Encoder/decoder for multidimensional sound fields

• US 5,583,962 A
• Filed: 09/04/1992
• Issued: 12/10/1996
• Est. Priority Date: 01/08/1991
• Status: Expired due to Term

• Current Assignee: Dolby Laboratories Licensing Corporation (Dolby Laboratories Incorporated)
• Original Assignee: Dolby Laboratories Licensing Corporation (Dolby Laboratories Incorporated)
• Inventors: Dolby, Ray M., Davis, Mark F., Todd, Craig C.
• Primary Examiner(s): Knepper, David D.
• Application Number:
  US07/927,429
• Time in Patent Office: 1,558 Days
• Field of Search: 395/2, 395/2.14-2.19, 395/2.38, 395/2.39, 381/29-40
• US Class Current: 704/229
• CPC Class Codes: G10L 19/008 Multichannel audio signal coding or decoding using interchannel correlation to reduce redundancy, e.g. joint-stereo, intensity-coding or matrixing
• CPC Class Codes: G11B 20/00 Signal processing not specific to the method of recording or reproducing; Circuits therefor
• CPC Class Codes: G11B 20/10527 Audio or video recording; Data buffering arrangements G11B20/12 - G11B20/18 take precedence
• CPC Class Codes: H04H 20/91 broadcasting computer programmes
• CPC Class Codes: H04S 3/008 in which the audio signals are in digital form, i.e. employing more than two discrete digital channels data reduction aspects thereof based on psychoacoustics G10L19/02

Abstract

Two or more audio channels (i.e.--stereo, 4-channel surround, etc.) are each divided into frequency subbands to be coarsely quantized. An adaptive bit allocation scheme is then applied to subbands which are combined across channels such that equivalent subbands (i.e.--same frequency band) from each channel are grouped together to form a steered subband. The power from each subband is averaged across the channels to form a steered subband level. Bits are conserved by forming a vector in which each channel is represented by the difference between the steered subband level (average) and the actual subband level. Subbands which are not steered are represented by the coarse quantization and are considered unsteered channel subbands. Subbands which are steered are represented by vectors and are considered composite channel subbands. Vectors may be represented using a lookup table and the steered subband levels may be calculated using alternatives such as a peak level rather than an average.

Claims

1. In an encoder for encoding two or more audio channels, the combination comprising:

subband means for generating subband signals, each subband signal representing spectral energy in a respective subband of a respective one of said audio channels,

composite means for forming one or more composite signals, each composite signal formed by combining subband signals in a respective subband of two or more of said audio channels, and

formatting means for assembling an output signal including information representing said one or more composite signals in a form comprising a coarse measure of composite signal contents and a corresponding finer measure of composite signal contents, and including information conveying spectral levels of each subband signal combined into a respective composite signal.

2. The combination of claim 1 further comprising means for encoding said one or more composite signals and subband signals not combined into a respective composite signal, wherein said composite means forms said one or more composite signals only when the amount of information required to encode subband signals generated by said subband means exceeds a limit, and wherein said composite signals are formed only to the extent that the amount of information saved by encoding said composite signals rather than the subband signals combined into said composite signals is sufficient to allow encoding using amount of information which does not exceed said limit.

3. The combination of claim 1 wherein said more finely quantized values are quantized using bits allocated from a common pool of bits.

4. The combination of claim 1 further comprising means for adding, prior to quantization, a noise-like signal to said subband signals and said one or more composite signals.

5. The combination of claim 4 wherein the mean amplitude of said noise-like signal substantially matches the expected quantizing error of said subband signals and said one or more composite signals.

6. The combination of claim 1 wherein said information conveying spectral levels for a respective composite signal includes an indication of amplitude or power level of each constituent subband signal.

7. The combination of claim 6 wherein said indication of amplitude or power level is either

a plurality of elements, each element representing the difference in amplitude or power level between a respective constituent subband signal and a level of said composite signal, or

a plurality of elements, each element representing the ratio of amplitude or power level between a respective constituent subband signal and a level of said composite signal, or

a plurality of elements, each element representing the absolute value of the amplitude or power level of each respective constituent subband signal.

8. The combination of claim 1 wherein said plurality of audio channels represent a sound field and said information conveying spectral levels for a respective composite signal includes sound field localization information for constituent subband signals combined into the respective composite signal.

9. The combination of claim 1 wherein said composite means includes means for compensating out-of-phase signal components between subband signals from which said one or more composite signals are formed.

10. The combination of claim 1 wherein said one or more composite signals and the subband signals not represented by a respective composite signal are represented in a form comprising one or more scale factors each associated with one or more scaled values, the combination further comprising means for adjusting either or both dynamic range and gain by manipulating the values of said one or more scale factors.

11. The combination of claim 1 wherein said composite means includes selection means for selecting each subband from which a respective one of said one or more composite signals is formed.

12. The combination of claim 11 wherein said selection means selects one or more predetermined subbands.

13. The combination of claim 11 wherein said selection means selects one or more of the highest frequency subbands.

14. The combination of claim 11 wherein said selection means selects one or more subbands such that the resulting one or more composite signals are least likely to be subject to errors caused by out-of-phase signal cancellation.

15. The combination of claim 11 further comprising means for allocating a limited number of bits to said one or more composite signals and to subband signals not represented by said one or more composite signals, wherein said selection means selects subbands whose subband signals, if not represented by a respective composite signal, would not be allocated a respective minimum number of bits.

16. The combination of claim 15 wherein said respective minimum number of bits is the number of bits required to render quantizing noise in a respective subband substantially inaudible.

17. The combination of claim 11 wherein said selection means selects a subband starting with subbands in which coding inaccuracies are least objectionable.

18. The combination of claim 11 wherein said selection means selects the highest frequency subband, reiteratively selecting the highest frequency subband not already selected until sufficient bits are made available to allocate at least said respective minimum number of bits to subband signals not represented by a respective composite signal.

19. The combination of claim 11 wherein said means for selecting further selects subband signals according to in which audio channels the subband signals are located.

20. In a decoder for decoding an encoded signal generated by an encoder, said encoded signal including subband information representing respective subbands of a plurality of audio channels and including spectral level information, each subband constituting a portion of the spectrum of said audio channels, said subband information representing one or more composite signals and a plurality of subband signals, each of said composite signals formed in said encoder by combining subband signals of two or more of said plurality of audio channels in a respective subband, the combination in said decoder comprising:

deformatting means for obtaining said subband information and said spectral level information from said encoded signal, wherein said subband information is represented in a form comprising a coarse measure of composite signal contents and a corresponding finer measure of composite signal contents, and said spectral level information conveys spectral levels of each subband signal combined in a respective composite signal,

reconstruction means for obtaining said one or more composite signals and said plurality of subband signals in response to said subband information, and for deriving subband signals in response to said one or more composite signals and said spectral level information, and

synthesis means for generating a plurality of output signals in response to said derived subband signals and said plurality of subband signals obtained from said subband information.

21. The combination of claim 20 further comprising means for substituting a noise-like signal for the least significant bits of said more finely quantized values.

22. The combination of claim 20 wherein said more finely quantized values include a noise-like signal added prior to their quantization in said encoder, wherein said combination further comprises means for generating a noise-like signal substantially the same as that added prior to quantization, and means for subtracting said noise-like signal from said more finely quantized values after dequantization.

23. The combination of claim 20 wherein said more finely quantized values are dequantized using bits allocated from a common pool of bits.

24. The combination of claim 20 wherein said spectral level information includes an indication of amplitude or power level of each constituent subband signal combined into a respective composite signal.

25. The combination of claim 20 wherein said spectral level information includes an indication of sound field localization for constituent subband signals combined into a respective composite signal.

26. The combination of claim 20 further comprising means for inverse out-of-phase compensation of signal components between subband signals from which composite signals are formed.

27. The combination of claim 20 wherein said encoded signal comprises subband information represented in a form comprising one or more scale factors each associated with one or more scaled values, said combination further comprising means for adjusting either or both dynamic range and gain of said subband information by manipulating the values of said one or more scale factors.

28. A method for use in the encoding of two or more audio channels, comprising:

generating subband signals of said audio channels, each subband signal representing spectral energy in a respective subband of a respective one of said channels,

forming one or more composite signals, each composite signal formed by combining subband signals in a respective subband of two or more of said audio channels, and

assembling an output signal including information representing said one or more composite signals in a form comprising a coarse measure of composite signal contents and a corresponding finer measure of composite signal contents, and including information conveying spectral levels of said subband signal combined into a respective composite signal.

29. The method of claim 28 wherein said plurality of audio channels represent a sound field and said information conveying spectral levels for a respective composite signal includes sound field localization information for constituent subband signal combined into the respective composite signal.

30. The method of claim 28 wherein forming one or more composite signals includes compensating out-of-phase signal components between subband signal from which said one or more composite signals are formed.

31. The method of claim 28 wherein said one or more composite signals and the subband signals not represented by a respective composite signal are represented in a form comprising one or more scale factors each associated with one or more scaled values, the method further comprising adjusting either or both dynamic range and gain by manipulating the values of said one or more scale factors.

32. The method of claim 28 further comprising encoding said one or more composite signals and subband signals not combined into a respective composite signal, wherein said one or more composite signals are formed only when the amount of information required to encode subband signals exceeds a limit, and wherein said composite signals are formed only to the extent that the amount of information saved by encoding said composite signals rather than the subband signals combined into said composite signals is sufficient to allow encoding using amount of information which does not exceed said limit.

33. The method of claim 28 wherein said more finely quantized values are quantized using bits allocated from a common pool of bits.

34. The method of claim 28 wherein a noise-like signal is added to said subband signal and said one or more composite signals prior to quantization.

35. The method of claim 34 wherein the mean amplitude of said noise-like signal substantially matches the expected quantizing error of said subband signals and said one or more composite signals.

36. The method of claim 28 wherein said forming one or more composite signals includes selecting each subband from which a respective one of said one or more composite signals is formed.

37. The method of claim 36 wherein one or more predetermined subbands are selected.

38. The method of claim 36 wherein one or more of the highest frequency subbands are selected.

39. The method of claim 36 wherein one or more subbands are selected such that the resulting one or more composite signals are least likely to be subject to errors caused by out-of-phase signal cancellation.

40. The method of claim 36 further comprising allocating a limited number of bits to said one or more composite signals and to subband signals not represented by said one or more composite signals, wherein subbands are selected whose subband signals, when not represented by a respective composite signal, would not be allocated a respective minimum number of bits.

41. The method of claim 40 wherein said respective minimum number of bits is the number of bits required to render quantizing noise in a respective subband substantially inaudible.

42. The method of claim 36 wherein a subband is selected starting with subbands in which coding inaccuracies are least objectionable.

43. The method of claim 36 wherein the highest frequency subband is selected, reiteratively selecting the highest frequency subband not already selected until sufficient bits are made available to allocate at least said respective minimum number of bits to subband signals not represented by a respective composite signal.

44. The method of claim 36 wherein subband signals are selected according to in which audio channels the subband signals are located.

45. The method of claim 28 wherein said information conveying spectra levels for a respective composite signal includes an indication of the amplitude or power level of said constituent subband signals.

46. The method of claim 45 wherein said indication of amplitude or power level is either

a plurality of elements, each element representing the difference in amplitude or power level between a respective constituent subband signal and a level of said composite signal, or

a plurality of elements, each element representing the ratio of amplitude or power level between a respective constituent subband signal and a level of said composite signal, or

a plurality of elements, each element representing the absolute value of the amplitude or power level of each respective constituent subband signal.

47. A method for use in decoding an encoded signal generated by an encoder, said encoded signal including subband information representing respective subbands of a plurality of audio channels and including spectral level information, each subband constituting a portion of the spectrum of said audio channels, said subband information representing one or more composite signals and a plurality of subband signals, each of said composite signals formed in said encoder by combining subband signals of two or more of said plurality of audio channels in a respective subband, the method comprising:

obtaining said subband information and said spectral level information from said encoded signal, wherein said subband information is represented in a form comprising a coarse measure of composite signal contents and a corresponding finer measure of composite signal contents, and said spectral level information conveys spectral levels of each subband signal combined in a respective composite signal,

obtaining said one or more composite signals and said plurality of subband signals in response to said subband information, and for deriving subband signals in response to said one or more composite signals and said spectral level information, and

generating a plurality of output signals in response to said derived subband signals and said plurality of subband signals obtained from said subband information.

48. The method of claim 47 further comprises substituting a noise-like signal for the least significant bits of said more finely quantized values.

49. The method of claim 47 wherein said more finely quantized values include a noise-like signal added prior to quantization in said encoder, wherein said method further comprises generating a noise-like signal substantially the same as that added prior to quantization, and subtracting said noise-like signal from said more finely quantized values after dequantization.

50. The method of claim 47 wherein said more finely quantized values are dequantized using bits allocated from a common pool of bits.

51. The method of claim 47 wherein said spectral level information includes an indication of amplitude or power level of each constituent subband signal combined into a respective composite signal.

52. The method of claim 47 wherein said spectral level information includes an indication of sound field localization for constituent subband signals combined into a respective composite signal.

53. The method of claim 47 further comprising inverse out-of-phase compensating of signal components between subband signals from which composite signals are formed.

54. The method of claim 47 wherein said encoded signal comprises subband information represented in a form comprising one or more scale factors each associated with one or more scaled values, said method further comprising adjusting either or both dynamic range and gain of said subband information by manipulating the values of said one or more scale factors.

55. In an encoder for encoding two or more audio channels, the combination comprising:

subband means for generating subband signals, each subband signal representing spectral energy in a respective subband of a respective one of said audio channels,

composite means for forming one or more composite signals, each composite signal formed by combining subband signals in a respective subband of two or more of said audio channels, wherein said composite means includes means for compensating out-of-phase signal components between subband signals from which said one or more composite signals are formed, and

formatting means for assembling an output signal including information representing said one or more composite signals and subband signals not combined into a respective composite signal.

56. The combination of claim 55 further comprising means for encoding said one or more composite signals and subband signals not combined into a respective composite signal, wherein said composite means forms said one or more composite signals only when the amount of information required to encode subband signals generated by said subband means exceeds a limit, and wherein said composite signals are formed only to the extent that the amount of information saved by encoding said composite signals rather than the subband signals combined into said composite signals is sufficient to allow encoding using amount of information which does not exceed said limit.

57. The combination of claim 55 or 56 wherein said composite means includes selection means for selecting each subband from which a respective one of said one or more composite signals is formed.

58. The combination of claim 57 wherein said selection means selects a subband starting with subbands in which coding inaccuracies are least objectionable.

59. The combination of claim 57 wherein said means for selecting further selects subband signals according to which audio channels the subband signals are located.

60. The combination of claim 55 wherein said information conveying spectral levels for a respective composite signal includes an indication of amplitude or power level of each constituent subband signal.

61. The combination of claim 60 wherein said indication of amplitude or power level is either

a plurality of elements, each element representing the difference in amplitude or power level between a respective constituent subband signal and a level of said composite signal, or

a plurality of elements, each element representing the ratio of amplitude or power level between a respective constituent subband signal and a level of said composite signal, or

a plurality of elements, each element representing the absolute value of the amplitude or power level of each respective constituent subband signal.

62. The combination of claim 55, 56 or 60 wherein said one or more composite signals and the subband signals not represented by a respective composite signal are represented in a form comprising one or more scale factors each associated with one or more scaled values, the combination further comprising means for adjusting either or both dynamic range and gain by manipulating the values of said one or more scale factors.

63. A method for encoding two or more audio channels comprising:

generating subband signals, each subband signal representing spectral energy in a respective subband of a respective one of said audio channels,

forming one or more composite signals, each composite signal formed by combining subband signals in a respective subband of two or more of said audio channels, wherein said combining includes compensating out-of-phase signal components between subband signals from which said one or more composite signals are formed, and

assembling an output signal including information representing said one or more composite signals and subband signals not combined into a respective composite signal.

64. The method of claim 63 further comprising encoding said one or more composite signals and subband signals not combined into a respective composite signal, wherein said one or more composite signals are formed only when the amount of information required to encode said subband signals exceeds a limit, and wherein said composite signals are formed only to the extent that the amount of information saved by encoding said composite signals rather than the subband signals combined into said composite signals is sufficient to allow encoding using amount of information which does not exceed said limit.

65. The method of claim 63 wherein said information conveying spectral levels for a respective composite signal includes an indication of amplitude or power level of each constituent subband signal.

66. The method of claim 65 wherein said indication of amplitude or power level is either

a plurality of elements, each element representing the difference in amplitude or power level between a respective constituent subband signal and a level of said composite signal, or

a plurality of elements, each element representing the ratio of amplitude or power level between a respective constituent subband signal and a level of said composite signal, or

a plurality of elements, each element representing the absolute value of the amplitude or power level of each respective constituent subband signal.

67. The method of claim 63 or 64 wherein said combining includes selecting each subband from which a respective one of said one or more composite signals is formed.

68. The method of claim 67 wherein said selecting selects a subband starting with subbands in which coding inaccuracies are least objectionable.

69. The method of claim 67 wherein said selecting selects subband signals according to in which audio channels the subband signals are located.

70. The method of claim 63, 64 or 65 wherein said one or more composite signals and the subband signals not represented by a respective composite signal are represented in a form comprising one or more scale factors each associated with one or more scaled values, the method further comprising adjusting either or both dynamic range and gain by manipulating the values of said one or more scale factors.