Adaptive transform coding by selecting optimum block lengths according to variatons between successive blocks

• US 5,235,623 A
• Filed: 11/14/1990
• Issued: 08/10/1993
• Est. Priority Date: 11/14/1989
• Status: Expired due to Term

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Sugiyama, Akihiko, Iwadare, Masahiro, Nishitani, Takao
• Primary Examiner(s): Kuntz, Curtis
• Assistant Examiner(s): GHEBRETINSAE, TEMESGHEN
• Application Number:
  US07/613,122
• Time in Patent Office: 1,000 Days
• Field of Search: 375/122, 358/138, 358/426, 341/61, 341/67, 341/118, 341/141, 381/29, 381/30, 381/31, 370/82, 370/83, 370/118
• US Class Current: 375/241
• CPC Class Codes: G10L 19/0212 using orthogonal transformation
• CPC Class Codes: G10L 19/022 Blocking, i.e. grouping of samples in time; Choice of analysis windows; Overlap factoring
• CPC Class Codes: H03M 7/3053 Block-companding PCM systems

Abstract

Subblocks of input digital samples are stored into a buffer at frame intervals and segmented into blocks having an integral multiple of the length of the subblock. Each block is encoded into transform coefficients and stored into a memory. Each coefficient is squared and those of the squared transform coefficients which correspond to high-frequency components of the input digital samples are summed and a minimum value is detected therefrom as corresponding to an optimum block length. Those transform coefficients which correspond to the optimum block length are selected from the memory and multiplexed with a signal representative of the optimum block length. In a modification, interblock differences are detected between successive transform coefficients of equal block length and squared. Those of the squared interblock differences which correspond to equal block length are summed, producing a set of squared sums for each block length. A variation is detected between a representative squared sum of a given set and a representative squared sum of a successive set to identify a block length which corresponds to the variation as an optimum block length.

Claims

1. An adaptive transform coding method comprising:

a) storing subblocks of input digital samples representing an audio-frequency signal in a buffer at predetermined intervals, reading the stored samples from the buffer and segmenting the samples from the buffer into a plurality of blocks each having an integral multiple of length of said subblock;

b) encoding each of said blocks into corresponding transform coefficients and storing said transform coefficients in a memory;

c) squaring each of said transform coefficients;

d) summing those of said squared transform coefficients which correspond to high-frequency components of said input digital samples to produce squared sums;

e) detecting a minimum of said squared sums as corresponding to an optimum block length;

g) selecting those of said transform coefficients which correspond to said optimum block length from said memory; and

h) multiplexing the selected transform coefficients with a block length signal representative of said optimum block length.

2. An adaptive transform coding method as claimed in claim 1, further comprising the steps of determining a step size from the transform coefficients selected by the step (g) and quantizing the selected transform coefficients according to said step size, and wherein the step (h) comprises multiplexing the quantized transform coefficients with a signal representative of said step size.

3. An adaptive transform coding method as claimed in claim 2, wherein the selected step size is determined by squaring said transform coefficients, taking an average value of a group of the squared transform coefficients of neighboring occurrences as a representative value of said group, providing interpolations between successive ones of said squared transform coefficients, and deriving said step size from the interpolated squared transform coefficients.

4. An adaptive transform coding method as claimed in claim 1, wherein the step (a) comprises organizing 2^a+b N samples read out of said buffer into (m+1) groups of 2^a blocks of 2^b N samples each, where a+b=m, and a is a variable in a range from m to 0 and b is a variable in a range from 0 to m, and where m is equal to or greater than 2, and N is an integer, each of said groups containing an equal number of samples of same arrivals.

5. An adaptive transform coding method as claimed in claim 1, further comprising deriving respective variances of said blocks, scaling said blocks according to the respective variances, storing said variances into a second memory, and selecting those variances corresponding to said optimum block length from said second memory, wherein the step (h) comprises multiplexing the selected transform coefficients with a signal representative of the selected variances.

6. An adaptive transform coding method as claimed in claim 1, further comprising generating a signal specifying a predetermined block length, wherein the step (g) comprises selecting those of said transform coefficients which correspond to said predetermined block length and the step (h) comprises multiplexing the transform coefficients selected as corresponding to said predetermined block length with a signal representative of said predetermined block length instead of with the signal representative of the optimum block length.

7. An adaptive transform coding method comprising:

a) storing subblocks of input digital samples representing an audio-frequency signal into a buffer at frame intervals, reading the stored samples from the buffer and segmenting the samples from the buffer into a plurality of blocks each having an integral multiple of the length of said subblock;

b) encoding each of said blocks into corresponding transform coefficients and storing said transform coefficients into a memory;

c) detecting interblock differences between successive transform coefficients of equal block length;

d) squaring each of said interblock differences;

e) summing those of said squared interblock differences which correspond to equal block length to produce a set of squared sums of said interblock differences for each block length;

f) detecting representative values of the squared sums from successive ones of said sets, detecting a variation between said representative values and identifying a block length corresponding to said variation as an optimum block length;

g) selecting those of said transform coefficients which correspond to said optimum block length from said memory; and

h) multiplexing the selected transform coefficients with a block length signal representative of said optimum block length.

8. An adaptive transform coding method as claimed in claim 7, wherein the step (f) comprises the steps of:

detecting a maximum value of the squared sums from each set; and

comparing said maximum values detected from a successive pair of said sets so that said compared values correspond respectively to longer and shorter block lengths and selecting one of the compared values as corresponding to said variation if the maximum value of the longer block length has a smaller maximum value than the maximum value of a shorter block length.

9. An adaptive transform coding method as claimed in claim 8, wherein the step (f) further comprises the steps of:

detecting those of said squared sums from each frame interval which correspond to a minimum block length;

dividing each of the detected squared sums with the detected squared sum of an adjacent occurrence to produce a plurality of ratios from each frame interval; and

detecting one of said ratios as corresponding to said variation if said ratio exceeds a prescribed value.

10. An adaptive transform coding method as claimed in claim 7, wherein the step (f) comprises the steps of:

detecting a maximum value of the squared sums from each set;

dividing one of the maximum values of a successive pair by the other maximum value of the pair and deriving a plurality of ratios from a plurality of said pairs; and

detecting a minimum value from said ratios as corresponding to said variation.

11. An adaptive transform coding method as claimed in claim 10, wherein the step (f) further comprises the steps of:

detecting those of said squared sums from each frame interval which correspond to a minimum block length;

dividing each of the detected squared sums with the detected squared sum of an adjacent occurrence to produce a plurality of ratios from each frame interval; and

detecting one of said ratios as corresponding to said variation if said ratio exceeds a prescribed value.

12. An adaptive transform coding method as claimed in claim 7, further comprising the steps of determining a step size from the transform coefficients selected by the step (g) and quantizing the selected transform coefficients according to said step size, and the step (h) comprises multiplexing the quantized transform coefficients with a signal representative of said step size.

13. An adaptive transform coding method as claimed in claim 12, wherein said selected step size is determined by squaring said transform coefficients, taking an average value of a group of the squared transform coefficients of neighboring occurrences as a representative value of said group, providing interpolations between successive ones of said squared transform coefficients, and deriving said step size from the interpolated squared transform coefficients.

14. An adaptive transform coding method as claimed in claim 7, wherein the step (a) comprises organizing 2^a+b N samples read out of said buffer into (m+1) groups of 2^a blocks of 2^b N samples each, where a+b=m, and a is a variable in a range from m to 0 and b is a variable in a range from 0 to m, and where m is equal to or greater than 2, and N is an integer, each of said groups containing an equal number of samples of same arrivals.

15. An adaptive transform coding method as claimed in claim 7, further comprising deriving respective variances of said blocks, scaling said blocks according to the respective variances, storing said variances into a second memory, and selecting those of said variances which correspond to said optimum block length from said second memory, wherein the step (h) comprising multiplexes multiplexing the selected transform coefficients with a signal representative of the selected variances.

16. An adaptive transform coding method as claimed in claim 7, further comprising generating a signal specifying a predetermined block length, wherein the step (g) comprises selecting those of said transform coefficients which correspond to said predetermined block length and the step (h) comprises multiplexing the transform coefficients selected as corresponding to said predetermined block length with a signal representative of the predetermined block length instead of with the signal representative of the optimum block length.

17. An adaptive transform coding method comprising:

a) storing subblocks of input digital samples representing an audio-frequency signal in a buffer at frame intervals, reading the stored samples from the buffer and segmenting the samples from the buffer into a plurality of blocks each having an integral multiple of the length of said subblock;

b) storing said blocks in a memory and detecting interblock differences between successive blocks of equal length;

c) squaring each of said interblock differences;

d) summing those of said squared interblock differences which correspond to equal block length to produce a set of squared sums of said interblock differences for each block length;

e) detecting representative values of the squared sums from successive ones of said sets, detecting a variation between said representative values and identifying a block length corresponding to said variation as an optimum block length;

f) selecting those of said blocks having said optimum block length from said memory;

g) linearly encoding the selected blocks into corresponding transform coefficients; and

h) multiplexing said transform coefficients with a block length signal representative of said optimum block length.

18. An adaptive transform coding method as claimed in claim 17, wherein the step (e) comprises the steps of:

detecting a maximum value of the squared sums from each set; and

comparing said maximum values detected from a successive pair of said sets so that said compared values correspond respectively to longer and shorter block lengths and selecting one of the compared values as corresponding to said variation if the maximum value of the longer block length has a smaller maximum value than the maximum value of a shorter block length.

19. An adaptive transform coding method as claimed in claim 18, wherein the step (e) further comprises the steps of:

detecting those of said squared sums from each frame interval which corresponds to a minimum block length;

dividing each of the detected squared sums with the detected squared sum of an adjacent occurrence to produce a plurality of ratios from each frame interval; and

detecting one of said ratios as corresponding to said variation if said ratio exceeds a prescribed value.

20. An adaptive transform coding method as claimed in claim 17, wherein the step (e) comprises the steps of:

detecting a maximum value of the squared sums from each set;

dividing one of the maximum values of a successive pair by the other maximum value of the pair and deriving a plurality of ratios from a plurality of said pairs; and

detecting a minimum value from said ratios as corresponding to said variation.

21. An adaptive transform coding method as claimed in claim 20, wherein the step (e) further comprises the steps of:

detecting those of said squared sums from each frame interval which correspond to a minimum block length;

dividing each of the detected squared sums with the detected squared sum of an adjacent occurrence to produce a plurality of ratios from each frame interval; and

detecting one of said ratios as corresponding to said variation if said ratio exceeds a prescribed value.

22. An adaptive transform coding method as claimed in claim 17, further comprising the steps of determining a step size from the selected transform coefficients and quantizing said transform coefficients according to said step size, and wherein the step (h) comprises multiplexing quantized transform coefficients with a signal representative of said step size.

23. An adaptive transform coding method as claimed in claim 22, wherein said step size is determined by squaring the selected transform coefficients, taking an average value of a group of the squared transform coefficients of neighboring occurrences as a representative value of said group, providing interpolations between successive ones of said squared transform coefficients, and deriving said step size from the interpolated squared transform coefficients.

24. An adaptive transform coding method as claimed in claim 17, wherein the step (a) comprises reading samples out of said buffer and organizing 2^a+b N samples read out of said buffer into (m+1) groups of 2^a blocks of 2^b N samples each, where a+b=m, and a is a variable in a range from m to O and b is a variable in a range from O to m, and where m is equal to or greater than 2, and N is an integer, each of said groups containing an equal number of samples of same arrivals.

25. An adaptive transform coding method as claimed in claim 17, further comprising deriving respective variances of said blocks, scaling said blocks according to the respective variances, storing said variances into a second memory, and selecting those of said variances which correspond to said optimum block length from said second memory, wherein the step (h) comprises multiplexing the selected transform coefficients with a signal representative of the selected variances.

26. An adaptive transform coding method as claimed in claim 17, further comprising generating a signal specifying a predetermined block length, wherein the step (f) comprises selecting those blocks having said predetermined block length and the step (h) comprises multiplexing the transform coefficients selected as corresponding to said predetermined block length with a signal representative of the predetermined block length instead of with the signal representative of the optimum block length.

27. A data compression system comprising:

a buffer;

buffer control means for storing subblocks of input digital samples representing an audio-frequency signal in said buffer at predetermined intervals, reading the stored samples from the buffer and segmenting the samples from the buffer into a plurality of blocks each having an integral multiple of the length of said subblock;

a transform encoder for encoding each of said blocks into corresponding transform coefficients;

a memory for storing said transform coefficients therein;

optimum block length determination means for squaring each of said transform coefficients, summing those of said squared transform coefficients which correspond to high-frequency components of said input digital samples to produce squared sums, detecting a minimum of said squared sums as corresponding to an optimum block length, and selecting those of said transform coefficients which correspond to said optimum block length from said memory;

multiplexer means for multiplexing the selected transform coefficients with a signal representative of said optimum block length into a multiplex signal;

demultiplexer means for demultiplexing the multiplex signal into said transform coefficients and said block length signal; and

a transform decoder coupled to said demultiplexer means for decoding said demultiplexed transform coefficients in a process inverse to said transform encoder at intervals corresponding to the block length indicated by said demultiplexed block length signal.

28. A data compression system as claimed in claim 27, further comprising:

means for deriving variances of said blocks respectively and scaling the blocks with the respective variances; and

a second memory for storing said variances,

wherein said optimum block length determination means selects those of said variances which correspond to said optimum block length, and said multiplexer means multiplexes a signal representative of the selected variance with said selected transform coefficients.

29. A data compression system as claimed in claim 27, further comprising means for determining a step size from the selected transform coefficients and means for quantizing said selected transform coefficients according to said step size and causing said multiplexer means to multiplex a signal representative of said step size with the quantized transform coefficients.

30. A data compression system as claimed in claim 29, wherein said step size determination means comprises means for squaring said selected transform coefficients and taking an average value of a group of the squared transform coefficients of neighboring occurrences as a representative value of said group and causing said multiplexer means to multiplex the selected transform coefficients with a signal representative of said average value as said step size representative signal, means for providing interpolations between successive ones of said squared transform coefficients, deriving said step size from the interpolated squared transform coefficients and applying a signal representative of the derived step size to said quantizer means.

31. A data compression system as claimed in claim 27, wherein said buffer control means organizes 2^a+b N samples read out of said buffer into (m+1) groups of 2^a blocks of 2^b N samples each, where a+b=m, and a is a variable in a range from m to 0 and b is a variable in a range from 0 to m, and where m is equal to or greater than 2, and N is an integer, each of said groups containing an equal number of samples of same arrivals.

32. A data compression system as claimed in claim 31, further comprising a normalizer for deriving respective variances of said blocks, and scaling said blocks according to the respective variances, a second memory for storing said variances therein, wherein said optimum block length determination means selects those variances corresponding to said optimum block length from said second memory, and wherein said multiplexer means multiplexes the selected transform coefficients with a signal representative of the selected variances.

33. A data compression system as claimed in claim 27, further comprising means for generating a signal specifying a predetermined block length and means for selecting those of said transform coefficients which correspond to said predetermined block length from said memory and causing said multiplexer means to multiplex the transform coefficients selected as corresponding to said predetermined block length with a signal representative of said predetermined block length instead of with the signal representative of the optimum block length.

34. A data compression system comprising:

a buffer;

buffer control means for storing subblocks of input digital samples representing an audio-frequency signal in said buffer at frame intervals, reading the stored samples from the buffer and segmenting the samples from the buffer into a plurality of blocks each having an integral multiple of the length of said subblock;

a transform encoder for encoding said blocks into respective transform coefficients;

a memory for storing said transform coefficients therein;

optimum block length determination means for detecting interblock differences between successive transform coefficients of equal block length, squaring each of said interblock differences, summing those squared interblock differences corresponding to equal block length to produce a set of said squared sums for each block length, detecting representative values of the squared sums from successive ones of said sets, detecting a variation between said representative values and identifying a block length corresponding to said variation as an optimum block length, and selecting those of the transform coefficients having said optimum block length from said memory;

multiplexer means for multiplexing the selected transform coefficients with a signal representative of said optimum block length into a multiplex signal;

demultiplexer means for demultiplexing the multiplex signal into said transform coefficients and said block length signal; and

a transform decoder coupled to said demultiplexer means for decoding said demultiplexed transform coefficients in a process inverse to said transform encoder at intervals corresponding to the block length indicated by said demultiplexed block length signal.

35. A data compression system as claimed in claim 34, wherein the optimum block length determination means comprises:

means for detecting a maximum value of the squared sums from each set; and

means for comparing said maximum values detected from a successive pair of said sets so that said compared values correspond respectively to longer and shorter block lengths and selecting one of the compared values as corresponding to said variation if the maximum value of the longer block length has a smaller maximum value than the maximum value of a shorter block length.

36. A data compression system as claimed in claim 35, wherein the optimum block length determination means further comprises:

means for detecting those of said squared sums from each frame interval which correspond to a minimum block length;

means for dividing each of the detected squared sums with the detected squared sum of an adjacent occurrence to produce a plurality of ratios from each frame interval; and

means for detecting one of said ratios as corresponding to said variation if said ratio exceeds a prescribed value.

37. A data compression system as claimed in claim 34, wherein the optimum block length determination means comprises:

means for detecting a maximum value of the squared sums from the different sets of each frame interval and deriving a plurality of said maximum values during each frame interval; and

means for forming the maximum value derived from each set with the maximum value of an adjacent set into a pair and deriving a plurality of successive pairs from successive frame intervals;

means for dividing one of the maximum values of each of said pairs by the other maximum value of the pair and deriving a plurality of ratios from said pairs; and

means for detecting a minimum value from said ratios as corresponding to said variation.

38. A data compression system as claimed in claim 37, wherein the optimum block length determination means further comprises:

means for detecting those of said squared sums from each frame interval which correspond to a minimum block length;

means for dividing each of the detected squared sums with the detected squared sum of an adjacent occurrence to produce a plurality of ratios from each frame interval; and

means for detecting one of said ratios as corresponding to said variation if said ratio exceeds a prescribed value.

39. A data compression system as claimed in claim 34, further comprising a normalizer for deriving respective variances of said blocks, and scaling said blocks according to the respective variances, a second memory for storing said variances therein, wherein said optimum block length determination means selects those variances corresponding to said optimum block length from said second memory, and wherein said multiplexer means multiplexes the selected transform coefficients with a signal representative of the selected variances.

40. A data compression system as claimed in claim 34, further comprising means for determining a step size from the selected transform coefficients and means for quantizing said selected transform coefficients according to said step size and causing said multiplexer means to multiplex a signal representative of said step size with the quantized transform coefficients.

41. A data compression system as claimed in claim 40, wherein said step size determination means comprises means for squaring said selected transform coefficients and taking an average value of a group of the squared transform coefficients of neighboring occurrences as a representative value of said group and causing said multiplexer means to multiplex the selected transform coefficients with a signal representative of said average value as said step size representative signal, means for providing interpolations between successive ones of said squared transform coefficients, deriving said step size from the interpolated squared transform coefficients and applying a signal representative of the derived step size to said quantizer means.

42. A data compression system as claimed in claim 34, wherein said buffer control means reads out the samples stored in said buffer and organizes 2^a+b N samples read out of said buffer into (m+1) groups of 2^a blocks of 2^b N samples each, where a+b=m, and a is a variable in a range from m to 0 and b is a variable in a range from 0 to m, and where m is equal to or greater than 2, and N is an integer, each of said groups containing an equal number of samples of same arrivals.

43. A data compression system as claimed in claim 34, further comprising means for generating a signal specifying a predetermined block length and means for selecting those of said transform coefficients which correspond to said predetermined block length from said memory and causing said multiplexer means to multiplex the transform coefficients selected as corresponding to said predetermined block length with a signal representative of said predetermined block length instead of with the signal representative of the optimum block length.

44. A data compression system comprising:

a buffer;

buffer control means for storing subblocks of input digital samples representing an audio-frequency signal into said buffer at predetermined intervals, reading the stored samples from the buffer and segmenting the samples from the buffer into a plurality of blocks each having an integral multiple of the length of said subblock;

a memory for storing said blocks therein;

optimum block length determination means for detecting interblock differences between successive blocks of equal block length, squaring each of said interblock differences, summing those squared interblock differences corresponding to equal block length to produce a set of said squared sums for each block length, detecting representative values of the squared sums from successive ones of said sets, detecting a variation between said representative values and identifying a block length corresponding to said variation as an optimum block length, and selecting those of the blocks having said optimum block length from said memory;

a transform encoder for encoding said selected blocks into respective transform coefficients;

multiplexer means for multiplexing the transform coefficients with a signal representative of said optimum block length into a multiplex signal;

demultiplexer means for demultiplexing the multiplex signal into said transform coefficients and said block length signal; and

a transform decoder coupled to said demultiplexer means for decoding said demultiplexed transform coefficients in a process inverse to said transform encoder at intervals corresponding to the block length indicated by said demultiplexed block length signal.

45. A data compression system as claimed in claim 44, wherein the optimum block length determination means comprises:

means for detecting a maximum value of the squared sums from the different sets of each frame interval; and

means for comparing between said maximum values detected from a successive pair of said sets so that said compared values correspond respectively to longer and shorter block lengths and selecting one of the compared values as corresponding to said variation if the maximum value of the longer block length has a smaller maximum value than the maximum value of a shorter block length.

46. A data compression system as claimed in claim 45, wherein the optimum block length determination means further comprises:

means for detecting those of said squared sums from each frame interval which correspond to a minimum block length;

means for dividing each of the detected squared sums with the detected squared sum of an adjacent occurrence to produce a plurality of ratios from each frame interval; and

means for detecting one of said ratios as corresponding to said variation if said ratio exceeds a prescribed value.

47. A data compression system as claimed in claim 44, wherein the optimum block length determination means comprises:

means for detecting a maximum value of the squared sums from the different sets of each frame interval and deriving a plurality of said maximum values during each frame interval; and

means for forming the maximum value derived from each set with the maximum value of an adjacent set into a pair and deriving a plurality of successive pairs from successive frame intervals;

means for dividing one of the maximum values of each of said pairs by the other maximum value of the pair and deriving a plurality of ratios from said pairs; and

means for detecting a minimum value from said ratios as corresponding to said variation.

48. A data compression system as claimed in claim 47, wherein the optimum block length determination means further comprises:

means for detecting those of said squared sums from each frame interval which correspond to a minimum block length;

means for dividing each of the detected squared sums with the detected squared sum of an adjacent occurrence to produce a plurality of ratios from each frame interval; and

means for detecting one of said ratios as corresponding to said variation if said ratio exceeds a prescribed value.

49. A data compression system as claimed in claim 44, further comprising a normalizer for deriving respective variances of said blocks, and scaling said blocks according to the respective variances, a second memory for storing said variances therein, wherein said optimum block length determination means selects those variances corresponding to said optimum block length from said second memory, and wherein said multiplexer means multiplexes the selected transform coefficients with a signal representative of the selected variances.

50. A data compression system as claimed in claim 44, further comprising means for determining a step size from the transform coefficients derived from said selected blocks and means for quantizing said transform coefficients according to said step size and causing said multiplexer means to multiplex a signal representative of said step size with the quantized transform coefficients.

51. A data compression system as claimed in claim 50, wherein said step size determination means comprises means for squaring said transform coefficients derived from said selected blocks and taking an average value of a group of the squared transform coefficients of neighboring occurrences as a representative value of said group and causing said multiplexer means to multiplex the transform coefficients derived from said selected blocks with a signal representative of said average value as said step size representative signal, means for providing interpolations between successive ones of said squared transform coefficients, deriving said step size from the interpolated squared transform coefficients and applying a signal representative of the derived step size to said quantizer means.

52. A data compression system as claimed in claim 44, wherein said buffer control means organizes 2^a+b N samples read out of said buffer into (m+1) groups of 2^a blocks of 2^b N samples each, where a+b=m, and a is a variable in a range from m to 0 and b is a variable in a range from 0 to m, and where m is equal to or greater than 2, and N is an integer, each of said groups containing an equal number of samples of same arrivals.

53. A data compression system as claimed in claim 44, further comprising means for generating a signal specifying a predetermined block length and means for selecting those of said transform coefficients which correspond to said predetermined block length from said memory and causing said multiplexer means to multiplex the transform coefficients selected as corresponding to said predetermined block length with a signal representative of said predetermined block length instead of with the signal representative of the optimum block length.