ITU-T P 810-1996 Modulated Noise Reference Unit (MNRU) - Telephone Transmission Quality Methods for Objective and Subjective Assessment of Quality (Study Group 12) 14 pp《调制噪声参考单元(M.pdf

1、INTERNATIONAL TELECOMMUNICATION UNION ITU-T TELECOMMUNICATION STANDARDIZATION SECTOR OF ITU P.81 O TELEPHONE TRANSMISSION QUALITY METHODS FOR OBJECTIVE AND SUBJECTIVE ASSESSMENT OF QUALITY (02/96) MODULATED NOISE REFERENCE UNIT (MNRU) ITU-T Recommendation P.81 O (Previously “CCITT Recommendation”) I

2、TU-T RECNN*P-BLO 96 48bS-591 ObLb44B 29T 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 Recommen- dations on them with a

3、 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 IT-T Study Groups which, in their turn, produce Recommendations on these topics. The approval of Recomm

4、endations 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.810 was revised by ITU-T Study Group 12 (1993-1996) and was approved under the WTSC Resolution No. 1 procedure on the 6th of February 1996. NOTE I

5、n this Recommendation, the expression “Administration” is used for conciseness to indicate both a telecommunication administration and a recognized operating agency. O I 1996 Ali rights reserved. No part of this publication may be reproduced or utilized in any form or by any means, electronic or mec

6、hanical, including photocopying and microfilm, without permission in writing from the ITU. ITU-T RECMN*P.BLO 96 YB62591 Ob16449 126 CONTENTS 1 Scope 2 References 3 Definitions 4 Abbreviations . 5 Conventions 6 General description 7 Performance specifications 7.1 General 7.2 Signal path 7.4 Combined

7、path 7.3 Noise path . Bibliography Recommendation P.810 (02/96) Page 1 2 2 2 2 3 3 3 4 8 9 9 1 - ITU-T RECMN*P=BLO 96 4862573 0636450 748 SUMMARY This Recommendation describes the Modulated Noise Reference Unit (MNRU). a standalone unit for introducing controlled degradations to speech signals. As s

8、uch, the MNRU has been used extensively in subjective performance evaluations of digital processes, both in conventional telephone bandwidth and in wideband (e.g. 70-7000 Hz) applications. Historically, the MNRU has been implemented in analogue hardware. The revisions encompassed in this version of

9、the Recommendation are the inclusion of descriptions of digital implementations of the MNRU. These descriptions are suitable for implementation in software or on digital hardware. One further revision is to note the need for a high-pass filter (removal of any DC component of the input speech materia

10、l) for all implementations. Existing analogue hardware implementations of the MNRU will continue to meet the specifications in this Recommendation, provided such filtering is applied externally. INTRODUCTION The MNRU was originally devised to produce distortion subjectively similar to that produced

11、by logarithmically companded PCM systems. This approach was based on the views: that network planning would require extensive subjective tests to enable evaluation of PCM system performance over a range of compandor characteristics, at various signal levels and in combination with various other tran

12、smission impairments (e.g. loss, idle circuit noise, etc.) at various levels; and that it would be as reliable and easier to define a reference distortion system, itself providing distortion perceptually similar to that of PCM systems, in terms of which the performance of PCM systems could be expres

13、sed. This requires extensive subjective evaluation of the reference system when inserted in one or more simulated telephone connections, but leads to the possibility of simplified subjective evaluation of new digital processing techniques. 1) 2) KEYWORDS Analogue MNRU, controlled degradation, digita

14、l MNRU, Modulated Noise Reference Unit (MNRU), subjective performance evaluation. ii Recommendation P.810 (OU96) ITU-T RECNN*P.BLO 96 4862593 0616453 884 Recommendation P.810) MODULATED NOISE REFERENCE UNIT (MNRU) (Malaga-Torremolinos, 1984; amended Melbourne, 1988; Helsinki, 1993; revised in 1996)

15、1 Scope The Modulated Noise Reference Unit (MNRU) is a standalone unit that is intended to introduce controlled degradations to speech signals. Various organizations (Administrations, scientific/industrial organizations), as well as the IT-T itself, have made extensive use of the MNRU concept for ev

16、aluating the subjective performance of digital processes (in arriving at Recommendations G.722 and the 32 kbit/s ADPCM algorithm of Recommendation G.726, for example). A modified version for use in evaluating codecs of wider bandwidth (70-7000 Hz) is now common practice. However, the actual devices

17、used, while based on common principles, may have differed in detail, and hence the subjective results obtained may also have differed. (Differences in subjective testing methodology are also relevant.) The purpose of this Recommendation is to define the narrow-band and wideband versions of the MNRU

18、as completely and in as much detail as possible in order to minimize the effects of the device, and of its objective calibration procedures, on the results of subjective tests. The need for a device, implemented in both hardware and software, that introduces controlled degradations to speech signals

19、 is exemplified by a number of factors: the use of digital processes (64 kbiils PCM A-law or p-law, AIDIA encoder pairs, A-lp-law or p-IA-law converters, digital pads based on 8-bit PCM words, 32 kbit/s ADPCM, etc.) in the international telephone network is now widespread; new digital processes are

20、being standardized, e.g. 64 kbit/s 7 kHz wideband ADPCM; there is a need for standard tools to measure the quantization distortion performance of digital processes for example, 32 kbit/s ADPCM (Recommendation G.726) and 64 kbit/s 7kHz wideband codec (Recommendation G.722), so that the tools can be u

21、sed for estimating the subjective transmission performance of international connections containing digital processes; an objective speech quality assessment method has not yet been established; at the present time, subjective tests incorporating reference system conditions represent the only suitabl

22、e method for measuring the speech transmission performance of digital processes; expressing results in terms of a common reference system may facilitate comparison of subjective test results obtained at different laboratories. It is recommended: 1) 2) NOTES 1 that a narrow-band MNRU be used as the r

23、eference system in terms of which subjective performance of telephone bandwidth digital processes are expressed; that a wideband MNRU be used as the reference system in terms of which subjective performance of wideband digital processes are expressed. The MNRU was originally constructed using analog

24、ue circuitry. Currently, however, it tends to be either constructed using digital circuitry or simulated on computers. Further information on the effects of the MNRU parameters is given in the bibliographic sources listed at the end of this Recommendation. The listening-only method presently propose

25、d when using the MNRU in subjective tests is described in Recommendation P.830. See 2.2P.80 for precautions concerning the use of listening-only tests. 2 1) Previously p.81. Recommendation PA10 (02/96) 1 ITU-T RECNN*P-BLO 96 4862593 0616452 710 3 Objective measurement methods that suitably reflect s

26、ubjective quantization distortion performance of various types of digital processes do not exist at present. (For example, the objective techniques of Recommendation G.712, based on sine wave and band-limited noise measurements, are designed for PCM and do not measure appropriately the distortion in

27、duced by other systems such as ADPCM.) The artificial voice described in Recommendation P.50 may be relevant. Even if an objective method is developed, subjective tests will be required to establish correlation of subjective resultdobjective results for particular digital process types. 2 References

28、 The following Recommendations contain provisions that, through reference in this text, constitute provisions of this Recommendation. At the time of publication, the editions indicated are valid. All Recommendations and other references are subject to revision; all users of this Recommendation are t

29、herefore encouraged to investigate the possibility of applying the most recent edition of the Recommendations listed below. A list of the currently valid ITU-T Recommendations is regularly published. - - - - - ITU-T Recommendation G. 191 (1993), Sofhvare tools for speech and audio coding standardiza

30、tion. ITU-T Recommendation P.50 (1993), Artificial voices. ITU-T Recommendation P.56 (1993), Objective measurement of active speech level. ITU-T Recommendation P.80 (1 993), Methods for subjective determination of transmission quality. ITU-T Recommendation P.830 (1996), Subjective perfarmance assess

31、ment of telephone-band and wideband digital codees. 3 Definitions For the purposes of this Recommendation, the following definitions apply: 3.1 dBov: dE3 relative to the overload of a digital system. 3.2 signai path: The path through the Modulated Noise Reference Unit that includes only the input si

32、gnal. 3.3 3.4 noise path: The path through the Modulated Noise Reference Unit that generates the modulated noise. combined path: The combined signal and noise paths through the Modulated Noise Reference Unit. 3.5 3.6 3.7 Q: The ratio, in dB, of speech power to modulated noise power. QN: Q for a narr

33、ow-band Modulated Noise Reference Unit. Qw: Q for a wideband Modulated Noise Reference Unit. 4 Ab brevia tions For the purposes of this Recommendation, the following abbreviations are used: ADPCM MNRU Modulated Noise Reference Unit PCM Pulse Code Modulation RMS Root Mean Square SNR Signal-to-Noise R

34、atio Adaptive Differential Pulse Code Modulation 5 Conventions Implementations of the MNRU in analogue hardware and on digital hardware or in software are described in this Recommendation. Throughout this Recommendation, references will be made to analogue hardware and to digital implementations. Wh

35、erever the digital implementation is mentioned, it should be understood that the description is applicable either to software or to digital hardware. 2 Recommendation P.810 (OU96) ITU-T RECMN*P-BLO 76 4862571 Ob36453 657 = 6 General description Simplified arrangements of the MNRU are shown in Figure

36、 1 a) for the analogue narrow-band version, Figure 1 b) for the analogue wideband version, Figure 2 a) for the digital narrow-band version, and Figure 2 b) for the digital wideband version. Speech signals entering from the left are split between two paths, a signal path and a noise path. The signal

37、path provides an undistorted (except for bandpass filtering) speech signal at the output. In the noise path, the speech signal instantaneously controls a multiplier with an applied gaussian noise “carrier“ that has a uniform spectrum between the cutoff frequencies shown for the noise source. The out

38、put of the multiplier, consisting of the noise modulated by the speech signal, is then added to the speech signal to produce the distorted signal. The attenuators and switches in the signal and noise paths allow independent adjustment of the speech and noise signal levels at the output. Typically, t

39、he system is so calibrated that the setting of the attenuator (in dB) in the noise path represents the ratio of instantaneous speech power to noise power, when both are measured at the output of the bandpass filter (terminal OT). Specifically, when both are set to O dB, the noise level measured at t

40、erminal OT, with separate resistive terminations at the terminals T1 and T2 (unlinked), should be the same as the speech level measured at terminal OT, with separate resistive terminations at the terminals T5 and T6 (unlinked): as a check on this it should be established that the ratio of speech-plu

41、s-modulated-noise power (measured at terminal OT) to the power of the input speech (measured at terminal IT) is 3 dB (see 7.3.1). For this Recommendation, the decibel representation of the ratio is called QN for the narrow-band version and Qw for the wideband version. Digital implementations of the

42、MNRU will be facilitated by noting that, when the signal path has unity gain, output speech-plus-modulated-noise generation can be expressed by: where x(i) is the input speech, N(i) is the random noise, Q is the ratio of speech power to modulated noise power (determined by the noise path gain), and

43、y(i) is the output speech-plus-modulated-noise. 7 Performance specifications 7.1 General The specifications in this subclause apply both to analogue hardware implementations and to digital implementations. Analogue MNRU hardware and digital implementations of the MNRU are described in parallel. Exis

44、ting analogue hardware meeting the specifications in this Recommendation and digital implementations of the MNRU were tested extensively by Study Group 12 during the 1989-1992 study period. These implementations were found to be equivalent in their subjective effects. As the MNFW is intended only as

45、 a reference degradation for subjective evaluation of digital processes, the experts in Study Group 12 have deemed subjective equivalence of MNRU implementations to be of paramount importance. NOTE - The implementations of the wideband MNRU described here were not tested. Comments on the wideband MN

46、RU found in clause 1, above, are noted here. For practical implementations, the actual signal levels and noise levels may be increased or decreased to meet special needs. In such cases, the level requirements detailed below will have to be modified accordingly. In particular, digital implementations

47、 will typically require speech material with a RMS level of -26.15 dBov (i.e. dB relative to the overload point of a digital system). However, if the language used in subjective testing has a peak-to-average level greater than 23 dB, the RMS level of input speech should be reduced appropriately. It

48、can be shown that a DC component in the input signal will generate an additive noise component on the output signal. This additive component of noise is not accounted for when one specifies a Q value. Thus, rather than having only a multiplicative noise at the output with SNR of Q, the output will h

49、ave another additive noise component, this component increasing the overall noise in the output, and decreasing the total signal to (additive-plus-multiplicative) noise ratio. Thus, MNRU implementations should include (or provide external) high-pass filtering at the input. Recommendation P.810 (02/96) 3 ITU-T RECMN*P-BLO 96 4862573 Ob36454 593 TL-W- + DC Removal Filter Input OT 0 outpu Filter 100-3400Hz -u- IT Input -+ DC Removal 0-20 IHz noise source OT c) Output - Filter a) Basic arrangement of narrowband analogue MNRU f-l Attenuator or amplifier T1207130-95/601 U 0