Method and apparatus for hybrid coding of speech at 4kbps

• US 6,233,550 B1
• Filed: 08/28/1998
• Issued: 05/15/2001
• Est. Priority Date: 08/29/1997
• Status: Expired due to Term

• Current Assignee: Regents of the University of California (University of California)
• Original Assignee: Regents of the University of California (University of California)
• Inventors: Shlomot, Eyal, Cuperman, Vladimir, Gersho, Allen, Li, Chunyan
• Primary Examiner(s): {haeck over (S)}mits, Ta̅livaldis I.
• Application Number:
  US09/143,265
• Time in Patent Office: 991 Days
• Field of Search: 704/208, 704/214, 704/219, 704/220
• US Class Current: 704/208
• CPC Class Codes: G10L 19/02 using spectral analysis, e.g. transform vocoders or subband vocoders
• CPC Class Codes: G10L 19/09 Long term prediction, i.e. removing periodical redundancies, e.g. by using adaptive codebook or pitch predictor
• CPC Class Codes: G10L 19/10 the excitation function being a multipulse excitation
• CPC Class Codes: G10L 19/18 Vocoders using multiple modes
• CPC Class Codes: G10L 2025/935 Mixed voiced class; Transitions

Abstract

A method and apparatus for encoding speech for communication to a decoder for reproduction of the speech where the speech signal is classified into steady state voiced (harmonic), stationary unvoiced, and “transitory” or “transition” speech, and a particular type of coding scheme is used for each class. Harmonic coding is used for steady state voiced speech, “noise-like” coding is used for stationary unvoiced speech, and a special coding mode is used for transition speech, designed to capture the location, the structure, and the strength of the local time events that characterize the transition portions of the speech. The compression schemes can be applied to the speech signal or to the LP residual signal.

Claims

1. A hybrid speech encoding method, comprising the steps of:

(a) classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) using harmonic coding to compress frames associated with at least one of said classes;

(c) coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding; and

(d) phase aligning a harmonic coded frame in a decoder when the preceding frame has been waveform encoded for pairs of adjacent frames comprising a waveform encoded frame adjacent to a harmonic coded frame.

2. A hybrid speech encoding method, comprising the steps of:

(a) classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) using harmonic coding to compress frames associated with at least one of said classes;

(c) coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding; and

(d) phase aligning a waveform encoded frame in a decoder when the preceding frame has been harmonic coded for pairs of adjacent frames comprising a waveform encoded frame adjacent to a harmonic coded frame.

3. A hybrid speech encoding method, comprising the steps of:

(a) classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) using harmonic coding to compress frames associated with at least one of said classes;

(c) coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding; and

(d) phase aligning the frame in an encoder to be waveform encoded when the subsequent frame is to be harmonic encoded for pairs of adjacent frames comprising a waveform encoded frame adjacent to a harmonic coded frame.

4. A hybrid speech encoding method, comprising the steps of:

(a) classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) using harmonic coding to compress frames associated with at least one of said classes;

(c) coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding; and

(d) phase aligning the frame in an encoder to be harmonic coded when the subsequent frame is to be waveform encoded for pairs of adjacent frames comprising a waveform encoded frame adjacent to a harmonic coded frame.

5. A hybrid speech encoding method, comprising the steps of:

(a) classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) using harmonic coding to compress frames associated with at least one of said classes;

(c) coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding;

(d) phase aligning a harmonic coded frame in a decoder when the preceding frame has been waveform coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame; and

(e) phase aligning the frame in an encoder to be waveform coded when the subsequent frame is to be harmonic coded for pairs of adjacent frames comprising a waveform coded frame followed by a harmonic coded frame.

6. A method for encoding speech in an encoder for communication to a decoder for reproduction thereof, said speech comprising a plurality of frames of speech, said method comprising the steps of:

(a) classifying each frame of speech into three or more classes wherein one or more of said classes is transitory in character;

(b) representing the speech in a frame of speech associated with at least one of said classes with a harmonic model;

(c) computing parameter values of said harmonic model where said parameter values are characteristic of the frame;

(d) quantizing said parameters for communication to said decoder;

(e) wherein one or more of said transitory classes is encoded using a coding technique selected from the group consisting of waveform-matching coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding; and

(f) phase aligning the reproduced speech across the boundary between two successive frames of speech where one frame of speech is waveform coded and the other frame of speech is harmonic coded.

7. A method as recited in claim 6, further comprising the step of phase aligning a harmonic coded frame of speech in the decoder when the preceding frame of speech has been waveform coded for pairs of adjacent frames of speech comprising a waveform coded frame of speech adjacent to a harmonic coded frame of speech.

8. A method as recited in 6, further comprising the step of phase aligning the frame in the encoder to be waveform coded when the subsequent frame is to be harmonic coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame.

9. A hybrid method of encoding speech in an encoder for transmission to a decoder for reproduction thereof, comprising the steps of:

(a) classifying frames of the speech signal into steady state voiced, stationary unvoiced, or transitory speech segments;

(b) coding a frame with harmonic coding if the frame is classified as steady state voiced speech;

(c) coding a frame with “

noise-like”

coding if the frame is classified as stationary unvoiced speech;

(d) coding a frame classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding;

(e) phase aligning the reproduced speech across the boundary between two successive frames of speech where one frame of speech is waveform coded and the other frame of speech is harmonic coded.

10. A method as recited in claim 9, further comprising the step of phase aligning a harmonic coded frame of speech in the decoder when the preceding frame of speech has been waveform coded for pairs of adjacent frames of speech comprising a waveform coded frame of speech adjacent to a harmonic coded frame of speech.

11. A method as recited in claim 9, further comprising the step of phase aligning the frame in the encoder to be waveform coded when the subsequent frame is to be harmonic coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame.

12. A hybrid speech encoder, comprising:

(a) means for classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) means for harmonic coding frames associated with at least one of said classes;

(c) means for coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding; and

(d) means for phase aligning a harmonic coded frame in a decoder when the preceding frame has been waveform encoded for pairs of adjacent frames comprising a waveform encoded frame adjacent to a harmonic coded frame.

13. A hybrid speech encoder, comprising:

(a) means for classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) means for harmonic coding frames associated with at least one of said classes;

(c) means for coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding;

(d) means for phase aligning a waveform encoded frame in a decoder when the preceding frame has been harmonic coded for pairs of adjacent frames comprising a waveform encoded frame adjacent to a harmonic coded frame.

14. A hybrid speech encoder, comprising:

(a) means for classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) means for harmonic coding frames associated with at least one of said classes;

(c) means for coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding; and

(d) means for phase aligning the frame in an encoder to be waveform encoded when the subsequent frame is to be harmonic encoded for pairs of adjacent frames comprising a waveform encoded frame adjacent to a harmonic coded frame.

15. A hybrid speech encoder, comprising:

(a) means for classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) means for harmonic coding frames associated with at least one of said classes;

(c) means for coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding; and

(d) phase aligning the frame in an encoder to be harmonic coded when the subsequent frame is to be waveform encoded for pairs of adjacent frames comprising a waveform encoded frame adjacent to a harmonic coded frame.

16. A hybrid speech encoder, comprising:

(a) means for classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) means for harmonic coding frames associated with at least one of said classes;

(c) means for coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding;

(d) means for phase aligning a harmonic coded frame in a decoder when the preceding frame has been waveform coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame; and

(e) means for phase aligning the frame in an encoder to be waveform encoded when the subsequent frame is to be harmonic coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame.

17. A hybrid speech encoder for communication to a decoder for reproduction of speech, said speech comprising a plurality of frames of speech, said encoder comprising:

(a) means for classifying each frame of speech into three or more classes wherein one or more of said classes is transitory in character;

(b) means for representing the speech in a frame of speech associated with at least one of said classes with a harmonic model;

(c) means for computing parameter values of said harmonic model where said parameter values are characteristic of the frame;

(d) means for quantizing said parameters for communication to said decoder;

(e) wherein one or more of said transitory classes is encoded using a coding technique selected from the group consisting of waveform-matching coding, anaylsis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding; and

(f) means for phase aligning the reproduced speech across the boundary between two successive frames of speech where one frame of speech is waveform coded and the other frame of speech is harmonic coded.

18. A hybrid speech encoder as recited in claim 17, further comprising means for phase aligning a harmonic coded frame of speech in the decoder when the preceding frame of speech has been waveform coded for pairs of adjacent frames of speech comprising a waveform coded frame of speech adjacent to a harmonic coded frame of speech.

19. A hybrid speech encoder as recited in claim 17, further comprising means for phase aligning the frame in the encoder to be waveform coded when the subsequent frame is to be harmonic coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame.

20. An apparatus for encoding speech for transmission to a decoder for reproduction thereof, comprising:

(a) means for classifying frames of the speech signal as steady state voiced, stationary unvoiced, or transitory speech;

(b) means for coding a frame with harmonic coding if the frame is classified as steady state voiced speech;

(c) means for coding a frame with “

noise-like”

coding if the frame is classified as stationary unvoiced speech;

(d) means for coding a frame classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding; and

(e) means for phase aligning the reproduced speech across the boundary between two successive frames of speech where one frame of speech is waveform coded and the other frame of speech is harmonic coded.

21. An apparatus as recited in claim 20, further comprising means for phase aligning a harmonic coded frame of speech in the decoder when the preceding frame of speech has been waveform coded for pairs of adjacent frames of speech comprising a waveform coded frame of speech adjacent to a harmonic coded frame of speech.

22. An apparatus as recited in claim 20, further comprising means for phase aligning the frame in the encoder to be waveform coded when the subsequent frame is to be harmonic coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame.

23. A hybrid speech encoder, comprising:

(a) a speech classifier, said speech classifier classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) a harmonic encoder, said harmonic encoder configured for harmonic coding of frames associated with at least one of said classes;

(c) a transitory encoder, said transitory encoder coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding;

(d) a first phase synchronizer, said first phase synchronizer phase aligning a harmonic coded frame in a decoder when the preceding frame has been waveform coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame; and

(e) a second phase synchronizer, said second phase synchronizer phase aligning the frame in an encoder to be waveform coded when the subsequent frame is to be harmonic coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame.

24. A hybrid speech encoder, comprising:

(a) a speech classifier, said speech classifier classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) a encoder for voiced signals;

(c) an encoder for unvoiced signals;

(d) an encoder for transitory signals;

(e) wherein at one of said encoders comprises a harmonic encoder, and wherein at least one of said encoders comprises an encoder selected from the group consisting of a waveform encoder, an analysis-by-synthesis encoder, a codebook excited linear prediction analysis-by-synthesis encoder, and a multipulse analysis-by-synthesis encoder;

(f) a first phase synchronizer, said first phase synchronizer phase aligning a harmonic coded frame in a decoder when the preceding frame has been waveform coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame; and

(g) a second phase synchronizer, said second phase synchronizer phase aligning the frame in an encoder to be waveform coded when the subsequent frame is to be harmonic coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame.

25. A hybrid speech compression system, comprising:

(a) a speech classifier, said speech classifier classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) a harmonic encoder, said harmonic encoder configured for harmonic coding of frames associated with at least one of said classes;

(c) a transitory encoder, said transitory encoder coding frames classified as transitory using a coding technique selected from the group consisting of waveform coding, analysis-by-synthesis coding, codebook excited linear prediction analysis-by-synthesis coding, and multipulse analysis-by-synthesis coding;

(d) a harmonic decoder;

(e) a transitory decoder, said transitory decoder decoding frames of speech classified as transitory using a decoding technique selected from the group consisting of waveform decoding, analysis-by-synthesis decoding, codebook excited linear prediction analysis-by-synthesis decoding, and multipulse analysis-by-synthesis decoding;

(f) a first phase synchronizer, said first phase synchronizer phase aligning a harmonic coded frame in a decoder when the preceding frame has been waveform coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame; and

(g) a second phase synchronizer, said second phase synchronizer phase aligning the frame in an encoder to be waveform coded when the subsequent frame is to be harmonic coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame.

26. A hybrid speech compression system, comprising:

(a) a speech classifier, said speech classifier classifying frames of speech signals as voiced, unvoiced, or transitory;

(b) an encoder for voiced signals;

(c) an encoder for unvoiced signals;

(d) an encoder for transitory signals;

(e) wherein at one of said encoders comprises a harmonic encoder, and wherein at least one of said encoders comprises an encoder selected from the group consisting of a waveform encoder, an analysis-by-synthesis encoder, a codebook excited linear prediction analysis-by-synthesis encoder, and a multipulse analysis-by-synthesis encoder;

(f) a decoder for speech signals classified as voiced signals;

(g) a decoder for speech signals classified as unvoiced signals;

(h) a decoder for speech signals classified as transitory signals;

(i) wherein at one of said decoders comprises a harmonic decoder, and wherein at least one of said decoders comprises a decoder selected from the group consisting of a waveform decoder, an analysis-by-synthesis decoder, a codebook excited linear prediction analysis-by-synthesis decoder, and a multipulse analysis-by-synthesis decoder;

(j) a first phase synchronizer, said first phase synchronizer phase aligning a harmonic coded frame in a decoder when the preceding frame has been waveform coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame; and

(k) a second phase synchronizer, said second phase synchronizer phase aligning the frame in an encoder to be waveform encoded when the subsequent frame is to be harmonic coded for pairs of adjacent frames comprising a waveform coded frame adjacent to a harmonic coded frame.