Device for the digital recording and reproduction of speech signals

• US 5,251,261 A
• Filed: 12/03/1990
• Issued: 10/05/1993
• Est. Priority Date: 06/15/1990
• Status: Expired due to Fees

• Current Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Original Assignee: US Philips Corporation (Koninklijke Philips N.V.)
• Inventors: Meyer, Peter, Hofmann, Rudolf
• Primary Examiner(s): Shaw, Dale M.
• Assistant Examiner(s): Tung, Kee M.
• Application Number:
  US07/623,870
• Time in Patent Office: 1,037 Days
• Field of Search: 381/29-40
• US Class Current: 704/211
• CPC Class Codes: G10L 19/00 Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis in musical instruments G10H
• CPC Class Codes: G10L 19/12 the excitation function being a code excitation, e.g. in code excited linear prediction [CELP] vocoders
• CPC Class Codes: H04M 1/6505 storing speech in digital form
• CPC Class Codes: H04M 1/651 with speech synthesis

Abstract

Devices for the digital recording and reproduction of speech signals are used, for example in answering apparatus. In order to reduce the quantity of data to be stored without noticeably affecting the acoustic quality of the speech, it is proposed to encode speech signals by means of a residual signal speech encoder.

Claims

1. A process for the digital recording and reproduction in a single device of speech signals characterized by a spectral envelope and a periodicity, comprising:

for the recording, combining digitized speech sampling values of the speech signal in a respective time frame, for each time frame, forming a digital overall code word which accommodates a first code word for the spectral envelope (Short Term Prediction, STP, code word), a second code word for the periodicity (Long Term Prediction, LTP, code word), and a third code word for a residual signal (Regular Pulse Excitation, RPE, code word), storing said digital overall code word in a memory, and reproducing the speech signals from the stored digital overall code word, and wherein the overall code word contains a STP parameter repetition symbol which indicates whether the overall code word contains an STP code word or not, said process further comprising;

forming the STP repetition symbol as a function of a comparison between speech sampling values of a first time frame and speech sampling values of a second, already encoded time frame.

2. A process as claimed in claim 1, further comprising, for the comparison of the speech sampling values, forming a sum of the square distances of an inverse Fourier transform of log-power density spectrum (cepstrum) of the time frames compared.

3. A process as claimed in claim 2 which further comprises, for the reproduction of encoded, stored speech signals, buffering the STP code word, and replacing the buffered STP code word by an STP code word present in the digital overall code word as a function of the STP parameter repetition contained in the digital overall code word.

4. A process as claimed in claim 1 which further comprises, for the reproduction of encoded, stored speech signals, storing the STP code word, and maintaining the stored STP code word or replacing the stored STP code word by and STP code word present in the digital overall code word as a function of the STP parameter repetition symbol contained in the digital overall code word.

5. A process as claimed in claim 1 which further comprises:

recognizing a pause in the speech signal and in response forming a digital overall code word which includes information describing the energy of noise in the pause, the STP parameters describing the spectral properties of the noise, and the time duration of the pause.

6. A process for the digital recording and reproduction in a single device of speech signals characterized by a spectral envelope and a periodicity, comprising:

for the recording, combining digitized speech sampling values of the speech signal in a respective time frame, for each time frame, forming a digital overall code word which accommodates a first code word for the spectral envelope (Short Term Prediction, STP, code word), a second code word for the periodicity (Long Term Prediction, LTP, code word), and a third code word for a residual signal (Regular Pulse Excitation, RPE, code word), storing said digital overall code word in a memory, and reproducing the speech signals from the stored digital overall code word, which process further comprises;

forming an LTP suppression symbol in the digital overall code word which symbol indicates whether the digital overall code word contains an LTP code word or not, the LTP suppression symbol being formed by comparing signals produced before and after an LTP analysis.

7. A process as claimed in claim 6 wherein said signal comparing comprises comparing a ratio of the signal energies after and before the LTP analysis with a presettable threshold value.

8. A process for the digital recording and reproduction in a single device of speech signals characterized by a spectral envelope and a periodicity, comprising:

for the recording, combining digitized speech sampling values of the speech signal in a respective time frame, for each time frame, forming a digital overall code word which accomodates a first code word for the spectral envelope (Short Term Prediction, STP, code word), a second code word for the periodicity (Long Term Prediction, LTP, code word), and a third code word for a residual signal (Regular Pulse Excitation, RPE, code word), storing said digital overall code word in a memory, and reproducing the speech signals from the stored digital overall code word, said process further comprising, forming as a part of the overall code word a symbol (unvoiced symbol) which indicates whether the overall code word comprises other code words besides the first (STP) code word, said symbol being formed as a function of a voiced/unvoiced decision.

9. A device for the recording and reproduction of a speech signal comprising:

short term prediction means for analyzing the speech signal in a first time frame so as to form a first code word that codes a spectral envelope of the speech signal in said first time frame,

long term prediction means for forming a second code word which codes a periodicity of the speech signal in said first time frame and for forming a residual signal representing a residue of the speech signal,

regular pulse excitation coding means for forming a third code word which codes the residual signal in said first time frame,

overall code word forming means coupled to the short term prediction means, the long term prediction means and the regular pulse excitation coding means thereby to form an overall code word for the first time frame, which overall code word comprises the first, second and third code words,

detection means fed by at least one of said short term prediction means, said long term prediction means and said regular pulse excitation coding means, the detection means detecting whether the speech signal in the first time frame meets a suppression condition and thereupon generating a suppression code,

means included in said code word forming means and fed by said detection means for, under control of said suppression code, suppressing any code word signal meeting the suppression condition from the overall code word, any overall code word containing at least a distinctive part of said suppression code,

storage means coupled to said overall code word forming means for storing the overall code word, and

speech synthesizing means for reading the overall code word from said storage means and for, in a second time frame, regenerating the speech signal from the overall code word.

10. A device according to claim 9 further comprising:

comparison means for making a comparison between a first sequence of sample values of the speech signal taken from the first time frame and a second sequence of sample values of the speech signal taken from a third time frame, the third time frame preceding the first time frame in time, the comparison means feeding

repetition symbol forming means for forming a repetition symbol indicating whether the comparison indicates that the first sequence and the second sequence meet a predetermined similarity criterion, the repetition symbol forming means feeding

the overall code word forming means for including the repetition symbol in the overall code word, and including the first code word in the overall code word only if the repetition symbol indicates that the similarity criterion is not met.

11. A device according to claim 10, wherein the comparison means comprise cepstrum distance determining means for determining a sum of squares of differences between first inverse fourier transform values of a first logarithm of a first absolute value of a first fourier transform of the first sequence and second inverse fourier transform values of a second logarithm of a second absolute value of a second fourier transform of the second sequence, the predetermined similarity criterion being that the sum of squares exceeds a predetermined threshold.

12. A device according to claim 10 wherein the speech synthesizing means comprise:

code word retrieval means fed by said storage means for retrieving the first code word and the repetition symbol from the overall code word, the code word retrieval means feeding

first code word buffering means for buffering the first code word, wherein the first code word buffering means update the first code word only if the repetition symbol indicates that the similarity criterion is not met.

13. A device according to claim 9 further comprising:

residue detection means for detecting whether the residue of the speech signal not accounted for by long term prediction meet a predetermined residue criterion, the residue detection means producing a long term suppression symbol indicating whether the residue criterion is met, the residue detection means feeding

the overall code word forming means for including the long term suppression symbol in the overall code word and for including the second code word in the overall code word only if the long term suppression symbol indicates that the residue criterion is not met.

14. A device according to claim 13, wherein the predetermined residue criterion is that a ratio between a first energy of an input signal of the long term prediction means and a second energy of the residue signal exceeds a predetermined threshold.

15. A device according to claim 13 wherein the speech synthesizing means comprise:

code word retrieval means fed by said storage means for retrieving the second code word and the long term suppression symbol from the overall code word, the code word retrieval means feeding

long term prediction synthesis means for synthesizing a regenerated long term predicted signal, the regenerated long term predicted signal being suppressed when the long term suppression symbol indicates that the residue criterion is met.

16. A device according to claim 9 further comprising:

voiced/unvoiced detection means for forming a voiced/unvoiced symbol which indicates whether the speech signal in the first time frame is voiced or unvoiced, the voiced/unvoiced detection means feeding

the overall code word forming means for including the voiced/unvoiced symbol in the overall code word and for allowing the second and third code word to be included in the overall code word only if the voiced/unvoiced symbol indicates that the speech signal is voiced.

17. A device according to claim 16 wherein the speech synthesizing means comprise:

code word retrieval means for retrieving the voiced/unvoiced symbol and the first code word, the code word retrieval means feeding

unvoiced synthesis means, the unvoiced synthesis means being activated when the voiced/unvoiced symbol indicates that the speech signal in the first time frame is unvoiced, the unvoiced synthesis means comprising a noise generator for generating a noise signal having a spectral envelope coded for by the first code word.

18. A device according to claim 9, further comprising:

a pause detector for detecting a pause in the speech signal and for forming a length code word

and an energy content code word estimating respectively a length and a noise energy content of the pause, the pause detector feeding

the overall code word forming means which derives a pause symbol indicating whether a pause is detected in the overall code word and, upon recognition of the pause, including the first code word, the length code word and the energy content code word in the overall code word.

19. A device according to claim 18 wherein the speech synthesizing means comprise:

code word retrieval means fed from said storage means for retrieving the pause symbol, the first code word, the length code word and the energy content code word, the code word retrieval means feeding

pause synthesis means for, when the pause symbol indicates a pause has occurred, forming, during a period of time determined from said length code word, a regenerated speech signal exclusively under control of the energy content code word and the first code word.

20. A device according to claim 9 which includes means for generating under predetermined conditions binary symbols forming the suppression code and which is a part of the overall code word and indicates the exclusion from said overall code word of at least one of said first, second and third code words.

21. A process for the digital recording and reproduction in a single device of speech signals characterized by a spectral envelope and a periodicity, comprising:

for the recording, combining digitized speech sampling values of the speech signal in a respective time frame, for each time frame, forming a digital overall code word which accommodates a first code word for the spectral envelope (Short Term Prediction, STP, code word), a second code word for the periodicity (Long Term Prediction, LTP, code word), and a third code word for a residual signal (Regular Pulse Excitation, RPE, code word), storing said digital overall code word in a memory, and reproducing the speech signals from the stored digital overall code word, said process further comprising;

forming an LTP suppression symbol as a part of the digital overall code word, said LTP suppression symbol indicating whether or not the digital overall code word contains an LTP code word, and in the reproduction of speech signals stored as digital overall code words, the performance of an LTP synthesis being determined by said LTP suppression symbol.

22. A process for the digital recording and reproduction in a single device of speech signals characterized by a spectral envelope and a periodicity, comprising:

for the recording, combining digitized speech sampling values of the speech signal in a respective time frame, for each time frame, forming a digital overall code word which accommodates a first code word for the spectral envelope (Short Term Prediction, STP, code word), a second code word for the periodicity (Long Term Prediction, LTP, code word), and a third code word for a residual signal (Regular Pulse Excitation, RPE, code word), storing said digital overall code word in a memory, and reproducing the speech signals from the stored digital overall code word, said process further comprising;

forming as a pary of the digital overall code word a symbol (unvoiced symbol) which indicates whether the overall code word comprises other code words besides the first (STP) code word, wherein for the reproduction of speech signals, stored as digital overall code words, the process further comprises;

bridging a device for synthesizing speech signals from the periodicity as a function of said unvoiced symbol in the digital overall code word so that the residual signal is reduced to a generated noise signal.

23. A process for the digital recording and reproduction in a single device of speech signals characterized by a spectral envelope and a periodicity, comprising:

for the recording, combining digitized speech sampling values of the speech signal in a respective time frame, for each time frame, forming a digital overall code word which accommodates a first code word for the spectral envelope (Short Term Prediction, STP, code word), a second code word for the periodicity (Long Term Prediction, LTP, code word), and a third code word for a residual signal (Regular Pulse Excitation, RPE, code word), storing said digital overall code word in a memory, and reproducing the speech signals from the stored digital overall code word, which process further comprises;

recognizing a pause in the speech signal and in response forming a digital overall code word which includes a pause symbol, a time code word (T), a residual noise code word (N1) and a residual STP code word, and wherein for the reproduction of stored digital speech signals, a timer is reset in response to the pause signal in the digital overall code word, and via the time code word (T), and during a period of time thus defined forming the speech signal exclusively from the residual noise code word (N1) and the residual STP code word of the digital overall code word.

24. A device for the recording and reproduction of a speech signal comprising:

short term prediction means for analyzing the speech signal in a first time frame so as to form a first code word that codes a spectral envelope of the speech signal in said first time frame,

long term prediction means for forming a second code word which codes a periodicity of the speech signal in said first time frame and forming a residual signal representing a residue of the speech signal,

regular pulse excitation coding means for forming a third code word which codes the residual signal in said first time frame,

overall code word forming means coupled to the short term prediction means, the long term prediction means and the regular pulse excitation coding means thereby to form an overall code word for the first time frame, which overall code word comprises the first, second and third code words,

means responsive to a signal supplied by at least one of said short term prediction means, said long term prediction means and said regular pulse excitation coding means for detecting whether the speech signal in the first time frame meets a suppression condition and if met generating a suppression symbol, and

said code word forming means being controlled by said suppression symbol to suppress from the overall code word formed any code word meeting said suppression condition thereby to reduce the length of the overall code word.

25. A device for the recording and reproduction of a speech signal comprising:

short term prediction means for analyzing the speech signal in a first time frame so as to form a first code word that codes a spectral envelope of the speech signal in said first time frame,

long term prediction means for forming a second code word which codes a periodicity of the speech signal in said first time frame and for forming a residual signal representing a residue of the speech signal,

regular pulse excitation coding means for forming a third code word which codes the residual signal in said first time frame,

overall code word forming means coupled to the short term prediction means, the long term prediction means and the regular pulse excitation coding means thereby to form an overall code word for the first time frame, which overall code word comprises the first, second and third code words, and

means responsive to signals supplied by said short term prediction means, said long term prediction means and said regular pulse excitation coding means for generating, as a part of the overall code word, suppression symbols indicating which, if any, of the first, second and third code words will be suppressed from the overall code word.

26. A device as claimed in claim 25 wherein said suppression symbols generating means comprises:

means for comparing a first sequence of sample values of the speech signal in the first time frame and a second sequence of sample values of the speech signal in a third time frame preceding the first time frame in time, whereby as a result of the comparison, an STP suppression symbol is formed by the suppression symbols generating means indicating the presence or absence of the first code word in the overall code word.

27. A device according to claim 26 wherein the speech synthesizing means comprise:

code word retrieval means fed by said storage means for retrieving the first code word and the STP symbol from the overall code word,

first code word storage means coupled to said code word retrieval means for buffering the first code word, wherein the first code word storage means update the first code word as a function of the STP suppression symbol in the overall code word.

28. A device as claimed in claim 26 wherein said suppression symbols generating means further comprises:

second means for comparing signals indicative of the first and second code words thereby to drive an LTP suppression symbol as a part of said overall code word, said LTP suppression symbol indicating the presence or absence of the second code word in the overall code word.

29. A device as claimed in claim 28 wherein the speech synthesizing means is responsive to STP and LTP suppression symbols indicating the absence of the first and second code words, respectively, from the overall code word thereby to omit a speech synthesis thereof.

30. A device as claimed in claim 28 wherein said suppression symbols generating means further comprises:

means responsive to an output of the first comparing means for generating a voiced/unvoiced symbol in the overall code word, said voiced/unvoiced symbol indicating whether the speech signal in the first time frame is voiced or unvoiced, and wherein

the speech synthesizing means includes means responsive to the voice/unvoiced symbol in the overall code word for generating a noise signal if said voiced/unvoiced symbol indicates that the speech signal in the first time frame is unvoiced.

31. A device as claimed in claim 28 further comprising:

means responsive to the speech signal and to an output of the first comparing means for detecting a pause in the speech signal and for deriving a time duration code word (T) and an energy content code word (N1) indicating the time duration and energy content, respectively, of the pause, and wherein

the overall code word forming means is controlled by the pause detecting means to derive a pause symbol in the overall code word whereby, if a pause is detected, the overall code word comprises the first code word, the time duration code word and the energy content code word, but omits the second and third code words.