Method and apparatus for objective speech quality measurements of telecommunication equipment

• US 5,621,854 A
• Filed: 12/12/1994
• Issued: 04/15/1997
• Est. Priority Date: 06/24/1992
• Status: Expired due to Term

• Current Assignee: Psytechnics Ltd. (NetScout Systems Incorporated)
• Original Assignee: British Telecommunications PLC (BT Group PLC)
• Inventors: Hollier, Michael P.
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): Smits, Talivaldis Ivars
• Application Number:
  US08/351,421
• Time in Patent Office: 855 Days
• Field of Search: 395/2.37, 395/2.42, 381/46, 381/47, 381/73.1
• US Class Current: 704/200.1
• CPC Class Codes: G10L 25/69 for evaluating synthetic or decoded voice signals
• CPC Class Codes: H04B 1/665 using psychoacoustic properties of the ear, e.g. masking effect
• CPC Class Codes: H04M 3/22 Arrangements for supervision, monitoring or testing
• CPC Class Codes: H04M 3/2236 Quality of speech transmission monitoring

Abstract

A telecommunications testing apparatus comprising a signal generator which generates a speech-like synthetic signal, which is supplied to the input of a telecommunication apparatus to be tested. The distorted output of the telecommunications apparatus is supplied to an analyzer, which derives, for both the undistorted test signal and the distorted signal from the telecommunications apparatus, a measure of the excitation of the human auditory system generated by both signals, taking into account both spectral masking and temporal masking phenomena. The difference between the two excitations is then calculated, and a measure of the loudness of the difference is derived which is found to indicate to a high degree of accuracy the human subjective response to the distortion introduced by the telecommunications system.

Claims

1. Test apparatus comprising:

a signal generator for supplying a test signal which has a spectral resemblance to human speech but which corresponds to more than one speaker, and

analysis means for receiving a distorted signal which corresponds to said test signal when distorted by equipment to be tested, and for analyzing said distorted signal to generate a distortion perception measure which indicates the extent to which the distortion of said signal will be perceptible to a human listener,

said analysis means including means to split the signal into a plurality of spectral bands having bandwidths which vary with frequency.

2. Apparatus as in claim 1 in which the analysis means includes means to filter the said test signal and/or distorted signal to generate a plurality of spectral component signals.

3. Apparatus as in claim 2, in which the analysis means is arranged to estimate, for each spectral component signal, the masking effect which that spectral component signal would produce on the human ear.

4. Apparatus as in claim 3 comprising means for generating a time sequence of successive spectral component values representative of component signal levels over a time interval, the time intervals being longer for lower frequency spectral component signals than for higher frequency spectral component signals.

5. Apparatus as in claim 1, in which the component signal bandwidths are, on a logarithmic frequency scale, approximately equal.

6. Apparatus as in claim 1 in which the component signal bandwidths are, on a Bark scale, approximately equal.

7. Apparatus as in claim 1, in which said analysis means is arranged to estimate the effect which said distortion would produce on the human ear taking into account the temporal persistence of said effect.

8. Apparatus as in claim 1, in which the analysis means is arranged to generate a time sequence of successive processed signal segments from said test signal and/or said distorted signal, the value of at least some signal segments being generated in dependence upon portion of said test signal and/or distorted signal which precede and/or succeed said signal segments.

9. Apparatus as in claim 1, in which the analysis means is arranged:

to decompose the distorted signal into a plurality of spectral component bands, the spectral component bands having bandwidths approximately equally spaced in pitch and being shaped to provide spectral masking;

to calculate the temporal masking of the signal due to preceding and/or succeeding temporal portions thereof;

to form, for each of the spectral component signals, a representation of the difference between the component signal of the distorted signal and a correspondingly calculated component of the test signal; and

to generate said distortion perception measure from said difference measure.

10. Apparatus as in claim 9 in which the analysis means is arranged to generate a measure of the amount of distortion from said difference signals.

11. Apparatus as in claim 9 in which the analysis means is arranged to generate a mean of the spectral and temporal distribution of the distortion from said difference signals.

12. Apparatus as in claim 9 in which the analysis means is arranged to form a weighted sum of the spectral component differences, weighted in accordance with the relative loudness of sounds of reference amplitude at pitches corresponding to said component signals, and to generate said distortion perception measure in dependence upon said weighted sum.

13. Apparatus as in claim 1, in which said analysis means is arranged to perform a spectral decomposition of said test and/or distorted signals into a plurality of spectral component signals, said decomposition generating, for each spectral component signal, a time sequence of spectral component values each representative of component signal values over a time interval, the time interval for lower frequency component signals exceeding that for higher frequency component signals.

14. Apparatus as in claim 1, in which the analysis means is arranged to filter said test signal and/or said distorted signal in accordance with a filter calculated to correspond to the transfer function of portions of the human auditory system between the equipment to be tested and the inner ear.

15. Apparatus as in claim 14, in which the analysis means is arranged to be capable of selecting one of a plurality of different said transfer functions corresponding, respectively, to different equipment to be tested.

16. Apparatus as in claim 1, further comprising an artificial ear structure for receiving said distorted signal as an acoustic signal and for acoustically processing said distorted signal prior to analysis by the analysis means.

17. Apparatus as in claim 1, in which the signal generator further comprises an artificial mouth structure for receiving said test signal from the signal generator in acoustic form and for acoustically processing said test signal prior to supply to said telecommunications apparatus.

18. Apparatus as in claim 1, in which the signal generator comprises a digital store for storing speech data, and means for reconstituting a speech signal from the stored speech data.

19. Apparatus as in claim 18, in which the stored speech data comprises digitised sound samples and the reconstituting means comprises a digital to analog convertor.

20. Apparatus as in claim 18, in which the store is arranged to store parameters for control of a voice synthesizer comprising said means for reconstituting the speech signal.

21. Apparatus as in claim 1, in which the signal generator is arranged to generate a test signal which comprises a sequence formed of a predetermined, small, number of speech segments, the speech signal comprising several different portions including said segments such that each segment is represented in several different temporal contexts within said sequence, so as to vary the effect one each segment of time varying distortions in the telecommunications apparatus.

22. Apparatus as in claim 21, in which the test signal generator is arranged to vary the context for different said speech segments by prefixing said segments with predetermined signal portions of several different levels within a sequence of said test signal.

23. Apparatus as in claim 21 in which said segments are present in a plurality of different combinations within said sequence.

24. Telecommunications testing apparatus comprising:

a signal generator for supplying a test signal which has a spectral resemblance to human speech but corresponds to a more than one speaker, and

analysis means for receiving a distorted signal which corresponds to said test signal when distorted by telecommunications apparatus to be tested, and for analyzing said distorted signal to generate a distortion perception measure which indicates the extent to which the distortion of said signal will be perceptible to a human listener;

in which the analysis means is arranged to estimate the effect which would be produced on the human auditory system by said test signal, and to estimate therefrom the effect which would be produced on the human auditory system by said distortion.

25. Apparatus according to claim 24, in which the analysis means includes means to estimate the effect which would be produced on the human auditory system by said distorted signal, and to determine the difference between the said effect and that due to the test signal, and to generate said distortion perception measure in dependence upon said difference.

26. Telecommunications testing apparatus comprising:

a signal generator for supplying a test signal which has a spectral resemblance to human speech but corresponds to a more than one speaker, and

analysis means for receiving a distorted signal which corresponds to said test signal when distorted by telecommunications apparatus to be tested, and for analyzing said distorted signal to generate a distortion perception measure which indicates the extent to which the distortion of said signal will be perceptible to a human listener;

in which the analysis means is arranged so as to generate said distortion perception to depend upon perceptual loudness of said distortion, and to depend nonlinearly upon the amplitude of said distortion.

27. Telecommunications testing apparatus comprising:

a signal generator for supplying a test signal which has a spectral resemblance to human speech but corresponds to a more than one speaker, and

analysis means for receiving a distorted signal which corresponds to said test signal when distorted by telecommunications apparatus to be tested, and for analyzing said distorted signal to generate a distortion perception measure which indicates the extent to which the distortion of said signal will be perceptible to a human listener;

in which the analysis means is arranged to generate a plurality of spectral component signals of said test signal and/or said distorted signal;

in which the component signals have different bandwidths, and

in which the component signal bandwidths are selected so they correspond to equal masking amplitudes for signals centred within each band.

28. Telecommunications testing apparatus comprising:

a signal generator for supplying a test signal, and analysis means arranged to;

(a) receive a distorted signal which corresponds to said test signal when distorted by telecommunications apparatus to be tested,

(b) to decompose said distorted signal into a plurality of spectral component bands, the spectral component bands having bandwidths approximately evenly spaced in pitch,

(c) to calculate the temporal masking of the distorted signal,

(d) to form for each of the spectral component signals, the difference between the distorted signal and a correspondingly calculated test signal component for a plurality of successive temporal portions of the test signal; and

(e) to generate a distortion perception measure which indicates the extent to which the distortion of said signal will be perceptible to a human listener by deriving a measure of the said spectral component differences over a plurality of said temporal portions and said spectral components, and a measure of the distribution of said differences over said temporal portions and spectral components.

29. Telecommunications testing apparatus comprising:

analysis means for receiving a distorted signal which corresponds to a test signal when distorted by telecommunications apparatus to be tested, and including means for generating a plurality of spectral component signals of said distorted signal, each said spectral component signal comprising a time sequence of successive spectral component values representative of component signal levels over a time interval, the time intervals being longer for lower frequency spectral component signals than for higher frequency spectral component signals,

in which the component signals have different bandwidths which vary with frequency.

30. Apparatus as in claim 29 in which the component signal bandwidths are, on a Bark scale, approximately equal.

31. Apparatus for testing equipment for handling speech signals, said apparatus comprising:

a signal generator for supplying a test signal, and

analysis means arranged to receive a distorted signal which corresponds to said test signal when distorted by the apparatus to be tested,

the analysis means including;

means to decompose said distorted signal into a plurality of spectral component bands, the spectral component bands having different bandwidths which vary with frequency;

means to calculate the temporal masking of the distorted signal by forming, for each of the spectral component signals, the difference between the distorted signal and a correspondingly calculated test signal component for a plurality of successive temporal portions of the test signal; and

means to generate a distortion perception measure which indicates the extent to which the distortion of said signal will be perceptible to a human listener by deriving measure of the said spectral component differences over a plurality of said temporal portions and said spectral components.

32. Apparatus as in claim 31, wherein the spectral component bands are approximately evenly spaced in pitch, and wherein the means to generate a distortion perception measurement produces a measure of the distribution of said differences over said temporal portions and spectral components.

33. Apparatus as in claim 31 in which the analysis means is arranged to estimate the effect which would be produced on the human auditory system by said test signal, and to estimate therefrom the effect which would be produced on the human auditory system by said distortion.

34. Apparatus as in claim 33, in which the analysis means is arranged to estimate the effect which would be produced on the human auditory system by said distorted signal, and to determine the difference between the said effect and that due to the test signal, and to generate said distortion perception measure in dependence upon said difference.

35. Apparatus as in claim 31, in which the analysis means is arranged to generate said distortion perception to depend upon perceptual loudness of said distortion, and to depend nonlinearly upon the amplitude of said distortion.

36. Apparatus as in claim 31 in which the analysis means is arranged to generate a measure of the spectral and temporal distribution of the distortion from said difference signal.

37. Apparatus as in claim 31 in which the analysis means is arranged to form a weighted sum of the spectral component differences, weighted in accordance with the relative loudness of sounds of reference amplitude at pitches corresponding to said component signals, and to generate said distortion perception measure in dependence upon said weighted sum.

38. Apparatus as in claim 31 in which said analysis means is arranged to generate, for each spectral component signal, a time sequence of spectral component values each representative of component signal values over a time interval, the time interval for lower frequency component signals exceeding that for higher frequency component signals.

39. Apparatus as in claim 31 in which the analysis means is arranged to filter said test signal and/or said distorted signal in accordance with a filter calculated to correspond to the transfer function of portions of the human auditory system between the telecommunications apparatus and the inner ear.

40. Apparatus as in claim 39 in which the analysis means is arranged to be capable of selecting one of a plurality of different said transfer functions corresponding, respectively, to different speech signal carrying equipment.

41. Apparatus as in claim 31 further comprising an artificial ear structure for receiving said distorted signal as an acoustic signal and for acoustically processing said distorted signal prior to analysis by the analysis means.

42. Apparatus as in claim 31 in which the signal generator further comprises an artificial mouth structure for receiving said test signal from the signal generator in acoustic form and for acoustically processing said test signal prior to supply to said equipment to be tested.

43. Apparatus as in claim 31 in which the signal generator comprises a digital store for storing speech data, and means for reconstituting a speech signal from the stored speech data.

44. Apparatus as in claim 43 in which the stored speech data comprises digitised sound samples and the reconstituting means comprises a digital to analogue convertor.

45. Apparatus as in claim 44 in which the store is arranged to store parameters for control of a voice synthesizer comprising said means for reconstituting the speech signal.

46. Apparatus as in claim 31 in which the signal generator is arranged to generate a test signal which comprises a sequence formed of a predetermined number of speech segments smaller than the number of commonly occurring human speech phonemes, the speech signal comprising several different portions including said segments such that each segment is represented in several different temporal contexts within said sequence, so as to vary the effect on each segment of time varying distortions in the equipment under test.

47. Apparatus as in claim 46 in which the test signal generator is arranged to vary the context for different said speech segments by prefixing said segments with predetermined signal portions of several different level within a sequence of said test signal.

48. Apparatus as in claim 47 in which said segments are present in a plurality of different combinations within said sequence.

49. Apparatus for testing humanly perceptible distortion of speech signal processing said apparatus comprising:

means for applying a test signal to speech signal processing circuits;

means for measuring the distribution and interaction over time and frequency of the spectral component differences within a distorted test signal output from said speech signal processing circuits; and

means for generating a distortion perception measure based on said distribution and interaction over time and frequency of the spectral component differences.

50. Apparatus as in claim 49 wherein said means for generating a distortion perception measure also bases such measure on a comparative effect of the detected distortion from the standpoint of the effect of the original signal.

51. Apparatus as in claim 49 wherein said means for generating a distortion perception measure of the distorted test signal also bases such measure on its perceptual loudness and nonlinearly on its amplitude.

52. Apparatus as in claim 49 wherein said means for generating a distortion perception measure of the distorted test signal also bases such measure on component signal bandwidths selected to correspond to equal masking amplitudes for signals centered within each band.

53. A method of testing speech signal handling equipment to be tested, said method comprising:

supplying a test signal to the speech handling equipment, said test signal having a spectral resemblance to human speech but which does not correspond to a single speaker conveying intelligent content,

analyzing the test signal as distorted by said speech signal handling equipment;

determining the extent to which the distortion of said signal will be perceptible to a human listener; and

generating a distortion perception measure indicative of said determined extent;

said step of analyzing including splitting the signal into a plurality of spectral bands having bandwidths which vary with frequency.

54. A method as in claim 53 wherein the analysis comprises the steps of filtering the said test signal and/or distorted signal to generate a plurality of spectral component signals of said test signal and/or said distorted signal.

55. A method as in claim 54 in which an estimation is made, for each spectral component signal, of the masking effect which that spectral component signal would produce on the human ear.

56. A method as in claim 54 comprising the step of generating a time sequence of successive spectral component values representative of component signal levels over a time interval, the time intervals being longer for lower frequency spectral component signals than for higher frequency spectral component signals.

57. A method as in claim 53 comprising the steps of:

generating said test signal,

passing said test signal through said equipment, and

analyzing the distorted signal produced at the output of said equipment.

58. A method as in claim 53 further comprising steps of:

analyzing the extent to which said test signal would be perceptible to a human listener,

deriving a measure of the difference between the perception of the test signal and of the distorted signal, and

deriving said distortion perception measure in dependence upon said difference.

59. A method of testing telecommunications apparatus comprising:

analyzing a speech-like test signal which has a spectral resemblance to human speech but does not correspond to a single speaker conveying intelligent content, as distorted by said telecommunications apparatus;

determining the extent to which the distortion of said signal will be perceptible to a human listener; and

generating a distortion perception measure indicative of said determined extent;

generating said test signal;

passing said test signal through said telecommunications test apparatus; and

analyzing the distorted signal produced at the output of said telecommunications test apparatus;

analyzing the extent to which said test signal would be perceptible to a human listener;

deriving a measure of the difference between the perception of the test signal and of the distorted signal; and

deriving said distortion perception measure in dependence upon said difference.

60. A method of testing equipment for handling speech signals, said method comprising the steps of:

supplying a test signal,

receiving a distorted signal which corresponds to said test signal when distorted by the equipment under test, and

decomposing said distorted signal into a plurality of spectral component bands, the spectral component bands having different bandwidths which vary with frequency;

calculating the temporal masking of the distorted signal to form, for each of the spectral component signals, the difference between the distorted signal and a correspondingly calculated test signal component for a plurality of successive temporal portions of the test signal; and

generating a distortion perception measure which indicates the extent to which the distortion of said signal will be perceptible to a human listener by deriving a measure of the said spectral component differences over a plurality of said temporal portions and said spectral components.

61. A method as in claim 60 wherein the spectral component bands are approximately evenly spaced in pitch, and wherein the distortion perception measurement means produces a measure of the distribution of said differences over said temporal portions and spectral components.

62. Method for testing humanly perceptible distortion of speech signal processing circuits, said method comprising:

applying a test signal to speech signal processing circuits;

measuring the distribution and interaction over time and frequency of the spectral component differences within a distorted test signal output from said speech signal processing circuits; and

generating a distortion perception measure based on said distribution and interaction over time and frequency of the spectral component differences.

63. Method as in claim 62 wherein said step of generating a distortion perception measure also bases such measure on a comparative effect of the detected distortion from the standpoint of the effect of the original signal.

64. Method as in claim 62 wherein said step of generating a distortion perception measure of the distorted test signal also bases such measure on its perceptual loudness and nonlinearly on its amplitude.

65. Method as in claim 62 wherein said step for generating a distortion perception measure of the distorted test signal also bases such measure on component signal bandwidths selected to correspond to equal masking amplitudes for signals centered within each band.