Method and apparatus for wavelet based data compression having adaptive bit rate control for compression of audio information

• US 5,845,243 A
• Filed: 02/03/1997
• Issued: 12/01/1998
• Est. Priority Date: 10/13/1995
• Status: Expired due to Term

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Original Assignee: US Robotics Mobile Communication Corporation
• Inventors: Yang, Jiankan J., Smart, Kevin
• Primary Examiner(s): Dorvil, Richemond
• Application Number:
  US08/794,121
• Time in Patent Office: 666 Days
• Field of Search: 704/229, 704/230, 704/203, 704/205, 704/500, 704/501, 704/504, 704/200, 704/201, 704/224
• US Class Current: 704/200.1
• CPC Class Codes: G06T 9/007 Transform coding, e.g. discrete cosine transform
• CPC Class Codes: G10L 25/27 characterised by the analysis technique
• CPC Class Codes: G11B 20/00007 Time or data compression or expansion audio compression based on psychoacoustics G10L19/00; data processing for reproducing audio data at different playback speeds G10L21/04; video compression H04N19/00; data compression per se H03M7/30
• CPC Class Codes: H04L 27/0004 using wavelets

Abstract

The compression method utilizes a combination of both lossy and lossless compression to achieve significant compression while retaining very high subjective quality of the reconstructed or decompressed signal. Methods and apparatus for compression and decompression of digital audio data are provided. In one preferred embodiment, the compression method or apparatus employs an approximation of a psychoacoustic model for wavelet packet decomposition and has a bit rate control feedback loop particularly well suited to matching the output bit rate of the data compressor to the bandwidth capacity of a communication channel.

Claims

1. A method for compressing digitally sampled audio data comprising the steps of:

a) determining a desired average bit rate;

b) performing a discrete wavelet transform on the digitally sampled data to obtain the resultant wavelet coefficients in such a manner that the resultant wavelet coefficients fall into critical bands that approximate a psychoacoustic model;

c) calculating a control parameter related to the fractional percentage of wavelet coefficients which must be eliminated to achieve the desired average bit rate;

d) using said control parameter to eliminate wavelet coefficients according to a predetermined criteria;

e) quantizing the wavelet coefficients using a selected quantization level;

f) entropy encoding the quantized coefficients; and

g) feeding the number of bits used to represent the entropy encoded coefficients back into the calculation of the control parameter used to eliminate wavelet coefficients so that the desired average bit rate is achieved.

2. A method for compressing digitally sampled audio data as recited in claim 1 further comprising the step of computing a masking threshold of quantization noise for each of the critical bands.

3. A method for compressing digitally sampled audio data as recited in claim 2 where the computing a masking threshold of quantization noise comprises deriving from the digitally sampled audio data the power for each of the critical bands.

4. A method for compressing digitally sampled audio data as recited in claim 2 further comprises the step of computing a number of symbols for masking the quantization noise.

5. A method for compressing digitally sampled audio data as recited in claim 4 further comprising the step of using the control parameter to compute a quantity of symbol quantization bits.

6. A method for compressing digitally sampled audio data as recited in claim 5 further comprising the step of scaling the quantity of symbol quantization bits to achieve the desired bit rate.

7. A method for compressing digitally sampled audio data as recited in claim 5 the control parameter is used to restrict the quantity of symbol quantization bits when a maximum bit rate is exceeded.

8. A method for compressing digitally sampled audio data as recited in claim 5 further comprising the step of entropy encoding the quantity of symbol quantization bits.

9. A method for compressing digitally sampled audio data as recited in claim 5 wherein the quantizing the wavelet coefficients further comprises quantizing and entropy encoding a gain of each of the critical packets when the quantity of symbol quantization bits is greater than zero.

10. A method for compressing digitally sampled audio data as recited in claim 9 wherein the quantizing the wavelet coefficients further comprises scaling the wavelet coefficients of each of the critical packets by the gain of each of the critical packets.

11. A method for compressing digitally sampled audio data as recited in claim 10 wherein the quantizing the wavelet coefficients is performed using the quantity of symbol quantization bits.

12. A method for compressing digitally sampled audio data as recited in claim 1 wherein the entropy encoding comprises run length encoding.

13. A method for compressing digitally sampled audio data as recited in claim 1 wherein the entropy encoding comprises Huffman encoding.

14. A method for compressing digitally sampled audio data as recited in claim 1 wherein the entropy encoding comprises a combination of entropy encoding methods.

15. A method for compressing digitally sampled audio data as recited in claim 14 wherein the combination of entropy encoding methods comprises a combination of run length encoding and Huffman encoding.

16. A method for compressing digitally sampled audio data which has been divided into data frames containing a predefined number of digital audio data samples, the method comprising the steps of:

a) performing a discrete wavelet transform on the data frame to obtain the corresponding wavelet coefficients;

b) decomposing the resultant wavelet coefficients into critical bands that approximate a psychoacoustic model;

c) calculating a control parameter used to eliminate wavelet coefficients in this frame in order to achieve a desired average bit rate;

d) selecting a quantization level for the wavelet coefficients based on a psychoacoustic model which uses one or more parameters derived from the data in the data frame;

e) quantizing the wavelet coefficients at the selected quantization level;

f) entropy encoding the quantized wavelet coefficients; and

g) feeding the number of bits used to represent the entropy encoded coefficients back into the calculation of the control parameter so that the desired average bit rate is achieved.

17. A method for compressing digitally sampled audio data as recited in claim 16 further comprising the step of computing a masking threshold of quantization noise for each of the critical bands.

18. A method for compressing digitally sampled audio data as recited in claim 17 where the computing a masking threshold of quantization noise comprises deriving from the digitally sampled audio data the power for each of the critical bands.

19. A method for compressing digitally sampled audio data as recited in claim 17 further comprising the step of computing a number of symbols for masking the quantization noise.

20. A method for compressing digitally sampled audio data as recited in claim 19 wherein the control parameter is used to compute a quantity of symbol quantization bits.

21. A method for compressing digitally sampled audio data as recited in claim 20 further comprising the step of scaling the quantity of symbol quantization bits to achieve the desired bit rate.

22. A method for compressing digitally sampled audio data as recited in claim 20 further comprising the step of restricting the quantity of symbol quantization bits when a maximum bit rate is exceeded.

23. A method for compressing digitally sampled audio data as recited in claim 20 further comprising the step of entropy encoding the quantity of symbol quantization bits.

24. A method for compressing digitally sampled audio data as recited in claim 20 wherein the quantizing the wavelet coefficients further comprises quantizing and entropy encoding a gain of each of the critical packets when the quantity of symbol quantization bits is greater than zero.

25. A method for compressing digitally sampled audio data as recited in claim 24 wherein the quantizing the wavelet coefficients further comprises scaling the wavelet coefficients of each of the critical packets by the gain of each of the critical packets.

26. A method for compressing digitally sampled audio data as recited in claim 25 wherein the scaling the wavelet coefficients comprises scaling the wavelet coefficients so they lie between -1.0 and +1.0.

27. A method for compressing digitally sampled audio data as recited in claim 25 wherein the quantizing the wavelet coefficients is performed using the quantity of symbol quantization bits.

28. A method of reconstructing digitally sampled audio information which has been previously compressed, the method comprising the steps of:

a) entropy decoding the compressed digitally sampled audio information;

b) decomposing the compressed digitally sampled audio information into critical bands that approximate a psychoacoustic model;

c) performing an inverse discrete wavelet transform; and

d) resealing the transformed data into the time domain.

29. A method of reconstructing digitally sampled audio information which has been previously compressed as recited in claim 28 further comprising the step of decoding a number of quantization bits for each of the critical bands.

30. A method of reconstructing digitally sampled audio information which has been previously compressed as recited in claim 29 further comprising the step of when the number of quantization bits is greater than zero for each of the critical bands, decoding quantized coefficients.

31. A method of reconstructing digitally sampled audio information which has been previously compressed as recited in claim 30 further comprising the steps of:

a) decoding a gain factor for each of the critical bands; and

b) scaling the quantized coefficients of each of the critical bands by the gain factor prior to performing the inverse discrete wavelet transform.

32. An article of manufacture comprising:

a) a computer usable medium having computer readable program code means embedded therein for compressing digitally sampled audio data, the computer readable program code means comprising;

i) means for determining a desired average bit rate;

ii) means for performing a discrete wavelet transform on the digitally sampled data to obtain the resultant wavelet coefficients in such a manner that the resultant wavelet coefficients fall into critical bands that approximate a psychoacoustic model;

iii) means for calculating a control parameter used to eliminate symbols representing wavelet coefficients in order to achieve the desired average bit rate;

iv) means for eliminating a plurality of wavelet coefficients according to a predetermined criteria;

v) means for quantizing the wavelet coefficients using a selected quantization level;

vi) means for entropy encoding the quantized coefficients; and

vii) means for feeding the number of bits used to represent the entropy encoded coefficients back into the calculation of the control parameter in order to achieve the desired average bit rate.

33. An article of manufacture as recited in claim 32 further comprising means for computing a masking threshold of quantization noise for each of the critical bands.

34. An article of manufacture as recited in claim 33 where the means for computing a masking threshold of quantization noise comprises means for deriving from the digitally sampled audio data the power for each of the critical bands.

35. An article of manufacture as recited in claim 33 further comprising means for computing a number of symbols for masking the quantization noise.

36. An article of manufacture as recited in claim 35 further comprising means for computing a quantity of symbol quantization bits.

37. An article of manufacture as recited in claim 36 further comprising means for scaling the quantity of symbol quantization bits to achieve the desired bit rate.

38. An article of manufacture as recited in claim 36 further comprising means for restricting the quantity of symbol quantization bits when a maximum bit rate is exceeded.

39. An article of manufacture as recited in claim 36 further comprising means for entropy encoding the quantity of symbol quantization bits.

40. An article of manufacture as recited in claim 36 wherein the means for quantizing the wavelet coefficients further comprises means for quantizing and entropy encoding a gain of each of the critical packets when the quantity of symbol quantization bits is greater than zero.

41. An article of manufacture as recited in claim 40 wherein the means for quantizing the wavelet coefficients further comprises means for scaling the wavelet coefficients of each of the critical packets by the gain of each of the critical packets.

42. An article of manufacture as recited in claim 41 wherein the means for quantizing the wavelet coefficients is performed using the quantity of symbol quantization bits.

43. An article of manufacture as recited in claim 32 wherein the means for entropy encoding comprises means for run length encoding.

44. An article of manufacture as recited in claim 32 wherein the means for entropy encoding comprises Huffman encoding.

45. An article of manufacture as recited in claim 32 wherein the means for entropy encoding comprises a combination of entropy encoding methods.

46. An article of manufacture as recited in claim 45 wherein the combination of entropy encoding methods comprises a combination of run length encoding and Huffman encoding.

47. An article of manufacture comprising:

a) a computer usable medium having computer readable program code means embedded therein for compressing digitally sampled data, the computer readable program code means comprising;

i) means for dividing the data into data frames containing a predefined number of digital data samples;

ii) means for performing a discrete wavelet transform on the data frame to obtain the corresponding wavelet coefficients in such a manner that the wavelet coefficients fall into critical bands that approximate a psychoacoustic model;

iii) means for calculating a control parameter used to adjust the instantaneous bit rate in order to achieve a desired average bit rate;

iv) means for eliminating a plurality of wavelet coefficients according to a predefined criteria;

v) means for selecting a quantization level for the wavelet coefficients based on the psychoacoustic model which uses one or more parameters derived from the data in the data frame;

vi) means for quantizing the wavelet coefficients at the selected quantization level;

vii) means for entropy encoding the quantized wavelet coefficients; and

viii) means for feeding the number of bits used to represent the entropy encoded coefficients back into the calculation of the control parameter in order to achieve the desired average bit rate.

48. An article of manufacture as recited in claim 47 further comprising means for computing a masking threshold of quantization noise for each of the critical bands.

49. An article of manufacture as recited in claim 48 wherein the means for computing a masking threshold of quantization noise comprises means for deriving from the digitally sampled audio data the power for each of the critical bands.

50. An article of manufacture as recited in claim 48 further comprising means for computing a number of symbols for masking the quantization noise.

51. An article of manufacture as recited in claim 50 further comprising means for computing a quantity of symbol quantization bits.

52. An article of manufacture as recited in claim 51 further comprising means for scaling the quantity of symbol quantization bits to achieve the desired bit rate.

53. An article of manufacture as recited in claim 51 further comprising means for restricting the quantity of symbol quantization bits when a maximum bit rate is exceeded.

54. An article of manufacture as recited in claim 51 further comprising means for entropy encoding the quantity of symbol quantization bits.

55. An article of manufacture as recited in claim 51 wherein the means for quantizing the wavelet coefficients further comprises means for quantizing and entropy encoding a gain of each of the critical packets when the quantity of symbol quantization bits is greater than zero.

56. An article of manufacture as recited in claim 55 wherein the means for quantizing the wavelet coefficients further comprises means for scaling the wavelet coefficients of each of the critical packets by the gain of each of the critical packets.

57. An article of manufacture as recited in claim 56 wherein the means for scaling the wavelet coefficients comprises means for scaling the wavelet coefficients so they lie between -1.0 and +1.0.

58. An article of manufacture as recited in claim 56 wherein the quantizing the wavelet coefficients is performed using the quantity of symbol quantization bits.

59. An article of manufacture comprising:

a) a computer usable medium having computer readable code means embedded therein for reconstructing digitally sampled information which has been previously compressed, the computer readable code means comprising;

i) means for entropy decoding the compressed digitally sampled audio information to obtain entropy decoded information comprising wavelet coefficients;

ii) means for decomposing the compressed digitally sampled audio information into critical bands that approximate a psychoacoustic model;

iii) means for performing an inverse discrete wavelet transform on the entropy decoded information contained within the critical bands to obtain inverse wavelet transformed data; and

iv) means for resealing the inverse wavelet transformed data into the time domain.

60. An article of manufacture as recited in claim 59 further comprising a means for decoding a number of quantization bits for each of the critical bands.

61. An article of manufacturers recited in claim 60 further comprising a means for decoding quantized coefficients when the number of quantization bits is greater than zero for each of the critical bands.

62. A method for compressing digitally sampled audio data comprising the steps of:

determining a desired average bit rate;

performing a discrete wavelet transform on the digitally sampled data to obtain resultant wavelet coefficients in a plurality of wavelet bands;

selecting the number of bits that will be used to quantize the wavelet coefficients in said plurality of wavelet bands;

encoding the wavelet coefficients in each of said plurality of wavelet bands with an entropy encoder to obtain a plurality of individually encoded wavelet bands;

assembling the individually encoded wavelet bands into a single compressed output data frame; and

feeding the number of bits used to represent the single compressed output data frame back into the selection of the number of bits used to quantize the wavelet coefficients in said plurality of wavelet bands.

63. A method for compressing digitally sampled audio data comprising the steps of:

performing a discrete wavelet transform on the digitally sampled data to obtain resultant wavelet coefficients in a plurality of wavelet bands;

selecting the number of bits that will be used to quantize the wavelet coefficients in said plurality of wavelet bands;

encoding the wavelet coefficients in each of said plurality of wavelet bands with an entropy encoder to obtain a plurality of individually encoded wavelet bands; and

assembling the individually encoded wavelet bands into a single compressed output data frame.

64. A computer-readable medium having computer-executable instructions comprising:

means for performing a discrete wavelet transform on digitally sampled audio data to obtain the resultant wavelet coefficients in such a manner that the resultant wavelet coefficients fall into critical bands that approximate a psychoacoustic model of a human ear;

means for quantizing said wavelet coefficients to a selected quantization level, said means for quantizing being adapted to quantize the wavelet coefficients in each of said critical bands to obtain a plurality of quantized critical bands;

means for individually entropy encoding each of said plurality of quantized critical bands; and

means for controlling the bit rate from said means for individually entropy encoding by controlling the selected quantization level based on the number of bits at the output of said means for individually entropy encoding.

65. A computer-readable medium having computer-executable instructions comprising:

means for performing a discrete wavelet transform on digitally sampled audio data to obtain the resultant wavelet coefficients in such a manner that the resultant wavelet coefficients fall into critical bands that approximate a psychoacoustic model of a human ear;

means for quantizing said wavelet coefficients to a selected quantization level, said means for quantizing being adapted to quantize the wavelet coefficients in each of said critical bands to obtain a plurality of quantized critical bands; and

means for individually entropy encoding each of said plurality of quantized critical bands.