting variable losses on the received and/or transmitted audio signals. + +**3.6 non-linear processor:** A device which reduces or cancels small echo signals by non-linear operation on the samples of the transmitted audio signal. A centre-clipper is a typical device of this kind. + +**3.7 supplementary howling control device:** A device which modifies some characteristics of the transmitted and/or received signals in order to improve the stability margin of the terminal. This function is typically implemented by a harmonic processor. To prevent network disturbances, such devices should be avoided in the case of terminals likely to be used on connections including network electric echo cancellers conforming to ITU-T G.168 [20] which are not able to work properly with time-variant echo paths (e.g. frequency shift). + +**3.8 noise reduction system:** A device which increases the signal-to-noise ratio (on the overall or a specific range of bandwidth of the signal). + +This Recommendation uses the following abbreviations: + +| | | +|-------|----------------------------------| +| AEC | Acoustic echo canceller | +| AGC | Automatic gain control | +| $a_H$ | Insertion attenuation range (dB) | + +| | | +|-----------|----------------------------------------------------------| +| $a_{H,S}$ | Attenuation range (dB) in the sending path | +| $a_{H,R}$ | Attenuation range (dB) in the receiving path | +| Ardt | Received speech attenuation (dB) during double talk | +| Asdt | Sent speech attenuation (dB) during double talk | +| CSS | Composite Source Signal (see ITU-T P.501[9])) | +| Drdt | Received speech distortion during double talk | +| Dsdt | Sent speech distortion during double talk | +| EEB | Early Energy Balance | +| EPDn | Round trip echo path delay – network interface | +| ERL | Echo return loss | +| ERLst | Temporally weighted echo return loss – single talk | +| ERLdt | Temporally weighted echo return loss – double talk | +| HFT | hands-free terminal | +| HFRP | Hands-Free Reference Point | +| LOT | Listening Only Test (see ITU-T P.832 [13]) | +| MRP | Mouth Reference Point | +| MOS | Mean Opinion Score (see ITU-T P.800 [19]) | +| OLR | Overall Loudness Rating | +| RLR | Receiving Loudness Rating | +| SLR | Sending Loudness Rating | +| TCLwdt | Weighted terminal coupling loss – double talk | +| TCLwst | Weighted terminal coupling loss – single talk | +| TCLst | Terminal Coupling Loss temporally weighted – single talk | +| TCLdt | Terminal Coupling Loss temporally weighted – double talk | +| $T_c$ | Convergence time | +| $T_{ic}$ | Initial convergence time | +| $T_R$ | Build-up Time | +| $T_{Rdt}$ | Build-up time – double talk | +| $T_{Rst}$ | Build-up time – single talk | +| $T_H$ | Hang-over Time | +| $T_S$ | Switching Time | +| TUT | Terminal Under Test | +| $V_{TH}$ | Threshold Level | + +# 4 Transmission characteristics + +## 4.1 Sending sensitivity + +The Sending LR (SLR) of an HFT should be about 5 dB higher than the SLR of the corresponding handset telephone (the actual value will depend on the type of handset used). + +NOTE 1 – The average acoustic speech level from a terminal user is about 3 dB higher when using a HFT than when using a handset telephone. + +The output level from the handset telephone in conversational use is about 1-2 dB lower than what is obtained in the LRGP position specified for loudness ratings measurements of handset telephones. For HFTs there is, however, no such difference. + +NOTE 2 – Hands-free terminals having a sending sensitivity that complies with this Recommendation can be assumed to fulfil ITU-T G.223 [2]. + +Furthermore, in order to avoid excessive crosstalk from the high-level speech currents and/or inadequate received volume from low-level speech currents, care should be taken to ensure that the variation of speech currents is not substantially greater than that from handset telephones. + +NOTE 3 – Necessary precautions should be taken so that the terminal user may be able to break the sending circuit if oscillations occur, or to provide suitable methods so that a device controlled by the voice may prevent oscillations. + +## 4.2 Receiving sensitivity + +The receiving sensitivity of a hands-free telephone without automatic gain control should be adjustable within a range of 15 to 30 dB. This range should span the value of the Receiving Loudness Rating (RLR) which is equal to that of the corresponding handset telephone, as well as an RLR value about 10 dB lower. + +NOTE 1 – Every precaution should be taken to ensure that the increase in gain due to the volume control does not allow the overhearing of other telephone conversations due to crosstalk. + +NOTE 2 – In principle, the RLR of the HFT should be equal to the RLR of the corresponding handset telephone in a quiet room. The range of room noise levels met in normal office use necessitates, however, an additional gain of at least 10 dB. + +For hands-free terminals equipped with an automatic gain control for the receive level (the gain being controlled by the incoming speech voltage), loudness ratings may not be applicable. In this case, the HFT should be designed so that the listening level at the maximum of the Overall Loudness Rating (OLR) of the connection for which the HFT is intended to be used can be preset to a value that may be considered as the best compromise between the levels required for listening in quiet and noisy rooms. + +NOTE 3 – The preferred listening level depends on the room noise level as well as on other external conditions. There is, furthermore, a great variance between individual listeners. + +The average preferred level for listening only appears to be a sound pressure level of about –29 dBPa for –49 dBPa(A) room noise, or –24 dBPa for –39 dBPa(A) room noise. However, to obtain maximum mean opinion scores in conversation tests, listening levels of about 5 to 10 dB higher may be required. + +## 4.3 Frequency response curves + +### 4.3.1 Sending + +Available information indicates that the optimum slope of the sending response curve when measured with the HFT on a table lies between 0 and +3 dB/octave, if the receiving response curve is flat. + +Only under highly reverberant conditions may a somewhat higher pre-emphasis increase the intelligibility. Therefore, if a frequency equalization is used to compensate the cable attenuation of the analogue part of a connection, the sending curve should not rise with frequency by more than 2-3 dB/octave. + +Below 300 Hz there should be a gradual roll-off. The slope may be steeper below 200 Hz. + +NOTE – The interval 200-300 Hz makes a significant contribution to the naturalness of the transmitted speech and should therefore be included in the transmission band of the HFT. + +Above 4000 Hz, a roll-off by at least –6 dB/octave (preferably –12 dB/octave) is appropriate for analogue terminals in order to avoid interference by crosstalk to adjacent channels in certain types of long-distance circuits. + +If the analogue terminals are intended to be connected through short lines to a digital connection, it is appropriate that the sensitivity above 4 kHz be as low as possible, to avoid spurious out-of-band signals. + +For (300-3400 Hz) digital terminals, information is available in ITU-T P.342 [3]. + +For wideband digital terminals, information is available in ITU-T P.341 [4]. + +### **4.3.2 Receiving** + +The receiving response curve should be substantially flat in the frequency range of 200-4000 Hz. + +The requirement refers to the sound pressure in the undisturbed field at the listener's position with a setup including the table as described in clause 5. + +## **4.4 Subjective evaluations of loudness ratings** + +Loudness rating should be determined in accordance with ITU-T P.78 [5]. + +NOTE – Some information about reference equivalents can be found in the Handbook on Telephonometry. + +### **4.4.1 Sending** + +The talking level for the measurement of Sending Loudness Rating (SLR) of an HFT should normally be the same as specified for measurements on handset telephones. + +It is not necessary for the talker during the test to shift between the reference microphone guard-ring and the guard-ring positioned relative to the HFT if the obstacle effect of the reference microphone can be assumed to be negligible. + +Normally the specified talking level and the use of a conventional test phrase or sentence should be sufficient to ensure that a voice-switched HFT will be in the sending condition during the determination of SLR. If this is not the case, the talking level may be increased by up to 5 dB, which may be compensated in the reference system to preserve the same listening level. + +If the sending sensitivity is controlled by the room noise level, the subjective measurement should be done in a quiet environment [ $< -59$ dBPa(A)]. Further information about the HFT performance may then be estimated by repeating the sending measurements with increasing levels of room noise, up to a maximum of $-34$ dBPa(A). + +### **4.4.2 Receiving** + +The talking level at the reference microphone for the measurement of RLR should normally be the same as specified for the measurement of handset telephones. This should normally ensure that when loudness balance is achieved between the reference system and the test system path, a signal of sufficient magnitude is present at the HFT to switch it into the receive condition. + +Problems can sometimes occur when approaching the balance condition from the condition of high attenuation in the balance attenuators, when the low level input signal may fail to switch the HFT into the receiving condition. If this does occur, the talking level may be increased by up to 5 dB in order to minimize the difference in loudness. + +NOTE – The listening level will thus also increase at balance, but in this case it will not be possible to correct it by changing the reference system attenuator. + +Obtaining the loudness balance for the receiving condition may be facilitated by use of a loudspeaking intermediate reference system. The specification of such a system is, however, outside the scope of this Recommendation. + +## 4.5 Objective evaluations of sensitivity and loudness ratings + +### 4.5.1 Sensitivity measurement + +Objective evaluations of loudspeaker and hands-free terminals concern: + +- the sending and receiving frequency sensitivity curves measurements; +- the objective determination of loudness ratings according to the method described in ITU-T P.79 [6]. + +#### 4.5.1.1 Sending sensitivity measurements + +The sending response curves of a hands-free telephone is recorded at the output terminals of the telephone with the same electrical connections as for the handset telephones. The acoustical input to the telephone microphone is supplied from an artificial mouth in the position shown in Figure 3. + +In such a case, the sending sensitivity of the local telephone system $S_{mJ}$ is expressed as dB relative to 1 V (electrical interface)/Pa (MRP) as follows: + +$$S_{mJ} = 20 \log V_s - 20 \log P_{MRP} + \text{Corr} - 24$$ + +where: + +$V_s$ is the measured voltage across the appropriate termination (unless stated otherwise, a 600 ohms termination). + +$P_{MRP}$ is the applied sound pressure at the MRP. + +$\text{Corr}$ is $20 \log (P_{MRP}/P_{HFRP})$ of the used artificial mouth. + +NOTE – The value of $\text{Corr}$ is the value given in the calibration chart of the artificial mouth (24.0 dB is the ideal value). + +#### 4.5.1.2 Receiving sensitivity measurements + +The receiving sensitivity of a loudspeaker and/or hands-free telephone is expressed as follows: + +$$S_{Je} = 20 \log_{10} \frac{p_R}{(1/2)E_J} \text{ dB rel 1 Pa/V}$$ + +where: + +$p_R$ is the sound pressure at point C in Figure 3 and $E_J$ is the e.m.f. in the 600 ohms source. + +### 4.5.2 Measure and computation of loudness ratings + +#### 4.5.2.1 Sending loudness rating + +The computation of the sending loudness rating may be performed according to ITU-T P.79 [6] by using the frequency sensitivity response measured between the electrical output of the set and the acoustical sound pressure at the MRP. + +NOTE – Appropriate types of signals can be found in ITU-T P.501 [9]. + +#### 4.5.2.2 Receiving loudness rating + +Objective measurements described in 4.5.1.2 are made with a free field microphone at point C (see Figure 3). + +Loudness ratings are computed following ITU-T P.79 [6]. + +When calculating the RLR of a hands-free or loudspeaking telephone, a correction factor of –14 dB applies and the $L_e$ factor used in the algorithm of ITU-T P.79 [6] is equal to 0. + +### 4.6 Switching parameters + +Most loudspeaker and hands-free terminals contain voice-switched circuits whose main function is to avoid singing through acoustic feedback. Such circuits insert a loss in either the sending or receiving direction in various ways. Switching from one direction to the other occurs when a signal above a given threshold is applied from the opposite direction, or when the control circuit, taking into account the relative levels and the nature of the signals in both directions, allows the switching. + +The fundamental voice-switching parameters of the switching function are defined as follows (see Figures 1 and 2): + +- Threshold level $V_{TH}$ – Minimum necessary signal level for removing insertion loss. +- Build-up time $T_R$ – Time from the input signal going above the threshold level until the time at which the output level reaches 3 dB below complete removal of the insertion loss. +- Hang-over time $T_H$ – Time from the input signal going below the threshold level until 3 dB of the switched loss is inserted in the output signal. +- Switching time $T_S$ – Time from one transmission direction to the other. $T_S$ is measured from the removal of the signal in the first direction until the level in the second direction reaches 3 dB below its final value. + +By a suitable choice of parameter values, the degradation of speech quality that is introduced by voice switching can be made negligible, while an inadequate choice of parameter values, Switching Times in particular, may lead to serious clipping effects and loss of initial or final consonants in the transmitted speech. + +The following values or guidance are recommended: + +Threshold levels should be chosen so that switching is not interrupted by random (environmental) noise sources at either end of the call. In addition, ambient room/network noise effects on threshold should not impair performance. Ambient noise levels can be used to improve threshold performance, as talkers tend to speak louder in a noisy environment than in a quiet one. + +Build-up time should be short enough so that the initial transient components of speech are not lost, but not so short that insertion loss removal would be noisy. The build-up time $T_R$ should be less than 15 ms, preferably below 10 ms. + +Hang-over time should be long enough to cover average pauses in speech so that intermittent unwanted switching does not occur before the initial talker is finished, but short enough to allow for reasonable break-in from the second talker. The hang-over time $T_H$ is defined in Table 4. + +Switching time from one active state to the other should be balanced to best simulate full duplex operation. Switching time is also dependent on both build-up time and hang-over time. The switching time $T_S$ should be approximately 100 ms. + +Measurements of voice switching characteristics may be divided into those dealing with: + +- Characteristics for single talk situations, in which two parties communicate by alternating speech spurts without interrupting each other. In this case, it may be assumed that the voice switch circuit returns to an idle state before being activated by an input signal in either direction. +- Characteristics for double talk conversation, in which both parties may interrupt each other by simultaneous talk, or where speech at one end of a connection is present simultaneously with noise at the other end. + +If single talk situation is of fundamental importance, it has been shown that the behaviour of hands-free terminals in double talk situation affects seriously the overall quality. + +![Figure 1/P.340: Timing diagram for voice switching characteristics. The top part shows a step function representing the input signal, with a high state 'H' and a low state 'L'. A dashed horizontal line represents the threshold level V_TH. The bottom part shows the output signal, which is a smoothed version of the input. It shows a build-up time T_R from the threshold level to a 3 dB level, and a hang-over time T_H from the 3 dB level back to the threshold level. The output signal also shows a 3 dB level during the hang-over period. The diagram is labeled T1212540-00.](fc857414626a8d94d132e12d9afe52a4_img.jpg) + +Input + +Output + +$T_R$ Build-up Time + +$T_H$ Hang-over Time + +$V_{TH}$ Threshold Level + +T1212540-00 + +Figure 1/P.340: Timing diagram for voice switching characteristics. The top part shows a step function representing the input signal, with a high state 'H' and a low state 'L'. A dashed horizontal line represents the threshold level V\_TH. The bottom part shows the output signal, which is a smoothed version of the input. It shows a build-up time T\_R from the threshold level to a 3 dB level, and a hang-over time T\_H from the 3 dB level back to the threshold level. The output signal also shows a 3 dB level during the hang-over period. The diagram is labeled T1212540-00. + +Figure 1/P.340 + +![Timing diagram for HFT switching. Input 1 shows a high-to-low transition. Input 2 shows a low-to-high transition. Output 1 shows a high-to-low transition with switching time T_s and a 3 dB point. A threshold V_TH is indicated for the inputs.](1d529a819ad929684331c55eed6673bb_img.jpg) + +The figure illustrates the timing characteristics of an HFT. Input 1 shows a signal transitioning from a high state (H) to a low state (L) across a threshold $V_{TH}$ . Input 2 shows a signal transitioning from a low state (L) to a high state (H) across the same threshold $V_{TH}$ . Output 1 shows a signal transitioning from a high state to a low state. The switching time $T_s$ is defined as the time interval between the 3 dB point of the output transition and the corresponding input transition. The diagram is labeled T1212380-00. + +Timing diagram for HFT switching. Input 1 shows a high-to-low transition. Input 2 shows a low-to-high transition. Output 1 shows a high-to-low transition with switching time T\_s and a 3 dB point. A threshold V\_TH is indicated for the inputs. + +Figure 2/P.340 + +# 5 Testing conditions + +For both subjective and objective measurements, physical test arrangements as described in this clause should be used. + +### 5.1 Test table + +During the measurements, the HFT is placed on a table defined as follows: + +The surface of the table should be hard (e.g. polished marine plywood or suitable hardwood), flat, rigid and horizontal to provide a sound-reflecting surface on which the HFT being tested rests. The dimensions of the table should be such that the surface area is about 1 m2 but not less than 0.96 m2 and the width not less than 800 mm [1]. + +NOTE – This arrangement should be used for all measurements, including the recording of frequency responses, although diffraction effects due to the table are likely to cause severe dips or peaks in the response curve. + +## 5.2 Test arrangement + +The physical test arrangements of one piece (single unit) HFTs for subjective and objective measurements are shown in Figure 3. + +If the projections of the housing are not rectangular, the point B is positioned at the crossing of the centre line through the housing and the outline of the vertical projection of the housing. + +The edge of the front of the box should be perpendicular to the line A-B. + +![Figure 3/P.340 – Physical test arrangements for subjective and objective measurements. The diagram shows two test setups. The left setup is for subjective measurements, showing a 'double unit HFT' and a 'single unit HFT' positioned at a 48.6° angle from a central 'Centre line through housing'. The distance from the 'Front edge of test table' to the 'Position of single unit HFT' is 600 mm, and the distance from the 'Front edge of test table' to the 'Position of double unit HFT' is 400 mm. The right setup is for objective measurements, showing an 'Outline of projection of housing' and a 'Position of artificial mouth, measuring microphone, or lip ring' (point C) at a 90° angle from the 'Edge of the table'. The distance from the 'Edge of the table' to the 'Position of artificial mouth...' is 400 mm, and the distance from the 'Edge of the table' to the 'Table surface' is 300 mm. The distance from the 'Position of artificial mouth...' to the 'Table surface' is 500 mm. The diagram is labeled T1205840-93.](ff7ce44f3fdd51bae7b231f34df07c6a_img.jpg) + +Figure 3/P.340 – Physical test arrangements for subjective and objective measurements. The diagram shows two test setups. The left setup is for subjective measurements, showing a 'double unit HFT' and a 'single unit HFT' positioned at a 48.6° angle from a central 'Centre line through housing'. The distance from the 'Front edge of test table' to the 'Position of single unit HFT' is 600 mm, and the distance from the 'Front edge of test table' to the 'Position of double unit HFT' is 400 mm. The right setup is for objective measurements, showing an 'Outline of projection of housing' and a 'Position of artificial mouth, measuring microphone, or lip ring' (point C) at a 90° angle from the 'Edge of the table'. The distance from the 'Edge of the table' to the 'Position of artificial mouth...' is 400 mm, and the distance from the 'Edge of the table' to the 'Table surface' is 300 mm. The distance from the 'Position of artificial mouth...' to the 'Table surface' is 500 mm. The diagram is labeled T1205840-93. + +**Figure 3/P.340 – Physical test arrangements for subjective and objective measurements** + +For more-than-one-piece HFTs, videophony and multimedia terminals, the test arrangement shall be modified to what it is stated in the instruction manual of the Terminal Under Test (TUT). The positioning of the terminal shall be referred to the point C defined and located as in Figure 3. Sending and receiving loudness ratings, sending and receiving sensitivity responses shall be tested and adjusted for this test arrangement. The axis of the artificial mouth shall be defined by manufacturer's declaration. + +If there is no manufacturer's declaration, the test arrangement of Figure 3 shall be implemented. + +NOTE – When a HATS is used, the test arrangement defined in ITU-T P.581 [21] applies. + +## 5.3 Calibration of the artificial mouth + +ITU-T P.51 [8] defines the tolerances that apply to the acoustic radiation of the mouth. To reduce the possible errors that could be due to these tolerances, the following procedure shall be used to calibrate the artificial mouth. + +The spectrum of acoustic signal produced by the artificial mouth is calibrated under free field conditions at the MRP. + +The level of the acoustic signal is adjusted to $-4.7$ dBPa at the MRP. + +The spectrum at MRP is then recorded and used as a reference. + +![Figure 4/P.340 – Physical test arrangement for calibration of the artificial mouth. The diagram shows a signal chain: a waveform box connected to a 'Spectrum shaper' box (containing a dashed line graph), which is connected to an 'Amplifier' box (containing a triangle symbol). The amplifier is connected to an 'Artificial mouth' (represented by a trapezoid). A 'Lip ring' is indicated by an arrow pointing to the front of the artificial mouth. A distance of '50 cm' is marked between the lip ring and an 'HFRP' (Head Form Reference Point). The HFRP is connected to a 'Microphone' (represented by a cylinder with horizontal lines). The microphone is labeled 'T1207610-95'.](9c6461e1e94afae4dec455e69a2ce152_img.jpg) + +Figure 4/P.340 – Physical test arrangement for calibration of the artificial mouth. The diagram shows a signal chain: a waveform box connected to a 'Spectrum shaper' box (containing a dashed line graph), which is connected to an 'Amplifier' box (containing a triangle symbol). The amplifier is connected to an 'Artificial mouth' (represented by a trapezoid). A 'Lip ring' is indicated by an arrow pointing to the front of the artificial mouth. A distance of '50 cm' is marked between the lip ring and an 'HFRP' (Head Form Reference Point). The HFRP is connected to a 'Microphone' (represented by a cylinder with horizontal lines). The microphone is labeled 'T1207610-95'. + +**Figure 4/P.340 – Physical test arrangement for calibration of the artificial mouth** + +The broadband signal level then is adjusted to $-28.7$ dBPa at the HFRP and the spectrum is not altered. + +The spectrum at the MRP and the actual level at the MRP (measured in third octaves) is used as reference to determine the sending sensitivity $S_{mj}$ (see Figure 4). + +NOTE – When a HATS is used, the calibration defined in ITU-T P.581 [21] applies. + +## 5.4 Test environment + +- 1) For the repeatability of the tests, the environment for most of the measurements shall be free field (anechoic) down to the lowest frequency of the 1/3 octave band centred on 200 Hz. + +Satisfactory free field conditions exist where errors, due to the departure from ideal conditions, do not exceed the values defined in Table 1, inside a sphere centred on point B (see Figure 5), with one metre radius, in absence of the table. + +**Table 1/P.340** + +| 1/3 octave band centre frequency (Hz) | Allowable departure (dB) | +|---------------------------------------|--------------------------| +| <630 | $\pm 1.5$ | +| 800 to 5000 | $\pm 1$ | +| >6300 | $\pm 1.5$ | + +The test signal level for verification of the free field is $-20$ dBPa. + +Measurements are made along the seven axes which are numbered ① to ⑦ in Figure 5, with the sound source placed at positions equivalent to B or C, as appropriate. Measurement points along each axis, taken from the front plan of the artificial mouth lip-ring are at the distances of 315 mm, 400 mm, 500 mm, 630 mm, 800 mm and 1000 mm. + +NOTE – Complementary information for test rooms is available in 10.3.1. + +- 2) The broadband noise level shall not exceed $-70$ dBPa(A). The octave band noise level shall not exceed the values specified in Table 2. + +![Three diagrams illustrating the determination of free field conditions. The top diagram is a 'Plan view' showing a square table with a dashed horizontal and vertical line intersecting at a central point labeled 'B'. Four arrows, numbered 1, 2, 3, and 4 in circles, originate from point B and point towards the corners. A label 'Virtual point corresponding to centrepont of the table' points to B. The middle diagram is 'End view (1)' showing a horizontal line for the 'Upper table surface'. A vertical dimension of 300 mm is shown from the surface to point 'C', which is labeled 'Lip ring'. A horizontal dimension of 800 mm is shown from the left edge to point 'C'. A diagonal dimension of 500 mm is shown from point 'C' to point '7' in a circle. A label 'Point C corresponds to the normal' is present. The bottom diagram is 'End view (2)' showing a horizontal line for the 'Upper table surface'. A vertical dimension of 300 mm is shown from the surface to point '6' in a circle. A horizontal dimension of 400 mm is shown from point 'B' to the base of the 300 mm dimension. An arrow numbered 5 in a circle points vertically upwards from point 'B'.](b6671cfafda3820aafe9a24fa7a4d8c7_img.jpg) + +Three diagrams illustrating the determination of free field conditions. The top diagram is a 'Plan view' showing a square table with a dashed horizontal and vertical line intersecting at a central point labeled 'B'. Four arrows, numbered 1, 2, 3, and 4 in circles, originate from point B and point towards the corners. A label 'Virtual point corresponding to centrepont of the table' points to B. The middle diagram is 'End view (1)' showing a horizontal line for the 'Upper table surface'. A vertical dimension of 300 mm is shown from the surface to point 'C', which is labeled 'Lip ring'. A horizontal dimension of 800 mm is shown from the left edge to point 'C'. A diagonal dimension of 500 mm is shown from point 'C' to point '7' in a circle. A label 'Point C corresponds to the normal' is present. The bottom diagram is 'End view (2)' showing a horizontal line for the 'Upper table surface'. A vertical dimension of 300 mm is shown from the surface to point '6' in a circle. A horizontal dimension of 400 mm is shown from point 'B' to the base of the 300 mm dimension. An arrow numbered 5 in a circle points vertically upwards from point 'B'. + +T1207620-95 + +Dimensions in millimetres. + +Points 1, 2, 3 and 4 are in the horizontal plane normally occupied by the table surface. + +Measurements of free field sound pressure are made in absence of the table. + +Axes used in the determination of free field conditions for 1 m radius sphere. + +**Figure 5/P.340 – Determination of the free field conditions** + +**Table 2/P.340 – Noise level** + +| Centre frequency (Hz) | Octave band pressure level (dBPa) | +|-----------------------|-----------------------------------| +| 63 | -45 | +| 125 | -60 | +| 250 | -65 | +| 500 | -65 | +| 1 k | -65 | +| 2 k | -65 | +| 4 k | -65 | +| 8 k | -65 | + +## 5.5 Test signals + +The test signals shall comply with the contents of ITU-T P.501 [9]. + +## 5.6 Test signal levels + +Measurement level, unless otherwise specified, corresponds to the standard level. This is -28.7 dBPa at HFRP for send, -30 dBm0 for receive. + +## 6 Conversational quality + +In the case of many of the hands-free terminals which have been hitherto available, the advantages over handsets have usually meant a sacrifice in terms of a reduction in speech transmission quality. The parameters relevant to a quality description were obtained from listening tests. The instrument-based parameters to be correlated were extracted. Hands-free terminals are not linear, time-invariant systems, and newly developed measurement engineering techniques for measuring transmission response must be additionally applied. A comprehensive definition of quality for hands-free terminals requires the adaptation or extension of current test measurements. + +Subjective evaluation for hands-free terminals are defined in ITU-T P.832 [13]. + +The parameters "background noise transmission" and "duplex capability" are the most important auditive characteristics and are crucial in deciding the subjective impression of quality of hands-free equipment. The most important value measurable in terms of instrumentation, correlating to these auditive values, is, in the case of duplex capability, the attenuation range, and for background noise transmission the attenuation distribution at idle. These parameters provide a first indication of the product quality of hands-free equipment and allow a classification of the equipment. Three types can be included in such a classification: + +### 1) *Full duplex capability* + +In duplex operation, attenuation of the conversation partner either does not occur or is unnoticeable. Background noise is transmitted in the send path (with or without room noise reduction). + +### 2) *Partial duplex capability* + +In duplex operation, one or both of the speech paths is attenuated. The conversation partner and background noise in the send path are still audible. + +### 3) *No duplex capability* + +When one conversation partner is talking, the other is fully attenuated. Background noise in the send direction is not transmitted. + +# 7 Parameters measured instrumentally + +Besides the transmission parameters defined in the preceding subclauses, the following objective parameters influence significantly the overall quality of the communication using hands-free terminals. + +- attenuation range, $a_H$ ; +- attenuation distribution at idle mode; +- hang-over time; +- dynamic compression; +- reverberance: impulse response, EEB; +- time-dependent frequency response; +- duplex behaviour; +- echo cancellation (see clause 10); +- quality of speech recognition. + +The appropriate test procedures are defined in ITU-T P.502 [12]. + +### 7.1 Attenuation range $a_H$ + +Attenuation range is determined by the difference in sensitivity response which results when one speech path is activated and when the duplex branch is activated. An activating signal (e.g. voiced sound of CSS) is used as the receive measurement signal, immediately followed by an activating signal in the send direction. In the analysis the level of the measurement output signal is represented versus time (time constant = 5 ms). The attenuation range is obtained from the difference between the maximum level at full activation and the minimum level obtained immediately after switch-over. + +## 7.2 Attenuation distribution at idle mode + +The attenuation range in idle is derived from the sum of the maximum attenuation in the send direction $a_{H,S}$ and the maximum attenuation in the receive direction $a_{H,R}$ . In the case of a system with no compressor, $a_H = a_{H,S} + a_{H,R} = \text{constant}$ applies. (In the case of a system including a compressor, the attenuation at maximum and minimum level due to the compressor must be obtained, see 7.4.) + +For measuring attenuation distribution a signal can be used which consists of a train of activating signals (e.g. voiced sound of CSS) and pauses in both speech directions, whereby extremely large variation in transition combinations are possible. The attenuation value can thus be obtained at the appropriate transition points from idle to activation for each of the speech paths. Again, as in 7.1, measurement signal level is represented versus time. + +Attenuation in the send direction $a_{H,S}$ can be read off from idle to send at the transition point, and is derived from the difference between the maximum level at full activation in the send direction and the minimum level obtained immediately after the send activation signal is present. The attenuation $a_{H,R}$ can be correspondingly read off at the transition idle/receive activation signal. + +If the switch-on time of the equipment is very fast, thus making reading off difficult, the measurement technique described in 7.3 (Hang-over time) can be used. The attenuation $a_{H,S}$ or $a_{H,R}$ of each speech path can be derived by calculating the difference in level at maximum and minimum attenuation in the second part of the signal (noise signal). + +## 7.3 Hang-over time + +The transition from activation to idle can be represented by feeding in an activation signal (e.g. voiced sound of CSS) in one direction, followed by a second signal in the same direction but of lower level, which does not activate the hands-free telephone (noise signal). The second part of the signal measured thus indicates the attenuation, from which the Hang-over time (switch-off time) can be determined. + +The duration of the voiced sound is 0.5 s in order to reach a final stable system condition. The level corresponds to standard level, as defined in 5.6. The second part of the signal (noise signal) has a duration of 1 s. The level must be selected low enough so as not to activate the equipment. The suggested levels to be applied are: $-58.7$ dBPa at HFRP for send, and $-60$ dBm0 for receive. + +## 7.4 Dynamic compression + +A companding or AGC response (e.g. the range of level adjustments as a function of input signal level) can be measured using an activation signal with steeply monotone rising or falling level response. The compressor is active during the period where the output signal has a constant level. The level limits within which the compressor operates can be obtained from the original signal. The compression range is obtained as the difference between the level limits. The attack time should be rather short and the release time should be sufficiently long. + +Appendix I gives some more information. + +Using a test signal with level steps, it is possible to determine time response of the companding or AGC (e.g. time duration for level adjustments). + +## 7.5 Reverberance + +The impulse response in the send direction is measured using maximum-length sequences, as described in [IV.5] of Appendix IV. The test signal should be composed of an activation signal (e.g. voiced sound of the CSS) and a maximal-length sequence of several periods (one segment is shown in Figure 6). The period length must be longer than the length at the impulse response of the system under test. The impulse response is calculated from the average of several periods, beginning with the second period. From this the Early Energy Balance (EEB) is calculated. The EEB represents fairly well the subjective impression of reverberance ([IV.6] of Appendix IV): + +$$EBB = 10 \cdot \log \left\{ \frac{\int_0^{35ms} h^2(t) dt}{\int_0^{5ms} h^2(t) dt} \right\} dB$$ + +![Figure 6/P.340 – Segment of maximum-length sequence. The graph shows a signal s(t)/% on the y-axis (ranging from -100 to 50) versus time t/ms on the x-axis (ranging from 0.0 to 5.0). The signal is a binary sequence alternating between approximately 50% and -50%.](df0685d2d1176d617ed1e642de4e5425_img.jpg) + +The figure is a plot of a signal $s(t)/\%$ over a time interval $t/ms$ . The y-axis ranges from -100 to 50, with major ticks at -100, -50, 0, and 50. The x-axis ranges from 0.0 to 5.0, with major ticks at 0.0, 1.0, 2.0, 3.0, 4.0, and 5.0. The signal is a binary sequence that alternates between approximately 50% and -50%. The sequence is periodic, with a period of approximately 0.5 ms. The signal is labeled $s(t)/\%$ and the time axis is labeled $t/ms$ . A small label $T1207680-95$ is present in the bottom right corner of the plot area. + +Figure 6/P.340 – Segment of maximum-length sequence. The graph shows a signal s(t)/% on the y-axis (ranging from -100 to 50) versus time t/ms on the x-axis (ranging from 0.0 to 5.0). The signal is a binary sequence alternating between approximately 50% and -50%. + +**Figure 6/P.340 – Segment of maximum-length sequence** + +## 7.6 Time-dependent transfer function + +In analysis, the output signals are referred to the corresponding original signals. For display of the time and frequency dependent structure of the output signal referred to the input signal, a spectrographic display may be selected where the x-axis represents time, the y-axis represents the frequency and the level is represented in different colours. In this way, any possible time or frequency variant structures can be revealed. + +## 7.7 Duplex behaviour + +Special duplex speech signals can be fed in both directions (send and receive) for investigating the duplex response. Appropriate signals are those of differently varying level, (e.g. speech sequence of fixed definition or sequences from CSS of fixed definition) fed in after each other with varying periodicity in the send and receive directions. + +A double talk sequence as described in ITU-T P.501 [9] should be used. The corresponding analysis techniques are given in ITU-T P.502 [12]. + +The output signals are referred to the corresponding original signals and spectrographically displayed as described in 7.6. This allows to represent how often gaps or signal interrupting occur during transmission. Conclusions can thus be reached regarding the duplex capability of the equipment. + +The duplex behaviour of a hands-free terminal is detailed in clause 9. + +## **7.8 Control characteristics of echo cancellation** + +Subjective quality is determined by the following parameters: + +- time-variable echo attenuation; +- echo attenuation over frequency; +- time response of NLP (Non-linear Processor), centre clipper; +- response in duplex operation; +- attenuation response with background noise. + +Clause 10 gives more details on these characteristics. + +## **7.9 Speech detection** + +Speech detection is measured in the send direction. From subjective listening and conversation tests ([IV.8] of Appendix IV), speech detection in the sending path appears as the more critical of the two transmission paths. + +For terminals incorporating level detection, the level measured is the minimum at which the equipment switches through. + +NOTE – Other levels and types of background noise may be used depending on the application and environment where the HFT is used. + +## **7.10 Speech detection with ambient noise** + +Measurement is done in the send direction. A 10 s sequence of a complex signal should be used. + +A minimum of four loudspeakers, fed with non-coherent noise signals, are simulating a room noise. They are positioned according to Figure 7. The resulting signal is measured at the point located in the middle of the test table (in absence of the table), the spectrum is a Hoth noise ([IV.7] of Appendix IV), and the level is set to $-44$ dBPa(A). + +![Figure 7/P.340 – Positioning of the equipment. The diagram consists of two parts: a top view and a side view. The top view shows a square room with side lengths of 1 m. A 'Hands-free telephone' is positioned at the center, marked by a solid dot. Four 'artificial mouth or measurement microphone' units, marked by circles with an 'X', are placed at the corners of the square. Diagonal lines from the corners to the center are labeled 1 m. A horizontal line from the left wall to the center is labeled 50 cm. The bottom-left corner unit is labeled 'positioning (top view)'. The side view shows the 'Hands-free telephone' on a 'Table'. An 'artificial mouth or measurement microphone' is also on the table, 30 cm from the telephone. A 'Loudspeaker for room noise simulation' is positioned 50 cm above the table. The telephone is marked with a solid dot, and the artificial mouth is marked with a circle and an 'X'. The label 'positioning (side view)' is at the bottom left.](cbdfdade780e677eb1c1aef3081ce9ef_img.jpg) + +Figure 7/P.340 – Positioning of the equipment. The diagram consists of two parts: a top view and a side view. The top view shows a square room with side lengths of 1 m. A 'Hands-free telephone' is positioned at the center, marked by a solid dot. Four 'artificial mouth or measurement microphone' units, marked by circles with an 'X', are placed at the corners of the square. Diagonal lines from the corners to the center are labeled 1 m. A horizontal line from the left wall to the center is labeled 50 cm. The bottom-left corner unit is labeled 'positioning (top view)'. The side view shows the 'Hands-free telephone' on a 'Table'. An 'artificial mouth or measurement microphone' is also on the table, 30 cm from the telephone. A 'Loudspeaker for room noise simulation' is positioned 50 cm above the table. The telephone is marked with a solid dot, and the artificial mouth is marked with a circle and an 'X'. The label 'positioning (side view)' is at the bottom left. + +**Figure 7/P.340 – Positioning of the equipment** + +## 7.11 Background noise transmission + +The sources of the transmitted noise may be: + +### – *Acoustical sources* + +Ambient noise in the test room (typically as defined in 5.4. If other situations are met, information should be given in the testing report); + +Noise intentionally produced in the test room (e.g. babble noise, car noise, etc.). + +### – *Electrical sources* + +Noise generated by the components of the terminals or/and network; + +Noise intentionally generated (e.g. comfort noise, etc.). + +Background noise transmission refers to the "noise intentionally produced in the test room" picked up by the microphone(s) of the hands-free terminal. See Table 3. + +**Table 3/P.340 – Parameters associated to background noise transmission, impacting on the quality** + +| Parameters | Examples of recommendations | +|-----------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| Absolute level | The transmitted background noise level should be as low as possible.
The background noise transmitted signal should not be interrupted from time to time. | +| Level fluctuations | The level fluctuations should not be more than $\pm 3$ dB, compared to steady state conditions.
NOTE – This does not apply if the residual background transmitted noise is masked (e.g. by the speech signal). | +| Additional parameters | Artefacts of noise reduction algorithms need to be avoided. | + +# 8 Behaviour of hands-free terminals in double talk situation + +## 8.1 Categorization + +Hands-free terminal should be categorized on the basis of "duplex capability". From [IV.9] and [IV.12] of Appendix IV, it appears that these characteristics are the most important on the perceived quality of the HFT, and that they may be characterized by the attenuation range. + +So, the following categories, numbered 1, 2 (2a, 2b, 2c) and 3, associate the behaviour of the hands-free terminal and the attenuation range. See Table 4. + +**Table 4/P.340** + +| Sending direction | Receiving direction | +|--------------------------------------------------------------|--------------------------------------------------------------| +| Behaviour 1: $a_{H,S,DT} \leq 3$ dB | Behaviour 1: $a_{H,R,DT} \leq 3$ dB | +| Behaviour 2a: $3 \text{ dB} < a_{H,S,DT} \leq 6 \text{ dB}$ | Behaviour 2a: $3 \text{ dB} < a_{H,R,DT} \leq 5 \text{ dB}$ | +| Behaviour 2b: $6 \text{ dB} < a_{H,S,DT} \leq 9 \text{ dB}$ | Behaviour 2b: $5 \text{ dB} < a_{H,R,DT} \leq 8 \text{ dB}$ | +| Behaviour 2c: $9 \text{ dB} < a_{H,S,DT} \leq 12 \text{ dB}$ | Behaviour 2c: $8 \text{ dB} < a_{H,R,DT} \leq 10 \text{ dB}$ | +| Behaviour 3: $a_{H,S,DT} > 12 \text{ dB}$ | Behaviour 3: $a_{H,R,DT} > 10 \text{ dB}$ | + +Annex A gives information on the categories and the test results used for their definition. + +## 8.2 Terminal coupling loss in double talk situation + +From the subjective tests it appears that the echo loss is also a parameter with a great influence on the quality perceived in double talk situation. + +From the tests described in Annex A, the following categorization has been established for echo loss in double talk: + +Behaviour 1: $TELRDT \geq 37 \text{ dB}$ + +Behaviour 2a: $37 \text{ dB} > TELRDT \geq 33 \text{ dB}$ + +Behaviour 2b: $33 \text{ dB} > TELRDT \geq 27 \text{ dB}$ + +Behaviour 2c: $27 \text{ dB} > TELRDT \geq 21 \text{ dB}$ + +Behaviour 3: $TELRDT < 21 \text{ dB}$ + +# 9 Parameters to be evaluated for each type of HFT + +The requirement of good duplex capability, the quality of room noise transmission in the send direction and other auditory quality factors result in specific requirements which should be met for each group. Different parameters are given priority in each type of equipment, according to the different quality level attaching to each type, each of which can be measured to different measurement regulations. + +Table 5 defines some parameters, guidance and limits. Complementary parameters and limits are given for information in Appendix I. + +**Table 5/P.340** + +| Behaviour 1
full duplex capability | Behaviour 2
partial duplex capability | Behaviour 3
no duplex capability | +|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------| +| Reverberance
The EEB value should be as small as possible. | Reverberance
The EEB value should be as small as possible. | Reverberance
The EEB value should be as small as possible. | +| Delay
Extra delay for processing is defined in clause 10. | Delay
Some extra delay for processing could be accepted; the maximum value is defined in clause 10. | Delay
No extra delay for processing. | +| Time and frequency variable transfer function
The transfer function is measured according to 7.6 and spectrographically displayed. The transfer function should show a constant frequency dependent response over time within $\pm 3$ dB, measured in 1/12 octaves.
In the case of time and frequency variable response, auditory evaluation of the system is required. | | | +| | Quality of speech detection in the send direction
The minimum threshold level should be $< -30$ dBPa at HFRP. | Quality of speech detection in the send direction
The minimum threshold level should be $< -30$ dBPa at HFRP. | +| | Hang-over time
Requirement: $T_H$ should be greater than 50 ms, preferably more than 100 ms. | Hang-over time
Requirement: $T_H$ should be greater than 250 ms. | + +**Table 5/P.340 (concluded)** + +| Behaviour 1
full duplex capability | Behaviour 2
partial duplex capability | Behaviour 3
no duplex capability | +|-----------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| | Dynamic compression
Control through the use of a compressor or AGC should be undetectable. The compressor control range should be small. | Dynamic compression
In the case of hands-free terminals without reliable noise detection, noise control through a compressor should not be annoying. | +| | | Room noise detection
Room noise detection must be reliable. Hoth noise ([IV.7] of Appendix IV) with a level $L_N$ at the hands-free equipment (HFRP) is generated. No accidental switch through the hands-free equipment should be registered at a level of $L_N = -50$ dBPa(A) over a period of 10 s. | + +# **10 Acoustic echo controllers and speech enhancement devices** + +## **10.1 Functional units** + +The functional units of an AEC are devices or parts of devices implemented in the processing unit, which contribute to the general function of acoustic echo control. There is no restriction on how to implement them. + +The functional units can be combined for better performance. They can use all the available signals in the terminal (for example, the individual signals coming from several microphones arranged in an acoustic array). Moreover, they can be mixed with other functions (for example sub-band speech coding) for efficient implementation, provided that they do not modify the proper characteristics of these functions when they are operating. + +A functional block-diagram of a typical processing unit is shown in Figure 8. + +![Functional block-diagram of a typical processing unit (AEC part). The diagram shows two main signal paths. The top path starts at 'From S_in', passes through a summing junction (with '+' and '-' inputs), then through 'Variable loss', 'Howling control device', and 'Non-linear processor' to 'To network interface'. The bottom path starts at 'From network interface', passes through 'Variable loss', then through a summing junction (with '+' and '-' inputs) to 'To R_out'. Control blocks include 'Acoustic echo estimator and other control circuit' and 'Loss controller'. Bypass signal paths (bp) are shown for testing purposes, connecting the summing junctions to the network interface and the output.](11edb7fcedf09ac6a817f8d7b8c61eec_img.jpg) + +The diagram illustrates the functional block-diagram of a typical processing unit (AEC part). It consists of two main signal paths and several control blocks. The top path represents the signal from the input (Sin) to the network interface. It starts with a summing junction where the input signal is combined with a feedback signal. This is followed by a 'Variable loss' block, a 'Howling control device', and a 'Non-linear processor'. The bottom path represents the signal from the network interface to the output (Rout). It starts with a 'Variable loss' block, followed by another summing junction. Control blocks include an 'Acoustic echo estimator and other control circuit' and a 'Loss controller'. Bypass signal paths (bp) are shown for testing purposes, connecting the summing junctions to the network interface and the output. The diagram is labeled T1507370-92. + +Functional block-diagram of a typical processing unit (AEC part). The diagram shows two main signal paths. The top path starts at 'From S\_in', passes through a summing junction (with '+' and '-' inputs), then through 'Variable loss', 'Howling control device', and 'Non-linear processor' to 'To network interface'. The bottom path starts at 'From network interface', passes through 'Variable loss', then through a summing junction (with '+' and '-' inputs) to 'To R\_out'. Control blocks include 'Acoustic echo estimator and other control circuit' and 'Loss controller'. Bypass signal paths (bp) are shown for testing purposes, connecting the summing junctions to the network interface and the output. + +**Figure 8/P.340 – Functional block-diagram of a typical processing unit (AEC part)** +(bp denotes bypass signal paths for testing purposes) + +NOTE – Interactions between terminal AEC and network signal processing equipment are considered in I.6.6 of Appendix I/G.168 [20]. + +## 10.2 Delay + +The values specified below correspond to the extra delay which can result from the AEC and noise reduction systems processing. In any case, compliance with transmission planning objectives must be achieved. + +General information about transmission delays can be found in ITU-T G.114 [15]; ITU-T G.131 [16] provides rules for echo control in the network. + +The sum group delays for mobile terminals, from the mouth reference point to the network interface and from the network interface to the ear reference point, of less than 20 ms delay is desirable (ITU-T G.174 [17]). + +### 10.2.1 Processing delay + +The echo cancellation and/or the noise reduction processes need some time. This time creates delay in the terminal, called below as "processing delay". + +Examples of values for mobile hands-free terminal are given in Appendix II. + +### 10.2.2 Round trip echo path delay (EPDn) – network interface + +Echo audibility is dependent upon the round trip delay in echo path. This requires calculation of impulse response (echo path). To that end, stationary broadband noise should be used. The terminal can be placed in nearly any reverberant or non-reverberant environment, as the first acoustic echo will be due to direct coupling. Acoustic noise should meet the requirements defined in 5.4. AEC is first trained using a signal at Rin. + +## 10.3 Acoustic echo control specifications + +Performance requirements can also be found in clause 8 and test methods in ITU-T P.502 [12]. + +### 10.3.1 Acoustic echo path + +The use of real rooms or enclosures with appropriate acoustic characteristics is recommended. Echo paths simulated by electronic devices like digital reverberators with non time-varying reflection patterns can be used as well if the terminal has internal accesses on the user side. In this latter case, the electronic simulator adjustments should comply with the values recommended for real rooms or enclosures; moreover, the shape of the simulated impulse response envelope should be similar to the real echo path impulse response. + +- For teleconference systems, the reverberation time averaged over the transmission bandwidth shall be typically 400 ms; the reverberation time in the lowest octave shall be no more than twice this average value; the reverberation time in the highest octave shall be not less than half this value. The volume of a typical test room shall be of the order of 90 m3. +- For hands-free terminals and videophones, the reverberation time averaged over the transmission bandwidth shall be typically 500 ms; the reverberation time in the lowest octave shall be no more than twice this average value; the reverberation time in the highest octave shall be not less than half this value. The volume of a typical test room shall be of the order of 50 m3. +- For mobile radio terminals an enclosure simulating the interior of a car can be used; a real car can be used as well. A typical average "reverberation time" is 60 ms. The volume of the enclosure shall be of the order of 2.5 m3. + +NOTE – It is recommended to avoid extremely long rooms (Length >> Width, Height) and rooms with extremely low ceilings (Height << Length, Width), and preferably also rooms with all the side dimensions nearly identical. + +Large, flat, parallel room limiting surfaces and surface areas that provide broadband sound reflection, particularly wall surfaces at an average room height (roughly 0.8 m to 1.8 m above the floor) should be avoided, since they can cause flutter echoes and flutter-echo-like disturbances (echoing, roughness), if the test setup is in an unfavourable position. + +Measuring the local frequency-dependent distribution of sound pressure levels within a selected room in the steady state can help to determine the optimum position of the test setup. + +As a general suggestion, the minimum distance between the test setup and room limiting surfaces should be 1 m, regardless of the acoustic properties of these surfaces. This can prevent disturbances due to initial reflections and a rise in sound pressure level that can occur locally at low frequencies. The same recommendation applies to geometrically large furniture surfaces that reflect sound. + +### 10.3.2 Parameters and recommended limits + +#### 10.3.2.1 Weighted terminal coupling loss – single talk (TCLwst) + +The weighted loss between the $R_{in}$ and $S_{out}$ network interfaces when the AEC is in normal operation, and when there is no signal coming from the local user1. + +Before each test the terminal is switched on. + +--- + +1 The weighting is made according to the rule specified in ITU-T G.122 (computation of talker echo loudness rating). Care must be taken to avoid possible masking of singing effects by the weighting (under study). + +The recommended values for each type of hands-free terminal can be found in the relevant ITU-T Recommendations (e.g. P.341 for wideband hands-free terminal, P.342 for digital hands-free terminal). + +#### **10.3.2.2 Weighted terminal coupling loss – double talk (TCLwdt)** + +The weighted loss between the $R_{in}$ and $S_{out}$ network interfaces when the AEC is in normal operation, and where the local user and the far-end user are active simultaneously1. + +The recommended values for each type of hands-free terminal can be found in clause 8. + +#### **10.3.2.3 Received speech attenuation during double talk (Ardt)** + +The received signal attenuation (at the $R_{out}$ point) which is inserted by the AEC during double-talk events. + +The frequency response on the receive side during double talk is under study. + +#### **10.3.2.4 Sent speech attenuation during double talk (Asdt)** + +The sent signal attenuation (at the $S_{out}$ point) which is inserted by the AEC during double-talk events. + +The frequency response on the send side during double talk is under study. + +#### **10.3.2.5 Received speech distortion during double talk (Drdt)** + +The total non-linear signal distortion at the $R_{out}$ point which can be produced by the AEC during double-talk events. + +For all the applications, the supplementary distortion at $R_{out}$ in comparison with single talk conditions should be low. The values of Drdt are under study. + +#### **10.3.2.6 Sent speech distortion during double talk (Dsdt)** + +The total non-linear signal distortion at the $S_{out}$ point which can be produced by the AEC during double-talk events. + +For all the applications, the supplementary distortion at $S_{out}$ in comparison with single talk conditions should be low. The values of Dsdt are under study. + +#### **10.3.2.7 Build-up time – single talk (TRnst)** + +The time interval between the onset of the received signal (similarly the transmitted signal) and the instant when the attenuation on the receive path (similarly on the send path) reaches [3] dB. For this purpose, the other side is quiet. + +##### **10.3.2.7.1 Receive side (TRst-r)** + +For all the applications, TRst\_r shall be no more than [20 ms]. + +##### **10.3.2.7.2 Send side (TRst-s)** + +For all the applications, TRst\_s shall be no more than [20 ms]. + +#### **10.3.2.8 Build-up time – double talk (TRdt)** + +The time interval between the onset of the received signal (similarly the sent signal) and the instant when the attenuation on the receive path (similarly on the send path) reaches the value Ardt (similarly Asdt). For this purpose, the signal in the opposite direction of transmission is held at a specified level. + +##### **10.3.2.8.1 Receive side (TRdt-r)** + +TRdt-r should be less than [20 ms], if the attenuation is more than 6 dB. + +##### **10.3.2.8.2 Send side (Tondt-s)** + +TRdt-s should be less than [20 ms], if the attenuation is more than 6 dB. + +#### **10.3.2.9 Convergence time (Tc) – Initial Convergence Time (Tic)** + +Convergence Time is the time interval between the instant when a specified test signal is applied to the $R_{in}$ port of the terminal (after all the functions of the AEC have been reset and then enabled), and the instant when the returned echo signal at the $S_{out}$ port is attenuated by at least a pre-defined amount. The local user is not active. + +Initial Convergence Time is the Convergence Time evaluated after switching on the HFT. + +#### **10.3.2.10 Hang-over Time after double talk (THdt)** + +The time elapsed between the end of a double talk event and the instant when the attenuation of the echo recovers a specified value (a signal is received continuously from the distant user). + +For all the applications, the attenuation of the signal at $S_{out}$ should be at least [20 dB] after THdt = [1] second. + +#### **10.3.2.11 Terminal coupling loss temporally weighted – single talk (TCLst)** + +The echo return loss from $R_{in}$ to $S_{out}$ is measured according to the procedure defined for ERLst in ITU-T Recommendation P.502 [12]. + +#### **10.3.2.12 Terminal coupling loss temporally weighted echo return loss – double talk (ERLdt)** + +The echo return loss from $R_{in}$ to $S_{out}$ is measured according to the procedure described for ERLdt in ITU-T Recommendation P.502 [12]. + +# **ANNEX A** + +## **Test results** + +The test results presented in this annex are extracted from [IV.10] and [IV.11] of Appendix IV. + +## **A.1 Categorization of hands-free terminals based on double talk performance** + +The results on attenuations in sending and receiving directions (Figures A.1, A.2) and on echo disturbance (Figures A.3, A.4 and A.5) have been determined from Third Party Listening Test results. + +The hands-free terminal is simulated. + +NOTE – It should be noted that the investigation to assess the echo disturbances during double talk are based on a one-way delay of 100 ms. + +### **A.1.1 Results of the listening only tests with variable attenuations in sending path** + +The test parameter is attenuation (dB) in the sending path. Long double talk sequences are used for this test. The tests have been made for different values of the attenuation, and the curves of Figure A.1 have been interpolated from these results. A long double talk sequence is judged. + +The upper curve shows the perceived quality (in MOS) on the loudness variation scale, the lower curve the overall quality. + +Figure A.1 shows the perceived quality (on a MOS scale) test conditions listening only tests. + +![Line graph titled 'overall quality and loudness variation SND' showing MOS vs. Attenuation in sending path (dB). The upper solid line represents 'interpolation loudness variation' and the lower dashed line represents 'interpolation overall quality'. Both curves decrease as attenuation increases from 'ref' to 24 dB.](fae82236e4211f753df5789eb276d3a4_img.jpg) + +The graph shows two curves representing perceived quality (MOS) as a function of attenuation in the sending path (dB). The x-axis ranges from 'ref' to 24 dB, and the y-axis ranges from 1.0 to 5.0 MOS. The upper curve (solid line) represents 'interpolation loudness variation', and the lower curve (dashed line) represents 'interpolation overall quality'. Both curves show a decreasing trend as attenuation increases. Error bars are present for each data point. + +| Attenuation in sending path (dB) | Interpolation Loudness Variation (MOS) | Interpolation Overall Quality (MOS) | +|----------------------------------|----------------------------------------|-------------------------------------| +| ref | 4.4 | 3.8 | +| 3 | 4.0 | 3.5 | +| 6 | 3.7 | 3.1 | +| 9 | 3.1 | 2.5 | +| 12 | 2.6 | 2.1 | +| 16 | 1.9 | 1.5 | +| 20 | 1.4 | 1.3 | +| 24 | 1.1 | 1.0 | + +Line graph titled 'overall quality and loudness variation SND' showing MOS vs. Attenuation in sending path (dB). The upper solid line represents 'interpolation loudness variation' and the lower dashed line represents 'interpolation overall quality'. Both curves decrease as attenuation increases from 'ref' to 24 dB. + +**Figure A.1/P.340 – Interpolation of test results for overall quality and loudness variation in sending direction (upper line: loudness variation)** + +### A.1.2 Results of the listening only tests with variable attenuations in receiving path + +The test parameter is attenuation (dB) in the receiving path. Long double talk sequences are used for this test. + +The tests have been made for different values of the attenuation, and the curve (perceived quality (in MOS) on the loudness variation scale) of Figure A.2 have been interpolated from these results. + +![Figure A.2/P.340: Interpolation of results for loudness variation in receiving direction. The graph shows MOS (Mean Opinion Score) on the y-axis (1.0 to 5.0) versus Attenuation in the receiving path (dB) on the x-axis (0 to 12 dB). A smooth curve represents the interpolation of loudness variation, with data points and error bars at 0, 2, 4, 6, and 12 dB. The curve starts at approximately 4.4 at 0 dB and decreases to about 2.0 at 12 dB.](24b549c0c4b79c515b244625cf4f4978_img.jpg) + +| Attenuation in the receiving path (dB) | MOS (Mean Opinion Score) | +|----------------------------------------|--------------------------| +| 0 | 4.4 | +| 2 | 4.2 | +| 4 | 3.9 | +| 6 | 3.0 | +| 12 | 2.0 | + +Figure A.2/P.340: Interpolation of results for loudness variation in receiving direction. The graph shows MOS (Mean Opinion Score) on the y-axis (1.0 to 5.0) versus Attenuation in the receiving path (dB) on the x-axis (0 to 12 dB). A smooth curve represents the interpolation of loudness variation, with data points and error bars at 0, 2, 4, 6, and 12 dB. The curve starts at approximately 4.4 at 0 dB and decreases to about 2.0 at 12 dB. + +**Figure A.2/P.340 – Interpolation of results for loudness variation in receiving direction** + +### A.1.3 Results of listening only tests on echo disturbance + +Test conditions (delay/TERL) for Figures A.3 and A.4 are derived from the curves 1% and 10%, as defined in ITU-T Recommendation G.131 [16]. + +Figure A.3 shows the results of the listening only tests in single talk. + +Figure A.4 shows the results of the listening only tests in double talk. + +For the results presented in Figure A.5, the delay is defined as a fix value of 100 ms. The variable parameter is TERL (dB). The tests are performed in single talk (lower line) and double talk (upper line). + +![Bar chart showing MOS for single talk conditions. The y-axis is MOS (1.0 to 5.0). The x-axis shows conditions: Reference, 10 ms 25/19 dB, 25 ms 33/27 dB, 50 ms 40/34 dB, 100 ms 47/41 dB, and 125 ms 48/42 dB. For each condition, two bars are shown: 1% curve (dark grey) and 10% curve (light grey). Error bars are present on all bars.](152efae989544ee653283e8de26cc9b1_img.jpg) + +**Echo (G.131, 1% and 10% curve, single talk)** + +| conditions | 1% curve (MOS) | 10% curve (MOS) | +|-----------------|----------------|-----------------| +| Reference | - | ~4.6 | +| 10 ms 25/19 dB | ~4.1 | ~3.7 | +| 25 ms 33/27 dB | ~4.2 | ~3.6 | +| 50 ms 40/34 dB | ~4.4 | ~3.8 | +| 100 ms 47/41 dB | ~4.5 | ~3.9 | +| 125 ms 48/42 dB | ~4.4 | ~3.5 | + +T1212400-00 + +Bar chart showing MOS for single talk conditions. The y-axis is MOS (1.0 to 5.0). The x-axis shows conditions: Reference, 10 ms 25/19 dB, 25 ms 33/27 dB, 50 ms 40/34 dB, 100 ms 47/41 dB, and 125 ms 48/42 dB. For each condition, two bars are shown: 1% curve (dark grey) and 10% curve (light grey). Error bars are present on all bars. + +**Figure A.3/P.340 – Echo disturbance for conditions according to G.131 +(single talk conditions, recordings in send direction of HFT, +judged at the handset)** + +![Bar chart showing MOS for double talk conditions. The y-axis is MOS (1.0 to 5.0). The x-axis shows conditions: Reference, 10 ms 25/19 dB, 25 ms 33/27 dB, 50 ms 40/34 dB, 100 ms 47/41 dB, and 125 ms 48/42 dB. For each condition, two bars are shown: 1% curve (dark grey) and 10% curve (light grey). Error bars are present on all bars.](fe655d77258397f7242c2df72b965b56_img.jpg) + +**Echo (G.131, 1 % and 10% curve, double talk)** + +| conditions | 1% curve (MOS) | 10% curve (MOS) | +|-----------------|----------------|-----------------| +| Reference | - | ~4.5 | +| 10 ms 25/19 dB | ~4.3 | ~4.1 | +| 25 ms 33/27 dB | ~4.5 | ~4.2 | +| 50 ms 40/34 dB | ~4.4 | ~4.3 | +| 100 ms 47/41 dB | ~4.5 | ~4.5 | +| 125 ms 48/42 dB | ~4.5 | ~4.2 | + +T1212410-00 + +Bar chart showing MOS for double talk conditions. The y-axis is MOS (1.0 to 5.0). The x-axis shows conditions: Reference, 10 ms 25/19 dB, 25 ms 33/27 dB, 50 ms 40/34 dB, 100 ms 47/41 dB, and 125 ms 48/42 dB. For each condition, two bars are shown: 1% curve (dark grey) and 10% curve (light grey). Error bars are present on all bars. + +**Figure A.4/P.340 – Echo disturbance for conditions according to G.131, (double talk +conditions, recordings in send direction of HFT, judged at the handset)** + +![Line graph titled 'Echo (100 ms, single talk-double talk)' showing MOS vs TELR (dB). The graph contains two lines: a solid line for 'interpolation single talk' and a dashed line for 'interpolation double talk'. Both lines show a downward trend as TELR increases from 41 dB to 20 dB. The dashed line is consistently higher than the solid line. Error bars are present for each data point.](f24d06b5e3b1d8ae12d4893e7619f6f5_img.jpg) + +| TELR (dB) | MOS (Single Talk - Solid Line) | MOS (Double Talk - Dashed Line) | +|-----------|--------------------------------|---------------------------------| +| 41 | ~3.9 | ~4.5 | +| 40 | ~3.4 | ~4.3 | +| 36 | ~2.7 | ~3.9 | +| 32 | ~2.3 | ~3.4 | +| 28 | ~2.1 | ~3.1 | +| 24 | ~1.9 | ~2.8 | +| 20 | ~1.8 | ~2.5 | + +Line graph titled 'Echo (100 ms, single talk-double talk)' showing MOS vs TELR (dB). The graph contains two lines: a solid line for 'interpolation single talk' and a dashed line for 'interpolation double talk'. Both lines show a downward trend as TELR increases from 41 dB to 20 dB. The dashed line is consistently higher than the solid line. Error bars are present for each data point. + +**Figure A.5/P.340 – Interpolation of LOT results for echo disturbances (upper line: double talk)** + +Table A.1 summarizes the results and defines the correlations between the individual parameters (attenuation in sending and receiving directions, echo loss) and the subjective judgment of the quality, in double talk situation. + +**Table A.1/P.340 – Values for parameters determining double talk performance (TELRDT, aHrdt, aHsdt) as a function of correlated MOS scores derived from LOT** + +| MOS | ≥4.0 | 4.0-3.5 | 3.5-3.0 | 3.0-2.5 | 2.5-2.0 | ≤2.0 | +|-------------------------|------|---------|---------|---------|---------|------| +| TELR DT [dB] | ≥37 | ≥33 | ≥27 | ≥21 | ≥13 | <13 | +| a Hsdt [dB] | ≤3 | ≤6 | ≤9 | ≤12 | ≤15 | >15 | +| a Hrdt [dB] | ≤3 | ≤5 | ≤8 | ≤10 | ≤12 | >12 | + +From Table A.1 has been derived the tables of clause 8 on the terminal categorization. + +## A.2 Relationship between Listening only tests, Double talk tests and Conversation tests: comparison of six different hands-free terminals + +Six hands-free telephones were used in all tests. Four of them are commercially available devices, the other two were laboratory prototypes. Table A.2 gives an overview of all devices under test: + +**Table A.2/P.340 – Technologies implemented in each type of hands-free terminal** + +| No. | Used algorithms | +|-----|----------------------------------------------------------------------------------| +| 1 | Level switching device, compander | +| 2 | Level switching device, echo canceller (70 ms), centre clipper | +| 3 | Echo canceller (254 ms), level switching device, centre clipper, frequency shift | +| 4 | Echo canceller | +| 5 | Compander, echo canceller, level switching device, background noise adaptation | +| 6 | Echo canceller (254 ms), speech controlled attenuation | + +Figures A.6 and A.7 give the test results on the quality parameter "overall quality" obtained with listening only tests and conversation tests. Globally it can be seen that the overall quality is judged more "severely" with listening only tests than with conversation tests, particularly for untrained subjects. + +For trained subjects there is a rather good correlation between the results of listening only tests and conversation tests, the MOS value obtained for conversation tests being slightly higher than for listening tests. + +![Bar chart showing Overall quality (MOS) for Untrained and Trained subjects across six HFT types (HFT1 to HFT6). The Y-axis ranges from 1.0 to 5.0. For Untrained subjects, MOS values are approximately: HFT1: 3.0, HFT2: 3.6, HFT3: 4.1, HFT4: 3.9, HFT5: 3.7, HFT6: 4.1. For Trained subjects, MOS values are approximately: HFT1: 2.3, HFT2: 1.1, HFT3: 3.0, HFT4: 2.3, HFT5: 3.0, HFT6: 4.0.](efbdff107f5a5aade487cf4ea0eae150_img.jpg) + +| HFT Type | Untrained (MOS) | Trained (MOS) | +|----------|-----------------|---------------| +| HFT1 | 3.0 | 2.3 | +| HFT2 | 3.6 | 1.1 | +| HFT3 | 4.1 | 3.0 | +| HFT4 | 3.9 | 2.3 | +| HFT5 | 3.7 | 3.0 | +| HFT6 | 4.1 | 4.0 | + +T1212430-00 + +Bar chart showing Overall quality (MOS) for Untrained and Trained subjects across six HFT types (HFT1 to HFT6). The Y-axis ranges from 1.0 to 5.0. For Untrained subjects, MOS values are approximately: HFT1: 3.0, HFT2: 3.6, HFT3: 4.1, HFT4: 3.9, HFT5: 3.7, HFT6: 4.1. For Trained subjects, MOS values are approximately: HFT1: 2.3, HFT2: 1.1, HFT3: 3.0, HFT4: 2.3, HFT5: 3.0, HFT6: 4.0. + +**Figure A.6/P.340 – Overall quality, conversational test, room 1 – Untrained and trained subjects separated** + +![Bar chart showing overall quality scores for HFT1 through HFT6. The y-axis ranges from 1.0 to 5.0. HFT1 and HFT2 have very low scores (around 1.1 and 1.05 respectively). HFT3 has the highest score (around 3.1), followed by HFT6 (around 3.2), HFT5 (around 2.2), HFT4 (around 2.0), and HFT2 (around 1.05). Error bars are shown for each data point.](bd4a5cf22b8d8060c4bf3577beb30df8_img.jpg) + +| HFT | Overall Quality (approx.) | +|------|---------------------------| +| HFT1 | 1.1 | +| HFT2 | 1.05 | +| HFT3 | 3.1 | +| HFT4 | 2.0 | +| HFT5 | 2.2 | +| HFT6 | 3.2 | + +T1212440-00 + +Bar chart showing overall quality scores for HFT1 through HFT6. The y-axis ranges from 1.0 to 5.0. HFT1 and HFT2 have very low scores (around 1.1 and 1.05 respectively). HFT3 has the highest score (around 3.1), followed by HFT6 (around 3.2), HFT5 (around 2.2), HFT4 (around 2.0), and HFT2 (around 1.05). Error bars are shown for each data point. + +**Figure A.7/P.340 – Overall quality, listening only test** + +Figures A.9 and A.10 present results on the parameter "double talk capability". + +It can be seen in Figure A.8, and particularly in Figure A.9, that HFT1 and HFT2 have a very low double talk capability. This poor behaviour is correlated to the loudness variation in double talk (see Figure A.11). It appears from these data that the loudness variation (or attenuation range) is an important factor to qualify the double talk capability. + +![Bar chart showing double talk capability scores for HFT1 through HFT6. The y-axis ranges from 1.0 to 5.0. HFT1 has the lowest score (around 2.2), followed by HFT2 (around 2.6). HFT3 has a score of around 3.8, HFT4 around 4.1, HFT5 around 3.6, and HFT6 around 4.0. Error bars are shown for each data point.](a72d02d4e6305689a79b46d36739cd51_img.jpg) + +| HFT | Double Talk Capability (approx.) | +|------|----------------------------------| +| HFT1 | 2.2 | +| HFT2 | 2.6 | +| HFT3 | 3.8 | +| HFT4 | 4.1 | +| HFT5 | 3.6 | +| HFT6 | 4.0 | + +T1212450-00 + +Bar chart showing double talk capability scores for HFT1 through HFT6. The y-axis ranges from 1.0 to 5.0. HFT1 has the lowest score (around 2.2), followed by HFT2 (around 2.6). HFT3 has a score of around 3.8, HFT4 around 4.1, HFT5 around 3.6, and HFT6 around 4.0. Error bars are shown for each data point. + +**Figure A.8/P.340 – Double talk capability, conversational test, room 1** + +![Bar chart showing double talk capability for HFT1-HFT6 in room 1. The y-axis ranges from 1.0 to 5.0. HFT1 is ~1.4, HFT2 is ~1.1, HFT3 is ~4.1, HFT4 is ~4.0, HFT5 is ~3.9, and HFT6 is ~3.8. Error bars are shown for each.](157e8c1d80a02c8eee893c2ea60a77a8_img.jpg) + +| HFT | Mean | Lower Error | Upper Error | +|------|------|-------------|-------------| +| HFT1 | 1.4 | 1.0 | 1.8 | +| HFT2 | 1.1 | 1.0 | 1.4 | +| HFT3 | 4.1 | 3.4 | 4.8 | +| HFT4 | 4.0 | 3.6 | 4.4 | +| HFT5 | 3.9 | 3.4 | 4.4 | +| HFT6 | 3.8 | 3.1 | 4.5 | + +T1212460-00 + +Bar chart showing double talk capability for HFT1-HFT6 in room 1. The y-axis ranges from 1.0 to 5.0. HFT1 is ~1.4, HFT2 is ~1.1, HFT3 is ~4.1, HFT4 is ~4.0, HFT5 is ~3.9, and HFT6 is ~3.8. Error bars are shown for each. + +**Figure A.9/P.340 – Double talk capability, double talk test, room 1** + +Figure A.10 presents results on the parameter "disturbance by echoes". + +![Bar chart showing disturbance by echoes for HFT1-HFT6 in room 1. The y-axis ranges from 1.0 to 5.0. HFT1 is 5.0, HFT2 is 5.0, HFT3 is ~4.3, HFT4 is ~2.7, HFT5 is 4.0, and HFT6 is ~4.5. Error bars are shown for each.](cfc852835f2d91bea8dc074568937e22_img.jpg) + +| HFT | Mean | Lower Error | Upper Error | +|------|------|-------------|-------------| +| HFT1 | 5.0 | 5.0 | 5.0 | +| HFT2 | 5.0 | 5.0 | 5.0 | +| HFT3 | 4.3 | 3.7 | 5.0 | +| HFT4 | 2.7 | 1.5 | 3.9 | +| HFT5 | 4.0 | 3.3 | 4.7 | +| HFT6 | 4.5 | 3.8 | 5.2 | + +T1212470-00 + +Bar chart showing disturbance by echoes for HFT1-HFT6 in room 1. The y-axis ranges from 1.0 to 5.0. HFT1 is 5.0, HFT2 is 5.0, HFT3 is ~4.3, HFT4 is ~2.7, HFT5 is 4.0, and HFT6 is ~4.5. Error bars are shown for each. + +**Figure A.10/P.340 – Disturbance by echoes, double talk test, room 1** + +Figure A.11 presents results on the parameter "loudness during double talk". + +![Bar chart showing Loudness during double talk for HFT1 through HFT6. The y-axis represents loudness from 1.0 to 5.0. HFT1 is ~1.1, HFT2 is ~1.3, HFT3 is ~2.8, HFT4 is ~3.9, HFT5 is ~2.9, and HFT6 is ~2.8. Error bars are shown for each bar. A small label 'T1212480-00' is at the bottom right.](32f80b2d70752613b6577797811f3a95_img.jpg) + +| HFT | Loudness (approx.) | +|------|--------------------| +| HFT1 | 1.1 | +| HFT2 | 1.3 | +| HFT3 | 2.8 | +| HFT4 | 3.9 | +| HFT5 | 2.9 | +| HFT6 | 2.8 | + +Bar chart showing Loudness during double talk for HFT1 through HFT6. The y-axis represents loudness from 1.0 to 5.0. HFT1 is ~1.1, HFT2 is ~1.3, HFT3 is ~2.8, HFT4 is ~3.9, HFT5 is ~2.9, and HFT6 is ~2.8. Error bars are shown for each bar. A small label 'T1212480-00' is at the bottom right. + +**Figure A.11/P.340 – Loudness during double talk, double talk test, room 2** + +From Figures A.10 and A.11, if we exclude HFT1 and HFT2, it can be seen that "echo disturbance" is correlated to the double talk capability. + +# APPENDIX I + +## Preliminary values + +This appendix gives some preliminary values and limits for parameters defined in clause 7. These parameters have been determined on the basis of results of subjective tests and are related to double talk performance and to the influence of companding or AGC devices on the speech quality. + +### I.1 Double talk performance + +| Parameters | From receive-to-send direction | From send-to-receive direction | +|-----------------------------------------------------|----------------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------| +| Threshold (minimum activation) Level to Switch Over | (for an activation level in receive direction of $-50$ dBV)
$L_{R-S} \leq -3$ dBPa at MRP | (for an activation level in send direction of $-9.7$ dBPa at MRP)
$L_{S-R} \leq -2$ dBV | +| Switching Times | $T_{S,R-S} \leq 100$ ms | $T_{S-10\text{dB},S-R} \leq 40$ ms
(the complete attenuation except a residual value of 10 dB is removed)
$T_{S,S-R} \leq 100$ ms | + +### I.2 Build-up time characteristics + +| Parameters | Sending direction | Receiving direction | +|----------------------------------------------------------|-------------------------------------------------------------------|-----------------------------------------------------------| +| Build-up time characteristics | $T_{R,idle-S} \leq 15$ ms | $T_{R,idle-R} \leq 10$ ms | +| Threshold level and
Build-up time for threshold level | $V_{Hidle-S} \leq -14$ dBPa (at MRP)
$T_{R,idle-S} \leq 15$ ms | $V_{Hidle-R} \leq -48.5$ dBV
$T_{R,idle-R} \leq 10$ ms | + +### I.3 Level adjustments by companding or AGC + +| | Parameters in sending direction | Parameters in receiving direction | +|--------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| Range of level adjustments as a function of the input signal level | For decreasing send levels
$L_S > -12$ dBPa (at MRP)
and for increasing send level
$L_S > -5$ dBPa (at MRP)
The deviation of the measured output signal level referred to the input signal level should be $\leq 3$ dB. | For decreasing receive levels
$L_R > -44$ dBV
and for increasing receive level
$L_R > -42$ dBV
The deviation of the measured output signal level referred to the input signal level should be $\leq 3$ dB. | +| Time duration for level adjustments | | | +| $\pm 5$ dB (comp. to $-4.7$ dBPa) | For a level variation of $-5$ dB:
No insertion of attenuation, duration for level adjustment
$t_{adj,S,-5} < 10$ ms
(tolerance scheme $\pm 3$ dB)
For a level variation of $+5$ dB:
$t_{adj,S,+5} < 15$ ms
(tolerance scheme $\pm 3$ dB) | For a level variation of $-5$ dB:
No insertion of attenuation, duration for level adjustment
$t_{adj,R,-5} < 20$ ms
(tolerance scheme $\pm 3$ dB)
For a level variation of $+5$ dB:
$t_{adj,R,+5} < 15$ ms
(tolerance scheme $\pm 3$ dB) | +| $\pm 10$ dB (comp. to $-4.7$ dBPa) | For a level variation of $-10$ dB:
Insertion of attenuation after hang-over time $130 \pm 20$ ms with $(20 \pm 3)$ dB/s
For a level variation of $+10$ dB:
Removal of attenuation after
$t_{adj,S,+10} < 15$ ms (tolerance scheme $\pm 3$ dB) | For a level variation of $-10$ dB:
No insertion of attenuation, duration for level adjustment
$t_{adj,R,-10} < 100$ ms for max volume
$t_{adj,R,-10} < 20$ ms for min. volume
(tolerance scheme $\pm 3$ dB)
For a level variation of $+10$ dB:
Duration for level adjustment
$t_{adj,R,+10} < 15$ ms
(tolerance scheme $\pm 3$ dB) | + +| | Parameters in sending direction | Parameters in receiving direction | +|------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| $\pm 15$ dB (comp. to $-4.7$ dBPa) | For a level variation of $-15$ dB:
Insertion of attenuation after
hang-over time: $130 \pm 20$ ms with
$(20 \pm 3$ dB)/s | For a level variation of $-15$ dB:
No insertion of attenuation,
duration for level adjustment
$t_{\text{adj,R,-15}} < 100$ ms for max.
volume
$t_{\text{adj,R,+15}} < 20$ ms for min. volume
(tolerance scheme $\pm 3$ dB) | +| | For a level variation of $+15$ dB:
Removal of attenuation after
$t_{\text{adj,S,+15}} < 15$ ms
(tolerance scheme $\pm 3$ dB) | For a level variation of $+15$ dB:
Duration for level adjustment
$t_{\text{adj,R,+15}} < 15$ ms
(tolerance scheme $\pm 3$ dB) | + +# APPENDIX II + +## Examples of processing delays for mobile hands-free terminal + +### II.1 Example 1 + +## **Tadd\_proc (additional time for processing)** + +Tadd\_proc delays providing minimal values for acceptable echo or/and noise reduction are defined as follows : + +- **Tadd\_proc\_AEC** for Echo Cancellation when using Hands-free mobile terminal, +**Tadd\_proc\_AEC: 28 (to 40) ms.** +- **Tadd\_proc\_NR** when using Noise Reduction (NR) and a coupling reduction processing for use of mobile handset, +**Tadd\_proc\_NR: 20 (to 32) ms.** +- **Tadd\_proc\_HF** for Hands Free when using AEC and NR when using Hands-free mobile terminal and if it is desired to add NR for listening comfort, +**Tadd\_proc\_HF: 36 (to 56) ms.** + +Tadd\_proc taking into account any additional and mandatory speech processing block (including noise bad effects compensation) where Tadd\_proc must be decomposed as follows: + +| | | +|--------------------------------------------------------|-------------------| +| Delay for signal block size or sub-band decomposition: | 16 to 24 ms | +| Delay for noise reduction: | 12 to 16 ms | +| Additional delay for computation: | 8 to 16 ms | +| Tadd_proc: | 36 to 56 ms | + +### II.2 Example 2 + +To fully exploit the benefits of block processing, it is important that the block length is sufficient: + +- to provide enough data for statistically good estimates of the properties of noisy speech; +- to efficiently handle reverberation times in normal car cabins; +- to provide adequate resolution in the frequency domain. + +It is also desirable, for reasons of efficiency in frequency domain processing, that the block length is a power of two. + +A block length of 256 samples, which corresponds to a 32 ms block at 8 ksamples/s, represents a good balance in these respects. + +With 32 ms blocks there will be an inherent delay of 32 ms. A reasonable figure for processing time of such blocks, using today's cost competitive DSPs, is 10 ms. + +An additional delay for hands-free signal processing could be a minimum of **32 ms + 10 ms = 42 ms**. + +# APPENDIX III + +## Recommended test bed + +This appendix gives some examples of recommended test beds that can be implemented for testing according to this Recommendation. + +### III.1 Example 1: Implemented test hybrid + +The test hybrid can be constructed from a passive 2-4 wire hybrid, mated to a digital echo canceller as described in Figure III.1. The digital echo canceller is described below. + +![Block diagram of the recommended test bed for a digital echo canceller. On the left, a 'Tip Ring interface & DC Feed' block is connected to an 'HFT' (Hands-Free Terminal) block. This interface is connected to a 'Passive two-four wire hybrid' (represented by a transformer symbol). The hybrid has two output lines: 'Line out' and 'Line in'. 'Line out' connects to a 'Sound card with Prog. DSP' block. 'Line in' connects to a 'Post Processing on PC for NLMS Echo Canceller' block. The 'Sound card with Prog. DSP' block is also connected to the 'Post Processing on PC' block. A small label 'T1212490-00' is present at the bottom right of the diagram.](0d0897dd7f03672d186db9d56ea0d6f5_img.jpg) + +Block diagram of the recommended test bed for a digital echo canceller. On the left, a 'Tip Ring interface & DC Feed' block is connected to an 'HFT' (Hands-Free Terminal) block. This interface is connected to a 'Passive two-four wire hybrid' (represented by a transformer symbol). The hybrid has two output lines: 'Line out' and 'Line in'. 'Line out' connects to a 'Sound card with Prog. DSP' block. 'Line in' connects to a 'Post Processing on PC for NLMS Echo Canceller' block. The 'Sound card with Prog. DSP' block is also connected to the 'Post Processing on PC' block. A small label 'T1212490-00' is present at the bottom right of the diagram. + +Figure III.1/P.340 + +The passive 2-4-wire hybrid is a transformer setup that can be balanced to provide a simulated 4-wire signal path that is then interfaced to the digital echo canceller. + +The digital echo canceller can be implemented through a post processing or real time application. In either case, simultaneous (full duplex) record-playback capability is needed. + +### III.2 Example 2: Adaptive Echo Canceller + +An NLMS (normalized least mean square) adaptive echo canceller is used to model the actual echo path formed by the 4- to 2-wire hybrid and the HFT. The block diagram is shown in Figure III.2. + +![Block diagram of 2- to 4-wire conversion. It shows an input x(n) entering an Echo canceller. The Echo canceller outputs y(n) to a summer (+) and also to a D/A converter within a Sound card. The summer also receives e(n) and outputs a signal that passes through a delay block Z^D to an A/D converter within the Sound card. The D/A converter's output goes to a summing junction (diamond) and then to a DUT. The A/D converter's output also goes to the summing junction. The summing junction's output is labeled T1212500-00.](e8e818455bb0d1a6153299a388b94868_img.jpg) + +Block diagram of 2- to 4-wire conversion. It shows an input x(n) entering an Echo canceller. The Echo canceller outputs y(n) to a summer (+) and also to a D/A converter within a Sound card. The summer also receives e(n) and outputs a signal that passes through a delay block Z^D to an A/D converter within the Sound card. The D/A converter's output goes to a summing junction (diamond) and then to a DUT. The A/D converter's output also goes to the summing junction. The summing junction's output is labeled T1212500-00. + +Figure III.2/P.340 – Block diagram of 2- to 4-wire conversion + +![Test bed diagram for acoustic measurements. It is divided into a 'Test Room' and an 'Analog Set'. In the Test Room, there is a 'Hands-free Terminal' on a 'Test Table', an 'Artificial Mouth or Meas. Mic (if required)', and a 'Head and Torso Simulator or equivalent'. Room noise can be added. The terminal is connected to a 'Digital Set' which contains an 'ISDN/Reference Codec' and a 'D.C. Feed'. The Digital Set connects to a '2 Wire/4 Wire Digital Hybrid'. The Hybrid has inputs R_in and S_out. R_in is connected to 'Sources' (represented by circles with waveforms) and 'Electrical equivalent of acoustic S_in'. S_out is connected to 'Data Acquisition' and 'Electrical equivalent of acoustic R_out'. The Data Acquisition unit also receives R_in and S_out from a 'Source' in the Analog Set. The diagram is labeled T1212510-00.](60ffbad6c0fb7371a57fe8f267d2d141_img.jpg) + +Test bed diagram for acoustic measurements. It is divided into a 'Test Room' and an 'Analog Set'. In the Test Room, there is a 'Hands-free Terminal' on a 'Test Table', an 'Artificial Mouth or Meas. Mic (if required)', and a 'Head and Torso Simulator or equivalent'. Room noise can be added. The terminal is connected to a 'Digital Set' which contains an 'ISDN/Reference Codec' and a 'D.C. Feed'. The Digital Set connects to a '2 Wire/4 Wire Digital Hybrid'. The Hybrid has inputs R\_in and S\_out. R\_in is connected to 'Sources' (represented by circles with waveforms) and 'Electrical equivalent of acoustic S\_in'. S\_out is connected to 'Data Acquisition' and 'Electrical equivalent of acoustic R\_out'. The Data Acquisition unit also receives R\_in and S\_out from a 'Source' in the Analog Set. The diagram is labeled T1212510-00. + +Figure III.3/P.340 – Test bed diagram + +# APPENDIX IV + +## Bibliography + +- [IV.1] European Committee for Standardization (CEN) Office Chair/Desk Working Position – Dimensions and Design Requirements, *CEN: prEN91/August 1981*. +- [IV.2] CCITT: A Method for Measuring the Sensitivity of a Loudspeaking Telephone Set, Annex 2 to Question 17/XII, *White Book*, Vol. V, ITU, Geneva 1969. +- [IV.3] ISO 1996-3:1987, *Acoustics – Description and measurement of environmental noise – Part 3: Application to noise limits*. +- [IV.4] BERANEK (L.L.): Noise and Vibration Control, *McGraw Hill*, pp. 564-566, New York 1971. +- [IV.5] XIANG (N.): Using M-sequences for Determining the Impulse Responses of LTI-systems, *Signal Processing*, No. 28, pp. 139-152, 1992. +- [IV.6] JORDAN (V.L.): Acoustical Criteria for Concert Hall Stages, *Applied Acoustics*, No. 15, pp. 321-328, 1982. +- [IV.7] HOTH (D.F.): Room Noise Spectra at Subscribers Telephone Locations. *J.A.S.A.*, Vol. 12. pp. 499-504, April 1941. +- [IV.8] ITU-T COM 12-60 (1995), Subjective Evaluation of Transmission Quality for Hands-free Telephones, Federal Republic of Germany. +- [IV.9] ITU-T COM 12-61 (1995), Supplementary Measurement Proposal for Certification Measurements of Hands-free Equipment, Federal Republic of Germany. +- [IV.10] ITU-T COM 12-6 (1997), Subjective evaluation of hands-free telephones using conversational tests, specific double talk tests and listening only tests, Federal Republic of Germany. +- [IV.11] ITU-T COM 12-102 (1999), Proposals for the definitions of different types of hands-free telephones based on double talk performance, Federal Republic of Germany. +- [IV.12] ITU-T SG 12-D.83 (November 1998) Advanced Measurements for Hands-free Telephones and Non-linear/Time variant Operating devices, Deutsche Telekom Berkom/Head Acoustics). +- [IV.13] ETSI 0358 601 (TR 101110) Digital Cellular Telecommunications system (Phase 2+). Characterisation, test methods and Quality assessment for hands-free mobiles. +- [IV.14] IEEE 1329 Standard Method for Measuring Transmission Performance of Hands-Free Telephone Sets. + + + +# SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|--------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series B | Means of expression: definitions, symbols, classification | +| Series C | General telecommunication statistics | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Construction, installation and protection of cables and other elements of outside plant | +| Series M | TMN and network maintenance: international transmission systems, telephone circuits, telegraphy, facsimile and leased circuits | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks and open system communications | +| Series Y | Global information infrastructure and Internet protocol aspects | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.342-200906-I_PDF-E/0dd5ee731e9d7e34e498b5c926110773_img.jpg b/marked/P/T-REC-P.342-200906-I_PDF-E/0dd5ee731e9d7e34e498b5c926110773_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..eda415f8a7442329faf82988736a7b6add5656eb --- /dev/null +++ b/marked/P/T-REC-P.342-200906-I_PDF-E/0dd5ee731e9d7e34e498b5c926110773_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:57383bbb00f999ff8b35f003b858edfe0d2fd778ccde4a7a5f7179e32165f8bf +size 57615 diff --git a/marked/P/T-REC-P.342-200906-I_PDF-E/14a22f23ced8ba1d63ece69861dbaacc_img.jpg b/marked/P/T-REC-P.342-200906-I_PDF-E/14a22f23ced8ba1d63ece69861dbaacc_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..36cf51f8ba07e44aa7a4b009f3ba08eb2b570b1c --- /dev/null +++ b/marked/P/T-REC-P.342-200906-I_PDF-E/14a22f23ced8ba1d63ece69861dbaacc_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:11780faeee75ba39ab4ed9de67a9e67abec0d9ba3068ebec661dea70c16e29d3 +size 4338 diff --git a/marked/P/T-REC-P.342-200906-I_PDF-E/2a77eb32ef4c4d8a5c1758a53a908336_img.jpg b/marked/P/T-REC-P.342-200906-I_PDF-E/2a77eb32ef4c4d8a5c1758a53a908336_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fb889b6d7640ccc68d8dc3f04be88de28b031e0c --- /dev/null +++ b/marked/P/T-REC-P.342-200906-I_PDF-E/2a77eb32ef4c4d8a5c1758a53a908336_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3657ff35cfe8698b9caeec1ea56ab0972f81dd6b50e7586952594f45f8c7d11f +size 110088 diff --git a/marked/P/T-REC-P.342-200906-I_PDF-E/625e10f48104ba2b06b2220a9b224712_img.jpg b/marked/P/T-REC-P.342-200906-I_PDF-E/625e10f48104ba2b06b2220a9b224712_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..729fe2d4130ea4a88bcb378ce1f6e828c2572fba --- /dev/null +++ b/marked/P/T-REC-P.342-200906-I_PDF-E/625e10f48104ba2b06b2220a9b224712_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c459ef2bd30343b36a7a4fb77515fcb9d3eac13ef88554994b7dc174981453db +size 22325 diff --git a/marked/P/T-REC-P.342-200906-I_PDF-E/cfef993dcc8fb513de79eb1f93cf26ae_img.jpg b/marked/P/T-REC-P.342-200906-I_PDF-E/cfef993dcc8fb513de79eb1f93cf26ae_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..00ab7bb152554af818df296036b462362caa0b7f --- /dev/null +++ b/marked/P/T-REC-P.342-200906-I_PDF-E/cfef993dcc8fb513de79eb1f93cf26ae_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5dcf66d212af61c9a7b86264c2fa574bc5242b6715dbdb9cce9cb6effd2007d8 +size 45100 diff --git a/marked/P/T-REC-P.342-200906-I_PDF-E/d3f6de4fe9f9138fc6afc584b5104433_img.jpg b/marked/P/T-REC-P.342-200906-I_PDF-E/d3f6de4fe9f9138fc6afc584b5104433_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..adc4ba70ba52eeccd4a9d5476c442aa894b6aa69 --- /dev/null +++ b/marked/P/T-REC-P.342-200906-I_PDF-E/d3f6de4fe9f9138fc6afc584b5104433_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e0fed7276c22f6e1aa6e1dbc87e74509dd9af991f577d7d9225e36586e8d5bb +size 28976 diff --git a/marked/P/T-REC-P.342-200906-I_PDF-E/d5a837fa4f4675e5ee596003cf55985c_img.jpg b/marked/P/T-REC-P.342-200906-I_PDF-E/d5a837fa4f4675e5ee596003cf55985c_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..74637425b7560e2c8ac979e9c8f5bfd5f9dba415 --- /dev/null +++ b/marked/P/T-REC-P.342-200906-I_PDF-E/d5a837fa4f4675e5ee596003cf55985c_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b8a2fd06dde43361973651d8c3000fa29576fedc78dc9e3cb3706611fec43044 +size 25362 diff --git a/marked/P/T-REC-P.342-200906-I_PDF-E/f4d72193f77f6646a2a1f4baaa927154_img.jpg b/marked/P/T-REC-P.342-200906-I_PDF-E/f4d72193f77f6646a2a1f4baaa927154_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3b6eff2934419116130d8f1016e087f3bbb50697 --- /dev/null +++ b/marked/P/T-REC-P.342-200906-I_PDF-E/f4d72193f77f6646a2a1f4baaa927154_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:18dbca62898406b78c4e0e620c3d39a494b5166151eb7ea4863b3a10f2d8a2f3 +size 108975 diff --git a/marked/P/T-REC-P.370-199608-I_PDF-E/raw.md b/marked/P/T-REC-P.370-199608-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..541dc3e161c76d039e8bcc023048b951e27aaaab --- /dev/null +++ b/marked/P/T-REC-P.370-199608-I_PDF-E/raw.md @@ -0,0 +1,755 @@ + + +![ITU logo: A globe with a lightning bolt and the letters ITU.](2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg) + +ITU logo: A globe with a lightning bolt and the letters ITU. + +INTERNATIONAL TELECOMMUNICATION UNION + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.370** + +(08/96) + +SERIES P: TELEPHONE TRANSMISSION QUALITY + +Subscribers' lines and sets + +# --- **Coupling Hearing Aids to Telephone sets** + +ITU-T Recommendation P.370 + +(Previously "CCITT Recommendation") + +--- + +# ITU-T P-SERIES RECOMMENDATIONS **TELEPHONE TRANSMISSION QUALITY** + +| | | +|-----------------------------------------------------------------------------------------------|--------------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | P.10-P.29 | +| Subscribers' lines and sets | P.300-P.399 | +| Subscribers' lines and sets | P.30-P.39 | +| Transmission standards | P.40-P.49 | +| Objective measuring apparatus | P.500-P.599 | +| Objective measuring apparatus | P.50-P.59 | +| Objective electro-acoustical measurements | P.60-P.69 | +| Measurements related to speech loudness | P.70-P.79 | +| Methods for objective and subjective assessment of quality | P.800-P.999 | +| Methods for objective and subjective assessment of quality | P.80-P.99 | + +*For further details, please refer to ITU-T List of Recommendations.* + +# FOREWORD + +The ITU-T (Telecommunication Standardization Sector) is a permanent organ of the International Telecommunication Union (ITU). The ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Conference (WTSC), which meets every four years, establishes the topics for study by the ITU-T Study Groups which, in their turn, produce Recommendations on these topics. + +The approval of Recommendations by the Members of the ITU-T is covered by the procedure laid down in WTSC Resolution No. 1 (Helsinki, March 1-12, 1993). + +ITU-T Recommendation P.370, was prepared by ITU-T Study Group 12 (1993-1996) and was approved under the WTSC Resolution No. 1 procedure on the 30th of August 1996. + +--- + +# NOTE + +In this Recommendation, the expression “Administration” is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +© ITU 1997 + +All rights reserved. No part of this publication may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying and microfilm, without permission in writing from the ITU. + +# CONTENTS + +*Page* + +| | | | +|-------|-------------------------------------------------------------------------------------------------------------------------|----| +| 1 | Scope ..... | 1 | +| 2 | Normative references ..... | 1 | +| 3 | Definitions and abbreviations..... | 2 | +| 3.1 | Definitions ..... | 2 | +| 3.2 | Abbreviations..... | 2 | +| 4 | Magnetic field strength around the earcap of telephone handsets which provide for coupling to hearing aids ..... | 2 | +| 4.1 | Introduction ..... | 2 | +| 4.2 | Magnetic field strength measurements and recommended values ..... | 3 | +| 4.2.1 | Calibration of acoustic receive level ..... | 3 | +| 4.2.2 | Magnetic field strength level..... | 4 | +| 4.2.3 | Linearity of the magnetic field strength ..... | 4 | +| 4.2.4 | Measurement of frequency characteristics..... | 4 | +| 4.3 | Probe coil..... | 5 | +| 4.3.1 | Dimensions ..... | 5 | +| 4.3.2 | Calibration of the probe coil ..... | 5 | +| 4.3.3 | Distortion ..... | 5 | +| 5 | Characteristics of telephone sets that provide additional amplification for the benefit of hearing impaired users ..... | 6 | +| 5.1 | Introduction ..... | 6 | +| 5.2 | Sending characteristics ..... | 6 | +| 5.2.1 | Sensitivity ..... | 6 | +| 5.2.2 | Frequency response..... | 6 | +| 5.3 | Receiving characteristics ..... | 7 | +| 5.3.1 | Sensitivity ..... | 7 | +| 5.3.2 | Frequency response..... | 7 | +| 5.3.3 | Restoring normal receive amplification..... | 7 | +| 5.4 | Sidetone ..... | 7 | +| 6 | Electrical coupling of telephone sets to hearing aids ..... | 8 | +| 6.1 | Introduction ..... | 8 | +| 6.2 | Electrical characteristics ..... | 8 | +| 6.2.1 | Point of connection in the telephone set ..... | 8 | +| 6.2.2 | Electrical characteristics at interconnection point..... | 8 | +| 6.3 | Interconnection with the telephone set ..... | 9 | +| 6.3.1 | Connector type..... | 9 | +| 6.3.2 | Connector position..... | 9 | +| 6.4 | Safety issues..... | 9 | +| | Annex A – Measurement of an acousto-magnetic adapter generating a magnetic field..... | 10 | +| A.1 | Measurement procedures ..... | 10 | +| A.2 | Magnetic field requirements ..... | 10 | +| A.3 | Physical properties..... | 10 | +| | Appendix I – Examples of frequency characteristics of hearing aids coupled to telephones..... | 10 | +| | Appendix II – Bibliography..... | 15 | + +# SUMMARY + +This Recommendation has been updated from Recommendation P.37, Coupling Hearing Aids to Telephone sets. + +Clause 4 covers inductive coupling and is technically similar to the original Recommendation P.37. Clause 4 however now includes the relevant figures that were previously referred out to other Recommendations. It presents recommended levels and frequency characteristics for the inductive field at the telephone earphone relative to a sound level presented at the same earphone. + +Clause 5 covers additional amplification that may be provided for the benefit of hearing impaired users, and only editorial changes have been made to the text originally contained in Recommendation P.37. Limits for the additional amplification and range of volume control are given. Frequency response is also covered. + +Clause 6 introduces the possibility of electrical coupling of a telephone set to a hearing aid. It is compatible with IEC Publication 118 Part 6, which covers the characteristics of hearing aid electrical input circuits. The electrical output level relative to the sound level in the same telephone's earphone is given in this Recommendation, together with frequency response, source impedance and maximum output level. + +Technically this Recommendation is compatible with similar documents recently drafted within ETSI covering the same subjects. + +# FOREWORD + +It is recognized that there is a sizeable proportion of telephone users that have difficulty in conversing over a telephone connection due to hearing loss. To alleviate these difficulties special means have been provided in many national systems to enable hearing impaired users to couple their hearing aids inductively to the telephone receiver, and a number of national/international specifications define characteristics for this form of coupling. Clause 4 addresses the requirements for successful inductive coupling of hearing aids to telephone sets. + +Furthermore, it is also recognized that many hearing impaired users are able to have satisfactory telephone conversations while coupling their hearing aids acoustically to the telephone receiver, or even using the telephone handset without a hearing aid. This latter situation is possible due to the fact that under good conditions a telephone connection can be louder than a face-to-face conversation over a one-metre air path by up to 30 dB. Provision of additional amplification in the mouth-to-ear path can greatly increase the proportion of telephone conversations involving hearing impaired users that are rated “good”. Clause 5 addresses this form of coupling. + +Certain national standards also exist to enable direct electrical connection of hearing aids to telephone apparatus. This form of coupling is addressed in clause 6, and has been written to be compatible with the IEC Publication 118-6, 1984, covering the characteristics of electrical input circuits for hearing aids. + +## NOTES + +1 Care should be taken when designing hearing aids to include sufficient immunity to radio frequency interference to avoid disturbances arising from the detection of radio signals emitted by cordless and mobile telephones. + +2 Tests have shown that telephone earphones having low acoustic impedance, which exhibit less low frequency loss due to leakage effects, couple acoustically very well into hearing aids by presenting a much flatter overall frequency characteristic to the ear of the user than conventional earphones having high acoustic impedance. + +# COUPLING HEARING AIDS TO TELEPHONE SETS + +(Geneva, 1996) + +# 1 Scope + +Clause 4 applies to telephone handsets having supra-aural earphones that can be connected to the Public Switched Telephone Network (PSTN) or the Integrated Services Digital Network (ISDN) which are intended for direct application to the ear (e.g. traditional handsets, operators' headsets) and which provide a magnetic field for coupling to hearing aids. It specifies the level, linearity and frequency dependence of the magnetic field strength produced by the handset and characteristics for the calibrated probe coil. + +Clause 5 specifies the electro-acoustic performance characteristics of telephony terminals which are intended for direct application to the ear (e.g. traditional handsets, operators' headsets) and which provide, at the earphone, additional amplification in the receiving direction compared with the Receiving Loudness Rating (RLR) specified in the requirements of the national system. + +Clause 6 specifies the electrical characteristics for the electrical coupling of the telephony function, implemented telecommunication terminal equipment, to hearing aids. It specifies the level and frequency response relative to the acoustic output at the earphone, also the noise and maximum level. + +Annex A specifies the measuring method for an acousto-magnetic adapter that converts the acoustic output of an associated telephone receiver to a magnetic field, in accordance with 4.2.1 and 4.2.2, that can be received by the magnetic pick-up coil in a hearing aid. + +# 2 Normative references + +The following Recommendations, and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; all users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. + +- IEC Publication 118-6: 1984, *Hearing aids – Part 6: Characteristics of electrical input circuits for hearing aids*. +- ITU-T Recommendation P.10 (1993), *Vocabulary of terms on telephone transmission quality and telephone sets*. +- IEC Publication 118-4: 1981, *Hearing aids – Part 4: Magnetic field strength in audio-frequency induction loops for hearing aid purposes*. +- IEC Publication 118-1: 1995, *Hearing aids – Part 1: Hearing aids with induction pick-up coil input*. +- ITU-T Recommendation P.340 (1996), *Transmission characteristics of hands-free telephones*. +- ITU-T Recommendation G.121 (1993), *Loudness Ratings (LRs) of national systems*. +- IEC Publication 130-8: 1976, *Connectors for frequencies below 3 MHz – Part 8: Concentric connectors for audio circuits in radio receivers*. +- IEC Publication 268-11: 1987, *Sound system equipment – Part 11: Application of connectors for the interconnection of sound system equipment*. +- ITU-T Recommendation P.57 (1993), *Artificial ears*. +- CCITT Recommendation P.48 (1988), *Specification for an intermediate reference system*. + +1) Previously Recommendation P.37. + +# 3 Definitions and abbreviations + +## 3.1 Definitions + +The relevant definitions given in Recommendation P.10 apply along with the following: + +**3.1.1 level of magnetic field strength:** the maximum level of the magnetic field strength and its measurement are given in 4.2.2. The units are amperes per metre (A/m). + +**3.1.2 plane of measurement:** A plane parallel to the earcap at a distance of 10 mm. + +**3.1.3 The adapter plane:** It is defined as the plane formed by the contacting points of a flat surface against the surface of the acousto-magnetic adapter opposite the earcap connection. + +## 3.2 Abbreviations + +For the purposes of this Recommendation, the following abbreviations are used. + +| | | +|-------|-------------------------------------------| +| AGC | Automatic Gain Control | +| DRP | Drum Reference Point | +| ERP | Ear Reference Point | +| HATS | Head and Torso Simulator | +| IEC | International Electrotechnical Commission | +| IRS | Intermediate Reference System | +| ISDN | Integrated Services Digital Network | +| LSTR | Listener Sidetone Rating | +| OLR | Overall Loudness Rating | +| $p_e$ | Sound pressure level at the ERP | +| PSTN | Public Switched Telephone Network | +| RLR | Receiving Loudness Rating | +| SLR | Sending Loudness Rating | +| STMR | Sidetone Masking Rating | +| TCL | Terminal Coupling Loss | + +# 4 Magnetic field strength around the earcap of telephone handsets which provide for coupling to hearing aids + +## 4.1 Introduction + +Magnetic induction systems incorporated in telephone handsets generate an alternating magnetic field with spatial characteristics which make the field detectable by hearing aids equipped with induction pick-up coils. + +Reception of an audio-frequency signal via an induction pick-up coil can often allow an acceptable signal-to-noise ratio to be achieved in cases where the acoustical reception would otherwise be degraded by background noise. + +The magnetic field strength which enables induction pick-up coils in hearing aids to function effectively must be high enough to produce an acceptable signal-to-noise ratio but not so high as to cause overloading of the hearing aid. + +The value of the magnetic field strength recommended in this Recommendation has been chosen so that these requirements are met as far as possible. (See Bibliography [1] for a discussion on end-to-end levels that can be expected for hearing aid users coupled inductively to a telephone set. This can be compared with the preferred listening levels for users having normal hearing [2]). + +Information on the measurement of an acoustic-magnetic adapter generating a magnetic field is found in Annex A. + +## 4.2 Magnetic field strength measurements and recommended values + +### 4.2.1 Calibration of acoustic receive level + +Using the measurement configuration shown in Figure 1 for analogue telephones, Figure 2 for digital telephones, the drive level of the oscillator shall be adjusted to produce a sound pressure level of $-14$ dBPa (80 dBSPL) at 1000 Hz. This drive level shall be used for determining the level and frequency characteristics of the magnetic field strength. + +![Figure 1/P.370: Setting the sound pressure level p_e in the artificial ear for an analogue telephone set. The diagram shows a 'Local telephone system' consisting of a 'Handset', 'Speech Circuit', 'Feed Circuit', and 'Signal Generator (600 ohm source)'. The 'Signal Generator' is connected to the 'Feed Circuit', which is connected to the 'Speech Circuit', which is connected to the 'Handset'. The 'Handset' is connected to an 'Art. Ear' (Artificial Ear). The 'Art. Ear' is connected to a 'Measuring amplifier', which is connected to a 'Voltage measuring device (e.g. voltmeter or signal analyser) calibrated to read sound pressure'. The 'Art. Ear' is also labeled with 'p_e = -14 dBPa'. The diagram is labeled 'T1207710-95/d01'.](367926125450c2bc3f4bdca9d59a62ba_img.jpg) + +Figure 1/P.370: Setting the sound pressure level p\_e in the artificial ear for an analogue telephone set. The diagram shows a 'Local telephone system' consisting of a 'Handset', 'Speech Circuit', 'Feed Circuit', and 'Signal Generator (600 ohm source)'. The 'Signal Generator' is connected to the 'Feed Circuit', which is connected to the 'Speech Circuit', which is connected to the 'Handset'. The 'Handset' is connected to an 'Art. Ear' (Artificial Ear). The 'Art. Ear' is connected to a 'Measuring amplifier', which is connected to a 'Voltage measuring device (e.g. voltmeter or signal analyser) calibrated to read sound pressure'. The 'Art. Ear' is also labeled with 'p\_e = -14 dBPa'. The diagram is labeled 'T1207710-95/d01'. + +FIGURE 1/P.370 + +Setting the sound pressure level $p_e$ in the artificial ear for an analogue telephone set + +![Figure 2/P.370: Setting the sound pressure level p_e in the artificial ear for a digital telephone set. The diagram shows a 'Digital Telephone' connected to an 'Interface', which is connected to a 'D/A Conversion Generator', which is connected to a 'Signal Generator'. The 'Signal Generator' is also connected to a 'Control'. The 'Digital Telephone' is connected to a 'Handset', which is connected to an 'Art. Ear' (Artificial Ear). The 'Art. Ear' is connected to a 'Measuring amplifier', which is connected to a 'Voltage measuring device (e.g. voltmeter or signal analyser) calibrated to read sound pressure'. The 'Art. Ear' is also labeled with 'p_e = -14 dBPa'. The diagram is labeled 'T1207720-95/d02'.](a234352dfaccdc24745c88eef7724cc6_img.jpg) + +Figure 2/P.370: Setting the sound pressure level p\_e in the artificial ear for a digital telephone set. The diagram shows a 'Digital Telephone' connected to an 'Interface', which is connected to a 'D/A Conversion Generator', which is connected to a 'Signal Generator'. The 'Signal Generator' is also connected to a 'Control'. The 'Digital Telephone' is connected to a 'Handset', which is connected to an 'Art. Ear' (Artificial Ear). The 'Art. Ear' is connected to a 'Measuring amplifier', which is connected to a 'Voltage measuring device (e.g. voltmeter or signal analyser) calibrated to read sound pressure'. The 'Art. Ear' is also labeled with 'p\_e = -14 dBPa'. The diagram is labeled 'T1207720-95/d02'. + +FIGURE 2/P.370 + +Setting the sound pressure level $p_e$ in the artificial ear for a digital telephone set + +### 4.2.2 Magnetic field strength level + +Place the centre of the calibrated probe coil (see 4.3) in the plane of measurement and orientate it in any direction for maximum coupling. Determine the magnetic field strength at 1000 Hz using the drive level as given in 4.2.1. + +Recommended range of values for the magnetic field strength is: + +$$-17 \text{ to } -30 \text{ dB relative to } 1 \text{ A/m.}$$ + +NOTE – For effective performance, hearing aids with magnetic pick-up coils primarily intended for coupling to magnetic loops in auditoria in accordance with IEC Publication 118-4 are likely to require a field strength in the upper end of the range, i.e. $-17$ to $-25$ dB relative to $1 \text{ A/m}$ . + +### 4.2.3 Linearity of the magnetic field strength + +With the probe coil positioned as in 4.2.2, increase the 1000 Hz drive level specified in 4.2.1 by 20 dB and measure the resulting magnetic field strength. + +The field strength should increase by $20 \text{ dB} \pm 1 \text{ dB}$ . + +### 4.2.4 Measurement of frequency characteristics + +With probe coil positioned as in 4.2.2, and the 1000 Hz drive level specified in 4.2.1, vary the frequency from 300 Hz to 5000 Hz for analogue telephones and to the upper frequency limit for digital telephones (4000 or 7000 Hz) as appropriate and measure the resulting field strength. The magnetic field strength frequency characteristics shall fit within the template shown in Figure 3. + +![Figure 3/P.370: Magnetic field strength characteristics. A graph showing the magnetic field strength (dB relative to level at 1000 Hz) versus frequency (Hz). The y-axis ranges from -20 to 20 dB, with major ticks at -20, -10, -3, 0, 3, 10, and 20. The x-axis ranges from 300 to 7000 Hz, with major ticks at 300, 500, 1000, 2000, 3400, 5000, and 7000. The graph features two solid lines representing the acceptable range and two dotted lines representing the preferred range. The upper solid line starts at 10 dB at 300 Hz, decreases to 3 dB at 1000 Hz, and then increases to 7 dB at 7000 Hz. The lower solid line starts at -10 dB at 300 Hz, increases to -3 dB at 500 Hz, remains at -3 dB until 1000 Hz, and then decreases to -18 dB at 7000 Hz. The preferred range is indicated by dotted lines at ±3 dB from the 0 dB reference line between 500 Hz and 3400 Hz.](58de972a8c79c238b69cbeeafcc8d5fb_img.jpg) + +| Frequency (Hz) | Upper Solid Line (dB) | Lower Solid Line (dB) | Upper Dotted Line (dB) | Lower Dotted Line (dB) | +|----------------|-----------------------|-----------------------|------------------------|------------------------| +| 300 | 10 | -10 | - | - | +| 500 | 5 | -3 | - | - | +| 1000 | 3 | -3 | 3 | -3 | +| 2000 | 4 | -9 | 3 | -3 | +| 3400 | 5 | -18 | 3 | -3 | +| 5000 | 6 | - | - | - | +| 7000 | 7 | - | - | - | + +Figure 3/P.370: Magnetic field strength characteristics. A graph showing the magnetic field strength (dB relative to level at 1000 Hz) versus frequency (Hz). The y-axis ranges from -20 to 20 dB, with major ticks at -20, -10, -3, 0, 3, 10, and 20. The x-axis ranges from 300 to 7000 Hz, with major ticks at 300, 500, 1000, 2000, 3400, 5000, and 7000. The graph features two solid lines representing the acceptable range and two dotted lines representing the preferred range. The upper solid line starts at 10 dB at 300 Hz, decreases to 3 dB at 1000 Hz, and then increases to 7 dB at 7000 Hz. The lower solid line starts at -10 dB at 300 Hz, increases to -3 dB at 500 Hz, remains at -3 dB until 1000 Hz, and then decreases to -18 dB at 7000 Hz. The preferred range is indicated by dotted lines at ±3 dB from the 0 dB reference line between 500 Hz and 3400 Hz. + +NOTE – Preferred frequency characteristics are within the dotted lines ( $\pm 3 \text{ dB}$ ). Range of acceptable characteristics is within the solid lines. + +FIGURE 3/P.370 +Magnetic field strength characteristics + +## 4.3 Probe coil + +### 4.3.1 Dimensions + +To minimize the loss of resolution when measuring the magnetic field strength, the following **maximum** dimensions are recommended for the calibrated coil: + +| | | | +|----------|----------------|--------------| +| Core: | length: | 13.5 mm | +| | cross-section: | 1.5 × 2.5 mm | +| Winding: | length: | 11 mm | +| | cross-section: | 2.5 × 3.5 mm | + +The winding shall be shorter than the core. + +#### NOTES + +1 The magnetic field may be non-homogeneous within distances comparable to the length of the probe coil. The introduction of magnetic core material may also redirect the magnetic field contours. Typically the sensitivity of the probe coil will increase with frequency at 6 dB/octave. + +2 The probe coil may be combined with frequency correcting elements to obtain a flat frequency response over the range of interest. + +3 The North American Standard EIA/TIA RS 504 [3], as included in the FCC Rules, Part 68.316 [4], refers to a commercially available probe coil meeting the requirements of this Recommendation. + +### 4.3.2 Calibration of the probe coil + +In order to calibrate the probe, a homogeneous magnetic field of known intensity shall be available. The magnetic field strength at the centre of a square loop of one turn with a side of “a” metres and carrying a current of “i” amperes is given by: + +$$H = \frac{2\sqrt{2}}{\pi} \cdot \frac{i}{a} \quad A/m$$ + +The dimension “a” should be 0.5 m or more to ensure that the field at the centre is sufficiently well defined in magnitude and direction. + +In practice, it may be advantageous to construct the loop having several turns to reduce the current from the source. Essentially, constant current conditions should be maintained over the test frequency range, for example driving the coil from a low impedance generator through a series resistor having at least 100 times the impedance of the coil over the frequency range of interest. If the current drive is monitored during the calibration process, any variations can be taken into account when deriving the probe coil sensitivity. + +The test space shall be remote from any field disturbing magnetic material or other material in which eddy currents can be induced, so causing a field disturbance. + +The sensitivity as a function of frequency of the probe coil shall be measured as the induced voltage over both leads of the probe coil with an accuracy of $\pm 0.5$ dB. This voltage is the standard in relation to the applied current per metre and shall be used for the measurement of the requirements specified in 4.2.2. + +The total harmonic distortion of the magnetic field shall be less than 1%. + +NOTE – Further helpful information is given in IEC Publication 118-1. + +### 4.3.3 Distortion + +The total harmonic distortion of the probe coil shall be less than 2%, when measuring the field strength up to +2 dB relative to 1 A/m. + +# **5 Characteristics of telephone sets that provide additional amplification for the benefit of hearing impaired users** + +## **5.1 Introduction** + +This clause specifies the requirements for telephones with receive amplification intended to aid the hearing impaired. A significant proportion of the population is disabled by varying degrees of hearing loss, often associated with a reduction of speech discrimination ability. Amplification can only replace the lost sensitivity. + +The group most likely to receive the greatest benefit from the use of a telephone with additional receive amplification are those with moderate to severe hearing losses in the range of 35 dB to 80 dB. + +In setting the numerical values in this Recommendation, consideration has been given to the fact that the sound pressure levels at the ear during a telephone conversation can be up to 30 dB above those occurring in normal face-to-face conversation (at one metre distance). As hearing impaired people do not necessarily have elevated thresholds of loudness discomfort, some form of output limitation will be required. Recent work has indicated that Automatic Gain Control (AGC) may provide a better automatic means of limitation than peak clipping. In addition, it is recognized that the frequency response to give maximum intelligibility to some hearing impaired people may require shaping. + +Two types of application are envisaged. Firstly, where the telephone will be used by persons with a range of hearing acuity, secondly, where the telephone is to be used in the main by a hearing impaired person. In the first, it may be preferable to arrange that the additional amplification is brought into use by the operation of a latching switch that automatically resets the gain to nominal level when the handset is replaced in its rest position. In the second case it will be advantageous to maintain the receive amplification (the level of which is selected by means of a volume control set by the hearing impaired user) when the handset is replaced. The use of voice switched attenuation, perhaps 10 dB to 12 dB, may be necessary to provide protection against instability, in particular for the first case, and could also improve the discrimination against the ambient noise received through the sidetone path. + +It is estimated that with the provision of additional amplification to the levels recommended below, possibly up to 80% of hearing impaired users would benefit even when not using their hearing aids to couple to the telephone set. If a hearing aid is worn in addition and inductive coupling is also provided, then the proportion of hearing impaired users who will be able to have satisfactory telephone conversations will increase further. It is, however, pointed out that with a high gain setting selected, the sound level and/or inductive field at the earphone may be considerably higher than normally experienced by hearing aid microphones and/or inductive pick-up coils and, on short telephone connections in particular, there is a very real danger of overloading the input stages of the hearing aid. + +The inclusion of additional receiving amplification does not reduce or replace existing technical standards that apply to a handset. Additional receiving amplification can be combined with other additional functionality, such as inductive coupling or additional earpieces, provided specifically for people with special needs. + +## **5.2 Sending characteristics** + +### **5.2.1 Sensitivity** + +It is recommended that when the user is talking, the Sending Loudness Rating (SLR) remains at a constant value set by the requirements of the national system, irrespective of any receive amplification selected. + +If voice switching is used in the interests of preserving stability margins and/or Terminal Coupling Loss (TCL) under difficult operating conditions, it is recommended that the switched attenuation be the minimum required for the purpose, e.g. approximately equal to that required to offset any increased receiving amplification selected. + +### **5.2.2 Frequency response** + +It is recommended that the sending frequency response is maintained according to the requirements of the national system irrespective of any increased receive amplification selected. + +## 5.3 Receiving characteristics + +### 5.3.1 Sensitivity + +In respect of the receive sensitivity, it is recommended that when no additional amplification is selected the Receiving Loudness Rating (RLR) requirements of the national system are met. + +With additional receive amplification selected, the RLR may be set up to 20 dB more negative (louder) than the nominal requirements of the national system. + +It shall be permissible to set the RLR up to 15 dB more positive (quieter) than the nominal requirements of the national system. + +If additional amplification is provided in association with voice switching functions then the principles outlined in 4.6/P.340 are recommended. The depth of switched attenuation should be kept to a minimum in the interests of good speech quality but should be sufficient to maintain stability, return loss and TCL requirements. + +### 5.3.2 Frequency response + +When no additional amplification is selected, it is recommended that the receive frequency response meets the requirements of the national system. + +With additional amplification selected, under some conditions and for particular hearing impaired users, it may be appropriate to provide special frequency shaping. For example, shaping could be provided to compensate for the loss of low frequencies caused by earcap leakage effects that occur with most types of telephone earphone. Earphones having low acoustic impedance are an exception to this (see Note 2, Foreword). An alternative shaping might be to provide additional gain at high frequencies to compensate for their loss over long local subscriber lines. Any frequency shaping provided could be by means of an additional control. + +NOTE – In practice, when coupling to a hearing aid inductively or acoustically, the low frequencies are normally missing, and for acoustic coupling to conventional earphones there is usually a slope of 12 to 18 dB per octave at frequencies below the first mechanical resonance, often around 1500 Hz. Some examples are shown in Appendix I. + +### 5.3.3 Restoring normal receive amplification + +**Option A** – Telephones to be used by any user + +At the termination of the telephone conversation any additional amplification selected shall be cancelled and the nominal RLR values restored by the action of replacing the handset to its rest position, or an equivalent action. + +**Option B** – Telephones to be used mainly by a hearing impaired user + +For the benefit of a hearing impaired user frequently using the terminal, the option may be provided for the additional amplification selected during a telephone conversation to be locked in position. If this option is provided, additional measures to ensure stability with maximum gain selected may be necessary. + +## 5.4 Sidetone + +It is recommended that sidetone levels (STMR, LSTR) indicated in Recommendation G.121 be provided when additional amplification is not in use. + +With additional amplification selected, it will normally only be possible to maintain recommended sidetone levels if voice switching techniques are used. + +# **6 Electrical coupling of telephone sets to hearing aids** + +## **6.1 Introduction** + +This clause covers the possibility of electrical coupling from a telephone set to a hearing aid. It is intended to be compatible with the IEC Publication 118-6, which covers the characteristics of electrical input ports for hearing aids. + +It is also envisaged that it could be used in other applications that require electrical connection to the PSTN, for example to couple into other non-telephony amplification systems such as public address, or recording systems. Another alternative would be the direct electrical connection to a hearing aid transducer or headset, given that the device had sufficient sensitivity. This clause may also find application for mobile apparatus and apparatus having cordless handsets that provide for electrical coupling to hearing aids. Some of the applications envisaged would need to address safety and/or protection issues and, although these and the applications themselves do not come within the scope of ITU Recommendations, attention is drawn to them here for information. + +## **6.2 Electrical characteristics** + +Unless otherwise stated all the requirements given below are for the volume control (if present) set to give the maximum output. + +### **6.2.1 Point of connection in the telephone set** + +It is recommended that the interconnection to the hearing aid be taken from the electrical path to the handset earphone so that the sidetone path is coupled into the hearing aid. It is also recommended that it shall not be possible to inject a signal into the telecommunication network from this point. + +### **6.2.2 Electrical characteristics at interconnection point** + +In what follows it is assumed that the load normally applied by a hearing aid input circuit will be 2000 ohms resistive. + +#### **6.2.2.1 Calibration of the acoustic receive level** + +Using the measurement configuration shown in Figure 1 for analogue telephones and Figure 2 for digital telephones, the drive level of the oscillator shall be adjusted to produce a sound level of $-14$ dBPa at 1000 Hz at the ERP (Ear Reference Point). This drive level shall be used for measuring the impedance, sensitivity and frequency characteristics recommended below. + +#### **6.2.2.2 Impedance** + +The modulus of the electrical source impedance between the frequencies of 300 and 4000 Hz shall be less than 1000 ohms for all settings of the volume control (if present). + +#### **6.2.2.3 Sensitivity** + +With the drive level of the oscillator set as in 6.2.2.1 above, at 1000 Hz, the electrical output at 1000 Hz shall be $-35 \pm 5$ dBV. With the volume control (if present) set to minimum the electrical output shall be between 15 dB and 30 dB lower than the output at the maximum setting. + +#### **6.2.2.4 Frequency characteristic** + +With the drive level of the oscillator set as in 6.2.2.1 above, the frequency response at the telephone electrical interconnection point, when loaded with $2\text{ k}\Omega$ , shall be equal to the sensitivity at 1000 Hz $\pm 3$ dB over the range 300 to 4000 Hz and $\pm 5$ dB over the range 100 to 300 Hz for all settings of the volume control (if present). + +#### **6.2.2.5 Noise** + +With the oscillator in 6.2.2.1 replaced by a 600 ohm resistor the noise level at the electrical interconnection point shall be less than $-70$ dBVp r.m.s. + +#### **6.2.2.6 Maximum output** + +With the oscillator in 6.2.2.1 set to give $+20$ dBV r.m.s. the maximum output at the electrical interconnection point shall be limited to $0$ dBV. + +#### **6.2.2.7 Insulation** + +In order to prevent direct electrical connection with the PSTN it is recommended that the interconnection point be electrically isolated by use of, for example, a transformer. It is recommended that the insulation provided should meet the requirements of the national system. + +## **6.3 Interconnection with the telephone set** + +### **6.3.1 Connector type** + +The connection shall be by means of a socket which is compatible with a miniature coaxial, stereo jack plug, type 130-8 IEC, conforming to IEC 268-11. Normally the socket shall be of 3.5 mm diameter. Alternatively, for compact telephone sets such as cordless or mobile telephones, the socket may be of 2.5 mm diameter. + +### **6.3.2 Connector position** + +It is recommended that the connector socket should be positioned on the telephone for high visibility and ease of inserting the plug. This means that the socket should be placed on the front edge, front top side or on the front sides of the subset housing. The socket should be placed so that the plug and lead will not interfere with the normal operation of the telephone, i.e. lifting/replacing the handset, dialling, inserting coins or cards, or any other functions. On public telephones it is important that the socket is not positioned on a horizontal surface making it susceptible to the ingress of fluids or other contaminants. Telephones and mobile telephones with cordless handsets shall not have the connector on the handset but on the body of the subset. + +One-piece telephones, cordless telephones and mobile telephones which have no separate handset, the socket should be positioned so that the plug and the lead will not interfere with the normal operation of the instrument, i.e. holding it, dialling, charging batteries, or any other function. + +## **6.4 Safety issues** + +It is recommended that other safety issues associated with the interconnection point should meet the requirements of the national system. + +# Annex A + +## Measurement of an acousto-magnetic adapter generating a magnetic field + +(This annex forms an integral part of this Recommendation) + +## A.1 Measurement procedures + +Measurements are made in accordance with this Recommendation. + +The output sound pressure level of the telephone receiver is measured against the artificial ear without the acousto-magnetic adapter being mounted. + +The characteristics of the magnetic field of the acousto-magnetic adapter are measured when mounted on the actual telephone receiver. + +NOTE – In reporting results, the type of telephone set used should be specified. + +## A.2 Magnetic field requirements + +The magnetic field produced by the adapter when fitted to a handset should meet the level and frequency characteristic requirements given in 4.2.1 and 4.2.2. + +## A.3 Physical properties + +Desirable physical properties of the acousto-magnetic adapter are: + +- easy to place on the earcap and remove again; +- a firm contact to the earcap so that the acousto-magnetic adapter and the telephone handset can be used as an integral unit; +- forming a good and well-defined acoustic coupling to the earcap (see Note); +- the surface of the acousto-magnetic adapter defining the adapter plane should be flat or should have a shape easily defining the adapter plane; +- the adapter plane should be approximately parallel to the earcap plane; +- the magnetic field produced by the adapter should be orientated so that the magnetic coupling to the hearing aid is only to a small extent dependent on the position of the hearing aid. + +NOTE – The inner diameter of an acoustic seal is recommended to be equal to the edge diameter of the IEC 318 artificial ear (see Recommendation P.57). + +# Appendix I + +## Examples of frequency characteristics of hearing aids coupled to telephones + +(This appendix does not form an integral part of this Recommendation) + +Figures I.1 to I.9 have been taken from [1] and illustrate the signals arriving at the Drum Reference Point (DRP) of a Head And Torso Simulator (HATS). Figure I.2 refers to the IRS (see Recommendation P.48) sending and receiving ends joined directly together and having an OLR by definition of 0 dB. All other figures are for a British Telecom Tribune set coupled to various UK hearing aids, (except where the Beocom telephone is mentioned) driven from the equivalent of the IRS send end with 10 dB additional loss. Thus the “connection” is what might be termed “typical” having an OLR approximating to 10 dB, i.e. similar to the ITU-T long-term aim for OLR. + +Note particularly the rising frequency characteristics under all coupled conditions. + +### Telephony + +| Sound level @ MRP | Loss | Sound level @ ERP | +|-------------------------|--------|----------------------------| +| 89.3 dB SPL (pure tone) | 1.1 dB | 88.2 dB SPL (1 kHz) | +| 88.8 dB SPL (Speech) | 5.5 dB | 83.3 dB SPL (0.2 to 4 kHz) | + +![Diagram of a telephone speech path showing Mouth, Handset (at MRP), IRS, Send, 10 dB loss, IRS, Rec., and Speech path.](2cde062fd82833415971a8bd1a2cafab_img.jpg) + +The diagram illustrates a telephone speech path. It starts with a 'Mouth' connected to a 'Handset' (labeled 'at MRP'). The signal then passes through an 'IRS' (Input Resistor) to a 'Send' block. A '10 dB' loss is indicated between the 'Send' block and another 'IRS' block. The signal then passes through the second 'IRS' to a 'Rec.' (Receiver) block, which is connected to a 'Speech path' (represented by a handset icon). The reference 'T1208470-96/d04' is noted at the bottom right. + +Diagram of a telephone speech path showing Mouth, Handset (at MRP), IRS, Send, 10 dB loss, IRS, Rec., and Speech path. + +### Metre air path + +| | | | +|---------------------------------------|---------|------------------------------------------------| +| 89.3 dB SPL (Speech)
(100-8000 Hz) | 30.2 dB | 59.1 dB SPL (Free field) | +| | 29.0 dB | 63.3 dB SPL
(including room & head effects) | +| 88.8 dB SPL (Speech)
(200-4000 Hz) | 28.9 dB | 62.9 dB SPL
(including room & head effects) | + +FIGURE I.1/P.370 + +**Speech path mouth to ear losses for a typical telephone connection, having an OLR of 10 dB, and a metre air path** + +![Graph comparing speech via a telephone (Loosely coupled, Sealed) and 1-metre air path across frequencies from 250 to 4000 Hz.](250cf77a1cd51989da09fca796b3e4ea_img.jpg) + +The graph plots dB SPL (Y-axis, 40 to 90) against Frequency (X-axis, 250 to 4000 Hz). It compares three speech transmission methods: + + +- Loosely coupled** (black line with diamond markers): Shows a rising trend from approximately 40 dB SPL at 250 Hz to a peak of about 88 dB SPL at 2500 Hz, followed by a sharp drop to 65 dB SPL at 4000 Hz. +- Sealed** (grey line with square markers): Shows a rising trend from approximately 60 dB SPL at 250 Hz to a peak of about 85 dB SPL at 630 Hz, then a slight decline to 65 dB SPL at 4000 Hz. +- 1 metre air path** (grey line with triangle markers): Shows a relatively flat trend, starting around 45 dB SPL at 250 Hz and ending around 40 dB SPL at 4000 Hz. + + The reference 'T1208480-96/d05' is noted at the bottom right. + +Graph comparing speech via a telephone (Loosely coupled, Sealed) and 1-metre air path across frequencies from 250 to 4000 Hz. + +FIGURE I.2/P.370 + +**Comparison of speech via a telephone and 1-metre air path** + +![Line graph showing dB SPL vs Frequency for BE17 over mic, BE17 over ear, and Room noise. The graph shows three data series: BE17 over mic (black diamonds), BE17 over ear (grey squares), and Room noise (grey triangles). The x-axis represents Frequency from 200 to 3150 Hz, and the y-axis represents dB SPL from 60 to 120 dB. BE17 over ear shows the highest output, peaking at approximately 115 dB SPL at 2000 Hz. BE17 over mic peaks at approximately 105 dB SPL at 2000 Hz. Room noise is relatively flat, peaking at approximately 85 dB SPL between 1250 and 1500 Hz.](b53846f262c6904a1b45abef2e95fbd8_img.jpg) + +| Frequency (Hz) | BE17 over mic (dB SPL) | BE17 over ear (dB SPL) | Room noise (dB SPL) | +|----------------|------------------------|------------------------|---------------------| +| 200 | 70 | 70 | 70 | +| 250 | 72 | 75 | 70 | +| 315 | 75 | 75 | 70 | +| 400 | 78 | 80 | 78 | +| 500 | 80 | 85 | 78 | +| 630 | 85 | 90 | 80 | +| 800 | 90 | 98 | 85 | +| 1000 | 98 | 105 | 85 | +| 1250 | 100 | 108 | 88 | +| 1500 | 102 | 109 | 85 | +| 2000 | 105 | 110 | 80 | +| 2500 | 108 | 115 | 83 | +| 3150 | 100 | 110 | 80 | +| 4000 | 78 | 88 | 70 | + +Line graph showing dB SPL vs Frequency for BE17 over mic, BE17 over ear, and Room noise. The graph shows three data series: BE17 over mic (black diamonds), BE17 over ear (grey squares), and Room noise (grey triangles). The x-axis represents Frequency from 200 to 3150 Hz, and the y-axis represents dB SPL from 60 to 120 dB. BE17 over ear shows the highest output, peaking at approximately 115 dB SPL at 2000 Hz. BE17 over mic peaks at approximately 105 dB SPL at 2000 Hz. Room noise is relatively flat, peaking at approximately 85 dB SPL between 1250 and 1500 Hz. + +- ◆— BE17 over mic. +- BE17 over ear +- ▲— Room noise + +FIGURE I.3/P.370 +**Acoustic coupling – effect of position** + +![Line graph showing dB SPL vs Frequency for Classic IE and Room noise. The graph shows two data series: Classic IE (black diamonds) and Room noise (grey squares). The x-axis represents Frequency from 200 to 3150 Hz, and the y-axis represents dB SPL from 60 to 120 dB. Classic IE shows a significant increase in output starting around 250 Hz, peaking at approximately 115 dB SPL at 1500 Hz, and then decreasing. Room noise is relatively flat, peaking at approximately 70 dB SPL between 1250 and 2000 Hz.](1a6a826cc13d4e964b7bda69508d78e6_img.jpg) + +| Frequency (Hz) | Classic IE (dB SPL) | Room noise (dB SPL) | +|----------------|---------------------|---------------------| +| 200 | 65 | 65 | +| 250 | 75 | 70 | +| 315 | 80 | 70 | +| 400 | 88 | 65 | +| 500 | 92 | 65 | +| 630 | 98 | 63 | +| 800 | 102 | 63 | +| 1000 | 105 | 63 | +| 1250 | 112 | 70 | +| 1500 | 115 | 70 | +| 2000 | 112 | 70 | +| 2500 | 108 | 68 | +| 3150 | 92 | 58 | +| 4000 | 85 | - | + +Line graph showing dB SPL vs Frequency for Classic IE and Room noise. The graph shows two data series: Classic IE (black diamonds) and Room noise (grey squares). The x-axis represents Frequency from 200 to 3150 Hz, and the y-axis represents dB SPL from 60 to 120 dB. Classic IE shows a significant increase in output starting around 250 Hz, peaking at approximately 115 dB SPL at 1500 Hz, and then decreasing. Room noise is relatively flat, peaking at approximately 70 dB SPL between 1250 and 2000 Hz. + +FIGURE I.4/P.370 +**Acoustic coupling – in the ear aid (Classic)** + +![Line graph showing dB SPL vs Frequency for Tribune KEMAR and Beocom KEMAR. The x-axis represents Frequency from 250 to 4000 Hz, and the y-axis represents dB SPL from 30 to 100. Tribune KEMAR (black diamonds) starts at ~35 dB SPL at 250 Hz, rises to ~95 dB SPL at 2000 Hz, and drops to ~72 dB SPL at 4000 Hz. Beocom KEMAR (grey squares) starts at ~55 dB SPL at 250 Hz, rises to ~98 dB SPL at 2500 Hz, and drops to ~85 dB SPL at 4000 Hz.](b2d16e07bfa79d67a8adabf7e26c7764_img.jpg) + +| Frequency (Hz) | Tribune KEMAR (dB SPL) | Beocom KEMAR (dB SPL) | +|----------------|------------------------|-----------------------| +| 250 | 35 | 55 | +| 315 | 45 | 68 | +| 400 | 55 | 75 | +| 500 | 60 | 78 | +| 630 | 68 | 80 | +| 800 | 75 | 81 | +| 1000 | 85 | 82 | +| 1250 | 90 | 83 | +| 1600 | 95 | 88 | +| 2000 | 95 | 98 | +| 2500 | 95 | 98 | +| 3150 | 90 | 98 | +| 4000 | 72 | 85 | + +Line graph showing dB SPL vs Frequency for Tribune KEMAR and Beocom KEMAR. The x-axis represents Frequency from 250 to 4000 Hz, and the y-axis represents dB SPL from 30 to 100. Tribune KEMAR (black diamonds) starts at ~35 dB SPL at 250 Hz, rises to ~95 dB SPL at 2000 Hz, and drops to ~72 dB SPL at 4000 Hz. Beocom KEMAR (grey squares) starts at ~55 dB SPL at 250 Hz, rises to ~98 dB SPL at 2500 Hz, and drops to ~85 dB SPL at 4000 Hz. + +- ◆— Tribune KEMAR +- Beocom KEMAR + +FIGURE I.5/P.370 +**Effect of low acoustic impedance on response** + +![Line graph showing dB SPL vs Frequency for Tribune/BE51/max and Beocom/BE51/Max. The x-axis represents Frequency from 200 to 3150 Hz, and the y-axis represents dB SPL from 80 to 130. Tribune/BE51/max (black diamonds) starts at ~80 dB SPL at 200 Hz, rises to ~125 dB SPL at 2000 Hz, and drops to ~100 dB SPL at 3150 Hz. Beocom/BE51/Max (grey squares) starts at ~88 dB SPL at 200 Hz, rises to ~128 dB SPL at 2500 Hz, and drops to ~108 dB SPL at 3150 Hz.](1c427123350e0e73e2a109b79069314b_img.jpg) + +| Frequency (Hz) | Tribune/BE51/max (dB SPL) | Beocom/BE51/Max (dB SPL) | +|----------------|---------------------------|--------------------------| +| 200 | 80 | 88 | +| 250 | 83 | 92 | +| 315 | 87 | 98 | +| 400 | 92 | 104 | +| 500 | 97 | 109 | +| 630 | 102 | 112 | +| 800 | 109 | 114 | +| 1000 | 115 | 117 | +| 1250 | 118 | 117 | +| 1600 | 118 | 117 | +| 2000 | 125 | 122 | +| 2500 | 128 | 128 | +| 3150 | 120 | 122 | +| 4000 | 100 | 108 | + +Line graph showing dB SPL vs Frequency for Tribune/BE51/max and Beocom/BE51/Max. The x-axis represents Frequency from 200 to 3150 Hz, and the y-axis represents dB SPL from 80 to 130. Tribune/BE51/max (black diamonds) starts at ~80 dB SPL at 200 Hz, rises to ~125 dB SPL at 2000 Hz, and drops to ~100 dB SPL at 3150 Hz. Beocom/BE51/Max (grey squares) starts at ~88 dB SPL at 200 Hz, rises to ~128 dB SPL at 2500 Hz, and drops to ~108 dB SPL at 3150 Hz. + +- ◆— Tribune/BE51/max +- Beocom/BE51/Max + +FIGURE I.6/P.370 +**Effect of low acoustic impedance on coupling to BE51 hearing aid** + +![Line graph showing dB SPL at DRP vs Frequency Hz for BE17 inductive coupling. The graph compares 'BE17 ind.' (black diamonds) and 'Room noise' (grey squares) across frequencies from 200 Hz to 3150 Hz. The BE17 ind. curve shows a peak around 1250-1500 Hz, while the Room noise curve is relatively flat.](3102c32204f998dba666e1e915d5babf_img.jpg) + +| Frequency Hz | BE17 ind. (dB SPL at DRP) | Room noise (dB SPL at DRP) | +|--------------|---------------------------|----------------------------| +| 200 | 65 | 48 | +| 250 | 73 | 51 | +| 315 | 79 | 47 | +| 400 | 86 | 48 | +| 500 | 91 | 49 | +| 630 | 96 | 56 | +| 800 | 101 | 57 | +| 1000 | 106 | 64 | +| 1250 | 108 | 71 | +| 1500 | 108 | 63 | +| 1800 | 107 | 62 | +| 2200 | 108 | 67 | +| 2800 | 105 | 62 | +| 3150 | 89 | 51 | + +T1208530-96/d10 + +Line graph showing dB SPL at DRP vs Frequency Hz for BE17 inductive coupling. The graph compares 'BE17 ind.' (black diamonds) and 'Room noise' (grey squares) across frequencies from 200 Hz to 3150 Hz. The BE17 ind. curve shows a peak around 1250-1500 Hz, while the Room noise curve is relatively flat. + +FIGURE I.7/P.370 +**Inductive coupling – BE17 aid over a typical connection** + +![Line graph showing dB SPL DRP vs Frequency Hz for inductively coupled Classic in the ear aid. The graph compares 'Classic ind./typical' (black diamonds) and '1-metre air path' (grey squares) across frequencies from 200 Hz to 3150 Hz. The Classic ind./typical curve shows a peak around 1500 Hz, while the 1-metre air path curve is relatively flat.](96a7eac66ef72bb016c280278506ac63_img.jpg) + +| Frequency Hz | Classic ind./typical (dB SPL DRP) | 1-metre air path (dB SPL DRP) | +|--------------|-----------------------------------|-------------------------------| +| 200 | 39 | 48 | +| 250 | 43 | 49 | +| 315 | 48 | 50 | +| 400 | 52 | 51 | +| 500 | 56 | 50 | +| 630 | 58 | 49 | +| 800 | 68 | 48 | +| 1000 | 76 | 46 | +| 1250 | 83 | 44 | +| 1500 | 84 | 43 | +| 1800 | 77 | 49 | +| 2200 | 72 | 53 | +| 2800 | 76 | 49 | +| 3150 | 74 | 45 | + +T1208540-96/d11 + +Line graph showing dB SPL DRP vs Frequency Hz for inductively coupled Classic in the ear aid. The graph compares 'Classic ind./typical' (black diamonds) and '1-metre air path' (grey squares) across frequencies from 200 Hz to 3150 Hz. The Classic ind./typical curve shows a peak around 1500 Hz, while the 1-metre air path curve is relatively flat. + +FIGURE I.8/P.370 +**Inductively coupled Classic in the ear aid over a typical connection** + +![Line graph showing dB SPL DRP vs Frequency Hz for BE31/ind./typical and 1-metre air path. The BE31/ind./typical curve peaks at 120 dB SPL DRP around 2000 Hz, while the 1-metre air path curve remains relatively flat around 50 dB SPL DRP.](5ed9189841659dfb01f809b8e3b21f74_img.jpg) + +| Frequency (Hz) | BE31/ind./typical (dB SPL DRP) | 1-metre air path (dB SPL DRP) | +|----------------|--------------------------------|-------------------------------| +| 200 | 80 | 50 | +| 315 | 95 | 50 | +| 500 | 105 | 50 | +| 800 | 115 | 50 | +| 1250 | 118 | 48 | +| 1600 | 118 | 45 | +| 2000 | 120 | 50 | +| 2500 | 120 | 55 | +| 3150 | 118 | 50 | +| 4000 | 102 | 45 | + +Line graph showing dB SPL DRP vs Frequency Hz for BE31/ind./typical and 1-metre air path. The BE31/ind./typical curve peaks at 120 dB SPL DRP around 2000 Hz, while the 1-metre air path curve remains relatively flat around 50 dB SPL DRP. + +FIGURE 1.9/P.370 +**Inductively coupled BE31 aid over a typical connection** + +# Appendix II + +## Bibliography + +(This appendix does not form an integral part of this Recommendation) + +- [1] BARNES (G.J.): What do we hear through the telephone?, *British Society of Audiologists*, Leeds, April 1996. +- [2] GLEISS, (N.): Preferred listening levels in telephony, *Tele Engl. Ed.*, No. 2, 1974. +- [3] EIA/TIA RS 504: *North American Standard*. +- [4] FCC Rules, Part 68.316. + + + +# ITU-T RECOMMENDATIONS SERIES + +- Series A Organization of the work of the ITU-T +- Series B Means of expression +- Series C General telecommunication statistics +- Series D General tariff principles +- Series E Telephone network and ISDN +- Series F Non-telephone telecommunication services +- Series G Transmission systems and media +- Series H Transmission of non-telephone signals +- Series I Integrated services digital network +- Series J Transmission of sound-programme and television signals +- Series K Protection against interference +- Series L Construction, installation and protection of cables and other elements of outside plant +- Series M Maintenance: international transmission systems, telephone circuits, telegraphy, facsimile and leased circuits +- Series N Maintenance: international sound-programme and television transmission circuits +- Series O Specifications of measuring equipment +- Series P Telephone transmission quality** +- Series Q Switching and signalling +- Series R Telegraph transmission +- Series S Telegraph services terminal equipment +- Series T Terminal equipments and protocols for telematic services +- Series U Telegraph switching +- Series V Data communication over the telephone network +- Series X Data networks and open system communication +- Series Z Programming languages \ No newline at end of file diff --git a/marked/P/T-REC-P.380-202207-I_PDF-E/raw.md b/marked/P/T-REC-P.380-202207-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..653941bad84439780fdd143bbf44965cb8690b6a --- /dev/null +++ b/marked/P/T-REC-P.380-202207-I_PDF-E/raw.md @@ -0,0 +1,321 @@ + + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.380** + +(07/2022) + +SERIES P: TELEPHONE TRANSMISSION QUALITY, +TELEPHONE INSTALLATIONS, LOCAL LINE +NETWORKS + +Voice terminal characteristics + +# --- **Electro-acoustic measurements on headsets** + +Recommendation ITU-T P.380 + +## ITU-T P-SERIES RECOMMENDATIONS + +## TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS + +| | | +|----------------------------------------------------------------------------------------------------|--------------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | P.10–P.19 | +| Voice terminal characteristics | P.30–P.39 | +| Reference systems | P.40–P.49 | +| Objective measuring apparatus | P.50–P.59 | +| Objective electro-acoustical measurements | P.60–P.69 | +| Measurements related to speech loudness | P.70–P.79 | +| Methods for objective and subjective assessment of speech quality | P.80–P.89 | +| Voice terminal characteristics | P.300–P.399 | +| Objective measuring apparatus | P.500–P.599 | +| Measurements related to speech loudness | P.700–P.709 | +| Methods for objective and subjective assessment of speech and video quality | P.800–P.899 | +| Audiovisual quality in multimedia services | P.900–P.999 | +| Transmission performance and QoS aspects of IP end-points | P.1000–P.1099 | +| Communications involving vehicles | P.1100–P.1199 | +| Models and tools for quality assessment of streamed media | P.1200–P.1299 | +| Telemeeting assessment | P.1300–P.1399 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | P.1400–P.1499 | +| Methods for objective and subjective assessment of quality of services other than speech and video | P.1500–P.1599 | + +For further details, please refer to the list of ITU-T Recommendations. + +## Recommendation ITU-T P.380 + +# Electro-acoustic measurements on headsets + +## Summary + +Recommendation ITU-T P.380 provides testing methods for headsets using the head and torso simulator. The Recommendation addresses the following topics: selection of artificial ears, classification of headsets, positioning of headsets on the head and torso simulator (HATS), test repeatability and contents of the measurement report. + +## History + +| Edition | Recommendation | Approval | Study Group | Unique ID* | +|---------|----------------|------------|-------------|---------------------------------------------------------------------------| +| 1.0 | ITU-T P.380 | 2003-11-13 | 12 | 11.1002/1000/7040 | +| 2.0 | ITU-T P.380 | 2022-07-29 | 12 | 11.1002/1000/15003 | + +## Keywords + +Artificial ear, head and torso simulator, headsets, testing methods. + +--- + +\* To access the Recommendation, type the URL in the address field of your web browser, followed by the Recommendation's unique ID. For example, . + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents/software copyrights, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the appropriate ITU-T databases available via the ITU-T website at . + +© ITU 2022 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## Table of Contents + +| | Page | +|---------------------------------------------------|------| +| 1 Scope ..... | 1 | +| 2 References..... | 1 | +| 3 Definitions ..... | 1 | +| 3.1 Terms defined elsewhere ..... | 1 | +| 3.2 Terms defined in this Recommendation..... | 1 | +| 4 Abbreviations and acronyms ..... | 2 | +| 5 Conventions ..... | 2 | +| 6 Types of headsets..... | 2 | +| 6.1 Receiver ..... | 2 | +| 6.2 Microphone..... | 2 | +| 7 Artificial ear selection..... | 3 | +| 8 Positioning and measurement of the headset..... | 3 | +| 8.1 Measurements..... | 3 | +| 8.2 Positioning..... | 3 | +| 9 Test repeatability ..... | 4 | +| 10 Measurement report..... | 5 | +| Bibliography..... | 6 | + + + +# Electro-acoustic measurements on headsets + +## 1 Scope + +This Recommendation is the result of a study held within ITU-T for defining the electro-acoustic testing methodologies for headsets, which provide the best correlation with the performance of headsets in real use, when using the couplers currently recommended in [ITU-T P.57]. + +The results of this round robin test, aiming to compare the acoustic behaviour of headsets placed on humans and on the head and torso simulator (HATS), can be found in the Bibliography. + +The recommended test methodology is based on the use of HATS, as this is the best approximation of acoustical conditions occurring in the real use of headsets. + +This Recommendation focuses specifically on headsets and overrules [ITU-T P.57] regarding the applicability rules of artificial ears to specific receivers, as long as these devices belong to the headsets. + +This Recommendation is complementary to the relevant [ITU-T P.64], [ITU-T P.79], etc., which specify the electro-acoustic and telephonometric testing methods that are applicable to the telephone devices. + +The recommendation of performance descriptors, such as masks or limit values, is left to the relevant performance standards. + +## 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- [ITU-T P.57] Recommendation ITU-T P.57 (2021), *Artificial ears*. +- [ITU-T P.64] Recommendation ITU-T P.64 (2022), *Determination of sensitivity/frequency characteristics of local telephone systems*. +- [ITU-T P.79] Recommendation ITU-T P.79 (2007), *Calculation of loudness ratings for telephone sets*. + +## 3 Definitions + +### 3.1 Terms defined elsewhere + +None. + +### 3.2 Terms defined in this Recommendation + +This Recommendation defines the following terms: + +- 3.2.1 CL: Centre of lips of head and torso simulator. + +**3.2.2 headset:** Apparatus for telephony comprising essentially a "hands-free" handset which is typically secured to the head of the wearer. It includes a telephone microphone and a telephone receiver. + +**3.2.3 recommended test position (RTP):** Corresponds to the position in which the headset should be placed on HATS, e.g., as instructed by the manufacturer. In all the cases, the RTP should resemble the RWP on humans. + +**3.2.4 recommended wearing position (RWP):** Corresponds to the position in which a headset should be placed on humans according to the intended use (e.g., as instructed by the manufacturer in the user manual, etc.). + +## **4 Abbreviations and acronyms** + +This Recommendation uses the following abbreviations and acronyms: + +| | | +|------|------------------------------| +| HATS | Head And Torso Simulator | +| RTP | Recommended Test Position | +| RWP | Recommended Wearing Position | +| RLR | Receive Loudness Rating | +| STMR | Sidetone Masking Rating | + +## **5 Conventions** + +None. + +## **6 Types of headsets** + +This Recommendation covers the headsets equipped with receivers and microphones as defined below. + +### **6.1 Receiver** + +With reference to the definitions provided in [ITU-T P.57], the following receiver types are covered: + +- Insert-type; +- Intra-concha; +- Supra-concha; +- Supra-aural. + +### **6.2 Microphone** + +This Recommendation covers headsets equipped with microphones positioned as follows: + +- In front of and around the mouth (e.g., long booms); +- At the cheek (e.g., short booms); +- At the ear; +- At the neck (e.g., hanging down or collar clip); +- At the chest (e.g., hanging down or clip). + +## 7 Artificial ear selection + +Type 4.3 artificial ear is preferred for insert-type headsets, intra-concha headsets, supra-concha headsets and supra-aural headsets. + +Type 4.4 artificial ear is preferred for insert-type headsets1, intra-concha headsets, supra-concha headsets and Supra-aural headsets. + +Type 3.3 artificial ear may be used for insert-type headsets2, intra-concha headsets, supra-concha headsets and supra-aural headsets. + +Type 3.4 artificial ear may be used for insert-type headsets2 and intra-concha3 headsets. + +In case artificial ears of Type 3 are used and obtain different measurement results against Type 4 artificial ears, the results from Type 4 artificial ear shall take precedence. + +## 8 Positioning and measurement of the headset + +### 8.1 Measurements + +The positioning methodology provided in this Recommendation has been validated [b-ITU-T SG12 Cont54], [b-ITU-T SG12 Cont47], and shall be used, for measuring the frequency response and loudness ratings [ITU-T P.79] for both receive and send directions. + +This positioning of the headset on the HATS is, however, also recommended for performing measurements related to other electro-acoustic parameters such as sidetone masking rating (STMR), D-factor, distortion, etc. + +Furthermore, special conditions apply to headsets in some respects. + +Some headsets are binaural devices. In case the required receive loudness rating (RLR) is only specified for a monaural device, the corresponding required RLR for a binaural device should be 6 dB higher (for each of the receivers measured separately). In a similar way, other receive related requirements should be adapted for binaural devices. + +### 8.2 Positioning + +For a given headset, one single position shall be defined to test all the electro-acoustic parameters on a given HATS. + +The reasons for this provision are the following: + +- This is close to what happens in real life, so this procedure is consistent with the basic goal of the Recommendation, e.g., resembling the real-use conditions; +- It is simple; +- It allows for sidetone (and other related) measurements to be performed. + +The positioning of the headset shall reflect the way it is intended by the manufacturer to be used in a real situation. + +--- + +1 The Type 4.4 artificial ear simulator cannot be used for insert-type headsets where the acoustic outlet of the receiver is placed beyond the reference plane in the ear canal of the Type 4.4 artificial ear. + +2 It should be noted that, in some cases, the use of Type 3.3 or Type 3.4 ear simulator with an insert headset could provide a seal greater than that occurring on humans, leading to an overestimation of the level of the frequency response curve in the low frequency region (below 1 kHz). This is especially true for headsets relying on a close sealing with the ear canal. + +3 Only for intra-concha headsets where the acoustic outlet of the receiver face towards the ear canal the Type 3.4 artificial ear can be used. + +For this reason, the manufacturer's user's guide should provide a recommended wearing position describing in a precise way how the device should be placed on the user's head. + +From this recommended wearing position (RWP), the recommended test position shall be derived, as close as possible to the RWP. + +Also, the RTP description should hopefully be provided by the manufacturer and shall state in which way the receiving part of the headset shall be placed against or inside the ear simulator, and describe the positioning of the microphone, once the receiver(s) of the headset has been placed on the ear simulator. + +The exact positioning of the microphone shall be specified by using geometric coordinates relative to the centre of the lips. + +NOTE – A good laboratory practice may refer to [ITU-T P.64], which defines "a set of Cartesian axes with origin at CL, the centre of the lips", and further defines the axes as follows: + +- x-axis: horizontal axis of the mouth, with positive direction into the mouth; +- y-axis: horizontal, perpendicular to the x-axis, with positive direction towards the side of the mouth on which the handset is held; +- z-axis: vertical, with positive direction upwards. + +The way in which the coordinates are defined and checked is left to the choice of the manufacturer, or of the test lab in case the RTP is not defined by the manufacturer, but it should be noted that the closer the microphone is to the mouth, the more sensitive the results will be to any inaccuracy of the geometrical positioning. + +As an essential complementary information, the manufacturer shall state what is the recommended orientation of the microphone towards the mouth. + +In the case where no RWP information is available for the headset under test, then a suitable RWP and the derived RTP shall be defined by the testing lab, aiming at best-guessing the expected real-use position. All the relevant parameters of these assumed positions shall be reported together with the measurement results. + +## 9 Test repeatability + +Due to the sensitivity of the test results to the headset positioning, the tests shall be repeated at least five times by completely repositioning the headset, following the rules described in this Recommendation. + +The test report shall individually provide the test results for each repetition, plus any additional statistical analysis as required. + +NOTE 1 – It is recommended that the test operators get acquainted with the specific headset characteristics by running some preliminary learning test sessions. + +NOTE 2 – The use of positioning jigs can improve the test repeatability so that no repetition is required. This generally applies to all tests from test repetitions carried out by a single operator to tests carried out at different laboratories. However, the jig must be customized or adjusted for the particular type of headset under test, fully respecting the positioning principles outlined for the RTP. The purpose of the jig is to reduce the variability of test results, especially for microphones, without changing the mean of the test results. Thus, care must be taken to validate the designed jig for its intended use, and it should be verified that the jig does not introduce a bias to the results (i.e., the jig should be verified by comparing the mean of results obtained with and without the jig). + +## **10 Measurement report** + +The following information shall be reported additionally to the generally requested information: + +- Precise description of the positioning of the headset used for the tests, along with pictures when relevant. +- The number of measurements taken into account in the calculation of the statistic parameters. + +## Bibliography + +- [b-ITU-T P.58] Recommendation ITU-T P.58 (2021), *Head and torso simulator for telephonometry*. +- [b-ITU-T SG12 Cont54] ITU-T SG12 Study Period 2001 Contribution 54 (2003), *Conclusions of the Round Robin Test on headsets: Receiving side*. +<> +- [b-ITU-T SG12 Cont47] ITU-TSG12 Study Period 2001 Contribution 47 (2002), *Rapporteur report – Workshop on Headsets/Sending Part*. +<> + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | Tariff and accounting principles and international telecommunication/ICT economic and policy issues | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling, and associated measurements and tests | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects, next-generation networks, Internet of Things and smart cities | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.382-202303-I_PDF-E/raw.md b/marked/P/T-REC-P.382-202303-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..0ba58703cdc024a941aba82462066f1769822edd --- /dev/null +++ b/marked/P/T-REC-P.382-202303-I_PDF-E/raw.md @@ -0,0 +1,996 @@ + + +# Recommendation**ITU-T P.382 (03/2023)** + +SERIES P: Telephone transmission quality, telephone installations, local line networks + +Voice terminal characteristics + +--- + +**Technical requirements and test methods for analogue wired multi-microphone headsets or headphones and corresponding universal interface of terminals** + +![ITU logo](0538daaa5583c23e17db3a12f2281a55_img.jpg) + +The logo of the International Telecommunication Union (ITU) is located in the bottom right corner. It features the letters "ITU" in a bold, blue, sans-serif font, superimposed on a stylized globe icon with intersecting lines. + +ITU logo + +## ITU-T P-SERIES RECOMMENDATIONS + +# TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS + +| | | +|----------------------------------------------------------------------------------------------------|--------------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | P.10–P.19 | +| Voice terminal characteristics | P.30–P.39 | +| Reference systems | P.40–P.49 | +| Objective measuring apparatus | P.50–P.59 | +| Objective electro-acoustical measurements | P.60–P.69 | +| Measurements related to speech loudness | P.70–P.79 | +| Methods for objective and subjective assessment of speech quality | P.80–P.89 | +| Voice terminal characteristics | P.300–P.399 | +| Objective measuring apparatus | P.500–P.599 | +| Measurements related to speech loudness | P.700–P.709 | +| Methods for objective and subjective assessment of speech and video quality | P.800–P.899 | +| Audiovisual quality in multimedia services | P.900–P.999 | +| Transmission performance and QoS aspects of IP end-points | P.1000–P.1099 | +| Communications involving vehicles | P.1100–P.1199 | +| Models and tools for quality assessment of streamed media | P.1200–P.1299 | +| Telemeeting assessment | P.1300–P.1399 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | P.1400–P.1499 | +| Methods for objective and subjective assessment of quality of services other than speech and video | P.1500–P.1599 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +# Recommendation ITU-T P.382 + +# Technical requirements and test methods for analogue wired multi-microphone headsets or headphones and corresponding universal interface of terminals + +## Summary + +Recommendation ITU-T P.382 specifies critical physical and electroacoustic characteristics for a universal headset interface with more than four terminals and provides corresponding test methods. Headset or headphone interfaces of diameter 3.5 mm and 2.5 mm have been widely used in digital mobile terminals during recent years. Nowadays, the consumer is free to choose either the headset or HP originally provided by the terminal manufacturer or others that are offered separately. However, the quality of service or quality of experience perceived by users is influenced by both the electrical performance of the interface and the compatibility between the terminal and the connected headset or headphone. + +## History + +| Edition | Recommendation | Approval | Study Group | Unique ID* | +|---------|----------------|------------|-------------|---------------------------------------------------------------------------| +| 1.0 | ITU-T P.382 | 2016-07-29 | 12 | 11.1002/1000/12971 | +| 2.0 | ITU-T P.382 | 2020-10-14 | 12 | 11.1002/1000/14466 | +| 3.0 | ITU-T P.382 | 2023-03-01 | 12 | 11.1002/1000/15465 | + +## Keywords + +Connector, headphone, headset, interface, terminals. + +--- + +\* To access the Recommendation, type the URL in the address field of your web browser, followed by the Recommendation's unique ID. For example, . + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had received notice of intellectual property, protected by patents/software copyrights, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the appropriate ITU-T databases available via the ITU-T website at . + +© ITU 2023 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## Table of Contents + +| | Page | +|----------------------------------------------------------------------------------------|------| +| 1 Scope ..... | 1 | +| 2 References..... | 1 | +| 3 Definitions ..... | 2 | +| 3.1 Terms defined elsewhere ..... | 2 | +| 3.2 Terms defined in this Recommendation ..... | 3 | +| 4 Abbreviations and acronyms ..... | 3 | +| 5 General description ..... | 4 | +| 6 Physical characteristics ..... | 4 | +| 6.1 General rules ..... | 4 | +| 6.2 Pin assignments ..... | 4 | +| 7 Terminal electrical interface specification..... | 5 | +| 7.1 Communication mode (terminal) ..... | 5 | +| 7.2 Multimedia playback (terminal) ..... | 8 | +| 7.3 Multimedia record (terminal) ..... | 9 | +| 8 Headset specification ..... | 12 | +| 8.1 Communication mode (headset)..... | 12 | +| 8.2 Multimedia playback mode (headset) ..... | 14 | +| 8.3 Multimedia record mode (headset)..... | 14 | +| 9 Function requirements for terminals with the universal headset interface ..... | 16 | +| Annex A – Interpolation method for diffuse-field correction ..... | 17 | +| Appendix I – Audio connectivity for sockets with five contact points ..... | 18 | +| I.1 Plug connector of diameter 2.5 mm with five poles..... | 18 | +| I.2 Socket connector of diameter 2.5 mm with five contact points ..... | 18 | +| I.3 Plug connector of diameter 3.5 mm with five poles..... | 18 | +| I.4 Socket connector of diameter 3.5 mm with five contact points ..... | 19 | +| Appendix II – Audio connectivity for sockets with six contact points..... | 20 | +| Appendix III – Audio connectivity for sockets with fewer than five contact points..... | 21 | +| Appendix IV – Other considerations ..... | 22 | +| Bibliography ..... | 23 | + + + +# Recommendation ITU-T P.382 + +# Technical requirements and test methods for analogue wired multi-microphone headsets or headphones and corresponding universal interface of terminals + +# 1 Scope + +This Recommendation specifies electrical requirements and test methods for a multi-microphone headset or headphone (HP) interface used in digital mobile terminals. It is also applicable to terminals with a digital headset interface combined with an external adapter for an analogue (universal) headset interface. The combination is measured as one device. + +The principle of this Recommendation is to ensure adequate compatibility between the digital mobile terminal and the universal and multi-microphone wired analogue headset or HP, and to provide better user experience. The universal multi-microphone headset or HP interface enables the separation of sales between digital mobile terminals and headsets or HPs. One benefit is that users are free to choose their favourite type of headset or HP in the market, which in the long run will reduce E-waste. + +Furthermore, the universal interface can be used as the electric coupling design in hands-free systems and hearing aids for wider harmonization. + +To provide instructions to manufacturers and encourage them to adopt the universal headset interface, the mechanical dimensions are shown in an appendix for at least configurations with five poles. + +This Recommendation applies to digital mobile terminals with a physical analogue audio output and input interface that needs to support headsets with more than one microphone. Other similar information and communication technology equipment may also lie within the scope of this Recommendation. + +This Recommendation does not apply to terminals designed solely for digital headset or HP usage. + +# 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- [ITU-T P.56] Recommendation ITU-T P.56 (2011), *Objective measurement of active speech level*. +- [ITU-T P.58] Recommendation ITU-T P.58 (2023), *Head and torso simulator for telephonometry*. +- [ITU-T P.64] Recommendation ITU-T P.64 (2022), *Determination of sensitivity/frequency characteristics of local telephone systems*. +- [ITU-T P.381] Recommendation ITU-T P.381 (2023), *Technical requirements and test methods for analogue wired headsets or headphones and corresponding universal interface of terminals*. +- [ITU-T P.501] Recommendation ITU-T P.501 (2020), *Test signals for use in telephony and other speech-based applications*. +- [IEC 60268-1] IEC 60268-1:1985, *Sound system equipment – Part1: General*. + +- [IEC 61260-1] IEC 61260-1:2014, *Electroacoustics – Octave-band and fractional-octave-band filters – Part I: Specifications*. +- [EN 50332-1] CENELEC EN 50332-1:2013, *Sound system equipment: Headphones and earphones associated with personal music players – Maximum sound pressure level measurement methodology – Part 1: General method for “one package equipment”*. +- [EN 50332-2] CENELEC EN 50332-2:2013, *Sound system equipment: Headphones and earphones associated with personal music players – Maximum sound pressure level measurement methodology – Part 2: Matching of sets with headphones if either or both are offered separately, or are offered as one package equipment but with standardised connectors between the two allowing to combine components of different manufacturers or different design*. + +# 3 Definitions + +## 3.1 Terms defined elsewhere + +This Recommendation uses the following terms defined elsewhere: + +**3.1.1 artificial ear** [ITU-T P.381]: A device that incorporates an acoustic coupler and a calibrated microphone for measuring sound pressure, and which has an overall acoustic impedance similar to that of the average adult ear over a given frequency band. + +NOTE – Definition based on that in [b-ITU-T P.10]. + +**3.1.2 codec** [ITU-T P.381]: Combination of an analogue-to-digital encoder and a digital-to-analogue decoder operating in opposite directions of transmission in the same equipment. + +NOTE – This definition is based on that in [b-ITU-T G.701]. + +**3.1.3 eardrum reference point (DRP)** [b-ITU-T P.10]: A point located at the end of the ear canal, corresponding to the eardrum position. + +**3.1.4 earphone** [b-IEC 60268-7]: Electroacoustic transducer by which acoustic oscillations are obtained from electric signals and intended to be closely coupled acoustically to the ear. + +**3.1.5 head and torso simulator (HATS) for telephonometry** [ITU-T P.381]: A manikin that extends downwards from the top of the head to the waist. It is designed to simulate the sound pick-up characteristics and the acoustic diffraction produced by the average adult, and to reproduce the acoustic field generated by the human mouth. + +NOTE – Definition is based on that in [b-ITU-T P.10]. + +**3.1.6 headphone (HP)** [ITU-T P.381]: An object based on the assembly of one or two earphones on a headband or chinband, the use of which may be optional (e.g., with intra-concha earphones). + +NOTE – Definition based on that in [b-IEC 60268-7]. + +**3.1.7 headset** [b-ITU-T P.10]: A device which includes telephone receiver and transmitter which is typically secured to the head or the ear of the wearer. + +**3.1.8 mean opinion score-listening-only quality objective (MOS-LQO)** [ITU-T P.381]: A score calculated by means of an objective model that aims to predict the quality for a listening-only test situation. + +NOTE 1 – Objective measurements made using the model given in [b-ITU-T P.863] give results in terms of MOS-LQO. + +NOTE 2 – This definition is based on that given in [b-ITU-T P.800.1]. + +**3.1.9 mouth reference point (MRP)** [b-ITU-T P.10]: Point 25 mm in front of and on the axis of the lip plane of the artificial mouth or a typical human mouth (see Figure A.1 of [ITU-T P.64]). + +**3.1.10 Receive** [ITU-T P.381]: The receiving direction of the signal transmission, usually from the measurement system to the device under test. + +**3.1.11 Send** [ITU-T P.381]: The sending direction of the signal transmission, usually from the device under test to the measurement system. + +## **3.2 Terms defined in this Recommendation** + +None. + +# **4 Abbreviations and acronyms** + +This Recommendation uses the following abbreviations and acronyms: + +| | | +|---------|-----------------------------------------------------| +| DC | Direct Current | +| DRP | ear Drum Reference Point | +| DUT | Device Under Test | +| EL | Echo Loss | +| FB | Fullband | +| FFT | Fast Fourier Transform | +| GND | Ground | +| HATS | Head and Torso Simulator | +| HP | Headphone | +| HTCL | Headset Terminal Coupling Loss | +| HTCLw | weighted Headset Terminal Coupling Loss | +| L | Left | +| LQO | Listening only Quality Objective | +| MIC | Microphone | +| MIC1 | primary Microphone | +| MIC2 | secondary Microphone | +| MOS-LQO | Mean Opinion Score-Listening only Quality Objective | +| MRP | Mouth Reference Point | +| NB | Narrowband | +| POI | Point of Interconnection | +| R | Right | +| RMS | Root Mean Square | +| SWB | Super-wideband | +| TCL | Terminal Coupling Loss | +| TCLw | weighted Terminal Coupling Loss | +| THD | Total Harmonic Distortion | +| WB | Wideband | + +# 5 General description + +Generally, if a headset or HP is used, the overall user experience during a call highly depends on both the terminal and the connected headset or HP. Although the acoustic quality of the headset or HP is usually the weak link, more consideration with regard to the physical and electrical performance of the universal interface is needed. + +This Recommendation specifies new connection points that are compatible with the universal concentric connector interface for successful interconnection of the digital mobile terminal and the headset or HP, including the plug connector and the socket connector. Normally, the socket connector is fixed inside the terminal, with the outside rim of the socket level with the surrounding shell of the terminal. + +# 6 Physical characteristics + +## 6.1 General rules + +Two types of concentric socket connectors of different diameters are recommended for use: 2.5 mm; and 3.5 mm. An isometric view of the plug and socket connectors is shown in Figure 6-1 of [ITU-T P.381]. + +If the terminal is equipped with a headset interface designed for both communication and audio playback and this multi-microphone interface is utilized, the fixed connector shall be a 3.5 mm diameter or 2.5 mm diameter concentric socket connector. Detailed dimensions for socket and the plug connectors with more than four contact points are given in Appendices I to III. Some terminals with curved edges may not work well with the dimensions given in Appendices I to III. + +NOTE – The contact points here do not include special points reserved for other functions. + +## 6.2 Pin assignments + +This clause gives an illustration of pin assignments of the socket connector with five contact points and those of the mated plug, as shown in Figure 1. + +![Diagram of a socket connector with five contact points showing pin assignments and plug details.](b0211cee4b20034939d883ac0d70f696_img.jpg) + +The diagram illustrates the pin assignments for a socket connector with five contact points. On the left, five pins are shown with their respective electrical connections: pin 4 is connected to the top terminal, pin 6 to the second terminal, pin 2 to the third terminal, pin 1 to the fourth terminal, and pins 3 and 5 are connected to the bottom terminal. On the right, a cross-sectional view of a plug is shown, detailing its components: Tip, Ring 1, Alternative Ring 2, Ring 3, and Sleeve isolation. The diagram is labeled P.382(16)\_F01. + +Diagram of a socket connector with five contact points showing pin assignments and plug details. + +**Figure 1 – Pin assignments of a socket connector with five contact points** + +The physical pinout order of the universal interface is important and should coordinate with the connected headset or HP. + +A socket connector with five contact points shall be compatible with both the three pole and four pole plugs specified in [ITU-T P.381]. + +It is recommended that terminals be able to identify different plugs intelligently and automatically. + +It is recommended that a normally open switch be used on the socket connector to detect when the plug is fully inserted. + +### 6.2.1 Recommended pin assignment + +Point 1 of the socket is to be connected to the tip of the plug, linking it to the left (L) audio channel of the receiver. Point 2 is to be connected to ring 2, linking it to the transducer secondary microphone+ (MIC2+) (right (R) recording channel) of the device under test (DUT). Point 3 is to be connected to ring 1, linking it to the R audio channel of the receiver. Points 4 and 5 are to be connected to the sleeve, linking it to ground (GND). Point 6 is to be connected to ring 3, linking it to the transducer primary microphone+ (MIC1+) (L recording channel) of the DUT. + +Referring to Figure 1: + +- 1 is the contact point of the tip, linking it to the L channel of the receiver (HP L audio); +- 2 is the contact point of ring 2, linking it to the transducer (MIC2+); +- 3 is the contact point of ring 1, linking it to the R channel of the receiver (HP R audio); +- 4 is the contact point of the sleeve, linking it to the GND; +- 5 is the bushing of the socket, linking it to the GND when it is made of conductive material; +- 6 is the contact point of ring 3, linking it to the transducer (MIC1+). + +In summary, the pole order from the tip to the sleeve of the headset plug is HP L/MIC2/HP R/MIC1/GND. + +The pinout order of HP L/MIC2/HP R/MIC1/GND has the advantage of electrostatic discharge protection and allows for both plastic and metallic converters. + +### 6.2.2 Alternative pin assignment + +Referring to Figure 1: + +- 1 is the contact point of the tip, linking it to the L channel of the receiver (HP L audio); +- 2 is the contact point of ring 1, linking it to the R channel of the receiver (HP R audio); +- 3 is the contact point of ring 2, linking it to the transducer (MIC2+); +- 4 is the contact point of the sleeve, linking it to the transducer (MIC1+); +- 5 is the bushing of the socket, linking it to the transducer (MIC1+); +- 6 is the contact point of ring 3, linking it to GND when it is made of conductive material. + +In summary, the alternative pole order from the tip to the sleeve of the headset plug is: HP L/HP R/MIC2/GND/MIC1. + +# 7 Terminal electrical interface specification + +## 7.1 Communication mode (terminal) + +Test set-ups and requirements in this mode are harmonized with the existing [ITU-T P.381] specification, so only the general test set-up is repeated here. See [ITU-T P.381] for requirements and specific test set-ups. + +### 7.1.1 Test set-up for the communication mode (terminal) + +Test set-ups are shown in Figures 7-1 and 7-2 of [ITU-T P.381]. + +#### 7.1.1.1 Input and output characteristics of the test system for connection to the headset connector + +The output of the test system connected to the interface in Send of the headset connector must be direct current (DC) resistant. The output impedance shall be between 1 $\Omega$ and 10 k $\Omega$ . The dynamic range of the test system shall be consistent with (or exceed) the level range provided by headset MICs. + +The input of the test system connected to the receiving interfaces of the headset connectors shall have an input impedance of 32 $\Omega$ . The dynamic range shall be consistent with (or exceed) the output level range provided by the electrical output of the digital mobile terminal headset outputs. + +The common GND impedance (between sending and receiving sides) for the test system shall be $\leq 0.05 \Omega$ . + +#### 7.1.1.2 Test signals and test signal levels + +If not specified otherwise, fullband (FB) real speech signals specified in [ITU-T P.501] are used for the measurements. Detailed information about the test signal used is specified in the corresponding clause of this Recommendation. + +All test signals used in Receive tests have to be band limited with $\geq 24$ dB/octave roll off, achieved by bandpass filter. The signal should be band limited between: + +- in narrowband (NB) mode, 100 Hz and 4 kHz; +- in wideband (WB) mode, 100 Hz and 8 kHz; +- in super-wideband (SWB) mode, 50 Hz and 16 kHz; +- in FB mode, 20 Hz and 20 kHz. + +For real speech, the test signal levels are in accordance with the ITU-T P.56 active speech level of the (band limited in receiving direction) test signal, calculated over the complete test sequence, if not specified otherwise. Other test signal levels are calculated as the average level over the complete test sequence length of the test signals. For the receiving direction, the signal should be band limited. + +If not stated otherwise, the nominal average signal levels for the measurements are as follows: + +- $-16$ dBm0 (absolute signal power level, in decibels, relative to a point of zero relative level) in Receive; +- $-60$ dBV (decibels relative to 1 V) in Send [typical equivalent signal level of a microphone (MIC) of the DUT corresponding to $-4.7$ dBPa (decibels at 1 Pa) at the mouth reference point (MRP)]; +- the Receive volume control is adjusted to the setting that produces the level closest to $-39$ dBV when considering binaural headsets. + +NOTE – If different network signal levels are used, it should be noted and stated in the test. The Lombard effect (increased talker speech level due to high background noise) is considered in the background noise tests. + +Some tests require exact synchronization of test signals in the time domain. Therefore, it is necessary to take the delays of the terminals into account. When analysing signals, any delays introduced by the test system, codecs and terminals have to be taken into account accordingly. + +Test set-ups and requirements in this mode are harmonized with the existing ITU-T P.381 specification. + +### 7.1.2 Delay for the communication mode (terminal) + +The detailed requirements are specified in clause 7.1.2.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.2.2 of [ITU-T P.381]. + +### 7.1.3 Level in Send for nominal speech input level for the communication mode (terminal) + +The detailed requirements are specified in clause 7.1.3.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.3.2 of [ITU-T P.381]. + +### 7.1.4 Level in Receive for nominal speech input level + +The detailed requirements are specified in clause 7.1.4.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.4.2 of [ITU-T P.381]. + +### **7.1.5 Level in Send for low and high speech input levels for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.5.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.5.2 of [ITU-T P.381]. + +### **7.1.6 Linearity in Receive for the communication mode (terminal)** + +The linearity in Receive is for further study + +### **7.1.7 Sending frequency response for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.7.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.7.2 of [ITU-T P.381]. + +### **7.1.8 Receiving frequency response for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.8.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.8.2 of [ITU-T P.381]. + +### **7.1.9 Sidetone loss for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.9.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.9.2 of [ITU-T P.381]. + +### **7.1.10 Sidetone delay for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.10.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.10.2 of [ITU-T P.381]. + +### **7.1.11 Noise in Send for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.11.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.11.2 of [ITU-T P.381]. + +### **7.1.12 Noise in Receive for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.12.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.12.2 of [ITU-T P.381]. + +### **7.1.13 Sending distortion for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.13.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.13.2 of [ITU-T P.381]. + +### **7.1.14 Receive distortion for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.14.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.14.2 of [ITU-T P.381]. + +### **7.1.15 Noise cancellation test in Send for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.15.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.15.2 of [ITU-T P.381]. + +### **7.1.16 One-way speech quality in Send for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.16.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.16.2 of [ITU-T P.381]. + +### **7.1.17 One-way speech quality in Receive for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.17.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.17.2 of [ITU-T P.381]. + +### **7.1.18 Terminal coupling loss for the communication mode (terminal)** + +The detailed requirements for weighted terminal coupling loss (TCLw) and terminal coupling loss (TCL) are specified in clause 7.1.18.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.18.2 of [ITU-T P.381]. + +### **7.1.19 Temporal echo effects for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.19.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.19.2 of [ITU-T P.381]. + +### **7.1.20 Double talk performance for the communication mode (terminal)** + +#### **7.1.20.1 Attenuation range in Send during double talk, $A_{H,s,dt}$** + +The detailed requirements are specified in clause 7.1.20.1.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.20.1.2 of [ITU-T P.381]. + +#### **7.1.20.2 Attenuation range in Receive during double talk, $A_{H,r,dt}$** + +The detailed requirements are specified in clause 7.1.20.2.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.20.2.2 of [ITU-T P.381]. + +#### **7.1.20.3 Detection of echo components during double talk** + +The detailed requirements are specified in clause 7.1.20.3.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.20.3.2 of [ITU-T P.381]. + +### **7.1.21 Activation in Send for the communication mode (terminal)** + +The detailed requirements are specified in clause 7.1.21.1 of [ITU-T P.381]. + +The test details are specified in clause 7.1.21.2 of [ITU-T P.381]. + +## **7.2 Multimedia playback (terminal)** + +Test set-ups and requirements in this mode are harmonized with the existing [ITU-T P.381] specification. + +### **7.2.1 Test set-up for the multimedia playback mode (terminal)** + +General test set-ups are described in clause 7.2.1 of [ITU-T P.381]. + +### **7.2.2 Output level for the multimedia playback mode (terminal)** + +The detailed requirements are specified in clause 7.2.2.1 of [ITU-T P.381]. + +The test details are specified in clause 7.2.2.2 of [ITU-T P.381]. + +### **7.2.3 Frequency response for the multimedia playback mode (terminal)** + +The detailed requirements are specified in clause 7.2.3.1 of [ITU-T P.381]. + +The test details are specified in clause 7.2.3.2 of [ITU-T P.381]. + +### 7.2.4 Noise for the multimedia playback mode (terminal) + +The detailed requirements are specified in clause 7.2.4.1 of [ITU-T P.381]. + +The test details are specified in clause 7.2.4.2 of [ITU-T P.381]. + +### 7.2.5 Distortion for the multimedia playback mode (terminal) + +The detailed requirements are specified in clause 7.2.5.1 of [ITU-T P.381]. + +The test details are specified in clause 7.2.5.2 of [ITU-T P.381]. + +### 7.2.6 Receiving crosstalk for the multimedia playback mode (terminal) + +The detailed requirements are specified in clause 7.2.6.1 of [ITU-T P.381]. + +The test details are specified in clause 7.2.6.2 of [ITU-T P.381]. + +## 7.3 Multimedia record (terminal) + +### 7.3.1 Test set-up for the multimedia record mode (terminal) + +The test set-up is shown in Figure 2. + +![Figure 2: Set-up for testing the electrical headset interface. The diagram shows a 'Wireless digital terminal' containing 'Headset signal processing', 'Audio recorder', and 'Signal storage'. This terminal is connected to a 'Test system' via 'Left/right channel' lines. The 'Test system' is also connected to 'Signal storage, e.g., memory card' via 'Left/right channel' lines. The 'Test system' is represented by a box with a sine wave symbol and a variable resistor symbol.](18722c46c9e8475524e634dedd08bac2_img.jpg) + +Figure 2: Set-up for testing the electrical headset interface. The diagram shows a 'Wireless digital terminal' containing 'Headset signal processing', 'Audio recorder', and 'Signal storage'. This terminal is connected to a 'Test system' via 'Left/right channel' lines. The 'Test system' is also connected to 'Signal storage, e.g., memory card' via 'Left/right channel' lines. The 'Test system' is represented by a box with a sine wave symbol and a variable resistor symbol. + +P.382(16)\_F02 + +Figure 2 – Set-up for testing the electrical headset interface + +#### 7.3.1.1 Input and output characteristics of the test system for connecting to the headset connector + +The output of the test system connected to the sending interface of the headset connector must be DC resistant. The output impedance shall be between 1 Ω and 10 kΩ. The dynamic range shall be consistent with (or exceed) the level range provided by headset MICs. + +The input of the test system connected to the Receive interfaces of the headset connectors shall have an input impedance of 32 Ω. The dynamic range shall be consistent with (or exceed) the output level range provided by the electrical output of the digital mobile terminal headset outputs. + +#### 7.3.1.2 Test signals and test signal levels + +All input signal levels are expressed in decibels relative to 1 V, as they are electrically output from the test system into each device headset socket MIC input channel. + +All output signal levels specified in this clause are relative to decibels relative to full scale, where 0 dBFS represents the root mean square (RMS) level of a full-scale sinusoidal signal. + +Programme simulation noise as specified in [IEC 60268-1] is used for the measurements. Detailed information about the test signal used is specified in the corresponding clause of this Recommendation. + +The programme simulation noise according to [EN 50332-1] and [EN 50332-2] is band limited by design and requires no filtering. + +All test signal levels are relative to the average level over the complete test sequence length of the test signals, if not specified otherwise. + +Some tests require exact synchronization of test signals in the time domain. Therefore, it is necessary to take the delays of the terminals into account. When analysing signals, any delays introduced by the test system, codecs and terminals have to be taken into account accordingly. + +### **7.3.2 Input level for the multimedia record mode (terminal)** + +#### **7.3.2.1 Requirement** + +The level is measured as the digital input level generated by the output stereo file of the mobile terminal recorder when applying a stereo signal to the headset interface. + +The input level shall be $-29.0 \text{ dBFS} \pm 3 \text{ dB}$ when playing programme simulation noise at $-60 \text{ dBV}$ at the mobile terminal inputs. + +For a stereo recording, the offset between L and R channel shall be $< 1 \text{ dB}$ . + +#### **7.3.2.2 Test** + +- 1) The test set-up is according to Figure 2. +- 2) The test signal used for the measurements shall be programme simulation noise providing sufficient signal energy to 22 kHz. The test signal level is $-60 \text{ dBV}$ . +- 3) For the calculation, the averaged level of the encoded file in the signal storage is used. The output level is determined up to 22 kHz. +- 4) The output level is expressed in decibels relative to full scale. + +### **7.3.3 Frequency response for the multimedia record mode (terminal)** + +#### **7.3.3.1 Requirement** + +The frequency response is evaluated from a digital output file of the mobile terminal recorder when applying a signal to the headset interface. + +The frequency response shall be within $\pm 2 \text{ dB}$ between 500 Hz and 18 kHz when programme simulation noise at $-60 \text{ dBV}$ is electrically presented at the mobile terminal inputs. The requirement applies to each stereo channel individually. + +#### **7.3.3.2 Test** + +- 1) The test set-up is according to Figure 2. +- 2) The test signal used for the measurements shall be programme simulation noise providing sufficient signal energy to 22 kHz. The test signal level is $-60 \text{ dBV}$ . +- 3) The sending frequency response is determined in one-12th octave bands, as given in [IEC 61260-1] for frequencies from 100 Hz to 18 kHz inclusive, measured at the point of interconnection (POI). In each one-12th octave band, the power density spectrum of the measured signal is compared to that of the input signal averaged over the complete test sequence length. +- 4) The sensitivity is expressed in decibels relative to full scale per decibel relative to 1 V. + +#### **7.3.4 Acoustic input range for the multimedia record mode (terminal)** + +#### **7.3.4.1 Requirement** + +The acoustic input range is evaluated from a digital stereo output file of the mobile terminal recorder when applying a signal to the headset interface. + +The input range shall be 50 dB when playing programme simulation noise at $-40 \text{ dBV}$ and at $-90 \text{ dBV}$ at the mobile terminal inputs. + +#### **7.3.4.2 Test** + +- 1) The test set-up is according to Figure 2. +- 2) The test signal used for the measurements shall be programme simulation noise. The test signal levels are $-40.0$ dBV and $-90.0$ dBV. +- 3) The recorded files are compared to the input and no temporal clipping should be observed for either input level. +- 4) The acoustic input range is expressed in decibels. + +### **7.3.5 Phase response for the multimedia record mode (terminal)** + +#### **7.3.5.1 Requirement** + +The phase response is evaluated from a digital stereo output file of the mobile terminal recorder when applying a stereo signal to the headset interface. + +The phase response shall be in phase, as confirmed by the transfer function, when programme simulation noise at $-60$ dBV is electrically presented at the mobile terminal inputs. + +#### **7.3.5.2 Test** + +- 1) The test set-up is according to Figure 2. +- 2) The test signal used for the measurements shall be programme simulation noise. The test signal level is $-60.0$ dBV. +- 3) The recorded file is compared to the input and the channel specific transfer functions individually computed. +- 4) The output L and R responses are compared to one another. + +### **7.3.6 Distortion at maximum acoustic input in multimedia record mode** + +#### **7.3.6.1 Requirement** + +The total harmonic distortion (THD) is evaluated from a digital output file of the mobile terminal recorder when applying a signal to the headset interface. + +The THD + noise shall be at least 30 dB between the signal level in the frequency range between 50 Hz and 18 kHz when a sinusoidal signal at $-40.0$ dBV is electrically presented at the mobile terminal inputs. + +#### **7.3.6.2 Test** + +- 1) The test set-up is according to Figure 2. +- 2) The test signal used for the measurements shall be a sinusoidal signal at 50 Hz and 1 kHz. The test signal level is $-40.0$ dBV and the duration shall be $<1$ s. A short activation and conditioning sequence of male or female ITU-T P.501 speech may be needed, in which case it shall be input prior to the sinusoidal signal at the nominal signal level. +- 3) The spectral distribution of the recorded file is analysed with a 4k-point fast Fourier transform (FFT), and the THD + noise is computed using a flat top window. +- 4) The output distortion response is expressed in decibels. + +### **7.3.7 Cross-talk for the multimedia record mode (terminal)** + +#### **7.3.7.1 Requirement** + +The cross-talk is evaluated from a digital stereo output file of the mobile terminal recorder when applying a stereo signal to the headset interface. + +The cross-talk shall be below 50 dB, as confirmed by the transfer function, when programme simulation noise at $-40$ dBV is electrically presented at the mobile terminal inputs. + +#### **7.3.7.2 Test** + +- 1) The test set-up is according to Figure 2. +- 2) The test signal used for the measurements shall be programme simulation noise. The test signal level is $-40.0$ dBV. +- 3) The recorded file is compared to the input and the channel specific transfer functions individually computed. +- 4) The cross-talk is calculated between MIC1 and MIC2, as well as vice versa, MIC2 versus MIC1. + +### **7.3.8 Time offset and sampling accuracy for the multimedia record mode (terminal)** + +#### **7.3.8.1 Requirement** + +The time offset is evaluated from a digital stereo output file of the mobile terminal recorder when applying a stereo signal to the headset interface. + +The recorded file's sinusoidal frequency shall deviate by $<0.1\%$ , when the input sinusoidal 1 kHz signal at $-60$ dBV is electrically presented at the mobile terminal inputs. + +#### **7.3.8.2 Test** + +- 1) The test set-up is according to Figure 2. +- 2) The test signal used for the measurements shall be a 1 kHz sinusoidal signal. The test signal level is $-60.0$ dBV. +- 3) The recorded file is compared to the input and the transfer function computed. The spectral distribution of the output signal is analysed with a 4k-point FFT and the windowing used is the flat top window. +- 4) The output L and R responses are individually computed. + +# **8 Headset specification** + +## **8.1 Communication mode (headset)** + +Test set-ups and requirements in this mode are mostly harmonized with the existing ITU-T P.381 specification, except for clause 8.1.7, which, due to the multiple MICs, needs a specific test. + +### **8.1.1 Test set-up (headset)** + +The test set-up is specified in clause 8.1.1 of [ITU-T P.381]. + +### **8.1.2 Sensitivity in Send (headset)** + +The detailed requirements are specified in clause 8.1.2.1 of [ITU-T P.381]. + +The test details are specified in clause 8.1.2.2 of [ITU-T P.381]. + +### **8.1.3 Sensitivity in Receive (headset)** + +The detailed requirements are specified in clause 8.1.3.1 of [ITU-T P.381]. + +The test details are specified in clause 8.1.3.2 of [ITU-T P.381]. + +### **8.1.4 Sending frequency response (headset)** + +The detailed requirements are specified in clause 8.1.4.1 of [ITU-T P.381]. + +The test details are specified in clause 8.1.4.2 of [ITU-T P.381]. + +### 8.1.5 Receiving frequency response (headset) + +The detailed requirements are specified in clause 8.1.5.1 of [ITU-T P.381]. + +The test details are specified in clause 8.1.5.2 of [ITU-T P.381]. + +### 8.1.6 Idle channel noise in Send (headset) + +The detailed requirements are specified in clause 8.1.6.1 of [ITU-T P.381]. + +The test details are specified in clause 8.1.6.2 of [ITU-T P.381]. + +### 8.1.7 Distortion in Send (headset) + +The detailed requirements are specified in clause 8.1.7.1 of [ITU-T P.381]. + +The test details are specified in clause 8.1.7.2 of [ITU-T P.381]. + +### 8.1.8 Headset terminal coupling loss (headset) + +#### 8.1.8.1 Requirement + +For speech transmission, the weighted headset terminal coupling loss (HTCLw) for NB, and headset terminal coupling loss (HTCL) for WB, SWB and FB is evaluated for the electrical interface MIC1 or MIC2 lines when applying a stereo signal to the electrical receiver of the headset interface. + +The coupling loss shall be <65 dB, as confirmed by each transfer function, when programme simulation noise at -22 dBV is electrically presented at the mobile terminal inputs. + +#### 8.1.8.2 Test + +- 1) The test set-up is according to Figure 3. +- 2) The test signal used for the measurements shall be programme simulation noise. The test signal level is -22.0 dBV. +- 3) The headset MIC output is compared to the headset receiver input and the channel specific transfer functions are individually computed in the frequency range between 50 Hz and 18 kHz. +- 4) The coupling loss is calculated between HP L input and MIC1 output, as well as HP R input and MIC2 output, respectively. Coupling loss between HP L input and MIC2 output, as well as HP R Input and MIC1 output, should be confirmed to also comply with the requirement. +- 5) For NB mode, HTCLw is calculated according to clause B.4 of [ITU-T G.122] (trapezoidal rule). In WB, SWB and FB mode, HTCL is calculated as the echo loss (EL) from 50 Hz to 18 kHz. For the calculation, the averaged test signal level at each frequency band takes as reference the averaged measured echo signal level in each frequency band. For the measurement, a time window has to be applied that is adapted to the duration of the actual test signal. The EL is calculated by the following equations. The form of the first is generalized from that specified in clause B.4 of [ITU-T G.122] to calculate EL based on tabulated data, which allows the calculation of EL within any frequency range between $f_0$ and $f_N$ . + +$$L_e = C - 10 \log_{10} \sum_{i=1}^N (A_i + A_{i-1})(\log_{10} f_i - \log_{10} f_{i-1}),$$ + +where $C = 10 \log_{10} 2 (\log_{10} f_N - \log_{10} f_0)$ , + +where + +$A_0$ is the output/input power ratio at frequency $f_0 = 50$ Hz; + +$A_1$ is the output/input power ratio at frequency $f_i$ ; + +$A_N$ is the output/input power ratio at frequency $f_N = 18$ kHz + +## 8.2 Multimedia playback mode (headset) + +Test set-ups and requirements in this mode are harmonized with the existing ITU-T P.381 specification. + +### 8.2.1 Test set-up (headset) + +The test set-up is specified in clause 8.2.1 of [ITU-T P.381]. + +### 8.2.2 Sensitivity for the multimedia playback mode (headset) + +The detailed requirements are specified in clause 8.2.2.1 of [ITU-T P.381]. + +The test details are specified in clause 8.2.2.2 of [ITU-T P.381]. + +### 8.2.3 Distortion for the multimedia playback mode (headset) + +The detailed requirements are specified in clause 8.2.3.1 of [ITU-T P.381]. + +The test details are specified in clause 8.2.3.2 of [ITU-T P.381]. + +### 8.2.4 Receiving crosstalk for the multimedia playback mode (headset) + +The detailed requirements are specified in clause 8.2.4.1 of [ITU-T P.381]. + +The test details are specified in clause 8.2.4.2 of [ITU-T P.381]. + +## 8.3 Multimedia record mode (headset) + +### 8.3.1 Test set-up (headset) + +The test set-up is shown in Figure 3. + +![Figure 3: Set-up for testing the headset. The diagram shows a headset with a microphone and earpiece connected to a 'Headset reference interface' box. This box is connected to a 'Test system' box. The test system has two output sources, one labeled -90° and the other 90°. The -90° source is connected to the earpiece, and the 90° source is connected to the microphone. A dashed circle with arrows indicates the acoustic path from the earpiece to the microphone, with a distance 'd' marked between them. The test system also has two input sources, one labeled -90° and the other 90°. The diagram is labeled P.382(16)_F03.](492256bb98eb10f95553a7b1983c4b96_img.jpg) + +The diagram illustrates the test set-up for a headset. A headset is shown with a microphone and an earpiece. The microphone is connected to a 'Headset reference interface' box, which is in turn connected to a 'Test system' box. The test system has two output sources, one labeled $-90^\circ$ and the other $90^\circ$ . The $-90^\circ$ source is connected to the earpiece, and the $90^\circ$ source is connected to the microphone. A dashed circle with arrows indicates the acoustic path from the earpiece to the microphone, with a distance $d$ marked between them. The test system also has two input sources, one labeled $-90^\circ$ and the other $90^\circ$ . The diagram is labeled P.382(16)\_F03. + +Figure 3: Set-up for testing the headset. The diagram shows a headset with a microphone and earpiece connected to a 'Headset reference interface' box. This box is connected to a 'Test system' box. The test system has two output sources, one labeled -90° and the other 90°. The -90° source is connected to the earpiece, and the 90° source is connected to the microphone. A dashed circle with arrows indicates the acoustic path from the earpiece to the microphone, with a distance 'd' marked between them. The test system also has two input sources, one labeled -90° and the other 90°. The diagram is labeled P.382(16)\_F03. + +Figure 3 – Set-up for testing the headset + +Requirements for HATS and ear simulators are specified in clause 8.2.1 of [ITU-T P.381]. + +#### 8.3.1.1 Input and output characteristics of the test system for connecting the headset + +The output impedance shall be $< 2\ \Omega$ . The RMS output voltage shall be adjusted not to overdrive an artificial mouth or loudspeaker adjacent to the HATS. The common GND impedance (between sending and receiving sides) for the test system shall be $\leq 0.05\ \Omega$ . + +#### 8.3.1.2 Test signals and test signal levels + +Sinusoidal signals are used, and signals analysed in one-12th octave bands. + +All test signal levels are referred to the average level of the test signals, averaged over the complete test sequence length, if not described otherwise. + +The nominal average signal level for the measurements is $-28.7$ dBPa at the MRP. + +Some tests require exact synchronization of test signals in the time domain. Therefore, it is necessary to take the delays of the terminals into account. When analysing signals, any delays introduced by the test system, codecs and terminals have to be taken into account accordingly. + +#### 8.3.1.3 Positioning of the headsets + +The same guidelines and requirements on headset positioning apply as in clause 8.1.1.3 of [ITU-T P.381]. + +#### 8.3.1.4 Position and calibration of reference loudspeaker + +The reference loudspeaker or mouth shall be calibrated and equalized at the appropriate distance for each test case. The calibration and equalization shall be performed using a free field microphone and is analysed in one-12th octave bands. During equalization, the HATS is removed from the sound field. + +### 8.3.2 Frequency response for the multimedia record mode (headset) + +#### 8.3.2.1 Requirement + +The frequency response is evaluated from the stereo signal of the headset reference interface when reproducing a mono signal at the output of the reference loudspeaker. A reference microphone is placed close to the MIC under test and this reference measurement is used to evaluate the frequency response. + +The frequency response shall be inside the mask in Table 1 for each stereo channel individually when sinusoidal signals at $-28.7$ dBPa are acoustically output from the reference loudspeaker. The requirement is only applicable for $0^\circ$ incidence. + +**Table 1** + +| Frequency (Hz) | Upper limit (dB) | Lower limit (dB) | +|----------------|------------------|------------------| +| 200 | 5 | -5 | +| 5 000 | 5 | -5 | +| 14 000 | 0 | -20 | + +NOTE 1 – All sensitivity values are expressed in decibels on an arbitrary scale. + +NOTE 2 – The limits for intermediate frequencies lie on a straight line drawn between the given values on a logarithmic (frequency) – linear (decibel) scale. + +#### 8.3.2.2 Test + +- 1) The test set-up is according to Figure 3. The reference loudspeaker is placed at $d = 50$ cm distance from the HATS MRP. A reference microphone is placed close to the MIC under test. +- 2) The headset is mounted on the HATS, which is rotated from $-90^\circ$ to $90^\circ$ in steps of $45^\circ$ and the MIC electrical signals are picked up by the test system. +- 3) For the test, sinusoidal signals are used in the bandwidth between 200 Hz and 14 kHz. The duration of each sine wave shall be $< 1$ s. The signals are played over the free-standing loudspeaker or artificial mouth at $-28.7$ dBPa, measured at the HATS MRP. +- 4) The sending frequency response is analysed with a 4k-point FFT and determined in one-12th octave bands, as given in [IEC 61260-1] for frequencies from 200 Hz to 14 kHz inclusive, measured at the POI. In each one-12th octave band, the power density spectrum of the + +measured signal is compared to the power density spectrum of the input signal using a flat top window, and averaged over the complete test sequence length. + +### **8.3.3 Maximum acoustic input for the multimedia record mode (headset)** + +#### **8.3.3.1 Requirement** + +The maximum acoustic input is evaluated from the individual L and R signals of the headset reference interface when reproducing a mono signal at the output of the reference loudspeaker. + +The maximum acoustic input shall be linear in level to the programme simulation noise signal, as the levels 0 dBPa and +16 dBPa are acoustically output from the reference loudspeaker. + +#### **8.3.3.2 Test** + +- 1) The test set-up is according to Figure 3. The reference loudspeaker is placed at $d = 10$ cm distance from the HATS MRP, then the HATS is rotated 90° L and R, respectively, so each headset MIC is measured in front of the reference loudspeaker. +- 2) Programme simulation noise as specified in [IEC 60268-1] is used for the measurements. Detailed information about the test signal used is specified in the corresponding clause of this Recommendation. The programme simulation noise according to [EN 50332-1] and [EN 50332-2] is band limited by design and requires no filtering. +- 3) The test levels are 0 dBPa and +16 dBPa, as measured at the headset MIC location. +- 4) The electrical MIC output signal is analysed with a 4k-point FFT, and the transfer function referenced to the acoustic spectral density and computed using a flat top window. + +### **8.3.4 Distortion at maximum acoustic input for the multimedia record mode (headset)** + +#### **8.3.4.1 Requirement** + +The distortion at maximum acoustic input is evaluated from the individual L and R signals of the headset reference interface when reproducing a mono signal at the output of the reference loudspeaker. + +The THD + noise shall be <30 dB, as sinusoidal signals at level +20.7 dBPa are acoustically output from the reference loudspeaker. + +#### **8.3.4.2 Test** + +- 1) The test set-up is according to Figure 3. The reference loudspeaker is placed at 10 cm distance from the HATS MRP, then the HATS is rotated 90° L and R, respectively, so that each headset MIC is measured in front of the reference loudspeaker. +- 2) For the test, a sinusoidal signal at frequencies of 315 Hz and 1 000 Hz is used. The duration of the sine wave shall be <1 s. The sinusoidal signal levels shall be 20.7 dBPa at the ear reference point. +- 3) The signal to harmonic distortion ratio is measured selectively up to 10 kHz. + +# **9 Function requirements for terminals with the universal headset interface** + +See clause 9 of [ITU-T P.381] for functional requirements and impedance ranges for the terminals that should be supported. + +# Annex A + +## **Interpolation method for diffuse-field correction** + +(This annex forms an integral part of this Recommendation.) + +For measurements requiring diffuse-field correction values for closer frequency spacing than one-12th octave bands, linear interpolation on a logarithmic scale from one-12th octave band interpolated values in Table 14b of [ITU-T P.58] shall be used. + +## Appendix I + +### Audio connectivity for sockets with five contact points + +(This appendix does not form an integral part of this Recommendation.) + +This appendix illustrates the dimensions of concentric plug and socket connectors with five contact points. + +### I.1 Plug connector of diameter 2.5 mm with five poles + +The 2.5 mm diameter plug connector with five poles is for further study. + +### I.2 Socket connector of diameter 2.5 mm with five contact points + +The 2.5 mm diameter socket connector with five poles is for further study. + +### I.3 Plug connector of diameter 3.5 mm with five poles + +Figure I.1 shows the shape and dimensions of the 3.5 mm diameter plug connector with five poles. The width of strip A along the axial direction is 0.15 mm and shall be free of sharp edges at its corners. Junctions B should be free of burr or flash. + +NOTE – "Flash" here refers to a rough edge or ridge on the surface. + +![Figure I.1: Shape and dimensions of the 3.5 mm diameter plug connector with five poles. The diagram shows a cross-section of the connector with various dimensions and callouts. Dimensions include 14.0, 8.5, 4.8, 0.7, 2.4 Max, 3.0±0.05, 2.9±0.05, 3.1±0.05, 3.4±0.05, 3.5±0.05, and R0.25. Callouts include (1) tip made of conductive material, (2) insulating rings, (3) ring 1 made of conductive material, (4) available spaces for a conductive ring 2, (5) ring 3 made of conductive material, (6) sleeve made of conductive material, and (7) illustration of the hand grip at the end of a plug. Asterisks (*) indicate dimensions to be found in ITU-T P.381. A general tolerance of ± 0.2 is specified for dimensions without individual tolerances.](8db6897b1527256adb62e789ab3e5197_img.jpg) + +Figure I.1: Shape and dimensions of the 3.5 mm diameter plug connector with five poles. The diagram shows a cross-section of the connector with various dimensions and callouts. Dimensions include 14.0, 8.5, 4.8, 0.7, 2.4 Max, 3.0±0.05, 2.9±0.05, 3.1±0.05, 3.4±0.05, 3.5±0.05, and R0.25. Callouts include (1) tip made of conductive material, (2) insulating rings, (3) ring 1 made of conductive material, (4) available spaces for a conductive ring 2, (5) ring 3 made of conductive material, (6) sleeve made of conductive material, and (7) illustration of the hand grip at the end of a plug. Asterisks (\*) indicate dimensions to be found in ITU-T P.381. A general tolerance of ± 0.2 is specified for dimensions without individual tolerances. + +- | | | +|------------------------------------------------------|--------------------------------------------| +| ① tip made of conductive material | ② insulating rings | +| ③ ring 1 made of conductive material | ④ available spaces for a conductive ring 2 | +| ⑤ ring 3 made of conductive material | ⑥ sleeve made of conductive material | +| ⑦ illustration of the hand grip at the end of a plug | | + +\*Dimensions intentionally left out, refer to drawings in [ITU-T P.381] for dimensional guidelines. + +**Figure I.1 – Shape and dimensions of the 3.5 mm diameter plug connector with five poles** + +The grip dimensions of the five pole connector shall be harmonized with the ITU-T P.381 dimensions for optimum compatibility. + +### I.4 Socket connector of diameter 3.5 mm with five contact points + +The socket should be able to mate and cooperate with the plug reliably. The dimensions (see Table I.1) and the positioning for each contact spring are illustrated in Figure I.2. Considering the tolerance of the plug dimension and positioning of the socket contact spring, in addition to the shift of the practical contact point location caused by the width of the spring, the minimum distance between the contact point of the ring 2 spring and that of the sleeve spring is recommended to be more than 2.4 mm. If bushing of the socket is made of conductive material, the contact area of the sleeve spring may exceed the given range indicated in Figure I.2, so bushing of the socket should not be longer than 2.4 mm. + +**Table I.1 – Specific spring contact dimensions for five contact points** + +| Dimension | Five pole socket: length (mm) | +|-----------|-------------------------------| +| Tip | 11.95 | +| Ring 1 | 7.6 | +| Ring 2 | Not defined | +| Ring3 | 3.7 | +| Sleeve | 1.0 | + +NOTE – Entries shall be contact dimensions with a plug inserted + +![Figure I.2: Dimensions of the 3.5 mm diameter socket with five contact points and positioning of each contact spring. The diagram shows a cross-section of the socket with five contact springs: Sleeve spring, Ring 3 spring, Ring 1 spring, Tip spring, and Ring 2 spring. The bushing of the socket is shown at the bottom. Dimensions are indicated by arrows and labels: Ring 1 ±0.2, Ring 3 ±0.2, Sleeve ±0.5, Ø 3.6±0.04, Ring 2 ±0.2, and Tip +0.3/-0.2. The diagram is labeled P.382(16)_FI.2.](cb74fd9f5ec715dd3e2e325b864b48bc_img.jpg) + +Figure I.2: Dimensions of the 3.5 mm diameter socket with five contact points and positioning of each contact spring. The diagram shows a cross-section of the socket with five contact springs: Sleeve spring, Ring 3 spring, Ring 1 spring, Tip spring, and Ring 2 spring. The bushing of the socket is shown at the bottom. Dimensions are indicated by arrows and labels: Ring 1 ±0.2, Ring 3 ±0.2, Sleeve ±0.5, Ø 3.6±0.04, Ring 2 ±0.2, and Tip +0.3/-0.2. The diagram is labeled P.382(16)\_FI.2. + +**Figure I.2 – Dimensions of the 3.5 mm diameter socket with five contact points and positioning of each contact spring** + +## Appendix II + +### **Audio connectivity for sockets with six contact points** + +(This appendix does not form an integral part of this Recommendation.) + +This appendix illustrates the optional dimensions of concentric plug and socket connectors with six contact points and is for further study. + +## Appendix III + +### **Audio connectivity for sockets with fewer than five contact points** + +(This appendix does not form an integral part of this Recommendation.) + +Refer to Appendices I to III of [ITU-T P.381] for illustrations of dimensions of concentric plug and socket connectors with three and four contact points. + +## **Appendix IV** + +## **Other considerations** + +(This appendix does not form an integral part of this Recommendation.) + +Refer to Appendix IV of [ITU-T P.381] for illustrations of other considerations when implementing a five pole plug and socket. + +# Bibliography + +- [b-ITU-T G.701] Recommendation ITU-T G.701 (1993), *Vocabulary of digital transmission and multiplexing, and pulse code modulation (PCM) terms.* +- [b-ITU-T P.10] Recommendation ITU-T P.10/G.100 (2017), *Vocabulary for performance, quality of service and quality of experience.* +- [b-ITU-T P.57] Recommendation ITU-T P.57 (2021), *Artificial ears.* +- [b-ITU-T P.380] Recommendation ITU-T P.380 (2022), *Electro-acoustic measurements on headsets.* +- [b-ITU-T P.800.1] Recommendation ITU-T P.800.1 (2016), *Mean opinion score (MOS) terminology.* +- [b-ITU-T P.863] Recommendation ITU-T P.863 (2018), *Perceptual objective listening quality prediction.* +- [b-IEC 60268-7] International Standard IEC 60268-7 (2010), *Sound system equipment – Part 7: Headphones and earphones.* + + + + + +# SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | Tariff and accounting principles and international telecommunication/ICT economic and policy issues | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling, and associated measurements and tests | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects, next-generation networks, Internet of Things and smart cities | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/012117dcead1d581e96cafc55090be70_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/012117dcead1d581e96cafc55090be70_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9aa6d3c1291760ff9eaae4de1e0021da1c1750d2 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/012117dcead1d581e96cafc55090be70_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6dcff3306cdc6e2be72b7264c7e9dfe0da429f5b3e68b45f60908fb1e7008b6f +size 77580 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/03498c9b76f980b32f2dfbb7c2e539d2_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/03498c9b76f980b32f2dfbb7c2e539d2_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..227f8591abd0bea95b760511201079b6d66efaed --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/03498c9b76f980b32f2dfbb7c2e539d2_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:525a04bc772433a62717df4f5afc10099280203466329aa4976c06ff15ce31e6 +size 77096 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/103781b498cc6846ffc11a041e14dc3e_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/103781b498cc6846ffc11a041e14dc3e_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..059d434aa06c5078a718e0458cd4febb5d4d13e5 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/103781b498cc6846ffc11a041e14dc3e_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5b64b75602df35372c923714146fd423dab04e80af40dc6b90336a5d1781d44b +size 79694 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/1c2028183a35357e7238438a4af9cab7_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/1c2028183a35357e7238438a4af9cab7_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..29d1dd1f56d170afadb4c5016db0603439763e45 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/1c2028183a35357e7238438a4af9cab7_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:039c6cc80a94b04382be96bafa0cf40c956f9657b9b23a457c87c381fb0fafb7 +size 59351 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/36f6f5bc2a35070ad6dc4312a93bcbe3_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/36f6f5bc2a35070ad6dc4312a93bcbe3_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6b76c5bbb6c86d254eda833a48211b9b8355f664 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/36f6f5bc2a35070ad6dc4312a93bcbe3_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:0b9534604e2e291b302d5bf99eaf1e1b0214279b797e6c48d80396a96ef75820 +size 129647 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/391ab9e5616ba6311161af4d7a93422b_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/391ab9e5616ba6311161af4d7a93422b_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..4247a20e2f9e9b7fcd19f15ac0e79f62840d5166 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/391ab9e5616ba6311161af4d7a93422b_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:91a9bc89ff0e519f4cf0614d50fe98f21fe02388c5e4f2a7a105d89b6bb79088 +size 35825 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/4134b40f9161ec8fcd3c28bdb48f80b0_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/4134b40f9161ec8fcd3c28bdb48f80b0_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cdf0714beaf685ca5340470e0051122630d63fe4 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/4134b40f9161ec8fcd3c28bdb48f80b0_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:06ea6bdcc5ec7ccb736875c82361d74aed87495d013dc3ec5c9a268d47677b48 +size 116399 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/5e7b93195badf89e1fb645ae53d619bb_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/5e7b93195badf89e1fb645ae53d619bb_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..c182fa487655de3626950dd922b71f6ca29a8727 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/5e7b93195badf89e1fb645ae53d619bb_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:63e9bc580e469b52e32e0046a254cfaf74beb1e0e58d96d0bfc73a0a1e5f470f +size 138606 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/5fdd55201f3f090a54d10ffe9de2f068_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/5fdd55201f3f090a54d10ffe9de2f068_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..dfff4852c4a56bde305b93da8c6997e860bfcaae --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/5fdd55201f3f090a54d10ffe9de2f068_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ed6d96f32070ec641b58109a8140396e25a830088ac903bba8476b8ad8ad10c0 +size 60129 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/60e9207be66a64332619bb4b667fe67b_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/60e9207be66a64332619bb4b667fe67b_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f03738470f5c77cf9f44495fa3050e6892f4b8b8 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/60e9207be66a64332619bb4b667fe67b_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f44b45ab1ddc1a663c98c5549fb2c70fa113344886023dd9c8882739a757e599 +size 59328 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/657acccf744d33f1fc3a1652741a256e_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/657acccf744d33f1fc3a1652741a256e_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..5e52e26d01a99ed068bbb3d17d90a5a7ed123151 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/657acccf744d33f1fc3a1652741a256e_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:de10bf78b6d64b139e16ca43ff64152fecdc4f18d3c1d0dde8bc7a53e7310bd6 +size 35671 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/65a73373b57df71e5c2ce1ce0eb7b65d_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/65a73373b57df71e5c2ce1ce0eb7b65d_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..9b4000fa0358b7de3f1fd3c8887a1d124fe3b8d6 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/65a73373b57df71e5c2ce1ce0eb7b65d_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9f2c9e9a56047c8ec0fc77edfa24018df327e26aa5e36ad9dc53581d21ff71f5 +size 45368 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/6fd373b1aea8f583da362efa5e3710df_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/6fd373b1aea8f583da362efa5e3710df_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6fe645fece10a56aee8a87d2f2e1a6eda13e6d37 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/6fd373b1aea8f583da362efa5e3710df_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:127ca98bf0270a928a74fe7585b15f2f8b54b274ceac7f416b0eca7c156002a3 +size 115199 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/76959415a1298a6f58ea1fff16fb01b6_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/76959415a1298a6f58ea1fff16fb01b6_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e3080505b28e853a474762a7b2689ede1812a89b --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/76959415a1298a6f58ea1fff16fb01b6_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:a3b1d109a76d3f6b7acae117583d2a7f3cf63178847ffd1303fd5751eafa7792 +size 119128 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/77a781dfb114c3e2b399f876f1808cfd_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/77a781dfb114c3e2b399f876f1808cfd_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..26aec89a1684b107a3ff5073b23d293e3e91db41 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/77a781dfb114c3e2b399f876f1808cfd_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:59e815047e21c521f80e2740f332794a0d3e87f7eee95ab2d336e3b1f02a8e03 +size 67252 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/8d5d2817d93a53bb48da3a869f6aa149_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/8d5d2817d93a53bb48da3a869f6aa149_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..0421c841ffe0e75dc41415f64746aecfd435db65 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/8d5d2817d93a53bb48da3a869f6aa149_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6ffb816326f7c5c254512c227e5a102be91702c38ed067574fded8c142d85445 +size 147197 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/95e259e8cb3519025066052af263f8c0_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/95e259e8cb3519025066052af263f8c0_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a4b7e9d61d0f399f26a9e08e5764d78e3e6992a4 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/95e259e8cb3519025066052af263f8c0_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:516639c2f7914df40f1848c70160cfbb915f5473c7cbcf3b4508740b1a5af4d4 +size 67540 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/97f61e67792478fb6ce089868e503063_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/97f61e67792478fb6ce089868e503063_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..fe5364cee023be2ba530a4810ebae9b17d33fd56 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/97f61e67792478fb6ce089868e503063_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:796784bdedc0f1d30e4a69ff8d3b595e959c9ca3acef75ab699f78f548aa1d7f +size 58786 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/9a19da4f7fccb96a934411c0bb5a386d_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/9a19da4f7fccb96a934411c0bb5a386d_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..494a0ef833aedb1544783daa75b741ed3e2a1e97 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/9a19da4f7fccb96a934411c0bb5a386d_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c6d43abfd9d7cf0e97b7c0478f7364a29ec4c6c81aea8f22f6d9cd1a5ea6f72d +size 46105 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/a7f9e5d1862146a136ba5ea891f15f17_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/a7f9e5d1862146a136ba5ea891f15f17_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3c308cdb7c498bfc5295ab4826e9e0b969ef994b --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/a7f9e5d1862146a136ba5ea891f15f17_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b4edd67722837cb1dbbf591c678595dcd78926b13c29694d45c56d6a7d083bcc +size 129794 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/bdd910852600bb450eb8544a4c88b280_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/bdd910852600bb450eb8544a4c88b280_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e7401c13f9170a5f14266c52617b033bcda673de --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/bdd910852600bb450eb8544a4c88b280_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3153163668eb3b78a73d1f2bcc59f24df1429494857d000ee098e35a1186e171 +size 69989 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/c2813655160a6d83a32289b367a84d1b_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/c2813655160a6d83a32289b367a84d1b_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f29aa38627ad47fed1302c86aa71dd31470de845 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/c2813655160a6d83a32289b367a84d1b_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:c9096a903ed879c30b3eda866797400fbbaac509022e8b8bf6878420cb61c5e2 +size 125509 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/caeb46d58919c727aa9d338ae9492077_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/caeb46d58919c727aa9d338ae9492077_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..6578c12faad49419fa6e14e004e97a0e26ebc3dd --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/caeb46d58919c727aa9d338ae9492077_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5673ab45570b4e08c77ddf8f80ec34def783cf5d746a287f89c83d4c36fc35b1 +size 49787 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/e4a14961bdc9882e296b25f7ae5f9760_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/e4a14961bdc9882e296b25f7ae5f9760_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..73d81a2d468b061feb0ecf52836d912fd5649894 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/e4a14961bdc9882e296b25f7ae5f9760_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:2e4d81e8ef67c13e5d9520d63ce6e3881a342246803b5327973e36d1b3f47f23 +size 18390 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/e999b14cdba9a4824937bb35d8489a03_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/e999b14cdba9a4824937bb35d8489a03_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..91982d5f5ce57a566b017b9e99cd998479d77052 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/e999b14cdba9a4824937bb35d8489a03_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:82da3b61809f8fd0b5d76a6b5d9fe0a8b537caaf605507c6146a3ef0115cb69b +size 63209 diff --git a/marked/P/T-REC-P.501-202504-I_PDF-E/f356ce635bbbc30c7b6db50ecdb5dab4_img.jpg b/marked/P/T-REC-P.501-202504-I_PDF-E/f356ce635bbbc30c7b6db50ecdb5dab4_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..a5d7301fccde4aefa22dd076e01ecab746513461 --- /dev/null +++ b/marked/P/T-REC-P.501-202504-I_PDF-E/f356ce635bbbc30c7b6db50ecdb5dab4_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:f7056b30b0f720df806557fdaaf660d0e1b9950a9c0dd3f583427fe0abb93af1 +size 135033 diff --git a/marked/P/T-REC-P.54-198811-I_PDF-E/raw.md b/marked/P/T-REC-P.54-198811-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..b84ad93f5fca5bccb157686fdb3a6d8c60e0aada --- /dev/null +++ b/marked/P/T-REC-P.54-198811-I_PDF-E/raw.md @@ -0,0 +1,44 @@ + + +![ITU logo](2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg) + +The logo of the International Telecommunication Union (ITU) features the letters 'ITU' in a bold, sans-serif font, superimposed on a stylized globe with intersecting lines. + +ITU logo + +INTERNATIONAL TELECOMMUNICATION UNION + +**ITU-T** + +**P.54** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +# **TELEPHONE TRANSMISSION QUALITY OBJECTIVE MEASURING APPARATUS** --- + +## **SOUND LEVEL METERS (APPARATUS FOR THE OBJECTIVE MEASUREMENT OF ROOM NOISE)** + +**ITU-T Recommendation P.54** + +(Extract from the *Blue Book*) + +--- + +## NOTES + +1 ITU-T Recommendation P.54 was published in Volume V of the *Blue Book*. This file is an extract from the *Blue Book*. While the presentation and layout of the text might be slightly different from the *Blue Book* version, the contents of the file are identical to the *Blue Book* version and copyright conditions remain unchanged (see below). + +2 In this Recommendation, the expression “Administration” is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +## **SOUND LEVEL METERS (APPARATUS FOR THE OBJECTIVE MEASUREMENT OF ROOM NOISE)** + +*(amended at Mar del Plata, 1968 and Geneva, 1972)* + +The CCITT recommends the adoption of the sound level meter specified in [1] in conjunction, for most uses, with the octave, half, and third octave filters in accordance with [2]. + +## **References** + +- [1] International Electrotechnical Commission Standard, *Sound level meters*, IEC Publication 651 (179), Geneva, 1979. +- [2] International Electrotechnical Recommendation, *Octave, half-octave and third-octave band filters intended for the analysis of sounds and vibrations*, IEC Publication 225, Geneva, 1966. \ No newline at end of file diff --git a/marked/P/T-REC-P.570-202503-I_PDF-E/raw.md b/marked/P/T-REC-P.570-202503-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..906effb72dac3d4282da9c9f6787ad2e563d9df9 --- /dev/null +++ b/marked/P/T-REC-P.570-202503-I_PDF-E/raw.md @@ -0,0 +1,2775 @@ + + +# Recommendation + +## **ITU-T P.570 (03/2025)** + +SERIES P: Telephone transmission quality, telephone installations, local line networks + +Objective measuring apparatus + +--- + +# **Artificial noise fields under laboratory conditions** + +![ITU logo](84a1d09fb489061482111515543b60dc_img.jpg) + +The logo of the International Telecommunication Union (ITU) is located in the bottom right corner. It features a blue globe with white lines representing latitude and longitude, and the letters 'ITU' in a bold, blue, sans-serif font superimposed on the globe. + +ITU logo + +## ITU-T P-SERIES RECOMMENDATIONS + +## **Telephone transmission quality, telephone installations, local line networks** + +| | | +|----------------------------------------------------------------------------------------------------|--------------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | P.10-P.19 | +| Voice terminal characteristics | P.30-P.39 | +| Reference systems | P.40-P.49 | +| Objective measuring apparatus | P.50-P.59 | +| Objective electro-acoustical measurements | P.60-P.69 | +| Measurements related to speech loudness | P.70-P.79 | +| Methods for objective and subjective assessment of speech quality | P.80-P.89 | +| Voice terminal characteristics | P.300-P.399 | +| Objective measuring apparatus | P.500-P.599 | +| Measurements related to speech loudness | P.700-P.709 | +| Methods for objective and subjective assessment of speech and video quality | P.800-P.899 | +| Audiovisual quality in multimedia services | P.900-P.999 | +| Transmission performance and QoS aspects of IP end-points | P.1000-P.1099 | +| Communications involving vehicles | P.1100-P.1199 | +| Models and tools for quality assessment of streamed media | P.1200-P.1299 | +| Telemeeting assessment | P.1300-P.1399 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | P.1400-P.1499 | +| Methods for objective and subjective assessment of quality of services other than speech and video | P.1500-P.1599 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +# Recommendation ITU-T P.570 + +# Artificial noise fields under laboratory conditions + +## Summary + +Recommendation ITU-T P.570 specifies various artificial noise field configurations for use in both objective and subjective evaluation of communication equipment. + +Furthermore, it provides guidance on capturing real-world noise environments, preprocessing of signals for reproduction of such environments, and describes calibration procedures and assessment methods for these artificial noise field configurations. + +## History \* + +| Edition | Recommendation | Approval | Study Group | Unique ID | +|---------|----------------|------------|-------------|--------------------| +| 1.0 | ITU-T P.570 | 2018-06-13 | 12 | 11.1002/1000/13624 | +| 2.0 | ITU-T P.570 | 2025-03-01 | 12 | 11.1002/1000/16215 | + +## Keywords + +Ambient noise, artificial noise fields, capturing environment noise, loudspeaker equalization, loudspeaker set-up. + +--- + +\* To access the Recommendation, type the URL in the address field of your web browser, followed by the Recommendation's unique ID. + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, and information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents/software copyrights, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the appropriate ITU-T databases available via the ITU-T website at . + +© ITU 2025 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## Table of Contents + +###### Page + +| | | | +|-------|------------------------------------------------------------------------------------|----| +| 1 | Scope ..... | 1 | +| 2 | References..... | 1 | +| 3 | Definitions ..... | 2 | +| 3.1 | Terms defined elsewhere ..... | 2 | +| 3.2 | Terms defined in this Recommendation..... | 2 | +| 4 | Abbreviations and acronyms ..... | 4 | +| 5 | Conventions ..... | 4 | +| 6 | How to use this Recommendation ..... | 5 | +| 7 | Techniques for capturing real-world noise environments ..... | 6 | +| 7.1 | Monaural recording ..... | 6 | +| 7.2 | Stereo recording..... | 6 | +| 7.3 | Binaural recording ..... | 6 | +| 7.4 | Sound field recording ..... | 8 | +| 7.5 | Multi-point sound field recording ..... | 9 | +| 7.6 | Recording format..... | 13 | +| 7.7 | Calibration of recording instrumentation ..... | 14 | +| 8 | Environments for reproducing artificial noise fields ..... | 14 | +| 8.1 | Loudspeaker configurations ..... | 15 | +| 8.2 | Loudspeaker calibration and equalization ..... | 15 | +| 8.3 | Reproducing artificial noise fields in a simulation set-up ..... | 42 | +| 9 | Techniques for synthesizing noise fields ..... | 42 | +| 9.1 | Equalization procedure ..... | 43 | +| 9.2 | Recommended set-ups..... | 43 | +| 10 | Techniques for assessing artificial noise fields ..... | 43 | +| 10.1 | Assessment based on spectral content..... | 43 | +| 10.2 | Assessment based on spatial coherence ..... | 43 | +| 10.3 | Assessment of uniformity..... | 44 | +| 11 | Documentation of measurements in artificial noise fields ..... | 44 | +| | Annex A – Loudspeaker performance ..... | 45 | +| | Appendix I – Requirements for test environments ..... | 46 | +| | Appendix II – Loudspeaker configurations and decoding of loudspeaker signals ..... | 48 | +| | Appendix III – Spatial sound field stability ..... | 54 | +| III.1 | Reproduction of monophonic signals in anechoic conditions..... | 54 | +| III.2 | Reproduction of monophonic signals in treated rooms..... | 66 | +| III.3 | Reproduction of stereophonic signals in anechoic conditions ..... | 66 | +| III.4 | Reproduction of stereophonic signals in treated rooms ..... | 66 | +| III.5 | Reproduction of binaural signals in anechoic conditions..... | 66 | + +| | Page | | +|-------------------|-------------------------------------------------------------------------------------|-----| +| III.6 | Reproduction of binaural signals in treated rooms..... | 66 | +| III.7 | Reproduction of sound field signals in anechoic conditions ..... | 66 | +| III.8 | Reproduction of sound field signals in treated rooms ..... | 66 | +| III.9 | Reproduction of multi-point signals in anechoic conditions and treated rooms ..... | 66 | +| Appendix IV – | Spatial reproduction stability ..... | 89 | +| IV.1 | Spatial reproduction accuracy of monophonic signals under anechoic conditions ..... | 89 | +| IV.2 | Spatial reproduction accuracy of monophonic signals in treated rooms ..... | 89 | +| IV.3 | Spatial Reproduction Accuracy of stereophonic signals in anechoic conditions ..... | 105 | +| IV.4 | Spatial Reproduction Accuracy of stereophonic signals in treated rooms ..... | 105 | +| IV.5 | Spatial Reproduction Accuracy of binaural signals in anechoic conditions .. | 105 | +| IV.6 | Spatial Reproduction Accuracy of binaural signals in treated rooms ..... | 105 | +| IV.7 | Spatial reproduction accuracy sound field signals in anechoic conditions .... | 112 | +| IV.8 | Spatial reproduction accuracy of sound field signals in treated rooms ..... | 112 | +| IV.9 | Spatial reproduction accuracy of multi-point reproduction..... | 112 | +| Appendix V – | Detailed description of the overall equalization procedure for a HATS ..... | 114 | +| Bibliography..... | | 115 | + +# **Introduction** + +This Recommendation addresses various loudspeaker set-ups that can be used for reproduction of typical environment noise scenarios. Its scope is limited to artificial noise fields under laboratory conditions; whereas actual performance testing of handset, headset or hands-free terminals when placed in an environment with background noise are covered by other relevant ITU Recommendations. + + + +# Artificial noise fields under laboratory conditions + +# 1 Scope + +The quality of ambient noise transmission is an important factor, which significantly contributes to the perceived overall quality of speech. Existing and even more the new generation of terminals, networks and system configurations including broadband services can be greatly improved with a proper design of terminals and systems in the presence of ambient noise. + +This Recommendation: + +- provides information about techniques needed for capturing real-world noise environments; +- outlines methods for preprocessing of signals for artificial noise fields (ANFs); +- specifies loudspeaker set-ups and a loudspeaker calibration procedure; +- describes the requirements for laboratory conditions; +- specifies techniques for assessing ANFs. + +The ANF established when applying this Recommendation can be used for the evaluation and optimization of terminals and of complex configurations including terminals, networks and other configurations. + +The ANF described in this Recommendation is applicable to: + +- objective performance evaluation of terminals in different ambient noise environments; +- speech-processing evaluation by using the preprocessed speech signal in the presence of ambient noise, recorded by a terminal; +- subjective evaluation of terminals by performing conversational tests, specific double talk tests or talking and listening tests in the presence of ambient noise; +- subjective evaluation in third party listening tests by recording the speech samples of terminals in the presence of ambient noise. + +# 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- [[ITU-T P.58](#)] Recommendation ITU-T P.58 (2023), *Head and torso simulator for telephonometry*. +- [[ITU-T P.581](#)] Recommendation ITU-T P.581 (2022), *Use of head and torso simulator for hands-free and handset terminal testing*. +- [[ITU-T P.1100](#)] Recommendation ITU-T P.1100 (2019), *Narrowband hands-free communication in motor vehicles*. +- [[ITU-T P.1110](#)] Recommendation ITU-T P.1110 (2019), *Wideband hands-free communication in motor vehicles*. + +- [ITU-T P.1120] Recommendation ITU-T P.1120 (2017), *Super-wideband and fullband stereo hands-free communication in motor vehicles*. +- [ITU-T P.1140] Recommendation ITU-T P.1140 (2022), *Speech communication requirements for emergency calls originating from vehicles*. +- [ITU-R BS.1116-3] Recommendation ITU-R BS.1116-3 (2015): *Methods for the subjective assessment of small impairments in audio systems*. +- [ETSI EG 202 396-1] ETSI EG 202 396-1, V1.2.2 (2008), *Speech processing, transmission and quality aspects (STQ); Speech quality performance in the presence of background noise; Part 1: Background noise simulation technique and background noise database*. +- [ETSI TS 103 224] Technical Specification ETSI TS 103 224, V1.6.1 (2022), *Speech processing, transmission and quality aspects (STQ); A sound field reproduction method for terminal testing including a background noise database*. +- [IEC 61094-4] IEC 61094-4:1995, *Measurement microphones – Part 4: Specifications for working standard microphones*. +- [IEC 61260-1] IEC 61260-1:2014, *Electroacoustics – Octave-band and fractional-octave-band filters – Part 1: Specifications*. +- [IEC TR 60268-13] IEC TR 60268-13:1998, *Sound system equipment – Part 13: Listening tests on loudspeakers*. +- [ISO 11904-1] ISO 11904-1:2002, *Acoustics – Determination of sound immission from sound sources placed close to the ear – Part 1: Technique using a microphone in a real ear (MIRE technique)*. + +# 3 Definitions + +Headset: device which includes telephone receiver and transmitter which is typically secured to the head or the ear of the wearer. + +## 3.1 Terms defined elsewhere + +This Recommendation uses the following term defined elsewhere: + +**3.1.1 hands-free terminal** [b-ITU-T P.10]: A telephone set that does not require the use of hands during the communications session; examples are headset, speakerphone and group-audio terminal. + +## 3.2 Terms defined in this Recommendation + +This Recommendation defines the following terms: + +**3.2.1 A-format**: A signal format used within a sound field microphone that is normally not accessible to the outside world. It consists of the four signals from an array of microphone capsules that are placed at the vertices of a regular tetrahedron. These signals are processed electronically to produce B-format. + +NOTE – See [b-Rumsey] for more information. + +**3.2.2 ambisonics**: A series of recording and replay techniques using multichannel mixing technology that can be used live or in the studio. By encoding and decoding sound information on a number of channels, a two-dimensional or three-dimensional sound field can be presented. + +NOTE – See [b-Rumsey] for more information. + +**3.2.3 artificial ear:** Device incorporating an acoustic coupler and a calibrated microphone for the measurement of the sound pressure and having an overall acoustic impedance similar to that of the median adult human ear over a given frequency band. + +NOTE – Based on [b-ITU-T P.10]. + +**3.2.4 B-format:** In the basic version, known as first-order ambisonics, a format in which sound information is encoded into four channels: *W*, *X*, *Y* and *Z*. The *W* channel is the non-directional mono component of the signal, corresponding to the output of an omni-directional microphone. The *X*, *Y* and *Z* channels are the directional components in three dimensions. They correspond to the outputs of three figure-of-eight microphones, facing forward, to the left, and upward respectively. + +NOTE – See [b-Rumsey] for more information. + +**3.2.5 C-format; UHJ:** A format that is a matrixed version of B-format for transmission over a reduced number of channels with associated decreases in decoding accuracy. + +NOTE – See [b-Rumsey] for more information. + +**3.2.6 D-format:** The format of the decoded signals to be applied to the loudspeakers in the ambisonic system. + +NOTE – See [b-Rumsey] for more information. + +**3.2.7 diffuse-field equalization:** Equalization of the head and torso simulator (HATS) sound pick-up, equalization of the difference, in decibels, between the spectrum level of the acoustic pressure at the eardrum reference point (DRP) and the spectrum level of the acoustic pressure at the HATS reference point (HRP) in a diffuse sound field with the HATS absent using the reverse nominal curve given in Table 3 of [ITU-T P.58]. + +NOTE – Based on [ITU-T P.1100]. + +**3.2.8 diffuse-field test point (DFTP):** The physical position of the measurement microphone during the loudspeaker equalization procedure. The centre of the head of the head and torso simulator (HATS) will be located at the DFTP when the actual measurements are performed. + +**3.2.9 eardrum reference point (DRP):** A point located at the end of the ear canal, corresponding to the eardrum position. + +NOTE – Based on [b-ITU-T P.10]. + +**3.2.10 free-field equalization:** The transfer characteristics of the artificial head are equalized in such a way that, for a frontal sound incidence in anechoic conditions, the frequency response of the artificial head is flat. This equalization is specific to the head and torso simulator (HATS) used. + +NOTE – Based on [ITU-T P.1100]. + +**3.2.11 free-field reference point:** A point located in the free sound field, at a distance of at least 1.5 m from a sound source radiating in free air [in the case of a head and torso simulator (HATS) in the centre of the artificial head with no artificial head present]. + +NOTE – Based on [ITU-T P.1100]. + +**3.2.12 hands-free reference point (HFRP):** A point located on the axis of the artificial mouth, at 50 cm from the outer plane of the lip ring, where the level calibration is made, under free-field conditions. + +NOTE 1 – The HFRP corresponds to measurement point 11, as defined in [ITU-T P.581]. + +NOTE 2 – Based on [ITU-T P.1100]. + +**3.2.13 head and torso simulator (HATS) for telephonometry:** Manikin extending downward from the top of the head to the waist, designed to simulate the sound pick-up characteristics and the acoustic diffraction produced by a median human adult and to reproduce the acoustic field generated by the human mouth. + +NOTE – Based on [b-ITU-T P.10]. + +**3.2.14 headset:** A device, comprising a telephone receiver and transmitter, that is typically secured to the head or the ear of the wearer. + +NOTE – Based on [b-ITU-T P.10]. + +**3.2.15 listening room:** A room that provides standard listening conditions in which audiovisual equipment, hi-fi systems, loudspeakers or recordings can be assessed subjectively. The acoustics are similar to those of a living room in a house rather than the 'deadness' of an anechoic room. + +NOTE –See [IEC TR 60268-13] for more information. + +**3.2.16 mouth reference point (MRP):** A point located 25 mm in front of and on the axis of the lip plane of a mouth simulator. + +NOTE – Based on [b-ITU-T P.10]. + +**3.2.17 pantophonic:** Two-dimensional representations of sound field. + +**3.2.18 periphonic:** Three-dimensional representations of sound field. + +# 4 Abbreviations and acronyms + +This Recommendation uses the following abbreviations and acronyms: + +| | | +|-------|------------------------------------------| +| ANF | Artificial Noise Field | +| DFTP | Diffuse-Field Test Point | +| DRP | Drum Reference Point | +| DUT | Device Under Test | +| EEP | Ear canal Entrance Point | +| FFT | Fast Fourier Transform | +| G-MOS | Global Mean Opinion Score | +| HATS | Head and Torso Simulator | +| HRP | Head and torso simulator Reference Point | +| MIRE | Microphone In Real Ear | +| MLS | Maximum Length Sequence | +| MOS | Mean Opinion Score | +| MRP | Mouth Reference Point | +| N-MOS | Noise Mean Opinion Score | +| PSD | Power Spectral Density | +| RFR | Room Frequency Response | +| RIR | Room Impulse Response | +| SNR | Signal to Noise Ratio | +| S-MOS | Speech Mean Opinion Score | + +# 5 Conventions + +dBPa: sound pressure level relative to 1 Pa, expressed in decibels. + +dBPa(A): A-weighted sound pressure level relative to 1 Pa, expressed in decibels. + +dB SPL: sound pressure level relative to 20 $\mu$ Pa, expressed in decibels; (94 dB SPL = 0 dBPa). + +# 6 How to use this Recommendation + +This Recommendation describes various methods for capturing real-world noise environments, as well as configurations and methods for reproducing such noise environments. The configurations that are described cover a reference set-up, such as an anechoic chamber, as well as an *in-situ* set-up, e.g., in a vehicle. This Recommendation also provides a detailed description of the capture of noise environments in a vehicle when using binaural recording techniques. + +As an alternative to using a binaural recording method for capturing noise environments in a vehicle, methods that use a single microphone e.g., the microphone of the hands-free terminal in the vehicle or a measurement microphone, are described in [ITU-T P.1100], [ITU-T P.1110], [ITU-T P.1120] and [ITU-T P.1140], which also specify configurations for specific applications of simulation set-ups intended for use in a vehicle. Note that this Recommendation also covers the loudspeaker configuration defined in [ETSI EG 202 396-1]. + +Table 1 lists the combinations of signals and configurations considered in this Recommendation. Each combination is evaluated with respect to four parameters. + +- **Spectral coherence (A)** – is used to evaluate how well the spectrum of the reproduced sound field correlates with the original spectrum of the recorded sound field. For the methods described in this Recommendation, the difference between the original spectrum and the reproduced spectrum is known to be within $\pm 3$ dB. +- **Temporal coherence (B)** – is used to evaluate how well the time signals of the reproduced sound field correlate with the original time signals of the recorded sound field. The actual performance depends on the individual method described in this Recommendation; hence detailed information is given in each method description. +- **Positional uniformity (C)** – is used to evaluate the sound field beyond the centre of calibration with respect to spectral coherence. +- **Perceptual characteristics (D)** – is used to evaluate characteristics such as spatial resolution, sound image localization and perceived acoustical space properties. See [ITU-R BS.1116-3] for more information on subjective assessment. + +Table 1 forms the basis for guidelines for selection of the proper configuration for testing the performance of a specific application e.g., handset, handheld hands-free, desktop hands-free, headset, vehicle mounted hands-free, objective performance, instrumental, subjective performance, perceptual performance and headphones with active noise control. As a general rule, the reproduction used for testing should be more advanced if more advanced signal processing is provided in the device under test (DUT), i.e., technologies that more accurately reproduce the original sound field should be used. + +"ABCD" indicates that all four parameters are supported by the configuration while a dash "-" indicates that this configuration is not applicable. + +**Table 1 – Characteristics of various reproduction set-ups** + +| No.
loudspeakers | Signal | | | | | | +|---------------------|--------|------------------|----------|-------------|-----------|-----------------| +| | 1D | 2D (pantophonic) | | | | 3D (periphonic) | +| | Mono | Stereo | Binaural | Multi point | Ambisonic | Ambisonic | +| 1 | – | – | – | – | – | – | +| 2 | – | – | – | – | – | – | +| 4 | AB | – | ABC | – | – | – | +| 8 | – | – | – | ABCD | – | – | +| More | – | – | – | – | – | – | + +# **7 Techniques for capturing real-world noise environments** + +In this clause, the techniques identified as applicable to the capture of real-world noise environments are described. + +## **7.1 Monaural recording** + +The simplest technique that can be applied for recording purposes is a single microphone used to capture the sound field. In general, such recording will lack any information about the directivity of the sound, but will give accurate information about the sound pressure level at the position in the sound field where the microphone was positioned during the recording. + +In order to have an optimal recording, the correct type of microphone must be utilized and the positioning, as well as orientation, of the microphone must be correct. For the applications outlined in this Recommendation, the choice will typically be to use a diffuse-field microphone. Furthermore, the appropriate equalization should be considered. + +## **7.2 Stereo recording** + +The stereo recording technique utilizes two microphones and preserves the sounds that have been picked up by the microphones as two separate, but simultaneously recorded, signals. This technique allows the actual physical and spatial relationships that existed at the time the event took place to be preserved in the recording. + +A large number of different microphone configurations for stereo recordings are used for various applications. However, they are all based on the fundamental principle that has been described in the previous paragraph. None of the recordings currently used within the scope of this Recommendation are stereo recordings; nonetheless, the stereo technique is described for completeness. + +## **7.3 Binaural recording** + +The general principle of the recording and playback technique is to provide an instrument (a head and torso simulator (HATS) or test subject wearing binaural probe microphones) that allows the recording as close to the original as possible of the signals typically received by the human ear in a sound field. A suitable description of the properties of the HATS to be used for the recording techniques can be found in [ITU-T P.58]. Since the directivity characteristics of the HATS are strongly directionally dependent, the output signals of the microphones of the HATS cannot be directly compared to a standard measurement microphone and an equalization procedure has to be applied in order to get comparable output signals. Various types of equalization procedure for HATS recordings can be used, as described in clauses 7.3.1 to 7.3.5. + +#### **7.3.1 Free-field equalization** + +The HATS is equalized in such a way that for frontal sound incidence in anechoic conditions, the frequency response of the head of the HATS is flat. + +### **7.3.2 Diffuse-field equalization** + +The HATS is equalized in a diffuse sound field. For random sound incidence, the frequency response of the HATS is flat. + +#### **7.3.3 Binaural equalization** + +Generally, the HATS used should comply with the requirements specified in [ITU-T P.58]. Although all equalization methods described in clause 7.3 can be used, the preferred method is diffuse-field equalization. This is because there are less-sharp peaks and dips in the transfer function of the HATS compared to the free-field transfer function (see also [ITU-T P.58]). + +All equalization techniques can be applied to the HATS as well as to a binaural probe microphone, which is worn by the user himself instead of using a HATS. If probe microphone techniques (so- + +called microphone in real ear (MIRE) techniques, see [ISO 11904-1]) are used, the same equalization technique as applied for the HATS needs to be used. The preferred recording methodology, however, is the HATS technique, since the performance characteristics of HATSs are well described and problems due to incorrect positioning, differences for different persons wearing the binaural microphone and equalization errors are unlikely. The equalization technique chosen for the recording process should be similar to that used in the equalization process for the reproduction set-up. The recording set-up is outlined in Figure 1. + +![Figure 1: Diagram comparing two methods of artificial noise field recording. On the left, a HATS (Head and Torso Simulator) is shown with two dashed lines leading from its ears to a 'Recorder' box. On the right, a 'Binaural microphone worn by subject' is shown, also with two dashed lines leading from its ears to a 'Recorder' box. A vertical dashed line separates the two methods. The label 'P.570(18)_F01' is at the bottom right.](bd0b93e7a46ede276d0a3b79ac487bd9_img.jpg) + +Figure 1: Diagram comparing two methods of artificial noise field recording. On the left, a HATS (Head and Torso Simulator) is shown with two dashed lines leading from its ears to a 'Recorder' box. On the right, a 'Binaural microphone worn by subject' is shown, also with two dashed lines leading from its ears to a 'Recorder' box. A vertical dashed line separates the two methods. The label 'P.570(18)\_F01' is at the bottom right. + +**Figure 1 – The different methods of artificial noise field recording addressed in this Recommendation. The methods are head and torso simulator and microphone in real ear** + +#### 7.3.4 Free-field equalization + +The free field is equalized for the free-field reference position in an anechoic room. The reference source is placed on the reference axis of the HATS, at a minimum distance of 1.5 m from the HATS lip-ring. When using a binaural microphone, this point corresponds to the lip plane of a human user wearing the binaural microphone. The measured free-field response of the HATS is: + +$$H_{\text{ff}}(0^\circ, 0^\circ, f)$$ + +From this the free-field equalization is calculated: + +$$H_{\text{ff-EQ}}(0^\circ, 0^\circ, f) = 1/H_{\text{ff}}(0^\circ, 0^\circ, f)$$ + +$H_{\text{ff-EQl}}(0^\circ, 0^\circ, f)$ is the free-field equalization of the left ear, $H_{\text{ff-EQr}}(0^\circ, 0^\circ, f)$ is the free-field equalization of the right ear. The equalization should be correct to within $\pm 0.5$ dB in the frequency range from 100 Hz to 10 kHz and to within $\pm 2.0$ dB in the frequency range from 10 kHz to 20 kHz measured in one-third-octave bands. + +NOTE – The free-field equalization of the HATS requires a very careful set-up of the measurement. Due to the directivity of the HATS, small deviations from the reference position ( $0^\circ, 0^\circ$ ) may result in incorrectly equalized transfer functions. + +#### 7.3.5 Diffuse-field equalization + +The diffuse field is equalized for the diffuse-field reference position in a diffuse sound field. Either an echoic room or an arrangement of multiple loudspeakers each fed by an uncorrelated noise signal and generating a truly diffusive sound field is used. The HATS is positioned at the centre of the diffusive sound field. The measured diffuse-field response of the HATS is: + +$$H_{\text{df}}(f)$$ + +From this the diffuse-field equalization is calculated: + +$$H_{\text{df-EQ}}(f) = 1/H_{\text{df}}(f)$$ + +$H_{\text{df-EQl}}(f)$ is the diffuse-field equalization of the left ear; $H_{\text{df-EQr}}(f)$ is the diffuse-field equalization of the right ear. The equalization should be correct to within $\pm 0.5$ dB in the frequency range from 100 Hz to 10 kHz and to within $\pm 2.0$ dB in the frequency range from 10 kHz to 20 kHz measured in one-third-octave bands. + +## 7.4 Sound field recording + +To present a specific sound field, a series of recording and replay techniques using multichannel mixing technology has been developed over the years. These techniques are based on encoding and decoding sound information on a number of channels – and the techniques are named ambisonics. + +The principles of ambisonics define a unified and scalable system that deals with spatial capture and coding of acoustical sound fields, as well as transmission and reproduction using loudspeakers. + +The principles can also be applied to synthesize alternative sound capture techniques or encode sources not captured or modelled using ambisonic techniques for subsequent processing and reproduction using ambisonic systems. + +Several good overviews (see [b-Malham], [b-Daniel] and [b-Rumsey]) of the principles for spherical harmonic decomposition are available (see [b-Gerzon, 1973] and [b-Gerzon, 1975]), and this Recommendation only gives an introduction to ambisonics. + +Ambisonic theory defines the decomposition of a sound field into an orthogonal set of Laplace spherical harmonic functions that characterize the wave front impacting the centre of a unit sphere. If a plane wave approaches with angles of incidence $\theta$ for the azimuth and $\varphi$ for the elevation with respect to the pole of a sphere, then its ambisonic components can be represented by four algebraic definitions: + +$$W = 1$$ + +$$X = \cos \theta \times \cos \varphi$$ + +$$Y = \sin \theta \times \cos \varphi$$ + +$$Z = \sin \varphi$$ + +The "order" defines the degrees of freedom used to define the entire surface of the unit sphere, so the $W$ component is the zeroth-order and the $X$ , $Y$ and $Z$ components provide the first-order definitions. + +For acoustic encoding purposes, these definitions approximate to $W$ being an omni-directional pressure component and $X$ , $Y$ and $Z$ being bi-directional (figure-of-eight) velocity components, all being coincident to each other. + +By convention, the $X$ component defines the front-back sensitivity, the $Y$ component defines the left-right sensitivity, and the $Z$ component defines the up-down sensitivity. Where all components are used, the resulting definition is termed periphonic as it includes the $Z$ height component. + +In practice, it is very difficult to achieve coincidence between separate microphone capsules having these responses. A sound field microphone provides a practical realization of such a system by using four capsules with well-matched cardioid ( $1 + \cos \theta$ ) or sub-cardioid ( $2 + \cos \theta$ ) responses that are positioned in a tetrahedral configuration. An associated preamplifier normalizes for level and applies fixed delay compensation, such that each capsule is virtually coincident, although the accuracy results in some phase differences between channels for frequencies above approximately 12 kHz. + +In ambisonic terminology, the four resulting signals from these capsules are called A-format, referenced to their positions, where the vertices of the tetrahedron formed by the centre of each capsule face are coincident with the vertices of a cube. The signals will be directed towards: + +$$L_{\text{FU}} = \text{Left/front/up}$$ + +$$R_{\text{FD}} = \text{Right/front/down}$$ + +$L_{BD} = \text{Left/back/down}$ + +$R_{BU} = \text{Right/back/up}$ + +which can be easily summed within the preamp to produce the equivalent $W$ , $X$ , $Y$ and $Z$ signals, called B-format. One important thing to note for B-format is that the $W$ signal typically has a higher amplitude than the $X$ , $Y$ and $Z$ signals, so it has been the convention with such sound field microphone systems that the preamp outputs are amplitude weighted to increase the levels of the $X$ , $Y$ and $Z$ channels by 3 dB to get a better signal-to-noise ratio or higher signal delivered to the recorder. This means that the relative intensities of each output channel will be: + +$$W = 1$$ + +$$X = \sqrt{2} (\cos \theta \times \cos \phi)$$ + +$$Y = \sqrt{2} (\sin \theta \times \cos \phi)$$ + +$$Z = \sqrt{2} (\sin \phi)$$ + +and must be accounted for when decoding. Where the sound field needs only be defined in the horizontal (azimuthal) plane, then the $Z$ component and $\cos \phi$ terms can be omitted and the resulting definition is termed pantophonic. + +#### 7.4.1 Practical applications and considerations + +When using ambisonic techniques to simulate real-world noise environments, there are a number of things that need to be considered when evaluating the pros and cons in relation to other sound field capture or reproduction methods. + +The scheme requires at least three signals for a pantophonic system or four signals for a periphonic system. Although this seems less efficient than other systems, the number of decoding options to simulate different spatial response patterns is much greater than for individual or pairs of microphones (e.g., mono, stereo [b-Alexander] or binaural). + +For capture, a portable system can easily be used either where a HATS cannot easily be set up or when there are practical reasons why a person with a MIRE system cannot be present during the recording. Although a portable set-up, including the sound field microphone (with shock-mount and windshield) and battery-powered preamplifier and multichannel recorder, is used very often in audiovisual production, it does attract attention more than, for example, MIRE, which can be an issue when discretion is needed while making recordings. + +## 7.5 Multi-point sound field recording + +The multi-point sound field recording technique is based on optimization of sound field reproduction at different points in space. The optimization criterion is based on minimization of the reproduction error at each microphone position. Based on this principle, the microphone locations and, as a consequence, the points in space for which the sound field reproduction is most accurate, can be chosen in a wide range. The advantage of the method is that these locations can be adapted to the type of device that is to be tested. For example, if the DUT incorporates a microphone array of the multi-point sound field, recording microphones can be positioned in the area of the microphones of the DUT. If a hands-free device is to be tested, the multi-point sound field recording microphones are positioned in the area of the hands-free device. + +The arrangements described in detail in clauses 7.5.1 to 7.5.3 are optimized for the testing of handset or headset terminals using a HATS according to [ITU-T P.58]. The procedure described here can be followed in the same way for other microphone set-ups. + +#### **7.5.1 Microphone set-up for handset, headset and handsfree terminals** + +When measuring handset, headset and handsfree terminals using HATS a multi-point recording of the sound field around the head of HATS is required. In clause 7.5.1.1 the microphone set-up for handset and headset terminals is described and in clause 7.5.1.2 the microphone set-up for handsfree terminals is described. + +#### **7.5.1.1 Microphone set-up for handset and headset terminals** + +Figure 2 shows the configuration of microphones located around an artificial head, which is also used for assessment of the system. The locations of the microphones define the sweet spots at which the reproduction of the recorded signals is optimal for all frequencies. As a result, the majority of these points are selected at relevant positions at which test device microphones are usually located (see Figure 2, top left). The exact positions of the eight recording microphones have been determined by extensive simulations and are given in Figure 2 (bottom). Eight additional positions can be identified by rotation of the microphone array (fine-tuning set in Figure 2, top right, darker spots). The fine-tuning position is defined by a clockwise rotation of $10^\circ$ of the microphones around the HATS axis of rotation. + +![Figure 2: Positions of the recording microphones. The figure consists of three diagrams. The top two are top-down views of a head model showing microphone positions. The middle diagram is a front view with vertical and horizontal dimensions for positions 1-8 and EEP. The bottom diagram is a side view with vertical dimensions for positions 1-8, EEP, and CL.](b2d16e07bfa79d67a8adabf7e26c7764_img.jpg) + +The figure illustrates the standard positions for recording microphones on a head model. The top two diagrams show top-down views of the head with microphone positions indicated by grey circles. The middle diagram is a front view showing the vertical and horizontal placement of the microphones. The bottom diagram is a side view showing the vertical placement relative to the ear canal entrance point (EEP) and the center line (CL). + +**Front View Dimensions (mm):** + +- Vertical positions (relative to mouth reference point): + - Pos 8: 42.5 + - Pos 7: 0 + - Pos 6: 42.5 + - Pos 5: 87.3 + - Pos 4: 102.9 + - EEP: 128 +- Horizontal positions (relative to center line): + - Pos 1: 106.5 + - Pos 2: 91.1 + - Pos 3: 87.3 + - Pos 4: 37.4 + - Center: 0 + - Pos 5: 87.3 + - Pos 6: 91.1 + - Pos 7: 106.5 + +**Side View Dimensions (mm):** + +- Vertical positions (relative to mouth reference point): + - Pos 6: 14.1 + - Pos 8: 0 + - Pos 1, 7: 30.8 + - EEP: 43 + - Pos 4: 43 + - Pos 3: 43 + - CL: 43 + - Pos 2, 5: 43 + +Figure 2: Positions of the recording microphones. The figure consists of three diagrams. The top two are top-down views of a head model showing microphone positions. The middle diagram is a front view with vertical and horizontal dimensions for positions 1-8 and EEP. The bottom diagram is a side view with vertical dimensions for positions 1-8, EEP, and CL. + +P.570(18)\_F02 + +**Figure 2 – Positions of the recording microphones. Vertical positions are related to the vertical position of the mouth reference point; EEP: ear canal entrance point** + +#### 7.5.1.2 Microphone set-up for handsfree terminals + +In general, different microphone arrays could be used for hands-free terminals as well as for handsets and headsets. However, to increase reusability and reduce efforts, the same microphone array can be used in both cases. The set-up of the array for measuring hands-free terminals is shown in Figure 3. + +For the hands-free equalization, the DUT is first positioned at its testing position, which is defined in the relevant standards. Then, the main microphone position of the terminal is determined. In the case of terminals using multi-microphone techniques terminals, the main microphone is chosen, and in case of array techniques the acoustical centre of the array (typically identical to the centre of the array) is used. + +In the set up for hand-held and tablet terminals, the microphone array is positioned such that microphone 5 is in top view, right-angled in front of the main microphone position at a 25 mm distance (Figure 3, right) and microphone 6 is at the height of the main microphone position (Figure 3, left). + +![Figure 3: Positions of the recording microphones in a hands-free set-up for hand-held and tablet terminals. The diagram shows two views of a microphone array. The left view shows the array from the side, with the DUT (Device Under Test) positioned above it. The right view shows the array from the top, with the DUT positioned in the center. Microphone 5 is positioned 25 mm to the right of the main microphone or acoustical centre. Other microphones are labeled Pos. 1 through Pos. 8.](715cc21be1e312bb659bf6de41740408_img.jpg) + +The diagram illustrates the microphone array setup for hands-free terminals. On the left, a side view shows the microphone array positioned below the DUT. The main microphone or acoustical centre is indicated by a red dot. Microphone 5 is positioned 25 mm to the right of the main microphone. On the right, a top view shows the microphone array positioned in front of the DUT. Microphone 5 is positioned 25 mm to the right of the main microphone. Other microphones are labeled Pos. 1 through Pos. 8. The text 'P.570(25)' is visible in the bottom right corner. + +Figure 3: Positions of the recording microphones in a hands-free set-up for hand-held and tablet terminals. The diagram shows two views of a microphone array. The left view shows the array from the side, with the DUT (Device Under Test) positioned above it. The right view shows the array from the top, with the DUT positioned in the center. Microphone 5 is positioned 25 mm to the right of the main microphone or acoustical centre. Other microphones are labeled Pos. 1 through Pos. 8. + +**Figure 3 – Positions of the recording microphones in a hands-free set-up for hand-held and tablet terminals** + +For desktop-operated hands-free terminals, microphone 5 of the array is positioned right-angled in front of the main microphone position at a 25 mm distance (Figure 4, right) and 25 mm above the table (Figure 4, left). The "fine-tuning set" is realized the same way as described in clause 5.2, rotating the microphone array clockwise by 10 degrees. + +![Figure 4: Positions of the recording microphones in a hands-free set up for desktop operated hands-free terminals. The diagram shows a Device Under Test (DUT) on a table. Microphones are positioned at various points: Pos. 8 (top left), Pos. 5 (top center), Pos. 6 (top right), Pos. 7 (top right), Pos. 4 (right side), Pos. 3 (bottom right), Pos. 2 (bottom right), and a 'Main microphone or acoustical centre' (red dot on the DUT). A 25 mm dimension line is shown between the DUT and the table. A keypad with numbers 1-9 is visible on the DUT. The text 'P.570(25)' is in the bottom right corner.](793eb94053441c45bcf1e1fad773a7eb_img.jpg) + +Figure 4: Positions of the recording microphones in a hands-free set up for desktop operated hands-free terminals. The diagram shows a Device Under Test (DUT) on a table. Microphones are positioned at various points: Pos. 8 (top left), Pos. 5 (top center), Pos. 6 (top right), Pos. 7 (top right), Pos. 4 (right side), Pos. 3 (bottom right), Pos. 2 (bottom right), and a 'Main microphone or acoustical centre' (red dot on the DUT). A 25 mm dimension line is shown between the DUT and the table. A keypad with numbers 1-9 is visible on the DUT. The text 'P.570(25)' is in the bottom right corner. + +**Figure 4 – Positions of the recording microphones in a hands-free set up for desktop operated hands-free terminals** + +Note that the DUT is absent during the equalization procedure itself. + +### 7.5.2 Calibration of recording instrumentation + +As stated before, it is fundamental for this method that, at the defined positions in the real sound field and at the corresponding positions in the reproduced sound field, the sound pressures over time are identical. Therefore, for every recording, the location, frequency or phase response and the directional sensitivity for each microphone have to be identical. As a consequence, the same microphone set-up is used for both the recording and equalization procedures. Every microphone is calibrated individually. + +### 7.5.3 Limitations caused by the selected microphone set-up + +If the sound field is reproduced perfectly at two closely located points, the distance between those two points corresponds to an upper critical frequency up to which the sound field in between those two points is also reproduced correctly. This upper critical frequency is comparable to the Nyquist-criterion and therefore can be estimated as + +$$f_{\text{lim}} = \frac{c}{2d_{\text{max}}} \quad (7-1)$$ + +where $d_{\text{max}}$ is the maximum distance between two microphones and $c$ is the velocity of sound. For the eight microphones in Figure 2, $f_{\text{lim}}$ is dependent on position and is about 1.7 kHz in the sparsely spaced microphone region and approximately 3 kHz in that of densely spaced microphones. Note that at the microphone positions themselves, the reproduction quality is optimal across the whole frequency range. In between these positions, accurate spatial reproduction can only be guaranteed up to $f_{\text{lim}}$ . + +## 7.6 Recording format + +The recording techniques described in this Recommendation should by default be conducted without any preprocessing and filtering. This implies that the recording should be without equalization and that filtering should only be applied to the extent that this is required by the process of storing the recording in either a digital or analogue representation. + +When a digital representation is used to store the recording, a sampling rate of at least 48 kHz should be used to enable the full audio band – 20 Hz to 20 kHz – to be utilized for the applications covered in this Recommendation. Different sampling rates can be used, but note that lower sampling rates reduce the effective frequency range of the recording. + +It is anticipated that the majority of recordings relevant to applications of this Recommendation will utilize a digital representation; hence no further description of an analogue representation is given here. + +The recording equipment should be able to handle the noise scenarios being recorded without introducing distortion and without adding any recording noise. + +The format in which the recording should be preserved is as an uncompressed audio file. Other formats can be used, but the recording should be structured according to the following schema. + +- Mono recording: The single microphone signal is stored in a single uncompressed audio file. +- Stereo recording: The two microphone signals are stored in a single uncompressed audio file. +- Binaural recording: The right signal and the left signal are stored in a single uncompressed audio file. It must be guaranteed that recorded real-field sound pressure levels can be reproduced correctly, e.g., by storing calibration data in the file header. +- Ambisonic recording: The B-format should be stored in a single uncompressed audio file. It must be guaranteed that recorded real-field sound pressure levels can be reproduced correctly, e.g., by storing calibration data in the file header. +- Multi-point sound field recording: Eight channels have to be stored with high accuracy, preferably 24-bit or more. It must be guaranteed that recorded real-field sound pressure levels can be reproduced correctly, e.g., by storing calibration data in the file header. + +Any calibration recording used must be stored in a separate uncompressed audio file and the format for the associated recording method must be applied. + +## 7.7 Calibration of recording instrumentation + +Noise environment recording methods described in this Recommendation require the recording instruments to be fully calibrated. When using a measurement microphone, equipment such as a HATS or MIRE only requires that a sensitivity level calibration be performed, typically by using a simple sound level calibrator. + +For recording instruments utilizing a sound field microphone, the supplier of such devices should provide information regarding the sensitivity of the individual microphones constituting the sound field microphone, as well as a suitable calibration procedure for the complete sound field microphone for verification purposes. + +For all recordings, ensure that the recording system has the same fixed delays and gains applied to each channel. Given the broad dynamic range of noise environments that might be encountered, and the associated instantaneous peak levels, it is preferable to record signals digitally with a resolution of at least 24-bit fixed point accuracy, which should allow a greater variety of environment levels to be recorded without needing to change the gain of the system. + +# 8 Environments for reproducing artificial noise fields + +A broad categorization of the testing environments in which noise sources or fields can be reproduced, independent of the source material or method of generation, can be defined as follows. + +- Reference set-up – acoustic presentation of noise sources utilizing transducers placed about a DUT within a reference physical environment having known acoustic properties (e.g., an anechoic chamber, a listening room, an acoustically treated office, a reverberant room). +- *In-situ* set-up – acoustic presentation of noise sources utilizing transducers placed about or within the physical environment in which a DUT will be utilized (e.g., a motor vehicle cab interior). +- Simulation set-up – "Virtual" presentation of noise sources within a reference modelled test environment (e.g., finite element model, boundary element model, analytical model). + +Each of these categories shares some common ways of defining how noise sources are presented for testing or auralization purposes. This is not intended to include loudspeakers producing individual sound events (e.g., competing speech) that are independent of the primary noise field simulation, although such independent sources can be assigned to loudspeakers that are part of the noise field system. + +## 8.1 Loudspeaker configurations + +For the creation of ANFs, an important part of the system is the actual loudspeaker configuration being used. This Recommendation addresses various loudspeaker set-ups; however, all set-ups require that the influence of the loudspeaker configuration e.g., the individual loudspeaker as well as the test environment, be well controlled and stable with respect to certain acoustical properties. + +To be able to establish a well-defined equalization of the loudspeaker configuration is an essential requirement. In the scope of this Recommendation, various equalization procedures are defined each serving different purposes. + +## 8.2 Loudspeaker calibration and equalization + +In clauses 8.2.1 to 8.2.9, different methods for equalization of loudspeaker configurations are described. For each recording method, a reproduction method and the associated equalization procedure is outlined. + +### 8.2.1 Loudspeaker equalization using the HATS method in a reference set-up + +The HATS-based equalization method described in this clause is intended for application to reproduction of binaural recordings when the loudspeaker configuration shown in Figure 5 is used. The distance between the centre of the acoustical sound field (the position used for equalization and where the test arrangement is typically placed) should be 2 m. The distance between each loudspeaker and the centre of the simulation set-up should be equal and the distance between two adjacent loudspeakers should be equal. The height of the four loudspeakers should be chosen so that the centre of the acoustical field is of the same height as the loudspeaker positions. The positioning of the subwoofer is mostly uncritical; however, placing the subwoofer exactly in the corner of a room should be avoided. + +![Diagram of the HATS method setup for loudspeaker equalization.](6f31cdb576d2f15c35c3f266e5f59211_img.jpg) + +The diagram illustrates the experimental setup for loudspeaker equalization using the HATS method. On the left, a 'Generator' unit is shown with five input channels labeled 'a' through 'e'. Each channel contains a symbol for a sine wave, a triangle representing an amplifier, and a 'τ' symbol representing a time delay. These channels are connected to four loudspeakers arranged in a square configuration: 'Left front', 'Right front', 'Left back', and 'Right back'. A central point represents the 'acoustical sound field' or 'test arrangement', indicated by three overlapping circles. Dashed lines connect each loudspeaker to this central point, with each line labeled '~2m' to indicate the distance. Below the square arrangement, an 'Analyser' unit is shown. Two dashed lines, labeled '(HATS R ear)' and '(HATS L ear)', point from the central area towards the analyser, representing the right and left ears of the HATS (Head and Torso Simulator) model. The diagram is labeled 'P.570(25)' in the bottom right corner. + +Diagram of the HATS method setup for loudspeaker equalization. + +**Figure 5 – The set-up for loudspeaker equalization using the HATS method and the physical set-up that is used when conducting the actual test of the device e.g., mobile phone** + +For the equalization, a HATS according to [ITU-T P.58] is used. The right and the left ear of the HATS should be equalized and the equalization shall correspond to the equalization used for the binaural recording of the sound field (free-field, diffuse-field, see clause 7). The HATS is positioned at the centre of the acoustical sound field (middle of the four loudspeakers, see Figure 5). The equalization is done binaurally e.g., the two left-hand speakers are equalized to the left ear of the HATS and the right-hand loudspeakers are equalized to the right ear of the HATS. In practice, the signal of the left-hand loudspeakers will also reach the right ear of the HATS, which impairs the equalization. The influence of this cross-talk cannot be eliminated, but it may be minimized as described in the equalization procedure. + +The equalization is done in several steps: + +- 1) separate equalization for each of the four loudspeakers; +- 2) separate level adjustment for each loudspeaker; +- 3) equalization for the two left-hand and the two right-hand loudspeakers; +- 4) equalization and level adjustment for the subwoofer; +- 5) delay compensation; +- 6) equalization for all loudspeakers (verification and reduction of cross-talk). + +Each step is described in clauses 8.2.1.1 to 8.2.1.6. + +#### 8.2.1.1 Step 1) – Separate equalization for each of the four loudspeakers + +For loudspeaker equalization, a pink noise signal is used – the same signal is inserted at a, b, c, d and e in Figure 5. The signal is sequentially applied to the left-front, the left-back, the right-back and the right-front loudspeaker. The pink noise signal applied to the four loudspeakers must cover the frequency range 120 Hz to 20 kHz. The frequency range below is cut off by a high-pass filter providing a roll-off characteristic of at least 18 dB/octave. For specifications of octave-band and fractional-octave-band filters, see [IEC 61260-1]. + +The filtered pink noise signal is applied at the front-left loudspeaker for at least 2 s. During this period, the output of the HATS left ear as well as the filtered pink noise signals are measured simultaneously. The power density spectrum for the ear signal – $S_{\text{Ls}}(f)$ – and for the original signal – $S_{\text{pn}}(f)$ – are determined and the averaged magnitude of the two power density spectra are subtracted as follows: + +$$H_{\text{Ls}}(f) = S_{\text{Ls}}(f)/S_{\text{pn}}(f) \quad (8-1)$$ + +The resulting frequency response – $H_{\text{Ls}}(f)$ – shall be flat between 120 Hz and 10 kHz within a tolerance of $\pm 3\text{dB}$ when determined in one-third octave. + +To fulfil this condition, a filter needs to be designed that compensates for the frequency response of the loudspeaker in the room. The compensating filter frequency response is calculated from the initially measured loudspeaker frequency response by using the inverse filter function: + +$$H_{\text{EQ}}(f) = 1/H_{\text{Ls}}(f) \quad (8-2)$$ + +The equalized noise signal is applied at the left-front loudspeaker. + +#### 8.2.1.2 Step 2) – Separate level adjustment for each loudspeaker + +After each of the loudspeakers is equalized separately, each sound pressure level should be adjusted. To reproduce the original level of any sound source with this arrangement, the level of each of the four loudspeakers should be 6 dB lower than that of the original sound source, due to the fact that the sound source is now reproduced by four loudspeakers. This level adjustment can be achieved by recording and analysing the played back signal for each loudspeaker as described in clause 8.2.1.1 and after that, adjusting the level to one 6 dB below the original. + +#### **8.2.1.3 Step 3) – Equalization for the two left-hand and the two right-hand loudspeakers** + +In this step, the two – separately equalized – left-hand or right-hand loudspeakers are equalized in combination (see Figure 5). Filtered pink noise is used as the test signal. The test signal is synchronously applied to the left-hand loudspeakers (front and back) and – similarly to 8.2.1.1 – measured with the HATS left ear and spectrally analysed in one-third octave. Again, the measured difference in the HATS left ear signal and the original signal should be within the tolerance of $\pm 3$ dB when determined in one-third octave. If the measured difference does not comply, the EQ filter for one or both of the left-hand loudspeakers is re-designed to pass the tolerance scheme. The right-hand loudspeakers are equalized by applying the same procedure, but in this situation the HATS right ear is used for the measurement instead of the left. + +NOTE – The tolerance scheme for a single loudspeaker might be violated, if the equalization for the combined left loudspeakers passes the tolerance scheme. + +#### **8.2.1.4 Step 4) – Equalization and level adjustment for the subwoofer** + +The subwoofer is equalized for the frequency range between 30 Hz and 120 Hz. The frequency range above is cut off by a low-pass filter providing a roll-off characteristic of at least 18 dB/octave. Pink noise is used as the test signal. Similar to the equalization for the four loudspeakers, the pink noise is applied to the subwoofer and measured at the right ear of the HATS. On the basis of the difference between the measured signal at the HATS ear and the original pink noises spectrum, a filter is designed with which this difference (analysed in one-third octave) is spectrally flat within a tolerance of $\pm 3$ dB. The level is adjusted to the level of the original sound source. + +#### **8.2.1.5 Step 5) – Delay adjustment** + +For the four-loudspeaker arrangement, a compensation for the individual loudspeaker delay to the centre of the acoustical field is required in order to keep the correlation of the recorded noise material with respect to the centre of the acoustical field. Incorrect delay adjustment might lead to comb filter effects, which leads to unnatural sounding of the background noise signals. The procedure consists of adding fixed delays to the four loudspeakers (no compensation is needed for the subwoofer) in order to increase the diffusivity of the sound field. Thus, in all types of rooms, the following delays are added to the individual loudspeakers: 0 ms, 11 ms, 17 ms, 29 ms. If not suitable, as room and loudspeaker properties, dimensions and positions are not always the same, these delay values may be replaced by better ones proven to lead to a better overall equalization. No delays higher than 50 ms shall be used. If the resulting overall frequency response is mostly flat (measured at both ears), an appropriate selection of delays can be assumed. + +#### **8.2.1.6 Step 6) – Overall equalization for all loudspeakers** + +The last step is to verify the equalization process, including the delay compensation. Therefore, a stationary, realistic pre-recorded broadband background noise signal covering the entire frequency range, e.g., car noise, is fed to the four-loudspeaker arrangement. With all equalizers, level and delay adjustments active, the signal is played back by all loudspeakers including the subwoofer (see Figure 5). The reproduced sound field is measured at both ears of the HATS. For each ear, the measurement is analysed in one-third octave and referenced to the spectrum of the original signal. The measured frequency response for the left ear and the right ear shall both be within a tolerance of $\pm 3$ dB. + +NOTE – If the result is outside the tolerance, out of the four, the loudspeaker contributing mostly to the equalization error needs to be checked. A detailed description of this procedure can be found in Appendix V. + +#### **8.2.2 Loudspeaker equalization using the microphone method in a reference set-up** + +The microphone-based equalization method described in this clause can be applied for reproduction of monaural recordings for various loudspeaker configurations. However, the following description is based on the same loudspeaker configuration shown in Figure 6. + +For equalization, a measurement microphone complying with [IEC 61094-4] is used. The measurement microphone is positioned at the centre of the acoustical sound field in a position equivalent to the centre of the HATS head when the HATS is positioned in the middle of the four loudspeakers – see Figure 6. Each loudspeaker is equalized individually as follows. + +![Figure 6: The set-up for loudspeaker equalization using the microphone method. The diagram shows a Generator with five outputs (a-e) connected to a series of blocks: a sine wave symbol, a triangle (amplifier), and a tau symbol (delay). These are connected to four loudspeakers labeled 'Left front', 'Right front', 'Left back', and 'Right back'. A measurement microphone is positioned at the center of the four loudspeakers, indicated by dashed lines and '~2m' distance markers. The microphone output is connected to an Analyser. The diagram is labeled P.570(25).](0c08e48c08f96934cd6bc6911f3069dc_img.jpg) + +Figure 6: The set-up for loudspeaker equalization using the microphone method. The diagram shows a Generator with five outputs (a-e) connected to a series of blocks: a sine wave symbol, a triangle (amplifier), and a tau symbol (delay). These are connected to four loudspeakers labeled 'Left front', 'Right front', 'Left back', and 'Right back'. A measurement microphone is positioned at the center of the four loudspeakers, indicated by dashed lines and '~2m' distance markers. The microphone output is connected to an Analyser. The diagram is labeled P.570(25). + +**Figure 6 – The set-up for loudspeaker equalization using the microphone method** + +The equalization is done in several steps: + +- 1) separate equalization for each of the four loudspeakers; +- 2) separate level adjustment for each loudspeaker; +- 3) equalization and level adjustment for the subwoofer; +- 4) delay compensation; +- 5) equalization for all loudspeakers (verification and reduction of cross-talk). + +Each step is described in clauses 8.2.2.1 to 8.2.2.5. + +#### 8.2.2.1 Step 1) – Separate equalization for each of the four loudspeakers + +For the equalization of the loudspeakers, a pink noise signal is used. The signal is sequentially applied to the left-front, the left-back, the right-back and the right-front loudspeaker. The pink noise signal applied to the four loudspeakers must cover the frequency range 120 Hz to 20 kHz. The frequency range below is cut off by a high-pass filter providing a roll-off characteristic of at least 18 dB/octave. + +The filtered pink noise signal is applied at the front-left loudspeaker for at least 2 s. During this period, the output of the measurement microphone as well as the filtered pink noise signal are measured simultaneously. The power density spectrum for the microphone signal – $S_{\text{Ls}}(f)$ – and for the original signal – $S_{\text{pn}}(f)$ – are determined and the averaged magnitude of the two power density spectra are subtracted as follows: + +$$H_{\text{Ls}}(f) = S_{\text{Ls}}(f)/S_{\text{pn}}(f) \quad (8-3)$$ + +The resulting frequency response – $H_{\text{Ls}}(f)$ – shall be flat between 120 Hz and 10 kHz within a tolerance of $\pm 3$ dB when determined in one-third octave. + +To fulfil this condition, a filter needs to be designed that compensates for the frequency response of the loudspeaker in the room. The compensating filter frequency response is calculated from the initially measured loudspeaker frequency response by using the inverse filter function: + +$$H_{EQ}(f) = 1/H_{Ls}(f) \quad (8-4)$$ + +The equalized noise signal is applied at the left-front loudspeaker. + +#### **8.2.2.2 Step 2) – Separate level adjustment for each loudspeaker** + +After each of the loudspeakers is equalized separately, each sound pressure level should be adjusted. To reproduce the original level of any sound source with this arrangement, the level of each of the four loudspeakers should be 6 dB lower than that of the original sound source, due to the fact that the sound source is now reproduced by four loudspeakers. This level adjustment can be achieved by recording and analysing the played back signal for each loudspeaker as described in clause 8.2.2.1 and after that, adjusting the level to one 6 dB below the original. + +#### **8.2.2.3 Step 3) – Equalization and level adjustment for the subwoofer** + +The subwoofer is equalized for the frequency range between 30 Hz and 120 Hz. The frequency range above is cut off by a low-pass filter providing a roll-off characteristic of at least 18 dB/octave. Pink noise is used as test signal. Similarly to the equalization for the four loudspeakers, the pink noise is applied to the subwoofer and measured with the measurement microphone at a position equivalent to the centre of the HATS head. On the basis of the difference between the measured signal at the microphone and the original pink noise spectrum, a filter is designed with which this difference (analysed in one-third octave) is spectrally flat within a tolerance of $\pm 3$ dB. The level is adjusted to the level of the original sound source. + +#### **8.2.2.4 Step 4) – Delay adjustment compensation** + +For the four-loudspeaker arrangement, a compensation of the individual loudspeaker delay to the centre of the acoustical field is required in order to keep the correlation of the recorded noise material with respect to the centre of the acoustical field. Incorrect delay adjustment might lead to comb filter effects that cause unnatural sounding of the background noise signals. The procedure consists of adding fixed delays to the four loudspeakers (no compensation is needed for the subwoofer) in order to increase the diffusivity of the sound field. Thus, in all types of rooms, the following delays are added to the individual loudspeakers: 0 ms, 11 ms, 17 ms, 29 ms. If not suitable, as room and loudspeakers properties, dimensions and positions are not always the same, these delay values may be replaced by ones proven to lead to a better overall equalization. No delays higher than 50 ms shall be used. If the resulting overall frequency response is mostly flat, an appropriate selection of delays can be assumed. + +#### **8.2.2.5 Step 5) – Overall equalization for all loudspeakers** + +The last step is to verify the equalization process, including the delay compensation. Therefore, the signal that was used in the equalization procedure is fed to the loudspeaker arrangement – including the subwoofer. With all equalizers, level and delay adjustments active, the signal is played back by all loudspeakers including the subwoofer (see Figure 6). The reproduced sound field is measured in one-third octave with the measurement microphone and referenced to the spectrum of the original signal. The measured frequency response for the loudspeakers as well as the subwoofer shall be within a tolerance of $\pm 3$ dB. + +#### **8.2.3 Loudspeaker configuration for a car using one microphone in a reference set-up** + +Some types of terminals are mostly used in a car environment, e.g., portable hands-free devices or hands-free terminals that are integrated into cars. For portable hands-free devices, the set-up and equalization process is similar to that described for an office except that the HATS is positioned at the driver's seat instead of at the centre of the four loudspeakers. In this Recommendation, the term "portable hands-free devices" refers to handsets as well as headsets. + +For hands-free terminals installed in a car, in principle, the same set-up is used for ANF reproduction as in the reference set-up described in clause 9.1. Four loudspeakers and a subwoofer are used. The + +subwoofer is positioned in the boot of a car. The two rear loudspeakers are positioned in the back of the car between the headrest and the window or the C-pillar. The two front loudspeakers are positioned either above the instrument panel or at the bottom in the foot well (only if the instrument panel position is impossible), depending on the tested car set-up. Figure 7 is a block diagram of the set-up. + +![Figure 7: The loudspeaker configuration for a car. The diagram shows a 'Generator' unit on the left with five input channels labeled a, b, c, d, and e. Each channel contains a sine wave symbol, a triangle (amplifier), and a tau symbol (delay). The outputs are connected to a car cabin diagram on the right. Channel 'a' connects to the 'Left front' speaker. Channel 'b' connects to the 'Right front' speaker. Channel 'c' connects to the 'Left back' speaker. Channel 'd' connects to the 'Right back' speaker. Channel 'e' connects to a subwoofer located in the rear of the cabin. The cabin diagram shows the front seats, rear seats, and the trunk area.](d17aa1fcc3b86503ad1dd0717a6c34c2_img.jpg) + +Figure 7: The loudspeaker configuration for a car. The diagram shows a 'Generator' unit on the left with five input channels labeled a, b, c, d, and e. Each channel contains a sine wave symbol, a triangle (amplifier), and a tau symbol (delay). The outputs are connected to a car cabin diagram on the right. Channel 'a' connects to the 'Left front' speaker. Channel 'b' connects to the 'Right front' speaker. Channel 'c' connects to the 'Left back' speaker. Channel 'd' connects to the 'Right back' speaker. Channel 'e' connects to a subwoofer located in the rear of the cabin. The cabin diagram shows the front seats, rear seats, and the trunk area. + +**Figure 7 – The loudspeaker configuration for a car** + +Note that the use of the four loudspeaker arrangement in a car is more sensitive to the positioning of the microphone than the simulation in an office. Due to the smaller dimensions of a car cabin and the surrounding reverberant windows, standing waves and interferences appear. + +Therefore, it is highly recommended that the same microphone be used for equalization and terminal testing. This requires that the microphone be separated from the hands-free terminal and used for system equalization, as well as in combination with the hands-free terminal. The microphone shall then be located in the position intended for use after installation. Equalization with the terminal microphone is described in clause 9.2.4. + +If the terminal microphone signal is not accessible for equalization and calibration, a pair of cardioid microphones should be used instead. They should be located as close as possible to the expected terminal microphone position to provide a simulated sound field that is as close as possible to the correct position. The equalization with a pair of cardioid microphones is described in clause 9.2.5. + +#### **8.2.4 Loudspeaker equalization in a car using a HATS** + +The procedure used for equalization in the reference set-up can be applied. For further details, see clause 8.2.1. + +### **8.2.5 Loudspeaker equalization in a car using the microphone of the hands-free terminal** + +For equalization, the microphone of the hands-free terminal should be located at the position where it will be used e.g., under the ceiling or in or on top of the rear view mirror. The equalization is done in several steps: + +- 1) separate equalization for each of the four loudspeakers; +- 2) separate level adjustment for each loudspeaker; + +- 3) equalization for the two left-hand and the two right-hand loudspeakers; +- 4) equalization and level adjustment for the subwoofer; +- 5) delay adjustment; +- 6) overall equalization. + +Each step is described in clauses 8.2.5.1 to 8.2.5.6. + +#### 8.2.5.1 Step 1 – Separate equalization for each of the four loudspeakers + +For the equalization of the loudspeakers, a stationary, realistic broadband noise (e.g., broadband car noise recorded under constant driving conditions) covering the frequency range 120 Hz to 20 kHz is used. The realistic broadband noise signal is sequentially applied to the left-front, the left-back, the right-back and the right-front loudspeaker. The frequency range below is cut off by a high-pass filter providing a roll-off characteristic of at least 18 dB/octave. + +The filtered signal is applied at the front-left loudspeaker for at least 10 s. During this period, the output of the microphone of the hands-free terminal as well as the filtered realistic broadband noise signal are measured simultaneously. The power density spectrum for the microphone signal – $S_{\text{Ls}}(f)$ – and for the original signal – $S_{\text{bbn}}(f)$ – are determined and the averaged magnitude of the two power density spectra are subtracted as follows: + +$$H_{\text{Ls}}(f) = S_{\text{Ls}}(f)/S_{\text{bbn}}(f) \quad (8-5)$$ + +The resulting frequency response – $H_{\text{Ls}}(f)$ – shall be flat between 120 Hz and 10 kHz within a tolerance of $\pm 3$ dB when determined in one-third octave. + +To fulfil this condition, a filter needs to be designed that compensates for the frequency response of the loudspeaker in the car. The compensating filter frequency response is calculated from the initially measured loudspeaker frequency response by using the inverse filter function: + +$$H_{\text{EQ}}(f) = 1/H_{\text{Ls}}(f) \quad (8-6)$$ + +The equalized noise signal is applied at the left-front loudspeaker. + +#### 8.2.5.2 Step 2 – Separate level adjustment for each loudspeaker + +After each of the loudspeakers is equalized separately, each sound pressure level should be adjusted. To reproduce the original level of any sound source with this arrangement, the level of each of the four loudspeakers should be 6 dB lower than that of the original sound source, due to the fact that the sound source is now reproduced by four loudspeakers. This level adjustment can be achieved by recording and analysing the played back signal for each loudspeaker as described in clause 8.2.5.1 and after that, adjusting the level to one 6 dB below the original. + +NOTE – For microphones with a strong directivity to the back of the car, an improved simulation can be achieved by adjusting the sound pressure level of the rear loudspeakers to a level approximately 5 dB higher and the sound pressure level of the front loudspeakers to a level approximately 3 dB lower. Be aware that the overall sound pressure level is still the same as for the original sound. + +#### 8.2.5.3 Step 3 – Equalization for the two front and the two rear loudspeakers + +In this step, the two – separately equalized – front or rear loudspeakers are equalized in combination (see Figure 8). A stationary, realistic broadband noise signal is used as the test signal. This test signal is used in order to achieve the same frequency-dependent correlation as for typical binaural recordings and to avoid disturbing influences due to the periodicity of the pink noise. The test signal is synchronously applied to the front loudspeakers and – similar to step 1 – measured with the microphone of the hands-free terminal and analysed in one-third octave. The measured difference in the microphone and original signals should be within the 6 dB tolerance. If the measured difference does not comply, the filter for one or both of the front loudspeakers should be re-designed to pass the tolerance scheme. The equalizations for the rear loudspeakers are performed in a similar manner. + +NOTE – The tolerance scheme for a single loudspeaker might be violated if the equalization for the combined front loudspeakers passes the tolerance scheme. + +![Figure 8: Diagram of a car loudspeaker configuration for equalization. A 'Generator' box on the left contains five channels (a-e), each with a sine wave symbol, an amplifier symbol, and a delay symbol (τ). These channels are connected to four car speakers: 'Left front', 'Right front', 'Left back', and 'Right back'. A 'Terminal microphone' is positioned in the center of the car's interior. An 'Analyser' box is connected to the output of the terminal microphone. The diagram is labeled P.570(25).](5414f65867392f05ba0063b208eeb5e1_img.jpg) + +Figure 8: Diagram of a car loudspeaker configuration for equalization. A 'Generator' box on the left contains five channels (a-e), each with a sine wave symbol, an amplifier symbol, and a delay symbol (τ). These channels are connected to four car speakers: 'Left front', 'Right front', 'Left back', and 'Right back'. A 'Terminal microphone' is positioned in the center of the car's interior. An 'Analyser' box is connected to the output of the terminal microphone. The diagram is labeled P.570(25). + +**Figure 8 – The loudspeaker configuration for a car when the microphone of the hands-free terminal is used during equalization** + +#### 8.2.5.4 Step 4 – Equalization and level adjustment for the subwoofer + +The subwoofer is equalized for the frequency range between 30 Hz and 120 Hz. The frequency range above is cut off by a low-pass filter providing a roll-off characteristic of at least 18 dB/octave. A stationary, realistic broadband noise is used as test signal. Similar to the equalization for the four loudspeakers, the broadband noise is applied to the subwoofer and measured with the microphone of the hands-free terminal. On the basis of the difference between the microphone signal and the original broadband noise spectrum, a filter is designed with which this difference (analysed in one-third octave) should be spectrally flat in a 6 dB tolerance scheme. The level is adjusted to that of the original sound source. + +#### 8.2.5.5 Step 5 – Delay adjustment + +The delay from each loudspeaker to the position of a hand-free terminal microphone should be measured and adjusted. The delay should not be assigned to one common delay for all loudspeakers, which would produce the highest signal correlation at the position of the microphone. The delays are adjusted to values that will not provide a high correlation at the microphone position, but do provide a mostly diffuse sound field. With this method, the sound directivity is not correctly reproduced; however, equalization with a single microphone is in any case not able to reproduce the sound directivity. The advantage of providing a diffuse sound field at the microphone position is that the spectral equalization matches very well. For the delay adjustment, the following rules should be followed: + +- the delay for front loudspeakers should be shorter than that for those at the rear; +- the minimum delay difference should be greater than the length of fast Fourier transform (FFT) used (typically > 20 ms); +- the delays, including acoustical delay, should be chosen so that they are not integral multiples of each other. + +#### 8.2.5.6 Step 6 – Overall equalization + +The last step is to verify the equalization process, including the delay compensation. Therefore, a stationary, realistic broadband background noise signal is fed to the four-loudspeaker arrangement. With all equalizers, level and delay adjustments active, the signal is played back by all loudspeakers including the subwoofer (see Figure 8). The reproduced sound field is measured with the microphone of the hands-free terminal. The measured frequency response at the microphone position should be within a tolerance of $\pm 3$ dB. The measured A-weighted signal level should match the original signal level within a tolerance of $\pm 1$ dB. + +NOTE – If the result is outside the tolerance, it needs to be checked which of the four loudspeakers mostly contributes to the equalization error. A detailed description of this procedure can be found in Appendix V. + +#### 8.2.6 Loudspeaker equalization using the multi-point method in a reference set-up with a predefined microphone and loudspeaker arrangement + +An overview of the benefits and limitations of the method described in this clause can be found in clause 6. + +#### 8.2.6.1 Loudspeaker configuration + +Figure 9 shows the set-up of the eight loudspeakers in the laboratory. The vertical position of the loudspeakers in general is not so critical. However, in order to achieve the best possible interlaboratory set-up accuracy, the vertical positions are specified as follows. For the first set of loudspeakers, the vertical position is adjusted so that the centre of the loudspeaker membrane is about $+15$ cm above the vertical position defined by the DRP. The second set of loudspeakers is adjusted to $-15$ cm below with respect to the DRP. The vertical positions of the loudspeakers alternate. In Figure 9, odd-numbered loudspeakers represent the first set and even numbered the second. The distance between the surface of the artificial head and the loudspeaker fronts should be at least 50 cm. The maximum distance should not extend 2.5 m. The distance between the loudspeakers and the HATS, as well as the horizontal distribution of the loudspeakers, can be selected depending on the room, hence the spacing between the loudspeakers does not have to be exactly equal. The set-up may be a square or a circle around the HATS or a set-up in between, depending on what fits the room best. Note that the maximum distance is also limited by the maximum sound pressure level of the loudspeakers, because for frequencies down to 50 Hz sound fields including levels of up to 105 dBSPL should be reproducible. + +![Diagram of a loudspeaker set-up around a central HATS (Head and Torso Simulator). Eight loudspeakers (LS 1 to LS 8) are arranged in a circular fashion. LS 1 is at the top, LS 2 at the top-right, LS 3 at the right, LS 4 at the bottom-right, LS 5 at the bottom, LS 6 at the bottom-left, LS 7 at the left, and LS 8 at the top-left. Dashed lines represent equalization paths from the HATS to various loudspeakers, labeled H12, H13, H14, H41, H48, and H47. The diagram is labeled P.570(25) at the bottom right.](fc0735d325f0ebd9214171975c68a888_img.jpg) + +The diagram shows a central circle representing the HATS with numbered points 1 through 8 on its perimeter. Surrounding this are eight loudspeaker icons labeled LS 1 through LS 8. LS 1 is at the top (12 o'clock), LS 2 at 1:30, LS 3 at 3 o'clock, LS 4 at 4:30, LS 5 at 6 o'clock, LS 6 at 7:30, LS 7 at 9 o'clock, and LS 8 at 10:30. Dashed arrows indicate transfer functions: $H_{12}$ , $H_{13}$ , and $H_{14}$ originate from point 1 towards LS 2, LS 3, and LS 4 respectively. Similarly, $H_{41}$ , $H_{48}$ , and $H_{47}$ originate from point 4 towards LS 1, LS 8, and LS 7 respectively. + +Diagram of a loudspeaker set-up around a central HATS (Head and Torso Simulator). Eight loudspeakers (LS 1 to LS 8) are arranged in a circular fashion. LS 1 is at the top, LS 2 at the top-right, LS 3 at the right, LS 4 at the bottom-right, LS 5 at the bottom, LS 6 at the bottom-left, LS 7 at the left, and LS 8 at the top-left. Dashed lines represent equalization paths from the HATS to various loudspeakers, labeled H12, H13, H14, H41, H48, and H47. The diagram is labeled P.570(25) at the bottom right. + +Figure 9 – Loudspeaker set-up and some equalization paths + +#### 8.2.6.2 Room requirements + +Rooms where the procedure has been verified for range from $1.80\text{ m} \times 2.40\text{ m} \times 2.1\text{ m}$ to $8.00\text{ m} \times 9.00\text{ m} \times 4.5\text{ m}$ ( $d \times b \times h$ ) in size. Rooms may range from anechoic rooms to acoustically treated office-type rooms. The most important criterion a room has to fulfil is the clarity 80 ( $C_{80}$ ). This + +parameter defines the signal energy of the first 80 ms of the impulse response in relation to the rest of the impulse response expressed in decibels. + +Suitable rooms shall have $C_{80} > 20\text{dB}$ . + +$$C_{80} = 10 \log_{10} \left( \frac{\int_{0\text{ms}}^{80\text{ms}} p^2(t) dt}{\int_{80\text{ms}}^{\infty} p^2(t) dt} \right) \quad (8-7)$$ + +#### 8.2.6.3 Equalization + +The complete equalization procedure consists of the following steps: + +- 1) separate level adjustment for every loudspeaker; +- 2) system identification; +- 3) recording using broad-band noise for later equalization check and error compensation; +- 4) preprocessing of the impulse responses for equalization; +- 5) calculation of the inversion filters; +- 6) first test of inversion with recorded noise; +- 7) adjusting the filter to compensate for inversion errors. + +For equalization, the same microphone array set-up and microphone positions should be used as for the recording set-up. Accordingly, the microphone array should be calibrated as described in clause 2.1.2. The equalization itself can then be performed completely automatically, independently of the microphone array set-up. See Figure 10. + +![Flowchart of the equalization procedure showing five sequential steps: Loudspeaker level adjustment, System identification, Preprocessing impulse responses, Calculation of the inversion filters, and Final equalization incl. correction of inversion errors.](94fd137860c16c8dfd75512f10161fe8_img.jpg) + +``` +graph LR; A[Loudspeaker level adjustment] --> B[System identification]; B --> C[Preprocessing impulse responses]; C --> D[Calculation of the inversion filters]; D --> E[Final equalization incl. correction of inversion errors]; +``` + +Flowchart of the equalization procedure showing five sequential steps: Loudspeaker level adjustment, System identification, Preprocessing impulse responses, Calculation of the inversion filters, and Final equalization incl. correction of inversion errors. + +P.570(25) + +**Figure 10 – Blocks of the equalization procedure** + +Each step is described in clauses 8.2.6.3.1 to 8.2.6.3.7. + +##### 8.2.6.3.1 Step 1 – Separate level adjustment for every loudspeaker + +First of all, the sound pressure level of each loudspeaker is adjusted so that the average sound pressure level across the whole frequency range and across the eight microphones is at least 70 dB SPL. This is needed only in order to ensure a sufficient signal to noise ratio (SNR) for measuring the impulse responses in clause 8.2.6.3.2. + +##### 8.2.6.3.2 Step 2 – System identification + +In this step, the impulse responses between all loudspeakers and all microphones are measured. Figure 9 shows the eight-microphone and eight-loudspeaker set-up. $H_{li}$ represents the impulse response (frequency response) from loudspeaker $l$ to microphone $i$ . + +The impulse response between every loudspeaker and every microphone must then be determined. There exist different possibilities for measuring impulse responses, e.g., using maximum length sequences (MLSs) or using swept-sine (sweeps). The advantage of sweeps is that non-linearities can easily be observed and that the SNR at lower frequencies is higher than with MLS. Using sweeps is therefore recommended for system identification. The sweep, denoted $S(f)$ , may be played back with a loudspeaker and $Y(f)$ is recorded with a microphone. The frequency response $H(f)$ can then be calculated from: + +$$H(f) = \frac{Y(f)}{S(f)} \quad (8-8)$$ + +For sufficient system identification, the frequency range in which the response is calculated should be at least 20 Hz to 20 kHz. + +##### 8.2.6.3.3 Step 3 – Recording using broad-band noise for later equalization check and error compensation + +In order to generate well-defined material for testing and optimizing the quality of the equalization procedure, a broad-band pink noise is recorded with the calibration set of microphones, as well as with the fine-tuning set. Therefore, a well-defined sound field has to be generated, e.g., by using eight uncorrelated broadband pink noise sources. In contrast to recordings done in real-world situations, both the original noise field and the simulated noise field are then known for exactly the same room. Hence, the reproduction should be perfect at the calibration set microphone positions. Moreover, the accuracy of sound field reproduction can be verified in the fine-tuning positions. Note that these positions are not incorporated into the equalization procedure. The broad-band noise level should be sufficiently high to allow for a good error calculation later on. The frequency range of the noise signal should cover all frequencies for which the system has to be equalized. A range from 20 Hz to 20 kHz is again recommended. + +##### 8.2.6.3.4 Step 4 – Preprocessing of the impulse responses for equalization + +The motivation for this preprocessing procedure is that, in rooms with a long reverberation time, it is difficult to compensate for this reverberation completely. + +In particular, sound fields at frequencies higher than $f_{lim}$ , which for the given microphone set-up is in the range between 2 kHz and 3 kHz, cannot be reproduced correctly. Hence, higher frequencies should not influence the quality of inversion. + +To cope with that, a low-pass filter with a time-variant cut-off frequency is applied to the measured impulse response. + +- 1) Filter the impulse response in the frequency domain by multiplying a window function which has the value 1 for frequencies smaller than $f_{pass}$ and 0 for frequencies greater than $f_{stop}$ . Between those two frequencies, the window function has a cosine (Hanning) characteristic. +- 2) The preprocessed impulse response is calculated in the time-domain as the weighted sum of the original impulse response and the low-pass-filtered impulse response from step (1). The time-variant weighting of the original impulse response is 1 up to $t_{min}$ after the start of the impulse response and falls to 0 with a cosine (Hanning) characteristic until $t_{max}$ after the start of the impulse response. The time-variant weighting of the low-pass-filtered impulse response from step (1) is 1 minus the weighting of the original impulse response. + +The values for these parameters are given in Table 2. + +**Table 2 – Parameters for preprocessing the impulse responses between loudspeakers and microphones** + +| | | | | +|------------|-------|-----------|------| +| $f_{pass}$ | 2 kHz | $t_{min}$ | 1 ms | +| $f_{stop}$ | 4 kHz | $t_{max}$ | 4 ms | + +The result of the above defined filter can be seen in Figure 11, where a measured impulse response can be compared with the corresponding processed impulse response. + +![Figure 11: Two side-by-side plots showing impulse responses. The left plot is the 'original (measured) impulse response' and the right plot is the 'processed impulse response'. Both plots show pressure (p/Pa) on the y-axis (ranging from -60 m to 60 m) against time (t/s) on the x-axis (ranging from 50 m to 0.1). Both plots show a sharp initial peak followed by a decaying oscillation.](183007754364096b2d89f42200cf870f_img.jpg) + +Figure 11: Two side-by-side plots showing impulse responses. The left plot is the 'original (measured) impulse response' and the right plot is the 'processed impulse response'. Both plots show pressure (p/Pa) on the y-axis (ranging from -60 m to 60 m) against time (t/s) on the x-axis (ranging from 50 m to 0.1). Both plots show a sharp initial peak followed by a decaying oscillation. + +P.570(25) + +**Figure 11 – Left: original (measured) impulse response; right: processed impulse response** + +This preprocessing introduces an error which is shown in Figure 12. This error should be compensated for in the equalization procedure later on. This will take place in step 7), in which an error compensation filter is calculated and applied to the equalization. + +![Figure 12: A plot showing the error in frequency response introduced by the processing of the impulse response. The y-axis is labeled L/dBPa and ranges from -10 to 5. The x-axis is labeled f/Hz and is logarithmic, ranging from 20 to 20 k. The curve shows a flat response around 0 dBPa until approximately 2000 Hz, where it drops sharply to about -10 dBPa at 5000 Hz, before rising back towards 0 dBPa at higher frequencies.](7848a2560bb726462c09c09bf727d81d_img.jpg) + +Figure 12: A plot showing the error in frequency response introduced by the processing of the impulse response. The y-axis is labeled L/dBPa and ranges from -10 to 5. The x-axis is labeled f/Hz and is logarithmic, ranging from 20 to 20 k. The curve shows a flat response around 0 dBPa until approximately 2000 Hz, where it drops sharply to about -10 dBPa at 5000 Hz, before rising back towards 0 dBPa at higher frequencies. + +P.570(25) + +**Figure 12 – Error in frequency response introduced by the processing of the impulse response. The curve has been subject to one-twelfth-octave smoothing** + +##### 8.2.6.3.5 Step 5 – Calculation of the inversion filters + +The inversion procedure is explained in clauses 8.2.6.3.5.1 and 8.2.6.3.5.2. After the theory has been introduced in clause 8.2.6.3.5.1, theory implementation guidelines appear in clause 8.2.6.3.5.2. + +###### 8.2.6.3.5.1 Theoretical explanation of the inversion procedure + +The starting point for developing the formulas needed for the inversion procedure is the expression of the signal $P_i(f)$ arriving at a single microphone $i$ as a linear combination of the signals $X_l(f)$ played back with loudspeaker $l$ multiplied by a filter $H_{li}(f)$ modelling the acoustic path between microphone and the loudspeaker. This can easily be seen in Figure 10. + +$$P_i(f) = \sum_{l=1}^L H_{li}(f) X_l(f) \quad (8-9)$$ + +This linear combination can be written as a linear system: + +$$\underbrace{\begin{pmatrix} H_{11} & H_{21} & \dots & H_{L1} \\ H_{12} & \ddots & & \vdots \\ \vdots & & \ddots & \vdots \\ H_{1N} & \dots & \dots & H_{LN} \end{pmatrix}}_{\mathbf{H}} \begin{pmatrix} X_1 \\ X_2 \\ \vdots \\ X_L \end{pmatrix} = \begin{pmatrix} P_1 \\ P_2 \\ \vdots \\ P_N \end{pmatrix} \quad (8-10)$$ + +As stated above, the aim of this procedure is to find the signals to be reproduced by the loudspeakers for a given set of spectra $P_i(f)$ recorded at the microphones. First of all, it can be seen that the number of loudspeakers must be at least as large as the number of microphones, otherwise the linear system is underdetermined. + +The goal of the equalization is finding compensating filters $\mathbf{C}$ . These filters are then used to generate loudspeaker signals that in combination reproduce a given sound pressure at the microphone positions. + +$$\underbrace{\begin{pmatrix} C_{11} & C_{12} & \dots & C_{1N} \\ C_{21} & \ddots & & \vdots \\ \vdots & & \ddots & \vdots \\ C_{L1} & \dots & \dots & C_{LN} \end{pmatrix}}_{\mathbf{C}} \begin{pmatrix} P_1 \\ P_2 \\ \vdots \\ P_N \end{pmatrix} = \begin{pmatrix} X_1 \\ X_2 \\ \vdots \\ X_L \end{pmatrix} \quad (8-11)$$ + +A solution of this linear system is given by: + +$$\mathbf{C} = (\mathbf{H}^* \mathbf{H})^{-1} \mathbf{H}^* \quad (8-12)$$ + +The direct solution of Equation (8-12) may introduce problems: at lower frequencies, the values of $H_{ij}$ are very similar to each other, which means that the matrix becomes ill-conditioned and the value of the condition number, + +$$\kappa = \frac{\sigma_{\max}(A)}{\sigma_{\min}(A)} \quad (8-13)$$ + +where $\sigma_{\max}(A)$ and $\sigma_{\min}(A)$ are the maximum and the minimum singular values of $A$ , rises. A large condition number may result in practically unfeasible high gains. + +Therefore, the Tikhonov-regularization [b-Kirkeby 1998] is used. + +$$\mathbf{C} = (\mathbf{H}^* \mathbf{H} + \beta \mathbf{I})^{-1} \mathbf{H}^* \quad (8-14)$$ + +Here a regularization factor $\beta$ is introduced that reduces poles and improves the condition of the matrix. The regularization factor $\beta$ is investigated in more detail in clause 8.2.6.3.5.2.2. + +If a mixture of different types of loudspeakers is used for reproduction, their various characteristics, e.g., their frequency response, can be considered with a weighted version of (8-14): + +$$\mathbf{C} = (\mathbf{H}^* \mathbf{W} \mathbf{H} + \beta \mathbf{I})^{-1} \mathbf{W} \mathbf{H}^* \quad (8-15)$$ + +The loudspeaker weighting matrix $\mathbf{W}$ is a frequency-dependent diagonal matrix. It is used as a high-pass for small loudspeakers which prevents damage to the speakers at low frequencies. $\mathbf{W}$ shall be 1 for the supported frequency range of the loudspeaker, i.e., above the $-3\text{dB}$ point of the frequency response, and fade to 0 with a cosine characteristic within a one-third-octave transition. + +###### 8.2.6.3.5.2 Implementation of inversion procedure + +All impulse responses for all combinations of loudspeakers and microphones are inverted individually. Each impulse response is first segmented into four subbands, as shown in Figure 13. Each subband is processed with a subset of impulse responses according to the microphones used in each subband, as defined in Table 3. The size of the matrix that has to be inverted is given in brackets, e.g., "Subband 1 ( $3 \times 8$ )". + +After finalizing the regularization process, the resulting inverted impulse responses are combined in the different frequency bands chosen by applying the appropriate frequency weighting (bandpass filtering) shown in the filter blocks, after the inversion blocks in Figure 13 are available. + +![Block diagram of the inversion procedure. The process starts with 'Irs (8x8-Matrix)' entering a 'Subband division' block. This branches into four parallel paths: Subband 1 (3x8, 50 Hz to 250 Hz), Subband 2 (6x8, 250 Hz to 1000 Hz), Subband 3 (7x8, 1000 Hz to 2000 Hz), and Subband 4 (7x8, 2000 Hz to 20000 Hz). Each subband path contains an 'Inversion' block followed by a 'Frequency-weighting' block (BP: 50 Hz-250 Hz, 250 Hz-1 kHz, 1 kHz-2 kHz, and 2 kHz-20 kHz respectively). The outputs of these four blocks are summed at a circular node with a plus sign. A text box next to the summing node states 'Output matrices are combined to final 8x8 output matrix'. The summed output then passes through a final 'Frequency-weighting' block (BP: 50 Hz to 20 kHz) to produce the 'Inverted Irs (8x8-Matrix)'.](339be989b91d5b1e73e5ecdc8401ca75_img.jpg) + +``` + +graph TD + Irs["Irs (8x8-Matrix)"] --> SubbandDivision["Subband division"] + SubbandDivision --> Subband1["Subband 1 (3x8) +50 Hz to 250 Hz"] + SubbandDivision --> Subband2["Subband 2 (6x8) +250 Hz to 1000 Hz"] + SubbandDivision --> Subband3["Subband 3 (7x8) +1000 Hz to 2000 Hz"] + SubbandDivision --> Subband4["Subband 4 (7x8) +2000 Hz to 20000 Hz"] + Subband1 --> Inversion1["Inversion"] + Subband2 --> Inversion2["Inversion"] + Subband3 --> Inversion3["Inversion"] + Subband4 --> Inversion4["Inversion"] + Inversion1 --> FW1["Frequency-weighting +BP: 50 Hz-250 Hz"] + Inversion2 --> FW2["Frequency-weighting +BP: 250 Hz-1 kHz"] + Inversion3 --> FW3["Frequency-weighting +BP: 1 kHz-2 kHz"] + Inversion4 --> FW4["Frequency-weighting +BP: 2 kHz-20 kHz"] + FW1 --> Sum((+)) + FW2 --> Sum + FW3 --> Sum + FW4 --> Sum + Sum --> FinalFW["Frequency-weighting +BP: 50 Hz to 20 kHz"] + FinalFW --> InvertedIrs["Inverted Irs (8x8-Matrix)"] + Sum -.-> Text["Output matrices are combined to +final 8x8 output matrix"] + +``` + +Block diagram of the inversion procedure. The process starts with 'Irs (8x8-Matrix)' entering a 'Subband division' block. This branches into four parallel paths: Subband 1 (3x8, 50 Hz to 250 Hz), Subband 2 (6x8, 250 Hz to 1000 Hz), Subband 3 (7x8, 1000 Hz to 2000 Hz), and Subband 4 (7x8, 2000 Hz to 20000 Hz). Each subband path contains an 'Inversion' block followed by a 'Frequency-weighting' block (BP: 50 Hz-250 Hz, 250 Hz-1 kHz, 1 kHz-2 kHz, and 2 kHz-20 kHz respectively). The outputs of these four blocks are summed at a circular node with a plus sign. A text box next to the summing node states 'Output matrices are combined to final 8x8 output matrix'. The summed output then passes through a final 'Frequency-weighting' block (BP: 50 Hz to 20 kHz) to produce the 'Inverted Irs (8x8-Matrix)'. + +P.570(25) + +**Figure 13 – Block diagram of the inversion procedure** + +###### 8.2.6.3.5.2.1 Different microphones for different frequency bands + +Generally, the greater the number of microphones leads, the worse the condition of the matrix to be inverted. For this reason, the number of microphones should be limited. On the other hand, spatial reproduction quality increases with a higher number of microphones. In order to further optimize the inversion quality, different numbers of microphones can be used for various frequency bands. For lower frequencies, the wavelengths are rather long and therefore microphones can be used that are far away from each other. As frequencies become higher, microphones should be used that are nearer to each other. + +Measurements have shown that the microphone combinations listed in Table 3 should be used. + +**Table 3 – Subband definitions and used microphones per subband** + +| Subband | Frequency range | Microphone positions | +|---------|-----------------------|----------------------| +| 1 | 50 Hz to 250 Hz | 1, 3, 7 | +| 2 | 250 Hz to 1 000 Hz | 1, 2, 3, 5, 7, 8 | +| 3 | 1 000 Hz to 2 000 Hz | 1, 2, 3, 4, 5, 7, 8 | +| 4 | 2 000 Hz to 20 000 Hz | 1, 3, 4, 5, 6, 7, 8 | + +###### 8.2.6.3.5.2.2 Search for best regularization factor + +The lower the regularization factor $\beta$ the more exact is the solution of the equation system. However, if the condition of the equation system is bad, the solution is unstable and may in consequence cause high loudspeaker output levels, possibly destroying the loudspeakers. Therefore, again a compromise has to be found by introducing an iterative process to find the best regularization factor. + +First the regularization factor is set to a very low value (e.g., $1^{-12}$ ). With the regularization factor chosen, the compensating filters are calculated. In conjunction with these filters and the knowledge of the maximum volume of signals to be reproduced, the maximum output level of the loudspeakers is calculated. If the loudspeakers reach their performance limit, the regularization factor is increased and the next iteration cycle is performed until an optimal value is found. + +###### 8.2.6.3.5.2.3 Search for the optimum regularization factor + +In order to find the optimum regularization factor, a (basic) iterative procedure is applied in all subbands individually. In the last subband with frequencies above 2 kHz, the wavelengths are shorter and a second iterative procedure refines the result of the first. + +###### 8.2.6.3.5.2.3.1 Basic methodology to find the optimal regularization factor + +The first iterative process used for the calculation of the inversion filters in each block of the equalization procedure (see Figure 10) is shown in Figure 14. + +![Flowchart of the basic iterative process for calculating regularization factors. The process starts with an initialization block defining variables: k (FFT-bin index), beta (regularization factor), C (compensation filter), H_p (FFT of preprocessed IRs), H_abs (absolute FFT coefficients of room IRs), W (loudspeaker weighting), and ||_abs (vector/matrix with abs. values). The main loop begins with calculating beta(k) = 1/2 * (beta_low(k) + beta_high(k)). Next, it calculates the inversion for every k: C = (H_p * W H_p + beta_s I)^-1 * W H_p*. Then, it finds the maximum volume over mics and LS for every k: max_mic, LS (|H_abs * (C_abs * X_abs)|). A decision diamond asks if beta_high(k) - beta_low(k) < 9e-14. If 'yes', the process ends. If 'no', it proceeds to another decision diamond 'Too loud?'. If 'yes' to 'Too loud?', it updates beta_low(k) = beta(k). If 'no' to 'Too loud?', it updates beta_high(k) = beta(k). Both updates loop back to the beta(k) calculation. A legend at the bottom defines '.' as value by value multiplication and '*' as matrix multiplication. The code P.570(25) is noted at the bottom right.](e64c7b989e5bdb2708cd7aefd18b06e1_img.jpg) + +$k$ : Index of FFT-bin + $\beta$ : Regularization factor + $C$ : Compensation filter + $H_p$ : FFT of preprocessed IRs + $H_{abs}$ : absolute FFT coefficients of room IRs + $W$ : Loudspeaker weighting + $||_{abs}$ : Vector/Matrix with abs. values + +$\beta(k) = \frac{1}{2} (\beta_{low}(k) + \beta_{high}(k))$ + +Calc. inversion for every $k$ : + $C = (H_p * W H_p + \beta_s I)^{-1} W H_p^*$ + +For every $k$ approximate maximum volume over mics and LS: + $\max_{mic, LS} (|H_{abs} \cdot (C_{abs} * X_{abs})|)$ + +$\beta_{low}(k) = \beta(k)$ + +$\beta_{high}(k) = \beta(k)$ + +yes + +Too loud? + +no + +no + +$\beta_{high}(k) - \beta_{low}(k) < 9e-14$ + +yes + +· value by value multiplication + \* matrix multiplication + +P.570(25) + +Flowchart of the basic iterative process for calculating regularization factors. The process starts with an initialization block defining variables: k (FFT-bin index), beta (regularization factor), C (compensation filter), H\_p (FFT of preprocessed IRs), H\_abs (absolute FFT coefficients of room IRs), W (loudspeaker weighting), and ||\_abs (vector/matrix with abs. values). The main loop begins with calculating beta(k) = 1/2 \* (beta\_low(k) + beta\_high(k)). Next, it calculates the inversion for every k: C = (H\_p \* W H\_p + beta\_s I)^-1 \* W H\_p\*. Then, it finds the maximum volume over mics and LS for every k: max\_mic, LS (|H\_abs \* (C\_abs \* X\_abs)|). A decision diamond asks if beta\_high(k) - beta\_low(k) < 9e-14. If 'yes', the process ends. If 'no', it proceeds to another decision diamond 'Too loud?'. If 'yes' to 'Too loud?', it updates beta\_low(k) = beta(k). If 'no' to 'Too loud?', it updates beta\_high(k) = beta(k). Both updates loop back to the beta(k) calculation. A legend at the bottom defines '.' as value by value multiplication and '\*' as matrix multiplication. The code P.570(25) is noted at the bottom right. + +**Figure 14 – Basic iterative process for calculating the regularization factors** + +To find a more exact solution of the equation system, an iterative procedure is used to find the lowest possible regularization factor $\beta$ . This process is needed to avoid unstable solutions which in consequence may cause high loudspeaker output levels, leading to distortions of loudspeaker signals. + +A summary of the search for the best regularization factor, as shown in Figure 11, follows. First, the variables $\beta_{low}$ and $\beta_{high}$ are initialized with a very small value for $\beta_{low}$ (e.g., $10^{-12}$ ) and a large value for $\beta_{high}$ (e.g., 10). Then, the regularization factor is chosen to be the mean of $\beta_{low}$ and $\beta_{high}$ , and the compensating filters are calculated. Using these filters, the maximum output level of the loudspeakers is approximated. This is accomplished by simulating a playback of a typical (loud) sound recording. The spectrum of this recording is represented by $X$ or $X_{abs}$ denoting its absolute version. + +NOTE – In general, different spectra can be used for $X$ . As a general rule, a spectrum representative for the maximum level and spectral content to be reproduced by the simulation system should be chosen. This can be achieved for instance by measuring typical noise situations which the system will be used for with a single measurement microphone. After measurement, the spectra of these measurements are averaged. The power spectral density (PSD) should be used for the frequency spectra. This will ensure that the calculated spectra are independent of FFT block size. The "too-loud" condition in Figure 14 is dependent on the loudspeakers. For many loudspeakers, a reasonable value for this condition is 50 dB re. 400 pPa2/Hz. + +As $X_{abs}$ is an $n$ -element vector, the matrix multiplication $C_{abs} * X_{abs}$ results in a virtual source signal that is used to simulate a playback by the loudspeakers. As the maximum level recorded at one microphone resulting from one loudspeaker needs to be determined (in contrast to the maximum at one microphone resulting from all loudspeakers playing at the same time), the value is calculated from $H_{abs}(C_{abs} * X_{abs})$ . The maximum of the resulting matrix is calculated to determine the maximum output level. + +If loudspeakers reach their performance limit, i.e., for levels greater than the "too-loud"-condition defined above, at the microphone positions, the regularization factor is increased and the next iteration cycle is performed until $\beta_{\text{low}}$ and $\beta_{\text{high}}$ converge and an optimal value for $\beta$ is found. + +Figure 15 provides an example of $\mathbf{X}$ where all eight loudspeakers in the set-up were fed with eight incoherent pink noise channels with the maximum output level of the loudspeakers with a fixed gain. Once the recording is done, the PSD function is calculated for the recordings of all eight microphones. PSD is used to ensure that the level at each frequency line does not change dependent on the frequency resolution in FFT. + +![Figure 15: A line graph showing the Power Spectral Density (PSD) in dB re. 400p Pa²/Hz on the y-axis (ranging from 20 to 100) against Frequency in kHz on the x-axis (ranging from 0 to 25). The plot shows a noisy blue line that starts at approximately 80 dB at 0 kHz, drops to around 50 dB by 5 kHz, and then fluctuates between 40 dB and 60 dB across the remaining frequency range, ending near 30 dB at 25 kHz. A label 'P.570(25)' is visible in the bottom right corner of the plot area.](16c69c0dacd3c57ae91acd114e5f5bd2_img.jpg) + +Figure 15: A line graph showing the Power Spectral Density (PSD) in dB re. 400p Pa²/Hz on the y-axis (ranging from 20 to 100) against Frequency in kHz on the x-axis (ranging from 0 to 25). The plot shows a noisy blue line that starts at approximately 80 dB at 0 kHz, drops to around 50 dB by 5 kHz, and then fluctuates between 40 dB and 60 dB across the remaining frequency range, ending near 30 dB at 25 kHz. A label 'P.570(25)' is visible in the bottom right corner of the plot area. + +**Figure 15 – An example of the maximum power spectral density across eight microphones, i.e., $\mathbf{X}$ , using incoherent pink noise individually for each of the eight loudspeakers** + +As an example, let us assume that the "best" $\beta$ is 0.5. This means that a $\beta < 0.5$ exceeds the maximum volume allowed and a $\beta > 0.5$ could be decreased because the "too loud" – condition was not reached. Table 4 shows how the algorithm is used to find the "best" $\beta$ of 0.5. + +**Table 4 – Regularization parameter during the optimization process** + +| Step | $\beta_{\text{high}}$ | $\beta_{\text{low}}$ | $\beta$ | | +|------|-----------------------|----------------------|--------------|--------------| +| 1 | 10 | $1^{-12}$ | $\sim 5$ | Not too loud | +| 2 | 5 | $1^{-12}$ | $\sim 2.5$ | Not too loud | +| 3 | 2.5 | $1^{-12}$ | $\sim 1.25$ | Not too loud | +| 4 | 1.25 | $1^{-12}$ | $\sim 0.625$ | Not too loud | +| 5 | 0.625 | $1^{-12}$ | $\sim 0.31$ | TOO LOUD! | +| 6 | 0.625 | 0.31 | $\sim 0.46$ | TOO LOUD! | +| ... | ... | ... | ... | ... | + +###### 8.2.6.3.5.2.3.2 Extended methodology to find the optimum regularization factor for frequencies above 2 kHz + +The error at microphones in-between the calibration set positions may become larger for frequencies higher than 2 kHz, since the wavelengths are quite short in this frequency range. This might become a problem if a microphone of the DUT is located at such a position. + +To overcome this, the regularization factor can be increased, which on one hand decreases the quality of reproduction at the calibration set positions, but on the other improves the variance of the error over all microphones of the calibration, as well as of the fine-tuning set. + +Therefore, a second iterative procedure is used to find the regularization factors with the lowest error variance, starting with the best $\beta$ values found during the first procedure. In each iteration, the $\beta$ values are smoothed over frequency and the filters are calculated. Then a playback is simulated by filtering a previously made reference noise recording with those calculated inverse filters **C** and convolving the result with the original impulse responses **H** before preprocessing for the calibration, as well as the fine-tuning set positions. + +For this reference noise recording, a well-defined sound field shall be generated by using eight uncorrelated broadband pink noise sources. The total signal level shall be about 70 dBSPL. This sound field shall then be recorded with microphones positioned in the calibration set and in the fine-tuning set. For hands-free or handset mode, different recordings have to be made. The frequency range of the noise signal shall cover all frequencies for which the system has to be equalized, which is the range from 20 Hz to 20 kHz. + +After that, the variance of the error between the reference noise recording and the simulated playback, both at calibration and fine-tuning position, are calculated by determining minimum and maximum deviation and calculating the difference as depicted in Figure 16. + +![Figure 16: Four plots showing error variance and its enhancement due to optimization. Top-left: Absolute error before optimization at 16 microphone positions, showing multiple overlapping colored lines. Top-right: Minimum and maximum error before optimization, with a red line for maximum and a green line for minimum. Bottom-left: Error range before optimization (difference between minimum and maximum), shown as a green line. Bottom-right: Final error range after optimization, shown as a green line with significantly reduced amplitude. All plots have frequency f/Hz on the x-axis (2000 to 4000) and error L in dB on the y-axis. The top plots use L/dB[Pa] and the bottom plots use L/dB[Pa0].](32f80b2d70752613b6577797811f3a95_img.jpg) + +Figure 16: Four plots showing error variance and its enhancement due to optimization. Top-left: Absolute error before optimization at 16 microphone positions, showing multiple overlapping colored lines. Top-right: Minimum and maximum error before optimization, with a red line for maximum and a green line for minimum. Bottom-left: Error range before optimization (difference between minimum and maximum), shown as a green line. Bottom-right: Final error range after optimization, shown as a green line with significantly reduced amplitude. All plots have frequency f/Hz on the x-axis (2000 to 4000) and error L in dB on the y-axis. The top plots use L/dB[Pa] and the bottom plots use L/dB[Pa0]. + +P.570(25) + +**Figure 16 – Example of the calculation of the error variance and its enhancement due to optimization (top left: absolute error before optimization at 16 microphone positions; top right: minimum and maximum of error; bottom left: error range before optimization (difference between minimum and maximum); bottom right: final error range after optimization)** + +In detail, the error range is calculated by the following steps: + +- 1) a playback is simulated using the non-preprocessed IRs and the filters calculated with current regularization factors; +- 2) the simulated signals from 1 as well as the expected or source signals are smoothed with one-twenty-fourth-octave resolution; +- 3) the differences between expected and simulated spectra are calculated; +- 4) minimum and maximum over all microphones are determined for every $k$ ; +- 5) then the error range is calculated for all $k$ : $E(k) = \text{MAX}(k) - \text{MIN}(k)$ + +The smoothing which is applied during the iterative process shown in Figure 17 is calculated as follows: + +- 1) calculate one- $n$ th-octave representation of the given vector; +- 2) perform cubic interpolation between the bins of the one- $n$ th-octave representation; +- 3) read values at FFT-bins from the interpolated curve. + +![Flowchart of the extended iterative process for calculating the regularization factor. The process starts with initialization: beta(k) = beta_min(k), beta_opt(k) = beta_min(k), and alpha(k) = 1.5. It proceeds through steps of updating smoothed parameters, calculating inversion, and determining error range. A nested loop checks for every k if the error range decreased to update the best regularization parameter. It then adjusts beta and checks against beta_max. The process iterates until the step size alpha becomes small enough (20log10(alpha) < 0.2).](0b7849dae424b0dd33e6386d2384643a_img.jpg) + +``` + +graph TD + Init["β(k) = βmin(k) for all k +βopt(k) = βmin(k) for all k +α(k) = 1.5 for all k"] + UpdateS1["Update smoothed regularization +parameters βs from β with +1/12th octave resolution"] + CalcC["Calc. inversion for every k: + $C = (H_p * WH_p + \beta_s I)^{-1} WH_p^*$ "] + DetermineE["Determine error range E(k) +Only if first iteration: Emin(k) = E(k), for all k"] + CheckK["Check for every k"] + DecisionE{"E(k) < Emin(k)"} + StoreBest["Store new best regularization +parameter and best error range +βopt(k) = β(k) +Emin(k) = E(k)"] + AdjustBeta["Adjust current regularization +parameter +β(k) = β(k) * α(k)"] + DecisionMax{"β(k) > βmax +or (βmax was reached once +and (E(k) > Emin(k)))"} + SetCurrent["Set current β to best +value found until now + $\beta(k) = \frac{\beta_{opt}(k)}{\alpha(k)}$ +Reduce stepsize + $\alpha(k) = \alpha^{3/4}(k)$ "] + DecisionAlpha{"for all k it is + $20\log_{10}(\alpha(k)) < 0.2$ "} + UpdateS2["Update smoothed regularization +parameters βs from β with +1/12th octave resolution"] + Output["βs(k)"] + + Init --> UpdateS1 + UpdateS1 --> CalcC + CalcC --> DetermineE + DetermineE --> CheckK + CheckK --> DecisionE + DecisionE -- no --> AdjustBeta + DecisionE -- yes --> StoreBest + StoreBest --> AdjustBeta + AdjustBeta --> DecisionMax + DecisionMax -- no --> CheckK + DecisionMax -- yes --> SetCurrent + SetCurrent --> CheckK + CheckK --> DecisionAlpha + DecisionAlpha -- yes --> UpdateS2 + UpdateS2 --> Output + DecisionAlpha -- no --> UpdateS1 + +``` + +Flowchart of the extended iterative process for calculating the regularization factor. The process starts with initialization: beta(k) = beta\_min(k), beta\_opt(k) = beta\_min(k), and alpha(k) = 1.5. It proceeds through steps of updating smoothed parameters, calculating inversion, and determining error range. A nested loop checks for every k if the error range decreased to update the best regularization parameter. It then adjusts beta and checks against beta\_max. The process iterates until the step size alpha becomes small enough (20log10(alpha) < 0.2). + +P.570(25) + +**Figure 17 – Extended iterative process for calculating the regularization factor** + +If the error range decreased, the regularization factor $\beta$ is stored as the new best value. Then $\beta$ is increased by a factor $\alpha$ , which is initialized at 1.5. If $\beta$ exceeds the upper threshold $\beta_{max} = 1$ , it is reset to a little less than the best value found and the factor is decreased. It is also reset if the deviation did not decrease and $\beta$ has already been reset once. The iteration ends if the multiplier becomes very small, i.e., $20\log_{10}(\alpha) < 0.2$ . + +##### 8.2.6.3.6 Step 6 – First test of inversion with recorded noise + +The quality of the equalization is assessed by playing back the noise recording which was made in clause 8.2.6.3.3. The reproduced sound field is recorded with the microphones of the calibration as well as with those of the fine-tuning set and compared to the reference noise recordings. The error $E(f)$ is calculated by first determining the amplitude error for each used microphone in one-ninety-sixth-octave bands, finding the minimum and the maximum error over frequency and calculating their mean. + +##### 8.2.6.3.7 Step 7 – Adjusting the filter to compensate for inversion errors + +Several factors that influence the result of the equalization have been presented. First of all, the preprocessing of the impulse responses has a negative impact on inversion quality. In addition, nonlinearities and arithmetic uncertainties introduce a deviation from the ideal output. As a last step, a correction filter $D(f)$ compensates for the impact of the impulse response preprocessing on the inversion quality as well as arithmetic uncertainties (e.g., the regularization factor). The frequency response of this filter is calculated as the reciprocal of the previously calculated mean reproduction error $E(f)$ : + +$$D(f) = \frac{1}{E(f)}. \quad (8-16)$$ + +For all frequencies below 1 800 Hz, the level is compensated for with the average of $D(f)$ in the frequency range between 50 Hz and 1 800 Hz. Above 1 800 Hz, a minimum phase version of $D(f)$ is applied to achieve causality and minimum additional delay. + +This correction function is applied to all filters in the frequency domain. In the transition from level compensation (for frequencies lower than 1 800 Hz) to minimum phase filter (for frequencies higher than 1 800 Hz), the application of fading with a cosine (Hanning) characteristic to reduce artefacts in the time domain is recommended. The bandwidth of this fading should be one-third octave. + +NOTE – Other implementations, e.g., those using a single minimum phase filter rather than having two steps in the procedure, could have negative influences on the phase for frequencies below 1 800 Hz. + +#### 8.2.6.4 Accuracy of the equalization + +After finalizing the equalization, the following criteria shall be met: + +- 1) *Spectral reproduction accuracy.* The spectrum averaged over the individual microphones and averaged in the frequency bands for which they have been equalized should be flat to within $\pm 3$ dB from 50 Hz to 10 kHz and within $\pm 6$ dB from 10 kHz to 16 kHz measured in one-third octave. The average spectrum accuracy, averaged over all microphones shall be within $\pm 3$ dB from 50 Hz to 20 kHz. + +In scenarios where frequencies down to $f_{\min}$ shall be reproduced the difference for all microphones should be within $\pm 3$ dB and shall be within a range of +3 and $-6$ dB, measured in one-third octave between $f_{\min}$ and 50 Hz. The average spectrum accuracy, averaged over all microphones shall be within $\pm 3$ dB from $f_{\min}$ to 20 kHz. + +- 2) *Level accuracy.* The level of the reproduced sound field at each microphone shall be accurate to within $\pm 1$ dB. +- 3) *Complex coherence.* The magnitude of the complex coherence (normalized cross-correlation spectrum, see [b-Carter] and [b-Priestley]) between the original signal and the reproduced signals at the fine-tuning position averaged over the individual microphones and averaged in the frequency bands for which they have been equalized, measured in one-third octave, should be $>0.9$ from 100 Hz to 1 kHz. The phase of the complex coherence should be accurate within $+45^\circ$ from 100 Hz to 1 kHz measured in one-third octave. + +NOTE – Complex coherence: in the frequency range from 100 Hz to 1 kHz the phase of the complex coherence for the microphones 3 to 8, which are used for the equalization in this frequency range, shall be accurate within $\pm 10$ degrees and within $\pm 30$ degrees in the range from 1 kHz to 1.5 kHz, both measured in 1/3 octaves. + +### 8.2.7 Loudspeaker equalization using a generalized multi-point noise simulation method with a flexible loudspeaker and microphone arrangement + +An overview of the benefits and limitations of the method described in this clause can be found in clause 6. + +This clause presents a generalization of the previously described Multi-point Method which allows the usage of the underlying equalization method for other, more flexible microphone or loudspeaker configurations. + +#### 8.2.7.1 Loudspeaker configuration + +This generalized multi-point noise simulation method requires at least two, preferably four, loudspeakers or more, which should not be placed close to each other but "as distributed as possible". In no case should the loudspeakers be in the microphone space, i.e., between the microphones. + +If the loudspeakers cannot reproduce frequencies down to 50 Hz, i.e., their $-3$ dB point of the frequency response is much higher than 50 Hz, an additional subwoofer is needed. In this case, the weighted version of the Thikonov-regularization (see Equation 8-14) is used during the inversion procedure, where the diagonal weighting matrix **W** accounts for the different frequency ranges of the loudspeakers. The loudspeakers that are used shall be categorized into three groups as described in Table 5. + +**Table 5 – Definitions of loudspeaker categories** + +| Loudspeaker category | Frequency range | +|----------------------|---------------------| +| Subwoofer | 40 Hz to 120 Hz | +| Large loudspeaker | 50 Hz to 22 000 Hz | +| Small loudspeaker | 120 Hz to 22 000 Hz | + +The values for the weighting matrix **W** are then 1 for the given frequency range and faded to 0 with a cosine characteristic within a one-third octave transition. + +#### 8.2.7.2 Microphone set-up + +No fixed microphone arrangement is required as it has to be arranged individually for every application. Either the output signal of the microphones of the DUT are used directly for equalization or additional measurement microphones are placed near those of the DUT. + +In any case, the same microphone arrangement must be used for recording of background noise (see the following) and equalization of the system. Microphone characteristics, like frequency or phase response and directionality, shall be the same for both use cases. + +The number of microphones in the calibration set, i.e., those used for the matrix inversion in clause 8.2.6.3.5, must not exceed the number of loudspeakers in the set-up excluding potential subwoofer. In general, it is recommended that there be more full-range loudspeakers in the set-up than microphones in the calibration set, to improve the accuracy of the equalization. + +If it is not possible to use the DUT microphones directly, it is recommended that one measurement microphone be placed near to each DUT microphone. If special consideration of the directivity characteristic of the DUT microphone is necessary, at least two measurement microphones should be placed in the DUT microphone surroundings in order to more accurately simulate the sound field characteristics in that area. + +Of course, the microphone configuration described in clause 7.5.1 may also be used. + +#### 8.2.7.3 Background noise recordings and reference noise + +Since the arrangement of the microphones is not fixed, but differs between different applications and (possibly) different DUTs, no default background noise scenarios can be provided. Instead, individual background noise recordings must be made for every application and microphone arrangement. + +The equalization step of clause 8.2.6.3 requires a reference noise for the calculation of the inversion filters. One portion of the individual background noise recordings must be selected as reference noise. The selected reference noise should be 5 s to 10 s long and, obviously, representative of the individual background noise recordings, as the selected reference noise will be used to test the equalization, as well as for determining the maximum output volume of the loudspeakers. + +#### 8.2.7.4 Adaptation of the equalization method in clause 8.2.6.3 + +The method described in clause 8.2.6.3 specifies performance of the following steps: + +- 1) \*separate level adjustment for every loudspeaker; +- 2) \*system identification; +- 3) (\*) preprocessing of the impulse responses; +- 4) calculation of the inversion filters: + - a) \*inversion procedure, + - b) different microphones for different frequency bands, + - c) \*search for the optimum regularization factor; +- 5) \*first test of equalization and filter adjustment for inversion error compensation. + +However, this generalization only makes use of the steps marked with an asterisk. If a step needs modification in comparison to clause 8.2.6.3 it is described in the following. The matrix size (used to identify the system and to calculate the matrix inversion for equalization) depends on the number of loudspeakers and microphones chosen for the background noise simulation. The following adaptations are required for this variation: + +In clause 8.2.6.3.3, the impulse responses are pre-processed with a low-pass filter with a time-variant cut-off frequency to avoid a degradation of the quality of the inversion by high-frequency components of the tails of the impulse responses. This step reduces the accuracy of the equalization for higher frequencies. In the variant described in this clause, step 3 is considered as optional: As long as the tails of the impulse responses are short, the step can be skipped. + +In clause 8.2.6.3.4, different microphones were selected to calculate the inversion filters for each frequency band based on the corresponding wavelength. For the proposed generalized multi-point noise simulation method, the microphone distances are unknown and therefore this type of optimization (i.e., step 4) b in the list in the first paragraph] is not applied. All microphones are used across the whole frequency range to calculate the inversion filters. + +In clause 8.2.6.3.4, the optimum regularization factor is searched subject to a limited loudspeaker output level to avoid distortion of the loudspeaker signals. "A spectrum representative for the maximum level and spectral content" (see the Note to clause 8.2.6.3.5.2.3.1) can be calculated from the reference noise selected. For this variant (i.e., step 4) c in the list in the first paragraph], the level of the filtered reference noise radiated by each loudspeaker to each microphone position shall not exceed the level which a pink noise of 50 dB re. 400 pPa2/Hz would generate in the same FFT-bin. + +The extended methodology that aims to reduce the error at the so-called fine-tuning positions is not applicable, because the locations of the DUT microphones are well known and therefore those for calibration can be located close to them. + +In clause 8.2.6.3.7, a correction filter $D(f)$ is applied to compensate for the arithmetic uncertainties (e.g., the regularization factor) of the inversion. For this generalized multi-point noise simulation method, the microphone distances are unknown and therefore the cut-off frequency is reduced to 50 Hz. + +After finalizing the equalization, the following criteria shall be met: + +- 1) *Spectral reproduction accuracy.* The spectrum averaged over the individual microphones and averaged in the frequency bands for which they have been equalized should be flat to within $\pm 3$ dB from 50 Hz to 10 kHz and within $\pm 6$ dB from 10 kHz to 16 kHz measured in one-third octave. The average spectrum accuracy, averaged over all microphones shall be within $\pm 3$ dB from 50 Hz to 20 kHz. + +In scenarios where frequencies down to $f_{\min}$ shall be reproduced the difference for all microphones should be within $\pm 3$ dB and shall be within a range of +3 and $-6$ dB, measured in one-third octave between $f_{\min}$ and 50 Hz. The average spectrum accuracy, averaged over all microphones shall be within $\pm 3$ dB from $f_{\min}$ to 20 kHz. + +- 2) *Level accuracy.* The level of the reproduced sound field at each microphone shall be accurate to within $\pm 1$ dB. +- 3) *Complex coherence.* The magnitude of the complex coherence (normalized cross-correlation spectrum, see [b-Carter] and [b-Priestley]) between the original signal and the reproduced signals at the fine-tuning position averaged over the individual microphones and averaged in the frequency bands for which they have been equalized, measured in one-third octave, should be $> 0.9$ from 100 Hz to 1 kHz. The phase of the complex coherence should be accurate within $+45^\circ$ from 100 Hz to 1 kHz measured in one-third octave + +NOTE – Complex coherence: in the frequency range from 100 Hz to 1 kHz the phase of the complex coherence should be accurate within $\pm 10$ degrees and within $\pm 30$ degrees in the range from 1 kHz to 1.5 kHz, both measured in one-third octave. + +#### 8.2.7.5 Special use case – inside vehicle + +Clauses 8.2.7.1 to 8.2.7.4 specify a generalization of the multi-point method, while this clause provides instructions on how to apply the method to loudspeaker equalization inside a car. + +##### 8.2.7.5.1 Loudspeaker configuration + +As the generalized multi-point noise simulation method is compatible with the loudspeaker arrangement of clause 8.2.3 (four loudspeakers plus subwoofer), the same loudspeaker arrangement can be used. As previously described, the required number of loudspeakers depends on the number of microphones. Accordingly, the number of loudspeakers shall be increased if more microphones are used in the calibration set. + +##### 8.2.7.5.2 Microphone set-up + +The microphones need to be arranged individually for each car interior and for each application. + +##### 8.2.7.5.3 Equalization + +Since the reverberation time in vehicle interiors is negligible compared to treated offices, the low-pass filtering (step 3) is omitted. + +#### 8.2.7.6 Special use case – binaural equalization + +The set-up, as illustrated in Figure 18, can be used as an alternative to the background noise simulation system according to clause 8.2.1. + +![Diagram of a binaural noise field simulation setup. A central head model with two microphones (left and right) is shown. Four loudspeakers (1, 2, 3, 4) are positioned around it. Dashed lines represent impulse responses: h1,L, h1,R, h3,L, and h3,R. A subwoofer is connected to a playback unit (channels 1 to 4). A recording system is connected to the microphones and the playback unit. The label P.570(25) is in the bottom right corner.](c1df61cc3717e878a48e530218403403_img.jpg) + +Diagram of a binaural noise field simulation setup. A central head model with two microphones (left and right) is shown. Four loudspeakers (1, 2, 3, 4) are positioned around it. Dashed lines represent impulse responses: h1,L, h1,R, h3,L, and h3,R. A subwoofer is connected to a playback unit (channels 1 to 4). A recording system is connected to the microphones and the playback unit. The label P.570(25) is in the bottom right corner. + +**Figure 18 – Binaural noise field simulation** + +##### 8.2.7.6.1 Loudspeaker configuration + +The loudspeaker set-up shall be the same as in clause 8.2.1, i.e., four large or small loudspeakers shall be used. Table 5 contains the definition of the loudspeaker categories. An additional subwoofer shall be used if four small loudspeakers are used. See also Annex A for performance requirements of the loudspeakers. + +##### 8.2.7.6.2 Microphone set-up + +The microphone configuration used is HATS according to [ITU T P.58]. The two signals for the equalization procedure are recorded at the DRP. A sound field correction corresponding to the correction of the binaural recordings shall be used, e.g., for the playback of diffuse-field equalized binaural recordings, the noise field equalization and calibration procedure shall also be conducted with diffuse-field correction. + +##### 8.2.7.6.3 Equalization + +The equalization and calibration procedure are performed with 2 microphones (the left and the right ear in HATS) and 5 loudspeakers for set-ups with subwoofer or 4 in case no subwoofer is used, respectively. + +### 8.2.8 Loudspeaker equalization using the generalized multi-point method in a reference set-up + +An overview of the benefits and limitations of the method described in this clause can be found in clause 6. + +The method described in clause 8.2.6 specifies performance of the following: + +- 1) \*separate level adjustment for every loudspeaker; +- 2) \*system identification; +- 3) preprocessing of the impulse responses; +- 4) calculation of the inversion filters: + - a) \*inversion procedure, + - b) different microphones for different frequency bands, + +- c) \*search for the optimum regularization factor; +- 5) \*first test of equalization and filter adjustment for inversion error compensation. + +However, the simple method makes use of the procedure marked with an asterisk. Details of each step are given in clauses 8.2.8.1 to 8.2.8.5. + +#### 8.2.8.1 Step 1 – Separate level adjustment for every loudspeaker + +The matrix inversion method equalizes the loudspeaker room responses, and the level and frequency response differences among loudspeakers are compensated for. Nevertheless, it may be more reliable if the loudspeakers produce more or less the same sound pressure level with a given input signal. For that, a sine-sweep signal having a maximum value of $-0.17$ dB relative to full scale in the waveform audio (WAV) file is played through each loudspeaker, and the average sound pressure level across eight microphones is calculated. The loudspeaker gain is adjusted until the average sound pressure level reaches $94$ dB SPL. The sound pressure level produced by each loudspeaker can further be increased if the reference (recorded) SPL is higher than the set-up is calibrated for. + +#### 8.2.8.2 Step 2 – System identification + +The room impulse response (RIR) functions between each loudspeaker and the microphones are measured. The level difference between the peak and noise part of impulse response functions is recommended to be higher than $60$ dB to achieve impulse response functions with good SNR. Figure 19 shows an example of an RIR. The SNR, i.e., peak-to-noise-floor ratio, of approximately $90$ dB is observed from the measured RIR. + +![Figure 19: A plot of a room impulse response function (RIR) in decibels. The y-axis is labeled 'Magnitude [dB re Pa/V]' and ranges from 0 to -200 in increments of 20. The x-axis is labeled 'Time [s]' and ranges from 0 to 3 in increments of 0.5. The plot shows a blue signal that starts at 0 dB at 0 seconds, decays rapidly to about -100 dB by 0.5 seconds, and then continues with a noisy decay, stabilizing around -140 dB from 1.5 seconds onwards. The signal exhibits significant high-frequency noise throughout the decay phase.](6a842e59ecf31b0d6f6721ac4e76f160_img.jpg) + +Figure 19: A plot of a room impulse response function (RIR) in decibels. The y-axis is labeled 'Magnitude [dB re Pa/V]' and ranges from 0 to -200 in increments of 20. The x-axis is labeled 'Time [s]' and ranges from 0 to 3 in increments of 0.5. The plot shows a blue signal that starts at 0 dB at 0 seconds, decays rapidly to about -100 dB by 0.5 seconds, and then continues with a noisy decay, stabilizing around -140 dB from 1.5 seconds onwards. The signal exhibits significant high-frequency noise throughout the decay phase. + +P.570(25) + +**Figure 19 – An example of a room impulse response function RIR displayed in decibels** + +For RIR measurement, each loudspeaker is excited by a sine sweep signal. Compared to other types of excitation signals, swept sine is able to separate non-linear distortion created by the loudspeakers from the linear part of the impulse response function [b-Müller]. For the sake of convenience, the same notation as used in clause 8.2.6 is used in the following. The deconvolution of input and output signals can be done by dividing the microphone signal $Y(f)$ by the loudspeaker excitation signal $S(f)$ in the frequency domain and subsequently taking the inverse FFT of the division. $H(f)$ denotes the measured room frequency response (RFR) function and $h(t)$ is the corresponding RIR. The frequency response functions are derived in the frequency range between $20$ Hz to $20$ kHz. + +#### 8.2.8.3 Step 4, a – Inversion procedure + +A detailed description of the matrix inversion method used by the multi-point method is provided in [b-Kirkeby 1996]. When playing a set of signals through loudspeakers, the resulting signals at microphone positions can be formulated as + +$$\begin{pmatrix} w_1 \\ \vdots \\ w_r \end{pmatrix} = \begin{pmatrix} H_{1,1} & \cdots & H_{1,l} \\ \vdots & \ddots & \vdots \\ H_{r,1} & \cdots & H_{r,l} \end{pmatrix} \begin{pmatrix} v_1 \\ \vdots \\ v_l \end{pmatrix} \quad (8-17)$$ + +where $v$ is the loudspeaker signal, $w$ is the reproduced signal measured at microphones, and $H_{r,l}$ is the RFR between the $l$ th loudspeaker and the $r$ th microphone. The pseudo inverse of the RFR matrix $\mathbf{H}$ can be calculated by making use of Thikonov regularization. + +$$\mathbf{C} = (\mathbf{H}^* \mathbf{H} + \beta \mathbf{I})^{-1} \mathbf{H}^* \quad (8-18)$$ + +Where $\mathbf{C}$ is the inverse matrix of the RFR matrix $\mathbf{H}$ , $\mathbf{I}$ is the identity matrix, and $\beta$ is a regularization parameter. The regularization parameter, i.e., $\beta$ , is introduced to make the matrix inversion more stable. The threshold for the regularization parameter in decibels, i.e., $20\log_{10}(1/\beta)$ , quantifies the amount of regularization. The matrix inversion procedure is applied for each frequency line separately, and the loudspeaker signals are calculated by Equation 8-19. + +$$\begin{pmatrix} v_1 \\ \vdots \\ v_r \end{pmatrix} = \begin{pmatrix} C_{1,1} & \cdots & C_{1,r} \\ \vdots & \ddots & \vdots \\ C_{l,1} & \cdots & C_{l,r} \end{pmatrix} \begin{pmatrix} u_1 \\ \vdots \\ u_r \end{pmatrix} \quad (8-19)$$ + +where $u$ is the reference signal recorded with an array of microphones in a reference sound field. + +Instead of performing the matrix multiplication in the frequency domain, the equalization filter $c(t)$ is calculated by taking the inverse FFT for each element of $\mathbf{C}$ . Subsequently, the convolution between the reference signals at the microphones and the equalization filters is done in the time domain, and the result of the convolution added to derive the resulting signal for each loudspeaker. + +#### 8.2.8.4 Step 4, c – Search for the optimum regularization factor + +An optimal regularization parameter can be found for a loudspeaker set-up before running actual device tests. It may be achieved by going through a number of regularization thresholds and checking the magnitude error, as well as coherence. The range of thresholds shall be between $-20$ dB (10) to $40$ dB (0.01). A step size of 5 or 10 dB can be selected. The magnitude error and coherence function can be compared across different thresholds in order to select an optimal threshold to be used via visual inspection. + +Audible artefacts may be observed when selecting higher thresholds, i.e., lower regularization parameter, even if the resulted errors are relatively small. Therefore, it is recommended that objective results be checked first, i.e., magnitude error and coherence, and for the selected threshold, the presence of audible artefacts should be examined by reproducing or listening to typical background noise under investigation. + +#### 8.2.8.5 Step 5 – First test of equalization and filter adjustment for inversion error compensation + +The matrix inversion method gives rise to very accurate sound reproduction at target positions where the sound field is optimized. However, the result at validation positions deviates more from the reference sound field, and the amount of deviation is dependent on how far the validation positions are from the target positions as well as the frequency. At low frequencies, the distance between the target and validation positions has less effect compared to high frequencies due to the wavelength corresponding to the frequency of interest. For the validation positions, far from the target positions, it is necessary to compensate for the average magnitude error across all microphones. If a few + +validation positions are far from the target positions and the rest are close to each other, the maximum and the minimum of magnitude error over frequency may be averaged in order to bring the magnitude error at far distances within the $\pm 3$ dB range. + +The magnitude error compensation is done once for a reference stimulus using incoherent pink noise, and the same compensation filter is used for other types of stimuli. The average error function across all microphones in one-twelfth-octave bands are linearly interpolated in frequency, and the inverse FFT of the interpolated response is calculated to derive the corresponding impulse response function. The resulting impulse response function is transformed to a minimum phase counterpart in order not to introduce further delays and non-causality in the processing. + +### **8.2.9 Loudspeaker equalization using the over-determined multi-point method in a reference set-up** + +An overview of the benefits and limitations of the method described in this clause can be found in clause 6. + +The optimal output of the matrix inversion method may be achieved by positioning the target microphones close to the built-in microphones of a device. When the microphone positions are further away from the device, the coherence between the reference and reproduced recordings may drop significantly at high frequencies, even though the magnitude error can be within a reasonable range. Decrease in coherence eventually means that perception in the reference sound field may not be similar to perception in that reproduced. It has been shown that the over-determined multi-point method, i.e., the number of loudspeakers higher than the number of calibration microphones, can derive the loudspeaker signals in such a way that the matrix inversion becomes more stable. The reproduction error in the validation microphones seems to decrease when using a higher number of loudspeakers than the number of microphones. For example, even if there are eight loudspeakers installed in the set-up, a couple of target microphone positions can be selected for the optimal sound reproduction. + +The procedure for the over-determined multi-point method is exactly the same as that for the generalized multi-point method in clause 8.2.8. + +## **8.3 Reproducing artificial noise fields in a simulation set-up** + +Pre-recorded noises to be used for various background noise scenarios can be found in the ETSI background noise database which can be accessed via ETSI. For more information, see [ETSI TS 103 224] clause 8. These recordings can be used for validation of various noise reproduction set-ups at various test laboratories. + +# **9 Techniques for synthesizing noise fields** + +While other techniques described in this Recommendation have the goal of reproducing sound fields recorded at a specific time and place, this clause describes how several sources can be superimposed at playback time to create an artificial sound field. + +Specifically, it is the intention of this clause to address the following situations. + +- A number of non-coherent sources are used to approximate a diffuse sound field of stationary noise (typically pink noise or [b-Hoth] noise signals). +- A number of non-coherent sources is used to approximate a sound field of non-stationary properties, e.g., several loudspeakers play different babble recordings comprising a number of talkers each, simulating being at a restaurant surrounded by a number of guests. +- A special case of the above is a mono recording that is played in all loudspeakers, but with significant delay (several seconds) between the reproduction channels. This decorrelates the signals so that interference effects are minimized (except for periodic signals). + +- One of the sound field reproduction methods described in this Recommendation (e.g., mono or binaural recordings reproduced on four loudspeakers) is altered to include significant delays (several seconds) between the reproduction channels. This decorrelates the signals so that interference effects are minimized (except for periodic signals). +- Additional sources are added to one of the sound field reproduction methods described in this Recommendation, e.g., a competing talker at a certain incidence angle is added on top of a background noise. + +Advantages of building up an artificial sound field from several sources include: + +- the different sources can be adjusted in level, time and space depending on the needs for the specific test, without re-recording the material; +- anechoic recordings reproduced *in situ* (thus including the target acoustics) have a potential to produce natural field properties; +- the signals are easy to capture (e.g., a mono recorder); +- there are no interference effects between channels and thus the robustness to position is high; +- the method is free from some artificial effects in the time domain (echoes, smearing of transients), that might occur for methods using small delays at playback time for partly decorrelating signals; +- the equalization procedure is straightforward. + +Considerations include: + +- although the final sound field can be representative of a real situation, it is after all artificially created; +- compared to advanced methods that recreate actual field properties (ambisonics etc.), the method described here is limited in the number of simultaneous incidence angles that are represented (to the number of loudspeakers used); +- for non-anechoic reproduction conditions, the reproducibility will be affected by the room characteristics; +- for non-anechoic mono recording conditions, the reverberation is also captured in mono; +- complex distributed sound fields are difficult to reproduce with this set-up. + +## 9.1 Equalization procedure + +Each loudspeaker is equalized using the single microphone technique outlined in clause 8.2.2, except that the final step (clause 8.2.2.5) is omitted. + +## 9.2 Recommended set-ups + +For consistency with other methods in this Recommendation, the loudspeaker angles and distances presented in Figures 4 and 6 and Annex A are recommended. + +# 10 Techniques for assessing artificial noise fields + +## 10.1 Assessment based on spectral content + +This is currently not covered in this Recommendation. + +## 10.2 Assessment based on spatial coherence + +This is currently not covered in this Recommendation. + +## **10.3 Assessment of uniformity** + +This is currently not covered in this Recommendation. + +# **11 Documentation of measurements in artificial noise fields** + +To document an artificial sound field that has been established by one of the methods described in this Recommendation, the following information shall be provided: + +- the loudspeaker configuration that has been used to create the sound field shall be documented, preferably with direct reference to Annex A; +- the equalization method that has been used for compensation of the loudspeakers shall be documented and specifically state whether a single measurement microphone, a binaural HATS or a microphone array have been used; +- the environment in which the loudspeaker configuration was located, e.g., an anechoic chamber, a treated office, an ordinary office or a vehicle. + +To further quantify the quality of the artificial sound field, the following information shall be provided: + +- the type of signal, as well as the level that was used for the equalization; +- the accuracy of the equalization that has been established. + +For further characterization of the artificial sound field additional parameters could be provided. Such parameters are: + +- the overall level as well as the spectrum of the ambient noise, when no signals are applied to the loudspeaker configuration; +- the reverberation time of the environment in which the loudspeaker configuration is located should be provided. + +# Annex A + +## Loudspeaker performance + +(This annex forms an integral part of this Recommendation.) + +When measured under free-field conditions and using a measurement microphone placed 1 m in front of the loudspeaker, the following specifications must be fulfilled: + +- a sound pressure level at 300 Hz of 100 dBSPL must be supported. +- the frequency response of the loudspeaker must be within $\pm 10$ dB in the frequency range 100 Hz to 20 kHz. +- the total harmonic distortion of the second and third harmonic must be less than 3% when measured at a constant level of 90 dBSPL over the frequency range 100 Hz to 20 kHz. +- the level measure at $\pm 10^\circ$ off-axis in the horizontal plan must be lower than 3 dB compared to the on-axis level, when measured in the frequency range 100 Hz to 20 kHz. + +When measured under free-field conditions and using a measurement microphone placed 1 m in front of the subwoofer, the following specifications must be fulfilled: + +- a sound pressure level at 50 Hz of 110 dBSPL must be supported; +- the frequency response of the loudspeaker must be within $\pm 0$ dB in the frequency range 50 Hz to 300 Hz; +- the total harmonic distortion of the second and third harmonic must be less than 10% when measured at a constant level of 90 dBSPL over the frequency range 50 Hz to 300 Hz; +- it is implicit that the directivity of the subwoofer will be acceptable for the application. + +# Appendix I + +## Requirements for test environments + +(This appendix does not form an integral part of this Recommendation.) + +The requirements that should be considered for a typical reference set-up, such as an anechoic chamber, a listening room, an acoustically treated office and a reverberant room, are summarized as follows. + +- For an anechoic chamber, no additional requirements should be considered. However, an anechoic chamber should be professionally designed, implying the use of very thick absorbing materials. Furthermore, it is important to realize that a "quiet" room is not necessarily anechoic. +- For a listening room, it will in general be difficult to outline what should be considered. For more specific information see [IEC TR 60268-13], which gives recommendations for establishing, conducting and evaluating listening tests on loudspeakers, as well as general guidelines. With regards to determining the reverberation time, [IEC TR 60268-13] calls for one-third-octave measurements. +- For an acoustically treated office, the following requirements should be considered: + - The floor area should be at least $3.5 \text{ m} \times 3.5 \text{ m}$ . + - The room height should be at least 2.5 m. + - No dimension should be an integral multiple of another. + - The floor should be carpeted. + - The ceiling should have acoustical damping. + - The reverberation time should be less than 0.7 s but higher than 0.2 s in the frequency range between 100 Hz and 8 kHz. For the multi-point reproduction technique, the $C_{80}$ criterion (see clause 8.2.6.2) should be applied. + - Strong reflections and standing waves caused by hard surfaces in the room should be avoided. One or two walls should be treated, e.g., with appropriate acoustical damping material. + - Choose loudspeaker and diffuse-field test point (DFTP) carefully and verify the position. The DFTP is at the geometric centre of the loudspeakers, but DFTP should be slightly off-centre, in all dimensions. +- For a reverberant room, no additional requirements should be considered. However, a "reverberant room" does imply that the walls and ceiling are of solid construction. Furthermore, replacing absorbent ceiling panels with 12 mm gypsum board is in many instances insufficient to render a room reverberant. + +Additionally, it is recommended that: + +- the operator be protected from test sounds that can be up to 90 dB SPL or even higher; +- a larger space be used than $3.5 \text{ m} \times 3.5 \text{ m}$ for ease of operator access; +- placing the loudspeakers in the corner of the room be avoided in order to reduce the presence of resonances; +- placing the loudspeakers in a perfect square be avoided; +- positioning the HATS or the measurement microphone in the exact centre of a square be avoided; + +- the influence of external noise be reduced by ensuring that the level is less than 30 dBSPL(A): + - separate a test room from operator areas, so that noise from personal computers (PCs) and other equipment, as well as the operator, can be controlled, + - separate an entire test suite from other workers as test sounds up to 90 dBSPL can occur and cause annoyance, + - minimize air-borne noise from ventilation by increasing the length of ventilation pipes and by lining them, + - minimize structure-borne noise from elevators, traffic, etc., + - seal openings under doors, cable penetrations, etc., + - sound isolating materials, + - locate a test room on solid ground, if possible, + - install floating floors, + - check for noisy devices that operate only occasionally. + +For operator convenience, the following are recommended: + +- provision of a clear view of the test space; + - window between control area and test room, + - camera and display screen; +- cable routing that is sensible: + - clear labels (tags or colour-code), + - position cables so tripping is prevented; +- appropriate lighting: + - sufficient light for operator to access equipment, + - low heat generation to ensure consistent measurement conditions; +- an intercom for communication between operator and the test room; +- an outline of a power on/off sequence for complete and reliable switching on/off equipment. + +# Appendix II + +## Loudspeaker configurations and decoding of loudspeaker signals + +(This appendix does not form an integral part of this Recommendation.) + +For the reference test environment category specified in Annex A, some compensation could be necessary to provide better unity between tests performed in, for example, a large anechoic chamber and a listening room conforming to the acoustic requirements of [ITU-R BS.1116-3]. For example, the absolute distance of each loudspeaker from the centre of an orthogonal array may be dependent on or restricted by the dimensions of the room. + +A simple way of uniquely specifying the orthogonal array without having to describe the required or implemented absolute position of each loudspeaker in relation to the test environment is to specify the Cartesian coordinates $(x, y, z)$ of each source about a unit scale. Examples of this definition for a number of arrays are shown in Figures II.1 to II.4. This positional specification can also be used to report additional loudspeakers used to reproduce signals independent of the simulated noise field (e.g., competing speech). + +An assumption here is that each loudspeaker source can provide the full range of frequencies required to accurately simulate a noise field, e.g., including enough low-frequency energy. This may not be feasible or practical to implement and so may require additional low frequency sources, i.e., so-called subwoofer/s. In this case, it may be possible to either have separate low-frequency extension speakers dedicated to each full-range speaker or, more likely, one or more subwoofers that are assigned the low frequency content of all source signals, depending on the format of the simulated noise recording. For practical purposes, any recommended specification of how the subwoofer is to be placed would be based primarily on the test environment and the system calibration procedure. + +![Diagram of a horizontal plane loudspeaker array in a square configuration. It shows a square with dashed lines and a circle inscribed within it. Four loudspeakers, labeled ch. A, ch. B, ch. C, and ch. D, are positioned at the corners of the square. A central oval represents the listener's position. The x and y axes are shown, with tick marks at 1 and -1. The coordinates for each channel are given in the table below.](20938559c8eb02f499e31059759fea6d_img.jpg) + +SQUARE + +| Channel | Co-ordinates $(x, y, z)$ | +|---------|--------------------------| +| A | 0.7071, 0.7071, 0.0000 | +| B | 0.7071, -0.7071, 0.0000 | +| C | -0.7071, -0.7071, 0.0000 | +| D | -0.7071, 0.7071, 0.0000 | + +Diagram of a horizontal plane loudspeaker array in a square configuration. It shows a square with dashed lines and a circle inscribed within it. Four loudspeakers, labeled ch. A, ch. B, ch. C, and ch. D, are positioned at the corners of the square. A central oval represents the listener's position. The x and y axes are shown, with tick marks at 1 and -1. The coordinates for each channel are given in the table below. + +P.570(18)\_FII.1 + +Figure II.1 – Horizontal plane loudspeaker array in a square configuration + +![Diagram of an octagonal loudspeaker array in the horizontal plane. The array is centered at the origin of an x-y-z coordinate system. Eight channels (ch. A through ch. H) are arranged in a circle. Channels A, C, E, and G are positioned on the axes at a distance of 1.0 from the center. Channels B, D, F, and H are positioned at 45-degree angles from the axes, also at a distance of 1.0 from the center. Each channel is represented by a speaker icon with an arrow indicating its orientation. A dashed circle connects all eight channels. A dashed rectangle encloses the outer four channels (A, C, E, G).](a5401b088b668d0554e8e96ea558121c_img.jpg) + +**OCTAGON 2** + +| Channel | Co-ordinates (x, y, z) | +|---------|--------------------------| +| A | 1.0000, 0.0000, 0.0000 | +| B | 0.7071, 0.7071, 0.0000 | +| C | 0.0000, 1.0000, 0.0000 | +| D | -0.7071, 0.7071, 0.0000 | +| E | -1.0000, 0.0000, 0.0000 | +| F | -0.7071, -0.7071, 0.0000 | +| G | 0.0000, -1.0000, 0.0000 | +| H | 0.7071, -0.7071, 0.0000 | + +Diagram of an octagonal loudspeaker array in the horizontal plane. The array is centered at the origin of an x-y-z coordinate system. Eight channels (ch. A through ch. H) are arranged in a circle. Channels A, C, E, and G are positioned on the axes at a distance of 1.0 from the center. Channels B, D, F, and H are positioned at 45-degree angles from the axes, also at a distance of 1.0 from the center. Each channel is represented by a speaker icon with an arrow indicating its orientation. A dashed circle connects all eight channels. A dashed rectangle encloses the outer four channels (A, C, E, G). + +P.570(18)\_FII.2 + +**Figure II.2 – Horizontal plane loudspeaker array in an octagonal configuration** + +![Three diagrams showing the horizontal and vertical plane loudspeaker array in a tetrahedral configuration. The top-left diagram shows the x-y axis with channels A, B, C, and D. The top-right diagram shows the x-z axis with channels A, B (left), C (right), and D. The bottom diagram shows the y-z axis with channels A (front), B (back), C (back), and D. Below the diagrams is a table with 4 columns: Channel, Co-ordinates (x, y, z).](a7f9e5d1862146a136ba5ea891f15f17_img.jpg) + +**TETRAHEDRON** +(x-y axis) + +**TETRAHEDRON** +(x-z axis) + +**TETRAHEDRON** +(y-z axis) + +| Channel | Co-ordinates (x, y, z) | +|---------|---------------------------| +| A | 0.9428, 0.0000, -0.3334 | +| B | -0.4714, 0.8164, -0.3334 | +| C | -0.4714, -0.8164, -0.3334 | +| D | 0.0000, 0.0000, 1.0000 | + +Three diagrams showing the horizontal and vertical plane loudspeaker array in a tetrahedral configuration. The top-left diagram shows the x-y axis with channels A, B, C, and D. The top-right diagram shows the x-z axis with channels A, B (left), C (right), and D. The bottom diagram shows the y-z axis with channels A (front), B (back), C (back), and D. Below the diagrams is a table with 4 columns: Channel, Co-ordinates (x, y, z). + +P.570(18)\_FII.3 + +**Figure II.3 – Horizontal and vertical plane loudspeaker array in a tetrahedral configuration** + +![Three diagrams showing a cube with 8 channels (A-H) in different orientations: x-y axis, x-z axis, and y-z axis. Below the diagrams is a table with 2 columns: Channel and Co-ordinates (x,y,z).](1dea2e515f706e8044ae79339efaa36c_img.jpg) + +The figure displays three views of a cube with 8 channels (A-H) positioned at its corners. Each view shows the cube's orientation and the corresponding coordinate axes. + +- CUBE (x-y axis):** Shows the cube in the x-y plane. The y-axis is vertical and the x-axis is horizontal. Channels are labeled as follows: ch. H (above), ch. D (below) at the top-left; ch. E (above), ch. A (below) at the top-right; ch. G (above), ch. C (below) at the bottom-left; ch. F (above), ch. B (below) at the bottom-right. +- CUBE (x-z axis):** Shows the cube in the x-z plane. The z-axis is vertical and the x-axis is horizontal. Channels are labeled as follows: ch. G (right), ch. H (left) at the top-left; ch. F (right), ch. E (left) at the top-right; ch. C (right), ch. D (left) at the bottom-left; ch. B (right), ch. A (left) at the bottom-right. +- CUBE (y-z axis):** Shows the cube in the y-z plane. The z-axis is vertical and the y-axis is horizontal. Channels are labeled as follows: ch. F (front), ch. G (back) at the top-left; ch. E (front), ch. H (back) at the top-right; ch. B (front), ch. C (back) at the bottom-left; ch. A (front), ch. D (back) at the bottom-right. + +| Channel | Co-ordinates (x,y,z) | +|---------|---------------------------| +| A | 0.5774, 0.5774, -0.5774 | +| B | 0.5774, -0.5774, -0.5774 | +| C | -0.5774, -0.5774, -0.5774 | +| D | -0.5774, 0.5774, -0.5774 | +| E | 0.5774, 0.5774, 0.5774 | +| F | 0.5774, -0.5774, 0.5774 | +| G | -0.5774, -0.5774, 0.5774 | +| H | -0.5774, 0.5774, 0.5774 | + +Three diagrams showing a cube with 8 channels (A-H) in different orientations: x-y axis, x-z axis, and y-z axis. Below the diagrams is a table with 2 columns: Channel and Co-ordinates (x,y,z). + +P.570(18)\_FII.4 + +**Figure II.4 – Horizontal and vertical plane loudspeaker array in a cubic configuration** + +The B-format signal can be decoded to represent any number of microphone responses or polar patterns from any direction within the approximated unit sphere (or unit circle if the Z-component is omitted) making the B-format very efficient and flexible for simulating coincident mono, stereo or multichannel recording techniques. For sound field reproduction over loudspeakers, it also allows wave front reconstruction, either pantophonic or periphonic, through decoding to orthogonal arrays of loudspeakers each reconstructing the wave field associated with its individual angle of incidence to the centre of the array. + +Several different "flavours" of loudspeaker decoding rules for first-order ambisonic signals have been proposed over the years, some taking advantage of psychoacoustic principles to improve spatial or immersive perception. However, the simplest form of decoding from B-format for an orthogonal loudspeaker array I which each loudspeaker is at the same distance from the origin of the array at an azimuth angle $\alpha$ and an elevation angle $\beta$ would describe the signal sent to each loudspeaker as: + +$$S_{\text{LS}(\alpha,\beta)} = \sqrt{2}W + X(\cos \alpha \times \cos \beta) + Y(\sin \alpha \times \cos \beta) + Z(\sin \beta)$$ + +where the $\sqrt{2}W$ term accounts for the weighting in the B-format encoding noted above. Power normalization dependent on the number and configuration of loudspeakers can be achieved by using: + +$$S_{\text{LS}(\alpha,\beta)} = \frac{\sqrt{2}}{N}W + \frac{d}{N}X(\cos \alpha \times \cos \beta) + \frac{d}{N}Y(\sin \alpha \times \cos \beta) + \frac{d}{N}Z(\sin \beta)$$ + +where $N$ is the number of speakers in the array and $d$ is the number of axes used in the array (e.g., for a square, pantophonic array $d = 2$ but for a cubic, periphonic array $d = 3$ ). + +As examples, Table II.1 shows the unit gains, with signs representing the phase that should be applied to the B-format channels before mixing to create the D-format signal for each loudspeaker in the loudspeaker arrays shown in Annex A. These gains have been calculated using $S_{\text{LS}(\alpha,\beta)}$ with the relevant values of $N$ , $d$ , $\alpha$ and $\beta$ also provided in Table II.1. + +$$S_{\text{LS}(\alpha,\beta)} = \frac{\sqrt{2}}{N}W + \frac{d}{N}X(\cos \alpha \times \cos \beta) + \frac{d}{N}Y(\sin \alpha \times \cos \beta) + \frac{d}{N}Z(\sin \beta)$$ + +**Table II.1 – Multiplication gains for mixing B-format signals to D-format** + +| Array | $N$ | $d$ | LS channel | $\alpha$ | $\beta$ | $W_{\text{gain}}$ | $X_{\text{gain}}$ | $Y_{\text{gain}}$ | $Z_{\text{gain}}$ | +|---------|-----|-----|------------|----------|---------|-------------------|-------------------|-------------------|-------------------| +| Square | 4 | 2 | A | 45 | 0 | 0.3536 | 0.3536 | 0.3536 | 0 | +| | 4 | 2 | B | 315 | 0 | 0.3536 | 0.3536 | -0.3536 | 0 | +| | 4 | 2 | C | 225 | 0 | 0.3536 | -0.3536 | -0.3536 | 0 | +| | 4 | 2 | D | 135 | 0 | 0.3536 | -0.3536 | 0.3536 | 0 | +| Hexagon | 6 | 2 | A | 0 | 0 | 0.2357 | 0.3333 | 0 | 0 | +| | 6 | 2 | B | 60 | 0 | 0.2357 | 0.1667 | 0.2887 | 0 | +| | 6 | 2 | C | 120 | 0 | 0.2357 | -0.1667 | 0.2887 | 0 | +| | 6 | 2 | D | 180 | 0 | 0.2357 | -0.3333 | 0 | 0 | +| | 6 | 2 | E | 240 | 0 | 0.2357 | -0.1667 | -0.2887 | 0 | +| | 6 | 2 | F | 300 | 0 | 0.2357 | 0.1667 | -0.2887 | 0 | +| Octagon | 8 | 2 | A | 0 | 0 | 0.1768 | 0.25 | 0 | 0 | +| | 8 | 2 | B | 45 | 0 | 0.1768 | 0.1768 | 0.1768 | 0 | +| | 8 | 2 | C | 90 | 0 | 0.1768 | 0 | 0.25 | 0 | +| | 8 | 2 | D | 135 | 0 | 0.1768 | -0.1768 | 0.1768 | 0 | +| | 8 | 2 | E | 180 | 0 | 0.1768 | -0.25 | 0 | 0 | +| | 8 | 2 | F | 225 | 0 | 0.1768 | -0.1768 | -0.1768 | 0 | +| | 8 | 2 | G | 270 | 0 | 0.1768 | 0 | -0.25 | 0 | +| | 8 | 2 | H | 315 | 0 | 0.1768 | 0.1768 | -0.1768 | 0 | +| Cube | 8 | 3 | A | 45 | -35.26 | 0.1768 | 0.2165 | 0.2165 | -0.2165 | +| | 8 | 3 | B | 315 | -35.26 | 0.1768 | 0.2165 | -0.2165 | -0.2165 | +| | 8 | 3 | C | 225 | -35.26 | 0.1768 | -0.2165 | -0.2165 | -0.2165 | +| | 8 | 3 | D | 135 | -35.26 | 0.1768 | -0.2165 | 0.2165 | -0.2165 | + +**Table II.1 – Multiplication gains for mixing B-format signals to D-format** + +| Array | $N$ | $d$ | LS channel | $\alpha$ | $\beta$ | $W_{\text{gain}}$ | $X_{\text{gain}}$ | $Y_{\text{gain}}$ | $Z_{\text{gain}}$ | +|--------------|-----|-----|-------------------|----------|---------|-------------------|-------------------|-------------------|-------------------| +| | 8 | 3 | E | 45 | 35.26 | 0.1768 | 0.2165 | 0.2165 | 0.2165 | +| | 8 | 3 | F | 315 | 35.26 | 0.1768 | 0.2165 | −0.2165 | 0.2165 | +| | 8 | 3 | G | 225 | 35.26 | 0.1768 | −0.2165 | −0.2165 | 0.2165 | +| | 8 | 3 | H | 135 | 35.26 | 0.1768 | −0.2165 | 0.2165 | 0.2165 | + +# Appendix III + +## Spatial sound field stability + +(This appendix does not form an integral part of this Recommendation.) + +### III.1 Reproduction of monophonic signals in anechoic conditions + +#### III.1.1 Introduction + +This appendix documents the spatial stability of the reproduced noise field that can be achieved using the microphone method as described in clause 8.2.2. All measurements were performed in an anechoic room. + +#### III.1.2 Description of the set-up + +The loudspeakers were positioned in a square with the microphone at the centre (and in the same plane) as shown schematically in Figure III.1 (around 2 m between the loudspeakers and the microphone); Figure III.2 is a photograph of the actual set-up. Particular attention was paid to not creating a perfect square and to not positioning the microphone at its exact centre. No subwoofer was used in the set-up, thus results are valid from 120 Hz to 20 kHz. + +![Schematic diagram of the set-up for loudspeaker equalization using the microphone method. A Generator with five outputs (a, b, c, d, e) is connected to five loudspeakers. Each output passes through a delay (τ) and an amplifier (triangle). The loudspeakers are arranged in a square: 'Left front', 'Right front', 'Left back', and 'Right back' at the corners, and one unlabeled speaker on the left side. A measurement microphone is at the center of the square. Dashed lines from the center to each corner speaker are labeled '~2m'. The microphone is connected to an Analyser. The diagram is labeled P.570(18)_FIII.1.](8bd7adf738ea4b7e65e0d6f67e122d55_img.jpg) + +The diagram illustrates the experimental setup for loudspeaker equalization. On the left, a 'Generator' unit has five outputs labeled 'a', 'b', 'c', 'd', and 'e'. Each output is connected to a processing block consisting of a delay element (τ) and an amplifier (represented by a triangle). The amplified signals are sent to five loudspeakers. Four of these speakers are positioned at the corners of a square, labeled 'Left front', 'Right front', 'Left back', and 'Right back'. A fifth speaker is positioned on the left side of the square. A measurement microphone is placed at the center of the square. Dashed lines connect the center microphone to each of the four corner speakers, with each line labeled '~2m', indicating the distance. The microphone is connected to an 'Analyser' unit. The diagram is identified by the label 'P.570(18)\_FIII.1' in the bottom right corner. + +Schematic diagram of the set-up for loudspeaker equalization using the microphone method. A Generator with five outputs (a, b, c, d, e) is connected to five loudspeakers. Each output passes through a delay (τ) and an amplifier (triangle). The loudspeakers are arranged in a square: 'Left front', 'Right front', 'Left back', and 'Right back' at the corners, and one unlabeled speaker on the left side. A measurement microphone is at the center of the square. Dashed lines from the center to each corner speaker are labeled '~2m'. The microphone is connected to an Analyser. The diagram is labeled P.570(18)\_FIII.1. + +Figure III.1 – The set-up for loudspeaker equalization using the microphone method + +![A photograph of an anechoic room with walls covered in large, tan-colored acoustic foam wedges. In the center of the room, a microphone is mounted on a stand on a small wooden platform. Four speakers are positioned around the microphone, two on the left and two on the right, each on its own stand. Cables are visible on the floor connecting the equipment.](bc6b860289610cf05b09018ab47c90aa_img.jpg) + +A photograph of an anechoic room with walls covered in large, tan-colored acoustic foam wedges. In the center of the room, a microphone is mounted on a stand on a small wooden platform. Four speakers are positioned around the microphone, two on the left and two on the right, each on its own stand. Cables are visible on the floor connecting the equipment. + +**Figure III.2 – Actual set-up using the microphone method in an anechoic room** + +In order to determine the spatial stability of the reproduced noise field, several measurements around the equalization centre (original position of the microphone for the equalization) were taken. These measurements were done according to Figure III.3. + +![A diagram showing a grid of measurement points (blue circles) for a reproduced noise field. The grid is centered at (0,0) and extends from -50 to 50 mm in both X and Y directions. Four speakers are shown at the corners of the grid: Front left, Front right, Rear left, and Rear right. The central microphone is also shown.](7b96fce298a23fd76a01ff6c176c1059_img.jpg) + +The diagram illustrates a measurement grid for a reproduced noise field. The grid is defined by X (mm) and Y (mm) axes, both ranging from -50 to 50 mm in increments of 10 mm. Blue circles represent the measurement points, arranged in a 11x11 grid. A central microphone is positioned at the origin (0,0). Four speakers are shown at the corners of the grid: Front left, Front right, Rear left, and Rear right. The label 'P.570(25)' is present near the bottom right speaker. + +A diagram showing a grid of measurement points (blue circles) for a reproduced noise field. The grid is centered at (0,0) and extends from -50 to 50 mm in both X and Y directions. Four speakers are shown at the corners of the grid: Front left, Front right, Rear left, and Rear right. The central microphone is also shown. + +**Figure III.3 – Measurement points (blue circles) for the reproduced noise field** + +The stability of the sound field in a square of 100 mm × 100 mm around the equalization centre ( $X = 0$ mm; $Y = 0$ mm) was measured. Note that the equalization centre corresponds to the original position at which the equalization was realized according to this Recommendation. Measurements were taken every 10 mm on both the $X$ - and $Y$ -axis (blue circles in Figure III.3) leading to 121 measurements for the full plan. + +The test signal that was used for all measurements is pink noise (high-pass filtered as required by this Recommendation) and its level was adjusted in order to get around 75 dBSPL (A) at the equalization centre. The reference used for the calculation of frequency responses is the signal measured at the equalization centre. This signal is used as the reference for the calculation of the frequency responses for all points of the grid defined in Figure III.3. + +The results presented in clause III.1.3 are magnitude measured in one-third octave across the full plan. + +#### III.1.3 Results comparison + +Two equalization approaches were tested and their results are compared in this clause. The only difference between these two approaches is the delay adjustment step (see clause 8.2.2.4). + +The first approach is slightly different to that described in this Recommendation, since it aims to perfectly synchronize the signals coming from the four loudspeakers at the measurement microphone. The advantage of this method is that it leads to a very good equalization at the measurement microphone (i.e., the equalization centre). This method is referred to as "anechoic room: synchronized delays" in this appendix. + +The second approach is that described in this Recommendation where fixed delays (0 ms, 11 ms, 17 ms, 29 ms) are applied to individual loudspeakers in order to increase the diffusivity of the noise field. For the actual examination, different delays were used, i.e., 0 ms, 13 ms, 22 ms, 45 ms, as in that configuration they gave a flatter response than using the recommended delays. This is illustrated in Figure III.4, where the result of equalization is shown in one-third octave for the recommended and modified sets of delays. This retained method is referred to as "anechoic room: fixed delays" in this appendix. + +![Figure III.4: Frequency response (dB) vs Frequency (Hz) plot comparing two equalization methods. The x-axis is logarithmic from 100 to 10,000 Hz. The y-axis is linear from -7 to 2 dB. Two lines are shown: a blue line for recommended delays (0, 11, 17, 29 ms) and an orange line for modified delays (0, 13, 22, 45 ms). The orange line is significantly flatter than the blue line, especially in the 100-1000 Hz range.](d3c5090abad168738b2dd27087f89fdc_img.jpg) + +The figure is a line graph showing the frequency response in dB on the y-axis (ranging from -7 to 2) against frequency in Hz on a logarithmic x-axis (ranging from 100 to 10,000). Two data series are plotted: a blue line representing 'Fixed delays: 0, 11, 17, 29 ms' and an orange line representing 'Fixed delays: 0, 13, 22, 45 ms'. The blue line shows significant fluctuations, with a sharp peak around 150 Hz and a sharp dip around 200 Hz, followed by smaller oscillations. The orange line is much flatter, staying mostly within the -1 dB to 0 dB range across the entire frequency spectrum. Two horizontal blue lines are drawn at approximately 2 dB and -4 dB. + +| Frequency (Hz) | Fixed delays: 0, 11, 17, 29 ms (dB) | Fixed delays: 0, 13, 22, 45 ms (dB) | +|----------------|-------------------------------------|-------------------------------------| +| 100 | -7 | -0.5 | +| 125 | 2 | -1.5 | +| 160 | -3 | -0.5 | +| 200 | -1 | -0.5 | +| 250 | -1.5 | -0.5 | +| 315 | -0.5 | -0.5 | +| 400 | -1.5 | -0.5 | +| 500 | -0.5 | -0.5 | +| 630 | -1.5 | -0.5 | +| 800 | -0.5 | -0.5 | +| 1000 | -1.5 | -0.5 | +| 1250 | -0.5 | -0.5 | +| 1600 | -1.5 | -0.5 | +| 2000 | -0.5 | -0.5 | +| 2500 | -1.5 | -0.5 | +| 3150 | -0.5 | -0.5 | +| 4000 | -1.5 | -0.5 | +| 5000 | -0.5 | -0.5 | +| 6300 | -1.5 | -0.5 | +| 8000 | -0.5 | -0.5 | +| 10000 | -1.5 | -0.5 | + +Figure III.4: Frequency response (dB) vs Frequency (Hz) plot comparing two equalization methods. The x-axis is logarithmic from 100 to 10,000 Hz. The y-axis is linear from -7 to 2 dB. Two lines are shown: a blue line for recommended delays (0, 11, 17, 29 ms) and an orange line for modified delays (0, 13, 22, 45 ms). The orange line is significantly flatter than the blue line, especially in the 100-1000 Hz range. + +P.570(18)\_FIII.4 + +**Figure III.4 – Results of equalization in one-third octave for all loudspeakers with fixed delays: for recommended delays, 0 ms, 11 ms, 17 ms, 29 ms (in blue); for the modified set of delays, 0 ms, 13 ms, 22 ms, 45 ms (in orange)** + +The results for both approaches are shown in Table III.1. Each row corresponds to a one-third octave band. The displayed results start with the one-third octave band centred on 126 Hz and ends with that centred on 19 953 Hz. + +Measurements were done every 10 mm on both the *X*- and *Y*-axis; however, in order to get better visual results, images were interpolated with a 1 mm step. Note that the 10 mm step is too large to capture the characteristics at high frequencies. + +**Table III.1 – Anechoic room results – left: synchronized delays; right: fixed delays** + +![Heatmap showing spatial stability of noise field at 126 Hz for synchronized delays. The plot shows a uniform dark red color across the entire 50x50 mm area, indicating high stability. The color scale on the right ranges from 0 dB (dark red) to -20 dB (dark blue). Heatmap showing spatial stability of noise field at 126 Hz for fixed delays. The plot shows a uniform dark red color across the entire 50x50 mm area, indicating high stability. The color scale on the right ranges from 0 dB (dark red) to -20 dB (dark blue). Heatmap showing spatial stability of noise field at 158 Hz for synchronized delays. The plot shows a uniform dark red color across the entire 50x50 mm area, indicating high stability. The color scale on the right ranges from 0 dB (dark red) to -20 dB (dark blue). Heatmap showing spatial stability of noise field at 158 Hz for fixed delays. The plot shows a uniform dark red color across the entire 50x50 mm area, indicating high stability. The color scale on the right ranges from 0 dB (dark red) to -20 dB (dark blue).](509a054727400bc6d424bb2a559b8cfc_img.jpg) + +| Anechoic room: Synchronized delays | Anechoic room: Fixed delays | +|------------------------------------------------------|------------------------------------------------------| +|

Spatial stability of noise field at 126 Hz dB

|

Spatial stability of noise field at 126 Hz dB

| +|

Spatial stability of noise field at 158 Hz dB

|

Spatial stability of noise field at 158 Hz dB

| + +Heatmap showing spatial stability of noise field at 126 Hz for synchronized delays. The plot shows a uniform dark red color across the entire 50x50 mm area, indicating high stability. The color scale on the right ranges from 0 dB (dark red) to -20 dB (dark blue). Heatmap showing spatial stability of noise field at 126 Hz for fixed delays. The plot shows a uniform dark red color across the entire 50x50 mm area, indicating high stability. The color scale on the right ranges from 0 dB (dark red) to -20 dB (dark blue). Heatmap showing spatial stability of noise field at 158 Hz for synchronized delays. The plot shows a uniform dark red color across the entire 50x50 mm area, indicating high stability. The color scale on the right ranges from 0 dB (dark red) to -20 dB (dark blue). Heatmap showing spatial stability of noise field at 158 Hz for fixed delays. The plot shows a uniform dark red color across the entire 50x50 mm area, indicating high stability. The color scale on the right ranges from 0 dB (dark red) to -20 dB (dark blue). + +![Heatmap showing spatial stability of noise field at 200 Hz for synchronized delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue). Heatmap showing spatial stability of noise field at 200 Hz for fixed delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue). Heatmap showing spatial stability of noise field at 251 Hz for synchronized delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue). Heatmap showing spatial stability of noise field at 251 Hz for fixed delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue). Heatmap showing spatial stability of noise field at 316 Hz for synchronized delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue). Heatmap showing spatial stability of noise field at 316 Hz for fixed delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue).](f117f4876cd1523b1717f3e38bd74229_img.jpg) + +| Anechoic room: Synchronized delays | Anechoic room: Fixed delays | +|------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 200 Hz dB

|

Spatial stability of noise field at 200 Hz dB

| +|

Spatial stability of noise field at 251 Hz dB

|

Spatial stability of noise field at 251 Hz dB

| +|

Spatial stability of noise field at 316 Hz dB

|

Spatial stability of noise field at 316 Hz dB

| + +Heatmap showing spatial stability of noise field at 200 Hz for synchronized delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue). Heatmap showing spatial stability of noise field at 200 Hz for fixed delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue). Heatmap showing spatial stability of noise field at 251 Hz for synchronized delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue). Heatmap showing spatial stability of noise field at 251 Hz for fixed delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue). Heatmap showing spatial stability of noise field at 316 Hz for synchronized delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue). Heatmap showing spatial stability of noise field at 316 Hz for fixed delays. The plot is almost entirely dark red, indicating values near 0 dB across the -50 to 50 mm range on both X and Y axes. A color bar on the right shows the scale from 0 dB (red) to -20 dB (blue). + +![Heatmap showing spatial stability of noise field at 398 Hz with synchronized delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field is almost entirely dark red, indicating near 0 dB variation across the entire area. Heatmap showing spatial stability of noise field at 398 Hz with fixed delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field is uniform dark red, indicating 0 dB variation. Heatmap showing spatial stability of noise field at 501 Hz with synchronized delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field is mostly dark red with a very slight variation in the center. Heatmap showing spatial stability of noise field at 501 Hz with fixed delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field is mostly dark red with a slight gradient towards a lighter red on the left side. Heatmap showing spatial stability of noise field at 631 Hz with synchronized delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field is dark red with a small, slightly lighter red spot in the center. Heatmap showing spatial stability of noise field at 631 Hz with fixed delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field shows a slight gradient from dark red on the right to a slightly lighter red on the left.](7e5f8a1e5575ef87a54c9c091eaaa66d_img.jpg) + +| Anechoic room: Synchronized delays | Anechoic room: Fixed delays | +|-------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 398 Hz

dB

|

Spatial stability of noise field at 398 Hz

dB

| +|

Spatial stability of noise field at 501 Hz

dB

|

Spatial stability of noise field at 501 Hz

dB

| +|

Spatial stability of noise field at 631 Hz

dB

|

Spatial stability of noise field at 631 Hz

dB

| + +Heatmap showing spatial stability of noise field at 398 Hz with synchronized delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field is almost entirely dark red, indicating near 0 dB variation across the entire area. Heatmap showing spatial stability of noise field at 398 Hz with fixed delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field is uniform dark red, indicating 0 dB variation. Heatmap showing spatial stability of noise field at 501 Hz with synchronized delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field is mostly dark red with a very slight variation in the center. Heatmap showing spatial stability of noise field at 501 Hz with fixed delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field is mostly dark red with a slight gradient towards a lighter red on the left side. Heatmap showing spatial stability of noise field at 631 Hz with synchronized delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field is dark red with a small, slightly lighter red spot in the center. Heatmap showing spatial stability of noise field at 631 Hz with fixed delays. The plot shows a 50x50 mm area with a color scale from -20 dB (blue) to 0 dB (red). The field shows a slight gradient from dark red on the right to a slightly lighter red on the left. + +![Heatmap showing spatial stability of noise field at 794 Hz with synchronized delays. The plot shows a relatively uniform red color across the 100x100 mm area, with slight orange tinting at the corners, indicating high stability. The color scale ranges from 0 to -20 dB. Heatmap showing spatial stability of noise field at 794 Hz with fixed delays. The plot shows a uniform dark red color across the 100x100 mm area, indicating very high stability. The color scale ranges from 0 to -20 dB. Heatmap showing spatial stability of noise field at 1000 Hz with synchronized delays. The plot shows a uniform red color across the 100x100 mm area, with slight orange tinting at the corners, indicating high stability. The color scale ranges from 0 to -20 dB. Heatmap showing spatial stability of noise field at 1000 Hz with fixed delays. The plot shows a uniform dark red color across the 100x100 mm area, indicating very high stability. The color scale ranges from 0 to -20 dB. Heatmap showing spatial stability of noise field at 1259 Hz with synchronized delays. The plot shows a uniform red color across the 100x100 mm area, with more noticeable orange/yellow tinting at the corners compared to lower frequencies, indicating high stability. The color scale ranges from 0 to -20 dB. Heatmap showing spatial stability of noise field at 1259 Hz with fixed delays. The plot shows a uniform dark red color across the 100x100 mm area, indicating very high stability. The color scale ranges from 0 to -20 dB.](c2e29ea8ce248c31df2f0aba049d018c_img.jpg) + +| Anechoic room: Synchronized delays | Anechoic room: Fixed delays | +|-------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 794 Hz dB

|

Spatial stability of noise field at 794 Hz dB

| +|

Spatial stability of noise field at 1000 Hz dB

|

Spatial stability of noise field at 1000 Hz dB

| +|

Spatial stability of noise field at 1259 Hz dB

|

Spatial stability of noise field at 1259 Hz dB

| + +Heatmap showing spatial stability of noise field at 794 Hz with synchronized delays. The plot shows a relatively uniform red color across the 100x100 mm area, with slight orange tinting at the corners, indicating high stability. The color scale ranges from 0 to -20 dB. Heatmap showing spatial stability of noise field at 794 Hz with fixed delays. The plot shows a uniform dark red color across the 100x100 mm area, indicating very high stability. The color scale ranges from 0 to -20 dB. Heatmap showing spatial stability of noise field at 1000 Hz with synchronized delays. The plot shows a uniform red color across the 100x100 mm area, with slight orange tinting at the corners, indicating high stability. The color scale ranges from 0 to -20 dB. Heatmap showing spatial stability of noise field at 1000 Hz with fixed delays. The plot shows a uniform dark red color across the 100x100 mm area, indicating very high stability. The color scale ranges from 0 to -20 dB. Heatmap showing spatial stability of noise field at 1259 Hz with synchronized delays. The plot shows a uniform red color across the 100x100 mm area, with more noticeable orange/yellow tinting at the corners compared to lower frequencies, indicating high stability. The color scale ranges from 0 to -20 dB. Heatmap showing spatial stability of noise field at 1259 Hz with fixed delays. The plot shows a uniform dark red color across the 100x100 mm area, indicating very high stability. The color scale ranges from 0 to -20 dB. + +![Heatmap showing spatial stability of noise field at 1585 Hz with synchronized delays. The plot shows a relatively uniform distribution of stability across the 50x50 mm area, with values mostly in the -2 to -6 dB range (orange/red). Heatmap showing spatial stability of noise field at 1585 Hz with fixed delays. The plot is almost entirely dark red, indicating very low spatial stability (near 0 dB) across the entire area. Heatmap showing spatial stability of noise field at 1995 Hz with synchronized delays. The plot shows a central region of high stability (red) surrounded by a ring of lower stability (yellow/green). Heatmap showing spatial stability of noise field at 1995 Hz with fixed delays. The plot is almost entirely dark red, indicating very low spatial stability (near 0 dB) across the entire area. Heatmap showing spatial stability of noise field at 2512 Hz with synchronized delays. The plot shows a central region of high stability (red) surrounded by a ring of lower stability (yellow/green), similar to the 1995 Hz plot. Heatmap showing spatial stability of noise field at 2512 Hz with fixed delays. The plot is almost entirely dark red, indicating very low spatial stability (near 0 dB) across the entire area.](ce857617d1453e7e90edfaae561e2f62_img.jpg) + +| Anechoic room: Synchronized delays | Anechoic room: Fixed delays | +|---------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------| +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 1585 Hz dB

|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 1585 Hz dB

| +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 1995 Hz dB

|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 1995 Hz dB

| +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 2512 Hz dB

|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 2512 Hz dB

| + +Heatmap showing spatial stability of noise field at 1585 Hz with synchronized delays. The plot shows a relatively uniform distribution of stability across the 50x50 mm area, with values mostly in the -2 to -6 dB range (orange/red). Heatmap showing spatial stability of noise field at 1585 Hz with fixed delays. The plot is almost entirely dark red, indicating very low spatial stability (near 0 dB) across the entire area. Heatmap showing spatial stability of noise field at 1995 Hz with synchronized delays. The plot shows a central region of high stability (red) surrounded by a ring of lower stability (yellow/green). Heatmap showing spatial stability of noise field at 1995 Hz with fixed delays. The plot is almost entirely dark red, indicating very low spatial stability (near 0 dB) across the entire area. Heatmap showing spatial stability of noise field at 2512 Hz with synchronized delays. The plot shows a central region of high stability (red) surrounded by a ring of lower stability (yellow/green), similar to the 1995 Hz plot. Heatmap showing spatial stability of noise field at 2512 Hz with fixed delays. The plot is almost entirely dark red, indicating very low spatial stability (near 0 dB) across the entire area. + +![Heatmap showing spatial stability of noise field at 3162 Hz with synchronized delays. The plot shows a large central red/orange region of high stability (0 to -2 dB) surrounded by concentric rings of decreasing stability (yellow, green, blue) reaching -20 dB at the edges. Axes range from -50 to 50 mm. Heatmap showing spatial stability of noise field at 3162 Hz with fixed delays. The entire plot area is dark red, indicating high stability (0 to -2 dB) across the entire -50 to 50 mm range. Axes range from -50 to 50 mm. Heatmap showing spatial stability of noise field at 3981 Hz with synchronized delays. The stability pattern is more complex than at 3162 Hz, with multiple red/orange regions of high stability separated by areas of lower stability (blue/green). Axes range from -50 to 50 mm. Heatmap showing spatial stability of noise field at 3981 Hz with fixed delays. The entire plot area is dark red, indicating high stability (0 to -2 dB) across the entire -50 to 50 mm range. Axes range from -50 to 50 mm. Heatmap showing spatial stability of noise field at 5012 Hz with synchronized delays. The stability pattern is highly complex, with many small red/orange regions of high stability separated by a dense network of blue/green lines of lower stability. Axes range from -50 to 50 mm. Heatmap showing spatial stability of noise field at 5012 Hz with fixed delays. The entire plot area is dark red, indicating high stability (0 to -2 dB) across the entire -50 to 50 mm range. Axes range from -50 to 50 mm.](8605b43090d582ea5046e1804178a7d8_img.jpg) + +| Anechoic room: Synchronized delays | Anechoic room: Fixed delays | +|-------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 3162 Hz dB

|

Spatial stability of noise field at 3162 Hz dB

| +|

Spatial stability of noise field at 3981 Hz dB

|

Spatial stability of noise field at 3981 Hz dB

| +|

Spatial stability of noise field at 5012 Hz dB

|

Spatial stability of noise field at 5012 Hz dB

| + +Heatmap showing spatial stability of noise field at 3162 Hz with synchronized delays. The plot shows a large central red/orange region of high stability (0 to -2 dB) surrounded by concentric rings of decreasing stability (yellow, green, blue) reaching -20 dB at the edges. Axes range from -50 to 50 mm. Heatmap showing spatial stability of noise field at 3162 Hz with fixed delays. The entire plot area is dark red, indicating high stability (0 to -2 dB) across the entire -50 to 50 mm range. Axes range from -50 to 50 mm. Heatmap showing spatial stability of noise field at 3981 Hz with synchronized delays. The stability pattern is more complex than at 3162 Hz, with multiple red/orange regions of high stability separated by areas of lower stability (blue/green). Axes range from -50 to 50 mm. Heatmap showing spatial stability of noise field at 3981 Hz with fixed delays. The entire plot area is dark red, indicating high stability (0 to -2 dB) across the entire -50 to 50 mm range. Axes range from -50 to 50 mm. Heatmap showing spatial stability of noise field at 5012 Hz with synchronized delays. The stability pattern is highly complex, with many small red/orange regions of high stability separated by a dense network of blue/green lines of lower stability. Axes range from -50 to 50 mm. Heatmap showing spatial stability of noise field at 5012 Hz with fixed delays. The entire plot area is dark red, indicating high stability (0 to -2 dB) across the entire -50 to 50 mm range. Axes range from -50 to 50 mm. + +![Heatmap showing spatial stability of noise field at 6310 Hz with synchronized delays. The plot shows a grid-like interference pattern with peaks (red, 0 dB) and troughs (blue, -20 dB) across the 100x100 mm area. Heatmap showing spatial stability of noise field at 6310 Hz with fixed delays. The plot is uniformly dark red, indicating high stability (0 dB) across the entire area. Heatmap showing spatial stability of noise field at 7943 Hz with synchronized delays. Shows a complex interference pattern with more frequent oscillations than the 6310 Hz plot. Heatmap showing spatial stability of noise field at 7943 Hz with fixed delays. The plot remains uniformly dark red (0 dB). Heatmap showing spatial stability of noise field at 10000 Hz with synchronized delays. Shows a very dense interference pattern with many small peaks and troughs. Heatmap showing spatial stability of noise field at 10000 Hz with fixed delays. The plot is still uniformly dark red, showing high stability even at high frequency.](23f81127919fb6dcb76edf3f78f56953_img.jpg) + +| Anechoic room: Synchronized delays | Anechoic room: Fixed delays | +|----------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 6310 Hz     dB

|

Spatial stability of noise field at 6310 Hz     dB

| +|

Spatial stability of noise field at 7943 Hz     dB

|

Spatial stability of noise field at 7943 Hz     dB

| +|

Spatial stability of noise field at 10000 Hz     dB

|

Spatial stability of noise field at 10000 Hz     dB

| + +Heatmap showing spatial stability of noise field at 6310 Hz with synchronized delays. The plot shows a grid-like interference pattern with peaks (red, 0 dB) and troughs (blue, -20 dB) across the 100x100 mm area. Heatmap showing spatial stability of noise field at 6310 Hz with fixed delays. The plot is uniformly dark red, indicating high stability (0 dB) across the entire area. Heatmap showing spatial stability of noise field at 7943 Hz with synchronized delays. Shows a complex interference pattern with more frequent oscillations than the 6310 Hz plot. Heatmap showing spatial stability of noise field at 7943 Hz with fixed delays. The plot remains uniformly dark red (0 dB). Heatmap showing spatial stability of noise field at 10000 Hz with synchronized delays. Shows a very dense interference pattern with many small peaks and troughs. Heatmap showing spatial stability of noise field at 10000 Hz with fixed delays. The plot is still uniformly dark red, showing high stability even at high frequency. + +![Heatmap showing spatial stability of noise field at 12589 Hz with synchronized delays. The plot shows a complex interference pattern with high stability (red) and low stability (blue) regions across a 50x50 mm area. The color scale ranges from -20 dB (blue) to 0 dB (red). Heatmap showing spatial stability of noise field at 12589 Hz with fixed delays. The plot is almost entirely red, indicating high spatial stability across the entire 50x50 mm area. The color scale ranges from -20 dB (blue) to 0 dB (red). Heatmap showing spatial stability of noise field at 15849 Hz with synchronized delays. Similar to the 12589 Hz plot, it shows a complex interference pattern. The color scale ranges from -20 dB (blue) to 0 dB (red). Heatmap showing spatial stability of noise field at 15849 Hz with fixed delays. The plot is almost entirely red, indicating high spatial stability across the entire 50x50 mm area. The color scale ranges from -20 dB (blue) to 0 dB (red). Heatmap showing spatial stability of noise field at 19953 Hz with synchronized delays. The plot shows a complex interference pattern. The color scale ranges from -20 dB (blue) to 0 dB (red). Heatmap showing spatial stability of noise field at 19953 Hz with fixed delays. The plot is almost entirely red, indicating high spatial stability across the entire 50x50 mm area. The color scale ranges from -20 dB (blue) to 0 dB (red).](de7b700c44b3911907685dffd54fc50f_img.jpg) + +| Anechoic room: Synchronized delays | Anechoic room: Fixed delays | +|--------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 12589 Hz dB

|

Spatial stability of noise field at 12589 Hz dB

| +|

Spatial stability of noise field at 15849 Hz dB

|

Spatial stability of noise field at 15849 Hz dB

| +|

Spatial stability of noise field at 19953 Hz dB

|

Spatial stability of noise field at 19953 Hz dB

| + +Heatmap showing spatial stability of noise field at 12589 Hz with synchronized delays. The plot shows a complex interference pattern with high stability (red) and low stability (blue) regions across a 50x50 mm area. The color scale ranges from -20 dB (blue) to 0 dB (red). Heatmap showing spatial stability of noise field at 12589 Hz with fixed delays. The plot is almost entirely red, indicating high spatial stability across the entire 50x50 mm area. The color scale ranges from -20 dB (blue) to 0 dB (red). Heatmap showing spatial stability of noise field at 15849 Hz with synchronized delays. Similar to the 12589 Hz plot, it shows a complex interference pattern. The color scale ranges from -20 dB (blue) to 0 dB (red). Heatmap showing spatial stability of noise field at 15849 Hz with fixed delays. The plot is almost entirely red, indicating high spatial stability across the entire 50x50 mm area. The color scale ranges from -20 dB (blue) to 0 dB (red). Heatmap showing spatial stability of noise field at 19953 Hz with synchronized delays. The plot shows a complex interference pattern. The color scale ranges from -20 dB (blue) to 0 dB (red). Heatmap showing spatial stability of noise field at 19953 Hz with fixed delays. The plot is almost entirely red, indicating high spatial stability across the entire 50x50 mm area. The color scale ranges from -20 dB (blue) to 0 dB (red). + +Obviously, spatial stability results for these two approaches are very different. For the fixed delays, the noise field is very stable on the measured plan. In one-third octave, a variation of less than 3 dB over the full frequency range (120 Hz to 20 kHz) can be observed. + +For the synchronized delays, the noise field is stable (less than 3dB of variation) up to around 1 kHz. For higher frequencies, attenuations occur. These attenuations follow a regular pattern that can be associated with a spatial comb filter. The size of this pattern decreases when frequency increases, i.e., when wavelength decreases. This behaviour is less desirable from a spatial stability perspective than that obtained using fixed delays, because the resulting noise field is not stable (up to 21 dB of attenuation), which is an issue when testing mobile phone noise reduction, for instance. + +So, the noise field stability obtained using fixed delays seems to be very good. It is true when analysed in third octave because signal spectrum is averaged over a large frequency range (increasing with frequency). When using a higher resolution analysis (i.e., FFT) a degradation of the spectrum can be observed. This phenomenon is illustrated in Figure III.5 where the signals equalized at $[X = 0 \text{ mm}; Y = 0 \text{ mm}]$ for both approaches are displayed on a linear scale. The frequency range is limited to 2 kHz for the sake of clarity. The original pink noise is also displayed (red curve) for comparison. It can be seen that the FFT spectrum obtained using synchronized delays method (green curve) is very close to the original noise spectrum. However, the spectrum obtained using fixed delays: 0 ms, 13 ms, 22 ms, 45 ms (orange curve) is affected by a strong comb effect filter. This comb effect filter is also present with the regular delays (blue curve) proposed in this Recommendation (0 ms, 11 ms, 17 ms, 29 ms). + +![Figure III.5: Fast Fourier transform spectrum plot showing Power spectrum (dBFS) on the y-axis (from -70 to -30) versus Frequency (Hz) on the x-axis (from 0 to 2 000). The plot compares four signals: Original pink noise (red line), Synchronized delays (green line), Fixed delays: 0, 13, 22, 45 ms (orange line), and Fixed delays: 0, 11, 17, 29 ms (blue line). The red and green lines are relatively flat, while the orange and blue lines show significant periodic dips (comb effect).](89b3389b2c4fa1b452cf2b5daa5f82ad_img.jpg) + +Figure III.5: Fast Fourier transform spectrum plot showing Power spectrum (dBFS) on the y-axis (from -70 to -30) versus Frequency (Hz) on the x-axis (from 0 to 2 000). The plot compares four signals: Original pink noise (red line), Synchronized delays (green line), Fixed delays: 0, 13, 22, 45 ms (orange line), and Fixed delays: 0, 11, 17, 29 ms (blue line). The red and green lines are relatively flat, while the orange and blue lines show significant periodic dips (comb effect). + +P.570(25) + +**Figure III.5 – Fast Fourier transform spectrum of original high-pass pink noise (red), equalized signal obtained using synchronized delays (green), equalized signal obtained using fixed delays: 0 ms, 13 ms, 22 ms, 45 ms (orange) and equalized signal obtained using fixed delays: 0 ms, 11 ms, 17 ms, 29 ms (blue)** + +Equalization using fixed delays has been tried with several other sets of delays and using other test signals than pink noise, but this comb effect was always there and as strong as illustrated in Figure III.5. + +#### III.1.4 Summary + +This clause documents the spatial stability of a noise field obtained using this Recommendation. As illustrated in clause III.1.3, the delay adjustment has a major impact on the equalized noise field. When using fixed delays, the noise field is very stable on the measured plan which is not the case when using + +synchronized delays. However, when using fixed delays, the reproduced noise field is subject to a comb effect filter that is noticeable with a frequency resolution higher than one-third octave. + +### **III.2 Reproduction of monophonic signals in treated rooms** + +This is currently not covered in this Recommendation. + +### **III.3 Reproduction of stereophonic signals in anechoic conditions** + +This is currently not covered in this Recommendation. + +### **III.4 Reproduction of stereophonic signals in treated rooms** + +This is currently not covered in this Recommendation. + +### **III.5 Reproduction of binaural signals in anechoic conditions** + +This is currently not covered in this Recommendation. + +### **III.6 Reproduction of binaural signals in treated rooms** + +This is currently not covered in this Recommendation. + +### **III.7 Reproduction of sound field signals in anechoic conditions** + +This is currently not covered in this Recommendation. + +### **III.8 Reproduction of sound field signals in treated rooms** + +This is currently not covered in this Recommendation. + +### **III.9 Reproduction of multi-point signals in anechoic conditions and treated rooms** + +#### **III.9.1 Introduction** + +This clause documents the reproduction accuracy of the reproduced noise field that can be achieved using the multi-point sound field recording and reproduction arrangement as described in clause 8.3.7. The measurements were performed in various types of rooms and for handset terminals as well as for hands-free terminals. + +#### **III.9.2 Results of the equalization of the multi-microphone method** + +The comparison with the original sound field is an inherent part of the equalization technique. Figures III.6 and III.7 show the results of the equalization process when applied to different types of rooms. The differences in the microphone levels versus frequency were measured, as well as the complex coherence between the recorded sound field and the simulated sound field after equalization at the relevant microphone positions used during calibration (Figure III.6) and used during fine-tuning (Figure III.7) for the four different types of rooms listed in Table III.2: + +**Table III.2 – Types of rooms in which the validation was performed** + +| Name | Reverberation time,
60 dB reduction, RT60
(ms) | C_{80} (dB) | +|-------------|---------------------------------------------------------------|---------------------------------| +| Room 1 | 27 | 50.3 | +| Room 2 | 123 | 36.8 | +| Room 3 | 98 | 45.6 | +| Room 4 | 264 | 20.3 | + +The room sizes were between $1.8 \text{ m} \times 2.4 \text{ m} \times 2.1 \text{ m}$ (smallest) and $8 \text{ m} \times 9 \text{ m} \times 4.5 \text{ m}$ (biggest) ( $d \times b \times h$ ). For all rooms, the equalization was performed within the requirements of this Recommendation. The absolute level difference was less than $\pm 1 \text{ dB}$ , the spectral accuracy of the simulated sound field was within the range from $-3 \text{ dB}$ to $+3 \text{ dB}$ in the frequency range from $50 \text{ Hz}$ to $10 \text{ kHz}$ , and the magnitude of the complex coherence was higher than $90\%$ in the frequency range from $100 \text{ Hz}$ to $1 \, 000 \text{ Hz}$ or even up to $2 \, 000 \text{ Hz}$ . + +NOTE – Channel 1 and 2 are not displayed here since they are on the "shadow side" of the head and only relevant for the equalization process. + +![Figure III.6: Spectral accuracy and complex coherence plots for four rooms (Room 1 to Room 4). Each room has two plots: the left plot shows spectral accuracy (L/dB) for Channels 3-8, and the right plot shows complex coherence magnitude (solid lines) and phase (dashed lines) for the same channels. The x-axis for all plots is frequency (f/Hz) on a logarithmic scale from 50 to 10,000. The left plots have a y-axis from -6 to 6 L/dB. The right plots have a left y-axis for phase (Phi/°) from -180 to 180 and a right y-axis for correlation (%) from 0 to 100. Room 1 has an RT60 of 27 ms, Room 2 has 123 ms, Room 3 has 98 ms, and Room 4 has 264 ms. The plots show that spectral accuracy is generally good (within ±2 dB) while coherence magnitude decreases and phase becomes more variable at higher frequencies, especially in rooms with longer reverberation times.](795f43a3b6469179db3a5670f31388b5_img.jpg) + +Figure III.6: Spectral accuracy and complex coherence plots for four rooms (Room 1 to Room 4). Each room has two plots: the left plot shows spectral accuracy (L/dB) for Channels 3-8, and the right plot shows complex coherence magnitude (solid lines) and phase (dashed lines) for the same channels. The x-axis for all plots is frequency (f/Hz) on a logarithmic scale from 50 to 10,000. The left plots have a y-axis from -6 to 6 L/dB. The right plots have a left y-axis for phase (Phi/°) from -180 to 180 and a right y-axis for correlation (%) from 0 to 100. Room 1 has an RT60 of 27 ms, Room 2 has 123 ms, Room 3 has 98 ms, and Room 4 has 264 ms. The plots show that spectral accuracy is generally good (within ±2 dB) while coherence magnitude decreases and phase becomes more variable at higher frequencies, especially in rooms with longer reverberation times. + +P.570(25) + +**Figure III.6 – Left: spectral accuracy between the simulated and reference sound fields; right: the complex coherence in magnitude (solid lines) and phase (dashed lines) at the positions of the calibration set. Results are in one-third octave** + +![A 4x2 grid of plots showing spectral accuracy and complex coherence for four rooms (Room 1 to Room 4) across eight channels. The left column shows spectral accuracy (L/dB) vs frequency (f/Hz) for RT60 values of 27 ms, 123 ms, 98 ms, and 264 ms. The right column shows complex coherence magnitude (solid lines) and phase (dashed lines) in degrees vs frequency (f/Hz).](a4cb63f5debf833e6069e99143a2b624_img.jpg) + +The figure displays eight plots arranged in a 4x2 grid, showing results for four rooms (Room 1, Room 2, Room 3, Room 4) across eight channels (Channel 3 to Channel 8). + +**Left Column: Spectral Accuracy** + +- Room 1 (RT60: 27 ms):** Spectral accuracy (L/dB) vs frequency (f/Hz). The y-axis ranges from -6 to 6 L/dB, and the x-axis ranges from 50 to 10 k Hz. The legend indicates Channels 3, 4, 5, 6, 7, and 8. +- Room 2 (RT60: 123 ms):** Spectral accuracy (L/dB) vs frequency (f/Hz). The y-axis ranges from -6 to 6 L/dB, and the x-axis ranges from 50 to 10 k Hz. The legend indicates Channels 3, 4, 5, 6, 7, and 8. +- Room 3 (RT60: 98 ms):** Spectral accuracy (L/dB) vs frequency (f/Hz). The y-axis ranges from -6 to 6 L/dB, and the x-axis ranges from 50 to 10 k Hz. The legend indicates Channels 3, 4, 5, 6, 7, and 8. +- Room 4 (RT60: 264 ms):** Spectral accuracy (L/dB) vs frequency (f/Hz). The y-axis ranges from -6 to 6 L/dB, and the x-axis ranges from 50 to 10 k Hz. The legend indicates Channels 3, 4, 5, 6, 7, and 8. + +**Right Column: Complex Coherence** + +- Room 1:** Complex coherence magnitude (solid lines) and phase (dashed lines) vs frequency (f/Hz). The left y-axis (Phi/°) ranges from -180 to 180, and the right y-axis (Correlation/%) ranges from 0 to 100. The x-axis ranges from 50 to 10 k Hz. +- Room 2:** Complex coherence magnitude (solid lines) and phase (dashed lines) vs frequency (f/Hz). The left y-axis (Phi/°) ranges from -180 to 180, and the right y-axis (Correlation/%) ranges from 0 to 100. The x-axis ranges from 50 to 10 k Hz. +- Room 3:** Complex coherence magnitude (solid lines) and phase (dashed lines) vs frequency (f/Hz). The left y-axis (Phi/°) ranges from -180 to 180, and the right y-axis (Correlation/%) ranges from 0 to 100. The x-axis ranges from 50 to 10 k Hz. +- Room 4:** Complex coherence magnitude (solid lines) and phase (dashed lines) vs frequency (f/Hz). The left y-axis (Phi/°) ranges from -180 to 180, and the right y-axis (Correlation/%) ranges from 0 to 100. The x-axis ranges from 50 to 10 k Hz. + +P.570(25) + +A 4x2 grid of plots showing spectral accuracy and complex coherence for four rooms (Room 1 to Room 4) across eight channels. The left column shows spectral accuracy (L/dB) vs frequency (f/Hz) for RT60 values of 27 ms, 123 ms, 98 ms, and 264 ms. The right column shows complex coherence magnitude (solid lines) and phase (dashed lines) in degrees vs frequency (f/Hz). + +**Figure III.7 – Left: spectral accuracy between the simulated and reference sound fields; right: the complex coherence in magnitude (solid lines) and phase (dashed lines) at the positions of the fine-tuning set. Results are in one-third octave** + +#### III.9.3 Terminal performance in original and simulated sound field + +##### III.9.3.1 Creation of the reference background noise sound field + +A reference sound field was generated by eight loudspeakers located in a reverberant room not used for sound field simulation. The loudspeakers were positioned at different locations to those used later for the sound field simulation. The only adjustment was of level so that all loudspeakers had approximately the same. + +Figure III.8 shows an example of a set-up for generating a hands-free reference sound field. As can be seen, the loudspeakers were positioned at different heights to make the sound field more realistic. + +![Figure III.8: A photograph of a laboratory setup for generating a reference sound field. A mobile robot with a microphone array is positioned in the center of a room. Four loudspeakers are visible, each with a large red arrow pointing towards the robot. The loudspeakers are at different heights: one on a tall stand to the left, one on a shorter stand to the left, one on a stand to the right, and one on the floor to the left. The room has a whiteboard and a desk with chairs in the background.](2c89962e75ab9e8ceddc51e6bddf3939_img.jpg) + +Figure III.8: A photograph of a laboratory setup for generating a reference sound field. A mobile robot with a microphone array is positioned in the center of a room. Four loudspeakers are visible, each with a large red arrow pointing towards the robot. The loudspeakers are at different heights: one on a tall stand to the left, one on a shorter stand to the left, one on a stand to the right, and one on the floor to the left. The room has a whiteboard and a desk with chairs in the background. + +P.570(18)\_FIII.8 + +**Figure III.8 – The set-up that was used for generating a reference sound field** + +Two different types of noise were used for sound field generation as follows. + +- Eight different car noise recordings were each played back over one loudspeaker. In the following, this recording is referred to as "car noise". The average level was 72.7 dBSPL(A) for handset terminals and 72.2 dBSPL(A) for hands-free terminals. +- A combination of four stationary noises like car noise with four non-stationary noises like cafeteria noise, played simultaneously. In the following this recording is referred to as "mix". The average level was 70.5 dBSPL(A) for handset terminals and 70.1 dBSPL(A) for hands-free terminals. + +Loudspeakers were positioned so as not to occupy the corresponding positions in the reproduction arrangement (see Figure III.9). Those sound fields were recorded using a microphone array as specified in clause 8.2.2. + +##### III.9.3.2 Devices used in the validation experiment + +Three different mobile phones with different properties were positioned in handset-mode. All phones were recent models employing at least two microphone noise cancellation techniques. All phones were wideband enabled and were evaluated in wideband mode – see Table III.3 for details. + +**Table III.3 – Different phones used in the experiments** + +| Name | Network mode | Size (inch) | +|---------|--------------|-------------| +| Phone 1 | 3G | 4.8 | +| Phone 2 | 4G | 5.5 | +| Phone 3 | 3G | 4.3 | +| Phone 4 | 3G | 4.7 | +| Phone 5 | 3G | 4.5 | + +##### III.9.3.3 Analysis of background noise transmission + +The background noise at the primary microphone position, as well as the background noise transmitted by the different phones, was measured in the reference sound field. These measurements served as the reference for all measurements in the different room where the sound field simulation was set up as specified in this Recommendation. + +After those measurements had been carried out, the playback system was equalized in four rooms as described in this Recommendation. In all rooms, the same measurements as in the reference situation were made with the simulated background noise using the microphone array recording employed previously. The results are discussed in the following. + +First the different spectra of the transmitted background noise are compared. The test signal consisted of a conditioning sequence of 30 s speech with background noise and 30 s of background noise. The analysis was performed on the background noise sequence only. + +When simulating the background noise, the simulated noise was picked up by a measurement microphone at the position of the main microphone of a mobile phone (as was also done in the reference recording). In the following, this signal is called "unprocessed". In addition to that, the sending signal of the mobile phone was also recorded. In the following, this signal is called "processed". The spectra of both signals were compared to those recorded in the original sound field. + +In a second step, the speech mean opinion score (S-MOS), noise mean opinion score (N-MOS) and global mean opinion score (G-MOS) values according to [b-ETSI TS 103 106] were calculated. + +In Figures III.9 to III.11, the bold solid graph represents the original spectrum whereas the other graphs represent the spectrum of reproduced sound field in different rooms. + +##### III.9.3.4 Handset measurements I – spectral comparisons + +![Four spectral comparison plots for handset measurements. Top-left: 'Car noise – processed' (L/dB[V] vs f/Hz). Top-right: 'Car noise – unprocessed' (L/dB[Pa] vs f/Hz). Bottom-left: 'Mix – processed' (L/dB[V] vs f/Hz). Bottom-right: 'Mix – unprocessed' (L/dB[Pa] vs f/Hz). Each plot compares Room 1 (RT60 = 27 ms), Room 4 (RT60 = 264 ms), Room 2 (RT60 = 123 ms), Room 3 (RT60 = 98 ms), and a Reference signal. The x-axis is frequency (f/Hz) from 50 to 10,000 Hz. The y-axis is level (L) in dB. Processed plots show smoother spectra, while unprocessed plots show more detailed spectral features.](a972ca1187ea7285bc4009055bcedd34_img.jpg) + +The figure consists of four subplots arranged in a 2x2 grid, showing spectral comparisons for handset measurements. The top-left plot is titled "Car noise – processed" and the top-right is "Car noise – unprocessed". The bottom-left is "Mix – processed" and the bottom-right is "Mix – unprocessed". Each plot shows the level (L) in dB versus frequency (f) in Hz on a logarithmic scale from 50 to 10,000 Hz. The y-axis for the processed plots ranges from -80 to -50 dB[V], and for the unprocessed plots from -70 to -20 dB[Pa]. The legend in each plot identifies the data for Room 1 (RT60 = 27 ms), Room 4 (RT60 = 264 ms), Room 2 (RT60 = 123 ms), Room 3 (RT60 = 98 ms), and a Reference signal. The processed plots show smoother spectra, while the unprocessed plots show more detailed spectral features. + +Four spectral comparison plots for handset measurements. Top-left: 'Car noise – processed' (L/dB[V] vs f/Hz). Top-right: 'Car noise – unprocessed' (L/dB[Pa] vs f/Hz). Bottom-left: 'Mix – processed' (L/dB[V] vs f/Hz). Bottom-right: 'Mix – unprocessed' (L/dB[Pa] vs f/Hz). Each plot compares Room 1 (RT60 = 27 ms), Room 4 (RT60 = 264 ms), Room 2 (RT60 = 123 ms), Room 3 (RT60 = 98 ms), and a Reference signal. The x-axis is frequency (f/Hz) from 50 to 10,000 Hz. The y-axis is level (L) in dB. Processed plots show smoother spectra, while unprocessed plots show more detailed spectral features. + +P.570(25) + +Figure III.9 – Phone 1, handset mode, comparison of one-twelfth-octave spectra for car noise and mix signal in different rooms (left: processed; right: unprocessed) + +![Four line graphs comparing one-twelfth-octave spectra for car noise and mix signals in different rooms (Room 1, Room 2, Room 3) and a reference. The graphs are arranged in a 2x2 grid: top-left is 'Car noise - processed' (L/dB[V]), top-right is 'Car noise - unprocessed' (L/dB[Pa]), bottom-left is 'Mix - processed' (L/dB[V]), and bottom-right is 'Mix - unprocessed' (L/dB[Pa]). All graphs show spectral levels across a frequency range from 50 Hz to 10 kHz. The 'unprocessed' graphs show more pronounced room-specific spectral differences compared to the 'processed' graphs, which appear flatter and closer to the reference.](e79b237b31616233842b1c75f51edeae_img.jpg) + +Four line graphs comparing one-twelfth-octave spectra for car noise and mix signals in different rooms (Room 1, Room 2, Room 3) and a reference. The graphs are arranged in a 2x2 grid: top-left is 'Car noise - processed' (L/dB[V]), top-right is 'Car noise - unprocessed' (L/dB[Pa]), bottom-left is 'Mix - processed' (L/dB[V]), and bottom-right is 'Mix - unprocessed' (L/dB[Pa]). All graphs show spectral levels across a frequency range from 50 Hz to 10 kHz. The 'unprocessed' graphs show more pronounced room-specific spectral differences compared to the 'processed' graphs, which appear flatter and closer to the reference. + +P.570(25) + +**Figure III.10 – Phone 2, handset mode, comparison of one-twelfth-octave spectra for car noise and mix signal in different rooms (left: processed; right: unprocessed)** + +![Four line graphs showing one-twelfth-octave spectra for car noise and mix signals in different rooms. The top-left graph shows 'Car noise - processed' with L/dB[V] on the y-axis. The top-right graph shows 'Car noise - unprocessed' with L/dB[Pa] on the y-axis. The bottom-left graph shows 'Mix - processed' with L/dB[V] on the y-axis. The bottom-right graph shows 'Mix - unprocessed' with L/dB[Pa] on the y-axis. All graphs have f/Hz on the x-axis (logarithmic scale from 50 to 10k). Each graph contains five data series: Room 4 (RT60 = 264 ms), Room 3 (RT60 = 98 ms), Reference, Room 1 (RT60 = 27 ms), and Room 2 (RT60 = 123 ms).](36f6f5bc2a35070ad6dc4312a93bcbe3_img.jpg) + +The figure consists of four subplots arranged in a 2x2 grid, comparing one-twelfth-octave spectra for car noise and mix signals across different rooms and processing states. + +- Top-left: Car noise – processed** +Y-axis: $L/\text{dB}[\text{V}]$ (ranging from -60 to -30). X-axis: $f/\text{Hz}$ (logarithmic scale from 50 to 10 k). The spectra for all rooms and the reference are nearly identical, peaking around 1000 Hz at approximately -42 dB[V]. +- Top-right: Car noise – unprocessed** +Y-axis: $L/\text{dB}[\text{Pa}]$ (ranging from -70 to -20). X-axis: $f/\text{Hz}$ (logarithmic scale from 50 to 10 k). The spectra show more variation between rooms, with Room 4 (RT60 = 264 ms) generally showing higher levels than the others. +- Bottom-left: Mix – processed** +Y-axis: $L/\text{dB}[\text{V}]$ (ranging from -60 to -30). X-axis: $f/\text{Hz}$ (logarithmic scale from 50 to 10 k). The spectra are very similar across all conditions, with a peak around 1000 Hz at approximately -42 dB[V]. +- Bottom-right: Mix – unprocessed** +Y-axis: $L/\text{dB}[\text{Pa}]$ (ranging from -70 to -20). X-axis: $f/\text{Hz}$ (logarithmic scale from 50 to 10 k). There is more variability here than in the processed mix, with Room 4 (RT60 = 264 ms) showing higher levels in the mid-frequency range. + +Common legend for all plots: + — Room 4 (RT60 = 264 ms) + — Room 3 (RT60 = 98 ms) + — Reference + — Room 1 (RT60 = 27 ms) + — Room 2 (RT60 = 123 ms) + +Four line graphs showing one-twelfth-octave spectra for car noise and mix signals in different rooms. The top-left graph shows 'Car noise - processed' with L/dB[V] on the y-axis. The top-right graph shows 'Car noise - unprocessed' with L/dB[Pa] on the y-axis. The bottom-left graph shows 'Mix - processed' with L/dB[V] on the y-axis. The bottom-right graph shows 'Mix - unprocessed' with L/dB[Pa] on the y-axis. All graphs have f/Hz on the x-axis (logarithmic scale from 50 to 10k). Each graph contains five data series: Room 4 (RT60 = 264 ms), Room 3 (RT60 = 98 ms), Reference, Room 1 (RT60 = 27 ms), and Room 2 (RT60 = 123 ms). + +**Figure III.11 – Phone 3, handset mode, comparison of one-twelfth-octave spectra for car noise and mix signal in different rooms (left: processed; right: unprocessed)** + +A further validation of the accuracy of the reproduction system when applied to a modern mobile phone is shown in Figures III.12, III.13 and III.14. The spectrograms of the same device tested in different rooms show a high degree of correlation, not only in the frequency domain, but also in the time domain. The different transient sounds, being part of the mix background noise, are well reproduced independently of the test room used. Note that the phone incorporates certain time-variant processing techniques, which in general may lead to small deviations in the result between tests. + +![Four spectrograms showing spectrum versus time analysis for phone 1. The top-left plot is labeled 'Reference'. The top-right plot is labeled 'Room 1 (27 ms)'. The bottom-left plot is labeled 'Room 2 (123 ms)'. The bottom-right plot is labeled 'Room 4 (264 ms)'. Each plot shows frequency (f/Hz) on the y-axis (logarithmic scale from 50 to 10k) versus time (t/s) on the x-axis (linear scale from 50 to 70). A color bar at the bottom of each plot indicates the level in L/dB[V] ranging from -70 to -45.](8d5d2817d93a53bb48da3a869f6aa149_img.jpg) + +Four spectrograms showing spectrum versus time analysis for phone 1. The top-left plot is labeled 'Reference'. The top-right plot is labeled 'Room 1 (27 ms)'. The bottom-left plot is labeled 'Room 2 (123 ms)'. The bottom-right plot is labeled 'Room 4 (264 ms)'. Each plot shows frequency (f/Hz) on the y-axis (logarithmic scale from 50 to 10k) versus time (t/s) on the x-axis (linear scale from 50 to 70). A color bar at the bottom of each plot indicates the level in L/dB[V] ranging from -70 to -45. + +Figure III.12 – Spectrum versus time analysis of the mix signal for three rooms for phone 1; reference recording at top left + +![Four spectrograms showing spectrum versus time analysis for phone 2. Top-left: Reference. Top-right: Room 1 (27 ms). Bottom-left: Room 2 (123 ms). Bottom-right: Room 3 (98 ms). Each plot shows frequency (f/Hz) from 50 to 10k on the y-axis and time (t/s) from 50 to 70 on the x-axis. A color scale at the bottom of each plot indicates power levels from -80 to -55 L/dB[V].](cd710dc9df8e7f47c4776e91bb425281_img.jpg) + +The figure displays four spectrograms arranged in a 2x2 grid, showing the spectrum versus time analysis of a mix signal for phone 2. The top-left plot is labeled 'Reference'. The top-right plot is labeled 'Room 1 (27 ms)'. The bottom-left plot is labeled 'Room 2 (123 ms)'. The bottom-right plot is labeled 'Room 3 (98 ms)'. Each plot has a y-axis labeled 'f/Hz' ranging from 50 to 10k, and an x-axis labeled 't/s' ranging from 50 to 70. A color scale at the bottom of each plot indicates power levels from -80 to -55 L/dB[V]. The spectrograms show various frequency components over time, with the reference signal showing a clear pattern of vertical lines, while the room signals show more complex patterns with additional horizontal and vertical features. + +Four spectrograms showing spectrum versus time analysis for phone 2. Top-left: Reference. Top-right: Room 1 (27 ms). Bottom-left: Room 2 (123 ms). Bottom-right: Room 3 (98 ms). Each plot shows frequency (f/Hz) from 50 to 10k on the y-axis and time (t/s) from 50 to 70 on the x-axis. A color scale at the bottom of each plot indicates power levels from -80 to -55 L/dB[V]. + +Figure III.13 – Spectrum versus time analysis of the mix signal for three rooms for phone 2; reference recording at top left + +![Figure III.14: Spectrum versus time analysis of the mix signal for three rooms for phone 3. The figure consists of four spectrograms arranged in a 2x2 grid. The top-left plot is labeled 'Reference' and shows a clear spectrogram with distinct horizontal bands of energy. The top-right plot is labeled 'Room 1 (27 ms)' and shows a similar spectrogram but with more background noise. The bottom-left plot is labeled 'Room 2 (123 ms)' and shows a spectrogram with significant reverberation and blurred energy bands. The bottom-right plot is labeled 'Room 4 (264 ms)' and shows a very noisy spectrogram with almost no discernible speech structure. Each plot has a frequency axis (f/Hz) from 50 to 10k on the right and a time axis (t/s) from 50 to 70 at the bottom. A color bar at the bottom of each plot indicates the level in L/dB[V] from -60 to -35.](a312ce7823f1d58ab6c3e86e6a7e9059_img.jpg) + +Figure III.14: Spectrum versus time analysis of the mix signal for three rooms for phone 3. The figure consists of four spectrograms arranged in a 2x2 grid. The top-left plot is labeled 'Reference' and shows a clear spectrogram with distinct horizontal bands of energy. The top-right plot is labeled 'Room 1 (27 ms)' and shows a similar spectrogram but with more background noise. The bottom-left plot is labeled 'Room 2 (123 ms)' and shows a spectrogram with significant reverberation and blurred energy bands. The bottom-right plot is labeled 'Room 4 (264 ms)' and shows a very noisy spectrogram with almost no discernible speech structure. Each plot has a frequency axis (f/Hz) from 50 to 10k on the right and a time axis (t/s) from 50 to 70 at the bottom. A color bar at the bottom of each plot indicates the level in L/dB[V] from -60 to -35. + +**Figure III.14 – Spectrum versus time analysis of the mix signal for three rooms for phone 3; reference recording at top left** + +#### III.9.3.5 Handset measurements I – MOS comparison + +Figure III.15 shows the mean opinion score (MOS) values measured in the simulation in relation to the MOS values measured in the original sound field for the handset case. High correlation between values determined in the reproduced sound field and original sound field can be observed. The correlation factors are given in Table III.4. + +![A 4x3 grid of scatter plots titled 'S-MOS, N-MOS, G-MOS Handset Wideband'. The columns are labeled 'G-MOS', 'N-MOS', and 'S-MOS'. The rows are labeled 'Room 1 (RT60=27ms)', 'Room 2 (RT60=123ms)', 'Room 3 (RT60=98ms)', and 'Room 4 (RT60=264ms)'. Each plot shows 'MOS-Value in Simulation' on the y-axis and 'MOS-Value in Reality' on the x-axis, both ranging from 1 to 5. A diagonal line represents perfect agreement. Data points are colored by phone: Phone1 (4.8\](ce88a251629679d7701df4d66e8e4e1e_img.jpg) + +A 4x3 grid of scatter plots titled 'S-MOS, N-MOS, G-MOS Handset Wideband'. The columns are labeled 'G-MOS', 'N-MOS', and 'S-MOS'. The rows are labeled 'Room 1 (RT60=27ms)', 'Room 2 (RT60=123ms)', 'Room 3 (RT60=98ms)', and 'Room 4 (RT60=264ms)'. Each plot shows 'MOS-Value in Simulation' on the y-axis and 'MOS-Value in Reality' on the x-axis, both ranging from 1 to 5. A diagonal line represents perfect agreement. Data points are colored by phone: Phone1 (4.8\ + +**Figure III.15 – Results of the S-MOS, N-MOS and G-MOS measurements for different phones, located in the handset position, in different rooms** + +**Table III.4 – Correlation of mean opinion score values for different phones in different rooms** + +| Name | G-MOS (%) | N-MOS (%) | S-MOS (%) | +|--------|-----------|-----------|-----------| +| Room 1 | 98.9 | 90.4 | 98.9 | +| Room 2 | 99.6 | 93.3 | 99.7 | +| Room 3 | 99.2 | 90.3 | 99.5 | +| Room 4 | 99.5 | 97.4 | 99.7 | + +#### III.9.3.6 Handset measurements II – spectral comparison + +Further experiments especially concentrating on the hands-free-mode were conducted in a second set of tests. For this purpose, two phones, numbered 4 and 5, both providing wideband transmission in hands-free mode, were used in addition to phone 1. Also, an additional new background noise was + +introduced. In the reference set-up, a cafeteria noise was played back by a loudspeaker located on the left side behind the HATS. This background noise is called "single speaker" in the following. This additional simulation was done in order to cover the situation of only one interfering sound source considered to be the most demanding for the simulation set-up. The average levels of this single speaker signal were 57.3 dBSPL(A) for the handset position and 58.1 dBSPL(A) for the hands-free position. Figure III.16 shows an example set-up for a handset measurement in room 2 (RT60: 123 ms). + +![Figure III.16: Example set-up for a handset measurement in room 2 (RT60: 123 ms). The image shows a grey HATS (Head and Torso Simulator) mannequin positioned on a stand. A black handset is mounted on the left side of the mannequin's head, facing forward. The background is a plain grey wall with a vertical wooden strip on the right side.](d4adfc6e910fa1c0c08db2ba4da84754_img.jpg) + +Figure III.16: Example set-up for a handset measurement in room 2 (RT60: 123 ms). The image shows a grey HATS (Head and Torso Simulator) mannequin positioned on a stand. A black handset is mounted on the left side of the mannequin's head, facing forward. The background is a plain grey wall with a vertical wooden strip on the right side. + +**Figure III.16 – Example set-up for a handset measurement in room 2 (RT60: 123 ms)** + +The results of these measurements are shown in Figures III.17 to III.19. In this experiment, only rooms 1, 2 and 4 were used for setting up the background noise field simulation. + +![Six line graphs comparing one-twelfth-octave spectra for car noise, mix, and single speaker in processed and unprocessed modes across different rooms (Room 1, Room 2, Room 4) and a reference. The graphs show sound level (L) in dB[V] or dB[Pa] versus frequency (f) in Hz.](b9745e4d501fcfe72053e7211ab39fed_img.jpg) + +The figure consists of six subplots arranged in a 3x2 grid, comparing the one-twelfth-octave spectra for different noise conditions and processing modes. The x-axis for all plots is frequency $f/\text{Hz}$ on a logarithmic scale from 50 to 10,000. The y-axis represents sound level $L$ in $\text{dB}[\text{V}]$ (left column) or $\text{dB}[\text{Pa}]$ (right column). + +Legend for all plots: + +- Room 1 ( $\text{RT}_{60} = 27 \text{ ms}$ ) +- Reference +- Room 4 ( $\text{RT}_{60} = 264 \text{ ms}$ ) +- Room 2 ( $\text{RT}_{60} = 123 \text{ ms}$ ) + +**Top Row: Car noise** + +- Left: Car noise – processed**: Y-axis from -80 to -50 $\text{dB}[\text{V}]$ . The spectra show a broad peak around 200-300 Hz, followed by a gradual decline. Room 4 (blue) shows a sharp drop-off above 5,000 Hz. +- Right: Car noise – unprocessed**: Y-axis from -70 to -10 $\text{dB}[\text{Pa}]$ . The spectra show a more pronounced decline with frequency, with Room 4 (blue) dropping off significantly after 2,000 Hz. + +**Middle Row: Mix** + +- Left: Mix – processed**: Y-axis from -70 to -40 $\text{dB}[\text{V}]$ . The spectra show multiple peaks and valleys, with Room 2 (cyan) and Room 4 (blue) showing more pronounced features than the others. +- Right: Mix – unprocessed**: Y-axis from -70 to -20 $\text{dB}[\text{Pa}]$ . The spectra show a general downward trend with frequency, with Room 4 (blue) showing a sharp drop-off above 2,000 Hz. + +**Bottom Row: Single speaker** + +- Left: Single speaker – processed**: Y-axis from -80 to -50 $\text{dB}[\text{V}]$ . The spectra show a broad peak around 200-300 Hz, followed by a gradual decline. Room 4 (blue) shows a sharp drop-off above 5,000 Hz. +- Right: Single speaker – unprocessed**: Y-axis from -70 to -30 $\text{dB}[\text{Pa}]$ . The spectra show a more pronounced decline with frequency, with Room 4 (blue) dropping off significantly after 2,000 Hz. + +Six line graphs comparing one-twelfth-octave spectra for car noise, mix, and single speaker in processed and unprocessed modes across different rooms (Room 1, Room 2, Room 4) and a reference. The graphs show sound level (L) in dB[V] or dB[Pa] versus frequency (f) in Hz. + +P.570(25) + +**Figure III.17 – Phone 1, handset mode, comparison of one-twelfth-octave spectra (left: processed; right: unprocessed)** + +![Six line graphs comparing one-twelfth-octave spectra for car noise, mix, and single speaker conditions. Each graph shows Level (L) in dB versus frequency (f) in Hz for four rooms: Room 1 (RT60 = 27 ms), Reference, Room 4 (RT60 = 264 ms), and Room 2 (RT60 = 123 ms). The left column shows processed results (L/dB[V]) and the right column shows unprocessed results (L/dB[Pa]).](accc4268081de4e491f669f7576054e2_img.jpg) + +The figure consists of six subplots arranged in a 3x2 grid, comparing the one-twelfth-octave spectra for different acoustic conditions. Each plot shows the sound level (L) in dB versus frequency (f) in Hz on a logarithmic scale from 50 Hz to 10 kHz. The legend for all plots is: Room 1 (RT60 = 27 ms) in green, Reference in red, Room 4 (RT60 = 264 ms) in blue, and Room 2 (RT60 = 123 ms) in cyan. + +- Top Left: Car noise – processed** (Y-axis: L/dB[V], -70 to -50). The spectra show a broad peak around 1 kHz, with the Reference and Room 1 curves being the highest, followed by Room 2, and Room 4 being the lowest. +- Top Right: Car noise – unprocessed** (Y-axis: L/dB[Pa], -70 to -10). The spectra show a similar trend to the processed version but with higher overall levels. +- Middle Left: Mix – processed** (Y-axis: L/dB[V], -80 to -40). The spectra show more complex features with multiple peaks and dips, particularly in the 1 kHz to 5 kHz range. +- Middle Right: Mix – unprocessed** (Y-axis: L/dB[Pa], -60 to -20). The spectra show similar complex features to the processed version but with higher overall levels. +- Bottom Left: Single speaker – processed** (Y-axis: L/dB[V], -80 to -50). The spectra show distinct peaks and dips, with the Reference and Room 1 curves being the highest, followed by Room 2, and Room 4 being the lowest. +- Bottom Right: Single speaker – unprocessed** (Y-axis: L/dB[Pa], -80 to -30). The spectra show similar features to the processed version but with higher overall levels. + +Six line graphs comparing one-twelfth-octave spectra for car noise, mix, and single speaker conditions. Each graph shows Level (L) in dB versus frequency (f) in Hz for four rooms: Room 1 (RT60 = 27 ms), Reference, Room 4 (RT60 = 264 ms), and Room 2 (RT60 = 123 ms). The left column shows processed results (L/dB[V]) and the right column shows unprocessed results (L/dB[Pa]). + +P.570(25) + +**Figure III.18 – Phone 5, handset mode, comparison of one-twelfth-octave spectra (left: processed; right: unprocessed)** + +![Six line graphs comparing one-twelfth-octave spectra for car noise, mix, and single speaker in processed and unprocessed states across different rooms (Room 1, Room 2, Room 4) and a reference. The graphs show sound level (L/dB) versus frequency (f/Hz) from 50 Hz to 10 kHz. The left column shows processed results in L/dB[V] and the right column shows unprocessed results in L/dB[Pa].](e1d063dc653d70930d6a3476d1d731cb_img.jpg) + +The figure consists of six subplots arranged in a 3x2 grid, comparing the one-twelfth-octave spectra for different acoustic scenarios. The left column displays processed results (L/dB[V]) and the right column displays unprocessed results (L/dB[Pa]). The x-axis for all plots is frequency (f/Hz) on a logarithmic scale from 50 to 10,000 Hz. The y-axis represents sound level (L/dB) on a linear scale. + +Legend for all plots: + +- Room 1 (RT60 = 27 ms) - Green line +- Reference - Red line +- Room 4 (RT60 = 264 ms) - Blue line +- Room 2 (RT60 = 123 ms) - Cyan line + +**Top Row: Car noise** + +- Left: Car noise – processed**: Y-axis from -80 to -50 dB[V]. The spectra show a broad peak around 100-200 Hz and a sharp rise starting around 2000 Hz, peaking near 5000 Hz. +- Right: Car noise – unprocessed**: Y-axis from -70 to -20 dB[Pa]. The spectra show a similar trend to the processed results but with higher overall levels. + +**Middle Row: Mix** + +- Left: Mix – processed**: Y-axis from -80 to -50 dB[V]. The spectra show more complex features with multiple peaks and dips across the frequency range. +- Right: Mix – unprocessed**: Y-axis from -60 to -20 dB[Pa]. The spectra show a similar complex pattern to the processed results. + +**Bottom Row: Single speaker** + +- Left: Single speaker – processed**: Y-axis from -80 to -60 dB[V]. The spectra show distinct peaks and dips, particularly in the 100-1000 Hz range. +- Right: Single speaker – unprocessed**: Y-axis from -70 to -30 dB[Pa]. The spectra show a similar pattern to the processed results. + +Six line graphs comparing one-twelfth-octave spectra for car noise, mix, and single speaker in processed and unprocessed states across different rooms (Room 1, Room 2, Room 4) and a reference. The graphs show sound level (L/dB) versus frequency (f/Hz) from 50 Hz to 10 kHz. The left column shows processed results in L/dB[V] and the right column shows unprocessed results in L/dB[Pa]. + +P.570(25) + +**Figure III.19 – Phone 4, handset mode, comparison of one-twelfth-octave spectra (left: processed; right: unprocessed)** + +As seen already in experiment I, a good correlation between reference and reproduced sound field can be observed. As expected, a somewhat higher deviation can be observed for the sound field consisting just of a single speaker, but the transmitted spectra of the noise when using the reproduction arrangement is quite similar to the original. It needs to be taken into account that all terminals are using highly sophisticated, non-linear and time variant signal processing unknown to the laboratory. Therefore, it can be assumed that some of the time variant behaviour can be attributed to the terminals. + +#### III.9.3.7 Handset measurements II – Mean opinion score comparison + +As already shown in Experiment I, the correlation of the S-MOS, N-MOS and G-MOS prediction compared to the real background noise situation is very high using the sound field simulation technique in the different rooms. See Table III.5. + +**Table III.5 – Correlation of mean opinion score values for different phones in different rooms** + +| Name | G-MOS (%) | N-MOS (%) | S-MOS (%) | +|--------|-----------|-----------|-----------| +| Room 1 | 97.5 | 99.3 | 95.4 | +| Room 2 | 98.0 | 99.4 | 98.8 | +| Room 4 | 99.2 | 99.3 | 97.9 | + +![Figure III.20: Results of S-MOS, N-MOS and G-MOS measurements for different phones, located in the handset position, in different rooms. The figure is a 3x3 grid of scatter plots. The columns are labeled G-MOS, N-MOS, and S-MOS. The rows are labeled Room 1 (RT60=27ms), Room 2 (RT60=123ms), and Room 4 (RT60=264ms). The x-axis for all plots is 'MOS-Value in Reality' and the y-axis is 'MOS-Value in Simulation', both ranging from 1 to 5. A diagonal line y=x is present in each plot. Data points are shown for three phones: Phone1 (4.8\](7f228001c8c028ac077fb5def89326e8_img.jpg) + +Figure III.20: Results of S-MOS, N-MOS and G-MOS measurements for different phones, located in the handset position, in different rooms. The figure is a 3x3 grid of scatter plots. The columns are labeled G-MOS, N-MOS, and S-MOS. The rows are labeled Room 1 (RT60=27ms), Room 2 (RT60=123ms), and Room 4 (RT60=264ms). The x-axis for all plots is 'MOS-Value in Reality' and the y-axis is 'MOS-Value in Simulation', both ranging from 1 to 5. A diagonal line y=x is present in each plot. Data points are shown for three phones: Phone1 (4.8\ + +**Figure III.20 – Results of S-MOS, N-MOS and G-MOS measurements for different phones, located in the handset position, in different rooms** + +#### III.9.3.8 Hands-free measurements – spectral comparison + +As described for the handset tests, an original sound field was created. However, instead of recording the sound field with the microphone array at the HATS position, the sound field was recorded at the hands-free position. This position was also used for sound field equalization in the different rooms. The experiment was conducted in three different rooms (numbered 1, 2 and 4). The three different background noises described previously were used. The sound field at the main hands-free microphone position, as well as the sending signal of the mobile phone in hands-free mode, were analysed. The results of the tests are shown in Figures III.22, III.23 and III.24. + +![A photograph showing a white robot head with a circular microphone array on its ear area, positioned in front of a wall covered in acoustic foam. To the right, a mobile phone is mounted on a stand, and a microphone array is visible in the foreground, partially obscuring the view.](b6609bd81fdcfe0d03eb93d51999c052_img.jpg) + +The image shows a white robot head, resembling a human head, positioned on the left side of the frame. It has a circular microphone array on its ear area. The robot is facing right. In the background, there is a wall covered with acoustic foam. To the right of the robot, a mobile phone is mounted on a stand. In the foreground, there is a microphone array, which is partially obscuring the view of the robot and the background. The setup is for hands-free measurements. + +A photograph showing a white robot head with a circular microphone array on its ear area, positioned in front of a wall covered in acoustic foam. To the right, a mobile phone is mounted on a stand, and a microphone array is visible in the foreground, partially obscuring the view. + +**Figure III.21 – Example of a hand-held hands-free set-measurement (the array is only shown to illustrate the set-up, it is not present during hands-free testing)** + +![Six spectral plots comparing processed and unprocessed audio for Car noise, Mix, and Single speaker across four rooms with different RT60 values.](d155a1511db57a83e3f7f7c6172e1939_img.jpg) + +The figure displays six spectral plots arranged in a 3x2 grid. The left column shows 'processed' signals (units in dB[V]) and the right column shows 'unprocessed' signals (units in dB[Pa]). The rows represent three scenarios: 'Car noise', 'Mix', and 'Single speaker'. Each plot graphs level (L/dB) against frequency (f/Hz) from 50 Hz to 10 kHz. Every plot contains four colored traces: green for Room 1 (RT60 = 27 ms), red for Reference, dark blue for Room 4 (RT60 = 264 ms), and light blue for Room 2 (RT60 = 123 ms). The 'Car noise' plots show a peak around 500-1000 Hz. The 'Mix' plots show more jagged spectral peaks across the mid-range. The 'Single speaker' plots show highly variable peaks and troughs characteristic of speech spectra. + +Six spectral plots comparing processed and unprocessed audio for Car noise, Mix, and Single speaker across four rooms with different RT60 values. + +P.570(25) + +**Figure III.22 – Phone 1, hands-free mode, comparison of one-twelfth-octave spectra + (left: processed; right: unprocessed)** + +![Six line graphs comparing one-twelfth-octave spectra for 'Car noise', 'Mix', and 'Single speaker' in processed and unprocessed modes across different rooms (Room 1, Room 2, Room 4) and a reference. The x-axis is frequency (f/Hz) from 50 to 10,000. The y-axis is level (L) in dB[V] or dB[Pa].](2cca454bbea734b85b2a785da04388f9_img.jpg) + +The figure consists of six subplots arranged in a 3x2 grid, comparing one-twelfth-octave spectra for different acoustic scenarios. The left column shows 'processed' results and the right column shows 'unprocessed' results. Each plot compares four conditions: Room 1 (RT60 = 27 ms), Reference, Room 4 (RT60 = 264 ms), and Room 2 (RT60 = 123 ms). + +- Top Left: Car noise – processed** +Y-axis: L/dB[V] (-70 to -30). The spectra for all conditions are very similar, showing a peak around 100 Hz and a general downward trend towards higher frequencies. +- Top Right: Car noise – unprocessed** +Y-axis: L/dB[Pa] (-70 to -10). The spectra show more variation between rooms, with Room 1 (green) being notably higher than the others in the mid-frequency range. +- Middle Left: Mix – processed** +Y-axis: L/dB[V] (-70 to -30). The processed results show good consistency across the different room conditions. +- Middle Right: Mix – unprocessed** +Y-axis: L/dB[Pa] (-70 to -20). There is more variability here, with Room 1 (green) showing higher levels than the reference and other rooms in the 1000-5000 Hz range. +- Bottom Left: Single speaker – processed** +Y-axis: L/dB[V] (-70 to -40). The spectra are relatively flat across the frequency range, with all conditions showing similar levels. +- Bottom Right: Single speaker – unprocessed** +Y-axis: L/dB[Pa] (-70 to -30). This plot shows significant differences between rooms. Room 1 (green) has much higher levels than the reference and other rooms across most of the frequency spectrum. + +Six line graphs comparing one-twelfth-octave spectra for 'Car noise', 'Mix', and 'Single speaker' in processed and unprocessed modes across different rooms (Room 1, Room 2, Room 4) and a reference. The x-axis is frequency (f/Hz) from 50 to 10,000. The y-axis is level (L) in dB[V] or dB[Pa]. + +P.570(25) + +**Figure III.23 – Phone 5, hands-free mode, comparison of one-twelfth-octave spectra (left: processed; right: unprocessed)** + +![Six line graphs comparing one-twelfth-octave spectra for car noise, mix, and single speaker scenarios. The left column shows 'processed' results in L/dB[V] and the right column shows 'unprocessed' results in L/dB[Pa]. Each graph plots sound level against frequency (f/Hz) from 50 to 10,000 Hz. Four data series are shown: Room 1 (RT60 = 27 ms) in green, Reference in red, Room 4 (RT60 = 264 ms) in blue, and Room 2 (RT60 = 123 ms) in cyan. The graphs show that the processed spectra are very similar to the reference, while the unprocessed spectra show more deviation, especially in the single speaker scenario.](b828d89aa6500ddee2da97f19ff86357_img.jpg) + +The figure consists of six subplots arranged in a 3x2 grid. The rows represent different noise scenarios: 'Car noise', 'Mix', and 'Single speaker'. The columns represent processing status: 'processed' (left) and 'unprocessed' (right). Each plot shows the sound pressure level (L) in dB against frequency (f) in Hz on a logarithmic scale from 50 to 10,000 Hz. The y-axis for the left column is L/dB[V] and for the right column is L/dB[Pa]. Each plot contains four data series: Room 1 (RT60 = 27 ms) in green, Reference in red, Room 4 (RT60 = 264 ms) in blue, and Room 2 (RT60 = 123 ms) in cyan. The processed plots show that the spectra for different rooms are very similar to the reference spectrum, while the unprocessed plots show more variation and deviation from the reference, particularly in the single speaker scenario. + +Six line graphs comparing one-twelfth-octave spectra for car noise, mix, and single speaker scenarios. The left column shows 'processed' results in L/dB[V] and the right column shows 'unprocessed' results in L/dB[Pa]. Each graph plots sound level against frequency (f/Hz) from 50 to 10,000 Hz. Four data series are shown: Room 1 (RT60 = 27 ms) in green, Reference in red, Room 4 (RT60 = 264 ms) in blue, and Room 2 (RT60 = 123 ms) in cyan. The graphs show that the processed spectra are very similar to the reference, while the unprocessed spectra show more deviation, especially in the single speaker scenario. + +P.570(25) + +**Figure III.24 – Phone 4, hands-free mode, comparison of one-twelfth-octave spectra (left: processed, right: unprocessed)** + +By comparing the different spectra of the transmitted background noise, determined for the two situations – when the hands-free device is exposed to the original sound field and the reproduced sound field – it can be seen that the spectra are very similar. As expected, the deviations observed for the single speaker (considered the worst-case scenario) are somewhat higher, but still very comparable to the recorded original sound field. + +#### III.9.3.9 Hands-free measurements – mean opinion score comparison + +As already shown in the handset experiments, the correlation of the S-MOS, N-MOS and G-MOS prediction compared to the real background noise situation is very high using the sound field simulation technique in the different rooms for the hands-free situation. See Table III.6 and Figure III.25. + +**Table III.6 – Correlation of mean opinion score values for different phones in hands-free mode** + +| Name | G-MOS (%) | N-MOS (%) | S-MOS (%) | +|--------|-----------|-----------|-----------| +| Room 1 | 98.2 | 99.7 | 98.2 | +| Room 2 | 99.1 | 99.1 | 98.7 | +| Room 4 | 97.7 | 99.3 | 96.5 | + +![Figure III.25: Results of the S-MOS, N-MOS AND G-MOS measurements for different phones located in the hands-free position, in different rooms. The figure is a 3x3 grid of scatter plots. The columns are labeled G-MOS, N-MOS, and S-MOS. The rows are labeled Room 1 (RT60=27ms), Room 2 (RT60=122ms), and Room 4 (RT60=264ms). Each plot shows MOS-Value in Simulation (y-axis) vs MOS-Value in Reality (x-axis), with a diagonal line representing perfect correlation. Data points are colored by phone: Phone1 (4.8\](5c1059b19e142ca5a6d58f18f3767b97_img.jpg) + +Figure III.25: Results of the S-MOS, N-MOS AND G-MOS measurements for different phones located in the hands-free position, in different rooms. The figure is a 3x3 grid of scatter plots. The columns are labeled G-MOS, N-MOS, and S-MOS. The rows are labeled Room 1 (RT60=27ms), Room 2 (RT60=122ms), and Room 4 (RT60=264ms). Each plot shows MOS-Value in Simulation (y-axis) vs MOS-Value in Reality (x-axis), with a diagonal line representing perfect correlation. Data points are colored by phone: Phone1 (4.8\ + +**Figure III.25 – Results of the S-MOS, N-MOS AND G-MOS measurements for different phones located in the hands-free position, in different rooms** + +#### III.9.4 Summary + +The results provided in this appendix show a high accuracy of the sound field reproduction technique described in this Recommendation for both, handset and hands-free reproduction set-ups using sophisticated modern mobile terminals employing the latest state of the art processing techniques. + +# Appendix IV + +## Spatial reproduction stability + +(This appendix does not form an integral part of this Recommendation.) + +### IV.1 Spatial reproduction accuracy of monophonic signals under anechoic conditions + +This is currently not covered in this Recommendation. + +### IV.2 Spatial reproduction accuracy of monophonic signals in treated rooms + +#### IV.2.1 Introduction + +This clause investigates the spatial stability of the reproduced noise field achieved using the microphone method (clause 8.3.2) in an acoustically treated room, whose dimensions were $3.5 \text{ m} \times 5.0 \text{ m} \times 2.7 \text{ m}$ . The room had been fitted with sound-absorbing materials to control the reverberation time, as well as standing waves, in the room. + +#### IV.2.2 Description of the experiment + +The loudspeakers were positioned in a square with a microphone at the centre (and in the same plane) as depicted in Figure IV.1 (around 2 m between the loudspeakers and the microphone). In this experiment, the subwoofer was not part of the set-up, thus results are valid from 120 Hz to 20 kHz. + +![Diagram of the experimental setup for loudspeaker equalization using the microphone method. A square arrangement of four loudspeakers (Left front, Right front, Left back, Right back) is shown, with a measurement microphone at the center. Each loudspeaker is connected to a generator via a series of blocks: a sine wave symbol, a triangle symbol, and a tau symbol. The distance from the center microphone to each loudspeaker is labeled as ~2m. The microphone is connected to an analyser. A red 'X' is placed over the left back loudspeaker. The diagram is labeled P.570(18)_FIV.1.](53f4e16be332aa0952e18cbb5259c31c_img.jpg) + +The diagram illustrates the experimental setup for loudspeaker equalization. On the left, a 'Generator' block contains five input channels labeled 'a' through 'e'. Each channel has a sine wave symbol, a triangle symbol, and a $\tau$ symbol. These channels are connected to four loudspeakers arranged in a square: 'Left front', 'Right front', 'Left back', and 'Right back'. A red 'X' is placed over the 'Left back' loudspeaker. A measurement microphone is positioned at the center of the square, indicated by a dashed line and a circle with a cross. The distance from the center to each loudspeaker is labeled as $\sim 2\text{m}$ . The microphone is connected to an 'Analyser' block at the bottom left. The diagram is labeled 'P.570(18)\_FIV.1'. + +Diagram of the experimental setup for loudspeaker equalization using the microphone method. A square arrangement of four loudspeakers (Left front, Right front, Left back, Right back) is shown, with a measurement microphone at the center. Each loudspeaker is connected to a generator via a series of blocks: a sine wave symbol, a triangle symbol, and a tau symbol. The distance from the center microphone to each loudspeaker is labeled as ~2m. The microphone is connected to an analyser. A red 'X' is placed over the left back loudspeaker. The diagram is labeled P.570(18)\_FIV.1. + +Figure IV.1 – Set-up for loudspeaker equalization using the microphone method + +In order to determine the spatial stability of the reproduced noise field, in a first step, several measurements were taken around the equalization centre (original position of the microphone for the equalization) without the HATS. These measurements (blue circles) were taken according to Figure IV.2. The points later occupied by the HATS were omitted. + +![A grid showing measurement points (blue circles) for a reproduced noise field. The grid is labeled with X (mm) on the vertical axis and Y (mm) on the horizontal axis. The origin (0,0) is marked with a red circle. Four speaker icons are positioned at the corners: Front left, Front right, Rear left, and Rear right. The grid extends from -150 to 150 mm on the Y-axis and -125 to 200 mm on the X-axis. A small label 'P.570(25)' is present near the bottom right speaker icon.](4e525151b89a2f87f31d869491a817d8_img.jpg) + +The figure displays a measurement grid for a reproduced noise field. The horizontal axis is labeled **Y (mm)** and ranges from -150 to 150 in increments of 25. The vertical axis is labeled **X (mm)** and ranges from -125 to 200 in increments of 25. Blue circles represent measurement points, which are mostly concentrated in the upper half of the grid (X > 0). A single red circle at the origin (0, 0) indicates the equalization centre. Four speaker icons are placed at the corners of the grid area, labeled **Front left**, **Front right**, **Rear left**, and **Rear right**. The label **P.570(25)** is located below the **Rear right** speaker icon. + +A grid showing measurement points (blue circles) for a reproduced noise field. The grid is labeled with X (mm) on the vertical axis and Y (mm) on the horizontal axis. The origin (0,0) is marked with a red circle. Four speaker icons are positioned at the corners: Front left, Front right, Rear left, and Rear right. The grid extends from -150 to 150 mm on the Y-axis and -125 to 200 mm on the X-axis. A small label 'P.570(25)' is present near the bottom right speaker icon. + +**Figure IV.2 – Measurement points (blue circles) for the reproduced noise field, without HATS; the measurement taken at the equalization centre is marked in red** + +In a second step, the microphone was replaced by the HATS and several measurements were taken around it. These measurements were taken according to Figure IV.3. + +![Figure IV.3: Measurement points (blue circles) for the reproduced noise field, with the HATS. The diagram shows a top-down view of a HATS (Head and Torso Simulator) model with a grid of measurement points. The grid is defined by X (mm) and Y (mm) axes. The X-axis ranges from -150 to 150 mm, and the Y-axis ranges from -125 to 200 mm. Measurement points are marked with blue circles at 25 mm intervals. Four loudspeakers are positioned around the HATS: Front left, Front right, Rear left, and Rear right. The HATS model is shown in grey, and the measurement points are blue circles. The text 'P.570(25)' is visible near the bottom right.](efedc464aaba06e85f269d26c98fee90_img.jpg) + +Figure IV.3: Measurement points (blue circles) for the reproduced noise field, with the HATS. The diagram shows a top-down view of a HATS (Head and Torso Simulator) model with a grid of measurement points. The grid is defined by X (mm) and Y (mm) axes. The X-axis ranges from -150 to 150 mm, and the Y-axis ranges from -125 to 200 mm. Measurement points are marked with blue circles at 25 mm intervals. Four loudspeakers are positioned around the HATS: Front left, Front right, Rear left, and Rear right. The HATS model is shown in grey, and the measurement points are blue circles. The text 'P.570(25)' is visible near the bottom right. + +**Figure IV.3 – Measurement points (blue circles) for the reproduced noise field, with the HATS** + +In both cases, the stability of the sound field in a rectangle of 325 mm × 300 mm around the equalization centre ( $X = 0$ mm; $Y = 0$ mm), which is also the HRP, was checked. The equalization centre corresponds to the original position where the equalization was performed; in Figure IV.2 this position is marked by a red circle. More measurements were taken in front of the HATS, because this area is usually of more interest than the area behind the HATS. Measurements were done every 25 mm on the $X$ - and $Y$ -axis (blue circles in Figures IV.2 and IV.3) where possible, leading to 149 measurements for each configuration. Note that no measurement could be done at position ( $X = -75$ mm; $Y = -50$ mm), because the handset positioner was set up on the left side of the HATS. + +The results presented in clause IV.2.3 are spatial frequency responses in one-third octave bands. The test signal that was used for all measurements was pink noise (high-pass filtered as required by this Recommendation) and its level was adjusted in order to get around 75 dBPa(A) at the equalization centre. The reference used for the calculation of frequency responses was the signal measured at the equalization centre [( $X = 0$ mm; $Y = 0$ mm) see Figure IV.2]. This signal was the reference for the calculation of the frequency responses for all grid points shown in Figures IV.2 and IV.3. + +#### IV.2.3 Results + +Clause 8.3.2 recommends adding fixed delays (0 ms, 11 ms, 17 ms, 29 ms) to individual loudspeakers in order to increase the diffusivity of the noise field. However different delays, i.e., 0 ms, 13 ms, 22 ms, 45 ms were applied, as, in this specific configuration (room and loudspeaker set-up), they gave better results than those recommended. + +The spatial noise field stability results for both cases (with and without HATS) are shown in Table IV.1 Each row corresponds to a one-third octave band. The results displayed start with the one-third octave band centred on 126 Hz and end with that centred on 19 953 Hz. + +The difference between these two configurations (centre minus left column results) is given in the right column. These additional results allow easier verification of the impact of placing the HATS in the noise field. + +Measurements were done every 25 mm on *X*- and *Y*-axis; however, in order to get better visual results, images were interpolated with a 1 mm step. + +**Table IV.1 – Spatial noise field stability results – left: without HATS; centre: with HATS; right: their differences** + +![Heatmap showing spatial stability of noise field at 126 Hz without HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow (high stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 126 Hz with HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow (high stability). A grey silhouette of a head is overlaid on the plot. The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing the difference in spatial stability of noise field at 126 Hz between with and without HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow (high stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB.](a05ad7b3a9ed890d8ef5f717f1b912b1_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|------------------------------------------------------|------------------------------------------------------|--------------------------------------------------------------------------| +|

Spatial stability of noise field at 126 Hz dB

|

Spatial stability of noise field at 126 Hz dB

|

Spatial stability difference, with and without HATS, at 126 Hz dB

| + +Heatmap showing spatial stability of noise field at 126 Hz without HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow (high stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 126 Hz with HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow (high stability). A grey silhouette of a head is overlaid on the plot. The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing the difference in spatial stability of noise field at 126 Hz between with and without HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow (high stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. + +![Heatmap showing spatial stability of noise field at 158 Hz without HATS. The Y-axis (top) ranges from -150 to 150 mm, and the X-axis (left) ranges from -100 to 200 mm. A color scale on the right ranges from -12 dB (blue) to 6 dB (red). A dark blue silhouette of a head and torso is centered in the plot, surrounded by a color gradient representing stability levels. Heatmap showing spatial stability of noise field at 158 Hz with HATS. Similar to the 'Without HATS' plot, but with a grey oval representing the HATS head positioned over the central blue silhouette. The color gradient shows the impact of the HATS on the noise field stability. Heatmap showing the difference in spatial stability at 158 Hz between 'With HATS' and 'Without HATS'. The color scale indicates the change in dB. The central silhouette is dark blue, indicating minimal difference in that region, while the surrounding areas show varying levels of change in stability. Heatmap showing spatial stability of noise field at 200 Hz without HATS. The axes and color scale are identical to the 158 Hz plots. The color gradient shows the spatial stability pattern at this higher frequency. Heatmap showing spatial stability of noise field at 200 Hz with HATS. Includes the grey HATS head oval. The plot illustrates how the HATS affects the noise field stability at 200 Hz compared to the 'Without HATS' condition. Heatmap showing the difference in spatial stability at 200 Hz between 'With HATS' and 'Without HATS'. The plot highlights the areas where the presence of the HATS causes the most significant changes in noise field stability at 200 Hz.](2fe41075ebec62bef1594e5ec1f2885c_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|----------------------------------------------------------------------------------|----------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 158 Hz dB

|

Spatial stability of noise field at 158 Hz dB

|

Spatial stability difference, with and without HATS, at 158 Hz dB

| +|

Spatial stability of noise field at 200 Hz dB

|

Spatial stability of noise field at 200 Hz dB

|

Spatial stability difference, with and without HATS, at 200 Hz dB

| + +Heatmap showing spatial stability of noise field at 158 Hz without HATS. The Y-axis (top) ranges from -150 to 150 mm, and the X-axis (left) ranges from -100 to 200 mm. A color scale on the right ranges from -12 dB (blue) to 6 dB (red). A dark blue silhouette of a head and torso is centered in the plot, surrounded by a color gradient representing stability levels. Heatmap showing spatial stability of noise field at 158 Hz with HATS. Similar to the 'Without HATS' plot, but with a grey oval representing the HATS head positioned over the central blue silhouette. The color gradient shows the impact of the HATS on the noise field stability. Heatmap showing the difference in spatial stability at 158 Hz between 'With HATS' and 'Without HATS'. The color scale indicates the change in dB. The central silhouette is dark blue, indicating minimal difference in that region, while the surrounding areas show varying levels of change in stability. Heatmap showing spatial stability of noise field at 200 Hz without HATS. The axes and color scale are identical to the 158 Hz plots. The color gradient shows the spatial stability pattern at this higher frequency. Heatmap showing spatial stability of noise field at 200 Hz with HATS. Includes the grey HATS head oval. The plot illustrates how the HATS affects the noise field stability at 200 Hz compared to the 'Without HATS' condition. Heatmap showing the difference in spatial stability at 200 Hz between 'With HATS' and 'Without HATS'. The plot highlights the areas where the presence of the HATS causes the most significant changes in noise field stability at 200 Hz. + +![Heatmap showing spatial stability of noise field at 251 Hz without HATS. The plot shows a central dark blue region (low stability) surrounded by orange/yellow regions (high stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 251 Hz with HATS. The plot includes a grey head silhouette. The stability pattern is similar to the 'Without HATS' case but slightly altered around the head. Axes and color bar are the same. Heatmap showing the spatial stability difference at 251 Hz. The difference is mostly yellow/orange, indicating a slight increase in stability with HATS in the central region. Axes and color bar are the same. Heatmap showing spatial stability of noise field at 316 Hz without HATS. Similar to the 251 Hz case, showing a central dark blue region. Axes and color bar are the same. Heatmap showing spatial stability of noise field at 316 Hz with HATS. Includes a grey head silhouette. The stability pattern is similar to the 'Without HATS' case. Axes and color bar are the same. Heatmap showing the spatial stability difference at 316 Hz. The difference is mostly yellow/orange, indicating a slight increase in stability with HATS in the central region. Axes and color bar are the same.](5dd117defe429c6c5620c4bb9ce20b0a_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|---------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 251 Hz    dB

|

Spatial stability of noise field at 251 Hz    dB

|

Spatial stability difference, with and without HATS, at 251 Hz    dB

| +|

Spatial stability of noise field at 316 Hz    dB

|

Spatial stability of noise field at 316 Hz    dB

|

Spatial stability difference, with and without HATS, at 316 Hz    dB

| + +Heatmap showing spatial stability of noise field at 251 Hz without HATS. The plot shows a central dark blue region (low stability) surrounded by orange/yellow regions (high stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 251 Hz with HATS. The plot includes a grey head silhouette. The stability pattern is similar to the 'Without HATS' case but slightly altered around the head. Axes and color bar are the same. Heatmap showing the spatial stability difference at 251 Hz. The difference is mostly yellow/orange, indicating a slight increase in stability with HATS in the central region. Axes and color bar are the same. Heatmap showing spatial stability of noise field at 316 Hz without HATS. Similar to the 251 Hz case, showing a central dark blue region. Axes and color bar are the same. Heatmap showing spatial stability of noise field at 316 Hz with HATS. Includes a grey head silhouette. The stability pattern is similar to the 'Without HATS' case. Axes and color bar are the same. Heatmap showing the spatial stability difference at 316 Hz. The difference is mostly yellow/orange, indicating a slight increase in stability with HATS in the central region. Axes and color bar are the same. + +![Heatmap showing spatial stability of noise field at 398 Hz without HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 398 Hz with HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing the spatial stability difference, with and without HATS, at 398 Hz. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 501 Hz without HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 501 Hz with HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing the spatial stability difference, with and without HATS, at 501 Hz. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB.](88ffa3d2537be96753fe9243edefd095_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 398 Hz dB

|

Spatial stability of noise field at 398 Hz dB

|

Spatial stability difference, with and without HATS, at 398 Hz dB

| +|

Spatial stability of noise field at 501 Hz dB

|

Spatial stability of noise field at 501 Hz dB

|

Spatial stability difference, with and without HATS, at 501 Hz dB

| + +Heatmap showing spatial stability of noise field at 398 Hz without HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 398 Hz with HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing the spatial stability difference, with and without HATS, at 398 Hz. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 501 Hz without HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 501 Hz with HATS. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing the spatial stability difference, with and without HATS, at 501 Hz. The plot shows a dark blue region (low stability) in the center, surrounded by yellow/orange regions (higher stability). The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. + +![Heatmap showing spatial stability of noise field at 631 Hz without HATS. The plot shows a central dark blue region (low stability) surrounded by yellow/orange regions (higher stability). Axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. Color scale ranges from -12 dB to 6 dB. Heatmap showing spatial stability of noise field at 631 Hz with HATS. The plot includes a grey HATS head model. The stability pattern is similar to the 'Without HATS' case but slightly altered around the head. Axes and color scale are the same. Heatmap showing the difference in spatial stability at 631 Hz between 'With HATS' and 'Without HATS'. The difference is mostly yellow/green, indicating minimal change. Axes and color scale are the same. Heatmap showing spatial stability of noise field at 794 Hz without HATS. Similar to the 631 Hz case, showing a central low-stability region. Axes and color scale are the same. Heatmap showing spatial stability of noise field at 794 Hz with HATS. Includes the grey HATS head model. Stability pattern is consistent with the 631 Hz 'With HATS' case. Axes and color scale are the same. Heatmap showing the difference in spatial stability at 794 Hz between 'With HATS' and 'Without HATS'. The difference is mostly yellow/green, indicating minimal change. Axes and color scale are the same.](618c67672da762d0498f228041be5a45_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 631 Hz dB

|

Spatial stability of noise field at 631 Hz dB

|

Spatial stability difference, with and without HATS, at 631 Hz dB

| +|

Spatial stability of noise field at 794 Hz dB

|

Spatial stability of noise field at 794 Hz dB

|

Spatial stability difference, with and without HATS, at 794 Hz dB

| + +Heatmap showing spatial stability of noise field at 631 Hz without HATS. The plot shows a central dark blue region (low stability) surrounded by yellow/orange regions (higher stability). Axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. Color scale ranges from -12 dB to 6 dB. Heatmap showing spatial stability of noise field at 631 Hz with HATS. The plot includes a grey HATS head model. The stability pattern is similar to the 'Without HATS' case but slightly altered around the head. Axes and color scale are the same. Heatmap showing the difference in spatial stability at 631 Hz between 'With HATS' and 'Without HATS'. The difference is mostly yellow/green, indicating minimal change. Axes and color scale are the same. Heatmap showing spatial stability of noise field at 794 Hz without HATS. Similar to the 631 Hz case, showing a central low-stability region. Axes and color scale are the same. Heatmap showing spatial stability of noise field at 794 Hz with HATS. Includes the grey HATS head model. Stability pattern is consistent with the 631 Hz 'With HATS' case. Axes and color scale are the same. Heatmap showing the difference in spatial stability at 794 Hz between 'With HATS' and 'Without HATS'. The difference is mostly yellow/green, indicating minimal change. Axes and color scale are the same. + +![Heatmap showing spatial stability of noise field at 1000 Hz without HATS. The plot shows a color gradient from blue (-12 dB) to red (6 dB) across a grid of positions from -100 to 200 mm on the X-axis and -150 to 150 mm on the Y-axis. A dark blue silhouette of a head is visible in the center. Heatmap showing spatial stability of noise field at 1000 Hz with HATS. Similar to the 'Without HATS' plot but with a grey oval representing the HATS (Head and Shoulders Transfer Function) overlaid on the head silhouette. Heatmap showing the difference in spatial stability of noise field at 1000 Hz between 'With HATS' and 'Without HATS'. The plot shows the change in dB across the same grid. Heatmap showing spatial stability of noise field at 1259 Hz without HATS. The plot shows a color gradient from blue (-12 dB) to red (6 dB) across a grid of positions. A dark blue silhouette of a head is visible in the center. Heatmap showing spatial stability of noise field at 1259 Hz with HATS. Similar to the 'Without HATS' plot but with a grey oval representing the HATS overlaid on the head silhouette. Heatmap showing the difference in spatial stability of noise field at 1259 Hz between 'With HATS' and 'Without HATS'. The plot shows the change in dB across the same grid.](46e78ee78a1eb29cc373cdb1d855f939_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|-------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 1000 Hz dB

|

Spatial stability of noise field at 1000 Hz dB

|

Spatial stability difference, with and without HATS, at 1000 Hz dB

| +|

Spatial stability of noise field at 1259 Hz dB

|

Spatial stability of noise field at 1259 Hz dB

|

Spatial stability difference, with and without HATS, at 1259 Hz dB

| + +Heatmap showing spatial stability of noise field at 1000 Hz without HATS. The plot shows a color gradient from blue (-12 dB) to red (6 dB) across a grid of positions from -100 to 200 mm on the X-axis and -150 to 150 mm on the Y-axis. A dark blue silhouette of a head is visible in the center. Heatmap showing spatial stability of noise field at 1000 Hz with HATS. Similar to the 'Without HATS' plot but with a grey oval representing the HATS (Head and Shoulders Transfer Function) overlaid on the head silhouette. Heatmap showing the difference in spatial stability of noise field at 1000 Hz between 'With HATS' and 'Without HATS'. The plot shows the change in dB across the same grid. Heatmap showing spatial stability of noise field at 1259 Hz without HATS. The plot shows a color gradient from blue (-12 dB) to red (6 dB) across a grid of positions. A dark blue silhouette of a head is visible in the center. Heatmap showing spatial stability of noise field at 1259 Hz with HATS. Similar to the 'Without HATS' plot but with a grey oval representing the HATS overlaid on the head silhouette. Heatmap showing the difference in spatial stability of noise field at 1259 Hz between 'With HATS' and 'Without HATS'. The plot shows the change in dB across the same grid. + +![Heatmap of spatial stability of noise field at 1585 Hz without HATS. The plot shows a color scale from -12 dB (dark blue) to 6 dB (dark red). The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A large dark blue rectangular region is centered around X=0, Y=0, indicating low stability in that area. Heatmap of spatial stability of noise field at 1585 Hz with HATS. This plot is similar to the 'Without HATS' version but includes a grey silhouette of a HATS head positioned over the central low-stability region. The surrounding noise field pattern remains largely consistent with the 'Without HATS' case. Heatmap showing the difference in spatial stability at 1585 Hz between the 'With HATS' and 'Without HATS' conditions. The color scale is the same. The plot is predominantly yellow and light green, indicating very small differences (near 0 dB) across most of the field, except for the central masked area. Heatmap of spatial stability of noise field at 1995 Hz without HATS. The pattern is similar to the 1585 Hz case, with a central dark blue region of low stability and higher stability (orange/red) towards the edges of the measured area. Heatmap of spatial stability of noise field at 1995 Hz with HATS. Includes the grey HATS head silhouette over the central region. The noise field pattern at this higher frequency shows slightly more variation than at 1585 Hz but maintains the same general structure. Heatmap showing the difference in spatial stability at 1995 Hz. Like the 1585 Hz difference plot, it shows minimal variation between the two conditions, with values mostly clustered around the 0 dB (yellow) mark outside the central masked region.](8d91c13d9c584af8129133d91ecb4bb9_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|-------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 1585 Hz dB

|

Spatial stability of noise field at 1585 Hz dB

|

Spatial stability difference, with and without HATS, at 1585 Hz dB

| +|

Spatial stability of noise field at 1995 Hz dB

|

Spatial stability of noise field at 1995 Hz dB

|

Spatial stability difference, with and without HATS, at 1995 Hz dB

| + +Heatmap of spatial stability of noise field at 1585 Hz without HATS. The plot shows a color scale from -12 dB (dark blue) to 6 dB (dark red). The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A large dark blue rectangular region is centered around X=0, Y=0, indicating low stability in that area. Heatmap of spatial stability of noise field at 1585 Hz with HATS. This plot is similar to the 'Without HATS' version but includes a grey silhouette of a HATS head positioned over the central low-stability region. The surrounding noise field pattern remains largely consistent with the 'Without HATS' case. Heatmap showing the difference in spatial stability at 1585 Hz between the 'With HATS' and 'Without HATS' conditions. The color scale is the same. The plot is predominantly yellow and light green, indicating very small differences (near 0 dB) across most of the field, except for the central masked area. Heatmap of spatial stability of noise field at 1995 Hz without HATS. The pattern is similar to the 1585 Hz case, with a central dark blue region of low stability and higher stability (orange/red) towards the edges of the measured area. Heatmap of spatial stability of noise field at 1995 Hz with HATS. Includes the grey HATS head silhouette over the central region. The noise field pattern at this higher frequency shows slightly more variation than at 1585 Hz but maintains the same general structure. Heatmap showing the difference in spatial stability at 1995 Hz. Like the 1585 Hz difference plot, it shows minimal variation between the two conditions, with values mostly clustered around the 0 dB (yellow) mark outside the central masked region. + +![Heatmap showing spatial stability of noise field at 2512 Hz without HATS. The plot shows a color scale from -12 dB (blue) to 6 dB (red) across a grid of positions on X and Y axes. A dark blue region indicates lower stability in the center. Heatmap showing spatial stability of noise field at 2512 Hz with HATS. The plot includes a grey HATS head model. The stability pattern is similar to the 'Without HATS' case but slightly altered around the head. Heatmap showing the difference in spatial stability at 2512 Hz between 'With HATS' and 'Without HATS'. The plot shows the change in dB across the grid. Heatmap showing spatial stability of noise field at 3162 Hz without HATS. Similar to the 2512 Hz case, showing a central region of lower stability. Heatmap showing spatial stability of noise field at 3162 Hz with HATS. Includes the grey HATS head model. The stability distribution is comparable to the 2512 Hz 'With HATS' case. Heatmap showing the difference in spatial stability at 3162 Hz between 'With HATS' and 'Without HATS'. The plot shows the change in dB across the grid.](a680643d7564c58bfaae20cb643d5da7_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|----------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 2512 Hz    dB

|

Spatial stability of noise field at 2512 Hz    dB

|

Spatial stability difference, with and without HATS, at 2512 Hz    dB

| +|

Spatial stability of noise field at 3162 Hz    dB

|

Spatial stability of noise field at 3162 Hz    dB

|

Spatial stability difference, with and without HATS, at 3162 Hz    dB

| + +Heatmap showing spatial stability of noise field at 2512 Hz without HATS. The plot shows a color scale from -12 dB (blue) to 6 dB (red) across a grid of positions on X and Y axes. A dark blue region indicates lower stability in the center. Heatmap showing spatial stability of noise field at 2512 Hz with HATS. The plot includes a grey HATS head model. The stability pattern is similar to the 'Without HATS' case but slightly altered around the head. Heatmap showing the difference in spatial stability at 2512 Hz between 'With HATS' and 'Without HATS'. The plot shows the change in dB across the grid. Heatmap showing spatial stability of noise field at 3162 Hz without HATS. Similar to the 2512 Hz case, showing a central region of lower stability. Heatmap showing spatial stability of noise field at 3162 Hz with HATS. Includes the grey HATS head model. The stability distribution is comparable to the 2512 Hz 'With HATS' case. Heatmap showing the difference in spatial stability at 3162 Hz between 'With HATS' and 'Without HATS'. The plot shows the change in dB across the grid. + +![Heatmap showing spatial stability of noise field at 3981 Hz without HATS. The plot shows a blue silhouette of a head on a yellow-orange background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 3981 Hz with HATS. The plot shows a grey silhouette of a head with HATS on a yellow-orange background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing the difference in spatial stability at 3981 Hz between 'With HATS' and 'Without HATS'. The plot shows a blue silhouette on a yellow background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 5012 Hz without HATS. The plot shows a blue silhouette of a head on a yellow-orange background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 5012 Hz with HATS. The plot shows a grey silhouette of a head with HATS on a yellow-orange background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing the difference in spatial stability at 5012 Hz between 'With HATS' and 'Without HATS'. The plot shows a blue silhouette on a yellow background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB.](7f6144f829587f68880ebd63c0d7e03d_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|----------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 3981 Hz    dB

|

Spatial stability of noise field at 3981 Hz    dB

|

Spatial stability difference, with and without HATS, at 3981 Hz    dB

| +|

Spatial stability of noise field at 5012 Hz    dB

|

Spatial stability of noise field at 5012 Hz    dB

|

Spatial stability difference, with and without HATS, at 5012 Hz    dB

| + +Heatmap showing spatial stability of noise field at 3981 Hz without HATS. The plot shows a blue silhouette of a head on a yellow-orange background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 3981 Hz with HATS. The plot shows a grey silhouette of a head with HATS on a yellow-orange background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing the difference in spatial stability at 3981 Hz between 'With HATS' and 'Without HATS'. The plot shows a blue silhouette on a yellow background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 5012 Hz without HATS. The plot shows a blue silhouette of a head on a yellow-orange background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 5012 Hz with HATS. The plot shows a grey silhouette of a head with HATS on a yellow-orange background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing the difference in spatial stability at 5012 Hz between 'With HATS' and 'Without HATS'. The plot shows a blue silhouette on a yellow background. The X-axis ranges from -100 to 200 mm, and the Y-axis ranges from -150 to 150 mm. A color bar on the right indicates values from -12 to 6 dB. + +![Heatmap showing spatial stability of noise field at 6310 Hz without HATS. The plot shows a blue region of low stability in the center, surrounded by orange and red areas of higher stability. The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 6310 Hz with HATS. The plot includes a grey head silhouette. The stability pattern is similar to the 'Without HATS' case. Axes and color bar are the same. Heatmap showing the difference in spatial stability at 6310 Hz between 'With HATS' and 'Without HATS'. The plot is mostly yellow, indicating minimal difference. Axes and color bar are the same. Heatmap showing spatial stability of noise field at 7943 Hz without HATS. Similar to the 6310 Hz case, with a central blue region of low stability. Axes and color bar are the same. Heatmap showing spatial stability of noise field at 7943 Hz with HATS. Includes grey head silhouette. Stability pattern is consistent with the 'Without HATS' case. Axes and color bar are the same. Heatmap showing the difference in spatial stability at 7943 Hz between 'With HATS' and 'Without HATS'. The plot is mostly yellow, indicating minimal difference. Axes and color bar are the same.](4227fcec3bcdf58025548f8fc54e7056_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|-----------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 6310 Hz     dB

|

Spatial stability of noise field at 6310 Hz     dB

|

Spatial stability difference, with and without HATS, at 6310 Hz     dB

| +|

Spatial stability of noise field at 7943 Hz     dB

|

Spatial stability of noise field at 7943 Hz     dB

|

Spatial stability difference, with and without HATS, at 7943 Hz     dB

| + +Heatmap showing spatial stability of noise field at 6310 Hz without HATS. The plot shows a blue region of low stability in the center, surrounded by orange and red areas of higher stability. The axes range from -150 to 150 mm on the Y-axis and -100 to 200 mm on the X-axis. A color bar on the right indicates values from -12 to 6 dB. Heatmap showing spatial stability of noise field at 6310 Hz with HATS. The plot includes a grey head silhouette. The stability pattern is similar to the 'Without HATS' case. Axes and color bar are the same. Heatmap showing the difference in spatial stability at 6310 Hz between 'With HATS' and 'Without HATS'. The plot is mostly yellow, indicating minimal difference. Axes and color bar are the same. Heatmap showing spatial stability of noise field at 7943 Hz without HATS. Similar to the 6310 Hz case, with a central blue region of low stability. Axes and color bar are the same. Heatmap showing spatial stability of noise field at 7943 Hz with HATS. Includes grey head silhouette. Stability pattern is consistent with the 'Without HATS' case. Axes and color bar are the same. Heatmap showing the difference in spatial stability at 7943 Hz between 'With HATS' and 'Without HATS'. The plot is mostly yellow, indicating minimal difference. Axes and color bar are the same. + +![Heatmap showing spatial stability of noise field at 10000 Hz without HATS. The plot shows a blue silhouette of a head on a yellow-orange background. The color scale ranges from -12 dB (dark blue) to 6 dB (dark red). The stability is highest (blue) around the head and decreases (yellow/orange) towards the edges. Heatmap showing spatial stability of noise field at 10000 Hz with HATS. Similar to the 'Without HATS' plot but with a grey oval representing the HATS (Head and Shoulders Transfer Function) overlaid on the head silhouette. The color scale and axes are the same. Heatmap showing the difference in spatial stability at 10000 Hz between 'With HATS' and 'Without HATS'. The plot shows a blue silhouette on a yellow background, indicating the change in stability. The color scale and axes are the same. Heatmap showing spatial stability of noise field at 12589 Hz without HATS. The plot shows a blue silhouette of a head on a yellow-orange background. The color scale ranges from -12 dB (dark blue) to 6 dB (dark red). The stability is highest (blue) around the head and decreases (yellow/orange) towards the edges. Heatmap showing spatial stability of noise field at 12589 Hz with HATS. Similar to the 'Without HATS' plot but with a grey oval representing the HATS overlaid on the head silhouette. The color scale and axes are the same. Heatmap showing the difference in spatial stability at 12589 Hz between 'With HATS' and 'Without HATS'. The plot shows a blue silhouette on a yellow background, indicating the change in stability. The color scale and axes are the same.](5ce8c700bb8e85f351d4018c4c272c97_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|--------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------|------------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 10000 Hz dB

|

Spatial stability of noise field at 10000 Hz dB

|

Spatial stability difference, with and without HATS, at 10000 Hz dB

| +|

Spatial stability of noise field at 12589 Hz dB

|

Spatial stability of noise field at 12589 Hz dB

|

Spatial stability difference, with and without HATS, at 12589 Hz dB

| + +Heatmap showing spatial stability of noise field at 10000 Hz without HATS. The plot shows a blue silhouette of a head on a yellow-orange background. The color scale ranges from -12 dB (dark blue) to 6 dB (dark red). The stability is highest (blue) around the head and decreases (yellow/orange) towards the edges. Heatmap showing spatial stability of noise field at 10000 Hz with HATS. Similar to the 'Without HATS' plot but with a grey oval representing the HATS (Head and Shoulders Transfer Function) overlaid on the head silhouette. The color scale and axes are the same. Heatmap showing the difference in spatial stability at 10000 Hz between 'With HATS' and 'Without HATS'. The plot shows a blue silhouette on a yellow background, indicating the change in stability. The color scale and axes are the same. Heatmap showing spatial stability of noise field at 12589 Hz without HATS. The plot shows a blue silhouette of a head on a yellow-orange background. The color scale ranges from -12 dB (dark blue) to 6 dB (dark red). The stability is highest (blue) around the head and decreases (yellow/orange) towards the edges. Heatmap showing spatial stability of noise field at 12589 Hz with HATS. Similar to the 'Without HATS' plot but with a grey oval representing the HATS overlaid on the head silhouette. The color scale and axes are the same. Heatmap showing the difference in spatial stability at 12589 Hz between 'With HATS' and 'Without HATS'. The plot shows a blue silhouette on a yellow background, indicating the change in stability. The color scale and axes are the same. + +![Heatmap showing spatial stability of noise field at 15849 Hz without HATS. The plot shows a central dark blue region of high stability surrounded by orange/yellow areas. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB. Heatmap showing spatial stability of noise field at 15849 Hz with HATS. A grey head silhouette is overlaid on the central blue region. Pattern is similar to the 'Without HATS' case. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB. Heatmap showing the difference in spatial stability at 15849 Hz. The plot is predominantly yellow/orange, indicating small differences. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB. Heatmap showing spatial stability of noise field at 19953 Hz without HATS. Similar pattern to 15849 Hz but with more fine-grained variations in the orange/yellow regions. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB. Heatmap showing spatial stability of noise field at 19953 Hz with HATS. Grey head silhouette overlaid on central blue region. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB. Heatmap showing the difference in spatial stability at 19953 Hz. Predominantly yellow/orange with some minor variations. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB.](3b5bef09e08b5c16e542a13b933ab9cd_img.jpg) + +| Without HATS | With HATS | Difference, with and without HATS | +|-----------------------------------------------------------------------------------------|-----------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------| +|

Spatial stability of noise field at 15849 Hz    dB

|

Spatial stability of noise field at 15849 Hz    dB

|

Spatial stability difference, with and without HATS, at 15849 Hz    dB

| +|

Spatial stability of noise field at 19953 Hz    dB

|

Spatial stability of noise field at 19953 Hz    dB

|

Spatial stability difference, with and without HATS, at 19953 Hz    dB

| + +Heatmap showing spatial stability of noise field at 15849 Hz without HATS. The plot shows a central dark blue region of high stability surrounded by orange/yellow areas. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB. Heatmap showing spatial stability of noise field at 15849 Hz with HATS. A grey head silhouette is overlaid on the central blue region. Pattern is similar to the 'Without HATS' case. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB. Heatmap showing the difference in spatial stability at 15849 Hz. The plot is predominantly yellow/orange, indicating small differences. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB. Heatmap showing spatial stability of noise field at 19953 Hz without HATS. Similar pattern to 15849 Hz but with more fine-grained variations in the orange/yellow regions. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB. Heatmap showing spatial stability of noise field at 19953 Hz with HATS. Grey head silhouette overlaid on central blue region. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB. Heatmap showing the difference in spatial stability at 19953 Hz. Predominantly yellow/orange with some minor variations. X-axis: -100 to 200 mm, Y-axis: -150 to 150 mm. Color scale: -12 to 6 dB. + +For the full frequency range (120 Hz to 20 kHz) similar level variations were observed with and without HATS. The variation was around 9 dB (from -3 dB to 6 dB) for the case without HATS and around 11 dB (from -4 dB to 7 dB) for the case with HATS. + +Without the HATS, the noise field is generally stable but still exhibits some variations. See Table IV.2. With the HATS present, results are similar up to 501 Hz but start to diverge for higher frequencies. This can be explained by the presence of the HATS in the noise field which creates some perturbations. However, results from right column indicate that these perturbations remain limited as the range of the difference (with HATS minus without HATS) is contained in [-4 dB; 5 dB]. + +**Table IV.2 – Sound field uniformity without and with HATS** + +| Frequency range | Microphone method
(without HATS) | Microphone method
(with HATS) | Likely cause | +|---------------------|-------------------------------------|----------------------------------|-----------------------| +| Low (120 Hz-400 Hz) | -2 dB to 6 dB | -3 dB to 7 dB | Room modes | +| Mid (500 Hz-3 kHz) | -3 dB to 5 dB | -4 dB to 7 dB | Reflections from HATS | +| High (4 kHz-20 kHz) | 0 dB to 5 dB | -2 dB to 6 dB | Reflections from HATS | + +Furthermore, no influence of the HATS positioner (located on the left side) on the noise field, was observed because the results rather symmetrical (relative to the *X*-axis). + +### IV.3 Spatial Reproduction Accuracy of stereophonic signals in anechoic conditions + +This is currently not covered in this Recommendation. + +### IV.4 Spatial Reproduction Accuracy of stereophonic signals in treated rooms + +This is currently not covered in this Recommendation. + +### IV.5 Spatial Reproduction Accuracy of binaural signals in anechoic conditions + +This is currently not covered in this Recommendation. + +### IV.6 Spatial Reproduction Accuracy of binaural signals in treated rooms + +#### IV.6.1 Introduction + +This clause investigates the spatial stability of the reproduced noise field achieved using the HATS method in clause 8.3.1 in an acoustically treated room. The room dimension was 3.5 m × 5.0 m × 2.7 m and had been equipped with sound absorbing materials to control the reverberation time as well as standing waves in the room. + +#### IV.6.2 Description of the experiment + +The positioning of the loudspeakers was a square and HATS at the centre as described in Figure IV.4 (around 2 m between the loudspeakers and the HRP). In this experiment, the subwoofer was not part of the set-up, thus results are valid from 120 Hz to 20 kHz. + +![Diagram of the HATS method setup for loudspeaker equalization. A 'Generator' block on the left has five channels (a-e), each with a sine wave symbol, a gain symbol, and a delay symbol. Channel 'c' is crossed out with a red 'X'. Channels 'a' and 'b' are connected to 'Left front' and 'Right front' loudspeakers respectively. Channels 'd' and 'e' are connected to 'Left back' and 'Right back' loudspeakers. A central 'HATS' head model is shown with four loudspeakers at its corners. Dashed lines from the head's ears to the loudspeakers are labeled '~2m'. An 'Analyser' block at the bottom receives signals from '(HATS R ear)' and '(HATS L ear)'. The label 'P.570(18)_FIV.4' is in the bottom right.](f2a6c46fa05d5588ce4c6e1a3ee194c7_img.jpg) + +Diagram of the HATS method setup for loudspeaker equalization. A 'Generator' block on the left has five channels (a-e), each with a sine wave symbol, a gain symbol, and a delay symbol. Channel 'c' is crossed out with a red 'X'. Channels 'a' and 'b' are connected to 'Left front' and 'Right front' loudspeakers respectively. Channels 'd' and 'e' are connected to 'Left back' and 'Right back' loudspeakers. A central 'HATS' head model is shown with four loudspeakers at its corners. Dashed lines from the head's ears to the loudspeakers are labeled '~2m'. An 'Analyser' block at the bottom receives signals from '(HATS R ear)' and '(HATS L ear)'. The label 'P.570(18)\_FIV.4' is in the bottom right. + +**Figure IV.4 – Set-up for loudspeaker equalization using the HATS method** + +In order to determine the spatial stability of the reproduced noise field, several measurements around the head of the HATS were realized. These measurements (blue circles) were done in the plane of the mouth according to Figure IV.5. + +![Diagram showing measurement points (blue circles) for the reproduced noise field in the plane of the HATS mouth. A central grey circle represents the HATS head. A grid of blue circles is overlaid on the head, with X (mm) on the horizontal axis (from -150 to 150) and Y (mm) on the vertical axis (from -125 to 200). Four loudspeakers are positioned at the corners: 'Front left', 'Front right', 'Rear left', and 'Rear right'. The label 'P.570(25)' is in the bottom right.](fe8840a79cf6b30f8b2284064f3d2a36_img.jpg) + +Diagram showing measurement points (blue circles) for the reproduced noise field in the plane of the HATS mouth. A central grey circle represents the HATS head. A grid of blue circles is overlaid on the head, with X (mm) on the horizontal axis (from -150 to 150) and Y (mm) on the vertical axis (from -125 to 200). Four loudspeakers are positioned at the corners: 'Front left', 'Front right', 'Rear left', and 'Rear right'. The label 'P.570(25)' is in the bottom right. + +**Figure IV.5 – Measurement points (blue circles) for the reproduced noise field, in the plane of the HATS mouth** + +The stability of the sound field in a rectangle of 325 mm × 300 mm around the HRP ( $X = 0$ mm; $Y = 0$ mm) was checked. More measurements were taken in front of the HATS because this area is usually of more interest than the area behind the HATS. Measurements were done every 25 mm on the $X$ - and $Y$ -axis (blue circles in Figure IV.5) where possible, leading to 149 measurements for the full plane. It can be noticed that no measurement could be done at position ( $X = -75$ mm; $Y = -50$ mm) because the handset positioner was set up on the left side of the HATS. + +The results presented in clause IV.6.3 are spatial frequency responses in one-third octave bands. The test signal that was used for all measurements was binaural pub noise (Pub\_Noise\_binaural\_V2 described in clause 8.1 of [ETSI EG 202 396-1] and high-pass filtered as required by this Recommendation). The reference used for the calculation of frequency responses was the signal measured at the position ( $X = 125$ mm; $Y = 0$ mm) Figure IV.5. This signal was the reference for the calculation of the frequency responses for all points of the grid defined in Figure IV.5. + +#### IV.6.3 Results + +Clause 8.3.1 recommends adding fixed delays (0 ms, 11 ms, 17 ms, 29 ms) to individual loudspeakers in order to increase the diffusivity of the noise field. However different delays, i.e., 0 ms, 13 ms, 22 ms, 45 ms were applied, as, in this specific configuration (room and loudspeaker set-up), they gave better results than those recommended. + +The spatial noise field results are listed in Table IV.3. Each cell corresponds to a one-third octave band. The results displayed start with the one-third octave band centred on 126 Hz and end with that centred on 19 953 Hz. + +Measurements were done every 25 mm on the $X$ - and $Y$ -axis; however, in order to get better visual results, images were interpolated with a 1 mm step. + +**Table IV.3 – Sound field uniformity with a head and torso simulator** + +![Heatmap showing spatial stability of noise field at 126 Hz. The plot shows a head and torso simulator (HATS) in the center. The color scale ranges from -12 dB (blue) to 6 dB (red). The field is relatively uniform, mostly yellow/orange, indicating stability within ±2 dB. Heatmap showing spatial stability of noise field at 158 Hz. Similar to the 126 Hz plot, but with more pronounced gradients, especially on the left side (green/blue), indicating less uniformity at this higher frequency.](0c54c3b831bfdbbb0e00adbbb0e65391_img.jpg) + +| HATS method | | +|--------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------| +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 126 Hz dB

|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 158 Hz dB

| + +Heatmap showing spatial stability of noise field at 126 Hz. The plot shows a head and torso simulator (HATS) in the center. The color scale ranges from -12 dB (blue) to 6 dB (red). The field is relatively uniform, mostly yellow/orange, indicating stability within ±2 dB. Heatmap showing spatial stability of noise field at 158 Hz. Similar to the 126 Hz plot, but with more pronounced gradients, especially on the left side (green/blue), indicating less uniformity at this higher frequency. + +![Heatmap showing spatial stability of noise field at 200 Hz. The plot shows a head silhouette with a color scale from -12 dB (blue) to 6 dB (red). The stability is relatively uniform across the field. Heatmap showing spatial stability of noise field at 251 Hz. Similar to the 200 Hz plot, showing a head silhouette and a color scale from -12 dB to 6 dB. Heatmap showing spatial stability of noise field at 316 Hz. The plot shows a head silhouette and a color scale from -12 dB to 6 dB. Heatmap showing spatial stability of noise field at 398 Hz. The plot shows a head silhouette and a color scale from -12 dB to 6 dB. Heatmap showing spatial stability of noise field at 501 Hz. The plot shows a head silhouette and a color scale from -12 dB to 6 dB. Heatmap showing spatial stability of noise field at 631 Hz. The plot shows a head silhouette and a color scale from -12 dB to 6 dB.](2543ad3dd579d09cfec98c9849226c43_img.jpg) + +| HATS method | | +|--------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------| +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 200 Hz dB

|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 251 Hz dB

| +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 316 Hz dB

|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 398 Hz dB

| +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 501 Hz dB

|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 631 Hz dB

| + +Heatmap showing spatial stability of noise field at 200 Hz. The plot shows a head silhouette with a color scale from -12 dB (blue) to 6 dB (red). The stability is relatively uniform across the field. Heatmap showing spatial stability of noise field at 251 Hz. Similar to the 200 Hz plot, showing a head silhouette and a color scale from -12 dB to 6 dB. Heatmap showing spatial stability of noise field at 316 Hz. The plot shows a head silhouette and a color scale from -12 dB to 6 dB. Heatmap showing spatial stability of noise field at 398 Hz. The plot shows a head silhouette and a color scale from -12 dB to 6 dB. Heatmap showing spatial stability of noise field at 501 Hz. The plot shows a head silhouette and a color scale from -12 dB to 6 dB. Heatmap showing spatial stability of noise field at 631 Hz. The plot shows a head silhouette and a color scale from -12 dB to 6 dB. + +![Heatmap showing spatial stability of noise field at 794 Hz. The plot shows a HATS head silhouette on a grid from -100 to 200 mm on the X-axis and -150 to 150 mm on the Y-axis. The color scale ranges from -12 dB (blue) to 6 dB (red). The field is relatively stable (yellow/green) around the head. Heatmap showing spatial stability of noise field at 1000 Hz. Similar to the 794 Hz plot, showing stability around the HATS head silhouette. The field is mostly yellow, indicating good stability. Heatmap showing spatial stability of noise field at 1259 Hz. The field shows more variation (more red and blue areas) compared to lower frequencies, indicating decreased stability. Heatmap showing spatial stability of noise field at 1585 Hz. The field is predominantly yellow, indicating stable conditions across the measured area. Heatmap showing spatial stability of noise field at 1995 Hz. The field is mostly yellow with some green areas, showing acceptable stability. Heatmap showing spatial stability of noise field at 2512 Hz. The field shows significant variation with large red and blue regions, indicating poor spatial stability at this higher frequency.](0f4879d336bad8b1801dee947a7dc617_img.jpg) + +| HATS method | | +|---------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------| +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 794 Hz dB

|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 1000 Hz dB

| +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 1259 Hz dB

|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 1585 Hz dB

| +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 1995 Hz dB

|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 2512 Hz dB

| + +Heatmap showing spatial stability of noise field at 794 Hz. The plot shows a HATS head silhouette on a grid from -100 to 200 mm on the X-axis and -150 to 150 mm on the Y-axis. The color scale ranges from -12 dB (blue) to 6 dB (red). The field is relatively stable (yellow/green) around the head. Heatmap showing spatial stability of noise field at 1000 Hz. Similar to the 794 Hz plot, showing stability around the HATS head silhouette. The field is mostly yellow, indicating good stability. Heatmap showing spatial stability of noise field at 1259 Hz. The field shows more variation (more red and blue areas) compared to lower frequencies, indicating decreased stability. Heatmap showing spatial stability of noise field at 1585 Hz. The field is predominantly yellow, indicating stable conditions across the measured area. Heatmap showing spatial stability of noise field at 1995 Hz. The field is mostly yellow with some green areas, showing acceptable stability. Heatmap showing spatial stability of noise field at 2512 Hz. The field shows significant variation with large red and blue regions, indicating poor spatial stability at this higher frequency. + +![Heatmap showing spatial stability of noise field at 3162 Hz. The plot shows a head silhouette with a color scale from -12 dB (blue) to 6 dB (red). The stability is relatively high (yellow/orange) around the head. Heatmap showing spatial stability of noise field at 3981 Hz. Similar to the 3162 Hz plot, showing high stability around the head silhouette. Heatmap showing spatial stability of noise field at 5012 Hz. The stability appears more uniform across the area compared to lower frequencies. Heatmap showing spatial stability of noise field at 6310 Hz. The stability is high and relatively uniform around the head. Heatmap showing spatial stability of noise field at 7943 Hz. There is more variation in stability, with some areas showing lower values (green/cyan). Heatmap showing spatial stability of noise field at 10000 Hz. The stability is high around the head but shows more variation than at lower frequencies.](41b5ef075822191f9a9b9430c34de2dc_img.jpg) + +| HATS method | | +|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +|

Position on Y axis (mm)
-150 -100 -50 0 50 100 150

Position on X axis (mm)
200 150 100 50 0 -50 -100

Spatial stability of noise field at 3162 Hz     dB

|

Position on Y axis (mm)
-150 -100 -50 0 50 100 150

Position on X axis (mm)
200 150 100 50 0 -50 -100

Spatial stability of noise field at 3981 Hz     dB

| +|

Position on Y axis (mm)
-150 -100 -50 0 50 100 150

Position on X axis (mm)
200 150 100 50 0 -50 -100

Spatial stability of noise field at 5012 Hz     dB

|

Position on Y axis (mm)
-150 -100 -50 0 50 100 150

Position on X axis (mm)
200 150 100 50 0 -50 -100

Spatial stability of noise field at 6310 Hz     dB

| +|

Position on Y axis (mm)
-150 -100 -50 0 50 100 150

Position on X axis (mm)
200 150 100 50 0 -50 -100

Spatial stability of noise field at 7943 Hz     dB

|

Position on Y axis (mm)
-150 -100 -50 0 50 100 150

Position on X axis (mm)
200 150 100 50 0 -50 -100

Spatial stability of noise field at 10000 Hz     dB

| + +Heatmap showing spatial stability of noise field at 3162 Hz. The plot shows a head silhouette with a color scale from -12 dB (blue) to 6 dB (red). The stability is relatively high (yellow/orange) around the head. Heatmap showing spatial stability of noise field at 3981 Hz. Similar to the 3162 Hz plot, showing high stability around the head silhouette. Heatmap showing spatial stability of noise field at 5012 Hz. The stability appears more uniform across the area compared to lower frequencies. Heatmap showing spatial stability of noise field at 6310 Hz. The stability is high and relatively uniform around the head. Heatmap showing spatial stability of noise field at 7943 Hz. There is more variation in stability, with some areas showing lower values (green/cyan). Heatmap showing spatial stability of noise field at 10000 Hz. The stability is high around the head but shows more variation than at lower frequencies. + +![Heatmap showing spatial stability of noise field at 12589 Hz. The plot shows a head and torso simulator (HATS) in the center. The color scale ranges from -12 dB (blue) to 6 dB (red). The field is relatively uniform around the simulator. Heatmap showing spatial stability of noise field at 15849 Hz. Similar to the 12589 Hz plot, but with more pronounced variations in the noise field around the HATS simulator. Heatmap showing spatial stability of noise field at 19953 Hz. The field shows significant asymmetry and variation, particularly on the left side of the simulator.](6eafd5abfb13fb29e6d819d2a1a3f6b2_img.jpg) + +| HATS method | | | +|----------------------------------------------------------------------------------------------------------------------|--|--| +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 12589 Hz dB

| | | +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 15849 Hz dB

| | | +|

Position on Y axis (mm)

Position on X axis (mm)

Spatial stability of noise field at 19953 Hz dB

| | | + +Heatmap showing spatial stability of noise field at 12589 Hz. The plot shows a head and torso simulator (HATS) in the center. The color scale ranges from -12 dB (blue) to 6 dB (red). The field is relatively uniform around the simulator. Heatmap showing spatial stability of noise field at 15849 Hz. Similar to the 12589 Hz plot, but with more pronounced variations in the noise field around the HATS simulator. Heatmap showing spatial stability of noise field at 19953 Hz. The field shows significant asymmetry and variation, particularly on the left side of the simulator. + +For the full frequency range (120 Hz to 20 kHz) a level variation of around 15 dB (from –8 dB to 7 dB) was observed. + +The noise field is quite stable for frequencies up to 1 000 Hz and then starts to get less stable for higher frequencies. This can be explained by the presence of the HATS in the noise field, which creates some perturbations. See Table IV.4 + +**Table IV.4 – Sound field uniformity with a head and torso simulator** + +| Frequency range | HATS method | Likely cause | +|---------------------|---------------|-----------------------| +| Low (120 Hz-400 Hz) | –5 dB to 3 dB | Room modes | +| Mid (500 Hz-3 kHz) | –5 dB to 6 dB | Reflections from HATS | +| High (4 kHz-20 kHz) | –8 dB to 7 dB | Reflections from HATS | + +Note that for frequencies higher than 6 310 Hz the noise field is not symmetrical (relative to the X-axis), which could be explained by the presence of the HATS positioned on the left side. + +### IV.7 Spatial reproduction accuracy sound field signals in anechoic conditions + +This is currently not covered in this Recommendation. + +### IV.8 Spatial reproduction accuracy of sound field signals in treated rooms + +This is currently not covered in this Recommendation. + +### IV.9 Spatial reproduction accuracy of multi-point reproduction + +To verify that the multi-point equalization procedure works for different rooms, it was applied to two different types: + +- a semi-anechoic room with an RT60 of about 140 ms; +- a room without special acoustical treatment, with an RT60 of 310 ms. + +In Figure IV.6, the differences in the microphones levels between the simulated and the real sound fields at the fine-tuning microphone positions are shown after equalization was done. This equalization was successful because in both rooms, the error, especially the mean error, is within $\pm 3$ dB. + +If the error is outside tolerance, it is associated with a microphone not equalized for level in this frequency range. + +![Figure IV.6: Two line graphs comparing sound field differences in a semi-anechoic room and a reverberant room. The x-axis is frequency (f/Hz) from 20 to 20,000. The y-axis is level difference (L/dB) from -6 to 6. Both graphs show multiple colored lines representing individual microphone errors and a thicker black line representing the mean error. Shaded regions indicate the ±3 dB tolerance limits. In the semi-anechoic room, the mean error stays within the tolerance. In the reverberant room, the mean error also stays within the tolerance, though with more fluctuations. A small label 'P.570(25)' is at the bottom right.](1326773262ea98ffa5f525b048dc04e3_img.jpg) + +P.570(25) + +Figure IV.6: Two line graphs comparing sound field differences in a semi-anechoic room and a reverberant room. The x-axis is frequency (f/Hz) from 20 to 20,000. The y-axis is level difference (L/dB) from -6 to 6. Both graphs show multiple colored lines representing individual microphone errors and a thicker black line representing the mean error. Shaded regions indicate the ±3 dB tolerance limits. In the semi-anechoic room, the mean error stays within the tolerance. In the reverberant room, the mean error also stays within the tolerance, though with more fluctuations. A small label 'P.570(25)' is at the bottom right. + +**Figure IV.6 – Differences between real and simulated sound field after equalization at fine tuning microphone positions. The wider curve represents the mean error. The violation of the tolerance at 1 250 Hz is caused by a microphone that is not equalized in this frequency range** + +As another analysis, the coherence between the microphone signals in the real sound field and in the reproduced sound field was calculated. Again, the microphone positions at the fine-tuning positions were compared. + +The results show that there is a strong correlation between both sound fields up to 3 kHz. Magnitude and phase show a good correspondence. See Figure IV.7. + +![Figure IV.7: Coherence between microphone signals at the fine-tuning set microphone positions. The figure consists of two side-by-side plots. The left plot is titled 'Semi-anechoic room' and the right plot is titled 'Reverberant room'. Both plots show coherence magnitude (solid line) and phase (dashed line) versus frequency (f/Hz) on a logarithmic scale from 50 to 5,000 Hz. The left y-axis represents phase (Phi/°) from -180 to 180, and the right y-axis represents magnitude (Pa*Pa) from 0 to 1. In both rooms, the magnitude is high (around 0.8-1.0) from 100 Hz to about 2,000 Hz, then drops sharply. The phase is near 0° across the frequency range.](1c356032f49001c090473e76823759c4_img.jpg) + +Figure IV.7: Coherence between microphone signals at the fine-tuning set microphone positions. The figure consists of two side-by-side plots. The left plot is titled 'Semi-anechoic room' and the right plot is titled 'Reverberant room'. Both plots show coherence magnitude (solid line) and phase (dashed line) versus frequency (f/Hz) on a logarithmic scale from 50 to 5,000 Hz. The left y-axis represents phase (Phi/°) from -180 to 180, and the right y-axis represents magnitude (Pa\*Pa) from 0 to 1. In both rooms, the magnitude is high (around 0.8-1.0) from 100 Hz to about 2,000 Hz, then drops sharply. The phase is near 0° across the frequency range. + +P.570(25) + +**Figure IV.7 – Coherence between microphone signals at the fine-tuning set microphone positions – solid line: magnitude; dashed line: phase** + +The final evaluation was done with a comparison criterion based on spherical correlation. A spherical microphone array was used to compare a real sound field to a simulated sound field in the spherical harmonics domain. The results are shown in Figure IV.8. + +![Figure IV.8: Spherical correlation of original and recorded fields. The figure consists of two side-by-side plots. The left plot shows sound pressure (p/Pa) versus frequency (f/Hz) on a logarithmic scale from 20 to 2,000 Hz. The right plot shows sound pressure per degree (p/Degree) versus frequency (f/Hz) on the same scale. Both plots compare 'Reverberant' (green line) and 'Semi-anechoic' (red line) fields. In the left plot, the semi-anechoic field is higher than the reverberant field above 500 Hz. In the right plot, both fields show high variability, with the semi-anechoic field generally having higher peaks.](8aef7252a3c6e5078b68c083fbc8bf3e_img.jpg) + +Figure IV.8: Spherical correlation of original and recorded fields. The figure consists of two side-by-side plots. The left plot shows sound pressure (p/Pa) versus frequency (f/Hz) on a logarithmic scale from 20 to 2,000 Hz. The right plot shows sound pressure per degree (p/Degree) versus frequency (f/Hz) on the same scale. Both plots compare 'Reverberant' (green line) and 'Semi-anechoic' (red line) fields. In the left plot, the semi-anechoic field is higher than the reverberant field above 500 Hz. In the right plot, both fields show high variability, with the semi-anechoic field generally having higher peaks. + +P.570(25) + +**Figure IV.8 – Spherical correlation of original and recorded fields** + +First of all, it can be seen from the results of the error and coherence analyses that both rooms exhibit approximately the same performance. A first view suggests that the limiting frequency is around 500 Hz, which would be rather low. However, bear in mind that the spherical microphone array used had a diameter of 25 cm with an even microphone distribution. In contrast, the microphone array used for equalization had a non-equal microphone distribution with a higher number of microphones on the right hand of the HATS where the handset terminal is placed. As a consequence, the result of the spherical correlation analysis leads to a limiting frequency in a sphere of 25 cm diameter around the microphone array with the HATS of about 500 Hz. However, the accuracy of the sound field reproduction on the right side of the HATS in the area where the handset is located is much higher, as seen in the coherence analysis shown in Figure IV.8. In addition to the correct sound field reproduction in this area, the sound pressure at the different microphones is correct up to 20 kHz, which cannot be guaranteed with ambisonics. + +# Appendix V + +## **Detailed description of the overall equalization procedure for a HATS** + +(This appendix does not form an integral part of this Recommendation.) + +If the result of the equalization is outside tolerance, it needs to be checked whether one or more loudspeakers contribute to the violation. In general, some experimental work is required to do some root cause analysis. + +Some examples of how to perform this are as follows. + +- Check the individual contribution of each loudspeaker to the overall response. The need to introduce high equalization gain, to have many equalizers and to have a high Q-factor in parametric equalizers are typically indications of poorly positioned loudspeakers. +- Reposition the loudspeakers to attempt to find a spot where less equalization is needed. +- Change the delays introduced into the loudspeaker-paths to find an optimum. + +# Bibliography + +- [b-ITU-T P.10] Recommendation ITU-T P.10/G.100 (2017), *Vocabulary for performance, quality of service and quality of experience*. +- [b-ETSI TS 103 106] ETSI TS 103 106, V1.5.1 (2018), *Speech and multimedia transmission quality (STQ); Speech quality performance in the presence of background noise: Background noise transmission for mobile terminals-objective test methods*. +- [b-Alexander] Alexander, R.C. (1999). *The inventor of stereo: The life and works of Alan Dower Blumlein*. Boston, MA: Focal Press. 421 pp. +- [b-Carter] Carter, C.G. (1987). Coherence and time-delay estimation. *Proc. IEEE* **75**(6), pp. 236-255. +- [b-Daniel] Daniel, J. (2000), [Représentation de champs acoustiques. application à la transmission et à la reproduction de scènes sonores complexes dans un contexte multimédia](https://www.researchgate.net/publication/242669363_Representation_de_champs_acoustiques_application_a_la_transmission_et_a_la_reproduction_de_scenes_sonores_complexes_dans_un_contexte_multimedia) [Representation of acoustic fields, application to the transmission and reproduction of complex sound scenes in a multimedia context], PhD thesis, 1996-2000 Université Paris 6. +[https://www.researchgate.net/publication/242669363\\_Representation\\_de\\_champs\\_acoustiques\\_application\\_a\\_la\\_transmission\\_et\\_a\\_la\\_reproduction\\_de\\_scenes\\_sonores\\_complexes\\_dans\\_un\\_contexte\\_multimedia](https://www.researchgate.net/publication/242669363_Representation_de_champs_acoustiques_application_a_la_transmission_et_a_la_reproduction_de_scenes_sonores_complexes_dans_un_contexte_multimedia) +- [b-Gerzon, 1973] Gerzon, M.A. (1973). Periphony: With-height sound reproduction. *J. Audio Eng. Soc.*, **21**(1), pp.2-10. +- [b-Gerzon, 1975] Gerzon, M.A. (1975). The design of precisely coincident microphone arrays for stereo and surround sound. Paper L-20 presented at the *50th AES Convention*, London. +- [b-Hoth] Hoth, D.F. (1941). Room noise spectra at subscribers' telephone locations. *J. Acoust. Soc. Am.*, **12**, pp. 99-504 +- [b-Kirkeby, 1996] Kirkeby, O., Nelson, P. A., Orduna-Bustamante, F., Hamada, H. (1996). Local sound field reproduction using digital signal processing. *J. Acoust. Soc. Am.* **100**(3), pp.1584-1593. +- [b-Kirkeby, 1998] Kirkeby, O., Nelson, P. A., Hamada, H., Orduna-Bustamante, F. (1998), Fast deconvolution of multichannel systems using regularization. *IEEE Transactions on speech and audio processing*, **6**(2), pp.189-194. +- [b-Malham] Malham, D.G. (1998). Spatial hearing mechanisms and sound reproduction, Technical report, University of York. +[https://web.archive.org/web/20180603185342/https://www.york.ac.uk/inst/mustech/3d\\_audio/ambis2.htm](https://web.archive.org/web/20180603185342/https://www.york.ac.uk/inst/mustech/3d_audio/ambis2.htm) +- [b-Müller] Müller, S; Massarani, P. (2001). Transfer function measurement with sweeps. *J. Audio Eng. Soc.*, **49**(6), pp. 443-471. +- [b-Priestley] Priestley, M.B., (1981). *Spectral analysis and time series*. 2 vols. London: Academic Press. 890 pp. +- [b-Rumsey] Rumsey, F. (2001). *Spatial audio*. Oxford: Focal. 240 pp. + + + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | Tariff and accounting principles and international telecommunication/ICT economic and policy issues | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling, and associated measurements and tests | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects, next-generation networks, Internet of Things and smart cities | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.61-198811-I_PDF-E/raw.md b/marked/P/T-REC-P.61-198811-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..4cd9327e3015a2988665d61e5765af1bf39bbe00 --- /dev/null +++ b/marked/P/T-REC-P.61-198811-I_PDF-E/raw.md @@ -0,0 +1,64 @@ + + +![ITU logo](2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg) + +The logo of the International Telecommunication Union (ITU) features a globe with a lightning bolt superimposed on it, and the letters 'ITU' in a bold, sans-serif font. + +ITU logo + +INTERNATIONAL TELECOMMUNICATION UNION + +**ITU-T** + +**P.61** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +# **TELEPHONE TRANSMISSION QUALITY OBJECTIVE ELECTRO-ACOUSTICAL MEASUREMENTS** --- + +### **METHODS FOR THE CALIBRATION OF CONDENSER MICROPHONES** + +**ITU-T Recommendation P.61** + +(Extract from the *Blue Book*) + +--- + +## NOTES + +1 ITU-T Recommendation P.61 was published in Volume V of the *Blue Book*. This file is an extract from the *Blue Book*. While the presentation and layout of the text might be slightly different from the *Blue Book* version, the contents of the file are identical to the *Blue Book* version and copyright conditions remain unchanged (see below). + +2 In this Recommendation, the expression “Administration” is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +### **METHODS FOR THE CALIBRATION OF CONDENSER MICROPHONES** + +*(amended at Malaga-Torremolinos, 1984)* + +Primary and secondary calibrations of condenser microphones can be carried out using the methods described below. + +#### **1 Primary calibration by the reciprocity method** + +The recommended procedure for primary calibration of condenser microphones is the reciprocity calibration technique. A precision method for reciprocity pressure calibration is described in [1]. A simplified method, suitable for calibration over the frequency range of interest for telephonometric measurements, is given in [2]. Although the methods described are specifically for one-inch microphones, similar methods are applicable to half-inch microphones. Methods suitable for half-inch microphones are under study by IEC. + +A precision method for free-field reciprocity calibration is given in [3]. Alternatively, the free-field correction curves given in [4] may be applied to the pressure calibration of one-inch condenser microphones to determine their free-field responses. The reciprocity free-field calibration method may in principle be extended to half-inch microphones. Free-field correction curves have not been standardized for half-inch microphones. + +#### **2 Secondary calibration by the comparison method** + +The secondary calibration of a condenser microphone may be achieved by direct comparison with a physically identical microphone having a known calibration. The procedure used is a modification of the “two microphones and auxiliary sound source” method described in [1] to [3]. The output of the calibrated microphone is first determined for a given drive level applied to the auxiliary sound source. The calibrated microphone is then replaced by the microphone to be calibrated, and its output is determined for the same drive level applied to the auxiliary sound source. The difference in level (in dB) between the outputs of the two microphones is then applied to the known calibration of the first microphone to determine the calibration of the second. The procedure is repeated at each frequency of interest. + +#### **3 Secondary calibration using pistonphones and other sound level calibrators** + +Secondary calibrations can also be made using pistonphones and other sound level calibrators which produce a known sound level. Such devices are often used to check the calibration of a microphone at a single frequency1). Care should be taken to follow the manufacturer's instructions when using such devices; in particular, it may be necessary to apply corrections for barometric pressure, coupler volume, microphone type, etc. Standardization of these calibrators is currently under study by the IEC. + +## **References** + +- [1] International Electrotechnical Commission *Precision method for pressure calibration of one-inch standard condenser microphones by the reciprocity technique*, IEC publication 327, Geneva, 1971. +- [2] International Electrotechnical Commission *Simplified method for pressure calibration of one-inch condenser microphones by the reciprocity technique*, IEC publication 402, Geneva, 1972. +- [3] International Electrotechnical Commission *Precision method for free-field calibration of one-inch standard condenser microphones by the reciprocity technique*, IEC publication 486, Geneva, 1974. +- [4] International Electrotechnical Commission *Values for the difference between free-field and pressure sensitivity levels for one-inch standard condenser microphones*, IEC Publication 655, Geneva, 1979. + +--- + +1) Calibrations with an accuracy of $\pm 0.3$ dB are possible. \ No newline at end of file diff --git a/marked/P/T-REC-P.700-202106-I_PDF-E/raw.md b/marked/P/T-REC-P.700-202106-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..9d402f24082122c74329858cc37c6352df9ac93f --- /dev/null +++ b/marked/P/T-REC-P.700-202106-I_PDF-E/raw.md @@ -0,0 +1,1827 @@ + + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.700** + +(06/2021) + +SERIES P: TELEPHONE TRANSMISSION QUALITY, +TELEPHONE INSTALLATIONS, LOCAL LINE +NETWORKS + +Measurements related to speech loudness + +# --- **Calculation of loudness for speech communication** + +Recommendation ITU-T P.700 + +## ITU-T P-SERIES RECOMMENDATIONS + +## TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS + +| | | +|----------------------------------------------------------------------------------------------------|--------------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | P.10–P.19 | +| Voice terminal characteristics | P.30–P.39 | +| Reference systems | P.40–P.49 | +| Objective measuring apparatus | P.50–P.59 | +| Objective electro-acoustical measurements | P.60–P.69 | +| Measurements related to speech loudness | P.70–P.79 | +| Methods for objective and subjective assessment of speech quality | P.80–P.89 | +| Voice terminal characteristics | P.300–P.399 | +| Objective measuring apparatus | P.500–P.599 | +| Measurements related to speech loudness | P.700–P.709 | +| Methods for objective and subjective assessment of speech and video quality | P.800–P.899 | +| Audiovisual quality in multimedia services | P.900–P.999 | +| Transmission performance and QoS aspects of IP end-points | P.1000–P.1099 | +| Communications involving vehicles | P.1100–P.1199 | +| Models and tools for quality assessment of streamed media | P.1200–P.1299 | +| Telemeeting assessment | P.1300–P.1399 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | P.1400–P.1499 | +| Methods for objective and subjective assessment of quality of services other than speech and video | P.1500–P.1599 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +## Recommendation ITU-T P.700 + +# Calculation of loudness for speech communication + +## Summary + +Recommendation ITU-T P.700 describes a unified method required for calculating loudness, allowing comparison of narrowband (300-3.4 kHz), wideband (100-8 kHz), super-wideband (50-14 kHz) and fullband (10-20 kHz) telephony for all types of terminals including handset, hands-free and conference terminals. + +The model described in this Recommendation is consistent when switching from one bandwidth to another and independent of the listening situation (e.g., handset, headset, hands-free) with regards to producing a constant perceived loudness. + +Compared with loudness rating models, such as the one presented in Recommendation ITU-T P.79, the present method predicts the *absolute* loudness, considers auditory masking and is applicable to a wide range of acoustic levels. + +This Recommendation incorporates an annex and a number of appendices that contain test vectors for the validation of loudness model implementations and descriptions and results of the loudness experiments that form the basis for this Recommendation. + +## History + +| Edition | Recommendation | Approval | Study Group | Unique ID* | +|---------|----------------|------------|-------------|---------------------------------------------------------------------------| +| 1.0 | ITU-T P.700 | 2019-06-29 | 12 | 11.1002/1000/13931 | +| 2.0 | ITU-T P.700 | 2021-06-13 | 12 | 11.1002/1000/14664 | + +## Keywords + +Loudness, speech, telephony. + +--- + +\* To access the Recommendation, type the URL in the address field of your web browser, followed by the Recommendation's unique ID. For example, . + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents/software copyrights, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the appropriate ITU-T databases available via the ITU-T website at . + +© ITU 2021 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## Table of Contents + +###### Page + +| | | | +|------|------------------------------------------------------------------------------|----| +| 1 | Scope..... | 1 | +| 2 | References..... | 1 | +| 3 | Definitions ..... | 2 | +| 3.1 | Terms defined elsewhere ..... | 2 | +| 3.2 | Terms defined in this Recommendation..... | 2 | +| 4 | Abbreviations and acronyms ..... | 2 | +| 5 | Conventions ..... | 3 | +| 6 | Existing loudness model structures ..... | 3 | +| 6.1 | Zwicker model for stationary sound..... | 4 | +| 6.2 | Moore and Glasberg model for stationary sound ..... | 4 | +| 6.3 | Zwicker and Fastl model for non-stationary sound ..... | 4 | +| 6.4 | Glasberg and Moore for non-stationary sound ..... | 4 | +| 7 | Reasons for a new universal loudness model ..... | 5 | +| 8 | Loudness model for receiving loudness ..... | 5 | +| 8.1 | Recording of speech signals ..... | 5 | +| 8.2 | Loudness calculation ..... | 6 | +| 8.3 | Application for terminals..... | 7 | +| 9 | Loudness model for sending loudness ..... | 8 | +| 10 | Loudness model for sidetone loudness ..... | 9 | +| 11 | Loudness model for listener sidetone loudness ..... | 9 | +| | Annex A – Validation and test vectors for loudness model implementation..... | 10 | +| A.1 | Introduction to test vectors for validation ..... | 10 | +| A.2 | Validation of loudness model implementation..... | 10 | +| A.3 | Validation of speech pause detection and loudness model implementation .. | 11 | +| | Appendix I – Result of loudness experiment A ..... | 13 | +| I.1 | Introduction to experiment A ..... | 13 | +| I.2 | Subjective experiment ..... | 13 | +| I.3 | Comparison results ..... | 17 | +| I.4 | Conclusion..... | 22 | +| | Appendix II – Result of loudness experiment B ..... | 24 | +| II.1 | Introduction to loudness experiment B..... | 24 | +| II.2 | Measurement set-up..... | 24 | +| II.3 | Auditory loudness assessment..... | 28 | +| II.4 | Instrumental loudness methods ..... | 30 | +| II.5 | Auditory results ..... | 32 | +| II.6 | Direct loudness model output..... | 36 | +| II.7 | Conclusions ..... | 39 | + +| | Page | +|---------------------------------------------------------------------|-------------| +| Appendix III – Result of loudness experiment C..... | 41 | +| III.1    Introduction to loudness experiment C..... | 41 | +| III.2    Subjective experiment ..... | 44 | +| III.3    Results and comparison to ISO 532-1 objective results..... | 47 | +| III.4    Conclusion..... | 51 | +| Appendix IV – Nominal transmission paths ..... | 52 | +| IV.1    Nominal receive paths ..... | 52 | +| Bibliography..... | 54 | + +# Introduction + +The loudness of speech in communication systems is one of the main parameters relevant for good user experience in telephone calls. A loudness model predicts the loudness of speech perceived by the user, independent of the listening situation and independent of the type of terminal used. In contrast to loudness rating as defined in Recommendation ITU-T P.79, the loudness described in Recommendation ITU-T P.700 focuses on the absolute loudness of speech and not on the attenuation or amplification introduced by the various components involved in the transmission of speech from mouth to ear. Therefore, loudness does not replace loudness rating, but is used in a complementary manner to describe perceptually perceived loudness. The loudness model has been validated on a range of conditions that include various types of signal processing used in terminals for which loudness rating calculations have not been validated. + + + +## Recommendation ITU-T P.700 + +# Calculation of loudness for speech communication + +# 1 Scope + +This Recommendation describes a loudness model which applies for all telephony audio bandwidths from narrowband, wideband, super-wideband (SWB) to fullband, and for all types of terminals, including handset, headset, hands-free and conference terminals. + +It is foreseen that the current ITU-T loudness rating model, as described in [ITU-T P.79], will be used in parallel with the loudness model given in this Recommendation for an extended period of time, since it is anticipated that network planning will continue to have a preference towards loudness rating calculations according to [ITU-T P.79]. + +# 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a standalone document, the status of a Recommendation. + +- [ITU-T G.160] Recommendation ITU-T G.160 (2012), *Voice enhancement devices*. +- [ITU-T G.191] Recommendation ITU-T G.191 (2019), *Software tools for speech and audio coding standardization*. +- [ITU-T P.48] Recommendation ITU-T P.48 (1988), *Specification for an intermediate reference system*. +- [ITU-T P.56] Recommendation ITU-T P.56 (2011), *Objective measurement of active speech level*. +- [ITU-T P.57] Recommendation ITU-T P.57 (2021), *Artificial ears*. +- [ITU-T P.58] Recommendation ITU-T P.58 (2021), *Head and torso simulator for telephonometry*. +- [ITU-T P.64] Recommendation ITU-T P.64 (2019), *Determination of sensitivity/frequency characteristics of local telephone systems*. +- [ITU-T P.76] Recommendation ITU-T P.76 (1988), *Determination of loudness ratings; fundamental principles*. +- [ITU-T P.78] Recommendation ITU-T P.78 (1996), *Subjective testing method for determination of loudness ratings in accordance with Recommendation P.76*. +- [ITU-T P.79] Recommendation ITU-T P.79 (2007), *Calculation of loudness ratings for telephone sets*. +- [ITU-T P.341] Recommendation ITU-T P.341 (2011), *Transmission characteristics for wideband digital loudspeaking and hands-free telephony terminals*. +- [ITU-T P.381] Recommendation ITU-T P.381 (2020), *Technical requirements and test methods for the universal wired headset or headphone interface of digital mobile terminals*. + +- [ITU-T P.382] Recommendation ITU-T P.382 (2020), *Technical requirements and test methods for multi-microphone wired headset or headphone interfaces of digital wireless terminals*. +- [ITU-T P.501] Recommendation ITU-T P.501 (2020), *Test signals for use in telephony and other speech-based applications*. +- [ITU-T P.581] Recommendation ITU-T P.581 (2014), *Use of head and torso simulator for hands-free and handset terminal testing*. +- [ITU-T P.830] Recommendation ITU-T P.380 (1996), *Subjective performance assessment of telephone-band and wideband digital codecs*. +- [ITU-T P.1401] Recommendation ITU-T P.1401 (2020), *Methods, metrics and procedures for statistical evaluation, qualification and comparison of objective quality prediction models*. + +# 3 Definitions + +### 3.1 Terms defined elsewhere + +This Recommendation uses the following terms defined elsewhere: + +**3.1.1 loudness** [b-ITU-T P.10]: Loudness belongs to a category of intensity sensations. Loudness is that attribute of auditory sensation in terms of which sounds can be ordered on a scale extending from quiet to loud. Loudness takes into account the spectral and temporal sensitivity of the human ear. Generally, masking effects in time and frequency are taken into account. The loudness level measure according to [b-Zwicker 3] was created to characterize the loudness sensation of tones. The loudness calculation procedure for stationary signals is defined in [b-ISO 532]. For the calculation of the loudness of time variant signal, different models are known. + +### 3.2 Terms defined in this Recommendation + +None. + +## 4 Abbreviations and acronyms + +This Recommendation uses the following abbreviations and acronyms: + +| | | +|-------|----------------------------------------------| +| AGC | Automatic Gain Control | +| ANOVA | Analysis of Variance | +| DECT | Digital Enhanced Cordless Telecommunications | +| DRP | Eardrum Reference Point | +| ERB | Equivalent Rectangular Bandwidth | +| ERL | Equal Reference Level | +| ERP | Ear Reference Point | +| FIR | Finite Impulse Response | +| HATS | Head and Torso Simulator | +| IRS | Intermediate Reference System | +| JLR | Junction Loudness Rating | +| LTL | Long-Term Loudness | +| MOS | Mean Opinion Score | + +| | | +|--------|----------------------------------------------------------------------------------| +| MRP | Mouth Reference Point | +| NOSFER | Nouveau Système Fondamental Européen Référence pour la transmission téléphonique | +| OLR | Overall Loudness Rating | +| POI | Point of Interconnection | +| RLR | Receive Loudness Rating | +| RMSE | Root Mean Square Error | +| SLR | Sending Loudness Rating | +| SNR | Signal-to-Noise Ratio | +| SPL | Sound Pressure Level | +| STL | Short-Term Loudness | +| SWB | Super-Wideband | +| VoIP | Voice Over IP | + +# 5 Conventions + +This Recommendation uses the following conventions: + +phon: A unit used to express loudness, where the phon scale is equal to scale of dBSPL for a pure tone of 1000 Hz. + +P.Loudness: The term P.Loudness is used as an abbreviation for the loudness prediction model that is defined in this Recommendation. This term is used extensively throughout the appendices of this Recommendation. The term is introduced to maintain clarity in the text, especially when comparisons with other loudness prediction models are performed. + +sone: A subjective scale to express loudness, where the value of 1 sone is attributed to the loudness of a pure tone of frequency 1000 Hz at 40 dBSPL. Thus, a sound with loudness equal to 2 sone will be perceived with a "strength", or sensation, twice more important than a sound with a loudness of 1 sone + +## 6 Existing loudness model structures + +All of the main existing loudness models are completely or partly based on the original Zwicker loudness model [b-Zwicker 1]. Zwicker proposed a sophisticated model that predicts loudness perception (in sone), not only as a function of intensity, but also depending on the spectral shape of a stationary sound, using findings from both physiological acoustics and psychoacoustics. This model takes into account the following facts: + +- the hearing threshold; +- the change in loudness with level; +- the spectral masking of frequency components; +- the effect of spectral loudness summation. + +All models use the same overall structure detailed in Figure 1. The general algorithm is summarized by the following steps: first, pre-filtering to account for outer and middle ear transmission, then construction of excitation patterns according to an auditory frequency scale (Bark/equivalent rectangular bandwidth (ERB)), next transformation of the excitation into specific loudness and finally the summation of the specific loudness across the auditory frequency scale. + +![Flowchart of the general structure of a loudness model based on the model proposed by Zwicker. The process starts with a 'Time signal' represented by a waveform graph. This signal is processed through four sequential steps: (1) Outer and middle ear filtering, (2) Transformation of spectrum to excitation pattern, (3) Transformation of excitation pattern to specific loudness, and (4) Integration of specific loudness to overall loudness. The final output is 'Loudness', with the reference 'P.700(19)_F01'.](d4af765160d04ecef538e5066006dc77_img.jpg) + +``` + +graph TD + A[Time signal] --> B["(1) Outer and middle ear filtering"] + B --> C["(2) Transformation of spectrum to excitation pattern"] + C --> D["(3) Transformation of excitation pattern to specific loudness"] + D --> E["(4) Integration of specific loudness to overall loudness"] + E --> F[Loudness P.700(19)_F01] + +``` + +Flowchart of the general structure of a loudness model based on the model proposed by Zwicker. The process starts with a 'Time signal' represented by a waveform graph. This signal is processed through four sequential steps: (1) Outer and middle ear filtering, (2) Transformation of spectrum to excitation pattern, (3) Transformation of excitation pattern to specific loudness, and (4) Integration of specific loudness to overall loudness. The final output is 'Loudness', with the reference 'P.700(19)\_F01'. + +**Figure 1 – General structure of a loudness model based on the model proposed by Zwicker** + +Depending on the type of signal being processed, there exist two main families of model: the models for stationary sounds and the ones for non-stationary sounds. + +For the non-stationary sounds, the basic principle is the same as for stationary sounds. However, the models have been extended to better cope with time-varying sounds. This is done by modelling the post-masking effect and the temporal integration of loudness [b-Zwicker 3], [b-Widmann]; hence, the loudness is calculated as a function of time and not in a global way. + +In the following clauses the methods that have been considered during the development of this Recommendation are briefly explained. + +### 6.1 Zwicker model for stationary sound + +This is the original model of Zwicker [b-Zwicker 1], [b-Zwicker 2], also adopted as an international standard by ISO [b-ISO 532]. + +### 6.2 Moore and Glasberg model for stationary sound + +Moore and Glasberg [b-Moore 1], [b-Moore 2], [b-Glasberg 2] have updated the Zwicker model to incorporate more recent findings in psychoacoustics. The three main differences with the Zwicker model are: the calculation of the auditory filter, outer and middle ear filtering and excitation pattern calculation. The model was adopted in the standard of the United States of America [b-ANSI S3.4-2007]. This model is denoted ANSI S3.4 in the remainder of this Recommendation. + +### 6.3 Zwicker and Fastl model for non-stationary sound + +This model was elaborated by Zwicker and Fastl in 1999 [b-Zwicker 3]. In this model, a new stage was added to model how loudness varies with time and the post-masking effect. To derive a single value of loudness for the overall signal, Zwicker and Fastl recommend to use some statistical indicator such as N4, N5 or N7, which is the loudness value reached 4, 5 or 7% of the time, respectively. Zwicker recommend using N7 for speech signals [b-Zwicker 3]. + +### 6.4 Glasberg and Moore for non-stationary sound + +This model was elaborated by Glasberg and Moore in 2002 [b-Glasberg 1]. The model was modified for stationary sound to obtain loudness as a function of time, called instantaneous loudness, and this would correspond to the overall activity inside the auditory nerve measured for a very short period of + +time. Then, the short-term loudness (STL) was calculated from the instantaneous loudness by taking into account the temporal masking and the temporal integration. STL corresponds to the loudness perceived during a short segment of sound (a syllable, for example). Derived also from the STL, long-term loudness (LTL) is used to describe the loudness sensations that are built rather slowly. According to the literature [b-Glasberg 1], the overall loudness could be calculated by considering the maximum of STL or the average of LTL. [b-Rennies 1] and [b-Rennies 2] showed that the averaged LTL better describes speech loudness. + +# 7 Reasons for a new universal loudness model + +There are several reasons why it is necessary to assess the loudness, and not only loudness rating, for speech/audio terminals: + +- The technology of speech and audiovisual/multimedia terminals has strongly evolved during recent years. In particular, most of this type of equipment implement speech enhancement functions which are non-linear and time variant such as background noise cancellation, improved double talk behaviour and automatic gain control (AGC), providing stereophonic and/or three dimensional (3D) acoustical outputs. The real speech signals or specific sound programs also impact the output level of the terminal, in particular on the level perceived by the user: + - Loudness ratings are determined according to [ITU-T P.79]. The algorithm was initially determined for narrowband speech and adaptations have been done for wideband speech. It is not the intention to develop such adaptations for SWB and fullband speech. + - Loudness rating is a fundamental parameter for the transmission plans and is determined with a reference input signal at a reference level. + - The loudness rating is currently assessed on a long-term signal spectrum (currently 30 s of speech). +- Some test signals (see [ITU-T P.501]) also give the possibility to compute loudness rating on a per utterance basis (in order to check the variation of loudness rating over time); however, there is no standardized approach to this use. +- Even if the terminal provides binaural listening, the loudness rating is based on a monaural model. +- The use of loudness rating is a fundamental parameter for transmission planning, while absolute loudness measurement is of interest for other purposes. + +# 8 Loudness model for receiving loudness + +### 8.1 Recording of speech signals + +For the calculation of receiving loudness, appropriately captured speech signals are indispensable. In order to ensure high-quality source material and reproducible results, only speech samples of clause 7.3 (including compressed speech), Annex B.3, Annex C or Annex D of [ITU-T P.501] shall be used. This source signal is denoted as $x(k)$ in the following. + +If not stated otherwise, the uncompressed, British English single-talk sequence described in clause 7.3 of [ITU-T P.501] shall be used. + +For the recording of transmitted speech signals in handset and hands-free mode, a head and torso simulator (HATS) according to [ITU-T P.58] equipped with artificial ears of type 3.3 or 3.4 according to [ITU-T P.57] shall be used. Measurements of handsets shall follow the guidelines in [ITU-T P.64]. Measurements conducted in hands-free mode shall comply with the guidelines in [ITU-T P.581] or [ITU-T P.341]. Measurement of headsets shall comply with the guidelines in [ITU-T P.381] or + +[ITU-T P.382]. Monaurally recorded signals are denoted as $y(k)$ in the following, binaurally recorded signal as **$y(k)$** (bold). A general overview of the recording set-up is shown in Figure 2. + +If not stated otherwise, the HATS signals shall be diffuse-field equalized using a filter following the inverse of the eardrum reference point (DRP) to diffuse-field response in Annex A of [ITU-T P.58]. + +NOTE – Since the loudness model according to [b-ISO 532-1] also supports free-field equalized recordings, this type of equalization is also possible, provided the free-field option is selected for the loudness calculation. + +![Figure 2: Recording set-up for receive loudness calculation based on HATS. The diagram shows a source speech signal x(k) entering a 'Network' block, which then connects to a 'Communication terminal' (represented by a telephone icon). The signal is then transmitted to a 'HATS' (Head and Torso Simulator) model, represented by a person's head and shoulders. Red curved arrows indicate sound waves entering the ears. Below the HATS, two red arrows labeled 'R' (Right) and 'L' (Left) point to 'Binaural recording y(k)'. The text 'P.700(21)_F02' is at the bottom right.](a33da0f14e456f92539ce3e9b7d81f9a_img.jpg) + +Figure 2: Recording set-up for receive loudness calculation based on HATS. The diagram shows a source speech signal x(k) entering a 'Network' block, which then connects to a 'Communication terminal' (represented by a telephone icon). The signal is then transmitted to a 'HATS' (Head and Torso Simulator) model, represented by a person's head and shoulders. Red curved arrows indicate sound waves entering the ears. Below the HATS, two red arrows labeled 'R' (Right) and 'L' (Left) point to 'Binaural recording y(k)'. The text 'P.700(21)\_F02' is at the bottom right. + +**Figure 2 – Recording set-up for receive loudness calculation based on HATS** + +### 8.2 Loudness calculation + +The underlying loudness model used for the calculation is described in [b-ISO 532-1]. Before inserting the signal into the model, it has to be calibrated to the physical unit pascal. + +The loudness analysis is carried out over the entire recording, including possible speech pauses, resulting in a STL vs time in sone. In order to obtain a single loudness value per recording, the curve is aggregated vs time by calculating the average across active speech parts. + +If the recording contains speech pauses, leading or trailing silence periods, these shall be excluded from the sone vs time data. An automated speech activity classification algorithm for pulse-code modulation (PCM) audio is described in Appendix II of [ITU-T G.160] and shall be applied to the sone vs time data in this case. This classifier assumes that speech pauses of durations of less than 400 ms are still considered as active speech. + +The aggregated loudness $N$ (unit sone) shall be reported as the main result per evaluated speech signal. Optionally, the result may also be reported as loudness level $LN$ (unit phon) by transforming loudness $N$ via Equation (1) to (3) in clause 5.3 of [b-ISO 532-1]. Figure 3 illustrates the different steps in the loudness calculation method. For binaural recordings, activity classification, loudness-vs-time calculation and average across active speech frames shall be carried out separately for each recording channel. + +![Figure 3: Loudness calculation for recorded speech signals. The flowchart shows the input signal y(k) branching into two parallel paths. The top path goes through 'ISO 532-1' to produce 'Loudness vs time [sone]'. The bottom path goes through 'Appendix II of [ITU-T G.160]' to produce 'Frame activity'. Both paths converge into a 'Skip silence/pauses' block. This is followed by an 'Average' block, then a 'Binaural summation' block. From 'Binaural summation', one output is 'Loudness [sone]' and the other goes through a 'Sone-to-phon' block to produce 'Loudness level [phon]'. The text 'P.700(21)_F03' is at the bottom right.](d17f75945bbb3feb84a153ecfedb9b81_img.jpg) + +Figure 3: Loudness calculation for recorded speech signals. The flowchart shows the input signal y(k) branching into two parallel paths. The top path goes through 'ISO 532-1' to produce 'Loudness vs time [sone]'. The bottom path goes through 'Appendix II of [ITU-T G.160]' to produce 'Frame activity'. Both paths converge into a 'Skip silence/pauses' block. This is followed by an 'Average' block, then a 'Binaural summation' block. From 'Binaural summation', one output is 'Loudness [sone]' and the other goes through a 'Sone-to-phon' block to produce 'Loudness level [phon]'. The text 'P.700(21)\_F03' is at the bottom right. + +**Figure 3 – Loudness calculation for recorded speech signals** + +Under certain recording conditions, adequate classification of active speech frames may be considerably impaired by e.g., background noise or temporal clipping of the speech signal. In such cases, the determination of speech activity can also be carried out on the underlying source signal of + +the measurement (introduced as $x(k)$ in clause 8.1). The delay between each recording channel (left, right) and source signal shall be less than 1 ms, i.e., the user shall apply an adequate compensation of the delay. The usage of a different signal as a reference for the speech activity classification shall be reported. + +![Figure 4: Loudness calculation for recorded speech signals with reference signal. The diagram shows two input signals, y(k) and x(k). y(k) is processed by 'ISO 532-1' to produce 'Loudness vs time [sone]'. x(k) is processed by 'Appendix II of [ITU-T G.160]' to produce 'Frame activity'. Both signals are then processed by 'Skip silence/pauses'. The output of 'Skip silence/pauses' is then processed by 'Average', which produces 'Loudness [sone]'. This 'Loudness [sone]' is then processed by 'Binaural summation' to produce the final 'Loudness [sone]'. The 'Binaural summation' output is also processed by 'Sone-to-phon' to produce 'Loudness level [phon]'. A label 'P.700(21)_F04' is present at the bottom right of the diagram.](bd0b93e7a46ede276d0a3b79ac487bd9_img.jpg) + +``` + +graph LR + yk[y(k)] --> ISO5321[ISO 532-1] + ISO5321 -- "Loudness vs time [sone]" --> SkipSilence[Skip silence/pauses] + xk[x(k)] --> AppendixII["Appendix II of [ITU-T G.160]"] + AppendixII -- "Frame activity" --> SkipSilence + SkipSilence --> Average[Average] + Average --> BinauralSummation[Binaural summation] + BinauralSummation --> LoudnessSone[Loudness [sone]] + BinauralSummation --> SoneToPhon[Sone-to-phon] + SoneToPhon --> LoudnessLevelPhon[Loudness level [phon]] + P70021_F04[P.700(21)_F04] + +``` + +Figure 4: Loudness calculation for recorded speech signals with reference signal. The diagram shows two input signals, y(k) and x(k). y(k) is processed by 'ISO 532-1' to produce 'Loudness vs time [sone]'. x(k) is processed by 'Appendix II of [ITU-T G.160]' to produce 'Frame activity'. Both signals are then processed by 'Skip silence/pauses'. The output of 'Skip silence/pauses' is then processed by 'Average', which produces 'Loudness [sone]'. This 'Loudness [sone]' is then processed by 'Binaural summation' to produce the final 'Loudness [sone]'. The 'Binaural summation' output is also processed by 'Sone-to-phon' to produce 'Loudness level [phon]'. A label 'P.700(21)\_F04' is present at the bottom right of the diagram. + +**Figure 4 – Loudness calculation for recorded speech signals with reference signal** + +### 8.3 Application for terminals + +#### 8.3.1 Overview + +Loudness calculations according to [b-ISO 532-1] is only specified for mono signals and corresponds to a diotic presentation of a stimulus to a (virtual) listener. Annex D of [b-ISO 532-1] only provides minimum information for the usage of binaural signals. For the application in communication terminals, a simple loudness summation model for the left and right ears is assumed in the following. The overall perceived loudness $N$ (in sone) is defined as the sum of the loudness values of the left ( $N_L$ ) and right ( $N_R$ ) ears, as shown in Equation 1. + +$$N = N_L + N_R \quad (1)$$ + +#### 8.3.2 Terminals with monaural playback + +For terminals which are applicable to only one ear (e.g., handsets, monaural headsets), loudness and loudness level are determined as follows: + +- The signal under test is present only at one artificial ear and is recorded according to clause 8.1. For sake of simplicity, it is assumed that the right ear is used in the following. +- The diotic loudness (in sone) for this ear signal is calculated according to clause 8.2 as $N'_R$ . Due to the assumed diotic playback, the resulting diotic loudness actually represents the doubled monaural loudness for this ear: $N'_R = 2 \cdot N_R$ . +- For the opposite left ear, it is assumed that the loudness is zero (e.g., $N_L = 0$ ). +- The overall monaural loudness $N$ is calculated with the aforementioned definitions according to Equation 2. + +$$N = N_L + N_R = 0 + 0.5 \cdot N'_R \quad (2)$$ + +- The overall monaural loudness level $LN$ is calculated by transforming the overall monaural loudness $N$ according to clause 5.3 of [b-ISO 532-1]. + +#### 8.3.3 Terminals with binaural playback + +For terminals which are applicable to both ears (e.g., binaural headsets, headphones, hands-free), loudness and loudness level are determined as follows: + +- The signal under test is present at both artificial ears and is recorded according to clause 8.1. +- The loudness (in sone) is calculated for the left and right ear signals separately according to clause 8.2 as $N'_L$ and $N'_R$ . Due to the assumed diotic playback of each individual ear signal, + +the resulting loudness actually represents the doubled monaural loudness value, i.e., $N'_L = 2 \cdot N_L$ and $N'_R = 2 \cdot N_R$ . + +- The overall binaural loudness $N$ is calculated with the aforementioned definitions according to Equation 3. + +$$N = N_L + N_R = 0.5 \cdot (N'_L + N'_R) \quad (3)$$ + +- The overall binaural loudness level LN is calculated by transforming the overall binaural loudness $N$ according to clause 5.3 of [b-ISO 532-1]. + +NOTE – The binaural loudness summation in Equation 3 can also be interpreted as the average between the diotic loudness values calculated for the left and the right ear separately. However, the calculation should always be reported as binaural loudness summation, since it only compensates for the diotic presentation mode of [b-ISO 532-1]. + +## 9 Loudness model for sending loudness + +The sending loudness is calculated by assuming a connection with a second terminal having the following receive characteristics: + +- 6.28 dBPa/V for all applicable frequencies, for calculation of narrowband sending loudness; +- 4.83 dBPa/V for all applicable frequencies, for calculation of wideband, SWB and fullband sending loudness. + +NOTE 1 – Loudness calculations can, in principle, only be made for signals in the acoustical domain. The sending loudness of a terminal can, however, be defined by assuming a connection with a nominal receiving terminal; see Appendix IV. + +NOTE 2 – The derivation of this assumed nominal receiving terminal is based on a monaural listening situation. This approach was chosen to be compatible with the definitions of the sending loudness rating (SLR) measurements according to [ITU-T P.79], which also includes an assumed monaural receiving terminal. + +![Diagram of recording set-up and calibration for send loudness calculation. A person (HATS) provides a source speech signal x(k) at the mouth reference point (MRP) to a communication terminal. The terminal is connected to a network, which outputs a signal y'(k) at the point of interconnection (POI). This signal is then processed by a receive sensitivity block to produce the final output y(k).](f57a881fcbcaf494a7dedc4f61224991_img.jpg) + +``` +graph LR; HATS[HATS] -- "Source speech signal at MRP x(k)" --> CT[Communication terminal]; CT --> Network[Network]; Network -- "y'(k) POI" --> RS[Receive sensitivity]; RS --> yk["y(k) P.700(21)_F05"]; +``` + +Diagram of recording set-up and calibration for send loudness calculation. A person (HATS) provides a source speech signal x(k) at the mouth reference point (MRP) to a communication terminal. The terminal is connected to a network, which outputs a signal y'(k) at the point of interconnection (POI). This signal is then processed by a receive sensitivity block to produce the final output y(k). + +(POI – point of interconnection) + +**Figure 5 – Recording set-up and calibration for send loudness calculation** + +A general overview of the recording set-up is shown in Figure 5. The following test procedure shall be used: + +- The test set-up is the same as for SLR measurements for the respective terminal type (e.g., [ITU-T P.311] for wideband handsets and headsets and [ITU-T P.341] for wideband digital loudspeaking and hands-free telephony terminals). +- As a test signal, the British English single-talk sequence described in clause 7.3 of [ITU-T P.501] shall be used at an [ITU-T P.56] active speech level of $-4.7$ dBPa at the mouth reference point (MRP); similarly to clause 8, this source signal is denoted as $x(k)$ in the following. + +- The test signal $x(k)$ is applied to the terminal using the equalized artificial mouth and recorded at an electrical reference point corresponding to the point of interconnection (POI), represented in volt. The signal is denoted as $y'(k)$ and is scaled to a representation in pascal, using the reference receiving path described above. The scaling factors are either: + - $10^{\frac{6.28}{20}} = 2.06$ for narrowband; or + - $10^{\frac{4.83}{20}} = 1.74$ for wideband, SWB and fullband. +- The scaled signal is denoted $y(k)$ in the following. +- The sending loudness is now calculated according to the description for receiving loudness in clause 8.2. Due to assumed monaural listening at the nominal receiving terminal, the resulting diotic loudness has to be divided by factor two (see also clause 8.3.2). + +## 10 Loudness model for sidetone loudness + +For further study. + +# 11 Loudness model for listener sidetone loudness + +For further study. + +# Annex A + +## Validation and test vectors for loudness model implementation + +(This annex forms an integral part of this Recommendation.) + +### A.1 Introduction to test vectors for validation + +This annex provides reference results for speech samples scaled at certain active speech levels according to [ITU-T P.56]. In addition, several band-pass filters are applied in order to simulate typical bandwidth limitations in narrowband, wideband, SWB and fullband mode. Examples with and without active speech classification are provided. + +## A.2 Validation of loudness model implementation + +#### A.2.1 Introduction + +The examples provided in this clause illustrate the usage of the loudness model for different bandwidth limitations and (active) speech levels. Note that in contrast to clause 8.2, the entire signal is analysed, hence no speech pauses are excluded from the calculation. + +#### A.2.2 Speech material + +For all reference loudness values, the speech signal *FB\_male\_female\_single-talk\_seq.wav* of clause 7.3.2 of [ITU-T P.501] is used. The speech signal includes short speech pauses which are longer than 400 ms. However, the analysis is carried out over the entire signal (including trailing/leading pauses). + +#### A.2.3 Processing + +For each bandwidth mode, the source speech signal is filtered with one or more finite impulse response (FIR) filters according to [ITU-T G.191]: + +- Narrowband: LP35 and MSIN (16 kHz sampling rate); +- Wideband: P341 (16 kHz sampling rate); +- SWB: 14KBP (32 kHz sampling rate); +- Fullband: 20KBP (48 kHz sampling rate). + +After filtering, each signal is calibrated to active speech levels according to [ITU-T P.56] ranging from 40 dBSPL to 90 dBSPL in six steps of 10 dB. + +Then the loudness model which has been implemented according to clause 8 is tested with each of the 24 resulting level/bandwidth conditions with diffuse-field correction. Diotic presentation is assumed for the mono signal. + +#### A.2.4 Reference results + +Table A.1 shows the loudness results in sone and phon for the signal processing described in clause A.2.3. Loudness values [sone] that are within $\pm 4\%$ of the values stated in Table A.1 and loudness levels that are within $\pm 0.5$ phon of the values stated in Table A.1 are considered to be compliant with the specifications of this Recommendation. + +**Table A.1 – Reference results for loudness and loudness level** + +| Active speech level
[dBSPL] | Filter | Loudness
[sone] | Loudness level
[phon] | +|----------------------------------------|---------------|----------------------------|----------------------------------| +| 40 | Narrowband | 0.93 | 38.96 | +| | Wideband | 1.03 | 40.46 | +| | SWB | 1.05 | 40.76 | +| | Fullband | 1.05 | 40.75 | +| 50 | Narrowband | 2.17 | 51.17 | +| | Wideband | 2.44 | 52.85 | +| | SWB | 2.50 | 53.23 | +| | Fullband | 2.50 | 53.21 | +| 60 | Narrowband | 4.57 | 61.93 | +| | Wideband | 5.15 | 63.65 | +| | SWB | 5.30 | 64.07 | +| | Fullband | 5.30 | 64.06 | +| 70 | Narrowband | 9.06 | 71.80 | +| | Wideband | 10.24 | 73.56 | +| | SWB | 10.55 | 73.99 | +| | Fullband | 10.54 | 73.97 | +| 80 | Narrowband | 17.42 | 81.23 | +| | Wideband | 19.74 | 83.03 | +| | SWB | 20.31 | 83.44 | +| | Fullband | 20.29 | 83.43 | +| 90 | Narrowband | 32.98 | 90.43 | +| | Wideband | 37.38 | 92.24 | +| | SWB | 38.35 | 92.61 | +| | Fullband | 38.33 | 92.60 | + +### A.3 Validation of speech pause detection and loudness model implementation + +#### A.3.1 Introduction + +In general, if the recording contains speech pauses, leading or trailing silence periods, these shall be excluded from the temporal aggregation. An automated speech activity classification algorithm for this purpose is described in Appendix II of [ITU-T G.160] and shall be used in this case (see clause 8.2). This classifier assumes that speech pauses of durations less than 400 ms are still considered as active speech. + +The source code for ITU-T G.160, Appendix II, is available as an electronic attachment to [ITU-T G.160]. The main program for ITU-T G.160 Appendix II, named "g160app2.c", can be generalized by utilizing a fixed frame size of 10 ms (instead of hardcoded frame length of 80 samples, which refers to a fixed sampling rate of 8 kHz). + +For validation of the loudness model described in this Recommendation, the default setting shall be used for all parameters. + +#### A.3.2 Speech material + +For all reference loudness values, the first 20.0 s of the speech signal *FB\_male\_female\_double-talk\_seq.wav* of clause 7.3.5 of [ITU-T P.501] is used. This signal contains short words only and provides a speech activity of approximately 25 per cent. + +#### A.3.3 Processing + +In contrast to clause A.2.3, the source speech signal is considered only in fullband mode, i.e., the filter 20KBP (48 kHz sampling rate) according to [ITU-T G.191] is applied. + +After filtering, the signal is calibrated to active speech levels according to [ITU-T P.56] ranging from 40 dBSPL to 90 dBSPL in six steps of 10 dB. + +Then the loudness model according to clause 8 is carried out for each of the 6 resulting level conditions with diffuse-field correction. Diotic presentation is assumed for the mono signal. + +#### A.3.4 Reference results + +Table A.2 shows the loudness results in sone and phon for the signal processing described in clause A.3.3. Loudness values [sone] that are within $\pm 4\%$ of the values stated in Table A.1 and loudness levels that are within $\pm 0.5$ phon of the values stated in Table A.1 are considered to be compliant with the specifications of this Recommendation. + +**Table A.2 – Reference results for loudness and loudness level (excluding speech pauses)** + +| Active speech Level [dBSPL] | Filter | Loudness [sone] | Loudness level [phon] | +|-----------------------------|----------|-----------------|-----------------------| +| 40 | Fullband | 1.45 | 45.36 | +| 50 | Fullband | 3.35 | 57.46 | +| 60 | Fullband | 6.97 | 68.02 | +| 70 | Fullband | 13.69 | 77.75 | +| 80 | Fullband | 26.18 | 87.10 | +| 90 | Fullband | 48.87 | 96.11 | + +# Appendix I + +## Result of loudness experiment A + +(This appendix does not form an integral part of this Recommendation.) + +### I.1 Introduction to experiment A + +This appendix presents the result of a subjective loudness experiment performed for clean speech, music and noisy speech stimuli. A mobile phone supporting narrowband, wideband and SWB has been used for recording the stimuli. This experiment should be seen in the context of similar experiments which have been conducted during the process of developing this Recommendation. Some of the characteristics of these experiments are summarized in the following part of the introduction. + +Orange used one-critical band noises centred at 1 kHz as the reference stimuli and compared the subjective rating of speech samples with that of the reference stimuli [b-AES E-Library], [b-Edjekouane]. The ANSI S3.4 [b-ANSI S3.4-2007] implementation with the modification of the loudness growth function, i.e., P.Loudness, is suggested for predicting the loudness of speech samples. The suggested loudness growth function is dependent on the telephone mode, i.e., hands-free and handset. HEAD acoustics utilized 3-Bark band noises centred at 1 kHz including a roll-off characteristic as the reference in order to minimize the perceptual interaction between loudness and annoyance created by the tonal perception of the one-critical-band noise [b-AES E-Library]. + +These investigations used one session for the reference stimuli and the other for speech samples. Separating the two sessions based on the type of stimuli may cause the range effect as well as the stimulus spacing and frequency. The experiment with the 3-Bark band noise cannot come up with the subjective loudness level directly, and rather the method requires a loudness equalization procedure between the reference and the speech samples. If there is a systematic offset introduced in the objective loudness calculation, the loudness equalization process introduces almost the same offset in both type of stimuli. The resulting comparison between the subjective responses and the objective loudness values may not reveal this systematic offset. + +For this reason, the current investigation included references as well as speech samples in the same listening test session. Diotic 1 kHz tones are used as the reference, and it allows the estimation of the subjective loudness level directly from the listening test results. Since this approach may discover the systematic offset when calculating the loudness of different stimuli, the calculated error between the subjective and the objective loudness levels may be greater than those reported in [b-AES E-Library]. + +### I.2 Subjective experiment + +The detailed procedure for the subjective evaluation is already reported in [b-AES E-Library]. Due to the limited time assigned for the listening test, the number of speech samples are reduced to four for the clean speech samples and two for the noisy speech samples. During the preparation of the noisy speech samples, the level of noise is calculated at the eardrum position rather than at the centre of the head in the free field. This resulted in increasing the signal-to-noise ratios (SNRs) in the noisy speech samples by approximately 9 dB. Thus, the actual SNRs used for the listening test are 9 and 19 dB instead of 0 and 10 dB. + +#### I.2.1 Subjective loudness function + +Reimes et al. [b-AES E-Library] compared the predicted loudness values using the individual loudness function and the averaged one. They concluded that there is no significant difference between them. Thus, the average loudness function is used in the current investigation. Figure I.1 shows the average subjective loudness values as a function of the 1-kHz tone level. The error bars represent the 95% confidence intervals across 18 subjects. A two-factor analysis of variance + +(ANOVA) with the telephone mode (handset and hands-free) and the reference levels all constituting within-subjects factors confirmed the significant main effect of telephone mode ( $p = 0.009$ ) and the significant interaction with the reference level ( $p < 0.001$ ). This may be triggered by the perceived loudness difference between hands-free mode and handset mode on average. When a stimulus is presented in a session with other, louder stimuli, there is a tendency of judging the perceived loudness of the stimulus to be softer. Consequently, the loudness functions are fitted separately for hands-free and handset mode. The solid lines in Figure I.1 show the fitted loudness functions. + +![Figure I.1: Subjective loudness growth function and the corresponding fitted sigmoid function. The graph plots Loudness level [phon] (y-axis, 0 to 100) against Loudness point (x-axis, 0 to 50). Two data series are shown: Hands-free (blue circles and line) and Handset (red circles and line). Both series show a sigmoidal growth pattern, with the Hands-free mode consistently showing higher loudness levels than the Handset mode across the range of loudness points. Error bars are present for each data point.](ed75e80b1e08237f7e90b65357de84d5_img.jpg) + +Figure I.1: Subjective loudness growth function and the corresponding fitted sigmoid function. The graph plots Loudness level [phon] (y-axis, 0 to 100) against Loudness point (x-axis, 0 to 50). Two data series are shown: Hands-free (blue circles and line) and Handset (red circles and line). Both series show a sigmoidal growth pattern, with the Hands-free mode consistently showing higher loudness levels than the Handset mode across the range of loudness points. Error bars are present for each data point. + +**Figure I.1 – Subjective loudness growth function and the corresponding fitted sigmoid function** + +The sigmoid function used for the fitting is defined in Equation I.1, and it is the same as the one suggested in [b-AES E-Library]. + +$$L(p) = p_{min} + \frac{(p_{max} - p_{min})}{1 + e^{a \cdot (p_0 - p)}} \quad (I.1)$$ + +Table I.1 summarizes the parameters estimated by the non-linear regression. The minimum value of $P_{min}$ is set to $-100$ for the regression. + +**Table I.1 – Fitted parameters for the sigmoid function in hands-free mode and handset mode** + +| | Hands-free | Handset | +|-----------------|--------------------------------------------------------------------|--------------------------------------------------------------------| +| Arithmetic mean | $P_{min}: -100.0$
$P_{max}: 109.3$
$a: 0.049$
$p_0: -2.3$ | $P_{min}: -100.0$
$P_{max}: 111.1$
$a: 0.044$
$p_0: -2.2$ | + +The subjective loudness point values of the speech and music samples from the listening test are transformed to the loudness levels using Equation I.1 and the parameters in Table I.1. + +#### I.2.2 Loudness of clean samples + +Figure I.2 shows the effect of different presentation levels in the perceived loudness of clean sound samples. Due to limited space in the report, only results for the EVS244 codec are displayed here. + +The first four sound samples are speech samples. Sample 5 is classical music, and Sample 6 is pop music. A four-factor ANOVA with the telephone modes, the codecs, the sound pressure levels (SPLs) and the sound samples all constituting within-subjects factors confirmed the highly significant main effect of SPLs ( $p < 0.001$ ). + +![Figure I.2: Two line graphs showing subjective loudness ratings for six samples across four SPL levels (45, 55, 65, 75 dB) for Hands-free and Handset modes.](e9f324c9d7f5305a55cb156855e95b95_img.jpg) + +Figure I.2 consists of two line graphs side-by-side. The left graph is titled 'Hands-free - EVS244' and the right graph is titled 'Handset - EVS244'. Both graphs plot 'Loudness point' (y-axis, 10 to 40) against 'Sample Nr.' (x-axis, 1 to 6). Four data series are shown in each graph, corresponding to different sound pressure levels (SPLs): 45 dB (blue dashed line with circles), 55 dB (green solid line with circles), 65 dB (red dashed line with triangles), and 75 dB (black dashed line with crosses). Error bars are included for each data point. + +| Sample Nr. | 45 dB | 55 dB | 65 dB | 75 dB | +|------------|-------|-------|-------|-------| +| 1 | 16 | 23 | 30 | 38 | +| 2 | 15 | 23 | 30 | 38 | +| 3 | 15 | 22 | 29 | 37 | +| 4 | 15 | 21 | 25 | 34 | +| 5 | 18 | 24 | 30 | 38 | +| 6 | 15 | 23 | 30 | 38 | + + + +| Sample Nr. | 45 dB | 55 dB | 65 dB | 75 dB | +|------------|-------|-------|-------|-------| +| 1 | 13 | 20 | 25 | 30 | +| 2 | 15 | 22 | 27 | 33 | +| 3 | 15 | 23 | 26 | 32 | +| 4 | 13 | 20 | 25 | 31 | +| 5 | 16 | 24 | 28 | 32 | +| 6 | 15 | 22 | 28 | 34 | + +Figure I.2: Two line graphs showing subjective loudness ratings for six samples across four SPL levels (45, 55, 65, 75 dB) for Hands-free and Handset modes. + +**Figure I.2 – Subjective loudness ratings in different presentation levels** + +The effect of telephone mode is displayed in Figure I.3, and there is a highly significant main effect ( $p < 0.001$ ). This effect of telephone mode is mainly caused by the fact that in handset mode, the subjects are exposed to the stimuli monaurally through their preferred ear. The typical range of binaural gain, i.e., the loudness advantage by listening with two ears compared to one ear, is between 3 and 10 dB [b-Moore 3]. Apart from the effect of binaural listening, the spectral shape of the stimuli is dependent on the telephone mode since the transfer function from the phone to the eardrum is affected by the telephone mode. For example, in handset mode, the outer-ear transfer function of the HATS is not present. + +![Figure I.3: A line graph comparing subjective loudness ratings for six samples between Hands-free and Handset modes at 75 dB SPL.](50fecd0e7c9bf4ebf321d8367d42cc94_img.jpg) + +Figure I.3 is a line graph titled 'EVS244 - 75 dB'. It plots 'Loudness point' (y-axis, 10 to 40) against 'Sample Nr.' (x-axis, 1 to 6). Two data series are shown: 'Hands-free' (blue dashed line with circles) and 'Handset' (green solid line with circles). Error bars are included for each data point. + +| Sample Nr. | Hands-free | Handset | +|------------|------------|---------| +| 1 | 38 | 30 | +| 2 | 38 | 33 | +| 3 | 37 | 32 | +| 4 | 34 | 31 | +| 5 | 38 | 32 | +| 6 | 38 | 34 | + +Figure I.3: A line graph comparing subjective loudness ratings for six samples between Hands-free and Handset modes at 75 dB SPL. + +**Figure I.3 – Subjective loudness ratings in the two telephone modes, i.e., hands-free and handset, when the level of stimuli is 75 dB** + +The effect of codecs is less visible compared to the other factors. Figure I.4 compares the subjective results across different codecs. It is noticeable that the fullband samples are judged to be softer compared to the other samples in hands-free mode ( $p < 0.001$ ). The fullband samples are the original + +samples without passing through the phone, and therefore they do not include the outer-ear transfer function, which typically increases the perceived loudness of speech samples. + +![Two line graphs showing subjective loudness ratings for different codecs in hands-free and handset modes at 75 dB stimulus level.](df0685d2d1176d617ed1e642de4e5425_img.jpg) + +The figure consists of two line graphs side-by-side. The left graph is titled 'Hands-free - 75 dB' and the right graph is titled 'Handset - 75 dB'. Both graphs plot 'Loudness point' (y-axis, 10 to 40) against 'Sample Nr.' (x-axis, 1 to 6). Six codecs are compared: AMR122 (blue dashed line with circles), AMRWB1265 (green solid line with circles), AMRWB2385 (red solid line with triangles), EVS132 (black dashed line with crosses), EVS244 (cyan dashed line with diamonds), and FB (magenta solid line with stars). In both modes, the loudness ratings are relatively stable across the six samples, with most codecs showing ratings between 30 and 40. The FB codec consistently shows lower loudness ratings, around 30-35. The EVS244 codec generally shows higher loudness ratings, around 35-40. The EVS132 codec shows a slight increase in loudness in the handset mode compared to the hands-free mode. + +| Sample Nr. | Hands-free - 75 dB (Loudness point) | Handset - 75 dB (Loudness point) | +|------------|-------------------------------------|----------------------------------| +| 1 | ~38 | ~30 | +| 2 | ~40 | ~32 | +| 3 | ~35 | ~30 | +| 4 | ~35 | ~32 | +| 5 | ~38 | ~32 | +| 6 | ~38 | ~35 | + +Two line graphs showing subjective loudness ratings for different codecs in hands-free and handset modes at 75 dB stimulus level. + +Figure I.4 – Subjective loudness ratings in different codecs when the level of stimuli is 75 dB + +#### I.2.3 Loudness of noisy speech samples + +When judging the loudness of a speech sample in noise, i.e., partial loudness, subjects are instructed to focus on the loudness of the target speech while ignoring the background noise. The perceived loudness is expected to decrease when the level of background noise increases. Figure I.5 displays the partial loudness ratings of speech sample 1 as a function of codecs. In hands-free mode, the perceived partial loudness seems to be almost the same for the two SNR conditions. However, in handset mode, i.e., monaural listening, the effect of SNR seems to be more visible. This effect of telephone mode in partial loudness is confirmed by a five-factor ANOVA with the SNRs, the telephone modes, the codecs, the SPLs and the sound samples all constituting within-subjects factors. The interaction between the SNRs and the telephone modes is significant ( $p = 0.001$ ). + +![Four line graphs showing subjective loudness ratings for speech sample 1 across six codecs. The graphs are titled 'Hands-free - 45 dB - speech sample 1', 'Handset - 45 dB - speech sample 1', 'Hands-free - 75 dB - speech sample 1', and 'Handset - 75 dB - speech sample 1'. Each graph plots 'Loudness point' (y-axis, 0-40) against 'Codec Nr.' (x-axis, 1-6). Two data series are shown: 9 dB (blue dashed line with circles) and 19 dB (red dashed line with triangles). In all cases, the 19 dB series is higher than the 9 dB series. Loudness generally decreases from codec 4 to 6. The 75 dB condition results in higher loudness points than the 45 dB condition. A label 'P.700(21)_FI.5' is present in the bottom right of the fourth graph.](645bea0b27d63e4a9a300af5793ae7d2_img.jpg) + +Four line graphs showing subjective loudness ratings for speech sample 1 across six codecs. The graphs are titled 'Hands-free - 45 dB - speech sample 1', 'Handset - 45 dB - speech sample 1', 'Hands-free - 75 dB - speech sample 1', and 'Handset - 75 dB - speech sample 1'. Each graph plots 'Loudness point' (y-axis, 0-40) against 'Codec Nr.' (x-axis, 1-6). Two data series are shown: 9 dB (blue dashed line with circles) and 19 dB (red dashed line with triangles). In all cases, the 19 dB series is higher than the 9 dB series. Loudness generally decreases from codec 4 to 6. The 75 dB condition results in higher loudness points than the 45 dB condition. A label 'P.700(21)\_FI.5' is present in the bottom right of the fourth graph. + +**Figure I.5 – Subjective loudness ratings in different codecs for speech sample 1** + +### I.3 Comparison results + +#### I.3.1 Loudness algorithms + +The following describes the objective models used for predicting the perceived loudness. + +Diotic listening (hands-free mode): + +- P.Loudness [b-AES E-Library]: MATLAB implementation provided by Orange; +- [b-ISO 532-1]: based on Zwicker's loudness model and supports both stationary and time-varying model; +- [b-ISO 532-2]: based on the Moore-Glasberg model and intended only for stationary signals; +- short-term time-varying-loudness model from Moore-Glasberg: see [b-Glasberg 1]. + +Monaural listening (handset mode): + +- P.Loudness: perfect loudness summation rule; +- [b-ISO 532-1]: perfect loudness summation rule; +- [b-ISO 532-2]: supports the latest development on binaural loudness including the concept of binaural inhibition [b-Moore 3]; +- short-term time-varying-loudness model from Moore-Glasberg: the latest development on binaural loudness including the concept of binaural inhibition, [b-Glasberg 1] and [b-Moore 3], is implemented. + +Target loudness in noise, i.e., partial loudness, the partial loudness model proposed by Moore et al. [b-Moore 2] is implemented in the following methods: + +- [b-ISO 532-2]; +- short-term time-varying-loudness model from Moore-Glasberg. + +For the time-varying loudness models, the arithmetic mean and N5 are used to determine a single value from the time history of calculated loudness values. The first 200 ms of samples are skipped for this aggregation in order to avoid the influence of temporal loudness integration, i.e., the slow monotonic increase of loudness in the first part of loudness time history. + +#### I.3.2 Clean stimuli result + +The general tendency of music samples on perceived loudness does not seem to be very different from that of speech samples according to the previous clause. Hence, the comparison between the subjective and objective loudness values includes both speech samples and music samples in the same analysis. + +In the current investigation, the subjective loudness level is measured by including diotic 1-kHz tones in the experiment. The goal of the analysis is to quantify the errors in terms of loudness level in an absolute manner compared to [b-AES E-Library], in which the relative loudness errors are calculated by employing the concept of equal noise level. For example, the method in [b-AES E-Library] cannot estimate the absolute loudness difference between N5 and the arithmetic mean of time-varying loudness values since the iterative loudness equalization process compares the calculated loudness of the reference noise and the speech samples with the same algorithm. The loudness difference between N5 and arithmetic mean affects both the reference noises and the speech samples to almost the same degree. In the current investigation, root mean square error (RMSE) is used to quantify the absolute loudness error. + +Figure B.6 compares the subjective loudness level with the predicted loudness level using the P.Loudness model provided by Orange. In hands-free mode, the free-field pressure is used as the input to the loudness, and it is calculated by applying the HATS free-field equalization function to the stimuli. In handset mode, the stimuli, which are recorded by the HATS, are directly used as the input to the loudness calculation because the algorithm requires the eardrum pressure. In general, the model overestimates the loudness values for hands-free mode and underestimates them for handset mode. The RMSE value of hands-free mode is slightly higher compared to handset mode. + +![Two scatter plots comparing subjective and predicted loudness levels. The left plot is for 'Hands-free' mode with an RMSE of 10.57. The right plot is for 'Handset' mode with an RMSE of 8.90. Both plots show a positive correlation between subjective and objective loudness levels in phon.](f880daa1cde4b71fcbaff2df81803a65_img.jpg) + +The figure consists of two side-by-side scatter plots. Both plots have 'Subjective loudness level [phon]' on the x-axis and 'Objective loudness level [phon]' on the y-axis, with axes ranging from 30 to 100. A dashed diagonal line represents the ideal 1:1 relationship. The left plot, titled 'P.Loudness - Hands-free', shows data points generally above the diagonal line, indicating overestimation, with an RMSE of 10.57. The right plot, titled 'P.Loudness - Handset', shows data points generally below the diagonal line, indicating underestimation, with an RMSE of 8.90. A small label 'P.700(21)\_FI.6' is located at the bottom right of the figure. + +Two scatter plots comparing subjective and predicted loudness levels. The left plot is for 'Hands-free' mode with an RMSE of 10.57. The right plot is for 'Handset' mode with an RMSE of 8.90. Both plots show a positive correlation between subjective and objective loudness levels in phon. + +**Figure I.6 – Subjective loudness level vs predicted loudness level using P.Loudness** + +Figure I.7 and Figure I.8 demonstrate how the ISO 532-1 algorithm works. Since the model requires the input signal to be free-field equalized, the HATS free-field equalization function has been applied + +to the signal prior to feeding it to the model. For handset mode, the calculated loudness is divided by 2 for the perfect loudness summation rule. The arithmetic mean aggregation seems to predict the subjective loudness level better compared to N5, and it performs better than the stationary loudness model. In general, the model is more accurate for handset mode compared to hands-free mode. + +![Two scatter plots comparing subjective and predicted loudness levels using the ISO 532-1 stationary model. The left plot is for 'Hands-free' mode with an RMSE of 9.84. The right plot is for 'Handset' mode with an RMSE of 5.88. Both plots show a positive correlation between subjective and objective loudness levels, with a dashed diagonal line representing perfect prediction.](8765ca4dff337b88338ec4e3fa1e14c9_img.jpg) + +Figure I.7 displays two scatter plots comparing subjective loudness level [phon] (X-axis) versus objective loudness level [phon] (Y-axis) using the ISO 532-1 stationary model. + +- ISO 532-1 Stationary - Hands-free:** The plot shows a scatter of data points with an RMSE of 9.84. The axes range from 30 to 100 phon. +- ISO 532-1 Stationary - Handset:** The plot shows a scatter of data points with an RMSE of 5.88. The axes range from 30 to 100 phon. + +P.700(21)\_FI.7 + +Two scatter plots comparing subjective and predicted loudness levels using the ISO 532-1 stationary model. The left plot is for 'Hands-free' mode with an RMSE of 9.84. The right plot is for 'Handset' mode with an RMSE of 5.88. Both plots show a positive correlation between subjective and objective loudness levels, with a dashed diagonal line representing perfect prediction. + +**Figure I.7 – Subjective loudness level vs predicted loudness level using ISO 532-1 stationary model** + +![Four scatter plots comparing subjective and predicted loudness levels using the ISO 532-1 time-varying model. The top row shows 'Mean' aggregation: Hands-free (RMSE: 5.64) and Handset (RMSE: 3.29). The bottom row shows 'N5' aggregation: Hands-free (RMSE: 10.87) and Handset (RMSE: 6.62). All plots show a positive correlation between subjective and objective loudness levels, with a dashed diagonal line representing perfect prediction.](6e5a85131eedf6b98db62877ee64506e_img.jpg) + +Figure I.8 displays four scatter plots comparing subjective loudness level [phon] (X-axis) versus objective loudness level [phon] (Y-axis) using the ISO 532-1 time-varying model. + +- ISO 532-1 - Hands-free - Mean:** The plot shows a scatter of data points with an RMSE of 5.64. The axes range from 30 to 100 phon. +- ISO 532-1 - Handset - Mean:** The plot shows a scatter of data points with an RMSE of 3.29. The axes range from 30 to 100 phon. +- ISO 532-1 - Hands-free - N5:** The plot shows a scatter of data points with an RMSE of 10.87. The axes range from 30 to 100 phon. +- ISO 532-1 - Handset - N5:** The plot shows a scatter of data points with an RMSE of 6.62. The axes range from 30 to 100 phon. + +P.700(21)\_FI.8 + +Four scatter plots comparing subjective and predicted loudness levels using the ISO 532-1 time-varying model. The top row shows 'Mean' aggregation: Hands-free (RMSE: 5.64) and Handset (RMSE: 3.29). The bottom row shows 'N5' aggregation: Hands-free (RMSE: 10.87) and Handset (RMSE: 6.62). All plots show a positive correlation between subjective and objective loudness levels, with a dashed diagonal line representing perfect prediction. + +**Figure I.8 – Subjective loudness level vs predicted loudness level using ISO 532-1 time-varying model** + +Figure I.9 and Figure I.10 display the performance of the Moore-Glasberg model on predicting the perceived loudness values. The Moore-Glasberg model can take the eardrum pressure as the input signal, and therefore the stimuli are used directly for the calculation. For handset mode, the model automatically calculates the binaural loudness by providing one of the ear signals with zero values. [b-ISO 532-2] is based on the Moore-Glasberg stationary loudness model and does not perform well compared to other algorithms. The short-term Moore-Glasberg time-varying model performs the best among the Moore-Glasberg models when it is combined with the arithmetic mean aggregation. The model seems to overestimate the loudness values. + +![Two scatter plots comparing subjective and predicted loudness levels for ISO 532-2. The left plot is for 'Hands-free' mode with an RMSE of 14.07. The right plot is for 'Handset' mode with an RMSE of 12.31. Both plots show a positive correlation between subjective and objective loudness levels, with data points generally above the 1:1 line, indicating overestimation.](4e5abec2ae85a6a3a1366c722daceccd_img.jpg) + +The figure consists of two side-by-side scatter plots. Both plots have 'Subjective loudness level [phon]' on the x-axis and 'Objective loudness level [phon]' on the y-axis, with axes ranging from 30 to 100. A dashed diagonal line represents the 1:1 relationship. The left plot, titled 'ISO 532-2 - Hands-free', shows data points with an RMSE of 14.07. The right plot, titled 'ISO 532-2 - Handset', shows data points with an RMSE of 12.31. In both cases, the data points are clustered around the diagonal line but tend to be above it, indicating that the model tends to overestimate the loudness level. + +Two scatter plots comparing subjective and predicted loudness levels for ISO 532-2. The left plot is for 'Hands-free' mode with an RMSE of 14.07. The right plot is for 'Handset' mode with an RMSE of 12.31. Both plots show a positive correlation between subjective and objective loudness levels, with data points generally above the 1:1 line, indicating overestimation. + +P.700(21)\_FI.9 + +**Figure I.9 – Subjective loudness level vs predicted loudness level using ISO 532-2** + +![Four scatter plots showing subjective vs. predicted loudness for MG model. Top-left: MG - Hands-free - Mean (RMSE: 9.74). Top-right: MG - Handset - Mean (RMSE: 6.72). Bottom-left: MG - Hands-free - N5 (RMSE: 15.07). Bottom-right: MG - Handset - N5 (RMSE: 14.14). All plots have axes from 30 to 100 phon.](4cec89a753c447a050c0171c274f2acb_img.jpg) + +The figure consists of four scatter plots arranged in a 2x2 grid, each showing the relationship between subjective and predicted loudness levels in phon. A dashed diagonal line represents the ideal 1:1 prediction. The Root Mean Square Error (RMSE) is provided for each plot. + +| Plot Title | RMSE | +|------------------------|-------| +| MG - Hands-free - Mean | 9.74 | +| MG - Handset - Mean | 6.72 | +| MG - Hands-free - N5 | 15.07 | +| MG - Handset - N5 | 14.14 | + +Four scatter plots showing subjective vs. predicted loudness for MG model. Top-left: MG - Hands-free - Mean (RMSE: 9.74). Top-right: MG - Handset - Mean (RMSE: 6.72). Bottom-left: MG - Hands-free - N5 (RMSE: 15.07). Bottom-right: MG - Handset - N5 (RMSE: 14.14). All plots have axes from 30 to 100 phon. + +P.700(21)\_FI.10 + +**Figure I.10 – Subjective loudness level vs predicted loudness level using Moore-Glasberg STL model** + +#### I.3.3 Noisy speech result + +The partial loudness algorithm requires two input signals, i.e., a clean speech signal and a noise alone signal. For the current investigation, the noise alone signals are not recorded by the mobile phone, rather, the raw background noise signal is directly mixed with the recorded speech signals. This is mainly due to the unknown effect of AGC in the mobile phone. For this reason, the partial loudness scenario in the investigation fits better with a situation where there is background noise present in the receiving part. + +The general tendency of the partial loudness model is similar to the prediction results of the Moore-Glasberg model for the clean stimuli. As shown in Figure I.11 and Figure I.12, the arithmetic mean aggregation of the short-term time-varying loudness seems to perform the best in the prediction. The overestimation is clearer for the stationary model and for N5 of the time-varying model. The results confirm that the partial loudness algorithm works as well as the loudness models for the clean stimuli and thus can be used practically. + +![Two scatter plots comparing subjective and predicted loudness levels for the ISO 532-2 algorithm. The left plot is for 'Hands-free' with an RMSE of 11.72. The right plot is for 'Handset' with an RMSE of 10.20. Both plots show a positive correlation between subjective and objective loudness levels, with data points clustered around a dashed diagonal line.](a2dcc4a0703102026ec86e82caa4985e_img.jpg) + +Partial - ISO 532-2 - Hands-free + +RMSE: 11.72 + +Objective loudness level [phon] + +Subjective loudness level [phon] + +Partial - ISO 532-2 - Handset + +RMSE: 10.20 + +Objective loudness level [phon] + +Subjective loudness level [phon] + +P.700(21)\_FI.11 + +Two scatter plots comparing subjective and predicted loudness levels for the ISO 532-2 algorithm. The left plot is for 'Hands-free' with an RMSE of 11.72. The right plot is for 'Handset' with an RMSE of 10.20. Both plots show a positive correlation between subjective and objective loudness levels, with data points clustered around a dashed diagonal line. + +**Figure I.11 – Subjective loudness level vs predicted loudness level using the partial loudness algorithm implemented in ISO 532-2** + +![Four scatter plots comparing subjective and predicted loudness levels for the Moore-Glasberg algorithm for short-term time-varying loudness. The plots are arranged in a 2x2 grid: 'Hands-free - Mean' (RMSE: 7.44), 'Handset - Mean' (RMSE: 4.15), 'Hands-free - N5' (RMSE: 12.10), and 'Handset - N5' (RMSE: 12.86). Each plot shows a positive correlation between subjective and objective loudness levels, with data points clustered around a dashed diagonal line.](7119b28e39fa3784606bf8b8f44e4f9d_img.jpg) + +Partial - MG - Hands-free - Mean + +RMSE: 7.44 + +Objective loudness level [phon] + +Subjective loudness level [phon] + +Partial - MG - Handset - Mean + +RMSE: 4.15 + +Objective loudness level [phon] + +Subjective loudness level [phon] + +Partial - MG - Hands-free - N5 + +RMSE: 12.10 + +Objective loudness level [phon] + +Subjective loudness level [phon] + +Partial - MG - Handset - N5 + +RMSE: 12.86 + +Objective loudness level [phon] + +Subjective loudness level [phon] + +P.700(21)\_FI.12 + +Four scatter plots comparing subjective and predicted loudness levels for the Moore-Glasberg algorithm for short-term time-varying loudness. The plots are arranged in a 2x2 grid: 'Hands-free - Mean' (RMSE: 7.44), 'Handset - Mean' (RMSE: 4.15), 'Hands-free - N5' (RMSE: 12.10), and 'Handset - N5' (RMSE: 12.86). Each plot shows a positive correlation between subjective and objective loudness levels, with data points clustered around a dashed diagonal line. + +**Figure I.12 – Subjective loudness level vs predicted loudness level using the partial loudness algorithm implemented in Moore-Glasberg for short-term time-varying loudness** + +### I.4 Conclusion + +A subjective loudness experiment was performed both for clean speech and music stimuli as well as noisy speech stimuli. A mobile phone supporting the option of narrowband, Wideband and SWB telephony audio bandwidths was used for recording the stimuli. The level of the stimuli and the + +background noise was controlled to simulate different exposure levels of the stimuli. For converting the subjective loudness judgements into loudness level, 1 kHz tones with different levels are presented diotically. + +For the clean stimuli, ISO 532-1 time-varying model with the arithmetic mean aggregation seems to perform the best in comparison with other models. The N5 aggregation overestimates the loudness values slightly. This may be caused by the fact that N5 is the 95th percentile and may give rise to noticeably higher values compared to the arithmetic mean for non-stationary time signals. + +For the noisy stimuli, only high SNRs were tested. For these SNRs, the arithmetic mean of Moore-Glasberg STL seems to achieve the best performance, and the resulting RMSEs are comparable with the ones for the clean stimuli. + +The [b-ISO 532-1] model can be recommended for testing the loudness of clean sound samples in telecommunication devices based on the experimental conditions employed in the current investigation. However, for the partial loudness paradigm, the arithmetic mean of the Moore-Glasberg STL model can be recommended despite the fact that the method is not standardized. + +In most of the conditions, hands-free mode resulted in higher RMSEs compared to handset mode. Therefore, it may be interesting to investigate the influence of binaural loudness summation. The current investigation focused on evaluating absolute loudness errors, but it would be necessary to discuss in the working group whether relative loudness errors are good enough for practical applications. + +# Appendix II + +## Result of loudness experiment B + +(This appendix does not form an integral part of this Recommendation.) + +### II.1 Introduction to loudness experiment B + +One of the initial proposals for P.Loudness included loudness calculation methods for handset and hands-free scenarios. However, the original experiments presented in [b-AES E-Library] for handset and hands-free only contained coded speech signals. Neither acoustic playback via loudspeaker systems (which may include strong non-linear effects) nor any other typical signal processing (e.g., speech compression) were involved. Additionally, the analysis of terminals evaluated using common HATS set-ups as used in terminal testing was not yet considered. + +In previous contributions [b-AES E-Library], two measurement series were introduced to cover these issues. Handset and hands-free application modes were evaluated as typical downlink scenarios. Multiple technologies such as digital enhanced cordless telecommunications (DECT), voice over IP (VoIP), 3G and 4G as well as certain state-of-the-art codecs were covered by the measurements. + +All recordings were then evaluated in listening tests obtaining absolute loudness values. An evaluation framework as described in [b-AES E-Library] was used here. This contribution presents results including 20 test subjects per mode (previously only 15). Additionally, several suggestions made in [b-AES E-Library] are taken into account in order to obtain comparable results over different experiments, e.g., by using multiple reference sounds. + +In this update, the instrumental loudness assessment is extended with regard to the additional reference sounds. In addition, a simplified method for the assessment on the proposed equal-reference-level scale is introduced. + +### II.2 Measurement set-up + +According to [b-AES E-Library], the measurement set-up should contain a wide range of real devices, including the audio bandwidths narrowband, Wideband, SWB and fullband. + +Since there are only a few real terminals currently available capable of SWB and fullband, a mobile mock-up in handset and (handheld) hands-free mode was additionally used for the recording procedure. Figure II.1 exemplarily depicts the usage of the mock-up device for both scenarios. The mock-up consists of two different micro-loudspeakers on front (handset) and back (hands-free) side. In these cases, the test signals are coded and decoded by certain codecs in advance to simulate a typical device. In addition, a modified version of the mock-up was used. The frequency response of the device was equalized to obtain a flat frequency response in handset and hands-free mode. + +![Figure II.1 shows two mock-up head and torso simulators (HATS) used for acoustic testing. The left image shows the HATS mounted in a handset mode, with a mobile device held in a clamp against the ear. The right image shows the HATS mounted in a handheld hands-free mode, with the mobile device held in a clamp against the mouth.](84cebc242b1af66723a02cdf67fc98c9_img.jpg) + +Figure II.1 shows two mock-up head and torso simulators (HATS) used for acoustic testing. The left image shows the HATS mounted in a handset mode, with a mobile device held in a clamp against the ear. The right image shows the HATS mounted in a handheld hands-free mode, with the mobile device held in a clamp against the mouth. + +**Figure II.1 – Mock-up mounted in handset (left) and handheld hands-free (right) mode** + +As described in [b-AES E-Library], a wide but also typical level/loudness range should be considered in the evaluation. Varying levels should mainly be obtained by using volume control of the terminal. The electrical level for the insertion into the device was fixed to $-16.0 \text{ dBm0}$ . + +For the current evaluation, each device was first measured with the maximum volume setting. To avoid overly extreme playback levels in the listening test, the volume was decreased to $90 \text{ dB SPL}$ for handset/monaural and $85 \text{ dB SPL}$ for hands-free/binaural recordings. These conditions are labelled as MAX in the following. + +Per device, two additional volume settings were recorded. Similar to the original work in [b-AES E-Library], level offsets of $-5 \text{ dB}$ and $-15 \text{ dB}$ relative to the MAX condition were collected. Since not all devices were able to reach the maximum level of $85/90 \text{ dB SPL}$ , a wide and continuous range of loudness conditions is obtained by this method. + +In contrast to the original work, a modified active speech level calculation according to [ITU-T P.56] is used for determining physical level in $\text{dBPa}$ and $\text{dB SPL}$ . This method is capable of excluding speech pauses from the level calculation and thus does not require exact cropping of samples. Since the calculation of level according to [ITU-T P.56] is only defined for signals including an amplitude of $\pm 1.0$ (digital full scale), a simple modification is applied: + +- If the signal to be analysed includes samples of absolute amplitude lower than $1.0 \text{ Pa}$ , the standard ITU-T P.56 algorithm is utilized to obtain the active speech level directly in $\text{dBPa}$ or $\text{dB SPL}$ . +- If the signal to be analysed includes samples of absolute amplitude higher than $1.0 \text{ Pa}$ , a scaling factor $a$ is determined to reduce the maximum amplitude to $1.0$ . Then the standard ITU-T P.56 algorithm is evaluated on the scaled version of the signal. The resulting active speech level is corrected by $20 \cdot \log_{10}(a)$ to obtain a value in $\text{dBPa}$ or $\text{dB SPL}$ . + +In this work, neither music/audio material nor noisy speech is used. Even though both stimuli type will become interesting especially in (mobile) SWB/ fullband scenarios, the current investigations are intentionally limited to this certain scenario (noise-free transmitted speech). + +#### II.2.1 Speech sequence + +As described in [b-AES E-Library], German speech material according to [ITU-T P.501] was used for the recordings. The language was selected because of the mother tongue of the test subjects. + +To include another typical processing, compression of the input signal was also evaluated. [ITU-T P.501] provides exemplarily compressed speech for British English but not for German material. Since the origin of the processing as well as an adequate description of the compression method is missing, an open source implementation was used for this purpose [b-github]. + +Figure II.2 illustrates the speech sequence used for the measurement for all devices (handset and hands-free). To collect a reasonable number of samples for an auditory evaluation, four samples per condition were used. For the handset listening test, only the even (s2) and for the hands-free scenario the odd (s1) sentences were selected. + +![Figure II.2: Waveform plot showing speech sequences for handset (grey) and hands-free (blue) conditions. The plot displays amplitude over time (t/s) from 30 to 60 seconds. The y-axis ranges from -1.25 to 1.25. The sequences are labeled m1s1, m1s2, m2s1, m2s2, f1s1, f1s2, f2s1, f2s2. The blue regions indicate hands-free samples, and the grey regions indicate handset samples.](9ac04bd96ff7aa9e60639b5f5a63ed40_img.jpg) + +ITU-T P.501 + +P.700(19)\_FI1.2 + +Figure II.2: Waveform plot showing speech sequences for handset (grey) and hands-free (blue) conditions. The plot displays amplitude over time (t/s) from 30 to 60 seconds. The y-axis ranges from -1.25 to 1.25. The sequences are labeled m1s1, m1s2, m2s1, m2s2, f1s1, f1s2, f2s1, f2s2. The blue regions indicate hands-free samples, and the grey regions indicate handset samples. + +**Figure II.2 – Sentences used for handset (grey) and hands-free (blue)** + +In contrast to the work in [b-AES E-Library], in this series rather short speech samples were used. The source file used for the measurements was composed according to the guidelines of [b-3GPP SA4]. Here the sentences are centred in a time window of 4.0 s. Each sample includes a speech portion between 2.0 and 3.0 s (activity between 50% and 75%). + +#### II.2.2 Handset mode + +Several terminals according to Table II.1 were evaluated for generating the handset conditions. As references to the previous experiments in [b-AES E-Library], the pure speech files were also included in the corpus. + +The handset recordings were conducted using diffuse-field equalization for both artificial ears. Even though no signal is active on the left ear, it was recorded and played back in the listening test to obtain the same/realistic idle noise on both ears. Similar to [b-AES E-Library], diffuse-field equalized playback was also used for the presentation in the listening test. + +**Table II.1 – Devices and operational modes used for handset evaluation** + +| DUT No. | Type/Net | Codec | Normal/Compressed | Vol. setting | +|---------|----------|-------------|-------------------|--------------| +| 1 | DECT | G.726 | ✓ / ✓ | MAX, -15 | +| 2 | DECT | G.722 | ✓ / ✓ | -15 | +| 3 | 3G | AMR-NB | ✓ / ✓ | -5, -15 | +| 4 | 3G | AMR-WB | ✓ / ✓ | -5, -15 | +| 5 | 4G | EVS-SWB | ✓ / ✓ | MAX, -5, -15 | +| 6 | VoIP | G.711u | ✓ / ✓ | MAX, -5, -15 | +| 7 | VoIP | G.722 | ✓ / ✓ | MAX, -5, -15 | +| 8 | Mock-up | No Codec-NB | ✓ / ✓ | MAX | + +**Table II.1 – Devices and operational modes used for handset evaluation** + +| DUT No. | Type/Net | Codec | Normal/Compressed | Vol. setting | +|---------|-------------------|-------------|-------------------|--------------------| +| 9 | Mock-up | No Codec-WB | ✓ / ✓ | MAX | +| 10 | Mock-up | Opus-SWB | ✗ / ✓ | MAX | +| 11 | Mock-up | EVS-SWB | ✓ / ✗ | MAX | +| 12 | Mock-up | Opus-FB | ✗ / ✓ | –5 | +| 13 | Mock-up | EVS-FB | ✓ / ✗ | –5 | +| 14 | Mock-up equalized | No Codec-NB | ✓ / ✓ | MAX | +| 15 | Mock-up equalized | No Codec-WB | ✓ / ✓ | MAX | +| 16 | Mock-up equalized | EVS-SWB | ✓ / ✓ | MAX, –5, –15 | +| 17 | Mock-up equalized | Opus-FB | ✓ / ✓ | MAX, –5, –15 | +| 18 | Reference | No Codec | ✓ / ✓ | 45, 50, 55, 60, 75 | + +#### **II.2.3 Hands-free mode** + +Similar to handset scenarios, several terminals were evaluated according to Table II.2. Since for this application there are even fewer SWB/ fullband -capable devices available, a loudspeaker in conjunction with pre-coded signals was additionally used here. As for the handset scenario, the clean speech sentences were also included as a reference. + +For the binaural hands-free recordings, free-field equalization was selected for both artificial ears. + +**Table II.2 – Devices and operational modes used for hands-free evaluation** + +| DUT No. | Type/Net | Codec | Normal/Compressed | Vol. setting | +|---------|--------------|-------------|-------------------|--------------| +| 1 | DECT | G.726 | ✓ / ✗ | MAX | +| 2 | DECT | G.722 | ✗ / ✓ | MAX, –5 | +| 3 | Mobile 3G | AMR-NB | ✓ / ✓ | MAX, –5, –15 | +| 4 | Mobile 3G | AMR-WB | ✓ / ✓ | MAX, –5, –15 | +| 5 | Mobile 4G | EVS-SWB | ✓ / ✓ | –5, –15 | +| 6 | Desktop VoIP | G.711u | ✗ / ✓ | MAX, –15 | +| 7 | Desktop VoIP | G.722 | ✓ / ✓ | MAX, –15 | +| 8 | Loudspeaker | No Codec-NB | ✓ / ✓ | MAX | +| 9 | Loudspeaker | No Codec-WB | ✓ / ✓ | MAX | +| 10 | Loudspeaker | Opus-SWB | ✓ / ✗ | –15 | +| 11 | Loudspeaker | EVS-SWB | ✓ / ✓ | MAX, –5 | +| 12 | Loudspeaker | Opus-FB | ✓ / ✗ | MAX | +| 13 | Loudspeaker | EVS-FB | ✓ / ✓ | –5, –15 | +| 14 | Car HFT 3G | AMR-NB | ✓ / ✓ | MAX, –15 | +| 15 | Car HFT 3G | AMR-WB | ✓ / ✓ | MAX, –15 | +| 16 | Mock-up | No Codec-NB | ✓ / ✗ | MAX | + +**Table II.2 – Devices and operational modes used for hands-free evaluation** + +| DUT No. | Type/Net | Codec | Normal/Compressed | Vol. setting | +|---------|-------------------|-------------|-------------------|---------------| +| 17 | Mock-up | No Codec-WB | ✗ / ✓ | MAX | +| 18 | Mock-up | Opus-SWB | ✗ / ✓ | MAX | +| 19 | Mock-up | Opus-FB | ✓ / ✓ | MAX, -5 | +| 20 | Mock-up equalized | No Codec-NB | ✓ / ✗ | MAX | +| 21 | Mock-up equalized | No Codec-WB | ✓ / ✗ | MAX | +| 22 | Mock-up equalized | Opus-FB | ✓ / ✗ | MAX | +| 23 | Mock-up equalized | EVS-FB | ✗ / ✓ | MAX | +| 24 | Reference | No Codec | ✓ / ✓ | 40, 55, 70 dB | + +| | | | | +|----|-----------------|----|---------------| +| 24 | Extremely loud | 24 | Extrem laut | +| 23 | | 23 | | +| 22 | | 22 | | +| 21 | | 21 | | +| 20 | Very loud | 20 | Sehr laut | +| 19 | | 19 | | +| 18 | | 18 | | +| 17 | | 17 | | +| 16 | Rather loud | 16 | Ziemlich laut | +| 15 | | 15 | | +| 14 | | 14 | | +| 13 | | 13 | | +| 12 | Moderately loud | 12 | Mittelmäßig | +| 11 | | 11 | | +| 10 | | 10 | | +| 9 | | 9 | | +| 8 | Quiet | 8 | Leise | +| 7 | | 7 | | +| 6 | | 6 | | +| 5 | | 5 | | +| 4 | Very quiet | 4 | Sehr leise | +| 3 | | 3 | | +| 2 | | 2 | | +| 1 | | 1 | | +| 0 | Not audible | 0 | Nicht gehört | + +**Figure II.3 – The loudness scale in English (left) and in German (right) for auditory assessment** + +### II.3 Auditory loudness assessment + +The categorical 24-point loudness scale as shown in Figure II.3 and already used in [b-AES E-Library] was used for the loudness function assessment as well as for the main evaluation of the speech samples. + +#### II.3.1 Reference sounds and individual loudness functions + +The principle of loudness functions is to provide a mapping between auditory results (in averaged points) and a level and/or loudness unit. This transformation is conducted with a reference sound and not with speech stimuli themselves. + +In the discussion on recent loudness experiments, the selection of this reference sound was identified as a crucial issue in the auditory assessment procedure. In the original work of [b-AES E-Library], a noise burst of 1-Bark bandwidth at 1 kHz was introduced. The basic idea here was to avoid the usage of a sine tone, but to use a signal where the level directly corresponds to the unit phon. However, as already discussed in [b-Hots 2] and [b-Hots 1], the relation between sub-band level and phon seems unsustainable. In addition, the noise signal includes steep spectral slopes, which cause edge tones and a strong tonal character in the time domain. + +In the work presented in [b-AES E-Library] as well as in previous experiments [b-AES E-Library], a more suitable noise signal of 3-Bark bandwidth at 1 kHz was used for the determination of loudness functions. The intention here was to use a sound which has a similar loudness impression as speech and is easier to judge for the listener. The disadvantage of this reference sound is that there is no relation given between level and loudness. Auditory and instrumental results can only be provided and compared on an equal-reference-level scale (introduced as "equal-noise-level" in [b-AES E-Library]). + +The original work in psychoacoustics and loudness assessment always refers to a sine tone as the reference. A sine tone of frequency 1 kHz at 40 dB SPL level is defined as 1 sone or 40 phon. Thus, it seems reasonable to include this reference sound as well. The disadvantage of a sine tone is the tonal character and the increased annoyance at higher levels, which may diverge from the perceived impression of louder speech signals. + +In consequence and as an extension to the work already conducted in [b-AES E-Library], all three reference sounds are considered in this contribution for the analysis. Figure II.4 shows the spectra of the three sounds. + +![Figure II.4: FFTs reference signals. A line graph showing the frequency spectrum (L [dBFS] vs Frequency [Hz]) for three reference sounds: Sinus (blue), 1 Bark (orange), and 3 Bark (green). The Sinus signal is a narrow peak at 1000 Hz. The 1 Bark signal is a broader peak centered at 1000 Hz. The 3 Bark signal is a wide, noisy band centered around 1000 Hz, with a peak level around -20 dBFS.](1cd6a2d6e58e0e6d75c879a999be06ef_img.jpg) + +The graph displays the Fast Fourier Transform (FFT) of three reference signals. The y-axis represents the level $L$ in $\text{dB}_{\text{FS}}$ , ranging from -50 to -10. The x-axis represents the frequency in Hz, ranging from 200 to 4000. The legend identifies three series: Sinus (blue line), 1 Bark (orange line), and 3 Bark (green line). The Sinus signal shows a sharp peak at 1000 Hz reaching approximately -12 dB $_{\text{FS}}$ . The 1 Bark signal shows a broader peak at 1000 Hz reaching approximately -18 dB $_{\text{FS}}$ . The 3 Bark signal shows a wide, noisy band centered around 1000 Hz, with a peak level around -20 dB $_{\text{FS}}$ . + +Figure II.4: FFTs reference signals. A line graph showing the frequency spectrum (L [dBFS] vs Frequency [Hz]) for three reference sounds: Sinus (blue), 1 Bark (orange), and 3 Bark (green). The Sinus signal is a narrow peak at 1000 Hz. The 1 Bark signal is a broader peak centered at 1000 Hz. The 3 Bark signal is a wide, noisy band centered around 1000 Hz, with a peak level around -20 dBFS. + +**Figure II.4 – Reference stimuli of current (green) and original (orange) work (FFT – fast Fourier transform)** + +For the determination of loudness functions per test subject, reference sounds with a duration of 4.0 s were used. Overall, ten discrete noise levels between 45 and 90 dB SPL (handset) and 40 and 85 + +dB SPL (hands-free) were used in steps of 5 dB for each reference stimulus. The ten levels were evaluated six times, which led to 60 trials for the loudness function assessment. An additional initial training phase with four samples according to [b-AES E-Library] were included. + +The reference sounds were randomized in a special presentation order. The absolute level difference between two consecutive stimuli was ensured to be lower than 20 dB, which prevents too large jumps in level. + +For the transformation from points to equal noise level, a sigmoid function according to Equation II.1 is fitted against the averaged auditory data: + +$$L(p) = p_{min} + \frac{(p_{max} - p_{min})}{1 + e^{a \cdot (p_0 - p)}} \quad (\text{II.1})$$ + +As already mentioned in [b-AES E-Library], the sigmoid function always provides a positive gradient and thus is more reliable than a polynomial fit. The obtained fitting curves for all reference sounds and both modes are shown in Figure II.5 (with inverted axes). All curves provide a reasonable extrapolation to the minimum and maximum point scale. The parameters determined finally for all reference sounds are given in Table II.3. + +![Figure II.5: Derived sigmoidal functions for handset (left) and hands-free (right) scenarios. The figure contains two line graphs. The left graph is titled 'Loudness functions (HS)' and the right graph is titled 'Loudness functions (HF)'. Both graphs plot 'Loudness points' (y-axis, 0.0 to 20.0) against 'Reference level [dB SPL]' (x-axis, 45 to 85). Each graph shows three sigmoid curves: a blue line for 'Sigmoid function Noise3Bark', an orange line for 'Sigmoid function Noise1Bark', and a green line for 'Sigmoid function sine'. In the handset scenario (HS), the curves start around 5.0 at 45 dB SPL and rise to about 16.0 at 85 dB SPL. In the hands-free scenario (HF), the curves start around 7.0 at 45 dB SPL and rise to about 18.0 at 85 dB SPL. A small label 'P.700(19)_FII.5' is visible in the bottom right of the HF graph.](0aa15f5c9c3edae230985491199cfe8b_img.jpg) + +Figure II.5: Derived sigmoidal functions for handset (left) and hands-free (right) scenarios. The figure contains two line graphs. The left graph is titled 'Loudness functions (HS)' and the right graph is titled 'Loudness functions (HF)'. Both graphs plot 'Loudness points' (y-axis, 0.0 to 20.0) against 'Reference level [dB SPL]' (x-axis, 45 to 85). Each graph shows three sigmoid curves: a blue line for 'Sigmoid function Noise3Bark', an orange line for 'Sigmoid function Noise1Bark', and a green line for 'Sigmoid function sine'. In the handset scenario (HS), the curves start around 5.0 at 45 dB SPL and rise to about 16.0 at 85 dB SPL. In the hands-free scenario (HF), the curves start around 7.0 at 45 dB SPL and rise to about 18.0 at 85 dB SPL. A small label 'P.700(19)\_FII.5' is visible in the bottom right of the HF graph. + +Figure II.5 – Derived sigmoidal functions for handset (left) and hands-free (right) scenarios (HS – handset; HF – hands-free) + +### II.4 Instrumental loudness methods + +Several common loudness methods and temporal aggregations are evaluated in the following analysis. Recent studies such as [b-Schlittenlacher] indicate that aggregation of loudness vs time can improve the prediction performance of loudness models. Thus, methods that produce loudness-vs-time curves in particular are investigated in detail. + +Since the speech samples are centred and arranged in windows of 4.0 s (see Figure II.2), initial and trailing silence/idle noise portions are included in the samples. For the calculation of level, the method according to ITU-T P.56 is used to exclude these silence periods. For all loudness calculations, only active parts were taken into account. The voice activity detection used is based on the frame classification algorithm of [ITU-T G.160] Appendix II. + +Table II.3 – Sigmoid function parameters for fitting handset and hands-free experiments + +| Parameter | 1-Bark noise | | 3-Bark noise | | Sine tone | | +|-----------|--------------|------------|--------------|------------|-----------|------------| +| | Handset | Hands-free | Handset | Hands-free | Handset | Hands-free | +| $P_{min}$ | –180.0 | –170.0 | –19.092 | –78.325 | –180.0 | –3.062e–17 | + +**Table II.3 – Sigmoid function parameters for fitting handset and hands-free experiments** + +| Parameter | 1-Bark noise | | 3-Bark noise | | Sine tone | | +|-----------|--------------|--------|--------------|---------|-----------|-------| +| $P_{max}$ | 112.083 | 99.495 | 101.020 | 104.489 | 112.380 | 93.02 | +| $P_0$ | -9.012 | -7.277 | 3.404 | -2.685 | -8.424 | 6.875 | +| A | 0.094 | 0.113 | 0.150 | 0.096 | 0.093 | 0.211 | + +#### II.4.1 Loudness according to Zwicker + +The time-varying loudness for non-stationary sounds according to Zwicker was standardized e.g., as DIN 45631/A1 [b-DIN 45631] or recently as [b-ISO 532-1]. Only one Zwicker method should be studied in this evaluation, so the latest official version of ISO 532-1 is used. It is expected that the difference to DIN 45631/A1 will be negligible. + +The current ISO 532-1 model does not include a special binaural loudness assessment. Left and right channel are calculated separately, then an ideal loudness summation is conducted. + +#### II.4.2 Loudness according to Moore-Glasberg + +Several methods according to the approach of Moore-Glasberg can be considered for loudness calculation. + +##### II.4.2.1 ANSI S3.4-2007 + +The loudness for stationary sounds according to Moore-Glasberg is standardized in [b-ANSI S3.4-2007] and in [b-ISO 532-2]. In contrast to the time-varying loudness of Zwicker, this method does not provide a loudness-vs-time curve, so the output of this analysis is directly a loudness value in sone and/or phon. The P.Loudness proposal is based on a modification of this method. + +The loudness according to ANSI S3.4-2007 could also be evaluated per block, since an average fast Fourier transform (FFT) is anyway needed for the calculation. In the following, a block size of 4096 samples and 50% overlap is used to calculate a loudness-vs-time curve. As in [b-ISO 532-1], no binaural processing is included for the loudness assessment. + +##### II.4.2.2 Loudness for non-stationary sounds – version 2002 + +The well-known time-varying loudness of Moore-Glasberg according to [b-Glasberg 1] is not standardized, but the reference implementation provided at [b-APG] can be used to evaluate loudness for short- and long-term frames over time. + +As in the ISO 532-1 standard, no binaural processing is included for the loudness assessment. + +##### II.4.4.3 Binaural loudness assessment – version 2016 + +A recent enhancement of the aforementioned calculation method is provided in [b-Glasberg 3]. While the basic loudness assessment method in this new approach is almost identical to the one presented in [b-Glasberg 1], this model considers binaural inhibition between the left and right ears. For this study, the reference implementation provided at [b-APG] is used. + +#### II.4.3 Temporal aggregation of loudness vs time + +In this evaluation, three temporal aggregation functions for a loudness-vs-time curve are used to obtain a single value from the loudness distribution vs. time: + +- 1) arithmetic mean (Avg.) +- 2) 95% percentile (N5) +- 3) average loudness level $LL(p)$ according to [b-Fiebig] ( $N(i)$ represents loudness in sone at $i$ -th frame): + +$$LL(p) = \frac{1}{\log_{10}(2)} \cdot \sqrt{\frac{1}{K} \cdot \sum_{k=1}^K N(i)^{\frac{1}{\log_{10}(2)}}}$$ + +#### II.4.4 Binaural loudness summation + +All analyses described above are in general carried out in binaural mode: + +- In the handset mode, the left channel/ear only includes an idle channel. For the instrumental calculation, only the right channel was evaluated for the loudness models that only support perfect loudness summation. Whenever the loudness method under test implicitly applies loudness summation (i.e., assumes diotic listening of single channel file), the sone output value was multiplied by 0.5 in order to address this issue. In the case of Moore-Glasberg loudness with binaural inhibition, the left channel containing idle noise was explicitly inserted into the signal under test; +- In hands-free mode, the left and right channel are evaluated. In some implementations, each channel was regarded as a diotically presented signal. In these cases, loudness in sone for both channels was multiplied by 0.5 and then summed in order to obtain the overall loudness. Again, Moore-Glasberg loudness with binaural inhibition was the only model directly supporting two-channel recordings. + +### II.5 Auditory results + +Overall, 20 normal-hearing test persons conducted the listening tests for handset and hands-free modes. Each participant listened to all samples, which led to 20 votes per sample and 80 votes per condition (four samples per condition). Experts, experienced listeners and naïve subjects participated in the test. Due to time and privacy constraints, no audiometric pre- or post-screening or data rejection was performed. + +#### II.5.1 Equal reference level + +The basic idea behind the assessment of the loudness functions is to obtain auditory votes for a signal with known loudness (either sone or phon). The original work in [b-AES E-Library] assumed that the level (in dB SPL) of a noise signal of 1-Bark bandwidth around 1.0 kHz can be directly interpreted in the unit phon. As already discussed in [b-AES E-Library], this approach does not reflect up-to-date research on loudness perception. + +Due to this fact and to the different reference signals used for the determination of loudness functions, the assessed points per test stimulus cannot be directly transformed into sone or phon anymore. Results of the proposed test design are provided on an equal reference level (ERL) scale. + +Based on the results found in [b-AES E-Library], listening test results in points are first averaged per sample. Then these average loudness values are transformed with the sigmoid functions according to Equation 1 and Table II.3 to the ERL domain. All results in the following clauses will thus be provided on the dB ERL scale. + +NOTE – In case of the sine tone as the reference signal, the loudness results could also be interpreted directly in unit phon (when strictly following the definitions of psychoacoustics). However, to keep text and graphs in this document consistent over all reference sounds, dB ERL will be used for all further investigations. + +#### II.5.2 Equal reference level of arbitrary loudness models + +In order to compare results of this test design with instrumental loudness models, which usually provide output values in phon or sone, a transformation to the ERL domain must be conducted. In previous work [b-AES E-Library], it was proposed to scale the reference sound iteratively and target at the same loudness model output as the signal under test. + +Even though this method provides correct results, it may cause unnecessary computational effort. As an enhancement, a pre-computed mapping is proposed to convert between model output in sone/phon and auditory results on the ERL scale. + +The transformation function is determined by calculating sone/phon values of a certain reference sound, a loudness model and a temporal aggregate for a given level range (e.g., in steps of 1 dB). For the transformation, the mapping function according to Equation II.2 is proposed. Here $N$ denotes the calculated loudness in sone of a given model, single value aggregate and reference sound. + +$$ERL [dB] = y_{min} + (y_{max} - y_{min}) \cdot \ln((N - x_m) \cdot s) \quad (II.2)$$ + +The left graph of Figure II.6 illustrates the principle for handset mode, 3-Bark noise as reference signal and the time-varying loudness model of Moore-Glasberg with short-term smoothing. Here the coefficients can be determined by least mean square error optimization as: + +$$y_{min}, y_{max}, x_m, s = [69.647, \quad 88.542, \quad -0.826, \quad 0.103]$$ + +![Figure II.6: Transformation of instrumental (left) and auditory (right) results to ERL scale. The left graph shows 'Equal reference level [dB SPL]' vs 'TVL - ST Avg loudness [Sone]' for handset mode (HS Noise3Bark), with a legend for 'Fit TVL - ST Avg' (blue line) and 'Loudness TVL - ST Avg' (red dots). The right graph shows 'Equal reference level [dB SPL]' vs 'Loudness points' for hands-free mode (HF, Noise3Bark), with a legend for 'Avg' (black line) and 'Sigmoid fitting' (red line). Both graphs show a sigmoidal relationship between loudness and reference level.](74448f9178da618d823e5a5dadc56fb5_img.jpg) + +Figure II.6: Transformation of instrumental (left) and auditory (right) results to ERL scale. The left graph shows 'Equal reference level [dB SPL]' vs 'TVL - ST Avg loudness [Sone]' for handset mode (HS Noise3Bark), with a legend for 'Fit TVL - ST Avg' (blue line) and 'Loudness TVL - ST Avg' (red dots). The right graph shows 'Equal reference level [dB SPL]' vs 'Loudness points' for hands-free mode (HF, Noise3Bark), with a legend for 'Avg' (black line) and 'Sigmoid fitting' (red line). Both graphs show a sigmoidal relationship between loudness and reference level. + +**Figure II.6 – Transformation of instrumental (left) and auditory (right) results to ERL scale (HS handset; HF – hands-free)** + +As introduced in clause II.3, the auditory result points can be transformed to the ERL domain by an inverse loudness function. For the example described above, the corresponding curve is shown in the right side of Figure II.6. + +#### II.5.3 Comparison of auditory and instrumental results + +For each mode, results can be evaluated by multiple loudness models, single value aggregation methods and reference sounds. In order to provide an adequate overview, only the best-performing combinations for each loudness model are provided. + +For the determination of the prediction accuracy, epsilon-insensitive root mean square error (RMSE\*), according to [ITU-T P.1401], is used. This epsilon-insensitive error metric takes the uncertainty of the auditory data into account, i.e., using the 95% confidence interval as a threshold for the error calculation. Note that due to the non-linear transformation of the auditory points to dB ERL, the confidence interval is asymmetric. + +As introduced in clause II.2, the active speech level can also be evaluated as a measure of loudness. For the handset and hands-free scenarios, the sine tone ERL provides best results (2.7 dB for handset, 2.6 dB for hands-free). Scatter plots of auditory vs instrumental data are provided in Figure II.7. + +![Figure II.7: Comparison of auditory data vs ASL ITU-T P.56 of HS – handset (left) and HF – hands-free (right).](103781b498cc6846ffc11a041e14dc3e_img.jpg) + +Figure II.7 consists of two scatter plots comparing auditory data (x-axis, $dB_{ERL}$ ) against Active Speech Level (y-axis, $dB_{SPL}$ ) for two conditions: Handset (HS) on the left and Hands-free (HF) on the right. Both plots show a strong positive correlation between the two measurements. + +**Left Plot: ActiveSpeechLevel (HS, per condition)** + +- rank order $r_s = 0.914$ +- RMSE = 3.788 +- RMSE mapped = 2.624 +- RMSE\* = 2.726 + +Legend: + +- Mapping function ( $r = 0.962$ ) +- Equal Ref. level-Sine ( $r = 0.959$ ) + +**Right Plot: ActiveSpeechLevel (HF, per condition)** + +- rank order $r_s = 0.981$ +- RMSE = 4.210 +- RMSE mapped = 1.962 +- RMSE\* = 2.601 + +Legend: + +- Mapping function ( $r = 0.990$ ) +- Equal Ref. level-Sine ( $r = 0.977$ ) + +P.700(21)\_FII.7 + +Figure II.7: Comparison of auditory data vs ASL ITU-T P.56 of HS – handset (left) and HF – hands-free (right). + +**Figure II.7 – Comparison of auditory data vs ASL ITU-T P.56 of HS – handset (left) and HF – hands-free (right)** + +##### II.5.3.1 Handset mode + +Figure II.8 provides the best results for the handset mode: + +- ISO 532-1 (upper left): + - RMSE\* = 1.42 dB + - Single value calculation: average + - Reference sound: sine tone +- P.Loudness (lower left): + - RMSE\* = 1.42 dB + - Single value calculation: n/a + - Reference sound: 3-Bark noise +- Moore-Glasberg binaural loudness (LT), 2016 (upper right): + - RMSE\* = 1.97 dB + - Single value calculation: average + - Reference sound: 3-Bark noise +- Moore-Glasberg monaural loudness (ST), 2002 (lower right): + - RMSE\* = 1.73 dB + - Single value calculation: average + - Reference sound: 1-Bark noise + +![Four scatter plots comparing Auditory dB_ERL (x-axis) vs Instrumental dB_ERL (y-axis) for different loudness models: HEAD_ISO5321_Avg (HS, per condition), TVLBinaural_LT_Avg (HS, per condition), P.Loudness (HS, per condition), and MooreGlasberg_ST_Avg (HS, per condition). Each plot shows a dashed red line for the mapping function and blue dots for data points. The top-left plot (ISO 532-1) has a rank order r_s = 0.949, RMSE = 2.452, RMSE mapped = 2.034, and RMSE* = 1.418. The top-right plot (TVL Binaural) has a rank order r_s = 0.971, RMSE = 2.905, RMSE mapped = 1.854, and RMSE* = 1.965. The bottom-left plot (P.Loudness) has a rank order r_s = 0.984, RMSE = 2.443, RMSE mapped = 1.395, and RMSE* = 1.424. The bottom-right plot (Moore-Glasberg) has a rank order r_s = 0.947, RMSE = 2.766, RMSE mapped = 1.961, and RMSE* = 1.728. All plots show a strong positive correlation between auditory and instrumental loudness.](5fdd55201f3f090a54d10ffe9de2f068_img.jpg) + +Four scatter plots comparing Auditory dB\_ERL (x-axis) vs Instrumental dB\_ERL (y-axis) for different loudness models: HEAD\_ISO5321\_Avg (HS, per condition), TVLBinaural\_LT\_Avg (HS, per condition), P.Loudness (HS, per condition), and MooreGlasberg\_ST\_Avg (HS, per condition). Each plot shows a dashed red line for the mapping function and blue dots for data points. The top-left plot (ISO 532-1) has a rank order r\_s = 0.949, RMSE = 2.452, RMSE mapped = 2.034, and RMSE\* = 1.418. The top-right plot (TVL Binaural) has a rank order r\_s = 0.971, RMSE = 2.905, RMSE mapped = 1.854, and RMSE\* = 1.965. The bottom-left plot (P.Loudness) has a rank order r\_s = 0.984, RMSE = 2.443, RMSE mapped = 1.395, and RMSE\* = 1.424. The bottom-right plot (Moore-Glasberg) has a rank order r\_s = 0.947, RMSE = 2.766, RMSE mapped = 1.961, and RMSE\* = 1.728. All plots show a strong positive correlation between auditory and instrumental loudness. + +**Figure II.8 – Comparison of auditory data vs loudness models (HS – handset)** + +##### II.5.3.2 Hands-free mode + +Figure II.9 provides the best results for the hands-free mode: + +- ISO 532-1 (upper left): + - RMSE\* = 1.73 dB + - Single value calculation: average + - Reference sound: sine tone +- P.Loudness (lower left): + - RMSE\* = 1.70 dB + - Single value calculation: n/a + - Reference sound: 3-Bark noise +- Moore-Glasberg binaural loudness (LT), 2016 (upper right): + - RMSE\* = 1.86 dB + - Single value calculation: LL(p) + - Reference sound: 3-Bark noise +- Moore-Glasberg monaural loudness (LT), 2002 (lower right): + - RMSE\* = 1.91 dB + - Single value calculation: N5 + - Reference sound: 3-Bark noise + +![Four scatter plots comparing instrumental dB_ERL vs auditory dB_ERL for different loudness models: HEAD_ISO5321_Avg, TVLBinaural_LT_LLp, P.Loudness, and MooreGlasberg_LT_N5. Each plot shows data points for 'Equal Ref. Level-Sine' or 'Equal Ref. Level-Noise3Bark' and a 'Mapping function' line. Statistics like rank order r_s, RMSE, RMSE mapped, and RMSE* are provided for each model.](fab98272229adb4efe6412d8e556fb67_img.jpg) + +**HEAD\_ISO5321\_Avg (HF, per condition)** + +rank order $r_s = 0.987$ +RMSE = 3.184 +RMSE mapped = 1.723 +RMSE\* = 1.731 + +**TVLBinaural\_LT\_LLp (HF, per condition)** + +rank order $r_s = 0.984$ +RMSE = 2.920 +RMSE mapped = 1.970 +RMSE\* = 1.861 + +**P.Loudness (HF, per condition)** + +rank order $r_s = 0.984$ +RMSE = 2.878 +RMSE mapped = 1.845 +RMSE\* = 1.701 + +**MooreGlasberg\_LT\_N5 (HF, per condition)** + +rank order $r_s = 0.982$ +RMSE = 2.963 +RMSE mapped = 2.103 +RMSE\* = 1.905 + +P.700(21)\_FII.9 + +Four scatter plots comparing instrumental dB\_ERL vs auditory dB\_ERL for different loudness models: HEAD\_ISO5321\_Avg, TVLBinaural\_LT\_LLp, P.Loudness, and MooreGlasberg\_LT\_N5. Each plot shows data points for 'Equal Ref. Level-Sine' or 'Equal Ref. Level-Noise3Bark' and a 'Mapping function' line. Statistics like rank order r\_s, RMSE, RMSE mapped, and RMSE\* are provided for each model. + +**Figure II.9 – Comparison of auditory data vs loudness models (HF – hands-free)** + +### II.6 Direct loudness model output + +In the discussion of previous studies [b-AES E-Library], it was noted that representation of results on an equal-reference scale is not the preferred representation for several parties. The usage of the psychoacoustic unit phon seems to be more convenient here. Considering the original idea of the individual loudness functions in [b-AES E-Library], i.e., directly converting points to phon, all result comparisons can also be represented directly with the model output in phon. + +The psychoacoustic definition (40 dBSPL at 1 kHz corresponds to 40 phon) implies that only the equal-reference-level calculation of the sine tone can be used to convert from points to phon. However, bypassing the transformation of the model output (and thus, the context of the auditory test) may obviously lead to similar correlations, but also to some shifts and offsets. + +Since all results presented in the previous clauses were calculated in sone, a generic function according to Figure II.10 is used to convert from sone to phon. Certain loudness models may internally use slightly different implementations, but in general, the illustrated relation is a valid approximation for most model outputs. + +![A semi-logarithmic plot titled 'sone to phon' showing the conversion from sone to phon. The x-axis is 'sone' on a logarithmic scale from 10^-1 to 10^2. The y-axis is 'phon' on a linear scale from 0 to 100. A blue curve starts at approximately (0.1, 15) and rises to (100, 100). The plot is labeled P.700(19)_FII.10.](35135a8841884b7bb6fa58c3d9665c13_img.jpg) + +sone to phon + +$$b = (\log(10) \cdot 0.1 \cdot 0.27)^{-1}$$ + +$$d = e^{2.4/b}$$ + +$$a = e^{40/b} - d$$ + +$$phon = b \cdot \log(a \cdot sone + d)$$ + +A semi-logarithmic plot titled 'sone to phon' showing the conversion from sone to phon. The x-axis is 'sone' on a logarithmic scale from 10^-1 to 10^2. The y-axis is 'phon' on a linear scale from 0 to 100. A blue curve starts at approximately (0.1, 15) and rises to (100, 100). The plot is labeled P.700(19)\_FII.10. + +**Figure II.10 – Conversion from sone to phon** + +#### II.6.1 Handset mode + +Figure II.11 provides the best results for the handset mode in phon. For the ISO 532-1 and binaural Moore-Glasberg model, prediction performance slightly decreases. For monaural Moore-Glasberg, the comparison on the phon scale performs even better than on the equal-reference scale. On the other hand, the P.Loudness model obtains overly pessimistic sone/phon values. + +![Four scatter plots comparing auditory data (dB_EARL) against instrumental phon for different loudness models: HEAD_ISO5321_LLp, TVLBinaural_ST_Avg, P.Loudness, and MooreGlasberg_ST_LLp. Each plot includes a mapping function (red dashed line) and Equal Ref. Level-Sine data points (blue circles with error bars). RMSE* values are provided for each model: 3.019, 2.188, 18.356, and 1.348 respectively. The MooreGlasberg_ST_LLp model shows the lowest RMSE* value.](7e97a53863a8df3ff6b3eac89564ee58_img.jpg) + +The figure consists of four subplots arranged in a 2x2 grid, each showing the relationship between 'Auditory $dB_{EARL}$ ' (y-axis) and 'Instrumental phon' (x-axis) for different loudness models. Each plot includes a red dashed line representing the 'Mapping function' and blue circles with error bars representing 'Equal Ref. Level-Sine' data points. A black dashed line represents the 1:1 reference line. The Root Mean Square Error (RMSE\*) is indicated in each plot. + +| Model | RMSE* | +|------------------------------------------|--------| +| HEAD_ISO5321_LLp (HS, per condition) | 3.019 | +| TVLBinaural_ST_Avg (HS, per condition) | 2.188 | +| P.Loudness (HS, per condition) | 18.356 | +| MooreGlasberg_ST_LLp (HS, per condition) | 1.348 | + +Four scatter plots comparing auditory data (dB\_EARL) against instrumental phon for different loudness models: HEAD\_ISO5321\_LLp, TVLBinaural\_ST\_Avg, P.Loudness, and MooreGlasberg\_ST\_LLp. Each plot includes a mapping function (red dashed line) and Equal Ref. Level-Sine data points (blue circles with error bars). RMSE\* values are provided for each model: 3.019, 2.188, 18.356, and 1.348 respectively. The MooreGlasberg\_ST\_LLp model shows the lowest RMSE\* value. + +**Figure II.11 – Comparison of auditory data vs loudness models in phon (HS – handset)** + +#### II.6.2 Hands-free mode + +Figure II.12 provides the best results for the hands-free mode in phon. Again, the prediction accuracy of ISO 532-1 slightly decreases, but is still adequate. Since most of the listening test conditions did not include many level differences between the left and right ears, it is not surprising that monaural and binaural Moore-Glasberg loudness performs very similarly. Both models over-predict loudness in a similar way. The binaural model performs slightly better, possibly due to the binaural inhibition algorithm. The P.Loudness model for hands-free also over-predicts loudness. + +![Four scatter plots comparing auditory data (dB_EQL) vs instrumental phon for different loudness models: HEAD_ISO5321_Avg (HF), TVLBinaural_ST_Avg (HF), P.Loudness (HF), and MooreGlasberg_ST_Avg (HF). Each plot shows a dashed red line for the mapping function and blue dots for equal reference level-sine data. RMSE* values are provided for each model: 3.518, 7.181, 10.129, and 6.879 respectively. The x-axis is 'Instrumental phon' and the y-axis is 'Auditory dB_EQL', both ranging from 40 to 100. A dashed black line represents the 1:1 relationship.](25fab88fda841d5eec316c440656187e_img.jpg) + +Four scatter plots comparing auditory data (dB\_EQL) vs instrumental phon for different loudness models: HEAD\_ISO5321\_Avg (HF), TVLBinaural\_ST\_Avg (HF), P.Loudness (HF), and MooreGlasberg\_ST\_Avg (HF). Each plot shows a dashed red line for the mapping function and blue dots for equal reference level-sine data. RMSE\* values are provided for each model: 3.518, 7.181, 10.129, and 6.879 respectively. The x-axis is 'Instrumental phon' and the y-axis is 'Auditory dB\_EQL', both ranging from 40 to 100. A dashed black line represents the 1:1 relationship. + +**Figure II.12 – Comparison of auditory data vs loudness models in phon (HF – hands-free)** + +### II.7 Conclusions + +This appendix presents updated auditory and instrumental results of recently conducted listening tests. The test corpus is based on realistic binaural recordings of several terminals including an acoustic bandwidth from narrowband to fullband. + +The number of test subjects was increased from 15 to 20, which led to more accurate auditory results. In addition to the recently introduced 3-Bark reference noise, sine tone and 1-Bark noise were added to the assessment of individual loudness functions. In combination with the proposed transformation of the equal-reference level, multiple single value calculations are possible for each loudness model. + +As a prediction performance indicator, RMSE\* was introduced in this contribution in order to address the auditory uncertainty of the data. In the evaluation of loudness models, the well-known active speech level according to [ITU-T P.56] provides already a "baseline" performance of approximately 2.7 dB for both modes. + +The best-performing configurations of each loudness model category (Zwicker approach, Moore-Glasberg monaural and binaural, P.Loudness) provide accurate predictions with RMSE\* between 1.4 + +and 1.9 dB. However, compared to the active speech level calculation, the "gain" of loudness methods is only in the range of approximately 1 dB. + +Even though the P.Loudness proposal provides adequate prediction results (at least on the equal-reference-scale), the source prefers the usage of the standardized ISO 532-1 loudness model. Based on the observations presented in this study, several advantages are obvious: + +- The best RMSE\* for ISO 532-1 loudness model with average single value and sine tone as reference sound. +- It performs best on equal-reference-scale as well as on absolute phon scale. +- Performs best for both handset and hands-free mode. Thus, this method would solve the issue of not having two but just one single loudness model that works for both applications. +- In addition, this specific method is already widely tested by other standardization groups. It provides reference C-code, which is available for standard-conforming implementations. + +As an alternative model the binaural loudness model according to Moore-Glasberg described in [b-ISO 532-2] from 2017 could be considered. Even though the performance metrics were slightly worse, this model is currently the only one that considers binaural inhibition. Possibly, it is best prepared for more advanced hands-free scenarios, which contain increased spatial aspects. Such conditions were not part of the current evaluation. + +# Appendix III + +## Result of loudness experiment C + +(This appendix does not form an integral part of this Recommendation.) + +### III.1 Introduction to loudness experiment C + +#### III.1.1 Background + +The P.Loudness work item aims at specifying a loudness prediction model which is suitable for speech communication systems. It is not intended as a direct replacement of [ITU-T P.79] loudness ratings (SLR, receive loudness rating (RLR), junction loudness rating (JLR), overall loudness ratings (OLR)), which are a measure of sensitivity, but intended to predict the absolute loudness perceived by e.g., a user of a telephone. + +Generic loudness algorithms for a variety of signal types have existed for a long time and the focus of ITU-T SG12 is to validate their applicability for speech specifically and, in particular, to study ISO 532-1. + +#### III.1.2 Motivation + +One interesting usage of a loudness predictor is for assessing the output capability of mobile phones in handheld hands-free mode (speakerphone), where low frequencies are difficult to reproduce due to size constraints. The situation of comparing two mobile phones in speakerphone mode was addressed at a similar sound pressure level but with a different frequency response, or to compare two tunings of the same mobile phone. + +Loudness balancing has been used extensively in the past, however, such experiments were repeated but with consideration of: + +- Fullband context; +- no send-side filtering; +- sufficient cut-off frequency for high-pass filters. + +Recent studies have covered these conditions, but in combination with other conditions rather than being separately studied. + +If an objective model can correctly predict results also using the loudness-balancing test paradigm, it provides additional validation for that model. + +#### III.1.3 Earlier experiments within the P.Loudness work item + +Investigation of loudness of speech and other signals with standard narrowband, wideband and SWB filters/codecs was reported by Orange in [b-AES E-Library] and other contributions. This gives e.g., useful information about the effect of the low-pass at 14, 7 and 3.5 kHz and also the effect of high-pass at ~300 Hz. It also provides data at different listening levels. + +HEAD acoustics reported on loudness of speech in [b-AES E-Library] and other contributions, for a variety of speech bandwidths and also with real terminals and speech with dynamic compression. This gives useful validation data for a wide set of conditions. + +Brüel & Kjær, Orange and Delta reported on loudness of speech and noisy speech for a variety of conditions in [b-AES E-Library] and other contributions, also giving useful validation data. + +The subjects in the auditory experiments above rated the perceived loudness on a numeric loudness scale with an additional category labelling. Reference sounds of 1 kHz (tone or narrowband noise) were included to align with the phon scale directly or indirectly. + +At an earlier stage, Ericsson reported on a loudness-balancing experiment to assess the difference between narrowband, wideband, SWB and fullband in [b-AES E-Library]. The user interface was MUSHRA-like with faders for the volume of each sample. The present study is similar but with plus/minus buttons instead of faders, and other high-pass conditions. + +#### II.1.4 Historical survey + +For the purpose of putting the study into context, a small part of telephony-loudness history is provided in the following, including an analysis of the Geneva CCITT/ITU-T library. + +In the early days of telephony, achieving adequate loudness was a major concern and not always a simple task. The assessment of terminals was made by comparing the terminal to a reference connection with controlled characteristics. A talker would speak with reasonably consistent speech level and a listener would adjust an attenuator until equal loudness was perceived when comparing to the reference. The final value of the attenuator would be the test result, or a combination attenuator reading from various experiments. For more information on such methodologies, see [ITU-T P.78]. + +Instrumental measurements were later introduced using artificial mouths and ears. + +One interesting study is found in ITU-T SG12 contribution 56 from 1977, by France [b-CCITT 1977 SG12-C-0056]. The loudness-balancing concept was implemented instrumentally using the 1973 Zwicker loudness model ISO 532B. A good match to subjective data was reported; see Figure III.1. + +For all the measurements made, the statistical distribution of the differences between the subjective and the objective measurements has the following characteristics: + +Mean value: - 0.06 dB +Standard deviation: 0.66 dB + +![Scatter plot showing the relationship between subjective and objective measurements for loudness balancing. The x-axis is labeled 'Subjective measurements' and the y-axis is labeled 'Objective determinations', both in dB. A diagonal line represents perfect agreement. Data points are clustered around this line, showing good correlation.](1c4139270d1edff9631800a63eabf7f5_img.jpg) + +The figure is a scatter plot titled 'Objective determinations' on the y-axis and 'Subjective measurements' on the x-axis, both in dB. The y-axis ranges from -2 to 15 with major ticks at -2, 0, 5, 10, and 15. The x-axis ranges from -2 to 15 with major ticks at -2, 0, 5, 10, and 15. A solid diagonal line represents the line of perfect agreement (y=x). Data points, represented by small black dots, are scattered around this line, showing a strong positive correlation between subjective and objective measurements. The points are mostly clustered between -2 and 15 dB on both axes. + +Scatter plot showing the relationship between subjective and objective measurements for loudness balancing. The x-axis is labeled 'Subjective measurements' and the y-axis is labeled 'Objective determinations', both in dB. A diagonal line represents perfect agreement. Data points are clustered around this line, showing good correlation. + +**Figure III.1 – Result of the loudness-balancing concept by ITU-T SG12** + +This is continued in e.g., SG12 contribution 15 from 1982, from France [b-CCITT 1982 SG12-C-0015]. Again, a good match to subjective data was reported over a variety of handsets. However, at this time, speakerphone with micro-speakers was, for natural reasons, not included. + +It was found that although the ISO 532B method did produce a good result, it was considered "overkill" at the time. [ITU-T P.79] still states that "comprehensive models" are "unnecessarily complicated": + +"This model does not claim to represent accurately all the features that relate to the perception of the loudness of speech; for example, the effects of interfrequency masking are ignored and it does not predict the increasing importance of the lower frequencies as the intensity of the sound is increased from the threshold. It is possible to construct models that represent more of the features fairly well, but no completely comprehensive model is known. Such models are unnecessarily complicated for calculating loudness ratings. The most important restriction with respect to this model is that it should be used to make comparisons at the constant listening level indicated in [ITU-T P.76]." + +With various reasonable assumptions, it was possible to reduce the Zwicker-based calculations to what is now known as SLR, RLR, JLR and OLR. The sound pressure in each 1/3rd-octave band presented to the ear of a user is predicted by measuring the device under test and assuming certain network and far-end terminal characteristics; the weighting factors $W_i$ are used to consider a typical speech spectrum, the B-party frequency response. The frequency-dependent loudness perception (at the levels of interest) and the $m$ (loudness growth) factor is used to convert to a quantity that allows the contributions from different frequency bands to be properly summed (assuming certain typical presentation levels). With a minimum of calculations, a simplified version of the critical band summation in ISO R532B and other such loudness models was created in [ITU-T P.79]; see Equation 5-1 in Figure III.2. Finally, the result was converted back to decibels, revealing how much loss is needed to match a certain reference connection, as in the original subjective loudness-balancing method. + +$$LR = -\frac{10}{m} \cdot \log_{10} \sum_{i=N_1}^{N_2} 10^{0.1 \cdot m(S_i - W_i)}$$ + +**Figure III.2 – ITU-T P.79 Equation 5-1** + +To determine what weighting factors are to be used, loudness-balancing experiments were conducted. One example is found in SG12 contribution 10, from 1981, where China reported on experiments with high-pass and low-pass filters inserted in series with send and receive Nouveau Systeme Fondamental Europeen Reference Pour La Transmission Telephonique (NOSFER) characteristics. + +In conclusion, a small part of telephony-loudness history has been covered here. It is also worth mentioning that much early speech research is not described here, e.g., contributions by Stevens [b-Stevens] and by Fletcher and Galt, see e.g., [b-Fletcher]. + +In recent times, advanced loudness models based on Zwicker and Moore-Glasberg including not only spectral masking, but also temporal masking effects have been published. Recently, ISO published such a model in [b-ISO 532-1], based on Zwicker. + +![Figure III.3: Block diagrams and frequency response graph. (a) Recording: NOSFER sending system connected to a Tape recorder. (b) Loudness balance: Tape recorder connected to a network of filters (hid. loss, bal. att., HP/LF filters, 20 dB ampl.) and a 25 dB attenuator, then to a NOSFER sending system. Below is a graph of Level [dB] vs. Frequency [Hz] showing LP and HP filter responses.](62039fd2a1ef7b03f84c0fe5d8911860_img.jpg) + +a) Recording + +b) Loudness balance + +Figure III.3 is a composite figure. Part (a) shows a block diagram of a recording setup where a NOSFER sending system is connected to a tape recorder. Part (b) shows a block diagram of a loudness balance setup. A tape recorder is connected to a network of components: a hidden loss filter, a balance attenuator, HP/LF filters, and a 20 dB amplifier. This network is connected to a 25 dB attenuator, which is then connected to a NOSFER sending system. Below these diagrams is a graph showing the frequency response of the filters. The y-axis is Level [dB] from 0 to -40, and the x-axis is Frequency [Hz] on a logarithmic scale from 100 to 10000. Two curves are shown: LP (Low Pass) and HP (High Pass). The LP curve starts at 0 dB at 100 Hz and rolls off at 20 dB/decade after 1000 Hz. The HP curve starts at -40 dB at 100 Hz and rises at 20 dB/decade until it reaches 0 dB at 10000 Hz. + +Figure III.3: Block diagrams and frequency response graph. (a) Recording: NOSFER sending system connected to a Tape recorder. (b) Loudness balance: Tape recorder connected to a network of filters (hid. loss, bal. att., HP/LF filters, 20 dB ampl.) and a 25 dB attenuator, then to a NOSFER sending system. Below is a graph of Level [dB] vs. Frequency [Hz] showing LP and HP filter responses. + +Figure III.3 – From contribution 10, 1981, China + +![Figure III.4: Frequency response graphs for NOSFER send and receive characteristics. (a) Send characteristic showing a flat response from 200 Hz to 4 kHz with a 10 dB ripple. (b) Receive characteristic showing a flat response from 100 Hz to 10 kHz with a peak around 8 kHz.](4666738057044ad78ced4dbbc0c1bfb3_img.jpg) + +Figure III.4 consists of two graphs, (a) and (b). Graph (a) is labeled 'Send' and shows Level [dB] on the y-axis and Frequency [f] [Hz] on the x-axis. The x-axis has markers at 200, 4k, and 10k. The curve is relatively flat between 200 Hz and 4 kHz, with a vertical double-headed arrow indicating a 10 dB variation. Graph (b) is labeled 'Receive' and shows Level [dB] on the y-axis and Frequency [f] [Hz] on the x-axis. The x-axis has markers at 100, 1k, and 10k. The curve is relatively flat between 100 Hz and 10 kHz, with a slight peak around 8 kHz. + +Figure III.4: Frequency response graphs for NOSFER send and receive characteristics. (a) Send characteristic showing a flat response from 200 Hz to 4 kHz with a 10 dB ripple. (b) Receive characteristic showing a flat response from 100 Hz to 10 kHz with a peak around 8 kHz. + +Figure III.4 – NOSFER send and receive characteristics, from loudness rating and other measures of loudness, Brüel & Kjær + +### III.2 Subjective experiment + +#### III.2.1 Test design + +For each subtest, four sound files were available for playback by the subject: + +- original fullband (reference); +- three high-pass filtered versions (3rd-order Butterworth at 250, 500 and 1000 Hz). + +The reference was placed on the left of the user interface screen and the 250, 500 and 1000 Hz conditions followed from left to right, in randomized order for each subtest. + +A graphical user interface provided Play buttons for all files and +/- gain buttons for the high-pass filtered files. The resulting gain was not visible to the test subject. + +The reference file was fixed at 65 dB active speech level (per [ITU-T P.56]). + +The filtered files were first presented at a random level and then adjusted by the test subject in 1 dB steps until equal loudness of all four samples was perceived. The subject could repeat playback as wished. The final gain for each file was logged. The peak level was also logged for each rated file. + +The presentation was diotic (headphone presentation with same signal to both ears). + +#### III.2.2 Initial level randomization + +The goal was to randomize the initial level so that the user sometimes needed to raise and sometimes needed to reduce the level of the samples, avoiding guiding the listener in a certain direction. + +To this end, some values were calculated for each file using ISO 532-1 [b-Head acoustics], from 40 to 90 dBSPL presentation level. For conversion, back and forth between sone and SPL, a sigmoid function and a logarithmic function were fitted for each stimulus, to allow interpolation and conversion in both directions as needed, following [b-AES E-Library]. + +$$N(\text{Level}) [\text{sone}] = p_{\min} + \frac{(p_{\max} - p_{\min})}{1 + e^{a \cdot (p_0 - \text{Level})}} \quad (\text{III.1})$$ + +$$\text{Level}(N) [\text{dBSPL}] = y_{\min} + (y_{\max} - y_{\min}) \cdot \ln((N - x_m) \cdot s) \quad (\text{III.2})$$ + +The non-active parts of the files were removed using the speech activity detection from ITU-T P.56, before calculating the overall mean sone value. The ISO 532-1 parameters were: + +- type: 'time\_varying' +- field: 'D' + +From the baseline of expected "correct" level alignment (based on [b-ISO 532-1]), a further random gain was applied, up to $\pm 6$ dB uniform distribution. The reference file was always kept at 65 dB active speech level. + +The result was that samples were initially presented with a random loudness offset and the task of the test subject was to press "+" and "-" to bring the samples back to equal loudness, without seeing how much gain or loss they inserted. Hence, the protocol was similar to [b-AES E-Library] but with invisible gain/loss values. + +#### III.2.3 Headphone equalization and level adjustment + +Open-back headphones were placed on an ITU-T P.58 compliant HATS and the frequency response was measured using programme simulation noise (see [ITU-T P.381]), in 1/12th octaves. The [ITU-T P.58] Annex A DRP to diffuse-field correction was used. The headphones were re-seated five times and the curves were dB-averaged. The curves were then smoothed, averaging five bands (two below, two above) to avoid creating sharp peaks by the equalization. A 1000-tap FIR filter was fitted to the desired response. As a validation, the equalized headphone was then also measured using speech files and the response curve checked to be flat with 1/3rd band analysis. + +Finally, the playback level was adjusted to give 65 dB active speech level after diffuse-field equalization. (It was first planned to run the experiment at 73 dB active speech level but after a pilot test and to better match hands-free use cases, it was decided to use 65 dB ASL, as seen in some other contributions within P.Loudness.) + +#### III.2.4 Test environment + +Tests were conducted in an office-type lab room with ambient noise level below 30 dBSPL(A) background noise (typical 25 dB(A)). A computer with an RME soundcard, and a graphical user interface for user interaction, was used. + +#### III.2.5 Test sequence + +The ID, age and gender of the subject were collected. The instructions were given with text and graphics. A familiarization test preceded the actual test. The complete test duration was about 20 minutes, with some variation depending on the subject. + +#### III.2.6 Speech samples + +[ITU-T P.501] clause 7.3 British English samples were used, with three male plus three female single sentences plus one sentence for familiarization. The materials were selected from clause 7.3, rather than from Annex B, to include talkers with a low fundamental frequency. It is also the same material used in some other related studies, and in the currently specified loudness rating test in many standards. + +#### III.2.7 Assessors and screening + +Twenty-two naïve assessors participated in the test. They had no known hearing impairment, based on self-reporting. The subjects were all fluent in the English language and intelligibility was not considered to be an issue in the testing (clear and clean speech at high SNR). + +Post-screening criteria were based on the assumption that a 1000 Hz high-pass filter should warrant a higher compensation gain than 250 Hz. + +The confidence intervals were slightly reduced after screening. Remaining assessors are described in Tables III.1 and III.2 below. Heat maps of results are presented with and without screening, in Figure III.7 and Figure III.8. + +**Table III.1 – Assessors** + +| No of assessors | Total | Female | Male | +|-----------------|-------|---------|----------| +| | 18 | 7 (39%) | 11 (61%) | + +**Table III.2 – Age** + +| Age of assessors | Mean | Std dev | Min | Max | +|------------------|------|---------|-----|-----| +| | 26.3 | 7.6 | 19 | 44 | + +#### III.2.8 Instructions and familiarization + +Only written instructions were given. Subjects were permitted to ask questions at any time, but none of them asked any questions. + +The test included a test round for familiarization, using a different talker and sentence, from the same ITU-T P.501 corpus. + +#### III.2.9 Comments from subjects + +Subjects could comment after the test. Some expressed difficulty in deciding whether they perceived equal loudness or not. In a pilot test using trained subjects, a comment was made that for a high-pass filtered stimulus with the same overall loudness as the flat sample, the mids and highs were detected to be on a higher level and such a trained subject could then be tempted to rate this higher band level + +rather than the overall loudness. This could apply to loudness comparisons in general, whether or not it is a side-by-side comparison of actual products or whether it is a lab test such as this one. + +### III.3 Results and comparison to ISO 532-1 objective results + +#### III.3.1 Test results + +As in the classical loudness-balancing studies, the result is the gain needed for a filtered file to match the loudness of the original unfiltered file. The results are presented on a sample basis rather than condition basis, since the high-pass filtering affected different source material differently (e.g., female and male talkers). + +![Scatter plot showing Subjective [dB] vs Instrumental ISO 532-1 [dB] for high-pass filtered speech. The plot includes data points with 95% confidence intervals, a linear fit line (y = 1.0779 + 1.1899 * x), a 3rd-order polynomial fit curve, and a dashed black line representing the perfect fit (y=x).](91f1a66918183ac1f85c438198c09376_img.jpg) + +The figure is a scatter plot with error bars representing 95% confidence intervals. The x-axis is labeled 'Instrumental ISO 532-1 [dB]' and ranges from 0 to 14. The y-axis is labeled 'Subjective [dB]' and also ranges from 0 to 14. A dashed black line represents the perfect fit (y=x). A solid blue line represents a linear fit with the equation $y = 1.0779 + 1.1899 \cdot x$ . A red curve represents a 3rd-order polynomial fit. The data points are clustered between x=0 and x=10, with some points showing larger error bars at higher instrumental values. + +| Instrumental ISO 532-1 [dB] | Subjective [dB] | +|-----------------------------|-----------------| +| 0.5 | 1.2 | +| 1.0 | 2.5 | +| 1.2 | 3.5 | +| 1.5 | 2.8 | +| 1.8 | 1.5 | +| 2.2 | 3.8 | +| 3.2 | 5.2 | +| 3.5 | 6.0 | +| 4.0 | 5.8 | +| 5.5 | 6.2 | +| 6.2 | 8.2 | +| 6.5 | 9.8 | +| 6.8 | 10.0 | +| 7.0 | 11.0 | +| 8.5 | 10.5 | +| 10.0 | 11.5 | + +Scatter plot showing Subjective [dB] vs Instrumental ISO 532-1 [dB] for high-pass filtered speech. The plot includes data points with 95% confidence intervals, a linear fit line (y = 1.0779 + 1.1899 \* x), a 3rd-order polynomial fit curve, and a dashed black line representing the perfect fit (y=x). + +**Figure III.5 – Gain applied when loudness balancing high-pass filtered speech against the unfiltered reference. Subjective results with 95% confidence intervals are plotted vs ISO 532-1 instrumental results. Blue line: linear fit, red curve: 3rd-order polynomial fit, dashed black line: perfect fit** + +![Scatter plot showing Subjective [dB] vs Instrumental ISO 532-1 [dB] for high-pass filtered speech. The plot includes data points for female (red) and male (blue) talkers with 95% confidence intervals. A linear fit (blue line) and a 3rd-order polynomial fit (red curve) are shown, along with a dashed black line representing perfect fit. The linear fit equation is y = 1.0779 + 1.1899 * x.](119006c9a930f46881484f2229d3b927_img.jpg) + +Figure III.6 is a scatter plot showing the relationship between Subjective [dB] (Y-axis) and Instrumental ISO 532-1 [dB] (X-axis) for high-pass filtered speech. The X-axis ranges from 0 to 14, and the Y-axis ranges from 0 to 14. A dashed black line represents the perfect fit (y=x). A blue line represents the linear fit, with the equation $y = 1.0779 + 1.1899 \cdot x$ . A red curve represents the 3rd-order polynomial fit. Data points are plotted for various stimuli, with error bars indicating 95% confidence intervals. Red markers represent female talkers, and blue markers represent male talkers. The stimuli labels are: t1.HP250.wav, t2.HP250.wav, t3.HP250.wav, t4.HP250.wav, t5.HP250.wav, t6.HP250.wav, t1.HP500.wav, t2.HP500.wav, t3.HP500.wav, t4.HP500.wav, t5.HP500.wav, t6.HP500.wav, t1.HP1000.wav, t2.HP1000.wav, t3.HP1000.wav, t4.HP1000.wav, t5.HP1000.wav, t6.HP1000.wav. + +Scatter plot showing Subjective [dB] vs Instrumental ISO 532-1 [dB] for high-pass filtered speech. The plot includes data points for female (red) and male (blue) talkers with 95% confidence intervals. A linear fit (blue line) and a 3rd-order polynomial fit (red curve) are shown, along with a dashed black line representing perfect fit. The linear fit equation is y = 1.0779 + 1.1899 \* x. + +**Figure III.6 – Gain applied when loudness-balancing high-pass filtered speech against the unfiltered reference. Subjective results with 95% confidence intervals are plotted vs ISO 532-1 instrumental results, the same as the previous figure but with stimuli labels added. Red markers represent female talkers, blue markers represent male talkers. Blue line: linear fit, red curve: 3rd-order polynomial fit, dashed black line: perfect fit** + +![Heat map titled 'Count of GaindB vs. GainToMatchRefLoudness' for all subjects. The y-axis is GaindB ranging from -2 to 16. The x-axis is GainToMatchRefLoudness with values: 0.538, 1.2, 1.24, 1.37, 1.51, 1.94, 2.26, 3.22, 3.25, 3.77, 4.05, 5.46, 6.13, 6.34, 6.64, 6.67, 8.32, 9.8. A color bar on the right indicates frequency from 0 (dark blue) to 8 (yellow).](a5401b088b668d0554e8e96ea558121c_img.jpg) + +This heatmap displays the frequency of gain values for all subjects. The distribution shows a positive correlation between GainToMatchRefLoudness and GaindB. High-frequency clusters (yellow/light green) are visible around (1.24, 2), (1.51, 2), (1.94, 2), (3.77, 8), (4.05, 7), and (8.32, 10). + +Heat map titled 'Count of GaindB vs. GainToMatchRefLoudness' for all subjects. The y-axis is GaindB ranging from -2 to 16. The x-axis is GainToMatchRefLoudness with values: 0.538, 1.2, 1.24, 1.37, 1.51, 1.94, 2.26, 3.22, 3.25, 3.77, 4.05, 5.46, 6.13, 6.34, 6.64, 6.67, 8.32, 9.8. A color bar on the right indicates frequency from 0 (dark blue) to 8 (yellow). + +**Figure III.7 – Heat map of gains applied when loudness balancing high-pass filtered speech +(subjective vs ISO 532-1), all subjects. Lighter colours represent higher occurrence** + +![Heat map titled 'Count of GaindB vs. GainToMatchRefLoudness' for non-rejected subjects. The axes and color scale are identical to Figure III.7.](03afcee7dbcfc0af9eae2f7bf5eb6712_img.jpg) + +This heatmap displays the frequency of gain values for non-rejected subjects. The overall pattern is very similar to Figure III.7, indicating that the exclusion of rejected subjects does not drastically change the observed relationship between the variables. Peak occurrences remain in the same regions. + +Heat map titled 'Count of GaindB vs. GainToMatchRefLoudness' for non-rejected subjects. The axes and color scale are identical to Figure III.7. + +**Figure III.8 – Heat map of gains applied when loudness balancing high-pass filtered speech +(subjective vs ISO 532-1), non-rejected subjects. Lighter colours represent higher occurrence** + +Statistics based on [ITU-T P.1401], unmapped results: + +- Pearson's correlation coefficient: 0.96 +- RMSE: 2.2 dB +- RMSE\*: 1.2 dB + +For mean opinion score (MOS) experiments, it is described in [ITU-T P.1401] how to further map the data using up to 3rd-order polynomial fit, before comparing to the objective data. This is motivated by the *context effect* in each experiment. + +In this experiment, the task for the subjects is different, namely, to balance two sounds to each other. It is assumed that there is no similar context effect in such an experiment and therefore there is no attempt to re-map the data. + +Talker gender analysis: + +- The under-prediction of the model seems to depend on talker gender. (Hypothesis that talker gender does not matter is rejected in a Student's t-test, p value 3.9e-4). +- The high-pass filtering causes a larger drop in loudness for male talkers than for female talkers. Since this drop in loudness is underestimated by ISO 532-1, the prediction error for high-pass filter conditions is larger for male talkers. + +The data from this experiment is in reasonably good agreement with data from previous experiments, as summarized in Table III.3. + +**Table III.3 – Comparison to other studies with high-pass filtered speech. The conditions were not identical to the present study and therefore comparisons must be made with caution** + +| Other study | Condition | Other study value | This subjective experiment, 3 rd -order Butterworth -3dB point, 65 dBSPL | ISO 532-1, 3 rd -order Butterworth -3dB point, 65 dBSPL | +|-------------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------|-----------------------------------|-------------------------------------------------------------------------------------|--------------------------------------------------------------------| +| [b-CCITT 1981 SG12-C-0010]: steep high-pass, using NOSFER in send and receive, unknown presentation level, loudness balancing | 250 Hz steep high-pass | 2.5 dB | 2.4 dB | 1.3 dB | +| | 500 Hz steep high-pass | 7.5 dB | 5.4 dB | 3.7 dB | +| | 1000 Hz steep high-pass | 16.5 dB | 10.0 dB | 7.3 dB | +| [ITU-T G.191]: mod intermediate reference system (IRS) filter, 67dBSPL, point scoring with ref 1 kHz tones | 300 Hz to 3.4 kHz (using flat receive-side modified IRS) | 3 phon | (Most similar condition 2.4 dB) | (Most similar condition 1.3 dB) | +| [b-AES E-Library]: ITU-T G.191 filter, preferred presentation level, loudness balancing | MSIN filter | 3.36 dB (female)
2.9 dB (male) | (Most similar condition 2.4 dB) | (Most similar condition 1.3 dB) | + +#### III.3.2 Check of impact of initial presentation gain offset + +As described in clause III.2.2, the initial balance in loudness was randomized before the subjects started their adjustment. When analysing the relation between subjective scores and the initial presentation gain offset, a very low correlation was found (Pearson's correlation coefficient 0.0471). This suggests that bias in the results due to the initial balance could indeed be avoided. + +![Scatter plot showing Subjective scores [dB] on the y-axis (0 to 20) versus Initial random gain [dB] on the x-axis (-6 to 6). The data points are scattered across the plot, showing no strong correlation.](4c547ec1af44f8fcdc8b1d67662ac30a_img.jpg) + +A scatter plot with 'Initial random gain [dB]' on the x-axis (ranging from -6 to 6) and 'Subjective [dB]' on the y-axis (ranging from 0 to 20). The plot contains numerous blue data points scattered across the entire area, with no discernible trend or correlation. + +Scatter plot showing Subjective scores [dB] on the y-axis (0 to 20) versus Initial random gain [dB] on the x-axis (-6 to 6). The data points are scattered across the plot, showing no strong correlation. + +**Figure III.9 – Subjective scores as a function of initial random gain – randomness is desirable** + +### III.4 Conclusion + +The gain necessary to restore the loudness of speech after high-pass filtering was investigated using a loudness-balancing test paradigm. The amount of gain needed to restore the loudness after high-pass filtering was in general higher for the auditory results compared to the ISO 532-1 algorithm, at the presentation level of 65 dBSPL (for the reference sample). The difference between subjective and instrumental scores is statistically significant (p value 1.27e-6). + +A linear fit of the subjective data (y) against the instrumental data (x) gives: + +$$y = 1.1 + 1.2x$$ + +However, as expected with a larger dataset, such data fitting curve should become bound to origo (0,0), since the balancing of two identical sound files should be rated as 0 dB by both subjective and instrumental methods. + +Our overall interpretation is that when using loudness balancing of speech at 65 dBSPL, the results suggest that the loudness drop due to a high-pass filter is detected by ISO 532-1, with a fair prediction accuracy. + +It should, however, be noted that the loudness-balancing test paradigm is likely not used in the training of ISO 532-1. It should also be noted that the observed differences between auditory and instrumental results are not severe and that the dataset is small. The results do not, on their own, warrant disqualifying the ISO algorithm for P.Loudness purposes. + +# Appendix IV + +## Nominal transmission paths + +(This appendix does not form an integral part of this Recommendation.) + +## IV.1 Nominal receive paths + +Nominal receive paths are defined for the purpose of calculating the sending loudness: + +- diffuse-field corrected sensitivity/frequency characteristics: + - narrowband: 6.28 dBPa/V for all applicable frequencies; + - wideband, SWB and fullband: 4.83 dBPa/V for all applicable frequencies. + +These receive characteristics correspond to RLR = 2 dB when using [ITU-T P.58] correction between diffuse field and DRP, [ITU-T P.57] correction from DRP to ear reference point (ERP) and RLR calculation according to [ITU-T P.79]. The quantities for the calculation are found in Tables IV.1 and IV.2. + +NOTE – The IRS [ITU-T P.48] and modified IRS [ITU-T P.830] paths have not been assumed in this case. Rather, a narrowband sending terminal is assumed to be connected to a receiving terminal which is not restricted in bandwidth but where the overall sensitivity is adjusted. + +**Table IV.1 – Calculation of RLR for the nominal narrowband receive path** + +| Band No. | Mid-frequency (Hz) | Receive char. diffuse field (dBPa/V) | Diffuse field to DRP (dB) from Table 3 of [ITU-T P.58] | DRP to ERP (dB) from Table 2a of [ITU-T P.57] | Receive char. ERP (dBPa/V) | Receive $W_n$ from Table 1 of [ITU-T P.79] | $x_i$
$10^{\frac{0.175}{10}(s_i - w_i)}$ | +|------------------------------------------------------------------------------|--------------------|---------------------------------------------|--------------------------------------------------------|-----------------------------------------------|-----------------------------------|--------------------------------------------|---------------------------------------------| +| 1 | 200 | 6.28 | 0.0 | 0.0 | 6.28 | 85.0 | 4.14e-02 | +| 2 | 250 | 6.28 | 0.5 | –0.3 | 6.48 | 74.7 | 6.32e-02 | +| 3 | 315 | 6.28 | 0.5 | –0.2 | 6.58 | 79.0 | 5.34e-02 | +| 4 | 400 | 6.28 | 1.0 | –0.5 | 6.78 | 63.7 | 9.97e-02 | +| 5 | 500 | 6.28 | 1.5 | –0.6 | 6.68 | 73.5 | 6.69e-02 | +| 6 | 630 | 6.28 | 2.0 | –0.7 | 7.08 | 69.1 | 8.12e-02 | +| 7 | 800 | 6.28 | 4.0 | –1.1 | 9.18 | 68.0 | 9.23e-02 | +| 8 | 1000 | 6.28 | 5.0 | –1.7 | 9.58 | 68.7 | 9.12e-02 | +| 9 | 1250 | 6.28 | 6.5 | –2.6 | 10.18 | 75.1 | 7.22e-02 | +| 10 | 1600 | 6.28 | 8.0 | –4.2 | 10.08 | 70.4 | 8.69e-02 | +| 11 | 2000 | 6.28 | 10.5 | –6.5 | 10.28 | 81.4 | 5.63e-02 | +| 12 | 2500 | 6.28 | 14.0 | –9.4 | 10.88 | 76.5 | 7.02e-02 | +| 13 | 3150 | 6.28 | 12.0 | –10.3 | 7.98 | 93.3 | 3.17e-02 | +| 14 | 4000 | 6.28 | 11.5 | –6.6 | 11.18 | 113.8 | 1.58e-02 | +| $RLR = -\frac{10}{0.175} \cdot \log_{10} \sum_{i=1}^{14} x_i = 2 \text{ dB}$ | | | | | | | | + +**Table IV.2 – Calculation of RLR for the wideband nominal receive path (also used for SWB and fullband)** + +| Band No. | Mid-frequency (Hz) | Receive char. diffuse field (dBPa/V) | Diffuse field to DRP (dB) from Table 3 of [ITU-T P.58] | DRP-to-ERP (dB) from Table 2a of [ITU-T P.57] | Receive char. ERP (dBPa/V) | Receive $W_R$ from Table A.2 of [ITU-T P.79] | $x_i$
$10^{\frac{0.175}{10}(S_i - W_i)}$ | +|------------------------------------------------------------------------------|--------------------|--------------------------------------|--------------------------------------------------------|-----------------------------------------------|----------------------------|----------------------------------------------|---------------------------------------------| +| 1 | 100 | 4.83 | 0.0 | 0.0 | 4.83 | 152.8 | 2.57e-03 | +| 2 | 125 | 4.83 | 0.0 | 0.0 | 4.83 | 116.2 | 1.12e-02 | +| 3 | 160 | 4.83 | 0.0 | 0.0 | 4.83 | 91.3 | 3.07e-02 | +| 4 | 200 | 4.83 | 0.0 | 0.0 | 4.83 | 85.3 | 3.91e-02 | +| 5 | 250 | 4.83 | 0.5 | -0.3 | 5.03 | 75.0 | 5.96e-02 | +| 6 | 315 | 4.83 | 0.5 | -0.2 | 5.13 | 79.3 | 5.04e-02 | +| 7 | 400 | 4.83 | 1.0 | -0.5 | 5.33 | 64.0 | 9.40e-02 | +| 8 | 500 | 4.83 | 1.5 | -0.6 | 5.23 | 73.8 | 6.31e-02 | +| 9 | 630 | 4.83 | 2.0 | -0.7 | 5.63 | 69.4 | 7.66e-02 | +| 10 | 800 | 4.83 | 4.0 | -1.1 | 7.73 | 68.3 | 8.71e-02 | +| 11 | 1000 | 4.83 | 5.0 | -1.7 | 8.13 | 69.0 | 8.61e-02 | +| 12 | 1250 | 4.83 | 6.5 | -2.6 | 8.73 | 75.4 | 6.81e-02 | +| 13 | 1600 | 4.83 | 8.0 | -4.2 | 8.63 | 70.7 | 8.20e-02 | +| 14 | 2000 | 4.83 | 10.5 | -6.5 | 8.83 | 81.7 | 5.31e-02 | +| 15 | 2500 | 4.83 | 14.0 | -9.4 | 9.43 | 76.8 | 6.62e-02 | +| 16 | 3150 | 4.83 | 12.0 | -10.3 | 6.53 | 93.6 | 2.99e-02 | +| 17 | 4000 | 4.83 | 11.5 | -6.6 | 9.73 | 114.1 | 1.49e-02 | +| 18 | 5000 | 4.83 | 11.0 | -3.2 | 12.63 | 144.6 | 4.90e-03 | +| 19 | 6300 | 4.83 | 8.0 | -3.3 | 9.53 | 165.8 | 1.84e-03 | +| 20 | 8000 | 4.83 | 6.5 | -16.0 | -4.67 | 166.7 | 1.00e-03 | +| $RLR = -\frac{10}{0.175} \cdot \log_{10} \sum_{i=1}^{20} x_i = 2 \text{ dB}$ | | | | | | | | + +# Bibliography + +- [b-ITU-T P.10] Recommendation ITU-T P.10/G.100 (2017), *Vocabulary for performance, quality of service and quality of experience*. +- [b-AES E-Library] Audio Engineering Society, *AES E-Library, Subjective and Objective Measurements of Speech Loudness in Hands-Free Telephony–Toward an Extended Loudness Model for Telephonometry*. +<> +- [b-ISO 532] ISO 532:1975 (2012), *Acoustics — Method for calculating loudness level*. +- [b-ISO 532-1] ISO 532-1:2017, *Acoustics – Methods for calculating loudness – Part 1: Zwicker method*. +- [b-ISO 532-2] ISO 532-2:2017, *Acoustics – Methods for calculating loudness – Part 2: Moore-Glasberg method*. +- [b-CCITT 1977 SG12-C-0056] COM12 – C 56 – E, France (1977), *Objective method for determining the send "relative equivalents" of telephone sets*. +- [b-CCITT 1981 SG12-C-0010] COM12 – C 10 – E, Ministry of Posts and Telecommunications of the People's Republic of China (1981): *High-pass and low-pass filters tests and parameters m, G and S' estimated therefrom*. +- [b-CCITT 1982 SG12-C-0015] COM12 – C 15 – E, France (1982), *Objective determination of local system sensitivity for send, receive and sidetone, and comparison with the results of subjective measurements of relative equivalents*. +- [b-DIN 45631] DIN 45631/A1:2010-03, *Calculation of loudness level and loudness from the sound spectrum – Zwicker method – Amendment 1: Calculation of the loudness of time-variant sound*. +- [b-3GPP SA4] DIN 45631/A1:2010-03, *Calculation of loudness level and loudness from the sound spectrum – Zwicker method – Amendment 1: Calculation of the loudness of time-variant sound 3GPP SA4, DESUDAPS-1: Common subjective testing framework for training and validation of SWB and FB P.835 test predictors (v1.1)*, available in S4-151492, San Jose del Cabo, Mexico., 26-30 October 2015. +- [b-APG] Auditory Perception Group (2019), *Auditory demonstrations and useful software*. +<> + +- [b-Edjekouane] Edjekouane, Plapous, Quinquis, Meunier (2015), *Loudness of speech transmitted via handsfree telephone systems – Perceptual measurements and loudness models in free field listening*. Acta Acust. united with Acust., Vol. 101, No. 6, pp. 1130–1144. +- [b-Fiebig] A. Fiebig (2015), *Cognitive stimulus integration in the context of auditory sensations and sound perceptions*. PhD thesis, Berlin. +- [b-Fletcher] H. Fletcher and R. H. Galt (1950), *The Perception of Speech and Its Relation to Telephony*. The Acoustical Society of America, Vol. 22, No. 2. +- [b-github] Github, *Simple Compressor* (2019), <> +- [b-Glasberg 1] Glasberg, Moore (2002), *A model of loudness application to time-varying sounds* +- [b-Glasberg 2] Glasberg, Moore (2006), *Prediction of absolute thresholds and equal-loudness contours using a modified loudness model*, J. Acoust. Soc. Am., Vol. 120 No. 2, pp. 585-588. +- , J. Audio Eng. Soc., Vol. 50, No. 5, pp. 331-342. +- [b-Glasberg 3] Glasberg (2016), *A loudness model for time-varying sounds incorporating binaural inhibition*, Trends Hear., Vol. 20, pp. 1–16. +- [b-Head acoustics] HEAD acoustics, Herzogenrath, Germany (2014), *Calculation of loudness level and stationary and time varying sounds – Executable program for ISO\_532-1*. +- [b-Hots 1] V. J. Hots (2013), *Loudness of sounds with a subcritical bandwidth: A challenge to current loudness models?* J. Acoust. Soc. Am., Vol. 134, No. 4, pp. 334–339. +- [b-Hots 2] V. J. Hots (2014), *Loudness of subcritical sounds as a function of bandwidth, centre frequency, and level*. J. Acoust. Soc. Am., Vol. 135, No. 3, pp. 1313–1320. +- [b-Moore 1] Moore, Glasberg (1996), *A revision of Zwicker's loudness model*, Acustica, Vol. 82, pp. 335-345. +- [b-Moore 2] Moore, Glasberg, Baer (1997), *A model for the prediction of thresholds, loudness and partial loudness*, J. Audio Eng. Soc., Vol. 45, No. 4, pp. 224–240. +- [b-Moore 3] Moore, Glasberg (2007), *Modelling binaural loudness*, J. Audio Eng. Soc., Vol. 121, No. 3, pp. 1604–1612. +- [b-Rennies 1] Rennies, Verhey, Fastl (2010), *Comparison of loudness models for time-varying sounds*. Acta Acust. united with Acust., Vol. 96, No. 2, pp. 383–396. +- [b-Rennies 2] Rennies, Verhey, Appell, Kollmeier (2013), *Loudness of complex time-varying sounds? A challenge for current loudness models*. In Proceedings of Meetings on Acoustics, Vol. 19, + +No. 1, p. 050189, Acoustical Society of America. + +- [b-Schlittenlacher] Schlittenlacher, Hashimoto, Kuwano, Namba (2014), *Overall loudness of short time-varying sounds*, Internoise, Melbourne, Australia. +- [b-Stevens] Stevens (1957), *On the psychophysical law*. Psychol. Rev. Vol. 64, No. 3, p. 153–181. +- [b-Widmann] Widmann, Lippold, Fastl (1998), *A Computer Program Simulating Post-Masking for Applications in Sound Analysis Systems*, In: Proceedings of NOISE-CON' 98, Ypsilanti Michigan USA, pp. 451–456. +- [b-Zwicker 1] Zwicker (1958), *Über psychologische und methodische Grundlagen der Lautheit*. Acta Acust. united with Acust., Vol. 8, Suppl. 1, pp. 237–258. +- [b-Zwicker 2] Zwicker, Fastl, Dallmayr (1984), *BASIC-program for calculating the loudness of sounds from their 1/3-oct. band spectra according to ISO 532B*, Acustica Vol. 55, pp. 63–67. +- [b-Zwicker 3] Zwicker, Fastl (1991), *Psychoacoustics – facts and models*, IBN 3-540-52600-5. + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | Tariff and accounting principles and international telecommunication/ICT economic and policy issues | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling, and associated measurements and tests | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects, next-generation networks, Internet of Things and smart cities | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.75-198811-I_PDF-E/raw.md b/marked/P/T-REC-P.75-198811-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..6c077220c9021f5e9a537e0f23fdb135fe98c25b --- /dev/null +++ b/marked/P/T-REC-P.75-198811-I_PDF-E/raw.md @@ -0,0 +1,50 @@ + + +![ITU logo](2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg) + +The logo of the International Telecommunication Union (ITU) features the letters 'ITU' in a bold, sans-serif font, superimposed on a stylized globe with intersecting lines. + +ITU logo + +INTERNATIONAL TELECOMMUNICATION UNION + +**ITU-T** + +**P.75** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**TELEPHONE TRANSMISSION QUALITY +MEASUREMENTS RELATED TO SPEECH +LOUDNESS** + +--- + +**STANDARD CONDITIONING METHOD FOR +HANDSETS WITH CARBON MICROPHONES** + +**ITU-T Recommendation P.75** + +(Extract from the *Blue Book*) + +--- + +# NOTES + +1 ITU-T Recommendation P.75 was published in Volume V of the *Blue Book*. This file is an extract from the *Blue Book*. While the presentation and layout of the text might be slightly different from the *Blue Book* version, the contents of the file are identical to the *Blue Book* version and copyright conditions remain unchanged (see below). + +2 In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +# **STANDARD CONDITIONING METHOD FOR HANDSETS WITH CARBON MICROPHONES** + +*(Geneva, 1972; amended at Malaga-Torremolinos, 1984, Melbourne, 1988)* + +**1** Since the characteristics of carbon microphones are strongly dependent on conditioning techniques, it is necessary to follow a consistent procedure prior to measuring sensitivity/frequency characteristics in order to obtain reproducible results. The CCITT recommends that for best reproducibility, automatic mechanical conditioning be used. The following steps are specified for the *standard conditioning method*: + +- a) Place the handset in a holding fixture with the handset clamped in a position corresponding to that in which the microphone is going to be measured [e.g. loudness rating guard-ring position (LRGP) according to Annex A of Recommendation P.76]. +- b) Connect the microphone or telephone set terminals as required to the d.c. feed circuit and appropriate terminating loading. +- c) Turn the feed current on. After 5 seconds, condition the microphone by rotating it smoothly. Rotation is made such that the plane of the granule bed moves through an arc of at least 180°. The procedure is repeated twice with the handset coming to rest finally in the test position. The time of each rotation cycle should lie within the range of 2 to 12 seconds. + +**2** When carrying out subjective tests with a carbon microphone telephone set, the conditioning of the handset should be done by the talker. This conditioning should conform to the conditioning for objective measuring as described under § 1 above insofar as it is practicable. \ No newline at end of file diff --git a/marked/P/T-REC-P.804-201710-I_PDF-E/raw.md b/marked/P/T-REC-P.804-201710-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..462c658320bf5d045d6f394c08d181fe9ce23efa --- /dev/null +++ b/marked/P/T-REC-P.804-201710-I_PDF-E/raw.md @@ -0,0 +1,701 @@ + + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.804** + +(10/2017) + +SERIES P: TELEPHONE TRANSMISSION QUALITY, +TELEPHONE INSTALLATIONS, LOCAL LINE +NETWORKS + +Methods for objective and subjective assessment of +speech and video quality + +# --- **Subjective diagnostic test method for conversational speech quality analysis** + +Recommendation ITU-T P.804 + +## ITU-T P-SERIES RECOMMENDATIONS + +### TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS + +| | | | +|----------------------------------------------------------------------------------------------------|---------------|--------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Voice terminal characteristics | Series | P.30 | +| | | P.300 | +| Reference systems | Series | P.40 | +| Objective measuring apparatus | Series | P.50 | +| | | P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of speech quality | Series | P.80 | +| Methods for objective and subjective assessment of speech and video quality | Series | P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | +| Transmission performance and QoS aspects of IP end-points | Series | P.1000 | +| Communications involving vehicles | Series | P.1100 | +| Models and tools for quality assessment of streamed media | Series | P.1200 | +| Telemeeting assessment | Series | P.1300 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | Series | P.1400 | +| Methods for objective and subjective assessment of quality of services other than speech and video | Series | P.1500 | + +For further details, please refer to the list of ITU-T Recommendations. + +## Recommendation ITU-T P.804 + +# Subjective diagnostic test method for conversational speech quality analysis + +## Summary + +Recommendation ITU-T P.804 describes a subjective methodology for assessing and diagnosing the quality of transmitted speech in a telephone conversation. In addition to a score for the overall conversation quality, the methodology yields overall quality scores for three perceivable phases in a telephone conversation: listening, speaking, and interaction, as well as scores for their corresponding seven perceptual dimensions. Four of the perceptual dimension scores represent degradation associated with the listening phase, two are associated with the speaking phase, and one is associated with the interaction phase. Each of the perceptual dimension scores are based on ratings of the amount of degradation present in one system condition. The method is designed to be used with naïve subjects. The dimension scores can be used to provide diagnostic information on the causes of system degradations. The method is meant as a complement to standard conversation tests. + +## History + +| Edition | Recommendation | Approval | Study Group | Unique ID* | +|---------|----------------|------------|-------------|---------------------------------------------------------------------------| +| 1.0 | ITU-T P.804 | 2017-10-29 | 12 | 11.1002/1000/13397 | + +## Keywords + +Conversational speech quality evaluation, diagnostic evaluation of conversational speech quality, multi-dimensional quality assessment, subjective testing. + +--- + +\* To access the Recommendation, type the URL in the address field of your web browser, followed by the Recommendation's unique ID. For example, . + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database at . + +© ITU 2017 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## Table of Contents + +| | Page | +|-----------------------------------------------------------------------------------------------------------------------|------| +| 1 Scope..... | 1 | +| 2 References..... | 1 | +| 3 Definitions ..... | 1 | +| 3.1 Terms defined elsewhere ..... | 1 | +| 3.2 Terms defined in this Recommendation..... | 1 | +| 4 Abbreviations and acronyms ..... | 1 | +| 5 Conventions ..... | 2 | +| 6 Introduction to conversational speech quality analysis ..... | 2 | +| 7 Test methodology ..... | 3 | +| 7.1 Dimension rating scales..... | 3 | +| 7.2 Test design..... | 4 | +| 7.3 Dimension rating scheme ..... | 5 | +| 7.4 Instruction and training..... | 6 | +| Annex A – Test instructions – translated from German – Analysis of speech quality in a conversational situation ..... | 8 | +| Appendix I – Examples for the speaking part of the test methodology – translated from German..... | 13 | +| Appendix II – Results from an initial pilot test and second retest ..... | 14 | +| Bibliography..... | 16 | + + + +## Recommendation ITU-T P.804 + +## Subjective diagnostic test method for conversational speech quality analysis + +## 1 Scope + +This Recommendation describes a subjective test methodology which is able to assess and diagnose the quality of speech in a "telephone conversation" scenario. Common conversation tests, as described in [ITU-T P.800] and [ITU-T P.805], provide valid methods for the overall conversational quality, but do not give insights into reasons for possible quality losses. In addition, common conversational tests lack analytic ability, since naïve participants concentrate on the conversation flow. To circumvent these problems, this Recommendation describes a test methodology that specifically allows participants to perceive each phase of a conversation separately, in addition to a natural conversation, and yields overall conversational quality scores as well as quality scores for each phase (listening, speaking, interaction). In addition, scores for seven underlying perceptual dimensions of conversational speech quality are provided. These scores enable the analysis of conversational speech quality for diagnosis and optimization. + +## 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- [ITU-T P.800] Recommendation ITU-T P.800 (1996), *Methods for subjective determination of transmission quality*. +- [ITU-T P.805] Recommendation ITU-T P.805 (2007), *Subjective evaluation of conversational quality*. +- [ITU-T P.806] Recommendation ITU-T P.806 (2014), *A subjective quality test methodology using multiple rating scales*. +- [ITU-T P.835] Recommendation ITU-T P.835 (2003), *Subjective test methodology for evaluating speech communication systems that include noise suppression algorithm*. + +## 3 Definitions + +### 3.1 Terms defined elsewhere + +None. + +### 3.2 Terms defined in this Recommendation + +None. + +## 4 Abbreviations and acronyms + +This Recommendation uses the following abbreviations and acronyms: + +- ACR Absolute Category Rating + +| | | +|------|---------------------------------| +| MDS | Multidimensional Scaling | +| MOS | Mean Opinion Score | +| PCA | Principal Component Analysis | +| RNVT | Random Number Verification Task | +| SCT | Short Conversation Test | +| SD | Semantic Differential | + +## 5 Conventions + +None. + +## 6 Introduction to conversational speech quality analysis + +To provide diagnostic information about the quality of transmitted speech, ITU-T recommends using multiple rating scales in subjective experiments in [ITU-T P.806]. The approach targets assessing perceptual dimensions to give deeper insight into possible quality loss. The recommended method refers to the passive listening-only situation and gives additional information to overall quality mean opinion score (MOS) absolute category rating (ACR)-experiments, as recommended in [ITU-T P.800]. However, the listening-only situation, only partly agrees with reality in telecommunications. Thus, conversation tests to assess the quality in an interactive situation have been designed and described in [ITU-T P.805]. The recommended methods do not provide diagnostic information. For this, a set of seven perceptual dimensions for a conversational situation are proposed in [b-Köster2014]. The proposed dimensions cover the three possible phases/situations of a conversation: listening, speaking, and interaction. See [b-Guéguin2008]. + +To identify the relevant proposed dimensions, each phase has been analysed in detail, applying a (I) pairwise similarity experiment with a following multidimensional scaling (MDS) and a (II) a semantic differential (SD) experiment with a following principal component analysis (PCA). + +Applying both methods in separate experiments in [b-Köster2014] and [b-Wältermann2010] resulted in the following set of perceptual dimensions for a conversational situation (see Table 1): the listening phase is comprised of four dimensions: *noisiness*, *discontinuity*, *coloration* and *loudness*; the speaking phase is comprised of two dimensions: *impact of one's own voice on speaking* and *degradation of one's own voice*; the interaction phase is comprised of only one dimension: *interactivity*. + +**Table 1 – Overview of the seven identified and proposed perceptual quality dimensions for a conversational situation** + +| Conversational phase | Perceptual dimension | Description | Possible source | +|--------------------------|--------------------------------|-------------------------------------------------------|-----------------------------------------------------| +| Listening phase | Noisiness | Background noise, circuit noise, coding noise | Coding, circuit or background noise | +| | Discontinuity | Isolated and non-stationary distortions | Packet loss | +| | Coloration | Frequency response distortions | Bandwidth limitations | +| | Loudness | Important for the overall quality and intelligibility | Attenuation | +| Speaking phase | Impact of one's own voice | How is the back-coupling of one's own voice perceived | Sidetone and echo | +| | Degradation of one's own voice | How is the back-coupling of one's own voice degraded | Frequency distortions of the sidetone and echo path | +| Interaction phase | Interactivity | Delayed and disrupted interaction | Delay | + +Since the proposed dimensions were identified in separate listening, speaking, and conversation tests, the dimensions were validated in a sophisticated conversational experiment in which each of the conversational phases, as well as all of the proposed perceptual dimensions were addressed [b-Köster2015a]. + +The outcome of the experiment showed that in traditional conversation scenarios the proposed dimensions are difficult to identify. Thus, this Recommendation describes a test method that specifically allows participants to perceive each phase separately, in addition to a natural conversation paradigm. In addition, the described method allows for directly quantifying the proposed seven dimensions within one single experiment. The method enables analysis of conversational speech quality for diagnosis and optimization and enables an increase in the number of conditions to be assessed. + +## 7 Test methodology + +### 7.1 Dimension rating scales + +The subjective method provides a means for quantifying seven quality relevant perceptual dimensions in a conversational situation (noisiness, discontinuity, coloration, loudness, impact of one's own voice on speaking, degradation of one's own voice, and interactivity) directly by means of seven descriptive scales. In addition, the overall conversational quality and the overall quality for each individual phase are gathered. + +Each dimension scale is dedicated to one particular dimension. The scales are labelled with the antonym-pairs describing the corresponding dimension. This enables for directly quantifying separate scores for each perceptual dimension present in a conversational situation. The overall rating scales and the graphical scale layout for the dimensions are shown in Figure 1. The continuous scales were chosen over traditional ACR scales because they showed to be more sensitive [b-Köster2015b]. While the labels on the left of the scales describe no impairment in the relating dimension, the labels on the right describe the maximum impairment. Thus the scales are considered to be unipolar. A detailed + +description of usage and definition of the scales, as given to test participants, can be found in Annex A. + +#### Overall Quality + +#### Overall quality + +![Overall quality dimension scale design](f519a5be118c846f631c992412353fb9_img.jpg) + +A horizontal dimension scale for Overall Quality. It features a central horizontal line with 11 vertical tick marks of varying lengths. Below the line, the labels 'very bad', 'bad', 'poor', 'fair', 'good', 'excellent', and 'ideal' are positioned at regular intervals. + +Overall quality dimension scale design + +#### Listening phase + +![Listening phase dimension scale design](8ee3b76dd49f31624d287885bc2c81ee_img.jpg) + +Four horizontal dimension scales for the Listening phase. Each scale has a central line with 11 vertical tick marks. The first scale is labeled 'uncolored' at the left and 'colored' at the right. The second scale is labeled 'continuous' at the left and 'discontinuous' at the right. The third scale is labeled 'not noisy' at the left and 'noisy' at the right. The fourth scale is labeled 'optimal loudness' at the left and 'sub-optimal loudness' at the right. + +Listening phase dimension scale design + +#### Speaking phase + +![Speaking phase dimension scale design](7c2f0efb2c5d10a52ce19ba33d9d3cec_img.jpg) + +Two horizontal dimension scales for the Speaking phase. Each scale has a central line with 11 vertical tick marks. The first scale is labeled 'own voice not degraded' at the left and 'own voice degraded' at the right. The second scale is labeled 'no impact on speaking' at the left and 'high impact on speaking' at the right. + +Speaking phase dimension scale design + +#### Interaction phase + +![Interaction phase dimension scale design](64bacf564ff025df294b6d30341c76df_img.jpg) + +A horizontal dimension scale for the Interaction phase. It features a central horizontal line with 11 vertical tick marks. Below the line, the labels 'easy to interact' and 'hard to interact' are positioned at the left and right ends respectively. + +Interaction phase dimension scale design + +P.804(17)\_F01 + +Figure 1 – Dimension scale design + +### 7.2 Test design + +The method follows common paradigms for subjective conversational tests as described in [ITU-T P.805]. For each condition, or transmission system properties under test, two participants in two separate rooms according to [ITU-T P.800] are required. The basic test setup can be seen in Figure 2. + +![Test method set-up diagram](ee8536b235eb6aad21e2048fd5308900_img.jpg) + +A block diagram of the test method set-up. It shows two participants, 'Room A participant 1' and 'Room B participant 2', each in a separate room. Each room contains a microphone and a speaker icon. The participants are connected to a central 'Network simulator' block via lines representing audio paths. + +Test method set-up diagram + +P.804(17)\_F02 + +Figure 2 – Test method set-up + +The test method specifically allows participants to perceive each phase separately, in addition to a natural conversation paradigm. Therefore, the test method to assess one condition is composed of three sessions: + +- 1 In the first session, the task of the two participants is to conduct a short conversation test (SCT) scenario according to [ITU-T P.805]. The SCTs were used because their tasks represent everyday-life situations and provide a reasonable degree of interaction while being limited to an acceptable test duration. Thus, this session represents a regular everyday-life conversational scenario of about 2-4 minutes. After each scenario, the participants are asked to judge the overall quality (according to [ITU-T P.800] using the scale shown in Figure 1), and then the seven perceptual dimensions representing all phases of a conversation. +- 2 The second session addresses the listening and speaking phases. One of the participants is asked to read two sentences out loud while the other participant listens to what is being read. The sentences and procedures of the speaking part are similar to [b-Appel2002] and [b-Köster2014]. The listening part is analog to [b-Wältermann2010]. After the first sequence, the participants change roles so that each participant has to both speak and listen. For each sequence, the participants are asked to judge the overall quality of the speaking as well as the two dimensions for the speaking phase and the overall quality of the listening as well as the four dimensions for the listening phase. +Examples for the speaking part can be found in Appendix I. +- 3 The third session addresses the interaction phase. This task is supposed to be sensitive for possible delays in the transmission system. Therefore, random number verification tasks (RNVTs) are used as per [ITU-T P.805]. The participants are asked to judge the overall quality of the interaction and the interactivity representing the interaction phase. + +An overview of the test procedure for both participants can be seen in Figure 3. + +![Figure 3: Overview of the test procedure. The diagram shows three sessions (Session 1, Session 2, Session 3) for two participants (Participant 1 and Participant 2). Session 1 is SCT (Short Conversation Test) with Overall quality and 7 perceptual dimensions. Session 2 involves Speaking and Listening phases with Overall quality and perceptual dimensions. Session 3 is RNVT (Random Number Verification Task) with Overall quality and 1 perceptual dimension. A vertical arrow on the left is labeled 'Condition i' and a horizontal arrow at the bottom right is labeled 'P.804(17)_F03'.](5a4e62bead259c258d069fd3663ea670_img.jpg) + +| | Participant 1 | Participant 2 | +|--------------------------------|-----------------------------|-------------------------| +| Condition i | Session 1 SCT ↔ SCT | | +| | Overall quality | Overall quality | +| | 7 perceptual dimensions | 7 perceptual dimensions | +| Session 2 Speaking → Listening | | | +| Overall quality (speaking) | Overall quality (listening) | | +| 2 perceptual dimensions | 4 perceptual dimensions | | +| Listening ← Speaking | | | +| Overall quality (listening) | Overall quality (speaking) | | +| 4 perceptual dimensions | 2 perceptual dimensions | | +| Session 3 RNVT ↔ RNVT | | | +| Overall quality | Overall quality | | +| 1 perceptual dimensions | 1 perceptual dimensions | | + +Figure 3: Overview of the test procedure. The diagram shows three sessions (Session 1, Session 2, Session 3) for two participants (Participant 1 and Participant 2). Session 1 is SCT (Short Conversation Test) with Overall quality and 7 perceptual dimensions. Session 2 involves Speaking and Listening phases with Overall quality and perceptual dimensions. Session 3 is RNVT (Random Number Verification Task) with Overall quality and 1 perceptual dimension. A vertical arrow on the left is labeled 'Condition i' and a horizontal arrow at the bottom right is labeled 'P.804(17)\_F03'. + +**Figure 3 – Overview of the test procedure. SCT – short conversation scenario, RNVT – random number verification task** + +## 7.3 Dimension rating scheme + +The dimension rating scheme of the test method is comparable to the scheme for analysing noisy signals [ITU-T P.835]. Each of the three separate sessions of the test method includes an assignment (speaking, listening, SCT, or RNVT) as well as an overall quality and a dimension assessment task. + +As these assessment tasks follow the same procedure, only the rating task for session I will be described in detail. + +After the overall quality assessment, according to [ITU-T P.800], the dimension scales (see Figure 1) are presented separately and consecutively. This is to reduce the bias due to the presentation order. Before participants are asked for their ratings, they are asked to conduct the given task once. Afterwards, the participants first give their judgments on the overall quality and second on the seven perceptual dimensions. The detailed rating schema for session I is shown in Figure 4. + +The conditions to be assessed are presented in randomized order. Additionally, the order of the dimension scales is permuted for each participant. The schema is shown in Table 2. For each participant the order of the scales is held constant to avoid confusion of the scales. + +![Figure 4: Condition and scale presentation and rating for session 1. A timeline diagram showing the sequence of events for a participant. It starts with 'Condition i' in a box. Below a horizontal timeline labeled 't', there are eight 'X' marks. Between each 'X' mark is a 'Voting interval' for a specific scale: 'overall quality', 'Dim. scale 1', 'Dim. scale 2', 'Dim. scale 3', 'Dim. scale 4', 'Dim. scale 5', 'Dim. scale 6', and 'Dim. scale 7'. Below each 'X' mark is a 'Presentation' event for the corresponding scale. The first 'X' is labeled 'Duration condition i' above it.](27b06ec9f42b5d727a2630f61a5f1861_img.jpg) + +Figure 4: Condition and scale presentation and rating for session 1. A timeline diagram showing the sequence of events for a participant. It starts with 'Condition i' in a box. Below a horizontal timeline labeled 't', there are eight 'X' marks. Between each 'X' mark is a 'Voting interval' for a specific scale: 'overall quality', 'Dim. scale 1', 'Dim. scale 2', 'Dim. scale 3', 'Dim. scale 4', 'Dim. scale 5', 'Dim. scale 6', and 'Dim. scale 7'. Below each 'X' mark is a 'Presentation' event for the corresponding scale. The first 'X' is labeled 'Duration condition i' above it. + +P.804(17)\_F04 + +**Figure 4 – Condition and scale presentation and rating for session 1** + +**Table 2 – Presentation order of the dimensions scales. Noi – noisiness, Dis - discontinuity, Col – coloration, Lou – loudness, Ios – impact of one's own voice on speaking, Dos – degradation of one's own voice, and Int – interactivity** + +| Participant | Dim scale 1 | Dim scale 2 | Dim scale 3 | Dim scale 4 | Dim scale 5 | Dim scale 6 | Dim scale 7 | +|-------------|-------------|-------------|-------------|-------------|-------------|-------------|-------------| +| 1 | Dis | Col | Noi | Lou | Ios | Dos | Int | +| 2 | Col | Noi | Lou | Ios | Dos | Int | Dis | +| 3 | Noi | Lou | Ios | Dos | Int | Dis | Col | +| 4 | Lou | Ios | Dos | Int | Dis | Col | Noi | +| 5 | Ios | Dos | Int | Dis | Col | Noi | Lou | +| 6 | Dos | Int | Dis | Col | Noi | Lou | Ios | +| 7 | Int | Dis | Col | Noi | Lou | Ios | Dos | +| ... | ... | ... | ... | ... | ... | ... | ... | + +### 7.4 Instruction and training + +A detailed written description of the test method is given to participants to ensure an equal level of knowledge (Annex A). First, the instruction gives an overview of the scales and how they should be used. It is explained, that in the experiment the characteristics of a conversation are supposed to be judged and that this judgment is done on seven scales. Each scale is labelled with an attribute at each end that describes the characteristic to be judged. The scales are described in detail using the highly correlated attributes according to the SD experiment conducted to extract the perceptual dimensions. Second, the test procedure (the three sessions and the relating tasks) is explained. These instructions can be found in Annex A. + +In the training the two participants run through one test sequence as described in Figure 3. This is done to ensure that the participants get to know the test procedure and become familiar with using the scales and the test method. + +In addition, optional training should be conducted to ensure that the test subjects get to know the test procedure as well as get familiarized with the usage of the scales and the test method. For this, the + +test procedure as well as the ratings (overall quality and dimensions) should be practiced. Thus, one possible training exercise could look like the following: + +The test subjects execute the test procedure (as described in Figure 3) twice. In the first run, the first session (SCT) is degraded with a condition related to the dimension noisiness. The second session (speaking and listening) is alternately degraded with conditions related to the dimensions discontinuity and impact of one's own voice. In the third session (RNVT), the test subjects will not be confronted with a degradation. In the second run, the first session is degraded with a condition related to the dimension Coloration. In the second session, the subjects will again be confronted with two alternating conditions, one related to the dimension loudness and one related to the dimension degradation of one's own voice. Finally, in the third session, the transmission system will introduce delay to trigger the dimension interactivity. The scheme of the training is illustrated in Table 3. + +With this possible training, the two test subjects are introduced to the test procedure and the seven perceptual dimensions. In addition, the test subjects get to know the characteristics of all perceptual dimensions and practice using the rating scales. The training of the third session (no degradation vs. *interactivity*) is particularly useful to ensure the test subject's sensitivity for a transmission delay. As possible conditions, the test conditions (or adapted conditions) described in Appendix II could be used. However, the training is just a recommendation; it might also be possible to run through the test procedure only once to ensure that the test subjects understand their tasks. + +**Table 3 – Scheme of possible training for the proposed test method** + +| | Session | Perceptual dimensions
Test subject 1 | Perceptual dimensions
Test subject 1 | +|-----------------------|------------------------|-------------------------------------------------|-------------------------------------------------| +| Training run 1 | 1 (SCT) | Noisiness | Noisiness | +| | 2 (listening/speaking) | Discontinuity | Impact of one's own voice | +| | 2 (speaking/listening) | Impact of one's own voice | Discontinuity | +| | 3 (RNVT) | none | none | +| Training run 2 | 1 (SCT) | Coloration | Coloration | +| | 2 (listening/speaking) | Loudness | Degradation of one's own voice | +| | 2 (speaking/listening) | Degradation of one's own voice | Loudness | +| | 3 (RNVT) | Interactivity | Interactivity | + +## Annex A + +### Test instructions – translated from German – Analysis of speech quality in a conversational situation + +(This annex forms an integral part of this Recommendation.) + +Thank you very much for taking part in this experiment! Please take the time to read the instructions thoroughly. Should you have any questions, please do not hesitate to ask the experiment supervisor. + +You are taking part in a conversational experiment, in which the properties and the perception of a conversation are to be evaluated. To do this, you and your conversational partner will be put into a conversational situation in which various conditions will be presented in three phases. These phases are as follows: (1) a short conversational scenario phase, (2) a speaking and listening phase, and (3) an interaction phase. After each phase, you are to evaluate your perception and the properties of the phase. The exact procedure for individual phases will be described in detail later and will be made clear in a test run. + +The characteristic of each condition is to be evaluated with the help of different scales. Please familiarize yourself with the scales and how to use them now. + +Every scale has a term at each endpoint (e.g., not noisy/noisy). You are to evaluate to what extent the characteristics of a condition can be described by the terms on the scale. There are two different kinds of scales. The scale for overall quality has describing terms for every point, while the scales for the characteristics (e.g., not noisy/noisy) only have terms at their endpoints. The usage of these two scale types is analogous and is explained in detail for the overall quality. + +The **overall quality** of a condition is to be evaluated. If you are of the opinion that the overall quality of the condition is bad, mark a cross at the position "bad" as shown next: + +![A horizontal quality scale from 'very bad' to 'ideal' with a cross on 'bad'.](b8261918596971c3801af23435c5ea50_img.jpg) + +A horizontal line representing a quality scale. Below the line are labels: "very bad", "bad", "poor", "fair", "good", "excellent", and "ideal". There are 10 tick marks of varying lengths along the line. A large 'X' is drawn over the tick mark corresponding to the label "bad". The label "P.804(17)\_FA.1" is located at the bottom right of the scale. + +A horizontal quality scale from 'very bad' to 'ideal' with a cross on 'bad'. + +If you think that the overall quality of the condition is excellent, mark a cross at the position "excellent" as shown next: + +![A horizontal quality scale from 'very bad' to 'ideal' with a cross on 'excellent'.](70f5f44cb855bbb0c203c00187b2113e_img.jpg) + +A horizontal line representing a quality scale, identical to the one above. Below the line are labels: "very bad", "bad", "poor", "fair", "good", "excellent", and "ideal". There are 10 tick marks of varying lengths along the line. A large 'X' is drawn over the tick mark corresponding to the label "excellent". The label "P.804(17)\_FA.2" is located at the bottom right of the scale. + +A horizontal quality scale from 'very bad' to 'ideal' with a cross on 'excellent'. + +For your evaluation, you can freely use the entire scale. The markings on the scale are there to provide points of reference to support your evaluation. You can even use the spaces in between the markings if you do not wish to settle for one of the markings. + +And you can always use the "overflow areas", beyond the terms, if you feel the terms are not sufficient for your evaluation, as shown next: + +![A horizontal scale for evaluating sound quality. The scale has seven tick marks. The first tick mark is labeled 'very bad' and the second is labeled 'bad'. These two are circled in red. The third is 'poor', the fourth is 'fair', the fifth is 'good', the sixth is 'excellent', and the seventh is 'ideal'. The label 'P.804(17)_FA.3' is at the bottom right.](71ab4df17511d75261da8d462d643b1a_img.jpg) + +A horizontal scale for evaluating sound quality. The scale has seven tick marks. The first tick mark is labeled 'very bad' and the second is labeled 'bad'. These two are circled in red. The third is 'poor', the fourth is 'fair', the fifth is 'good', the sixth is 'excellent', and the seventh is 'ideal'. The label 'P.804(17)\_FA.3' is at the bottom right. + +The usage of the scales is analogous for the different characteristics. Overall there are seven characteristics to evaluate. + +#### 1 Noisiness + +The first scale has the terms "not noisy" and "noisy" and appears as follows: + +![A horizontal scale for evaluating noisiness. The scale has 11 tick marks. The first tick mark is labeled 'not noisy' and the last is labeled 'noisy'. The label 'P.804(17)_FA.4' is at the bottom right.](fe753d01ad5fe6cf150018c958173c6d_img.jpg) + +A horizontal scale for evaluating noisiness. The scale has 11 tick marks. The first tick mark is labeled 'not noisy' and the last is labeled 'noisy'. The label 'P.804(17)\_FA.4' is at the bottom right. + +With this scale, you are to evaluate the **noisiness** of what you heard in a condition. The terms "not noisy" and "noisy" can be described with the terms *noisless* and *not hissing* and *noisy* and *hissing*, respectively. + +#### 2 Discontinuity + +The **discontinuity** of what you heard in a condition is to be evaluated with the second scale: + +![A horizontal scale for evaluating discontinuity. The scale has 11 tick marks. The first tick mark is labeled 'continuous' and the last is labeled 'discontinuous'. The label 'P.804(17)_FA.5' is at the bottom right.](4a8166688ed7276efb173f550ba47eb4_img.jpg) + +A horizontal scale for evaluating discontinuity. The scale has 11 tick marks. The first tick mark is labeled 'continuous' and the last is labeled 'discontinuous'. The label 'P.804(17)\_FA.5' is at the bottom right. + +The term "continuous" means that what you heard in a condition is completely *even, firm, not chopped* and *not frayed*. The term "discontinuous" can be described with terms like *uneven, wobbly, chopped* or *frayed*. + +#### 3 Coloration + +The **coloration** of what you heard in a condition is described by the third scale: + +![A horizontal scale for evaluating coloration. The scale has 11 tick marks. The first tick mark is labeled 'uncolored' and the last is labeled 'colored'. The label 'P.804(17)_FA.6' is at the bottom right.](a2f89834a149b852721f8887349fd6fa_img.jpg) + +A horizontal scale for evaluating coloration. The scale has 11 tick marks. The first tick mark is labeled 'uncolored' and the last is labeled 'colored'. The label 'P.804(17)\_FA.6' is at the bottom right. + +The term "uncolored" means that what you heard in a condition is *direct, close, full* and *not nasally*. The term "colored" means the heard sound is *indirect, far, thin* and *nasally*. + +#### 4 Loudness + +The fourth scale describes the **loudness** of the heard sound in a condition: + +![A horizontal scale for evaluating loudness. The scale has 11 tick marks. The first tick mark is labeled 'optimal loudness' and the last is labeled 'sub-optimal loudness'. The label 'P.804(17)_FA.7' is at the bottom right.](dfd60da6d5e7618f3122c129506e105e_img.jpg) + +A horizontal scale for evaluating loudness. The scale has 11 tick marks. The first tick mark is labeled 'optimal loudness' and the last is labeled 'sub-optimal loudness'. The label 'P.804(17)\_FA.7' is at the bottom right. + +If the heard sound is neither too loud nor too quiet then the volume is optimal. If that is not the case then the volume is not optimal. + +#### 5 Impact of one's own voice + +The **impact on your speaking through the hearing of your own voice** is to be evaluated with the fifth scale: + +![A horizontal scale with 10 tick marks. The left end is labeled 'no impact on speaking' and the right end is labeled 'high impact on speaking'. The code 'P.804(17)_FA.8' is at the bottom right.](ad29805cd4f64ad2828e14feb66de664_img.jpg) + +A horizontal scale with 10 tick marks. The left end is labeled 'no impact on speaking' and the right end is labeled 'high impact on speaking'. The code 'P.804(17)\_FA.8' is at the bottom right. + +In some conditions, you will be confronted with back-coupling of your own voice. This means that you will hear your own voice. With this characteristic you are to evaluate if this does or doesn't negatively impact your speaking. More to the point, you are to evaluate if speaking while hearing yourself is *not distracting, not irritating, fluid and does not take concentration* or if it is *distracting, irritating, not fluid and takes concentration*. + +#### 6 Degradation of one's own voice + +The sixth scale is for the evaluation of the **degradation of your own voice** while having back-coupling: + +![A horizontal scale with 10 tick marks. The left end is labeled 'own voice not degraded' and the right end is labeled 'own voice degraded'. The code 'P.804(17)_FA.9' is at the bottom right.](e95f47f7a4c01c8889d6d46919b4c73d_img.jpg) + +A horizontal scale with 10 tick marks. The left end is labeled 'own voice not degraded' and the right end is labeled 'own voice degraded'. The code 'P.804(17)\_FA.9' is at the bottom right. + +The term "own voice not degraded" means that you hear your own voice unaltered, e.g., you hear your voice *not distorted, echofree* and with *optimal volume*. The term "own voice degraded" can be described with terms like *distorted, echoed, thin or not optimal volume*. + +#### 7 Interaction + +The seventh scale is used to evaluate the **interactivity** of a condition: + +![A horizontal scale with 10 tick marks. The left end is labeled 'easy to interact' and the right end is labeled 'hard to interact'. The code 'P.804(17)_FA.10' is at the bottom right.](0bd23f00e0632855cfef9274f1ab93d8_img.jpg) + +A horizontal scale with 10 tick marks. The left end is labeled 'easy to interact' and the right end is labeled 'hard to interact'. The code 'P.804(17)\_FA.10' is at the bottom right. + +The term "easy to interact" means that the interaction between you and your conversational partner is easy. That means you have an *effective, pleasant, easy, and agile* interaction. The term "hard to interact" means that the interaction is *ineffective, unpleasant, hard and sluggish*. + +For each scale, the following applies: Use as described for the scale for "overall quality" (see above). + +Now that you have familiarized yourself with the meaning of the characteristics, the phases and the procedure of the test will be described. + +#### Phase 1 – Conversation + +##### Explanation + +In this phase you and your partner simulate a telephone conversation. One of you will be the caller, who wants something and calls a company/organization/institution. The other one will be the company/organization/institution. + +##### Explanation of symbols + +**As the caller, you have the following symbols:** + +![Green telephone handset icon with signal waves.](88db8f0ac1e6a237c23436f105949f30_img.jpg) + +Green telephone handset icon with signal waves. + +This symbol means: You are the caller. + +Please wait until the experiment supervisor asks you to start the first conversation. + +![Blue circle with a white lowercase 'i' inside.](c1b31d4aff04545839dcaa17acf93363_img.jpg) + +Blue circle with a white lowercase 'i' inside. + +Next to this symbol is the reason for your call. + +*E.g., I want to buy a ticket!* + +![Icon for conditions: a document with the word 'CONDITIONS' written on it.](ac852a162572ca8a8c8478c49b571af5_img.jpg) + +Icon for conditions: a document with the word 'CONDITIONS' written on it. + +Next to this symbol are the conditions that should be incorporated into the exchange of information. + +*E.g., I want to buy a ticket! → BUT cheaply if possible!* + +![Icon for recording information: a telephone handset with a downward arrow.](6cc4a2d5ea0462e4825d57bd689bd2b3_img.jpg) + +Icon for recording information: a telephone handset with a downward arrow. + +Next to this symbol you are to record all information that you need from your telephone partner. + +![Icon for providing information: a telephone handset with a circular arrow.](53cb558495f8cc93c4202314ef492966_img.jpg) + +Icon for providing information: a telephone handset with a circular arrow. + +Next to this symbol is all of the information that your partner needs and that you should provide at some point during the conversation. + +![Icon for a question: a person's head with a question mark.](f7bc9b0327ed4589a3faf9a7b3c92712_img.jpg) + +Icon for a question: a person's head with a question mark. + +Next to this symbol is a question to which neither you nor your conversational partner have any information. You are to discuss this question briefly and come to a satisfactory conclusion together. + +###### **If you get called, you have the following symbols:** + +![Icon for getting called: a telephone handset with signal waves.](e0f3606e9c9eee18bc68b713df4eda8a_img.jpg) + +Icon for getting called: a telephone handset with signal waves. + +This symbol means: You are getting called. Wait until you hear the ringtone and then pick up. + +![Icon for information: a blue circle with a white letter 'i'.](02e2819551a3fd5c0a5e941008b5c526_img.jpg) + +Icon for information: a blue circle with a white letter 'i'. + +Next to this symbol is information from which you are to sort out the information that your partner needs. + +*E.g., Prices for train tickets for adults, students, children, seniors, etc.* + +![Icon for recording information: a telephone handset with a downward arrow.](dedde70afe48651b0864f9bf0d2979c4_img.jpg) + +Icon for recording information: a telephone handset with a downward arrow. + +Next to this symbol you are to record all information that you need from your conversational partner. + +##### **Procedure** + +Read the information that you are given once before starting the conversation. + +If you are you the caller, call your partner. Play through the scenario and then hang up the conversation. Now click "Next". First, you evaluate the overall quality of the conversation and then the conversation regarding the seven characteristics. + +If you are you the one who gets called, please wait until you hear the ringtone. When you do, pick up. Play through the scenario with your partner and wait until they hang up the conversation. Click on "Next". First, you evaluate the overall quality of the conversation and then the conversation regarding the seven characteristics. + +#### **Phase 2 – Listening and speaking** + +##### **Explanation** + +In this phase either you read two sentences to your partner or your partner reads two sentences to you. Afterwards you swap. + +##### **Procedure** + +You should have two sentences before you: call your partner and read both sentences. Hang up the conversation. Click "Next". Now, evaluate the overall quality of your speaking experience. Afterwards, evaluate your speaking regarding the two characteristics: "impact of your own voice" and "degradation of your own voice". + +You should you have no sentences to read; wait until you are called and pick up the conversation. Listen to the two sentences and wait until your partner hangs up. Click "Next". Evaluate the "overall quality" of your listening experience, followed by the four conditions "noisiness", "coloration", "continuity" and "volume". + +#### **Phase 3 – Interaction** + +##### **Explanation** + +In this phase, you and your partner are to perform a number verification. + +##### **Procedure** + +You will see four series of numbers before you. Two are **bold** and two are not. You are to read the bold number sequences to your partner. Should the first of the four sequences on your screen be bold, please call your partner. Perform the number verification in an alternating fashion and when all four sequences have been read, hang up the conversation. Wait after every number in the sequence (e.g., 24) for a confirmation from your partner in the form of a "yes" when they have the same number or a "no" if they have a different number. After the conversation, click "Next". Please evaluate the "overall quality" of the interaction and afterwards evaluate the characteristic "interaction". + +##### **Overall procedure** + +Summarize how the test goes, for every condition, in the following way (11 conditions in total): + +call → phase 1 → hang up → Next → overall quality → evaluate 7 characteristics → call → phase 2 (speaking/listening) → hang up → Next → overall quality → 2/4 characteristics → call → phase 2 (speaking /listening) → hang up → Next → overall quality → 2/4 characteristics → call → phase 3 → hang up → Next → overall quality → interaction. + +With a few of the conditions you could have the feeling that you have evaluated them before. This is not the case. Please evaluate every condition independently from all your other evaluations. Try not to remember how you evaluated other "similar" conditions before; instead evaluate each characteristic of every condition individually. + +Please evaluate the scales quickly and intuitively. This experiment is purely subjective in its nature. There are no right or wrong answers. **The system, not you, is being tested.** Only your personal opinion matters for this experiment. + +If you have any additional questions, please don't hesitate to ask the experiment supervisor. Have fun! 😊 + +## Appendix I + +### **Examples for the speaking part of the test methodology – translated from German** + +(This appendix does not form an integral part of this Recommendation.) + +"Can you please give me the best connection between Munich and Berlin? I have to arrive on Saturday at 12.30 pm latest." + +"Please connect me with the complaints department. The repair on the main drain of my house has not been carried out properly." + +"Here is the local fire station. We are trying to find an emergency call that has been suspended without giving any specific information. It sounded like a local call." + +"I'm really sorry that I could not come to dinner at the weekend. Just before I wanted to leave, I had a small accident." + +"Dear mother, here in Palma it is beautiful. The weather is hot and sunny and the sea is unbelievable. This morning I made a hike." + +"On warm nights, I often lie in my bed and look through the open window. The moon shines down on me and it looks as if he smiles." + +"Tomorrow morning my daughter wants to bake a pie. She insists on making everything by hand. She is convinced that it tastes much better then." + +"What am I going to eat tonight? I still have a stew in the freezer, but the problem is that I eat it two to three times a week." + +"Please connect me with the repair department. My TV has been with them for three weeks now and I want to know when it's finally done." + +"I hate it when it rains on Monday. The streets are slippery and I have to be very careful when I walk to the station." + +## Appendix II + +### Results from an initial pilot test and second retest + +(This appendix does not form an integral part of this Recommendation.) + +To evaluate the test methodology a pilot test was conducted as described in [b-Köster2017]. Here, eleven test conditions were tested and analysed. The results showed a high reliability but also a need for analysing the inter-rater agreement of a second retest. For the latter, a retest was conducted [b-Köster2017]. Both tests provide subjective scores from the same lab, in the same country, using the same language and the same test conditions. In both tests, 36 naïve test subjects participated. + +Figures II.1 and II.2 show examples of phase quality scores for the eleven test-conditions derived from the presented test methodology in a similar lab in Germany using German (the data is taken from [b-Köster2017]). The two figures illustrate that subjects in two independent tests show a high degree of agreement on the phase quality scores for a common set of eleven conditions. For illustrative purposes, Figures II.1 and II.2 only show scores for the phase quality scores, but there was very good agreement for the dimension quality score as well. Table II.1 shows the correlation between the pilot and the retest scores for the sets of eleven conditions. + +Typical results for the test methodology should include means and standard deviations for each condition in the test computed over the number of votes per condition. + +Additional statistical analyses could include comparison of pairs of test conditions by student's t-test or evaluation of experimental factors by analysis of variance. Procedures and guidelines for the evaluation of individual subject's performance are included in the ITU-T Handbook "Practical procedures for subjective testing" [b-ITU-T\_Testing]. + +![Bar chart showing subjective quality ratings (MOS_CO, MOS_U, MOS_SP, MOS_IN) for eleven test conditions. The y-axis represents the Mean Opinion Score (MOS) from 1 to 5. The x-axis lists the conditions: (1) Clean, (2) Sidetone, (3) Delay, (4) Echo, (5) Packet-loss, (6) Noise, (7) Attenuation, (8) Codec, (9) Noise + echo, (10) Codec + sidetone, (11) Delay + packet-loss. Error bars represent 95% confidence intervals.](d6ef69cf73d5a878b050b2d7599e1816_img.jpg) + +| Condition | MOS_CO | MOS_U | MOS_SP | MOS_IN | +|--------------------------|--------|-------|--------|--------| +| (1) Clean | 4.0 | 4.2 | 4.1 | 4.1 | +| (2) Sidetone | 2.2 | 3.9 | 2.2 | 3.0 | +| (3) Delay | 3.5 | 4.0 | 3.8 | 3.3 | +| (4) Echo | 1.7 | 3.6 | 2.4 | 2.1 | +| (5) Packet-loss | 2.7 | 2.5 | 3.6 | 3.3 | +| (6) Noise | 2.0 | 2.3 | 2.8 | 2.9 | +| (7) Attenuation | 3.4 | 3.4 | 3.9 | 3.7 | +| (8) Codec | 1.8 | 1.8 | 3.6 | 2.6 | +| (9) Noise + echo | 1.7 | 2.3 | 1.6 | 2.3 | +| (10) Codec + sidetone | 1.7 | 1.8 | 1.6 | 2.3 | +| (11) Delay + packet-loss | 2.9 | 2.6 | 3.7 | 2.7 | + +Bar chart showing subjective quality ratings (MOS\_CO, MOS\_U, MOS\_SP, MOS\_IN) for eleven test conditions. The y-axis represents the Mean Opinion Score (MOS) from 1 to 5. The x-axis lists the conditions: (1) Clean, (2) Sidetone, (3) Delay, (4) Echo, (5) Packet-loss, (6) Noise, (7) Attenuation, (8) Codec, (9) Noise + echo, (10) Codec + sidetone, (11) Delay + packet-loss. Error bars represent 95% confidence intervals. + +P.804(17)\_FII.1 + +**Figure II.1 – Subjective quality ratings resulting from the first pilot test; the overall conversational quality (MOS\_CO), and the quality of the three conversational phases (MOS\_LI, MOS\_SP and MOS\_IN). The error-bars display the 95% confidence intervals** + +![Bar chart showing subjective quality ratings (MOS) for 11 conditions. The Y-axis represents the Mean Opinion Score (MOS) from 1 to 5. The X-axis lists 11 conditions: (1) Clean, (2) Sidetone, (3) Delay, (4) Echo, (5) Packet-loss, (6) Noise, (7) Attenuation, (8) Codec, (9) Noise + echo, (10) Codec + sidetone, and (11) Delay + packet-loss. For each condition, four bars are shown: MOS_CO (Overall Conversational Quality), MOS_U (Listening Phase), MOS_SP (Speaking Phase), and MOS_IN (Interaction Phase). Error bars represent 95% confidence intervals.](5ed9189841659dfb01f809b8e3b21f74_img.jpg) + +| Condition | MOS_CO | MOS_U | MOS_SP | MOS_IN | +|--------------------------|--------|-------|--------|--------| +| (1) Clean | 3.8 | 4.2 | 4.1 | 4.3 | +| (2) Sidetone | 2.1 | 3.4 | 2.3 | 2.8 | +| (3) Delay | 3.1 | 4.1 | 3.8 | 2.6 | +| (4) Echo | 1.8 | 3.1 | 1.8 | 2.3 | +| (5) Packet-loss | 3.4 | 3.1 | 4.0 | 3.6 | +| (6) Noise | 2.3 | 2.6 | 2.5 | 2.8 | +| (7) Attenuation | 3.7 | 3.8 | 3.8 | 4.1 | +| (8) Codec | 2.2 | 1.9 | 3.4 | 2.6 | +| (9) Noise + echo | 1.8 | 2.1 | 1.7 | 2.2 | +| (10) Codec + sidetone | 1.8 | 1.8 | 1.7 | 2.2 | +| (11) Delay + packet-loss | 2.7 | 2.8 | 3.6 | 2.6 | + +Bar chart showing subjective quality ratings (MOS) for 11 conditions. The Y-axis represents the Mean Opinion Score (MOS) from 1 to 5. The X-axis lists 11 conditions: (1) Clean, (2) Sidetone, (3) Delay, (4) Echo, (5) Packet-loss, (6) Noise, (7) Attenuation, (8) Codec, (9) Noise + echo, (10) Codec + sidetone, and (11) Delay + packet-loss. For each condition, four bars are shown: MOS\_CO (Overall Conversational Quality), MOS\_U (Listening Phase), MOS\_SP (Speaking Phase), and MOS\_IN (Interaction Phase). Error bars represent 95% confidence intervals. + +P.804(17)\_Fil.2 + +**Figure II.2 – Subjective quality ratings resulting from the retest; the overall conversational quality (MOS\_CO), and the quality of the three conversational phases (MOS\_LI, MOS\_SP and MOS\_IN). The error-bars display the 95% confidence intervals** + +**Table II.1 – Correlation between the pilot test and retest scores for the 11 conditions; CO – overall quality, LI – listening phase, SP – speaking phase, IN – interaction phase, Noi – noisiness, Dis – discontinuity, Col – coloration, Lou – loudness, Ios – impact of one's own voice on speaking, Dos – degradation of one's own voice, and Int – interactivity** + +| MOS_CO | MOS_LI | MOS_SP | MOS_IN | MOS_noi | MOS_dis | MOS_col | MOS_lou | MOS_ios | MOS_dos | MOS_int | +|--------|--------|--------|--------|---------|---------|---------|---------|---------|---------|---------| +| 0.92 | 0.91 | 0.98 | 0.92 | 0.83 | 0.89 | 0.90 | 0.93 | 0.99 | 0.97 | 0.92 | + +## Bibliography + +- [b-ITU-T\_Testing] ITU-T Handbook (2011), *Practical procedures for subjective testing*. +- [b-Appel2002] R. Appel and J.G. Beerends, *On the Quality of Hearing One's Own Voice*, *Journal of the Audio Engineering Society*, vol. 50, no. 4, pp. 237-248, 2002. +- [b-Guéguin2008] M. Guéguin, R. Le Bouquin-Jeannès, V. Gautier-Turbin, G. Faucon, and V. Barriac, *On the Evaluation of the Conversational Speech Quality in Telecommunications*: EURASIP J.Adv. Signal Process, 2008. +- [b-Köster2014] F. Köster and S. Möller, *Analyzing perceptual dimensions of conversational speech*, in Proc. *INTERSPEECH*, Singapore, Singapore, 2014, pp. 2041-2045. +- [b-Köster2015a] F. Köster and S. Möller, *Perceptual Speech Quality Dimensions in a Conversational Situation*, in Proc. *INTERSPEECH*, Dresden, Germany, 2015, pp. 2544-2548. +- [b-Köster2015b] F. Köster, D. Guse, M. Wältermann, and S. Möller, *Comparison between the discrete ACR scale and an extended continuous scale for the quality assessment of transmitted speech*, in *Fortschritte der Akustik – DAGA 2015: Plenarvortr. u. Fachbeitr. d. 40. Dtsch. Jahrestg. f. Akust.*, Nürnberg, 2015, pp. 150-153. +- [b-Köster2017] F. Köster, *Multidimensional Analysis of Conversational Telephone Speech*, Springer, Berlin, 2017. +- [b-Wältermann2010] M. Wältermann, A. Raake, and S. Möller, *Quality Dimensions of Narrowband and Wideband Speech Transmission*, *Acta Acustica united with Acustica*, 2010, pp. 1090-1103. + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | Tariff and accounting principles and international telecommunication/ICT economic and policy issues | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling, and associated measurements and tests | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects, next-generation networks, Internet of Things and smart cities | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.833.1-200904-I_PDF-E/raw.md b/marked/P/T-REC-P.833.1-200904-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..2cbbf83027a908c60df53d508a6d9c5858ebc2d8 --- /dev/null +++ b/marked/P/T-REC-P.833.1-200904-I_PDF-E/raw.md @@ -0,0 +1,343 @@ + + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.833.1** + +(04/2009) + +SERIES P: TERMINALS AND SUBJECTIVE AND +OBJECTIVE ASSESSMENT METHODS + +Methods for objective and subjective assessment of +speech quality + +--- + +**Methodology for the derivation of equipment +impairment factors from subjective +listening-only tests for wideband speech codecs** + +Recommendation ITU-T P.833.1 + +# ITU-T P-SERIES RECOMMENDATIONS **TERMINALS AND SUBJECTIVE AND OBJECTIVE ASSESSMENT METHODS** + +| | | | +|-----------------------------------------------------------------------------------------------|---------------|--------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Voice terminal characteristics | Series | P.30 | +| | | P.300 | +| Reference systems | Series | P.40 | +| Objective measuring apparatus | Series | P.50 | +| | | P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of speech quality | Series | P.80 | +| | | P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | +| Transmission performance and QoS aspects of IP end-points | Series | P.1000 | +| Communications involving vehicles | Series | P.1100 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +## **Recommendation ITU-T P.833.1** + +## **Methodology for the derivation of equipment impairment factors from subjective listening-only tests for wideband speech codecs** + +## **Summary** + +Recommendation ITU-T P.833.1 describes an extension of the methodology for deriving equipment impairment factors from subjective listening-only tests. It is intended that it primarily be applied to determining wideband equipment impairment factors $l_{e,wb}$ , capturing the degradation introduced by wideband speech codecs. The resulting wideband equipment impairment factors derived by this methodology are intended to be used on the extended transmission rating scale underlying the E-model, see Appendix II of Recommendation ITU-T G.107. They will reflect the auditory impairments of the corresponding equipment in a listening-only mode. + +## **Source** + +Recommendation ITU-T P.833.1 was approved on 29 April 2009 by ITU-T Study Group 12 (2009-2012) under Recommendation ITU-T A.8 procedure. + +## **Keywords** + +Codec, E-model, impairment factor method, wideband speech transmission. + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +### NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure e.g. interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database at . + +© ITU 2009 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +## CONTENTS + +| | Page | +|--------------------------------------------------------------------------------------|------| +| 1 Scope ..... | 1 | +| 2 References..... | 1 | +| 3 Definitions ..... | 2 | +| 4 Abbreviations and acronyms ..... | 2 | +| 5 Conventions ..... | 2 | +| 6 Selection of experiment parameters..... | 2 | +| 6.1 Reference conditions for wideband speech codecs without transmission errors..... | 3 | +| 6.2 Reference conditions for wideband speech codecs with transmission errors..... | 4 | +| 7 Test method ..... | 4 | +| 8 Derivation of wideband equipment impairment factors ..... | 5 | +| 8.1 Scale transformation (step 1)..... | 6 | +| 8.2 Linear interpolation of the test results (step 2)..... | 6 | +| 8.3 Additivity check (step 3) ..... | 7 | +| 8.4 Derivation of $I_{e,wb}$ values for transmission error conditions (step 4) ..... | 7 | +| 8.5 Additivity check (step 5) ..... | 7 | +| 9 Application of the derived wideband equipment impairment factor values..... | 7 | +| Bibliography..... | 8 | + +## Introduction + +ITU-T has introduced the impairment factor method in order to describe the joint effect of different types of degradations on overall transmission quality. According to this method, transmission impairments can be transformed into so-called "psychological factors", which are assumed to be additive on a psychological scale. The transmission rating scale (*R*-scale) underlying the E-model (see Recommendation ITU-T G.107) is assumed to reflect such a psychological scale. + +The impairment factor method allocates a value of impairment to each parameter describing the transmission channel, and then allows the simple addition of these impairments to determine the overall impairment. The overall transmission rating *R* is finally calculated by subtracting the overall impairment from the maximum value *Rmax*. Whereas the *R*-scale and the E-model have originally been formulated for narrow-band (300-3400 Hz) speech transmission only, Appendix II of Recommendation ITU-T G.107 presents an extension of this scale to wideband (50-7000 Hz) speech transmission. Although a complete wideband version of the E-model is not yet available, Appendix IV of Recommendation ITU-T G.113 provides values for wideband equipment impairment factors *Ie,wb*, which describe the degradation resulting from selected wideband speech codecs on the extended *R*-scale. This Recommendation describes a methodology for how *Ie,wb* values can be derived for other wideband speech codecs, from the results of listening-only tests. + +## Recommendation ITU-T P.833.1 + +## Methodology for the derivation of equipment impairment factors from subjective listening-only tests for wideband speech codecs + +## 1 Scope + +It is assumed that users of this Recommendation are familiar with [ITU-T P.833]. + +The extension of the methodology described in this Recommendation is intended to derive wideband equipment impairment factors $I_{e,wb}$ quantifying the impairment introduced by wideband (50-7000 Hz) speech codecs, including or not the effects of transmission errors, such as random bit errors, random packet loss or discard, or bursty packet loss or discard. Such wideband equipment impairment factors have been introduced as a simplified measure of the degradation introduced by wideband speech codecs on the integral transmission quality from mouth to ear. They are in no way an exact description of the effects related to each individual codec or codec tandem, which may be very diverse in their perceptual nature. Instead, they represent the relative degradation in comparison to other impairments occurring in a connection. + +In order to provide guidance on the quantitative amount of impairment introduced by such codecs, a framework of wideband equipment impairment factor values for several wideband speech codecs has been derived by ITU-T, see Appendix IV of [ITU-T G.113]. The derivation of these values was based on many subjective listening-only tests carried out in different test laboratories, so as to guarantee stable values for all the codecs under investigation so far, as well as their relations to each other. If new equipment impairment factor values for different codecs have to be derived, then the overall consistency with the established framework is of primary importance if results are to be obtained that are valid for network planning. The methodology described in this Recommendation was designed to fulfil this requirement. + +So far, wideband speech codecs are the only wideband elements for which impairment factors have been derived. For asynchronous tandems of multiple codecs of the same type, or of multiple codecs of different types, it is assumed that individual equipment impairment factors are additive. The overall equipment impairment factor for the chain of codecs is then calculated as a simple sum of all the individual $I_{e,wb}$ s. Experimental test data collected provide evidence that this simple additivity is not satisfied for all the potential combinations of codecs [b-Möller]. In some cases, order effects exist, i.e., the tandem of codec A followed by codec B results in a different degradation than codec B followed by codec A. These deviations from the pure additivity property are an item for further study. Impairment factor values for other wideband network elements, resulting in other types of degradations and to be used in a wideband version of the E-model, have not yet been defined and are for further study as well. + +## 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +[ITU-T G.107] Recommendation ITU-T G.107 (2008), *The E-model, a computational model for use in transmission planning*. + +- [ITU-T G.113] Recommendation ITU-T G.113 (2007), *Transmission impairments due to speech processing*. +- [ITU-T G.722] Recommendation ITU-T G.722 (1988), *7 kHz audio-coding within 64 kbit/s*. +- [ITU-T G.722.1] Recommendation ITU-T G.722.1 (2005), *Low-complexity coding at 24 and 32 kbit/s for hands-free operation in systems with low frame loss*. +- [ITU-T G.722.2] Recommendation ITU-T G.722.2 (2003), *Wideband coding of speech at around 16 kbit/s using Adaptive Multi-Rate Wideband (AMR-WB)*. +- [ITU-T P.341] Recommendation ITU-T P.341 (2005), *Transmission characteristics for wideband (150-7000 Hz) digital hands-free telephony terminals*. +- [ITU-T P.800] Recommendation ITU-T P.800 (1996), *Methods for subjective determination of transmission quality*. +- [ITU-T P.830] Recommendation ITU-T P.830 (1996), *Subjective performance assessment of telephone-band and wideband digital codecs*. +- [ITU-T P.833] Recommendation ITU-T P.833 (2001), *Methodology for derivation of equipment impairment factors from subjective listening-only tests*. + +## 3 Definitions + +None. + +## 4 Abbreviations and acronyms + +This Recommendation uses the following abbreviations and acronyms: + +| | | +|-------|---------------------------------------------| +| ADPCM | Adaptive Differential Pulse Code Modulation | +| CELP | Code Excited Linear Prediction | +| MLTC | Modified Lapped Transform Coding | +| MOS | Mean Opinion Score | + +## 5 Conventions + +None. + +# 6 Selection of experiment parameters + +The listening-only test from which $I_{e,wb}$ values are to be derived should fulfil the general requirements for listening-only tests given in [ITU-T P.800] and [ITU-T P.830], as well as the requirements set in [ITU-T P.833]. However, in contrast to [ITU-T P.833], the test should be carried out in a wideband or mixed-band mode, implying several changes given hereafter. + +In contrast to narrow-band, there is no standard wideband terminal which would be representative for most wideband connections. For the given purpose, the test should be carried out either with diotic or monotonic headphone presentation. The frequency response of the terminal should represent a wideband characteristic in the range 50 to 7000 Hz, as it is defined in [ITU-T P.341]. + +For reasons given in [ITU-T P.833], subjective tests aiming at deriving equipment impairment factors should include a number of reference conditions. These references are necessary for anchoring impairment factor values, and they will guarantee that new equipment impairment factor values fit into the existing system given in Appendix IV of [ITU-T G.113]. + +It is recommended that different types of wideband speech codecs be used as a reference for experiments deriving wideband equipment impairment factors (i.e., waveform codec as G.722 or hybrid codec as G.722.2). The exact reference conditions to be used vary depending on whether transmission errors are to be considered, and whether an additivity check has to be performed. + +In addition, a narrow-band codec may be used in case of mixed-band test. + +## 6.1 Reference conditions for wideband speech codecs without transmission errors + +When wideband equipment impairment factors for codecs disregarding transmission errors are determined, the set of 12 reference codec conditions given in Table 1 should be included in the subjective test conditions. This list has been chosen from the codecs for which values are already defined in Appendix IV to [ITU-T G.113], and they have been selected to cover the whole range of $I_{e,wb}$ values. + +**Table 1 – Reference conditions for wideband speech codecs without transmission errors** + +| No. | Abbreviation | Codec type | Reference | Operating rate (kbit/s) | $I_{e,wb}$ value | +|-----|---------------|--------------------|---------------|-------------------------|------------------| +| 1 | Clean | Linear PCM, 16 bit | – | – | 0 | +| 2 | G.722.2@23.05 | CELP | ITU-T G.722.2 | 23.05 | 1 | +| 3 | G.722.2@19.85 | CELP | ITU-T G.722.2 | 19.85 | 3 | +| 4 | G.722.2@15.85 | CELP | ITU-T G.722.2 | 15.85 | 7 | +| 5 | G.722.2@14.25 | CELP | ITU-T G.722.2 | 14.25 | 10 | +| 6 | G.722@64 | ADPCM | ITU-T G.722 | 64 | 13 | +| 7 | G.722.1@32 | MLTC | ITU-T G.722.1 | 32 | 13 | +| 8 | G.722.1@24 | MLTC | ITU-T G.722.1 | 24 | 19 | +| 9 | G.722@56 | ADPCM | ITU-T G.722 | 56 | 20 | +| 10 | G.722.2@8.85 | CELP | ITU-T G.722.2 | 8.85 | 26 | +| 11 | G.722@48 | ADPCM | ITU-T G.722 | 48 | 31 | +| 12 | G.722.2@6.6 | CELP | ITU-T G.722.2 | 6.6 | 41 | + +It is important to check the additivity of the newly derived wideband equipment impairment factor in the framework of other equipment impairment factor values defined so far. If such an additivity check is not performed, the property of a simple summation of equipment impairment factors in order to cater for codec tandems should not be regarded as valid. Table 2 gives a minimum number of additional reference conditions which should, in any case, be included in the test set to allow for a rough additivity check. It is preferable, however, to test inter-codec tandem operations with a larger set of similar conditions, including triple tandems in different codec orders. + +**Table 2 – Reference conditions for the additivity check in tandem operation of wideband speech codecs without transmission errors** + +| No. | Tandem operation | Reference codec type | Operating rate (kbit/s) | Ie,wb value | +|-----|---------------------------|----------------------|-------------------------|--------------------------------| +| 13 | G.722.2@19.85*(new codec) | CELP | 19.85 | $3 + Ie,wb(\text{new codec})$ | +| 14 | G.722.2@14.25*(new codec) | CELP | 14.25 | $10 + Ie,wb(\text{new codec})$ | +| 15 | G.722@64*(new codec) | ADPCM | 64 | $13 + Ie,wb(\text{new codec})$ | +| 16 | G.722.1@32*(new codec) | MLTC | 32 | $13 + Ie,wb(\text{new codec})$ | +| 17 | G.722.1@24*(new codec) | MLTC | 24 | $19 + Ie,wb(\text{new codec})$ | +| 18 | G.722@48*(new codec) | ADPCM | 48 | $31 + Ie,wb(\text{new codec})$ | +| 19 | (new codec)*G.722.2@19.85 | CELP | 19.85 | $Ie,wb(\text{new codec}) + 3$ | +| 20 | (new codec)*G.722.2@14.25 | CELP | 14.25 | $Ie,wb(\text{new codec}) + 10$ | +| 21 | (new codec)*G.722@64 | ADPCM | 64 | $Ie,wb(\text{new codec}) + 13$ | +| 22 | (new codec)*G.722.1@32 | MLTC | 32 | $Ie,wb(\text{new codec}) + 13$ | +| 23 | (new codec)*G.722.1@24 | MLTC | 24 | $Ie,wb(\text{new codec}) + 19$ | +| 24 | (new codec)*G.722@48 | ADPCM | 48 | $Ie,wb(\text{new codec}) + 31$ | + +NOTE – A\*B designates asynchronous tandeming of codecs A and B, B followed by A. + +## 6.2 Reference conditions for wideband speech codecs with transmission errors + +When wideband equipment impairment factors for codecs under the effects of transmission errors are determined, the same reference conditions as given in Table 1 should be applied. In addition to these conditions, supplementary conditions may be included in the test from the codecs listed in Appendix IV of [ITU-T G.113], including perceptively noticeable degradation due to transmission errors (random bit errors, random packet loss, bursty packet loss, or propagation errors in terms of error patterns). However, no *Ie,wb* values have yet been agreed upon for these codecs, so that these conditions cannot be considered as references. Preliminary values can be found, e.g., in [b-Möller], and the *Ie,wb* values obtained for these codecs may be compared to these values. Larger deviations from the values will have to be considered when interpreting the newly-derived *Ie,wb* values for codecs including transmission errors. + +## 7 Test method + +The test method generally follows the recommendations given in [ITU-T P.833]. Due to the lack of available references for codecs under transmission errors, the number of mandatory reference conditions is slightly different. Table 3 summarizes the test conditions to be included in the different parts of the experiment. + +The test conditions do not need to be included into one single test or test session; in fact, the number of test conditions in one session should be limited in order to avoid listener fatigue, following the recommendations given in [ITU-T P.800]. + +**Table 3 – Overview of test conditions for the different parts of the experiment** + +| Part | Purpose | Test conditions | Mandatory/
optional | Min. overall
$\Sigma$ test cond. | +|------|-----------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|------------------------|-------------------------------------| +| A | Determination of $Ie,wb$ for the new codec in error-free conditions | References 1-12. | Mandatory | 13 | +| | | New codec in single operation. | Mandatory | | +| | | New codec in single operation, at 2 alternative speech input levels. | Optional | | +| | | Additional wideband codec references. | Optional | | +| B | Additivity check | References 13-24. | Mandatory | 14 | +| | | New codec in double and triple self-tandem operation, i.e., (new codec)*(new codec) and (new codec)*(new codec)*(new codec). | Mandatory | | +| | | New codec in double and triple tandem operation with other wideband codecs, i.e., (new codec)*(other codec) and/or (new codec)*(new codec)*(other codec) and/or (new codec)*(other codec)*(other codec), potentially in different orders. | Optional | | +| C | Determination of $Ie,wb$ for the new codec in transmission error conditions | New codec in single operation in different transmission error conditions ( $m$ conditions). | Mandatory | $m$ | +| | | Additional references according to clause 6.2. | Optional | | + +## 8 Derivation of wideband equipment impairment factors + +The methodology for deriving wideband equipment impairment factors follows mostly [ITU-T P.833], however with some modifications of the calculations. It consists of three to five steps, depending on whether transmission errors are taken into account or not: + +Step 1: Scale transformation of the subjective test data. + +Step 2: Derivation of a stable $Ie,wb$ value for the codec under test, in single codec operation without transmission errors, via a linear interpolation of the test results. + +Step 3: Additivity check. + +If transmission errors are under investigation, the following additional steps are to be taken: + +Step 4: Derivation of stable $Ie,wb$ values for different transmission error conditions in single codec operation. + +Step 5: Additivity check. + +These steps are described below. + +## 8.1 Scale transformation (step 1) + +Mean opinion scores (MOSs) are determined for all test conditions, including the references listed in Table 3. These MOS results have to be transformed to the extended $R$ -scale. For this purpose, they are first transformed to the non-extended, narrow-band $R_{NB}$ -scale (range [0;100], subscript $NB$ ), using either a numeric or graphic inversion of the relationship between MOS and $R_{NB}$ given in the E-model: + +$$\begin{aligned} &\text{for } MOS=1.0: & R_{NB}=0 \\ &\text{for } 1.0 < MOS < 4.5: & MOS=1+0.035 \cdot R_{NB} + R_{NB} \cdot (R_{NB}-60) \cdot (100-R_{NB}) \cdot 7 \cdot 10^{-6} \quad (1) \\ &\text{for } MOS \geq 4.5: & R_{NB}=100 \end{aligned}$$ + +In case that the maximum $MOS_{max}$ value of the test is higher than 4.5, the range of MOS values of that test first has to be linearly transformed to a maximum value of 4.5, so that equation 1 can be applied: + +$$MOS_{norm,i} = \frac{MOS_i - 1}{MOS_{max} - 1} \cdot 3.5 + 1 \quad (2)$$ + +These values have to be transformed to the extended $R$ -scale (range [0;129]). For this purpose, a linear expansion of the obtained $R_{NB}$ values to the wideband $R_{WB}$ values (subscript $WB$ ) is carried out: + +$$R_{WB} = 1.29 \cdot R_{NB} \quad (3)$$ + +From the resulting values for $R_{WB}$ , the corresponding raw $Ie,wb,sub$ values (subscript $sub$ for "subjective test") can be calculated by defining the $R_{WB}$ -value for reference condition No. 1 (see Table 1) as an anchor, thus: + +$$Ie,wb,sub = R_{WB}(\text{condition No. 1}) - R_{WB}(\text{test condition}) \quad (4)$$ + +This equation results in $Ie,wb,sub$ for the reference condition No. 1 always set to 0. The outcome of step 1 is an $Ie,wb,sub$ value for each test condition. It reflects the specific test condition, and it is not necessarily consistent with wideband equipment impairment factors defined in Appendix IV to [ITU-T G.113]. + +### 8.2 Linear interpolation of the test results (step 2) + +For all 12 reference conditions of Table 1, as well as possibly for all supplementary reference conditions involving only codecs for which $Ie,wb$ values have already been defined, pairs of expected equipment impairment factors $Ie,wb,exp$ and observed values $Ie,wb,sub$ are now available. These pairs can be represented as a scatter plot. A linear interpolation using a straight line + +$$Ie,wb,sub = a \cdot Ie,wb,exp + b \quad (5)$$ + +can now be made. The coefficients $a$ and $b$ are determined numerically, approximating all the reference pairs in a least-squares sense. From this approximation, a stable equipment impairment factor value for the codec under test ( $Ie,wb = Ie,wb,exp$ ) can be derived. + +In rare cases, the linear transformation may result in a negative $Ie,wb$ value for the codec under investigation. This might occur if the related subjective ratings are close to or better than the one for reference condition No. 1 of Table 1. In this case, $Ie,wb$ should be set to zero instead. + +### 8.3 Additivity check (step 3) + +The equipment impairment factor derived in step 2 does not necessarily satisfy the additivity property of $I_{e,wb}$ s. This has to be checked for both tandems of the new codec alone and mixed tandems with codecs for which $I_{e,wb}$ values have already been defined. The procedure is the same as described in [ITU-T P.833]. If more than 4 out of 14 tandem conditions (2 pure tandems of the codec under investigation and 12 reference tandem conditions Nos 13-24, see Table 2) show major deviations from the interpolation line, the additivity property should not be regarded as having been satisfied. In this case, the wideband equipment impairment factor derived from the experiment will not properly represent the degradations occurring in tandem operations of the new codec. + +## 8.4 Derivation of $I_{e,wb}$ values for transmission error conditions (step 4) + +Using the interpolation line and the procedure of step 2, $I_{e,wb}$ values can be derived for all conditions of the new codec including transmission errors, as well as for all conditions of the codecs given in Appendix IV of [ITU-T G.113] which include transmission errors. Because no reference $I_{e,wb}$ values are defined for the latter codecs, it cannot be checked whether the inclusion of transmission errors may have an effect on the judgments, and consequently also on the interpolation line. However, the obtained values may be compared to the ones given in [b-Möller]. Any large deviations may indicate that the rating behaviour changed for the error conditions in comparison to the error-free case; the resulting $I_{e,wb}$ values for codecs under transmission errors should then be considered with care. + +Graphs or tables of $I_{e,wb}$ values for transmission error conditions can be used for a plausibility check. The minimum consistency to be reached is to have non-decreasing wideband equipment impairment factor values for increasing transmission error rates. + +### 8.5 Additivity check (step 5) + +A similar additivity check as in step 3 should be carried out on the wideband equipment impairment factor values derived in step 4, using all the available tandem conditions of the codec under investigation with transmission errors introduced and other codecs for which impairment factors have already been derived. + +## 9 Application of the derived wideband equipment impairment factor values + +Due to the inherent variability of subjective ratings, $I_{e,wb}$ values derived using this methodology cannot be expected to represent very exact quantitative measures of impairment that a subject in a specific situation would experience. Instead, they represent the additional degradation expected to be introduced by the respective wideband speech codec, expressed on the extended $R$ -scale. A value of $I_{e,wb} = 0$ signifies that the corresponding codec introduces no additional impairment compared to the clean (16 kHz sampling frequency, 16-bit quantization, linear PCM) channel. Unless additivity has been proven in steps 3 and 5, it cannot be concluded from the results how large the impairment would be when the codec is combined with other codecs. + +In addition, it is unclear how the codec degradation combines with other degradations which are currently covered by the (narrow-band) E-model, like ambient noise, circuit noise, non-optimal loudness, sidetone, talker or listener echo, or delay. These items may be covered by future wideband extensions of the E-model and are under study in ITU-T SG 12. Unless such a model is available, $I_{e,wb}$ values cannot be combined with other types of impairment factors in order to obtain an estimate of the integral wideband speech quality mouth-to-ear. + +## Bibliography + +- [b-Möller] Möller, S., Raake, A., Kitawaki, N., Takahashi, A., Wältermann, M., (2006), *Impairment Factor Framework for Wide-band Speech Codecs*, IEEE Transactions on Audio, Speech, and Language Processing, Vol. 14, No. 6, pp. 1969-1976. + + + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|---------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Terminals and subjective and objective assessment methods | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects and next-generation networks | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.840-200311-I_PDF-E/raw.md b/marked/P/T-REC-P.840-200311-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..6d70e8807632c43204ea93040532a5b3012cbc55 --- /dev/null +++ b/marked/P/T-REC-P.840-200311-I_PDF-E/raw.md @@ -0,0 +1,894 @@ + + +![ITU logo: A globe with a lightning bolt and the letters ITU.](2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg) + +ITU logo: A globe with a lightning bolt and the letters ITU. + +INTERNATIONAL TELECOMMUNICATION UNION + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.840** + +(11/2003) + +SERIES P: TELEPHONE TRANSMISSION QUALITY, +TELEPHONE INSTALLATIONS, LOCAL LINE +NETWORKS + +Methods for objective and subjective assessment of +quality + +--- + +**Subjective listening test method for evaluating +circuit multiplication equipment** + +ITU-T Recommendation P.840 + +--- + +# ITU-T P-SERIES RECOMMENDATIONS + +## TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS + +| | | | +|-----------------------------------------------------------------------------------------------|---------------|-----------------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Subscribers' lines and sets | Series | P.30
P.300 | +| Transmission standards | Series | P.40 | +| Objective measuring apparatus | Series | P.50
P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of quality | Series | P.80
P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +# **ITU-T Recommendation P.840** + +# **Subjective listening test method for evaluating circuit multiplication equipment** + +## **Summary** + +This Recommendation describes a subjective listening test method which can be used to evaluate the speech quality of Circuit Multiplication Equipment (CME). It is intended for use with CME systems, such as those described in ITU-T Recs G.763, G.767, G.768 (DCME), G.765 (PCME) and G.769/Y.1242 (IP-CME) which use digital speech interpolation (DSI) techniques. + +In this version, the scope is expanded to more recent speech coders implemented in CME. Updating the Recommendation includes tandemming situations and comfort noise test configuration. + +The new Appendix I gives guidance for conversation tests. + +## **Source** + +ITU-T Recommendation P.840 was approved by ITU-T Study Group 12 (2001-2004) under the ITU-T Recommendation A.8 procedure on 13 November 2003. + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications. The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure e.g. interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementors are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database. + +© ITU 2004 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +# CONTENTS + +| | | Page | +|------------|-------------------------------------------------------------------------|------| +| 1 | Scope ..... | 1 | +| 1.1 | Purpose ..... | 1 | +| 1.2 | Test philosophy ..... | 1 | +| 2 | References..... | 2 | +| 3 | Definitions ..... | 3 | +| 4 | Abbreviations..... | 3 | +| 5 | Source recordings ..... | 4 | +| 5.1 | Apparatus and environment..... | 4 | +| 5.2 | Speech material ..... | 4 | +| 5.3 | Procedure..... | 5 | +| 5.4 | Calibration signals and speech levels..... | 6 | +| 6 | Simulating system load..... | 6 | +| 6.1 | Requirements for a generic voice load simulator ..... | 6 | +| 6.2 | Determining load capacity of tested systems ..... | 7 | +| 6.3 | Controlling load applied to tested systems..... | 8 | +| 7 | Processing of the speech..... | 8 | +| 8 | Test design..... | 9 | +| 8.1 | Test No. 1 – Effect of applied load..... | 9 | +| 8.2 | Test No. 2 – Effect of digital errors in the DCME control channel ..... | 11 | +| 9 | Listening test procedure..... | 12 | +| 10 | Analysis of results ..... | 12 | +| Annex A | – Description of circuit multiplication equipment..... | 12 | +| A.1 | Definitions ..... | 12 | +| A.2 | Digital Speech Interpolation (DSI)..... | 15 | +| A.3 | Speech detection..... | 16 | +| A.4 | CME load..... | 16 | +| A.5 | Overload strategies ..... | 17 | +| A.6 | Silence reconstruction methods..... | 19 | +| A.7 | Circuit versus packet mode ..... | 19 | +| A.8 | Packet reconstruction..... | 20 | +| Annex B | – Speech material used to construct speech sequences ..... | 20 | +| Annex C | – Instructions on the use of a limited number of sentences..... | 22 | +| Appendix I | – Guidance for conversation tests..... | 22 | + + + +# **Subjective listening test method for evaluating circuit multiplication equipment1** + +# **1 Scope** + +## **1.1 Purpose** + +The purpose of this Recommendation is to describe a subjective listening test method which can be used to evaluate the speech quality of Circuit Multiplication Equipment (CME). It is intended for use with CME systems, such as those described in ITU-T Recs G.763, G.764, G.765, G.767, G.768 or as described in G.769/Y.1242, which use digital speech interpolation (DSI) techniques. For subjective test evaluations of CME systems not using digital speech interpolation techniques, channel transcoder systems described in ITU-T Rec. G.761 (*General characteristics of a 60-channel transcoder equipment*) or ITU-T Rec. G.762 (*General characteristics of a 48-channel transcoder equipment*), use ITU-T Rec. P.830. + +In addition to the listening tests described here, conversational tests are also recommended in order to thoroughly assess the combined effects of delay, echo, non-linearities, etc. Appendix I gives some guidance for conversation tests. + +For CME systems, degradations can include not only the effects of variable bit-rate coding, DSI gain (channel allocation), clipping, freezeout and noise contrast, but also those due to non-linearities in the speech detection system, such that the system may function differently for different speech input levels or activity factors. + +It may, therefore, be useful to perform additional tests such as those provided by ITU-T Rec. P.831. Listening tests play an important preliminary role in the assessment, and can supply useful information service to narrow the range of conditions needing a complete conversation test. This listening test method will not provide results for generating network application rules. Future improvements of the test will allow such results to be obtained. + +Evaluation of CME in tandem with other CME is considered in this revised version, but some effects of systems using encoding at different rates need to be completed. This Recommendation will subsequently be updated when information on these specific points becomes available. + +## **1.2 Test philosophy** + +In order for a test to satisfactorily evaluate CME performance, the test methodology should meet certain conditions. These are as follows: + +- i) the method should use principles, procedures, and instrumentation that are acceptable to ITU-T; +- ii) the method should be adaptable to different languages and should yield results that are comparable to those from other tests performed using this Recommendation; +- iii) the method should permit CME performance to be compared subjectively (or objectively) to reference conditions. Examples of suitable reference conditions are hypothetical reference connections (HRCs), white noise and speech correlated noise. The HRCs should model the facilities the CME is designed to replace, when these facilities are known. The results of the comparisons should permit making "equivalence statements" about the CME, + +--- + +1 The specifications in this Recommendation are subject to future enhancements and, therefore, should be regarded as provisional. + +e.g., a CME system is subjectively equivalent to x asynchronously tandemed 64 kbit/s PCM systems. Ideally, the method should yield results from which a network application rule can be derived; + +- iv) the CME should be tested with a realistic load simulator and circuit-under-test signal conditions applied. Most of the transitory impairments arise when the CME is operating in the range of applied load which forces the use of DSI. Therefore, to subjectively measure the effects of these impairments, it is necessary to vary the applied load on the CME up to and marginally beyond the maximum design load. The clipping produced by the speech detector is affected by the type of signal being transmitted on the circuit under test. Therefore, only a realistic speech signal which also contains appropriate additive noise should be used on the circuit under test; +- v) the methodology should, ideally, yield results which can be used to produce new opinion models or modify existing models. + +# 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +- ITU-T Recommendation G.703 (2001), *Physical/electrical characteristics of hierarchical digital interfaces*. +- ITU-T Recommendation G.704 (1998), *Synchronous frame structures used at 1544, 6312, 2048, 8448 and 44 736 kbit/s hierarchical levels*. +- ITU-T Recommendation G.711 (1988), *Pulse code modulation (PCM) of voice frequencies*. +- ITU-T Recommendation G.726 (1990), *40, 32, 24, 16 kbit/s Adaptive Differential Pulse Code Modulation (ADPCM)*. +- ITU-T Recommendation G.728 (1992), *Coding of speech at 16 kbit/s using low-delay code excited linear prediction*. +- ITU-T Recommendation G.729 (1996), *Coding of speech at 8 kbit/s using conjugate-structure algebraic-code-excited linear-prediction (CS-ACELP)*. +- ITU-T Recommendation G.732 (1988), *Characteristics of primary PCM multiplex operating at 2048 kbit/s*. +- ITU-T Recommendation G.733 (1988), *Characteristics of primary PCM multiplex operating at 1544 kbit/s*. +- ITU-T Recommendation G.761 (1988), *General characteristics of a 60-channel transcoder equipment*. +- ITU-T Recommendation G.762 (1988), *General characteristics of a 48-channel transcoder equipment*. +- ITU-T Recommendation G.763 (1998), *Digital circuit multiplication equipment using G.726 ADPCM and digital speech interpolation*. +- ITU-T Recommendation G.764 (1990), *Voice packetization – Packetized voice protocols*. +- ITU-T Recommendation G.765 (1992), *Packet circuit multiplication equipment*. + +- ITU-T Recommendation G.767 (1998), *Digital circuit multiplication equipment using 16 kbit/s LD-CELP, digital speech interpolation and facsimile demodulation/remodulation.* +- ITU-T Recommendation G.768 (2001), *Digital circuit multiplication equipment using 8 kbit/s CS-ACELP.* +- ITU-T Recommendation G.769/Y.1242 (2002), *Circuit multiplication equipment optimized for IP-based networks.* +- ITU-T Recommendation P.50 (1999), *Artificial voices.* +- ITU-T Recommendation P.59 (1993), *Artificial conversational speech.* +- ITU-T Recommendation P.800 (1996), *Methods for subjective determination of transmission quality.* +- ITU-T Recommendation P.810 (1996), *Modulated noise reference unit (MNRU).* +- ITU-T Recommendation P.830 (1996), *Subjective performance assessment of telephone-band and wideband digital codecs.* +- ITU-T Recommendation P.831 (1998), *Subjective performance evaluation of network echo cancellers.* + +# 3 Definitions + +This Recommendation defines the following term: + +**3.1 Q:** The ratio, in dB, of speech power to modulated noise power in the Modulated Noise Reference Unit, as described in ITU-T Rec. P.810. + +# 4 Abbreviations + +This Recommendation uses the following abbreviations: + +| | | +|--------|-----------------------------------------------------------------| +| A/D | Analogue-to-Digital | +| ADPCM | Adaptive Differential Pulse Code Modulation | +| BC | Bearer Channel | +| CME | Circuit Multiplication Equipment | +| CMS | Circuit Multiplication System | +| CUT | Circuit Under Test | +| D/A | Digital-to-Analogue | +| DCME | Digital Circuit Multiplication Equipment | +| DCMS | Digital Circuit Multiplication System | +| DLC | Dynamic Load Control | +| DSI | Digital Speech Interpolation | +| FIFO | First-In First-Out | +| HRC | Hypothetical Reference Connection | +| ICN | Ideal Circuit Noise | +| IP-CME | Circuit Multiplication Equipment optimized for IP-based network | +| IRS | Intermediate Reference System | + +| | | +|------|---------------------------------------------------| +| ISDN | Integrated Services Digital Network | +| LRE | Low Rate Encoding | +| MNRU | Modulated Noise Reference Unit (ITU-T Rec. P.810) | +| MOS | Mean Opinion Score | +| PCM | Pulse Code Modulation | +| SNR | Signal-to-Noise Ratio | +| STMR | SideTone Masking Rating | +| VBR | Variable Bit Rate | +| VLS | Voice Load Simulator | + +# 5 Source recordings + +## 5.1 Apparatus and environment + +See B.1.1/P.800 through B.1.3/P.800. + +## 5.2 Speech material + +The format of the speech material must be suitable for the opinion scale being used in the test. This will normally be the listening-quality scale, but optionally the listening-effort scale will be used (see clause 8). + +When the listening-quality scale is used, the following requirements apply: + +- i) The speech material should consist of short passages (called segments), chosen at random (from current non-technical literature or newspapers, for example), easy to understand, and more or less self-contained in meaning. +- ii) Each segment when spoken naturally should have a duration of not less than 9 seconds and not more than 11 seconds. +- iii) Each segment should consist of at least three "sentences" in a broad sense, that is, parts that can naturally be separated by pauses in speaking, but connected in meaning to what precedes and follows within the segment. +- iv) Within each segment there should be at least one pause which naturally, in view of the meaning and structure of the text, would last for 1 to 2 seconds. The other pauses must be of natural length, since unnaturally long or short pauses may well be interpreted as reducing the quality of the speech. The simplest way of ensuring this is to construct the script of each text with either a special mark or the beginning of a new paragraph at the point where the 1-to 2-second pause is desired. The talkers recording the segments can then be instructed to read consecutively, making sure that they pause for a second or two at the marked point, and pausing naturally at other points. + +When the listening-effort scale is used, the following requirements apply: + +- v) The speech material should consist of single meaningful sentences, easy to understand, chosen at random (from current non-technical literature or newspapers, for example) and assembled into groups (called segments). The number of sentences in every segment should be the same (three is the recommended number). +- vi) There should be no obvious connection of meaning between one sentence and another in the same segment. This precaution is to reduce the context information within the segment to a minimum, so as not to inflate the opinion scores artificially. + +- vii) The pauses between the sentences must be at least one second in length. This ensures that the listeners perceive the sentences as isolated from each other in meaning, and puts the CME system to the test in respect of the unbridged gaps in speech. Moreover, each segment should have a duration of not less than 9 seconds and not more than 11 seconds including the pauses. + +This structure can be attained by either: + +- giving timing cues to talkers at the recording stage (see 2.5.8.1 d of the *Handbook on Telephonometry*); or +- editing the recordings afterwards. + +Either of two approaches may be used for the listening-quality scale: + +- to have as many different segments as there are conditions (an example of suitable material from which segments may be constructed is contained in Annex B); or +- to have a more limited number, e.g., 10 segments per talker, where combinations of two segments can be used (this is shown in detail in Annex C). In this case, additional precautions would have to be taken in the analysis of variance of results from the tests. + +The first approach is essential when the listening-effort scale is used, because listening-effort scores are affected when the listener has heard the same sentences before. Enough segments must be available to cater for all the test conditions, plus a sufficient number for use in a practice session. + +## 5.3 Procedure + +A silent period containing only circuit noise of approximately one second should precede each segment and the segment should end with a similar silent period containing only the circuit noise. + +To facilitate the processing of the recorded speech through the CME, i.e., to allow for the starting and stopping of the recorders between segments and to allow time for adjusting the CME for the next test condition, segments should be separated by a 5-second gap on the tape. Therefore, the recorded source segments will have the pattern on the tape shown in Figure 1 (this is an example for the listening-quality scale; for the listening-effort scale all pauses must be at least 1 second). Note that if the speech is stored digitally (e.g., on a disk-based system), the segments will not require these 5-second gaps. + +![Diagram of a speech segment structure for listening-quality scales.](7c6d9bfe9c31ce872722d60b73d20df1_img.jpg) + +The diagram illustrates the temporal structure of a speech segment. It shows a sequence of elements: 'End of segment n-1', a '5 s gap', a '1 s silence', 'Sentence 1', a 'Normal pause', 'Sentence 2', a 'Natural 1-2 s pause', 'Sentence 3', another '1 s silence', a '5 s gap', and 'Start of segment n+1'. A bracket labeled 'Segment n' spans from the start of the first '1 s silence' to the start of the second '1 s silence', with a note below it stating '9 s < length < 11 s'. The label 'P.840\_F01' is in the bottom right corner. + +Diagram of a speech segment structure for listening-quality scales. + +**Figure 1/P.840 – Example of a speech segment in the format required for the listening-quality scales, containing three sentences** + +The recording procedure detailed in 2.5.8.1 (Listening Tests) of the *Handbook on Telephonometry* should be followed. Only the part dealing with recording through an IRS is required for this Recommendation. + +NOTE – When this recording is made through a physical IRS, the sidetone path of the IRS should be set to 12 dB STMR. This helps to stabilize the speech level of the talker. + +Segments should be played back to listeners complete with the silent period. After the segment has ended, a sufficient length of complete silence should be provided to permit the listener to vote. + +Talkers should pronounce the segment of sentences fluently but not dramatically and have no speech deficiencies such as "stutter". (See also B.1.6/P.800.) + +## 5.4 Calibration signals and speech levels + +See B.1.7 and B.1.8/P.800. + +# 6 Simulating system load + +## 6.1 Requirements for a generic voice load simulator + +Circuit Multiplication Equipment (CME), by definition, is used to gain an advantage in the number of circuits multiplexed onto a digital transmission facility. With this advantage, however, comes potential degradation of transmission quality when carried loads exceed that for which the CME was engineered. Thus, a rigorous performance evaluation of CME includes studying the behaviour of the CME under conditions of no load, engineered load, and overload. Because the transmission performance of CME under load depends critically upon the load characteristics, it is necessary to know and control simulated loads in order to properly assess CME performance. This clause describes the generic requirements for a voice load simulator for the purpose of facilitating CME performance evaluations under conditions that are meaningful. Use of voice load simulators with the generic requirements described here will also enable the comparison of results from different studies of various CME. + +Either the load simulator or the CME system itself must be programmed so as to record, for each individual segment played through it, the proportion of time during that segment for which the "live" channel was ON, or equivalent information. This quantity $x_i$ will be called the concomitant variable for segment $i$ (see 2.5.2 f of the *Handbook on Telephonometry*). + +NOTE 1 – The load simulator specified here is to be used for the performance evaluation of CME using Digital Speech Interpolation (DSI). This excludes Type A (see A.1), for which load is not an issue by virtue of the fixed time-slot assignment of the channels. + +NOTE 2 – The load simulator specified here is an "external" simulator that produces simulated speech signals so as to exercise many circuits being multiplexed onto a digital transmission facility. Some DCME frequently use "internal" load simulation of "trunk needs service" requests that simulate the output of multiple speech detector circuits and thus compete for transmission capacity, even though no simulated signals are actually transmitted; only the "live" channel under test is actually transmitting. This internal type of simulator can be very useful in the lab, but is not treated here because certain assumptions would have to be made regarding the performance characteristics of the associated speech detector simulation. + +### 6.1.1 Parameters + +A generic Voice Load Simulator (VLS) for CME performance evaluation has the following attributes (the parametric specification of which are detailed later in this clause): + +- talk-spurt characteristics; +- silence (gap) characteristics; +- background noise-fill for silent periods; +- spectral properties of the simulated speech; +- amplitude characteristics; +- physical interface, including idle-circuit specifications. + +The above are a minimum set of parameters that may have to be expanded as required; for example, time variation of the number of simulated calls might have to be studied, at which time a pertinent specification would have to be added. Also, only simulated speech signals are discussed. It may be desirable to add simulated tones, signalling frequencies, and voiceband data of various types at a later date. + +### 6.1.2 Requirements + +#### 6.1.2.1 General + +These requirements apply to generic VLS testing a CME. Accordingly, the CME must receive digital signals from the VLS that simulate multiple and independent sources of speech similar to that which is observed in telephone networks. To meet the "multiple and independent" condition, it will be assumed that the VLS output is to several T1 or E1 interfaces. + +#### 6.1.2.2 Talk-spurt characteristics + +The characteristics of talk-spurts are specified in ITU-T Rec. P.59. + +#### 6.1.2.3 Silence (gap) characteristics + +The characteristics of silences are specified in ITU-T Rec. P.59. + +#### 6.1.2.4 Background noise-fill for silent periods + +Noise should be inserted into the silent periods (gaps) so that the performance of DSI in the presence of noise can be studied. It is desirable to have the noise level adjustable; a default value of $-58 \text{ dBm0}$ is provisionally recommended. + +#### 6.1.2.5 Properties of the simulated speech + +The artificial voice signal of ITU-T Rec. P.50 shall be used as a basis for simulating the characteristics of human speech. This signal can then be switched on/off according to the talk-spurt and silence duration statistics described in ITU-T Rec. P.59. + +#### 6.1.2.6 Physical interface + +The load simulator should have T1 and/or E1 outputs which have physical, electrical, coding, frame structure, alignment, and signalling characteristics as per ITU-T Recs G.703, G.704, G.711 and G.732 (2048 kbit/s) or G.733 (1544 kbit/s). + +## 6.2 Determining load capacity of tested systems + +The average applied load equals the product of the number of circuits in use, $N$ , and the average speech activity. The load capacity of the tested system equals the maximum load that the system is designed to handle, $L_{\max}$ , where $L_{\max}$ equals the product of the maximum number of circuits, $N_{\max}$ , possible and the average speech activity. The load capacity can be determined by: + +- i) obtaining the manufacturer's specifications; +- ii) calculation. + +After the load capacity is determined, the partial loads at which the system will be tested can be determined. The partial loads are: + +$$L_i = c_i L_{\max}$$ + +where + +$$c_i = 0.0, 0.75, 1.0 \text{ and } 1.2.$$ + +## 6.3 Controlling load applied to tested systems + +The load applied to the CME can be changed by varying $N$ and the activity factor. For these tests the speech activity factor will be assumed constant at 28%. Therefore, to obtain a partial load $L_i$ , it is necessary to calculate the number of activity circuits which come closest to achieving this desired value. + +For example, if $L_{\max} = 48$ and if a partial load of $L_i = 0.75$ , $L_{\max}$ is desired and the speech activity factor of 28% is assumed, then the number of active circuits, $N_{\text{active}}$ is calculated thus: + +$$N_{\text{active}} = c_i \frac{L_{\max}}{(\text{speech activity factor})} = 0.75 \frac{48}{0.28} = 129 \text{ active circuits}$$ + +In this example, 129 circuits would carry speech load and the remainder would be idled. + +NOTE – The following items are for future study: + +- a) Should CME loads include voiceband data as well as speech? The effect of voiceband data traffic on speech quality is an important issue in the evaluation of CME performance. Data percentage is defined as follows: + +$$P_{\text{data}} = \frac{\text{Number of input circuits active with data}}{\text{Total number of active circuits}} \times 100\%$$ + +- b) Some administrations report that speech activity on their real circuits averages about 36% when using a highly sensitive speech detector having a short hangover time of about 30 ms. Is it desirable to modify the speech load requirements given in 6.1, and, if so, what values are recommended? +- c) Fractional values of speech load are given in 6.2. Some CME may operate so as to display significant changes in performance at different fractional load points. Should the fractional load points be changed to accommodate this type of operation, and, if so, what changes are recommended? + +# 7 Processing of the speech + +The CME testing laboratory will take the source recordings, replay them through the circuit under test of the agreed CME (using the calibration tone to set the agreed input level) while operating the CME at the agreed load, and record the output from the circuit under test in a predetermined arrangement (explained in clause 8). The recorded outputs will then be used to perform the listening test. The CME being tested must be connected to the load simulator and to the recording and playback equipment as shown in Figure 2. It may be necessary to make provision for special A/D and D/A interfaces to permit the selected load simulator and recording equipment to be connected to the CME. + +The idle circuit noise (ICN) must be added to the speech prior to processing through the CME. This noise is electrically summed with the source speech immediately before the A/D converter at the CME input, as shown in Figure 2. It is essential that the r.m.s. input level of the speech to the CME is adjusted using the 1 kHz tone on the source recording and not by measuring the level of the speech plus noise at the input of the A/D. + +All the processed outputs will be on the left channel of the recording medium. The corresponding original signal will be simultaneously recorded on the right channel. The 1 kHz tone will be available both in its original form (right channel) and as processed by passing through the CME under test (left channel). + +The 1 kHz tone on the source recording (see 5.4) will be used to adjust the r.m.s. input speech level to be 20, 30 or 38 dB below the peak overload point of the CME coder. + +# 8 Test design + +Two separate tests are proposed to evaluate different aspects of CME performance. The first verifies the effect of various loads on the performance. The second verifies the effect of errors in the CME digital control channel. + +In both these tests, white noise equivalent to $-68 \text{ dBmp}$ at the input to the IRS receive end should be added to reduce noise contrast effects at the onset of speech utterances. The IRS must be used with sidetone present ( $12 \text{ dB STMR}$ ), as this sounds more natural to the subjects. + +![Block diagram of the CME testing setup. On the left, a 'Source speech recording' is played back through a 'Recording medium' to produce 'Source speech playback'. This is summed with noise from a 'Noise generator (ICN)' in a 'Summing amplifier'. The output is fed into the 'Circuit under test (CUT)'. The CUT is connected to a 'Speech load simulator' via an 'A/D' converter. The CUT is part of a 'CME under test' block, which contains 'CME terminal', 'CME transmission facility', and 'CME terminal' components. Multiple 'Bearers channels' (labeled \kappa M < N^a) are shown entering the transmission facility from 'Circuits No. 1' through 'No. N'. The output of the transmission facility is connected to another 'CME terminal', which is connected to 'Circuits No. 1' through 'No. N'. These circuits are connected to a 'D/A' converter, which then connects to a 'Processed speech recorder'. The output of the recorder is played back through a 'Recording medium' to a 'Listening test system playback' unit, which is connected to a pair of headphones.](7efae06af3af43ffe5d4b956a679cf54_img.jpg) + +Block diagram of the CME testing setup. On the left, a 'Source speech recording' is played back through a 'Recording medium' to produce 'Source speech playback'. This is summed with noise from a 'Noise generator (ICN)' in a 'Summing amplifier'. The output is fed into the 'Circuit under test (CUT)'. The CUT is connected to a 'Speech load simulator' via an 'A/D' converter. The CUT is part of a 'CME under test' block, which contains 'CME terminal', 'CME transmission facility', and 'CME terminal' components. Multiple 'Bearers channels' (labeled $\kappa M < N^a$ ) are shown entering the transmission facility from 'Circuits No. 1' through 'No. N'. The output of the transmission facility is connected to another 'CME terminal', which is connected to 'Circuits No. 1' through 'No. N'. These circuits are connected to a 'D/A' converter, which then connects to a 'Processed speech recorder'. The output of the recorder is played back through a 'Recording medium' to a 'Listening test system playback' unit, which is connected to a pair of headphones. + +a) For an explanation of the factor $\kappa M$ , see A.4. + +NOTE – The "block" called "CME under test" will be duplicated for tandemming. + +**Figure 2/P.840 – Testing CME** + +## 8.1 Test No. 1 – Effect of applied load + +This test may be conducted twice, once to obtain a quality rating and (optionally) a second time to obtain a listening effort rating. If performing both listening effort and listening quality tests, then either the subjects or speech material must be different. If they are the same, the listening effort test must be performed first as it is essential in listening effort tests to use material which has not been heard before. + +The parameters for testing are as follows: + +- a) *CME parameters* + - 1) CMEs under test: $N$ (at least 1); + - 2) CME loads: four values (0, 0.75, 1.0, 1.2) (see 6.2); + - 3) speech activity factor: one value (28%); + - 4) active circuit speech characteristic: one value (see 6.1); + - 5) circuit under test (CUT), idle circuit noise (ICN): two values (–70 and –45 dBm0p); + - 6) input speech level to CUT: three values (20, 30 and 38 dB below CME coder peak overload); + - 7) output listening levels2: at least three values (preferred and preferred $\pm 10$ dB); + - 8) talkers: four talkers, i.e., 2 male and 2 female; + - 9) comfort noise : "adaptive method", when available; in other cases, the fixed method will be used. +- b) *Reference parameters* + - 1) original source segments: one value; + - 2) MNRU: four values (5-35 dB in 10 dB steps); + - 3) Babble noise. SNR(p): three values (20, 30 and 40 dB); + - 4) reference connections (HRCs): approximately four different cases to be decided by test team; + - 5) listening levels: three levels (see above); + - 6) talkers: four talkers, i.e., 2 male and 2 female. + +For the stated set of parameters the number of test conditions is: + +$$4 \times 2 \times 3 \times 3 \times 4 \times N = 288 \times N \text{ CME conditions}$$ + +plus + +$$12 \times 3 \times 4 = 144 \text{ reference conditions.}$$ + +This totals (assuming $N = 1$ CME): + +$$432 \text{ test conditions} + 36 \text{ practice} = 468 \text{ conditions.}$$ + +The set of test conditions should be divided into about 13 sequences (12 test + 1 practice) of 36 conditions with the conditions within each sequence put into a random order. Table 1 lists the conditions in a non-randomized sequence. + +The sequence in Table 1 will be repeated for each of 4 talkers and 3 listening levels to create 12 test sequences: A to L inclusive. A practice sequence P will also be created. The test sequence A to L inclusive plus P can then be ordered for playback in the listening test according to the procedure described in clause 9. + +Assuming each condition takes 15 s to present and obtain a vote, total test time is about 2 hours. + +- c) *Multi-hop configuration* + +Test configurations implementing CME in tandem (2 and 3) could also be tested. + +--- + +2 This is a set up using the calibration tone on the source tape, not by measuring the level of the processed speech. + +**Table 1/P.840 – Basic sequence (assumes 1 CME for testing)** + +| Condition | Load | ICN (dBm0p) | Input (Note) (dB) | Q (dB) | SNR (dB) | HRC | +|-----------|------|-------------|-------------------|--------|----------|----------| +| 1 | 0.00 | -70 | 20 | | | | +| 2 | 0.75 | -70 | 20 | | | | +| 3 | 1.00 | -70 | 20 | | | | +| 4 | 1.20 | -70 | 20 | | | | +| 5 | 0.00 | -45 | 20 | | | | +| 6 | 0.75 | -45 | 20 | | | | +| 7 | 1.00 | -45 | 20 | | | | +| 8 | 1.20 | -45 | 20 | | | | +| 9 | 0.00 | -70 | 30 | | | | +| 10 | 0.75 | -70 | 30 | | | | +| 11 | 1.00 | -70 | 30 | | | | +| 12 | 1.20 | -70 | 30 | | | | +| 13 | 0.00 | -45 | 30 | | | | +| 14 | 0.75 | -45 | 30 | | | | +| 15 | 1.00 | -45 | 30 | | | | +| 16 | 1.20 | -45 | 30 | | | | +| 17 | 0.00 | -70 | 38 | | | | +| 18 | 0.75 | -70 | 38 | | | | +| 19 | 1.00 | -70 | 38 | | | | +| 20 | 1.20 | -70 | 38 | | | | +| 21 | 0.00 | -45 | 38 | | | | +| 22 | 0.75 | -45 | 38 | | | | +| 23 | 1.00 | -45 | 38 | | | | +| 24 | 1.20 | -45 | 38 | | | | +| 25 | | | 20 | | | Original | +| 26 | | | 20 | 5 | | | +| 27 | | | 20 | 15 | | | +| 28 | | | 20 | 25 | | | +| 29 | | | 20 | 35 | | | +| 30 | | | 20 | | 20 | | +| 31 | | | 20 | | 30 | | +| 32 | | | 20 | | 40 | | +| 33 | | | 20 | | | HRC1 | +| 34 | | | 20 | | | HRC2 | +| 35 | | | 20 | | | HRC3 | +| 36 | | | 20 | | | HRC4 | + +ICN Idle Circuit Noise + +NOTE – dB below CME coder peak overload level. + +## 8.2 Test No. 2 – Effect of digital errors in the DCME control channel + +The preceding test was done assuming that the digital transmission facility is operated error-free. Under real conditions errors will occur and errors in the DCME "control channel" may cause momentary disruption of the voice circuits. To determine the effect of digital errors on performance, Test No. 1 should be repeated while random errors at a rate of $10^{-3}$ are injected into the control channel. For this test, only one listening level (preferred) is necessary, so the total number of test conditions is $N \times 96$ plus 48 reference conditions. With $N = 1$ , the test time is 36 minutes. + +NOTE – If this test is to use the listening-effort scale, either different subjects or different speech material must be used for the reason given in 8.1. + +# 9 Listening test procedure + +See B.2, B.3 and B.4/P.800. + +# 10 Analysis of results + +As well as the opinion score, $Y_{iq}$ or $Y_{le}$ , collected from each subject for each segment replayed, the associated value $x$ (see 6.1) for that segment will be also known and should also be taken into account in the analysis. The usual analysis of variance becomes in fact an analysis of co-variance, with $x$ as the concomitant variable. This increases the precision of the results and conclusions, effectively eliminating the uncertainty due to the random instantaneous fluctuation of system load about its intended mean, by adjusting all mean scores to what they would have been had that mean been maintained exactly. + +The same technique should be used in the test with digital errors, by regarding the actual error rate for each segment as the concomitant variable. + +## Annex A + +## Description of circuit multiplication equipment + +A detailed description of the characteristics of CME that can be evaluated with this methodology is given in this annex. Some definitions are given below. For a more comprehensive list of definitions related to CME, see clause 2/G.763. + +### A.1 Definitions + +For the purpose of this annex the following definitions apply: + +**A.1.1 circuit multiplication equipment (CME):** Digital Circuit Multiplication Equipment (DCME) is defined in ITU-T Rec. G.763 as "A general class of equipment which permits concentration of a number of 64 kbit/s PCM encoded input trunk channels on a reduced number of transmission channels.". After the first DCME Recommendation, ITU-T Rec. G.763, was approved. Recommendations for the new types of equipment similar to DCME, but having packetized type interfaces at the bearer side, PCME (ITU-T Rec. G.765) and IP-CME (ITU-T Rec. G.769/Y.1242), were approved. Since there was no appropriate term to cover DCME, PCME and IP-CME, Circuit Multiplication Equipment (CME), a more generic type of equipment, including all these types of equipment such as DCME, PCME, IP-CME and Channel Transcoder, having PCM trunks, compressing speech signals on channel basis in various ways and sending them out through various types of bearers including STM type, Packetized type, ATM type and IP type, was newly introduced. A working definition may be: any digital transmission method that derives more voicegrade circuits than is possible using equipment conforming to ITU-T Rec. G.711. For the purpose of this Recommendation, the term circuit may at times refer to a circuit between two switching points (trunk), or between the customer's premises and a switching point (loop). At other times, it may refer to an end-to-end digital connection. The circuit may also be physical or virtual. The term voicegrade means that the bandwidth of the circuit is nominally 3.1 kHz. We will attempt to avoid confusion by using suitable qualifiers, when necessary, to describe the kind of circuit we mean. + +Based on the above discussion, we conclude that there are three basic types of CME. These are: + +- *Type A* – Channel Transcoder, only using low rate encoding (LRE, < 64 kbit/s see A.1.4) to obtain a circuit multiplier larger than 1. Some LRE methods (e.g., 32 kbit/s ADPCM) are + +amenable to the subjective testing methods described in ITU-T Rec. P.830: other methods (e.g., 48 kbit/s vocoding) may require new subjective test methods. + +- *Type B* – Uses only digital speech interpolation (DSI, see A.1.5) to obtain a circuit multiplier larger than 1. DSI is defined in A.2. By definition, the digital coding used in Type B CME to drive a circuit operates at 64 kbit/s and conforms to ITU-T Rec. G.711. Thus, the coding provides a circuit multiplier of unity. During periods of CME overload, any of several overload strategies may be used to resolve the contention for channels. The three basic overload strategies are defined in A.5. For example, during momentary periods of overload, the channel coding rate may be reduced to increase the channel capacity. However, this recoding action is attributed to the DSI and the circuit multiplier larger than 1 thus obtained is credited to the DSI. +- *Type C* – Combination of Types A and B. This hybrid type employs LRE to obtain a circuit multiplier larger than 1 and then DSI to obtain an additional circuit multiplier larger than 1. For example, if the LRE conforms to ITU-T Rec. G.726 32 kbit/s ADPCM, then the coder has a circuit multiplier of $\kappa = 2$ . The DSI may increase this multiplier by a further factor of 2 or 3 depending upon the CME. The total multiplier, 4 to 6 is equal to the product of the LRE and DSI multipliers. + +This annex is applicable to Types B and C CME only. + +**A.1.2 circuit multiplication system (CMS):** A telecommunication system comprised of two or more CME terminals where each terminal contains a transmit unit and a receive unit connected by a digital transmission system providing a pool of bearer channels. The CMS may support: + +- i) 64 kbit/s clear channels for ISDN services (can be used in the bearer pool); +- ii) voiceband data (dial-up) including Group III facsimile; +- iii) voice services in the frequency range 300-3400 Hz, carried at 56 or 64 kbit/s; +- iv) 64 kbit/s clear (pre-assigned not ISDN dial-up); +- v) sub-64 kbit/s digital data. + +**A.1.3 circuit versus packet mode:** Internally the CME may employ a circuit or a packet mode for the transmission of speech or data. In the circuit mode, bearer channels are derived by providing suitable time slots on the transmission facility interconnecting the CME terminal equipment. In the packet mode, virtual bearer channels are created and the speech or data samples are put into one or more packets of fixed or variable length. The packets are addressed to the destination circuit and transmitted in a virtual channel on the transmission facility one at a time. Thus, in the circuit mode, the transmission facility can be thought of as carrying a number of channels multiplexed together while, in the packet mode, the facility is thought of as a single high-speed channel logically divided into virtual channels which transmits packets one at a time. + +**A.1.4 low rate encoding (LRE):** A voiceband signal encoding method, e.g., ADPCM, which results in a bit rate less than 64 kbit/s, e.g., either 40 kbit/s, 32 kbit/s, 24 kbit/s or optionally 16 kbit/s. Conversion between speech signals encoded in PCM at 64 kbit/s and those encoded in ADPCM, must be carried out by means of transcoding processes given in ITU-T Rec. G.726. + +**A.1.5 digital speech interpolation (DSI):** This is a technique whereby advantage can be taken of the inactive periods during a conversation, creating extra channel capacity. Speech activity is typically 30-40%, on average, which can produce a DSI gain of up to 3:1, but generally in the range of 2:1 to 2.5:1. + +**A.1.6 bearer channel (BC):** A bearer channel is a unidirectional, digital, transmission path from the transmit unit of one DCME to the receive unit of a second associated DCME used to carry concentrated traffic between two DCMEs. + +NOTE 1 – A number of bearer channels in each direction of transmission form the both-way link required between two DCMEs. This link may be, for example, a 2048 kbit/s system. + +NOTE 2 – A bearer channel may have any of the following instantaneous bit rates: either 64, 32, 24 or optionally 16 kbit/s. + +**A.1.7 LRE gain, DSI gain, CME gain:** LRE gain is the factor by which the 64 kbit/s rate of the incoming circuits is reduced when LRE is used for coding within the CME. For example, when a transcoder conforming to the 32 kbit/s portion of ITU-T Rec. G.726 is used, the LRE gain will equal 2. The LRE gain is 1 when no transcoding is used. + +DSI gain is the ratio of the number of active speech input circuits to the number of bearer channels used to transport this speech, where the same encoding rate is used for circuits and bearer channels. The DSI gain is constrained by the number of input circuits and the speech activity factor and other input speech characteristics. The DSI gain is 1 when DSI is not used. In packetized type of CME such as PCME or IP-CME, the reciprocal of averaged speech activity ratio can be considered as DSI gain. + +The CME gain is the product of the LRE and DSI gain factors. + +**A.1.8 dCME overload:** The condition when the number of instantaneously active input circuits exceeds the maximum number of bearer channels available for DSI. + +**A.1.9 freezeout:** The condition when an input circuit becomes active with speech and cannot be immediately assigned to a bearer channel for DCME, virtual packet channels of PCME or IP-CME, due to lack of availability of such channels. + +**A.1.10 freezeout fraction:** The percentage of speech lost, obtained by averaging overall input circuits for a given time interval, e.g., one minute. (See 2.23/G.763 and 15.2.3.2/G.763 for a precise definition of freezeout fraction.) + +**A.1.11 transmission overload:** The condition when the average bits per sample goes beyond the value set in accordance with the speech quality requirements. + +#### **A.1.12 clipping** + +- i) Speech detector clipping, the time it takes to recognize that speech is present, can cut off ("clip") the start of the speech utterance. +- ii) Competitive clipping is caused by the overload control strategy which allows freezeout to occur when bearer channels are temporarily unavailable. +- iii) Control channel clipping resulting from simultaneous activity detection in more than one input circuit within the same DCME frame or servicing higher order priority messages such as 64 kbit/s unrestricted circuit establishment. + +**A.1.13 variable bit rate (VBR):** The capability of the encoding algorithm to dynamically switch between different bit rates, e.g., between 32 and 24 kbit/s or between 24 and 16 kbit/s for speech traffic under control of the G.763 DCME. + +This provides an overload control strategy to cope with traffic peaks and hence freezeout problems. + +Temporary, additional bearer channels (overload channels) with reduced signal-to-noise ratio are created. Graceful degradation is provided through the use of rapid rotation where the lower encoding rate is cyclically moved through all bearer channels carrying speech, and is only applied to any given bearer channel for a short interval of time (typically $\leq 10$ ms for DCME or 16 ms for PCME). The frequency of occurrence of the lower bit rate is a function of the ensemble CME input circuit activity. + +**A.1.14 queueing:** A channel assignment strategy employing buffer memory in the DCME transmitter to assign a newly activated trunk channel to a bearer channel to become available on a first-in first-served basis. + +**A.1.15 dynamic load control (DLC):** An overload control strategy in which the DCMS signals to the associated switch that the traffic load the switch is generating, or is predicted to generate, cannot + +be transmitted satisfactorily by the DCMS and that the switch should reduce its demand on the DCMS. + +**A.1.16 load carrying capacity:** The load carrying capacity is defined as the maximum offered speech load plus "overhead" load (see A.1.17) that the CME channels can carry without entering the CME overload mode (see A.1.8). + +**A.1.17 applied and offered load:** The applied load consists of the speech bursts entering the CME on the active circuits. Thus the applied load is a function of the number of active circuits and the speech activity on the circuits. + +The offered load consists of the applied load plus any additional load (overhead) generated by the dCME messages and control information or PCME/IP-CME packet overheads. The offered load is the load presented to the CME bearer channels. If the offered load is less than the load-carrying capacity of the channels, then all the offered load is carried by the CME. The CME may employ variable bit rate coding so that the CME can momentarily increase the load-carrying capacity of the bearer channels (creation of overload channels) in order to accommodate the extra load. Dynamic load control may also be used to limit the applied load. However, if the offered load exceeds the capacity of the bearer channels then, depending upon the overload strategy of the CME, some of the offered load may be lost through competitive clipping (sample dropping). + +The instantaneous load is a function of the statistics of the input, speech and the CME overhead traffic, and is difficult to characterize mathematically. However, the long-term time average applied load can be calculated as follows: + +$$L_a = N \frac{\alpha}{\alpha + \beta}$$ + +where $L_a$ is the average applied load, $\alpha$ is the average speech burst length, $\beta$ is the average silence length, and $N$ is the number of circuits in use. The term $\alpha/(\alpha + \beta)$ is equal to the average speech activity. The applied load is measured at the circuit input to the CME. Thus, the average load on the CME can be externally controlled by varying the number of circuits in use, $N$ , or the speech activity factor, $\alpha/(\alpha + \beta)$ . + +Similarly, average offered load is a useful concept, and it can be calculated from this formula: + +$$L_o = N \frac{\alpha(k+1)}{\alpha + \beta} + G$$ + +where $L_o$ is the average load offered to the bearer channels, the term $k$ is a constant which is for the "stretching" effect that the CME speech detector has on the activity factor, and the term $G$ is a load factor that accounts for the system overhead traffic (e.g., control messages). Thus, the average offered load, $L_o$ , will always be larger than the average applied load, $L_a$ . + +### A.2 Digital Speech Interpolation (DSI) + +Digital speech interpolation is defined in A.1.5. A working definition of DSI may be: any method for assigning a voicegrade bearer channel on demand for the transmission of speech at the onset of the speech burst (talk-spurt). The bearer channel comes from a pool maintained by the DCME or from the packet link for PCME or IP-CME, and the speech comes from an active circuit connected to the CME. When the speech burst stops, the channel is either: + +- i) relinquished and put back into the pool; or +- ii) kept assigned to the circuit as long as the pool is not empty and the channel is not needed to service another circuit; + +and for DCME, + +- i) relinquished and stop sending packets; or + +- ii) keep sending packets as long as the packet link is not fully loaded and the packet link capacity is not needed to serve for another circuit. + +In the above context, the term "bearer channel" refers to the transmission path between the CME terminals which are used to carry the traffic on the circuits connected to the CME. + +### A.3 Speech detection + +To perform DSI the CME must contain a speech detector. The speech detector monitors the circuits and determines when speech is present and when it is not. When speech is declared present, the CME attempts to assign an available bearer capacity to the circuit. If no capacity is available, the CME then invokes its overload strategy. When the speech burst ends the speech detector may provide some "hangover" to avoid tail-end clipping of the burst. Hangover extends the effective length of the burst. + +"Fill-in" is another speech detector function sometimes employed to bridge, or eliminate, the silence gaps of less than a certain length between speech bursts. Fill-in does not extend the length of individual bursts the way hangover does, but requires a processing delay equal to the maximum filled-in gap. Both hangover and fill-in increase the activity factor of the speech on the bearer channels. + +To avoid front-end clipping of the speech burst, the speech detector usually employs delay of a few milliseconds to give it time to decide whether speech is present. + +Clipping or mutilation of the speech burst (both front-end and possibly tail-end) may occur because the speech detector makes false or late decisions. The operation of the speech detector, and thus the clipping performance of the CME, is a function of many factors characterizing the signal on the circuits, such as the signal level, signal-to-noise ratio, and echo path loss. + +### A.4 CME load + +The frequency of CME overloading is a function of the load on the system. The system load consists of speech bursts generated on the incoming circuits plus CME-generated overhead traffic. Since the speech burst activity on the circuits varies from moment to moment, the load also has short-term variations. + +In defining load we must distinguish between the applied load and the offered load, both defined in A.1.17. + +While the load varies continuously, subject to the statistics of the speech and the circuit activity, if we assume that the number of circuits in use, $N$ , is a constant over some period of time in which we are observing the operation of the CME, then the average applied and offered loads become useful concepts. Formulas for the average loads are defined in A.1.17. While these formulas are somewhat simplistic and do not capture the information concerning the variance of the load about the average, they do allow useful insight into the operation of the CME. + +The load carrying capacity of the CME channels is also an important consideration. The load carrying capacity is defined in A.1.16. If the offered load is less than the load carrying capacity of the channels, then all the offered load is carried by the CME. However, if the offered load exceeds the capacity of the channels then, depending upon the overload strategy of the CME (see A.5), variable bit rate coding will be used to momentarily increase the load carrying of the channels so that they can accommodate the extra load, or some of the offered load will be lost through sample dropping. Thus, overloading is defined to occur when the load exceeds the normal carrying capacity of the CME channels. + +In a sample dropping system, the load capacity is fixed and is simply $\kappa M$ , where $M$ is the number of 64 kbit/s equivalent channels provided, and $\kappa$ is the LRE factor which accounts for the difference in bit rates between the circuits (always 64 kbit/s) and the bearer channels. If 32 kbit/s LRE is used on + +the bearer channels, for example, then $\kappa = 2$ . If LRE is not used then $\kappa = 1$ . If variable bit rate (VBR) coding is used then the load capacity of the CME is not fixed, and sample dropping may be avoided by temporarily creating extra bearer channels. If the coding rate drops from 32 to 16 kbit/s, for example, then during the period VBR is active $\kappa = 4$ . + +In these examples, the number of bearer channels available to carry speech is assumed to be constant. However, in CME that carries voiceband data and other tones on the circuits, DSI cannot be used on these signals. The result is that these continuous signals capture channels for full-time use, reducing the pool of channels available for carrying speech. + +By using the average load equations and the concept of load capacity, we can illustrate in Figure A.1 the load curves for a sample dropping type-C CME. The slope of the offered load curves depends upon the speech activity factor, $\alpha/(\alpha + \beta)$ , and the speech detector "stretch" factor, $k$ . Load curves for three different activity factors are shown. If the number of circuits in use, $N$ , is less than $N_{\min} = \kappa M - G = 43$ then the DSI will never activate, even if the momentary speech activity factor goes to unity on all active circuits. Since the CME-carried load cannot exceed $\kappa M = 48$ , as the average offered load, $L_o$ , gets closer and closer to the maximum capacity, the frequency of sample dropping will increase as the moment-to-moment fluctuations in the speech activities push the offered load above the limit. + +Figure A.2 illustrates the load curves for a variable bit rate type-C system which recodes at 16 kbit/s during overload. In this example, when the offered load exceeds $\kappa M = 48$ , the coding rate is dropped from 32 to 16 kbit/s on the bearer channels. The capacity is thus increased to $\kappa M = 96$ . The extra capacity absorbs the momentary overload and prevents sample dropping (freezeout) from occurring. If the offered load exceeds 96, then sample dropping will have to occur because further VBR (e.g., down to 8 kbit/s) is not provided for in this example. + +Thus, in summary, as long as $N \leq N_{\min}$ the CME will not need to use the DSI function, because all circuits will have access to a bearer channel. Sample dropping will not occur until the offered load exceeds the load carrying capacity. In overload, the CME will decrease the coding rate, in which case $\kappa$ will increase, thus momentarily increasing the capacity of the CME, or the CME will start dropping samples or may queue the samples, in which case $\kappa$ will not change. + +### A.5 Overload strategies + +When a number of active circuits connected to the CME exceeds the number of available normal channels, the CME will experience momentary overloads; an increase in speech bursts will sometimes require more channels than are available. When this happens the CME must invoke its "overload strategy". The strategy is designed to deal with the issue of how best to share the channel pool. A number of basic strategies are possible: + +#### Type 1 – Competitive clipping or speech sample dropping + +In this strategy, defined in A.1.12, samples are dropped from the front end of the speech burst that unsuccessfully bids for a channel. Sample dropping continues until a channel is available or the burst ends normally. Perceptually, the effects of front-end sample dropping and front-end clipping, the latter caused by the speech detector, should be the same, even though they have different causes. Theoretically, however, they are not entirely the same because front-end clipping, due to the speech detector, is of short duration and is more likely to affect low-level parts of the signal, but front-end clipping, due to freezeout, is of longer duration. Freezeout, for cases other than front-end sample dropping, affects all levels with equal probability. + +![Figure A.1/P.840: Example of load curves for type-C, sample dropping, dCME. The graph plots Offered load (L_o) in Erlangs (y-axis, 0 to 60) against the Number of active 64 kbit/s circuits (N) (x-axis, 0 to 140). Three curves are shown for alpha/(alpha+beta) ratios of 0.8, 0.6, and 0.4. The maximum load capacity is kappa*M = 48 Erlangs. The minimum number of circuits required for DSI is N_min = kappa*M - G = 43. The region between N=0 and N=43 is labeled 'No DSI', and the region beyond N=43 is labeled 'DSI'. Assumptions: Sample dropping system, L_o = N * (alpha*(k+1)) / (alpha+beta) + G, M = 24 channels (64 kbit/s equivalents), k = 0.1 (speech detector 'stretch' factor), kappa = 2 (LRE circuit multiplier), G = 5 Erlangs overhead.](9ce50bc10864dc86e1cdee4be08f1897_img.jpg) + +Figure A.1/P.840: Example of load curves for type-C, sample dropping, dCME. The graph plots Offered load (L\_o) in Erlangs (y-axis, 0 to 60) against the Number of active 64 kbit/s circuits (N) (x-axis, 0 to 140). Three curves are shown for alpha/(alpha+beta) ratios of 0.8, 0.6, and 0.4. The maximum load capacity is kappa\*M = 48 Erlangs. The minimum number of circuits required for DSI is N\_min = kappa\*M - G = 43. The region between N=0 and N=43 is labeled 'No DSI', and the region beyond N=43 is labeled 'DSI'. Assumptions: Sample dropping system, L\_o = N \* (alpha\*(k+1)) / (alpha+beta) + G, M = 24 channels (64 kbit/s equivalents), k = 0.1 (speech detector 'stretch' factor), kappa = 2 (LRE circuit multiplier), G = 5 Erlangs overhead. + +P.840\_FA.1 + +**Figure A.1/P.840 – Example of load curves for type-C, sample dropping, dCME** + +![Figure A.2/P.840: Example of load curves for type-C, variable bit rate coding, dCME. The graph plots Offered load (L_o) in Erlangs (y-axis, 0 to 110) against the Number of active 64 kbit/s circuits (N) (x-axis, 0 to 160). Three curves are shown for alpha/(alpha+beta) ratios of 0.8, 0.6, and 0.4. The maximum load capacity with recoding is kappa*M = 96 Erlangs, and the maximum load capacity without recoding is kappa*M = 48 Erlangs. The region between N=0 and N=43 is labeled 'No DSI', and the region beyond N=43 is labeled 'DSI'. The region between the two maximum capacity curves is labeled 'Overloading', and the region below the lower curve is labeled 'No overloading'. Assumptions: Variable bit rate coding down to 16 kbit/s, M = 24 channels (64 kbit/s equivalents), k = 0.1 (speech detector 'stretch' factor), kappa = 2 (LRE circuit multiplier), G = 5 Erlangs (overhead).](45329c7d9aa2bd1290af5b2027f08d7e_img.jpg) + +Figure A.2/P.840: Example of load curves for type-C, variable bit rate coding, dCME. The graph plots Offered load (L\_o) in Erlangs (y-axis, 0 to 110) against the Number of active 64 kbit/s circuits (N) (x-axis, 0 to 160). Three curves are shown for alpha/(alpha+beta) ratios of 0.8, 0.6, and 0.4. The maximum load capacity with recoding is kappa\*M = 96 Erlangs, and the maximum load capacity without recoding is kappa\*M = 48 Erlangs. The region between N=0 and N=43 is labeled 'No DSI', and the region beyond N=43 is labeled 'DSI'. The region between the two maximum capacity curves is labeled 'Overloading', and the region below the lower curve is labeled 'No overloading'. Assumptions: Variable bit rate coding down to 16 kbit/s, M = 24 channels (64 kbit/s equivalents), k = 0.1 (speech detector 'stretch' factor), kappa = 2 (LRE circuit multiplier), G = 5 Erlangs (overhead). + +P.840\_FA.2 + +**Figure A.2/P.840 – Example of load curves for type-C, variable bit rate coding, dCME** + +#### Type 2 – Variable bit rate coding + +This strategy, defined in A.1.13, employs multiple bit rate speech coding algorithms (ITU-T Recs G.726, G.727, G.728 or G.723.1) to effectively multiply the number of bearer channels momentarily available to the circuits to carry the offered load. Since a lowering of the bit rate will have the effect of increasing the quantization noise produced by the coders, the perceptual effect of variable rate coding will be momentarily increased in quantizing noise, i.e., reductions in *Q* (for a discussion of *Q*, see clause 6/P.810). The subjective effect is minimized by rapidly moving the use of the lower bit rate coding among all active speech bearer channels. + +#### Type 3 – Queueing + +This strategy, defined in A.1.14, employs buffers (memories) for the speech burst samples to occupy while waiting for a channel. The perceptual effect of pure queueing, without buffer overflow, is a time shift of the speech bursts. No samples are lost, and there is no increase in noise. + +The impairment introduced can be called "silence duration modulation". From the listener's point of view, a given speech burst, when queued, will begin somewhat later in time relative to its predecessor burst than it would have without queueing. Also the successful burst may be perceived as beginning somewhat sooner. Since the buffers must, of necessity, be finite, this strategy cannot be employed alone, but must be coupled with either sample dropping or variable rate coding. Thus, a queueing system can have speech mutilation or recoding noise as well as time shifting. + +#### **Type 4 – Dynamic load control** + +An overload control strategy, defined in A.1.15, in which the CME signals to the associated switch that the traffic load which the switch is generating, or is predicted to generate, cannot be transmitted satisfactorily by the CME, and the switch should reduce its demand on the CME. + +### **A.6 Silence reconstruction methods** + +Since a CME with loading sufficient to require interpolation among bearer channels does not transmit silences between speech bursts at the receiving end, the silences must be artificially recreated. Several different methods for doing this are possible. The simplest is to insert a white noise at a fixed level in the receiver during silences. Careful selection of the level is necessary to avoid noise contrast, that is, an apparent and annoying contrast between the noise in the silences and the background noise during speech bursts. Instead of white noise, some types of noises, such as 1/f characteristic noise or Hoth noise, may be used to give a more moderated impression to the listeners. Adaptive methods attempt to adapt the noise level automatically to the circuit conditions, requiring careful filtering and estimation of source noise power. + +These adaptive methods, when available, shall be used when testing CME, instead of methods introducing a noise at a fixed level. + +### **A.7 Circuit versus packet mode** + +Internally the CME may employ a circuit or a packet mode for the transmission of speech bursts. In the circuit mode, bearer channels are derived by providing suitable time slots on the transmission facility interconnecting the CME terminal equipment. In the packet mode, the speech burst samples are put into one or more packets of fixed or variable length. Packet multiplexing, inserting multiple sub-packets corresponding to different trunk channels into a single standard packet, may also be employed. The packets are addressed to the destination circuit and transmitted over the transmission facility one at a time. Thus, in the circuit mode, the transmission facility can be thought of as carrying a number of channels multiplexed together, while in the packet mode, the facility is thought of as a single high-speed channel which transmits packets one at a time. + +In the packet mode, performance of the system depends on how the packets are serviced. Two servicing methods are: + +- a) All packets from all circuits enter a first-in first-out (FIFO) queue and are serviced by the high-speed channel one at a time. Each packet is treated independently. Each packet experiences a variable delay in arriving at the receiving end that is a function of the fill of the FIFO queue. If packets arrive too late, after a given reconstruction delay, they will be lost or discarded by the receiver. This is called packet dropping and it is a function of the system load. Packet dropping can cause speech mutilation at any point in the burst. It gives rise to "mid-burst" sample dropping. Packets can also be dropped in the FIFO queue if it experiences overflow. The fill of the queue is monitored and the overload strategy is invoked, when necessary, to prevent excessive packet dropping. +- b) Once a circuit has seized the high-speed channel for transmission of a packet, all the packets on the circuit for that burst are transmitted before the high speed line is free to transmit another circuit's packets. Thus the circuit is "cut-through" during the burst. Cut-through operation avoids mid-burst speech sample lost. However, since only one circuit at a + +time can use the high-speed channel, other circuits with packets to transmit must await their turn. The packets must be queued while they await the channel. Load-dependent queueing delays must be equalized at the receiving end. This is usually done by employing some form of time stamp on the packet. The possibility always exists that packet queues will overflow before the packets can be transmitted. When this happens, the overload strategy is invoked to prevent excessive packet dropping. + +Packet mode introduces more delay than a non-packet mode CME. The extra delay has three components. The first is the packetization time. Packetization time is a function of packet length and circuit coding rate. The second is the reconstruction delay. Reconstruction delay is chosen to minimize the probability of packet loss. The third is packet queueing delay. All three components may be load-dependent and thus variable. + +In summary, use of packet mode rather than circuit mode may introduce these additional performance-affecting aspects: + +- i) mid-burst sample dropping; +- ii) additional delay equal to the sum of the packetization and reconstruction delays; +- iii) packet queueing delay. + +### **A.8 Packet reconstruction** + +In a packet mode, system loss of a packet presents the receiver with a dilemma, namely, what to use in place of the speech samples carried in the lost packet. Several methods are employed and they have different performance consequences. One method is to insert samples in place of the lost speech samples. Another method repeats samples in a previous packet to replace the lost samples. Other methods are also employed. + +## **Annex B** + +## **Speech material used to construct speech sequences** + +### **ORWELL** + +George Orwell began his classic novel 1984 with, "It was a bright cold day in April," but he gave no further hint as to what the weather might be during the fateful year. From the succession of untoward weather events that marked 1983, many have come to believe that the world's weather has undergone an unprecedented change for the worse and that we might be heading for a series of natural disasters this year to match the demise of free democratic thought and speech described in Orwell's book. + +Since we do not have the ability to predict what individual weather events might occur during 1984, let us turn the calendar back a hundred years and see what happened throughout the country in 1884. The year opened with the arrival of arctic air from northern Canada which drove the thermometer down to $-40^{\circ}\text{F}$ at Rockford, Illinois, and to $-25^{\circ}\text{F}$ at Indianapolis, Indiana, both records that still stand. Sub-zero temperatures penetrated into the South, and a hard freeze hit citrus groves in Florida. + +In early February, heavy rains falling on a deep snow cover caused the Ohio River to flood. Crests were of record heights from Cincinnati to the river's mouth at Cairo, Illinois. + +Late February brought an outbreak of tornados in the South and the Ohio Valley, where some sixty individual funnels descended. More than 420 were killed, and more than 1000 injured. Nothing approached this visitation in severity or extent until the tornado outbreak in Durango, Colorado, for seventy-six days ending 16 April. + +In May, out-of-season rainstorms in the deserts of the Southwest caused widespread floods. Rail traffic from Salt Lake City to the south was interrupted for three weeks, and the Rio Grande River flooding at El Paso, Texas, caused \$ 1 million in damage. + +Heavy frosts occurred in late May, when the thermometer dropped to 22 °F in Massachusetts, and snow fell in Vermont on Memorial Day. + +California got more heavy rain in June; Los Angeles had 1.39 inches and San Francisco 2.57 inches, both all-time June records. And as a result of rain in Wisconsin the flooding Chippewa River did more than \$ 1.5 million in damages and left 2000 homeless at Eau Claire. + +The great Oregon snow blockade followed 34 inches of snowfall at Portland in the middle of December. Rail communication was cut off from the east and south for many days, and mail from California had to come by ocean steamer. + +If you think the weather that made so many headlines in 1983 was unprecedented, hark back to 1884. We do not know whether El Niño was active then or whether some other atmospheric or oceanic force was the culprit. All we can do now is wait and see what 1984 brings. + +#### FOG + +One of winter's most spectacular sights is a smokelike fog that rises from openings in the arctic ice fields and occasionally appears above the open waters of unfrozen lakes and harbours in our temperate zone. Various names for the phenomenon are "frost smoke", "sea smoke", "steam fog", "warm water fog", and "water smoke". The fog is caused by the passage of a stream of arctic or polar air with a temperature near zero Fahrenheit over unfrozen water. Within the lower forty-eight states, it occurs principally over unfrozen areas of the Great Lakes and over harbour waters of the north Atlantic coast. + +"Sea smoke" occurs because the vapour pressure at the surface of the water is greater than that in the air above. Water vapour evaporates into the air faster than the air can accommodate it. The excess moisture condenses and forms a layer of fog, like steam or smoke, rising off the water. Usually, a clear space exists between the water's surface and the bottom of the fog, and its upper limit is generally 10 to 25 feet. If an atmospheric inversion develops near the water's surface, the fog may be confined there and becomes thick, resulting in a hazard to navigation. + +If the air temperature is severely cold, -20 °F or below, the rising moisture may form ice crystals in the layer of air just above the water. This is called "frost smoke" and it makes a beautiful sight, especially when sunlight glitters on the thin ice needles. + +"Steam fog" can occur over lakes and streams in the autumn following a clear, still night during which the air has cooled. The differences in vapour pressures cause the warm water to steam into the cold air, and whole valleys and basins can be covered with a thin layer of fog while the hillside remains clear. + +## Annex C + +## Instructions on the use of a limited number of sentences + +If $N$ sentences per talker are used, there will be $N(N - 1)$ possible sentence combinations per talker. The first 16 results are tabulated below: + +| | | | | | | | | | | | | | | | | | +|------------|---|---|----|----|----|----|----|----|----|-----|-----|-----|-----|-----|-----|-----| +| $N$ | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | +| $N(N - 1)$ | 2 | 6 | 12 | 20 | 30 | 42 | 56 | 72 | 90 | 110 | 132 | 156 | 182 | 210 | 240 | 272 | + +Either of two reasons for wanting to limit the number of sentences can be put forth: + +- the wish to save time by not having to author lists of more than $2 \times 85$ sentence combinations per talker. Separate recording of all the combinations is, of course, still needed unless sophisticated editing equipment for digital types is at hand; or +- the need to organize the test in a way that fulfils the requirements for an analysis of variance. + +Depending on which of the motives above is invoked, different methods can be adopted. These are: + +- 1) All possible $N(N - 1)$ sentence combinations per talker are recorded. + - a) The same $N$ sentence is used for all 4 talkers. The same sentence pair should then not be used for the same test conditions from talkers to talker, in order to avoid possible systematic interaction between test conditions and phonetic content; or + - b) four different sets of $N$ sentence ( $N1$ , $N2$ , $N3$ and $N4$ ) are authored. Then no precautions corresponding to a are needed. However, interaction will still be possible and uncontrolled. +- 2) To allow for an analysis of variance, subjects must judge the same speech material for all test conditions and all talkers. The number of sentences will then be limited to $M \times 2$ where $M$ is the number of pairs that will be used in the test. If $M = 1$ the test may appear too tedious for the subjects and the phonetic coverage may be insufficient. If an analysis of variance is to be justified, and the test is still to be practically possible, an expansion of the number of presentations is, therefore, recommended. $M = 2$ or 3 should be enough. This will lengthen the test time for each subject, but experience shows that tasks of 2.5 hours per subject are quite possible. Adjustments for such an expansion must then be made when deciding the presentation order. + +# Appendix I + +## Guidance for conversation tests + +To be complete, a CME performance evaluation has to include objective tests and subjective tests. Subjective tests represent a very important part of the evaluation and they usually include conversational tests and listening tests (with naïve subjects, and expert listeners). Results of such tests are interesting as they represent the user's point of view. + +This appendix gives, as an example conversation test to be applied to CME using CS-ACELP coding, compliant to ITU-T Rec. G.729. + +Ten test conditions have been set up, combining the following parameters: + +- number of hops (1, 2, 3); + +- load level (10, 76%); +- noise type, comfort noise (fixed or adaptive). + +The comfort noise injected by the CME can be "adaptive", i.e., its level depending on the noise level measured at the first CME, or "fixed", i.e., independent of the noise level measured at the first CME. + +Guidance for conversation test plan may be found in ITU-T Rec. P.800. + +After each conversation, the testers are asked to answer a set of 3 questions on the perceived quality of their last communication. These questions deal with the naturalness of voice, the perception of impairment and the overall quality of the communication. + +Furthermore, after each conversation, the testers could eventually fill in an open questionnaire about the level of annoyance and the frequency of potential signal impairments like noises or clipping. + +The following graphs illustrate the results obtained for several test conditions: + +Figure I.1 illustrates MOS results from conversational tests for 1, 2 and 3 hops; results show the clear degradation implied by the number of hops on the three parameters (voice quality, impairment and overall quality). MOS difference obtained with 1 and 3 hops reaches 1 point. + +![Bar chart showing MOS scores for Voice, Impairments, and Quality across 1, 2, and 3 hops.](51536f203c07ce20ce119e7e09b11cb6_img.jpg) + +A bar chart showing MOS scores for three parameters: Voice, Impairments, and Quality. The Y-axis represents the MOS score, ranging from 1.0 to 5.0 in increments of 0.5. The X-axis lists the three parameters. For each parameter, there are three bars representing different hop counts: 1 hop (yellow), 2 hops (pink), and 3 hops (red). The scores generally decrease as the number of hops increases. For Voice, the scores are approximately 4.4, 3.7, and 3.5. For Impairments, the scores are approximately 4.8, 4.6, and 4.3. For Quality, the scores are approximately 4.1, 3.7, and 3.0. + +| Parameter | 1 hop | 2 hops | 3 hops | +|-------------|-------|--------|--------| +| Voice | 4.4 | 3.7 | 3.5 | +| Impairments | 4.8 | 4.6 | 4.3 | +| Quality | 4.1 | 3.7 | 3.0 | + +Bar chart showing MOS scores for Voice, Impairments, and Quality across 1, 2, and 3 hops. + +P.840\_F1.1 + +**Figure I.1/P.840 – MOS results from conversational tests for 1, 2 and 3 hops** + +Figure I.2 compares MOS scores obtained from conversational tests depending on load level. It exhibits tests results from conversational tests for 2 load levels. Voice quality and overall quality are both degraded when the higher load level is applied on the system. + +![Bar chart showing MOS scores for Voice, Impairments, and Quality across Load 1 and Load 2. The Y-axis represents MOS score from 1.0 to 5.0. Load 1 is represented by orange bars and Load 2 by purple bars. For Voice, Load 1 is approximately 4.4 and Load 2 is approximately 4.1. For Impairments, both Load 1 and Load 2 are approximately 4.9. For Quality, Load 1 is approximately 4.1 and Load 2 is approximately 3.7.](a2dcc4a0703102026ec86e82caa4985e_img.jpg) + +| Category | Load 1 | Load 2 | +|-------------|--------|--------| +| Voice | 4.4 | 4.1 | +| Impairments | 4.9 | 4.9 | +| Quality | 4.1 | 3.7 | + +Bar chart showing MOS scores for Voice, Impairments, and Quality across Load 1 and Load 2. The Y-axis represents MOS score from 1.0 to 5.0. Load 1 is represented by orange bars and Load 2 by purple bars. For Voice, Load 1 is approximately 4.4 and Load 2 is approximately 4.1. For Impairments, both Load 1 and Load 2 are approximately 4.9. For Quality, Load 1 is approximately 4.1 and Load 2 is approximately 3.7. + +P.840\_F1.2 + +**Figure I.2/P.840 – MOS results from conversational tests for loads 1 and 2** + +In the open questionnaire of conversational tests and the expert listening tests, subjects noted the characteristics of the eventual impairment (noise, cut, etc.), which occurred during the conversation. + +The voice degradations, which were noted, are of several types: + +- cut during the communication; +- modification of voice timbre; +- variation of voice level; +- variation of comfort noise level. + +From this test, it appears that voice quality depends particularly on three parameters: + +- the total number of hops; +- the bearer load; +- the type of generated comfort noise. + +The parameter, which has the most determining influence on the speech quality degradation, is the number of hops, i.e., the number of successive coding/decoding. From 2 hops, the voice timbre seems different, less clear and there are, in general, some variations of voice level. The quality decreases very quickly with the number of hops. + +The other influential parameter is the bearer overload which sometimes creates micro-cuts. In this situation, the voice is often found distorted and metallic. + +Concerning the comfort noise, it was observed, that the "adaptive" noise level is always preferred to the "fixed" noise: the overall quality of the communication is found to be improved when the inserted noise level is adaptive. When set to "fixed", comfort noise is found to be disturbing (i.e., too high or too low level). + + + +# SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|--------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series B | Means of expression: definitions, symbols, classification | +| Series C | General telecommunication statistics | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Construction, installation and protection of cables and other elements of outside plant | +| Series M | TMN and network maintenance: international transmission systems, telephone circuits, telegraphy, facsimile and leased circuits | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks and open system communications | +| Series Y | Global information infrastructure, Internet protocol aspects and Next Generation Networks | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.880-200405-I_PDF-E/raw.md b/marked/P/T-REC-P.880-200405-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..b589dd6e4afb1a30a13acaf4786ea904e5be744c --- /dev/null +++ b/marked/P/T-REC-P.880-200405-I_PDF-E/raw.md @@ -0,0 +1,269 @@ + + +![ITU logo](2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg) + +The logo of the International Telecommunication Union (ITU) features the letters 'ITU' in a bold, sans-serif font, superimposed on a stylized globe with intersecting lines. + +ITU logo + +INTERNATIONAL TELECOMMUNICATION UNION + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.880** + +(05/2004) + +SERIES P: TELEPHONE TRANSMISSION QUALITY, +TELEPHONE INSTALLATIONS, LOCAL LINE +NETWORKS + +Methods for objective and subjective assessment of +quality + +--- + +**Continuous evaluation of time varying speech +quality** + +ITU-T Recommendation P.880 + +--- + +# ITU-T P-SERIES RECOMMENDATIONS + +## TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS + +| | | | +|-----------------------------------------------------------------------------------------------|---------------|-----------------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Subscribers' lines and sets | Series | P.30
P.300 | +| Transmission standards | Series | P.40 | +| Objective measuring apparatus | Series | P.50
P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of quality | Series | P.80
P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +## **ITU-T Recommendation P.880** + +### **Continuous evaluation of time varying speech quality** + +## **Summary** + +This Recommendation describes a methodology called Continuous Evaluation of Time Varying Speech Quality (CETVSQ) that can be used for evaluating the impact of the time fluctuations of speech quality on the instantaneous perceived quality (that is perceived at any instant of a speech sequence) and on the overall perceived quality (at the end of the speech sequence). The method uses a two-part task: first, an instantaneous judgment on a continuous scale with a slider during the speech sequence, and second, an overall judgment on a standard five-category scale at the end of the speech sequence. + +### **Source** + +ITU-T Recommendation P.880 was approved on 14 May 2004 by ITU-T Study Group 12 (2001-2004) under the ITU-T Recommendation A.8 procedure. + +# FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications. The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure e.g. interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementors are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database. + +© ITU 2004 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +# CONTENTS + +| | Page | +|-----------------------------------|------| +| 1 Scope ..... | 1 | +| 2 References..... | 1 | +| 3 Abbreviations..... | 1 | +| 4 Description of the method ..... | 1 | +| 4.1 Origin and motivation..... | 1 | +| 4.2 Test preparation ..... | 2 | +| 4.3 Listening session ..... | 2 | +| 4.4 Statistical analysis ..... | 4 | +| BIBLIOGRAPHY ..... | 6 | + + + +# **Continuous evaluation of time varying speech quality** + +# **1 Scope** + +This Recommendation defines a method of subjective assessment of transmitted speech quality for long speech sequences containing quality/time fluctuations. This method is based both on a continuous rating during the listening of a speech sequence and on an overall rating at the end of the speech sequence. Therefore, in addition to the measure of the overall perceived quality (as with the generally recommended methods), it provides a measure of the instantaneous perceived quality, i.e., the quality perceived at any instant of a heard transmitted speech sequence. In its current version, this method is not applicable to the selection of speech codecs. However, it is a useful tool to diagnose the effects of impairments on instantaneous and overall perceived quality, especially for discontinuous impairments with temporal fluctuations (e.g., those due to IP packet losses, handover in mobile networks, etc.). It can also aid the development and the validation of objective measurement tools that aim to predict the speech quality by detecting and analysing various types of impairments present in a speech signal; in fact, the Continuous Evaluation of Time Varying Speech Quality (CETVSQ) method can provide instantaneous judgments, as well as an overall evaluation of the subjective quality. + +# **2 References** + +- [1] ITU-T Recommendation P.800 (1996), *Methods for subjective determination of transmission quality*. +- [2] ITU-R Recommendation BT-500-11 (2002), *Methodology for the subjective assessment of the quality of television pictures*. + +# **3 Abbreviations** + +This Recommendation uses the following abbreviations: + +| | | +|--------|------------------------------------------------------| +| ACR | Absolute Category Rating | +| ANOVA | ANalysis Of VAriance | +| CETVSQ | Continuous Evaluation of Time Varying Speech Quality | +| MOS | Mean Opinion Score | +| QoS | Quality of Service | +| SSCQE | Single Stimulus Continuous Quality Evaluation | + +# **4 Description of the method** + +## **4.1 Origin and motivation** + +The development of the Continuous Evaluation of Time Varying Speech Quality (CETVSQ) method is motivated by the fact that the Quality of Service (QoS) of new networks varies, even during a single conversation, due to specific impairments (such as, packet losses for IP, handovers in mobile networks, etc.). These impairments are characterized by transient quality artefacts (as opposed to "continuous" impairments such as signal-to-noise ratios) with a more or less high density. Short samples cannot account for this density. Moreover, because of the technical characteristics of mobile or IP networks, speech quality can vary strongly during the same communication. In order to assess speech quality, typical methods described in ITU-T + +Rec. P.800 [1] use short stimuli (8 s) in subjective listening tests. These methods are well suited for time-constant speech quality. However, unless one evaluates a very large number of samples, they cannot take into consideration realistic occurrences and distribution of these impairments. In addition, they cannot take into consideration long temporal quality fluctuations for which mnesic processes (memory effects) occur that impact the overall perceived quality. The absence of a methodology for assessing time-varying speech quality motivates this Recommendation. Therefore, an assessment of the impact on perceived quality of these kinds of degradations and of their time fluctuations during a specific communication, requires speech sequences longer than those generally used in standard subjective methods. Moreover, with a method of continuous judgment, it becomes possible to study the impact of transient degradations on perceived quality at any single instant during the listening sequence, and on overall perceived quality (at the end of the sequence). This method was inspired by the method Single Stimulus Continuous Quality Evaluation (SSCQE) used in the video domain (ITU-R BT.500-11 [2]). It has also been validated for speech quality through several previous studies ([B-1], [B-2], [B-3], [B-4]). + +## **4.2 Test preparation** + +### **4.2.1 Stimuli** + +The speech material should be simple, meaningful, and easy to understand. Short speech sequences should be avoided and, provisionally, speech sequence durations between 45 seconds and 3 minutes can be used. Source recordings, including recording environment and procedure, sending and recording system, talkers and speech levels could be the same as those described in B.1/P.800 [1]. + +### **4.2.2 Sources** + +The number and the choice of the conditions depend on the purpose of the test. The only limit is the one imposed by the test duration. If possible, it is recommended including some control conditions in the set of test conditions, i.e., conditions without any variations of physical parameters. + +## **4.3 Listening session** + +### **4.3.1 Listeners** + +An adequate number (at least 24) naive listeners shall participate in the test. All the listeners shall be native speakers of the language used for the test and no listener shall have participated in a subjective experiment in the previous six months. + +### **4.3.2 Audio presentation and testing environment** + +Audio presentation shall comply with the guidelines given in ITU-T Rec. P.800 [1]. These guidelines include the listening system, listening levels, and listening environment. + +### **4.3.3 Continuous judgement recording device and set-up** + +An electronic slider (e.g., variable resistor) connected to a computer should be used for recording the continuous quality assessment from the subjects. This device should have the following characteristics: + +- slider mechanism without any "re-set" position (i.e., no automatic return to a pre-defined position); +- linear range of travel of about 10 cm; +- fixed or desk-mounted position; +- "slider position" samples recorded twice a second (fast enough to accurately capture responses from the subjects); + +- "slider position" could be coded from 0 (bottom of scale) to a maximum of 100 (top of scale), for an acceptable resolution. The initial slider position should be at the midpoint of the scale. + +### 4.3.4 Evaluation task + +For each speech sequence, the subjects' task is twofold: a continuous evaluation while listening to the sequence, and an overall evaluation at the end of the sequence. For more details and results from previous studies, see [B-1] and [B-2]. + +#### a) Continuous evaluation + +Firstly, subjects are instructed to assess the speech quality of the sequence continuously by moving a slider along a continuous scale so that its position reflects their opinion on quality at that instant; the subjects can position the slider anywhere on the scale. Five labels are shown along the scale, i.e., Excellent, Good, Fair, Poor and Bad to help the subject associate the slider position with suitable ranges of speech quality. + +*Continuous-quality scale* + +![Diagram of a continuous quality scale with five labels: Excellent, Good, Fair, Poor, and Bad. The scale is a vertical line with tick marks. The label 'P.880_F01' is at the bottom right.](0f3e3ea50bcceb86f6c524ab2b6f3e7a_img.jpg) + +The figure shows a vertical line representing a continuous quality scale. There are five tick marks along the line, each corresponding to a label. From top to bottom, the labels are: Excellent, Good, Fair, Poor, and Bad. The label 'P.880\_F01' is positioned at the bottom right of the scale. + +Diagram of a continuous quality scale with five labels: Excellent, Good, Fair, Poor, and Bad. The scale is a vertical line with tick marks. The label 'P.880\_F01' is at the bottom right. + +**Figure 1/P.880 – Continuous scale used for the instantaneous judgment** + +#### b) Overall evaluation + +Secondly, at the end of each sequence, subjects are asked to rate its overall quality on the following 5-category listening-quality scale (the same MOS scale used in the ACR). + +*Overall-quality scale (ACR)* + +| Quality of the speech | Associated score | +|-----------------------|------------------| +| Excellent | 5 | +| Good | 4 | +| Fair | 3 | +| Poor | 2 | +| Bad | 1 | + +### 4.3.5 Test procedure + +Prior to the assessment of the test speech sequences, subjects undergo training by listening to a few selected sequences. Training sequences should cover different quality levels and different quality fluctuations representative of the range of temporal fluctuations and quality levels that the subjects will encounter during the actual test. + +Generally the test consists of a number of sessions, separated with breaks. The entire set of stimuli (sequences) is presented in a different random order to the different groups of subjects. + +### 4.3.6 Instructions to subject + +An example of typical instructions is given in Table 1 (for speech sequences of duration T seconds). The written instructions must be given (verbally as well, if necessary) prior to the beginning of the experiment. + +**Table 1/P.880 – Example of instructions that would be given to the listeners** + +| | +|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +|

In this test, you will be listening to T-s speech sequences via the telephone handset. The speech quality of each sequence can vary in time in different ways. For each sequence, your tasks are:

  1. 1) To give your opinion on speech quality, during the entire sequence, i.e., at any instant, by moving the slider on the table in front of you. The rating scale is continuous so that you can place the slider anywhere, with its position reflecting your opinion of the speech quality.

Labels are shown along the scale (Excellent, Good, Fair, Poor and Bad) and are provided to help you for positioning the slider. For example, if you think that the quality corresponds exactly to a position between Fair and Good, you will place the slider in the middle of the two corresponding labels.

However, you do not have to position the slider either directly in front of a label or exactly half-way between two labels unless one of those positions accurately represents your opinion.

Do not forget to move the slider as the quality varies.

  1. 2) At the end of each sequence, you are asked to give an overall quality score that should reflect your opinion on the speech quality of the entire sequence. You give this score by pressing the appropriate button to indicate your opinion on the following scale:

Overall opinion on the speech quality of the sequence you have just heard:

Excellent

Good

Fair

Poor

Bad

You will have five seconds to record your answer by pushing the button corresponding to your choice.

There will be a short pause before the presentation of next sequence.

We will begin with a short practice session to familiarize you with the test procedure. The actual tests will take place during sessions of 10 to 15 minutes.

Thank you for your help.

| +|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| + +## 4.4 Statistical analysis + +For each subject, if T-s corresponds to the duration of each sequence (in seconds), a data file of $2 \times \text{T-s}$ values is recorded (i.e., one instantaneous score every 500 ms during T seconds), plus one scalar value (i.e., the overall quality judgment). The $2 \times \text{T-s}$ instantaneous values (from $t = 0$ until $t = 2T - 1$ ) are subsequently linearly transformed into values from 1 to 5, using the relation $S(t) = 1 + 4 \text{ (slider position/maximum)}$ , where $S(t)$ is the instantaneous opinion score. + +For each sequence, a mean instantaneous judgment (and its standard deviation) is obtained by averaging individual instantaneous judgments over the subjects, at each instant $t$ (i.e., every 500 ms). For each sequence, a mean overall judgment MOS (and its standard deviation) is obtained by averaging individual overall judgments over the subjects on the ACR scale. + +Substantial deviations between the continuous score and the overall score are likely indications that transient impairments were experienced. This will depend on the recency, number of occurrences and duration of the impairments relative to the overall score judgement. Further study is required before this information can be used in transmission planning. + +Statistical analysis (e.g., ANOVA) can be performed to identify significant effects present in the different experimental conditions. + +*Remark:* The responses from some subjects can be discarded if those responses exhibit high variations and appear for more than 10% of the time out of two inter-subject standard deviations (calculated over all subjects). + +# BIBLIOGRAPHY + +- [B-1] Contribution ITU-T [COM 12-94] (1999), *Continuous assessment of time-varying subjective vocal quality and its relationship with overall subjective quality*, (Source FT R&D). +- [B-2] GROS (L.), CHATEAU (N.), Instantaneous and Overall Judgments for Time-Varying Speech Quality: Assessments and Relationships, *Acta Acustica*, 87, 367-377, 2001. +- [B-3] Contribution ITU-T [COM 12-D139] (2000), *Study of the relationship between instantaneous and overall subjective speech quality for time-varying quality speech sequences: influence of recency effect*, (Source FT R&D). +- [B-4] Contribution ITU-T [COM 12-22] (2000), *Assessment of time-varying speech quality: comparison between listening situation and conversational situation*, (Source FT R&D). + + + +# SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|--------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series B | Means of expression: definitions, symbols, classification | +| Series C | General telecommunication statistics | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Construction, installation and protection of cables and other elements of outside plant | +| Series M | TMN and network maintenance: international transmission systems, telephone circuits, telegraphy, facsimile and leased circuits | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks and open system communications | +| Series Y | Global information infrastructure, Internet protocol aspects and Next Generation Networks | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.912-201603-I_PDF-E/raw.md b/marked/P/T-REC-P.912-201603-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..2a6f7dda725270bedc1762e1c89eba45261fdda9 --- /dev/null +++ b/marked/P/T-REC-P.912-201603-I_PDF-E/raw.md @@ -0,0 +1,573 @@ + + +International Telecommunication Union + +**ITU-T** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +**P.912** + +(03/2016) + +SERIES P: TERMINALS AND SUBJECTIVE AND +OBJECTIVE ASSESSMENT METHODS + +Audiovisual quality in multimedia services + +--- + +**Subjective video quality assessment methods +for recognition tasks** + +Recommendation ITU-T P.912 + +ITU-T + +![ITU logo](1d7527f4316cfe2d342b08d1653d1592_img.jpg) + +The logo of the International Telecommunication Union (ITU) features a globe with a red lightning bolt striking across it. To the right of the globe, the text "International Telecommunication Union" is written in blue, with "ITU" in a larger, bold font above it. + +ITU logo + +International +Telecommunication +Union + +# ITU-T P-SERIES RECOMMENDATIONS + +# TERMINALS AND SUBJECTIVE AND OBJECTIVE ASSESSMENT METHODS + +| | | | +|--------------------------------------------------------------------------------------------------|---------------|--------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Voice terminal characteristics | Series | P.30 | +| | | P.300 | +| Reference systems | Series | P.40 | +| Objective measuring apparatus | Series | P.50 | +| | | P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of speech quality | Series | P.80 | +| | | P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | +| Transmission performance and QoS aspects of IP end-points | Series | P.1000 | +| Communications involving vehicles | Series | P.1100 | +| Models and tools for quality assessment of streamed media | Series | P.1200 | +| Telemeeting assessment | Series | P.1300 | +| Statistical analysis, evaluation and reporting guidelines of quality measurements | Series | P.1400 | +| Methods for objective and subjective assessment of quality of services other than voice services | Series | P.1500 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +## Recommendation ITU-T P.912 + +## Subjective video quality assessment methods for recognition tasks + +# Summary + +Recommendation ITU-T P.912 defines subjective assessment methods for evaluating the quality of one-way video used for target recognition tasks. "Target" refers to something in the video that the viewer needs to identify (e.g., a face, object, or number). Target recognition video (TRV) is video that is used as a tool in order to accomplish a specific goal through the ability to recognize specific targets of interest in a video stream. TRV can be used in various video services such as surveillance, human identification, licence plate identification, telemedicine, robot control, and remote monitoring and decision-making. + +## History + +| Edition | Recommendation | Approval | Study Group | Unique ID* | +|---------|----------------|------------|-------------|---------------------------------------------------------------------------| +| 1.0 | ITU-T P.912 | 2008-08-13 | 9 | 11.1002/1000/9514 | +| 2.0 | ITU-T P.912 | 2016-03-15 | 9 | 11.1002/1000/12774 | + +--- + +\* To access the Recommendation, type the URL in the address field of your web browser, followed by the Recommendation's unique ID. For example, . + +# FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications, information and communication technologies (ICTs). The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Assembly (WTSA), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSA Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +Compliance with this Recommendation is voluntary. However, the Recommendation may contain certain mandatory provisions (to ensure, e.g., interoperability or applicability) and compliance with the Recommendation is achieved when all of these mandatory provisions are met. The words "shall" or some other obligatory language such as "must" and the negative equivalents are used to express requirements. The use of such words does not suggest that compliance with the Recommendation is required of any party. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementers are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database at . + +© ITU 2016 + +All rights reserved. No part of this publication may be reproduced, by any means whatsoever, without the prior written permission of ITU. + +# Table of Contents + +| | | Page | +|-----|-----------------------------------------------------|------| +| 1 | Scope..... | 1 | +| 2 | References..... | 1 | +| 3 | Definitions ..... | 1 | +| 4 | Abbreviations and acronyms ..... | 2 | +| 5 | Source signal..... | 2 | +| 6 | Test methods and experimental design..... | 2 | +| 6.1 | Multiple choice method ..... | 3 | +| 6.2 | Single answer method..... | 3 | +| 6.3 | Timed task method ..... | 4 | +| 6.4 | Real-time vs. viewer-controlled viewing ..... | 4 | +| 6.5 | Scenes ..... | 5 | +| 6.6 | Experimental design ..... | 5 | +| 6.7 | Reference conditions ..... | 5 | +| 7 | Evaluation procedures ..... | 5 | +| 7.1 | Viewing and listening conditions ..... | 5 | +| 7.2 | Processing and playback system ..... | 6 | +| 7.3 | Subjects..... | 6 | +| 7.4 | Instructions to subjects and training session ..... | 6 | +| 8 | Statistical analysis and reporting of results ..... | 6 | +| 8.1 | Subject screening..... | 6 | +| 8.2 | Further statistical analysis ..... | 7 | +| | Appendix I – Crowdsourcing Environment ..... | 10 | +| I.1 | Introduction..... | 10 | +| I.2 | Definitions ..... | 10 | +| I.3 | Software..... | 10 | +| I.4 | Designing a task..... | 11 | +| I.5 | Distribution of the campaign ..... | 11 | +| I.6 | Data analysis ..... | 12 | +| | Bibliography..... | 13 | + + + +## Recommendation ITU-T P.912 + +## Subjective video quality assessment methods for recognition tasks + +# 1 Scope + +This Recommendation defines subjective assessment methods for evaluating the quality of one-way video used for target recognition tasks. "Target" refers to something in the video that the viewer needs to identify (e.g., a face, object, or number). Target recognition video (TRV) is video that is used as a tool in order to accomplish a specific goal through the ability to recognize specific targets of interest in a video stream. TRV can be used in various video services such as surveillance, human identification, licence plate identification, telemedicine, robot control, and remote monitoring and decision-making. + +This Recommendation considers three categories of target: + +- 1) human identification (including facial recognition); +- 2) object identification; +- 3) alphanumeric identification. + +Each of these areas requires specific video test material that spans realistic conditions with stimuli that are carefully chosen to allow multiple scenarios to be created repeatedly with different objects of interest, in different lighting conditions or with small changes in scene details. + +# 2 References + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. The reference to a document within this Recommendation does not give it, as a stand-alone document, the status of a Recommendation. + +[ITU-T P.910] Recommendation ITU-T P.910 (2008), *Subjective video quality assessment methods for multimedia applications*. + +[ITU-T P.913] Recommendation ITU-T P.913 (2016), *Methods for the subjective assessment of video quality, audio quality and audio-visual quality of Internet video and distribution quality television in any environment*. + +# 3 Definitions + +This Recommendation defines the following terms: + +**3.1 discrimination class (DC):** One of four levels of visual discrimination at which the target can be analysed. + +- Elements of the action – in a very broad and general sense, identification of the series of events that took place. +- Target presence – recognition or detection of the presence or absence of valid targets. +- Target characteristics – recognition of unique characteristics of the target (e.g., markings, scars, tattoos, dents, colour). +- Target positive recognition – recognition of a specific instance of the target (e.g., recognition of a person, a specific object, or an exact alphanumeric sequence). + +**3.2 pretest:** An experiment run on a small set of subjects to determine any problem with the experiment. The experiment is redesigned based on the pretest data and then the pretest data is discarded. + +**3.3 scenario group (SG):** A collection of scenes of the same basic scenario, with very slight differences between the scenes. + +# **4 Abbreviations and acronyms** + +This Recommendation uses the following abbreviations and acronyms: + +ACR Absolute Category Rating + +CPD Cycles Per Degree + +DC Discrimination Class + +PVS Processed Video Sequence + +SG Scenario Group + +TRV Target Recognition Video + +# **5 Source signal** + +Test sequences should follow the general principles stated in [ITU-T P.913] (or [ITU-T P.910] for those situations where it is more appropriate) and [b-ATIS-0100801.01], which specify that scenes should be consistent with the transmission service under test, and should span the full range of spatial and temporal information. It is critical for the nature of these evaluations that the stimuli used actually reflect the true operational parameters of the conditions under which the video material is collected. If the stimuli used cannot actually cover the entire range of scenarios possible for the application area for identification, the application description has to be explicitly limited. For example, the results should not be generalized. Unlike other subjective assessment methods developed for quality evaluations, this method is directed at the usefulness of the video material to complete a task and not the quality of the video itself. + +# **6 Test methods and experimental design** + +For video that is used to perform a specific task, it may not be appropriate to rate the quality of the video according to a subjective scale, such as the absolute category rating (ACR) [ITU-T P.913] (or [ITU-T P.910] for those situations where it is more appropriate). The goal of test methods for TRV is to assess the ability of a viewer to recognize the appropriate information in the video, regardless of the viewer's perceived quality of the viewing experience. To assess the quality level of TRV, methods that reduce subjective factors and measure the ability of a participant to perform a task are useful in that they avoid ambiguity and personal preference. + +The application of TRV is directly related to the ability of the user to recognize targets at increasing levels of detail. These levels are referred to as discrimination classes (DCs). When determining the DC for particular scenarios, it needs to be considered that, for a set distance from the camera to the object of interest, the DC directly correlates to decreasing video resolution of the target, and therefore the object is represented by fewer cycles per degree (CPD) of resolution. Fewer CPD of resolution also mean that the object subtends less of the information content of the video, making identification of the target more difficult. + +CPD, the key parameter, is affected by the resolution of the object and (potentially) the distance between the camera and the object [b-Leszczuk, 2011]. Consequently, it relates to the achievable DC. + +Examples of the achievable DC are shown in Figure 1. If the distance between the camera and the object is 50 m, "Target Positive Recognition" is possible; for 215 m "Target Characteristics"; but for 430 m only "Target Presence". + +![Figure 1: DC in testing methods for various distances between camera and object. The figure shows three images of a car at distances of 50m, 215m, and 430m from a camera. Below the images is a scale with a camera icon and car icons at each distance.](d4e9f8f6bf5d7853ecae9c9633900af1_img.jpg) + +The figure illustrates the relationship between distance and achievable detection capabilities. It consists of three photographs of a black car at a parking entrance, taken from a fixed camera position at distances of 50m, 215m, and 430m. Below the photographs is a horizontal scale bar. On the left is a camera icon. To its right are three car icons, each positioned above a distance label: '50m', '215 m', and '430m'. The car at 50m is clearly visible, while the car at 430m is very small and blurry. + +Figure 1: DC in testing methods for various distances between camera and object. The figure shows three images of a car at distances of 50m, 215m, and 430m from a camera. Below the images is a scale with a camera icon and car icons at each distance. + +**Figure 1 – DC in testing methods for various distances between camera and object** + +Experimental methods should consist of responding to questions relating to the content in the image or video. The parameter addressed by the question is the target to be recognized. + +## 6.1 Multiple choice method + +This method is appropriate for all DC levels and target categories (human, object and alphanumeric). For this method, the video is shown above a list of verbal labels representing the possible answers. After presenting the video, the viewers must choose the label closest to what they recognized in the clip. The use of fixed multiple choices eliminates any possible ambiguity that could arise from open questions, and allows for more accurate measurements. + +The number of choices offered to the viewer depends on the number of alternative scenes being presented. The use of "Unsure" as one of the listed choices is discouraged, but allowed. The experimenter should be aware that individual subjects tend to overuse the "Unsure" choice, leading to contamination of results. Consequently, special care must be taken when "Unsure" is one of the listed choices. + +An example of the test screen a viewer would see is shown in Figure 2. + +![Figure 2: Multiple choice method for recognition of objects in a video clip: test screen and list of possible visual recognition responses. The left side shows a video frame of a person holding an object. The right side shows a test screen with the question 'What was the person holding?' and six possible answers: Gun, Banana, Comb, Hairpiece, Knife, Potato.](e2eb8b8c35f32b665245d2c24d337dca_img.jpg) + +The figure shows a split-screen view. On the left is a video frame showing a person in a red shirt walking past a concrete wall, holding a small object. On the right is a user interface for a multiple-choice test. At the top, it says 'Please answer clip 1 of 20' and 'What was the person holding?'. Below this are six rectangular buttons arranged in a 2x3 grid, containing the following text: 'Gun', 'Banana', 'Comb' in the top row, and 'Hairpiece', 'Knife', 'Potato' in the bottom row. + +Figure 2: Multiple choice method for recognition of objects in a video clip: test screen and list of possible visual recognition responses. The left side shows a video frame of a person holding an object. The right side shows a test screen with the question 'What was the person holding?' and six possible answers: Gun, Banana, Comb, Hairpiece, Knife, Potato. + +**Figure 2 – Multiple choice method for recognition of objects in a video clip: test screen and list of possible visual recognition responses** + +## 6.2 Single answer method + +If there is an unambiguous answer to an identification question, the single answer method may be used. This method is appropriate for alphanumeric character recognition scenarios. A viewer is asked which letter(s) or number(s) was/were present in a specific area of the video, and the answer can be + +evaluated as either correct or incorrect. Alternatively, fuzzy logic may be used (e.g., Hamming distance or Levenstein distance), as seen in [b-Leszczuk, 2014]. + +Yes or No tests also fall under this method. A viewer may be asked if a certain object was present in the clip, for example. In this method, it is important to ensure that the procedure used to gather viewer responses is easy to understand, so that the test interface does not distract from the cognitive processing required for actual identification of the alphanumeric characters or object. Care must also be taken to avoid terminology that can differ from participant to participant. + +The use of "Unsure" as the third possible answer is discouraged, but allowed. The experimenter should be aware that individual subjects tend to overuse the "Unsure" choice, leading to contamination of results. Consequently, special care must be taken when "Unsure" is one of the listed choices. + +An example of an alphanumeric single answer viewer screen is shown in Figure 3. + +![Figure 3: Single answer method for unambiguous responses to video object recognition: test screen and keyboard for entry of alphanumeric character recognition. The image shows a rear view of a dark grey SUV on a road. To the right is a test interface with a text input field and a grid keyboard.](e9b30aeb317ed964fa6de36804acf24c_img.jpg) + +The image displays a test interface for video object recognition. On the left is a photograph of a dark grey SUV from a rear-quarter perspective. On the right is a software interface with a light beige background. At the top of the interface is the text "What was on the license plate?". Below this is a white rectangular text input field. Underneath the input field is a grid keyboard. The keyboard is organized into four rows: the first row contains digits 1 through 0; the second row contains uppercase letters A through J; the third row contains uppercase letters K through T; and the fourth row contains lowercase letters u through z. + +Figure 3: Single answer method for unambiguous responses to video object recognition: test screen and keyboard for entry of alphanumeric character recognition. The image shows a rear view of a dark grey SUV on a road. To the right is a test interface with a text input field and a grid keyboard. + +**Figure 3 – Single answer method for unambiguous responses to video object recognition: test screen and keyboard for entry of alphanumeric character recognition** + +## 6.3 Timed task method + +A viewer may be asked to watch for a particular action or object to be recognized in the video clip. When the viewer perceives that the target has occurred, a timer button can be pushed. In the timed task, the experimenter is able to determine whether the time falls within an acceptable time-frame for decision-making. These time-frames are defined by the field in which the video is used, e.g., a person responding to a riot who needs to identify whether crowd members have real weapons versus a person who is chasing a car and needs to read the licence plate. + +## 6.4 Real-time vs. viewer-controlled viewing + +Depending on the nature of the task, TRV test methods can be used, either in real time, without the ability to freeze or rewind, or they can be used for non real-time analysis. The experiment should mimic the real world application of the video. If the intended use of the video is analysis, the subject under test should have the capability to control the playing of the test clip. + +## 6.5 Scenes + +Since TRV is generally used to perform a recognition task, the scenes should contain targets consistent with the application under study. However, because the measurements are focused on a subject's ability to identify objects and actions, the possibility that a viewer may memorize the scene content and use other visual clues to remember the identity of the target must be addressed. Therefore, an individual scene may be replaced by a set of scenes containing multiple versions, with controlled differences between the versions. This is called a scenario group (SG). For example, the scenario could be that a person walks across the field of view carrying an object. The SG would consist of multiple shots using different objects or different people. The number of scenes in a SG should be large enough so that scene memorization is unlikely. An example of three scenes from one scenario group is shown in Figure 4. The scene content is almost identical except for the single change in the object being held. + +![Figure 4 shows three side-by-side images of a woman walking from left to right against a concrete wall. In the first image, she is holding a small white object. In the second image, she is holding a small black object. In the third image, she is holding a small black object, but her arm is extended forward, making the object appear larger and more prominent. The background is a plain concrete wall with a horizontal line and a shadow cast by the woman.](6d9013c24741e861f3c8e0a763b6da22_img.jpg) + +Figure 4 shows three side-by-side images of a woman walking from left to right against a concrete wall. In the first image, she is holding a small white object. In the second image, she is holding a small black object. In the third image, she is holding a small black object, but her arm is extended forward, making the object appear larger and more prominent. The background is a plain concrete wall with a horizontal line and a shadow cast by the woman. + +**Figure 4 – A scenario group of images: scene 1, scene 2, scene 3** + +The content of the scenes should be determined by experts in the application for which the video is used. These experts should identify critical *tasks*, critical *scenes* in which these tasks are accomplished and critical *parameters* of the scenes. These parameters are used in the design of the experiment to create the set of multiple choice answers. The scenes should be created in a way that the parameters of interest appear in the video at a resolution that can be realistically expected in practice; i.e., the parameters should occupy a realistic percentage of the field of view. + +## 6.6 Experimental design + +The experimenter should follow the guidelines outlined in [ITU-T P.913] (or [ITU-T P.910] for those situations where it is more appropriate). + +## 6.7 Reference conditions + +The experimenter should follow the guidelines outlined in [ITU-T P.913] (or [ITU-T P.910] for those situations where it is more appropriate). + +# 7 Evaluation procedures + +In clause 7, a laboratory test is described. Description of a crowdsourcing environment is described in Appendix I. + +## 7.1 Viewing and listening conditions + +The experimenter should follow the guidelines outlined in [ITU-T P.913] (or [ITU-T P.910] for those situations where it is more appropriate). + +## 7.2 Processing and playback system + +The experimenter should follow the guidelines outlined in [ITU-T P.913] (or [ITU-T P.910] for those situations where it is more appropriate). + +## 7.3 Subjects + +Subjects who are experts in the TRV application field should be used. For certain areas of application testing, where neither specific experience nor expertise is required, use of non-expert subjects is also permissible. Such non-experts must be motivated in other than a professional manner (e.g., they are paid). The validity of this approach is shown in [b-Leszczuk, 2012]. The number of subjects should follow the recommendations of [ITU-T P.913] (or [ITU-T P.910] for those situations where it is more appropriate). + +## 7.4 Instructions to subjects and training session + +The subject should be given the context of the task before the video clip is played, and told what they are looking for or trying to accomplish. If questions are to be answered about the content of the video, the questions should be posed before the video is shown, so that the viewer knows what the task is. + +It is safe to assume that there are no easy tasks. Even something as easy as recognizing a character must be described in detail. This means the instructions must clearly state what subjects must do if: + +- 1 they cannot recognize a character; +- 2 they have doubts; +- 3 they can recognize some, but not all, characters. + +The optimal training session must show all specific cases and the correct scoring behaviour (i.e., that desired by the experiment design). + +Especially difficult is to define a task for specialists, e.g., medical doctors. In this case, the running of a pretest on a small group before running any larger experiment is strongly recommended. A typical number of subjects for a pretest is approximately 20% of the total. A pretest group can consist of a single person. Specialists often change the task so that it fits a real situation typical for a particular specialist better. This can change the experimental conditions and finally harm the experiment itself. Therefore, it is very important in the pretest to clearly explain the task, the reason for running the test and the reasons why the experiment has been set up in a particular way. Feedback from the pretest is used to improve the experiment before running it with actual subjects. + +# 8 Statistical analysis and reporting of results + +The first step of the analysis is subject screening to eliminate those who did not pay attention or who did not understand the task. Further statistical analyses vary slightly depending on the scoring method. + +## 8.1 Subject screening + +This technique is optional. + +In order to detect abnormal subjects, it is not enough to compare the results obtained by one subject to the average obtained in the experiment, since in a typical experiment different subjects perform different tasks (see clause 6.5). Even with careful design, the tasks performed by one subject can be more difficult than the tasks performed by the average subject. An algorithm for solving the problem of different task difficulty performed by different subjects is proposed in [b-Janowski, 2012]. + +The algorithm proposed assumes that tasks can be partially ordered. For example, consider an experiment where the goal is to specify detection probability as a function of the compression bitrate. The processed video sequences (PVSs) obtained for the same source and lower bitrate are likely to + +have less information and likely the detection is at least not easier than for a higher bitrate. Also if an object covers fewer pixels on a screen, it is not easier to detect. + +Based on the above assumption, consider a list of tasks $T_1^{(l)}, \dots, T_{n_l}^{(l)}$ , which are ordered by difficulty. The list starts from the easiest, $T_1^{(l)}$ , to the most difficult task, $T_{n_l}^{(l)}$ . If a subject did not manage task $T_j^{(l)}$ and some tasks $T_{j+k}^{(l)}$ were correctly performed, this indicates that a subject did not pay enough attention or that another problem occurred. Moreover, if a subject did not manage task $T_j^{(l)}$ and another subject correctly solved task $T_{j+1}^{(l)}$ , the error is less serious compared with the situation where task $T_{j+10}^{(l)}$ was solved correctly. Therefore, the penalty for a subject from not solving a task that was solved by another subject must be a function of the difference in task difficulty. The final proposed equation is: + +$$ssq_{i,j,l} = (1 - r(i,j,l)) \sum_{m=1}^M \sum_{k=j+1}^{n_l} (k - j) r(m,k,l) \quad (1)$$ + +where $r(i,j,l)$ is 0 if the task $T_j^{(l)}$ was performed incorrectly by subject $i$ and 1 otherwise, and $M$ is total number of subjects. + +The result of equation (1) is 0 if task $j$ is performed correctly by subject $i$ . It is also 0 if all more difficult tasks were performed incorrectly by all subjects. A higher value indicates that more subjects or more difficult tasks were done correctly. + +The final value obtained for user $i$ is: + +$$Sq_i = \sum_{l=1}^L \sum_{j=1}^{n_l} ssq_{i,j,l} \quad (2)$$ + +where $L$ is the total number of partially ordered groups. + +A large value of $Sq_i$ indicates that a subject is irrelevant or that the task conditions and results must be double checked. For example, $Sq_i$ detects a subject in one experiment who confused the terms "radio" and "mobile phone". The value of $Sq_i$ depends on the experiment size and the length of the partial sorted groups within the experiment, therefore a threshold value cannot be specified. It must be used as an indicator for further investigation or, if no outliers are detected, proving that all subjects behave correctly and the results can be analysed further. + +Show $Sq_i$ distribution when the results are reported. + +## 8.2 Further statistical analysis + +The statistical analysis for each method varies slightly. + +For all conditions, a correlation and understanding of the number of CPD or area subtended of the target are taken into consideration to determine the correlation between success and CPD. + +For cases where there are multiple answers, a statistical validity indicator is required. + +### 8.2.1 Multiple choice + +For multiple-choice answers, the probability of an incorrect answer needs to be balanced against the ability to answer the questions correctly. The statistical metric in this situation requires an examination of the stability of the answers within and between subject performance metrics. "Unsure" answers should be pooled with those that are incorrect. + +##### 8.2.1.1 Recognition probability as a function + +For multiple choice, estimate the probability of correct answer as a function of independent variables, like bitrate or camera quality. Probability as a function of independent variables can be calculated by logistic regression [b-Agresti, 2002]. Logistic regression can be found in almost all statistical packages. The main difference between logistic and linear regression is that instead of modelling the + +response variable (i.e., 1=correct and 0=incorrect), logistic regression models the probability that the response variable has particular value. The simplest logistic regression is given by: + +$$\pi(x) = \frac{\exp(ax+b)}{1+\exp(ax+b)} \quad (3)$$ + +where $\pi(x) = P(Y = 1|X = x)$ i.e., the probability of a correct answer, in which the explanatory variable is $x$ ; $a$ and $b$ are model parameters. + +Much more complicated parameters and explanatory variables can be considered, e.g., if there are two explanatory variables $x$ and $y$ , a logit model could be: + +$$\pi(x) = \frac{\exp(ax+by+cxy+d)}{1+\exp(ax+by+cxy+d)} \quad (4)$$ + +Note that for such model both $x$ and $y$ must be normalized, such that the unit change of $x$ is similar (i.e., has a similar influence on probability) to the unit change of $y$ . + +If answer "Unsure" is used, it is analysed in the same way as any other answer. + +##### 8.2.1.2 Comparing different conditions + +A function showing the probability of correct recognition as a function of specific parameters is one way of analysing multiple-choice data. The second important way of analysing multiple-choice data is to compare two or more different conditions. For each condition, the numbers of correct and incorrect recognitions are collected. The goal is to determine whether the observed difference is statistically significant. + +The collected data can be represented by a matrix. Assuming that there are $k$ different conditions, the matrix has the form shown in Table 1. + +**Table 1 – Collected data matrix representation** + +| Condition | Correct | Incorrect | Sum | +|-----------|----------|-----------|----------| +| $A_1$ | $n_{11}$ | $n_{10}$ | $n_{1.}$ | +| $A_2$ | $n_{21}$ | $n_{20}$ | $n_{2.}$ | +| ... | ... | ... | ... | +| $A_k$ | $n_{k1}$ | $n_{k0}$ | $n_{k.}$ | +| Sum | $n_{.1}$ | $n_{.0}$ | $N$ | + +In order to answer the question whether all conditions are statistically the same, the $\chi^2$ -test is needed.1 The test is performed by comparing the detection probability obtained for each condition with the overall detection probability. The overall detection probability is calculated to be + +$$p = \frac{\sum_{i=1}^k n_{i1}}{N} \quad (5)$$ + +where $N = \sum_{i=1}^k (n_{i1} + n_{i0}) = \sum_{i=1}^k n_{i.}$ and $k$ represents the number of conditions. The $\chi^2$ -test is given by + +$$\chi^2 = \sum_{i=1}^k \frac{(n_{i0} - n_{i.}(1-p))^2}{n_{i.}(1-p)} + \sum_{i=1}^k \frac{(n_{i1} - n_{i.}p)^2}{n_{i.}p} \quad (6)$$ + +Commonly in experiments, the number of answers given for each condition, denoted here by $n_{i.}$ , is the same value; denote it $n$ . In such a case, equation (6) simplifies to: + +1 Note that $\chi^2$ test does not answer which condition is different, it only answers if all of them are the same or not. Comparing some conditions is described later in this clause. + +$$\chi^2 = \sum_{i=1}^k \frac{(n_{i0} - n(1-p))^2}{n(1-p)} + \sum_{i=1}^k \frac{(n_{i1} - np)^2}{np} \quad (7)$$ + +The value obtained is compared using the $\chi^2$ -distribution with $k - 1$ degrees of freedom. If the value obtained is greater than the value of the $\chi^2$ -distribution calculated for a specific significance level (typically 0.05), the hypothesis that all conditions are the same is rejected. + +After comparing many conditions, comparison of a subset of these conditions can be of interest. In this case, the significance level has to be correctly set. The reason why this has to be done can be explained in an example. Assume that 100 conditions are considered. The test shows that there is no statistical difference. Nevertheless, if the first condition is compared with all other conditions, statistically five of them could be statistically significantly different. The simplest way to correct the significance level is to use the Bonferroni correction, whose formula is: + +$$\alpha_{cor} = \alpha/N \quad (8)$$ + +where $\alpha_{cor}$ is the corrected significance level, $\alpha$ is the significance level used to compare the group, and $N$ is the total number of comparisons, typically $N = k(k - 1)$ , where $k$ is the number of conditions considered. + +#### 8.2.2 Single answer + +For single answer conditions, where answers are either correct or incorrect, a statistical metric to determine whether the subject performs above the random chance of answering correctly should be implemented. "Unsure" answers should be pooled with incorrect answers. + +For a single answer, the correctness of the answer can be analysed on a different scale. The simplest scale is 0-1 correct/incorrect. The correctness threshold can be different depending on the specific analysis. Since the final results are of the 0-1 type, the results obtained are similar to those for the multiple-choice case and the same analytical tools must be used. + +If the correctness of the answer is analysed, different models can be used. It is difficult to describe all options since the answer can be very different depending on the answer type. Most probably, correctness can be analysed by the generalized linear model described in [b-Agresti, 2002]. + +### 8.2.3 Timed task + +For timed tasks, the statistical analysis should incorporate two metrics that are finally correlated against each other to understand the impact of correctness versus time taken to perform the task. + +The timed factor is a straight average of time to identify the object, that is then weighted against the correctness of the answer. For the correctness factor, the same statistical analysis as for single answer conditions is also applied. + +For timed tasks, the statistical analysis must incorporate time as an explanatory variable. Time can be a numerical value "how long it took to finish the task, in seconds" or it could be "number of replays of the movie before a decision was made." The analysis must indicate the influence of time on the result obtained. + +# Appendix I + +## Crowdsourcing Environment + +(This appendix does not form an integral part of this Recommendation.) + +## I.1 Introduction + +One of the main problems of recognition tasks is the obvious limitation of source sequences reuse as described in clause 6.5. The best way to protect against source sequence remembrance is to prevent the showing of the same source sequence to the same subject more than once. Nevertheless, such a solution has an obvious drawback: it requires a much larger number of subjects. For laboratory tests, it is difficult to achieve a sufficient number of subjects. A natural solution is crowdsourcing, which gives access to thousands of potential subjects at the same time. + +The advantage of accessing large number of subjects comes with a price of lack of control over the subjects and environment. This Recommendation describes possible ways to increase the accuracy of results obtained and is based on a white paper of which more details can be found in [b-Hossfeld, 2014]. + +## I.2 Definitions + +**I.2.1 crowdsourcing:** Obtaining the needed service by a large group of people, most probably an on-line community. + +**I.2.2 test:** Subjective assessments in a crowdsourcing environment. + +NOTE – I.2.2 follows terminology presented in [b-Hossfeld, 2014]. + +**I.2.3 worker:** Person participating in a crowdsourcing test. + +NOTE – I.2.3 follows terminology presented in [b-Hossfeld, 2014]. + +**I.2.4 task:** Set of actions that a worker needs to perform to complete a subscribed part of the test. + +NOTE – I.2.4 follows terminology presented in [b-Hossfeld, 2014]. + +**I.2.5 question:** A single event that requires an answer for a worker. A task contains many questions. + +**I.2.6 campaign:** A group of similar tasks. It also contains a more detailed description of the part of the test that is under investigation, like workers' payment, and indicates subjective assessments in a crowdsourcing environment. A test can contain multiple campaigns. + +NOTE – I.2.6 follows terminology presented in [b-Hossfeld, 2014]. + +An example of using definitions I.2.1 to I.2.6 is as follows. A research question calls for a subjective experiment. It was decided to run the experiment as a crowdsourcing test. The test goal is to answer the given question. This test can be divided into multiple campaigns, namely preliminary, main and corrected. Each campaign contains multiple questions that have to be answered. Those questions are grouped into tasks. Each task is assigned to a single worker, but a single question can be asked in many different tasks, also some tasks can be identical, although they are performed by different workers. + +## I.3 Software + +In order to be able to run a crowdsourced test, a worker has to have access to the test environment. Implementation of the test as a web service, which can be easily accessed by anyone with an Internet connection, is advised. Of course, other solutions, like software code or an application, can also be used, but the number of workers willing to install additional software, compared to those willing to access a specific web page, is much smaller. Even the use of a very specific web browser plugin can restrict the number of participants significantly. + +Regardless of the test software, it is important to include a feedback channel, which can help in detecting errors or improving the test. + +## **I.4 Designing a task** + +The task preparation should take into account all lessons learned from any laboratory study, if such studies were conducted. Any additional questions asked by subjects should be addressed. Note that a worker cannot ask an additional question or at least it is not easy to do. Therefore, all problems should be solved before the task is sent to the workers. + +Also the task itself has to be easy. Any question asked should be tested against any misinterpretation. Consultation with non-native speakers to ask their opinion is a good idea, since it is probable that some workers do not speak English well. For the same reason, use simple English in all descriptions, questions and messages presented to workers. If possible, enrich the text with pictures. For example, if the task is to recognize an object, it is recommended that pictures of the object be added, not only written descriptions. + +It is important that the task be short, so that a single worker does not take a long time to complete it. The fact that the task is performed in a home environment needs to be taken into account. A long task can be easily interrupted by external events, like a phone call. A shorter task is more likely to be finished in a single session. A short task requires that detailed description be limited. This is also in line with the simplification of the description. It is also common to collect social information, which also should be limited to the most important questions. The answering interface should be prepared in such way that it can be filled in fast. + +The task should be short, but training cannot be skipped. It is very important for a worker to know that this question is just a training so he can explore the interface. Since each worker supplies very few answers, it is important to ensure that none of them are lost due to misunderstanding the interface. + +Even a correctly design interface can be problematic for some workers, and others can simply cheat or answer randomly. Therefore, unreliable workers need to be detected. Detecting unreliable workers should be included in the experimental design by adding specific questions. In a recognition test, extending the test by adding obvious questions is a good idea. Also, some repeated questions that detect random clickers are needed. An interesting idea is presented in [b-Gardlo, 2014] where an additional monitor test is used. Such a test has two advantages: the first is the detection of random clickers and the second is the detection of workers with very bad screens or lighting conditions. + +A task for a worker is similar to a typical subjective experiment for a subject. Nevertheless, the test software has to log much more information than software used in the laboratory study. In general, all information available should be recorded. Among required parameters are: response time; browser type and version; operating system; and screen resolution. This information is crucial for detecting unreliable workers. The most important step after data compilation is finished and before the main conclusions are taken is to clear the data of answers given by unreliable workers. Explain the plan carefully in advance and specify how detection is to be done. The detection method should be based on more data than just the answers to the main research questions. The logs created in the experiment design should be used. + +## **I.5 Distribution of the campaign** + +After creating the test platform, distribute it among subjects. There are two main ways to advertise a specific campaign. + +1 Using social media and mailing lists + +a Advantages: + +- i it is possible to get to specific group, e.g., policemen; +- ii quite often it does not include additional costs; + +- iii workers willing to make a task for free are most of the time honest. + - b Disadvantages: + - i the mailing list or social media generate(s) a very specific (probably biased) group of workers; + - ii since no payment is made for the task, a large number of tests will not be completed, unless the test is extremely short or involves gamification; + - iii the speed of collecting the data is, most of the time, very rapid just after announcement, but falls away rapidly, meaning that the web server can be overloaded; + - iv it is difficult to predict how many answers will be collected; + - v checking whether an individual ran the task once only is difficult. +- 2 Using specific services (called crowdsource platforms) gathering people willing to make micro tasks +- a Advantages: + - i the speed of collecting the data can be adjusted; + - ii the task is advertised constantly by the service; + - iii a large number of data can be collected in a short period of time. + - b Disadvantages: + - i some workers will use the test just to get money and their answers are random; + - ii every answer, even those given by workers answering randomly, costs some money; + - iii workers are pooled from a specific group of people willing to make money by doing micro tasks. + +## I.6 Data analysis + +Even with careful subject validation, assume that subjects are different. Since each sequence is validated by a diverse subgroup of subjects, the difference in recognition probability can be characterized only by subgroup of subjects, not by difference in conditions. Nevertheless, results presented in [b-Korshunov, 2012] show high correlation between results obtained in a laboratory environment and those from crowdsourcing. Such a result is possible only after removing unreliable subjects. + +# Bibliography + +- [b-ITU-T P.911] Recommendation ITU-T P.911 (1998), *Subjective audiovisual quality assessment methods for multimedia applications*. +- [b-ITU-T P.920] Recommendation ITU-T P.920 (2000), *Interactive test methods for audiovisual communications*. +- [ITU-R BT.500-13] Recommendation ITU-R BT.500-13 (2012), *Methodology for the subjective assessment of the quality of television pictures*. +- [b-ATIS-0100801.01] ATIS-0100801.01.1995 (R2011), *Digital Transport of Video Teleconferencing/Video Telephony Signals – Video Test Scenes for Subjective and Objective Performance Assessment* +- [b-Agresti, 2002] Agresti, A., *Categorical data analysis*, 2nd edition (2002), Hoboken, NJ: Wiley. +- [b-Gardlo, 2014] Gardlo, B.; Egger, S.; Seufert, M.; Schatz, R. (2014), Crowdsourcing 2.0: Enhancing execution speed and reliability of web-based QoE testing. In: *IEEE International Conference on Communications (ICC)*, pp.1070-1075, 10-14 June. +- [b-Hossfeld, 2014] Hoßfeld, T.; Hirth, M.; Redi, J.; Mazza, F.; Korshunov, P.; Naderi, B.; Seufert, M.; Gardlo, B.; Egger, S.; Keimel, C. (2014), *Best practices and recommendations for crowdsourced QoE - Lessons learned from the Qualinet Task Force Crowdsourcing*, European Network on Quality of Experience in Multimedia Systems and Services (COST Action IC 1003 Qualinet). +- [b-Janowski, 2012] Janowski, L. (2012), Task-based subject validation: Reliability metrics. In: *Fourth International Workshop on Quality of Multimedia Experience (QoMEX)*, pp. 182-187, 5-7 July. +- [b-Leszczuk, 2011] Leszczuk M., Janowski L., Romaniak P., Głowacz A., Mirek R. (2011), Quality assessment for a licence plate recognition task based on a video streamed in limited networking conditions, In: *Proceedings of the Multimedia Communications, Services and Security*, Poland, June, Springer. +<[http://link.springer.com/chapter/10.1007/978-3-642-21512-4\\_2](http://link.springer.com/chapter/10.1007/978-3-642-21512-4_2)> +- [b-Leszczuk, 2012] Leszczuk, M., Koń, A., Dumke, J., Janowski, L. (2012), Redefining the ITU-T recommendation P.912 for requirements of the subjects of quality assessments in recognition tasks, In: *Proceedings of the 5th International Conference on Multimedia Communications, Services and Security*, Krakow, Poland, June, Springer CCIS Vol. 287. +<[http://link.springer.com/chapter/10.1007%2F978-3-642-30721-8\\_19](http://link.springer.com/chapter/10.1007%2F978-3-642-30721-8_19)> +- [b-Leszczuk, 2014] Leszczuk, M. (2014), Optimising task-based video quality: A journey from subjective psychophysical experiments to objective quality optimisation, In: *Multimedia Tools and Applications*, United States, January, Springer Vol. 68. +<> +- [b-Moyer] Moyer, S., Hixson, J., Edwards, T., and Krapels, K. (2004), *Identification of Small Handheld Objects for Electro-Optic/FLIR Applications*, Optical Engineering. P. 0632201-1 – 12. +<> + +[b-O'Connor] O'Connor, J., O'Kane, B., Ayscue, K., Bonzo, D., and Nystrom, B. (1998), *Recognition of human activities using handheld thermal systems*, Proceedings of the SPIE Conference on Sensor Technology for Soldier Systems, Orlando, Florida, April, SPIE Vol. 3394. +<> + + + +## **SERIES OF ITU-T RECOMMENDATIONS** + +| | | +|-----------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Cable networks and transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Environment and ICTs, climate change, e-waste, energy efficiency; construction, installation and protection of cables and other elements of outside plant | +| Series M | Telecommunication management, including TMN and network maintenance | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Terminals and subjective and objective assessment methods | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks, open system communications and security | +| Series Y | Global information infrastructure, Internet protocol aspects and next-generation networks, Internet of Things and smart cities | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file diff --git a/marked/P/T-REC-P.916-201603-I_PDF-E/1c140ee5d1a87186977da94a96bfa700_img.jpg b/marked/P/T-REC-P.916-201603-I_PDF-E/1c140ee5d1a87186977da94a96bfa700_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..3f4b0dbb838d9df7edb32c5493d4db5963f5164b --- /dev/null +++ b/marked/P/T-REC-P.916-201603-I_PDF-E/1c140ee5d1a87186977da94a96bfa700_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9ef16413a59bc260d95a6f8e1a70fafa5f760e7ecf71ee9c3526020a10810f72 +size 27584 diff --git a/marked/P/T-REC-P.916-201603-I_PDF-E/5f2c99ae08864cf2d5c949947bac2b98_img.jpg b/marked/P/T-REC-P.916-201603-I_PDF-E/5f2c99ae08864cf2d5c949947bac2b98_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..577bf8221921baccdbcc3bc5fa4ffacc8f798b32 --- /dev/null +++ b/marked/P/T-REC-P.916-201603-I_PDF-E/5f2c99ae08864cf2d5c949947bac2b98_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:8e07b0b71cb48ffd63ebfb3fd47e0eb93d3e0f3392590734ff836f609d4873b7 +size 39616 diff --git a/marked/P/T-REC-P.916-201603-I_PDF-E/60e9207be66a64332619bb4b667fe67b_img.jpg b/marked/P/T-REC-P.916-201603-I_PDF-E/60e9207be66a64332619bb4b667fe67b_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..eb0760d0113981230902e82e1f411af52db5beb5 --- /dev/null +++ b/marked/P/T-REC-P.916-201603-I_PDF-E/60e9207be66a64332619bb4b667fe67b_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:9b33659ac921efd0447b3aa476d84d8f8201401273775ba9bb062fb03e244b54 +size 67275 diff --git a/marked/P/T-REC-P.916-201603-I_PDF-E/84a1d09fb489061482111515543b60dc_img.jpg b/marked/P/T-REC-P.916-201603-I_PDF-E/84a1d09fb489061482111515543b60dc_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f802537e0d518b939ca667439dc55a75350771e1 --- /dev/null +++ b/marked/P/T-REC-P.916-201603-I_PDF-E/84a1d09fb489061482111515543b60dc_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:77fdf36e5ef2a6a2f8b773a079111fca553fadb6d9498c977c1f9a90d1b6fac2 +size 6786 diff --git a/marked/P/T-REC-P.916-201603-I_PDF-E/aaf3e6e44cdeabd6d1df869c5f392ea1_img.jpg b/marked/P/T-REC-P.916-201603-I_PDF-E/aaf3e6e44cdeabd6d1df869c5f392ea1_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..eb34d21ac91e5bb77cd83227828c421c055bcf82 --- /dev/null +++ b/marked/P/T-REC-P.916-201603-I_PDF-E/aaf3e6e44cdeabd6d1df869c5f392ea1_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:52580e3fcfdf58dc606472c4efa64e2586b95e229edee8a9f87ee001f695e008 +size 9634 diff --git a/marked/P/T-REC-P.916-201603-I_PDF-E/d48475a25698b1c0592e4cfe07138f2a_img.jpg b/marked/P/T-REC-P.916-201603-I_PDF-E/d48475a25698b1c0592e4cfe07138f2a_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e9bb0f4a6508242a707f596284f9111db72edef7 --- /dev/null +++ b/marked/P/T-REC-P.916-201603-I_PDF-E/d48475a25698b1c0592e4cfe07138f2a_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:b4a5039452255e9de8d5ce831608ecb16f2abad06c45dd89f64349835d126d98 +size 109110 diff --git a/marked/P/T-REC-P.916-201603-I_PDF-E/d4c143a69ccd7e28fe8d01dbc9dfbcfa_img.jpg b/marked/P/T-REC-P.916-201603-I_PDF-E/d4c143a69ccd7e28fe8d01dbc9dfbcfa_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..f30c477baca54cdcdfe3609a862d5bd70ad460b7 --- /dev/null +++ b/marked/P/T-REC-P.916-201603-I_PDF-E/d4c143a69ccd7e28fe8d01dbc9dfbcfa_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:075ba6f1763a4f7e4da8f1905431ebac6ae73794e9d25e267afeb48973197805 +size 36635 diff --git a/marked/P/T-REC-P.916-201603-I_PDF-E/d5a837fa4f4675e5ee596003cf55985c_img.jpg b/marked/P/T-REC-P.916-201603-I_PDF-E/d5a837fa4f4675e5ee596003cf55985c_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..e80a6112d69b34aa4c0002fdb35e7473ca41da8a --- /dev/null +++ b/marked/P/T-REC-P.916-201603-I_PDF-E/d5a837fa4f4675e5ee596003cf55985c_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:3cb0091bd2b15b424ae7cddb0b21e44ac2710875c51b02bbf41bcc38ca145b23 +size 150531 diff --git a/marked/P/T-REC-P.916-201603-I_PDF-E/dd0f5301a5a6dd7c319701302110de88_img.jpg b/marked/P/T-REC-P.916-201603-I_PDF-E/dd0f5301a5a6dd7c319701302110de88_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..cc9665cc9f1aee2e72fb477544f1de667d9fa2a9 --- /dev/null +++ b/marked/P/T-REC-P.916-201603-I_PDF-E/dd0f5301a5a6dd7c319701302110de88_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:ad0e01b95ecbb981652389653b9ca8c8f3909180a60c9ce0102297d56809e4b3 +size 157680 diff --git a/marked/P/T-REC-P.916-201603-I_PDF-E/fa859e4e468bfb2710a94527f2c504af_img.jpg b/marked/P/T-REC-P.916-201603-I_PDF-E/fa859e4e468bfb2710a94527f2c504af_img.jpg new file mode 100644 index 0000000000000000000000000000000000000000..237bea09e76526f09cbef313e3bf752c25569691 --- /dev/null +++ b/marked/P/T-REC-P.916-201603-I_PDF-E/fa859e4e468bfb2710a94527f2c504af_img.jpg @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:23bb5de107ae229db1f78243f9fea28e4ec7c92ba2b8d5c9dd7c3d1106b5fe45 +size 59469 diff --git a/marked/P/T-REC-P.920-200005-I_PDF-E/raw.md b/marked/P/T-REC-P.920-200005-I_PDF-E/raw.md new file mode 100644 index 0000000000000000000000000000000000000000..a6cdac8ce93161eb9925a65f5ecd192b08de2cf0 --- /dev/null +++ b/marked/P/T-REC-P.920-200005-I_PDF-E/raw.md @@ -0,0 +1,465 @@ + + +![ITU logo](2dfa6ac3edfe874f68aa0cbccaa42322_img.jpg) + +The logo of the International Telecommunication Union (ITU) features the letters 'ITU' in a bold, sans-serif font, superimposed on a stylized globe with latitude and longitude lines. A lightning bolt symbol is positioned to the right of the globe. + +ITU logo + +INTERNATIONAL TELECOMMUNICATION UNION + +**ITU-T** + +**P.920** + +TELECOMMUNICATION +STANDARDIZATION SECTOR +OF ITU + +(05/2000) + +SERIES P: TELEPHONE TRANSMISSION QUALITY, +TELEPHONE INSTALLATIONS, LOCAL LINE +NETWORKS + +Audiovisual quality in multimedia services + +--- + +**Interactive test methods for audiovisual +communications** + +ITU-T Recommendation P.920 + +(Formerly CCITT Recommendation) + +--- + +# ITU-T P-SERIES RECOMMENDATIONS + +# TELEPHONE TRANSMISSION QUALITY, TELEPHONE INSTALLATIONS, LOCAL LINE NETWORKS + +| | | | +|-----------------------------------------------------------------------------------------------|---------------|---------------| +| Vocabulary and effects of transmission parameters on customer opinion of transmission quality | Series | P.10 | +| Subscribers' lines and sets | Series | P.30
P.300 | +| Transmission standards | Series | P.40 | +| Objective measuring apparatus | Series | P.50
P.500 | +| Objective electro-acoustical measurements | Series | P.60 | +| Measurements related to speech loudness | Series | P.70 | +| Methods for objective and subjective assessment of quality | Series | P.80
P.800 | +| Audiovisual quality in multimedia services | Series | P.900 | + +*For further details, please refer to the list of ITU-T Recommendations.* + +## **Interactive test methods for audiovisual communications** + +## **Summary** + +This ITU-T Recommendation is intended to define interactive evaluation methods for quantifying the impact of terminal and communication link performance on point-to-point or multipoint audiovisual communications. This methodology is based upon conversation opinion tests, and can be considered to be an extension of the methods defined in Annex A/P.800. + +###### **Source** + +ITU-T Recommendation P.920 was prepared by ITU-T Study Group 12 (1997-2000) and approved under the WTSC Resolution 1 procedure on 18 May 2000. + +## FOREWORD + +The International Telecommunication Union (ITU) is the United Nations specialized agency in the field of telecommunications. The ITU Telecommunication Standardization Sector (ITU-T) is a permanent organ of ITU. ITU-T is responsible for studying technical, operating and tariff questions and issuing Recommendations on them with a view to standardizing telecommunications on a worldwide basis. + +The World Telecommunication Standardization Conference (WTSC), which meets every four years, establishes the topics for study by the ITU-T study groups which, in turn, produce Recommendations on these topics. + +The approval of ITU-T Recommendations is covered by the procedure laid down in WTSC Resolution 1. + +In some areas of information technology which fall within ITU-T's purview, the necessary standards are prepared on a collaborative basis with ISO and IEC. + +## NOTE + +In this Recommendation, the expression "Administration" is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. + +## INTELLECTUAL PROPERTY RIGHTS + +ITU draws attention to the possibility that the practice or implementation of this Recommendation may involve the use of a claimed Intellectual Property Right. ITU takes no position concerning the evidence, validity or applicability of claimed Intellectual Property Rights, whether asserted by ITU members or others outside of the Recommendation development process. + +As of the date of approval of this Recommendation, ITU had not received notice of intellectual property, protected by patents, which may be required to implement this Recommendation. However, implementors are cautioned that this may not represent the latest information and are therefore strongly urged to consult the TSB patent database. + +© ITU 2001 + +All rights reserved. No part of this publication may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying and microfilm, without permission in writing from ITU. + +# CONTENTS + +| | Page | +|-----------------------------------------------------------------------------------------------------------------|------| +| 1 Scope..... | 1 | +| 2 References..... | 1 | +| 3 Experimental design ..... | 1 | +| 3.1 Basic approach and factors to be investigated ..... | 1 | +| 3.2 Stimuli used in conversational tests..... | 1 | +| 3.3 Test conditions and experimental design..... | 2 | +| 3.4 Subjects..... | 3 | +| 3.5 Subject training and reference connections ..... | 3 | +| 3.6 Ambient room and equipment characteristics..... | 3 | +| 4 Solicitation of opinions..... | 4 | +| Appendix I – Examples of tasks and stimuli for conversation..... | 5 | +| I.1 Stimuli for conversation..... | 5 | +| I.2 Tasks to evaluate the effects of speech delay on communication quality..... | 5 | +| I.3 Tasks to evaluate the effects of audiovisual delay and/or transmission errors on communication quality..... | 6 | +| I.4 Task to evaluate the synchronization between audio and video signals ..... | 6 | +| Appendix II – Protocols for the stimuli for conversation..... | 6 | +| II.1 Protocol for the Name-Guessing task ..... | 6 | +| II.2 Protocol for the Story-Comparison task ..... | 7 | +| II.3 Protocol for the Picture-Comparison task..... | 7 | +| II.4 Protocol for the building blocks task ..... | 7 | +| II.5 Protocol for the object-description task ..... | 8 | +| Appendix III – Test condition questionnaire..... | 9 | +| Appendix IV – Exit questions ..... | 10 | +| Appendix V – Bibliography ..... | 11 | + +# Introduction + +The audiovisual interactive test methods described in this ITU-T Recommendation are intended for quantifying the impact of terminal and communication link performance, that may affect the ability to conduct an interactive audiovisual communication. + +The efficacy of these tests strongly depends on the ability to reproduce in laboratory environments the conditions that are very close to the real situations. In this regard, particular care must be taken in choosing the tasks proposed to the subjects. In general, those tasks used in conversation tests for telephony assessment are not suited for audiovisual assessment because they often distract the subject's attention from the video screen. Therefore new tasks have been developed following the criteria illustrated in this ITU-T Recommendation. + +Substantial work has been done in this area, although all aspects of audiovisual quality are not yet completely understood. + +This ITU-T Recommendation reflects the current status of research on interactive audiovisual testing. + +As progress on this work continues, understanding of these interactive test methods will no doubt improve. As new knowledge is attained, this ITU-T Recommendation will be revised. + +## **Interactive test methods for audiovisual communications** + +# **1 Scope** + +This ITU-T Recommendation is intended to define interactive evaluation methods for quantifying the impact of coding artifacts, transmission delay and transmission impairments (e.g. packet loss, cell loss, digital channel errors) on point-to-point or multipoint audiovisual communications. This methodology is based upon conversation opinion tests, and can be considered to be an extension of methods defined in Annex A/P.800. + +This ITU-T Recommendation does not cover topics that are already described in other P-series ITU-T Recommendations such as the objective measurements of the link. + +# **2 References** + +The following ITU-T Recommendations and other references contain provisions which, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated were valid. All Recommendations and other references are subject to revision; all users of this Recommendation are therefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations and other references listed below. A list of the currently valid ITU-T Recommendations is regularly published. + +- ITU-T Recommendation P.800 (1996), *Methods for subjective determination of transmission quality*. +- ITU-R Recommendation BT.812 (1992), *Subjective assessment of the quality of alphanumeric and graphic pictures in Teletext and similar services*. +- ITU-T Recommendation P.911 (1998), *Subjective audiovisual quality assessment methods for multimedia applications*. +- IEC Publication 60651 (1979), *Sound level meters*. + +# **3 Experimental design** + +## **3.1 Basic approach and factors to be investigated** + +In order to quantify the impact of factors, such as transmission delay, that may affect the ability to conduct an interactive communication, the approach proposed in this ITU-T Recommendation is based on an active talker conversation assessment. Further, since it is necessary to express these opinions using a rating system, several single-stimulus rating scales are proposed. + +## **3.2 Stimuli used in conversational tests** + +In general, in conversational opinion tests it is desired to minimize the artificiality of the environment. However, at the same time, it is necessary to invoke some method to stimulate interactive communication utilizing the conditions which are being evaluated. In telephony assessments, it is common to use a set of photographs, or some other form of printed material, to achieve this objective. In audiovisual terminal performance assessments, however, such mechanisms are likely to distract a participant's attention from the video screen, thus possibly leading to an unnatural mode of communication for this type of terminal. + +For general applications, the following guidelines are provided for designing task-based tests: + +- the task should be designed such that, during their conversation, the subjects primarily maintain their attention on the audiovisual terminal; +- the task must have sufficient face value, that is, it must resemble real-life audiovisual communication to a sufficient degree. In particular, it is preferable that the task be performed by two subjects and not by one subject and an experimental leader; +- the task must yield reproducible quantitative results that represent adequate measures of communication efficiency. When time delays are involved, time measures should be among the results. + +A wide range of subjects, including elderly and hearing-impaired subjects, should be able to perform the task. + +It is preferable that the task is, in itself, sufficiently rewarding for the subjects. This has several advantages: the subjects learn the task faster and they are less susceptible to fatigue and loss of motivation. + +From past experiments, it has been found that lively audiovisual conversations can be stimulated if the participants in such a test know each other. Subsequently, the provision of written material can be used as a secondary, rather than primary, source of stimulation. Thus, unlike telephony, familiarity between pairs of conversing participants is highly desirable, if not essential. + +It is recognized, however, that for specific applications, the conversational tasks may have to be modified to take into account the services that the system under test is intended to provide. In order to permit meaningful measurements to be made of the factors being investigated, it is recommended that in such cases the conversational tasks be structured so as to represent the applications of interest, particularly as regards: + +- a) the rate of information exchange; and +- b) the degree of audio and video signal utilization. + +For example, to account for the attributes in the first category, tasks could range from predominately one-way communication, to free-conversation, to a rapid exchange of information, be it via video, audio or both signals. Similarly, to test attributes in the second category, tasks could range from the subjects working on a hard-copy document in front of them (minimum use of video information) to reading sign language over the video link (maximum use of video information). The actual tasks should combine attributes from both categories. These guidelines have been applied to develop the tasks illustrated in Appendix I, and the protocols for the tasks are detailed in Appendix II. + +## **3.3 Test conditions and experimental design** + +In general, at least one transmission impairment factor or test condition is likely to be evaluated in a test, in addition to a baseline (reference) condition where the impact of such factor is minimum (when using the reference condition, this should not be identified as such to the participants). However, because conversational tests are time-consuming, the total number of conditions ought to be reasonably constrained in order to minimize participant fatigue and maximize experimental accuracy. This requirement should be balanced against the need to ensure that the duration of each conversation/condition is at least five minutes long. + +As with conversational tests (audio communications), a Latin or Greco-Latin square may be found to be a suitable experimental design for this purpose. In such case, the square's rows may be associated with the test participants and the square's columns with the order in which the conditions in the test are being presented. + +Other treatments may also be appropriate depending on the factors being investigated. For example, past experiments have appeared to indicate that there may be an interaction between the audiovisual communication path quality and the perception of the impact of transmission delay. Consequently, it + +may be preferable to apply two treatments using a Latin-Latin square design, so that the letters of the first alphabet are associated with different values of transmission delay and the letters of the second alphabet are associated with different image/voice coding rates. + +Of course, other experimental designs including replicated block designs and Youden square designs may be suitable and could be left up to the experimenter to select in order to meet specific cost and accuracy objectives in view of the number of conditions of interest. + +Also any possible effects related to the order in which the tasks are performed must be taken into account. + +## **3.4 Subjects** + +At least 16 subjects should participate in a test, the exact number will be dictated by the experimental design and the accuracy required to the results. These subjects should be non-expert, and they should not be directly involved with either audio and/or video technology as part of their normal work. + +Nevertheless, in the early phases of the development of audiovisual communications systems and in pilot experiments carried out before a larger test, small groups of experts (4-8) or other critical subjects can provide indicative results with sufficient reliability. + +## **3.5 Subject training and reference connections** + +Before starting the experiment, a scenario of the intended application of the system under test should be given to the subjects. The range and type of impairments should be shown in a preliminary phase. During this phase, a first level of personal introduction may thus be allowed to take place over the communication link at the worst (or best) experimental condition, while further discussion pertinent to the tasks expected of the participants can be subsequently permitted at the best (or worst) experimental condition. + +Again, as with the main test, the particulars of the conditions should not be revealed to the test participants. + +## **3.6 Ambient room and equipment characteristics** + +Table 1 lists typical viewing and listening conditions as used in audiovisual quality assessment. The actual parameter settings used in the assessment should be specified. For the comparison of test results, all viewing and listening conditions must be fixed and equal over laboratories for the same kind of tests. + +Both the size and the type of monitor used should be appropriate for the application under investigation. When sequences are presented through a PC-based system, the characteristics of the display and audio transducers must be specified, e.g. dot pitch and of the monitor, type of video display card used, characteristics of either handsets, headphones or loudspeakers, etc. + +In particular, in case of loudspeaker presentation, the number and positions of the loudspeakers relative to the image should be reported. + +Operational parameters, such as signal level, for the test sequences shall match those of the alignment signal used to verify the viewing and listening conditions. Any operational adjustments performed so that source or processed sequences meet this requirement should be reported. + +Synchronization between audio and video signal should be measured according to ITU-T Recommendation P.911 and reported. + +**Table 1/P.920 – Typical viewing and listening conditions +as used in audiovisual quality assessment** + +| Parameter | Setting | +|---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|----------------------------------| +| Room size (Note 7) | Specify $L \times W \times H$ | +| Viewing distance (Note 5) | 1-8 H | +| Peak luminance of the screen | 100-200 $\text{cd/m}^2$ | +| Ratio of luminance of inactive screen to peak luminance | $\leq 0.05$ | +| Ratio of the luminance of the screen, when displaying only black level in a completely dark room, to that corresponding to peak white | $\leq 0.1$ | +| Ratio of luminance of background behind picture monitor to peak luminance of picture (Note 1) | $\leq 0.2$ | +| Chromaticity of background (Note 6) | D 65 | +| Background room illumination (Note 1) | $\leq 20$ lux | +| Background noise level (Note 2) | $\leq 30$ dBA | +| Listening level (Note 3) | $\sim 80$ dBA | +| Reverberation time (Note 4) | $< 500$ ms, $\forall f > 150$ Hz | +|

NOTE 1 – This value indicates a setting allowing maximum detectability of distortions, for some applications higher values are allowed or they are determined by the application.

NOTE 2 – If background noise levels used in the application are significantly higher, Hoth noise should be used for environments like office. The Hoth noise should be generated in a room with low background noise levels (\leq 30 dBA) and be measured in the acoustical domain. If for some special type of application another type of background noise is used, the spectral power density and dBA level should be specified.

NOTE 3 – This value indicates a setting allowing maximum detectability of distortions, for some applications lower or higher values are allowed. Level setting is measured using the maximum value over the audio sequence using IEC Publication 60651 fast averaging. When loudspeakers are used, the sound level might be adjusted according to the viewing distance.

NOTE 4 – This value is only relevant for loudspeaker presentation, larger reverberation times generally lead to a decrease in detectability of distortions.

NOTE 5 – For a given screen height, it is likely that the viewing distance preferred by the subjects increases when visual quality is degraded. Concerning this point, the preferred viewing distance should be predetermined for qualification tests. Viewing distance in general depends on the applications. Viewing distance should be defined taking into account not only the screen size, but also type of applications, type of screen and the goal of the experiment.

NOTE 6 – For PC monitors, the chromaticity of background may be adapted to the chromaticity of the monitor.

NOTE 7 – Room size is important only for loudspeaker presentation.

| | + +# 4 Solicitation of opinions + +As with telephone conversational assessments, each participant should also be separately solicited for his, or her, opinion after the completion of each condition. It is preferable to structure in advance the type of questioning and minimize the number of questions posed after each condition and to minimize uncontrolled variations. A possible test condition questionnaire that could be used for this purpose can be found in Appendix III. + +Several category judgement scales can be used to evaluate the audiovisual terminal performance. The sensitivity of these scales to various transmission impairment factors may be different. + +Examples of scales that may be used for this purpose are: + +- Overall Audiovisual Quality, Video Quality, and Audio Quality are generally assessed using the categories: Excellent, Good, Fair, Poor, or Bad. +- Effort Needed to Interrupt is generally assessed using the categories: No Effort, Minor Effort, Moderate Effort, Considerable Effort, or Extreme Effort. +- Communications Difficulty and Acceptability of Communication are generally assessed using a binary choice: Yes or No. + +Although selection of any particular scale may be subject to the goals of an individual experiment, it is important that these scales (and their associated wording, or translation) be used consistently by different laboratories. + +If the use of audiovisual terminals is novel for most participants, it is recommended that "Exit Questions" are presented after the last test condition has been rated. Such questions should attempt to capture any other factors that may have been inadvertently omitted from the experiment. A sample of such questions is given in Appendix IV. + +# APPENDIX I + +## **Examples of tasks and stimuli for conversation** + +### **I.1 Stimuli for conversation** + +The following tasks differ from each other in the degree to which free conversation can occur. The protocols for the tasks are described in more detail in Appendix II. + +Task 1: The Name-Guessing task. The name-guessing task is a question-answer game performed according to a fixed protocol. This results in a very restricted conversation. + +Task 2: The Story-Comparison task. In the story-comparison task, subjects have to discover a number of differences between two versions of a story. They are allowed unrestricted conversation. Prior to the test, both subjects have to read and memorize a short story. + +Task 3: The Picture-Comparison task. In this task, the subjects have to memorize a picture and subsequently determine whether they were given identical or different pictures. Conversation is not restricted. + +### **I.2 Tasks to evaluate the effects of speech delay on communication quality** + +In the following tasks the talk spurt increases from task 1) to task 6), whereas the conversation switching rate decreases. + +- 1) take turns in counting; +- 2) take turns reading random numbers aloud as quickly as possible; +- 3) take turns verifying random numbers aloud as quickly as possible; +- 4) words with missing letters are completed with letters supplied by the other talker; +- 5) take turns verifying city names as quickly as possible; +- 6) determine the shape of a figure described verbally; +- 7) free conversation. + +The previous tasks [with exception of task 1) and task 7)] cannot be used for audiovisual quality evaluations because most of them require the subjects to concentrate their attention on a sheet of paper and not on the screen. + +### **I.3 Tasks to evaluate the effects of audiovisual delay and/or transmission errors on communication quality** + +The following tasks should draw the attention of the assessors to the video signal. + +The protocols for the tasks 1) and 2) are described in more detail in Appendix II. + +- 1) one of the subjects shows and describes a plastic building block and the other one is required to reproduce it; +- 2) each subject in a pair is given a detailed object to describe to their partner; +- 3) one of the two subjects shows some exercises for the alleviation of neck pain and the other one is required to reproduce them. + +### **I.4 Task to evaluate the synchronization between audio and video signals** + +The following task is intended for drawing the attention of the assessors to the synchronization between audio and video signals: + +- One person claps while the other checks the synchronization between the movement and the sound. + +The previous task is not suitable for assessing lip synchronization. + +Additional tasks aimed at evaluating the synchronization between the two signals are for further studies. + +# **APPENDIX II** + +## **Protocols for the stimuli for conversation** + +### **II.1 Protocol for the Name-Guessing task** + +The following protocol for the Name-Guessing task can be used: + +"The first task, the name-guessing task, is a question-answer game performed according to a fixed protocol. In this task, one of the subjects receives three pieces of information: first, either the word "Brand" or the word "Person", indicating whether it is a brand name or a name of a (well-known) person that should be guessed; second, in the case of a brand, a description of the product, and in the case of a person, his or her profession; third, the name to be guessed. Thus, the subject would, for example, be presented with the text: + +- Brand; +- Cigarettes; +- Camel." + +"In guessing the name of the brand or person, the second subject has to adhere to the following protocol: the first question should be "Is it a brand or a person?"; the second, in the case of a brand, "What is the product?", and, in the case of a person "What is the profession?"; subsequently, one guess can be made; when the guess is false, consecutive letters of the name should be asked for and a guess can each time be made; this continues until the name is guessed correctly or until the entire name has been spelled out. The subjects being interrogated are allowed to consult their text during the conversation. Most subjects do not find this necessary, however, because they had no trouble retaining the limited amount of information contained in it. + +After the name is guessed, the total time required and the number of letters that are requested are scored. Several names, varying in length and difficulty, should be guessed in a given experimental condition. A linear regression analysis can then be applied to the data so that estimates are obtained of the time required for a direct guess (zero letters suggested) and the extra time per letter." + +### **II.2 Protocol for the Story-Comparison task** + +The following protocol for the Story-Comparison task can be used: + +"In the second task, subjects have to discover a number of differences between two versions of a story. They are allowed unrestricted conversation. Prior to the actual test, both subjects have to read and memorize a short story. They are given stories that are essentially the same but that contain a certain number of small but distinct differences. (For example stories of approximately 200 words, containing six differences.) In addition, they receive a list of questions about the story, which they have to answer for themselves. The questions are intended to improve retention of the story. After the memorization period, the subjects have to start a conversation with the aim of discovering as quickly as possible all differences between the two stories. They are not allowed to consult the text during the conversation. The subjects know how many differences there are and they get feedback informing them when they have detected a difference and how many differences they have detected thus far. The conversation continues until all differences are detected or until no differences have been detected for a certain time." + +"The starting time of the conversation and the times at which detection of a difference occurs are registered. In the analysis, the time interval between the start and the detection of the first difference, and the intervals between subsequent detections are determined. These intervals are then ordered according to their duration. This is done because the time interval, i.e. the time required for the detection of the next difference, depends both on the strategy used by the subject pair and on where the differences are located in the text. Reordering the intervals reduces the variance introduced by these factors. In addition, it allows evaluation of the effect being investigated as a function of the duration of the interval." + +### **II.3 Protocol for the Picture-Comparison task** + +The following protocol for the Picture-Comparison task can be used: + +"In this task, the subjects have to memorize a picture and subsequently determine whether they were given identical or different pictures. (Pictures of various subjects, such as landscapes, buildings and urban sites, can be used.) Conversation is not restricted. The subjects are not allowed to look at the picture during the conversation." + +The results that are scored are the total time required by the subjects to reach a decision and whether or not the decision is correct. As subjects normally need more time when the pictures are identical than when they are different, the results for both cases have to be separated. Though this task is somewhat similar to the story-comparison task, there is a difference in the degree to which free conversation will occur. This is because, in comparing stories, the subjects tend to borrow phrases from the text, known to both of them, whereas the task of comparing pictures forces them to use their own wordings." + +### **II.4 Protocol for the building blocks task** + +The following protocol for the building blocks task can be used: + +"One subject was given a bag of multicolour, interlocking construction blocks. The other subject was given a completed figure made from an identical set of blocks. The first subject's goal was to build the same figure with help from the other subject, and verify its correctness." + +The blocks chosen were designed for pre-school children (5 years old or less), with a minimum size of 3cm × 3cm × 2cm. Some example figures were birds, trucks, and arches, typically composed of ten or fewer pieces. Completed figures were easily held in one hand, leaving the other hand free to build or point out features. The resolution of the video system was sufficient to see the shape of the figures, but the borders between blocks of the same colour were not easily visible. Again, use of self- + +view was prohibited. The verification suggested in the instructions required both subjects to rely on the video channel, as well as the audio. + +The instructions to subjects had the same features as those of the "Describe an Object" task. Examples follow: + +#### **BUILD an OBJECT from DESCRIPTION – INSTRUCTIONS:** + +Your partner has been given an object to describe to you. You must build a similar object from loose blocks and your partner's instructions. Your partner will display the object to you and point out its features to help you build it. + +When you have completed the call, please answer the following questions. + +#### **DESCRIBE AN OBJECT TO BE BUILT – INSTRUCTIONS:** + +You have been given an object to describe to your test partner. Your partner must build a similar object from loose blocks and your instructions. You should hold the object in front of you and point out its features to help your partner. Confirm that your partner understands the description of each feature. + +When you have completed the call, please answer the following questions." + +### **II.5 Protocol for the object-description task** + +The following protocol for the object-description task can be used: + +"Each subject in a pair was given a detailed object to describe to their partner. Partners changed between the active listening role, and the describing role at the halfway point of the call. The touch screens in the boothcarrels gave instructions to place calls, and indicated the halfway point and end of the call with audible indicators and messages on the screen. + +The objects easily fit in one hand, so that the subject in the describing role could hold the object and point to a specific feature while narrating. The objects chosen had large features with different colours that could easily be seen on compressed video, as well as detailed features that the video system could not reproduce. Also, there was no self-view available to the subjects, so the listener would often need to direct the describer to hold the object higher, closer, etc., to improve their view. This combination of features helped to balance the reliance on the video and audio channels to convey the descriptions, and maintained good interactions throughout the call. + +The instructions to subjects were organised as follows. The subjects received their object for each call in a plastic bag. A card attached to the bag gave a short label for the task, and detailed instructions. After conducting the task a few times with different objects, the test subjects only needed to refer to the task label to understand their roles (but instructions were always included). + +#### **DESCRIBE FIRST, THEN LISTEN – INSTRUCTIONS:** + +You have been given an object to describe to your test partner. You may hold it in front of you and point out its features as you describe them. Confirm that your partner understands the description of each feature. + +Next, your test partner will describe a different object. You may ask questions about the object during the description process. + +When you have completed the exchange, please answer the following questions. + +#### **LISTEN FIRST, THEN DESCRIBE – INSTRUCTIONS:** + +First, your test partner will describe an object. You may ask questions about the object during the description process. + +Next, you have been given a different object to describe to your test partner. You may hold it in front of you and point out its features as you describe them. Confirm that your partner understands the description of each feature. + +When you have completed the exchange, please answer the following questions." + +# APPENDIX III + +## Test condition questionnaire + +Questionnaire to be used after each test condition: + +Administrator: I will ask you a few questions (again) relative to your opinion of the connection over which you just conversed. Are you ready? + +- 1) How would you rate the overall audiovisual quality? (Circle one) +Excellent      Good              Fair              Poor              Bad +- 2) How would you rate the video quality of the connection? (Circle one) +Excellent      Good              Fair              Poor              Bad +- 3) How would you rate the audio quality of the connection? (Circle one) +Excellent      Good              Fair              Poor              Bad +- 4) How would you judge the effort needed to interrupt the other party (or parties)? (Circle one) +No effort      Minor effort      Moderate effort      Considerable effort      Extreme effort +- 5) Did you have any difficulty during the connection? (Circle one) +                                                            Yes                      No +- 6) Was the connection acceptable? (Circle one) +                                                            Yes                      No + +Administrator: Thank you for your answers. We will now go ahead and re-establish our connection in a few minutes so that we can continue the experiment. As soon as the connection is set up, I shall let you know that you can go ahead and converse with the other party. + +# APPENDIX IV + +## **Exit questions** + +Additional questions to be used at the end of the test: + +Administrator: Since we are at the end of the test, I would like to ask you some supplementary questions. Are you ready? + +Could you tell me, in order of importance, which of the following factors would you consider sufficiently important that you would like to see improved. (Evaluator will read a number of factors here.) + +7a\_\_\_\_\_ 7b\_\_\_\_\_ 7c\_\_\_\_\_ 7d\_\_\_\_\_ + +Do you have any other comments relative to the entire test you wish to make known to us? + +--- + +--- + +--- + +--- + +--- + +Thank you for your cooperation in taking part in the test. I shall be over in a few minutes to wrap up the session and let you out of the building. + +# APPENDIX V + +## **Bibliography** + +- [1] KIRK (R.E.): Experimental Design, *Procedures for the Behavioural Sciences*, 2nd Edition, Brooks/Cole Publishing Co., California 1982. +- [2] BRONKHORST (A.W.) (J.A.) VERHAVE: The effect of audio-video desynchronization on communication efficiency in videotelephony, *Report IZF 1992 C-35* Study for PTT Telecom Netherlands, TNO Institute for Perception 1992. +- [3] HOTH (D.F.): Room noise spectra at subscribers' telephone locations, *J.A.S.A.*, Vol. 12, pp. 99-504, April 1941. + +## SERIES OF ITU-T RECOMMENDATIONS + +| | | +|-----------------|--------------------------------------------------------------------------------------------------------------------------------| +| Series A | Organization of the work of ITU-T | +| Series B | Means of expression: definitions, symbols, classification | +| Series C | General telecommunication statistics | +| Series D | General tariff principles | +| Series E | Overall network operation, telephone service, service operation and human factors | +| Series F | Non-telephone telecommunication services | +| Series G | Transmission systems and media, digital systems and networks | +| Series H | Audiovisual and multimedia systems | +| Series I | Integrated services digital network | +| Series J | Transmission of television, sound programme and other multimedia signals | +| Series K | Protection against interference | +| Series L | Construction, installation and protection of cables and other elements of outside plant | +| Series M | TMN and network maintenance: international transmission systems, telephone circuits, telegraphy, facsimile and leased circuits | +| Series N | Maintenance: international sound programme and television transmission circuits | +| Series O | Specifications of measuring equipment | +| Series P | Telephone transmission quality, telephone installations, local line networks | +| Series Q | Switching and signalling | +| Series R | Telegraph transmission | +| Series S | Telegraph services terminal equipment | +| Series T | Terminals for telematic services | +| Series U | Telegraph switching | +| Series V | Data communication over the telephone network | +| Series X | Data networks and open system communications | +| Series Y | Global information infrastructure and Internet protocol aspects | +| Series Z | Languages and general software aspects for telecommunication systems | \ No newline at end of file