Speech message code modifying arrangement

• US 4,709,390 A
• Filed: 05/04/1984
• Issued: 11/24/1987
• Est. Priority Date: 05/04/1984
• Status: Expired due to Term

• Current Assignee: Bell Telephone Laboratories, Inc. (Nokia Corporation)
• Original Assignee: American Telephone & Telegraph Company (AT&T, Inc.)
• Inventors: Caspers, Barbara E., Atal, Bishnu S.
• Primary Examiner(s): KEMENY, EMANUEL
• Application Number:
  US06/607,164
• Time in Patent Office: 1,299 Days
• Field of Search: 381/29-40, 381/51-53, 364/513.5
• US Class Current: 704/262
• CPC Class Codes: G10L 19/10 the excitation function being a multipulse excitation
• CPC Class Codes: G10L 21/04 Time compression or expansion

Abstract

Natural quality and bit-rate for LPC speech synthesis is improved by encoding the LPC residual signal in a prescribed multipulse format formed for each LPC frame. Voiced, unvoiced, and mixed (hiss plus periodic) excitation is inherent. The speaking-rate is changed by adding, deleting, or repeating pitch-periods, and the pitch (intonation) is changed by adding or deleting zeros in the multipulse excitation signal.

Claims

1. Apparatus for coding a speech pattern comprising:

means for partitioning said speech pattern into successive time frame portions;

means responsive to each successive time frame portion of the speech pattern for generating speech parameter signals comprising a set of linear predictive parameter type spectral representative signals and an excitation signal comprising a sequence of excitation pulses each of amplitude beta and location m within said time frame;

means responsive to the frame speech parameter signals for identifying successive intervals of said speech pattern as voiced or other than voiced, each voiced interval being a plurality of time frame portions coextensive with a pitch period of said speech pattern and each other than voiced interval comprising a time frame portion of said speech pattern; and

means for modifying the excitation signals of each successive identified voiced interval to compress the speech pattern excitation signals of said speech pattern;

said modifying means including;

means responsive to each other than voiced interval for forming an excitation signal comprising the sequence of excitation pulses of the time frame portion of the other than voiced interval;

means responsive to the occurrence of a succession of identified voiced intervals for forming an excitation signal comprising the sequence of excitation pulses of the pitch period of a selected one of said succession of identified voiced intervals; and

means for forming an excitation signal for each of the remaining voiced intervals of said succession of identified voiced intervals comprising a coded signal repeating the sequence of excitation signals of the pitch period of said selected identified voiced interval.

2. Apparatus for coding a speech pattern according to claim 1 wherein:

the means for selecting of one of a sequence of successive voiced excitation signal intervals comprises means for selecting the first of a succession of voiced excitation signal intervals; and

said substituting means comprises means for generating a predetermined code and for replacing the excitation signals of the remaining succession of voiced excitation intervals with said predetermined code.

3. Apparatus for coding a speech pattern according to claim 1 further comprising means for generating a signal for editing the excitation signals of said voiced excitation signal intervals;

and

said modifying means comprises means responsive to said predetermined pattern editing signal for altering the excitation signals of the voiced excitation signal intervals.

4. Apparatus for coding a speech pattern according to claim 3 wherein said editing signal comprises a signal for changing the duration of the voiced excitation signal intervals and said modifying means comprises means responsive to said duration change signal for altering the excitation signal of each voiced excitation signal interval to effect a change in speaking rate.

5. Apparatus for coding a speech pattern according to claim 3 wherein said editing signal comprises a succession of duration changing signals and said modifying means comprises means responsive to the succession of duration change signals for altering the excitation signal of the excitation signal intervals of said speech pattern to effect of a change in intonation of said speech pattern.

6. A method for altering a speech message comprising the steps of:

generating a time frame sequence of speech parameter signals representative of a speech message, each time frame speech parameter signal including a set of spectral representative signals and an excitation signal comprising a sequence of excitation pulses of varying amplitudes and varying locations within the time frame;

generating a sequence of speech message time frame editing signals;

identifying a succession of prescribed type excitation signal intervals, said succession being identified in response to groups of the time frame speech parameter signals having various pitch periods;

modifying the excitation and spectral representative signals of the frames of the prescribed type excitation signal intervals in response to said speech message editing signals; and

forming an edited speech message responsive to the modified excitation and spectral representative signals.

7. A method for altering a speech message according to claim 6 wherein:

the speech message editing signal generating step comprises generating a signal representative of a prescribed speaking rate;

said prescribed type of excitation signal interval is a voiced excitation signal interval;

and said modifying step comprises modifying the number of pitch periods employed to constitute each voiced excitation signal interval in response to said prescribed speaking rate editing signal.

8. A method for altering a speech message according to claim 6 wherein:

said prescribed type excitation signal interval is a voiced excitation signal interval;

the speech message editing signal generating step comprises generating a sequence of voiced interval duration changing signals; and

said modifying step comprises altering the duration of the succession of voiced excitation signal intervals responsive to said duration changing speech message editing signals to modify the intonation pattern of the speech message.

9. Apparatus for altering a speech message comprising:

means responsive to the speech message for generating a time frame sequence of speech parameter signals representative of a speech message, each time frame speech parameter signal including a set of spectral representative signals and an excitation signal of the multi-pulse type;

means responsive to the time frame speech parameter signals for identifying a succession of pitch period signal intervals;

means for generating a sequence of speech message time frame editing signals responsive in part to the identifying means;

means responsive to said speech message editing signals for increasing the repetitiveness of at least some of the excitation and spectral representative signals of the frames of the pitch period signal intervals; and

means responsive to the modified excitation and spectral representative signals for forming an edited speech message.

10. Apparatus for altering a speech message according to claim 9 wherein:

the speech message editing signal generating means comprises means for generating a signal representative of a prescribed speaking rate;

said prescribed type of excitation signal interval is a voiced excitation signal interval;

and said modifying means comprises means responsive to said prescribed speaking rate editing signal for changing the number of pitch periods representing each voiced excitation signal interval.

11. Apparatus for altering a speech message according to claim 9 wherein:

said prescribed type excitation signal interval is a voiced excitation signal interval;

the speech message editing signal generating means comprises means for generating a sequence of voiced interval duration changing signals; and

said modifying means comprises means responsive to said duration changing speech message editing signals for altering the duration of the succession of voiced excitation signal intervals to change the intonation pattern of the speech message.

12. A method for altering a speech message coded as a sequence of time frame spectral representative signals and multi-pulse excitation signals comprising the steps of:

generating a predetermined speech message editing signal;

identifying prescribed type intervals in the excitation signal sequence of the coded speech message; and

increasing the repetitiveness of the multi-pulse excitation signals of selected prescribed type intervals responsive to said speech message editing signal.

13. A method for altering a speech message according to claim 12 wherein:

said speech message editing signal comprises an interval repeat signal; and

said modifying step comprises detecting a sequence of successive prescribed type excitation signal intervals, selecting one of said successive prescribed type excitation signal intervals, and substituting the excitation signal of the selected interval for the excitation signals of the remaining intervals of the sequence responsive to said interval repeat signal.

14. A method for altering a speech message according to claim 13 wherein:

said speech message editing signal comprises a speaking rate change signal; and

said modifying step comprises detecting prescribed type excitation signal intervals in said coded speech message, and changing the number of time frames of the excitation signals of said detected intervals responsive to said speaking rate change signal.

15. A method for altering a speech message according to claim 13 wherein:

said speech message editing signal comprises a sequence of pitch frequency modifying signals; and

said modifying step comprises detecting the successive prescribed type excitation signal intervals, and changing the duration of the excitation signals of successive detected intervals responsive to the sequence of pitch frequency modifying signals.

16. A method for altering a speech message according to claims 12, 13, 14, or 15 wherein the prescribed type excitation signal intervals are voiced intervals of the speech message.

17. A method for altering a speech message according to claims 12, 13, 14, or 15 wherein each time frame excitation signal corresponds to the linear predictive residual of the time frame.

18. Apparatus for altering a speech message coded as a time frame sequence of spectral representative and multi-pulse excitation signals comprising:

means for generating a predetermined speech message editing signal;

means responsive to said speech message spectral representative and excitation signals for identifying prescribed type sequential intervals of the at-least-partially voiced type in the excitation signal sequence of the coded speech message; and

means responsive to said speech message editing signal for increasing the repetitiveness of the excitation signals of the identified prescribed type intervals by repeating a selected group of multi-pulse excitation signals representative of one such interval in the other sequential intervals to reduce the effective bit rate of the resulting coded speech message.

19. Apparatus for altering a speech message according to claim 18 wherein:

said speech message editing signal generating means comprises means for generating an interval repeat signal; and

said modifying means comprises means for detecting a sequence of successive voiced excitation signal intervals, means for selecting one of said successive prescribed type excitation signal intervals, and means responsive to said interval repeat signal for substituting the excitation signal of the selected interval for the excitation signals of the remaining intervals of the sequence.

20. Apparatus for altering a speech message according to claim 18 wherein:

said speech message editing signal generating means comprises means for generating a speaking rate change signal; and

said modifying means comprises means for detecting the prescribed type excitation signal intervals in said coded speech message, and means responsive to said speaking rate change signal for changing the number of time frame portions of the excitation signals of said detected intervals.

21. Apparatus for altering a speech message according to claim 18 wherein:

said speech message editing signal generating means comprises means for generating a sequence of pitch frequency modifying signals; and

said modifying means comprises means for detecting the successive prescribed type excitation signal intervals, and means responsive to said sequence of pitch frequency modifying signals for changing duration of the the excitation signals of successive detected intervals.

22. Apparatus for altering a speech message according to claim 21 in which modifying means comprises means for changing the number of zero-valued pulses in the multi-pulse excitation signal to change its duration.