Speech codec

• US 5,488,704 A
• Filed: 03/15/1993
• Issued: 01/30/1996
• Est. Priority Date: 03/16/1992
• Status: Expired due to Fees

• Current Assignee: Sanyo Electric Company Limited (Panasonic Holdings Corporation)
• Original Assignee: Sanyo Electric Company Limited (Panasonic Holdings Corporation)
• Inventors: Fujimoto, Mitsuo
• Primary Examiner(s): MacDonald, Allen R.
• Assistant Examiner(s): HAFIZ, TARIQ R
• Application Number:
  US08/031,808
• Time in Patent Office: 1,051 Days
• Field of Search: 395/2.2, 395/2.28, 395/2.31, 395/2.3
• US Class Current: 704/219
• 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: G10L 19/125 Pitch excitation, e.g. pitch synchronous innovation CELP [PSI-CELP]
• CPC Class Codes: G10L 25/93 Discriminating between voiced and unvoiced parts of speech signals G10L25/90 takes precedence

Abstract

A speech codec includes a drive sound source generating circuit in which, in a case of a voiced sound speech, a pulse pattern signal corresponding to a pitch-scale, drive sound source signals stored within a newest predetermined past time period and a noise signal are multiplied by predetermined gains, respectively, and then, added to each other so as to generate a drive voiced sound source and, in a case of an unvoiced sound, drive sound source signals stored within a newest predetermined past time period and a noise signal are multiplied by predetermined gains, respectively, and then, added to each other so as to generate a drive unvoiced sound source.

Claims

1. A speech codec, comprising:

pitch extracting means for extracting a pitch-scale on the basis of a signal of an inputted speech;

voiced sound/unvoiced sound determinating means for determinating whether said inputted speech is one of a voiced sound speech and unvoiced sound speech on the basis of the signal of said inputted speech;

drive sound source generating means for generating a drive sound source signal on the basis of said pitch-scale and a determination result of said voiced sound/unvoiced sound determinating means, said drive sound source generating means including drive voiced sound source generating means for generating drive voiced sound source signal when said inputted speech is the voiced sound speech, and unvoiced sound source generating means for generating a drive unvoiced sound source signal when said inputted speech is the unvoiced sound speech, wherein said drive voiced sound source generating means includes pulse pattern generating means for generating a pulse pattern signal corresponding to said pitch-scale;

first adaptive code book means for storing drive voiced sound source signals within a predetermined past time period;

first noise code book means for storing in advance a noise signal; and

first generating means for generating the drive voiced sound source signal on the basis of said pulse pattern signal, an output of said first adaptive code book means and an output of said first noise code book means;

speech synthesizing means for outputting a synthesized speech signal on the basis of one of said drive voiced sound source signal and said drive unvoiced sound source signal generated by said drive sound source generating means; and

code outputting means for selectively outputting a code corresponding to one of said drive voiced sound source signal and said drive unvoiced sound source signal at a timing that an error between said synthesized speech signal and the signal of said inputted speech becomes minimum by comparing the synthesized speech signal and the signal of said inputted speech.

2. A speech codec according to claim 1, wherein said first generating means includes means for generating said drive voiced sound source signal by multiplying said pulse pattern signal, the output of said first adaptive code book means and the output of said first noise code book means by predetermined gains, respectively, and then by mixing the same.

3. A speech codec according to claim 1, wherein said first generating means selectively utilizes any one of said first drive voiced sound source signals stored in said first adaptive code book means within a time range of a predetermined number of the signals associated with said pitch-scale.

4. A speech codec according to claim 1, wherein said drive unvoiced sound source generating means includes second adaptive code book means for storing the drive unvoiced sound source signals within a predetermined past time period;

second noise code book means for storing in advance a noise signal; and

second generating means for generating the drive unvoiced sound source signal on the basis of an output of said second adaptive code book means and an output of said second noise code book means.

5. A speech codec according to claim 4, wherein said second generating means includes means for generating said drive unvoiced sound source signal by multiplying, the output of said second adaptive code book means and the output of said second noise code book means by predetermined gains, respectively, and then by mixing the same.

6. A speech codec according to claim 1, wherein said speech synthesizing means includes voiced sound speech synthesizing means for outputting a voiced sound synthesized speech signal on the basis of said drive voiced sound source, and unvoiced sound speech synthesizing means for outputting an unvoiced sound synthesized speech signal on the basis of said drive unvoiced sound source, and said speech codec further comprising means for outputting said determination result by said voiced sound/unvoiced sound determinating means together with said code.

7. A speech codec according to claim 1, wherein said pitch extracting means includes predictive residual signal outputting means for evaluating a predictive residual signal on the basis of the signal of said inputted speech, and extracts an interval between a preceding pitch pulse position of the drive sound source signal from a succeeding pitch pulse position at which an amplitude value of said predictive residual signal becomes larger than a predetermined value within a predetermined range in the vicinity of a position separated from the preceding pitch pulse position by said pitch-scale.

8. A speech codec according to claim 7, wherein said pitch extracting means includes predictive residual signal outputting means for evaluating a predictive residual signal on the basis of the signal of said inputted speech, and phase equalization means for phase-equalizing said predictive residual signal, and extracts an interval between a preceding pitch pulse position of said drive sound source signal from a peak position of said phase-equalized predictive residual signal as said pitch-scale.

9. A speech codec, comprising:

pitch extracting means for extracting pitch-scale on the basis of a signal of an inputted speech;

analyzing means for analyzing a parameter of said inputted speech;

first speech synthesizing means for synthesizing a voiced sound synthesized speech signal on the basis of said pitch-scale and said parameter, wherein said first speech synthesizing means includes pulse pattern generating means for generating a pulse pattern signal in response to said pitch-scale;

first adaptive code book means for storing a drive voiced sound source signal within a predetermined past time period;

first noise code book means for storing in advance a noise signal; and

first generating means for generating said voiced sound synthesized speech signal on the basis of said pulse pattern signal, an output of said first adaptive code book means and an output of said first noise code book means;

second speech synthesizing means for synthesizing an unvoiced sound synthesized speech signal on the basis of the signal of said inputted speech and said parameter;

similarity determining means for determining that which one of said voiced sound synthesized speech signal and unvoiced sound synthesized speech signal is similar to said inputted speech;

selecting means for selecting one of said voiced sound synthesized speech signal and said unvoiced sound synthesized speech signal in response to an output of said similarity determining means; and

multiplexer means for multiplexing one of said voiced sound synthesized speech signal and said unvoiced sound synthesized speech signal selected by said selecting means and said parameter.

10. A speech codec according to claim 9, wherein said analyzing means includes LPC analyzing means for outputting an LPC parameter of said inputted speech.

11. A speech codec according to claim 9, wherein said first generating means includes first mixing means for mixing said pulse pattern signal, said output of said first adaptive code book means and said output of said first noise code book means with multiplying the same by predetermined gains, respectively;

and first synthesizing filter means for receiving an output of said first mixing means.

12. A speech codec according to claim 9, wherein said first generating means selectively utilizes any one of the past drive voiced sound source signals that are stored in said first adaptive code book means and exists within a time range in association with said pitch-scale.

13. A speech codec according to claim 9, wherein said second speech synthesizing means includes second adaptive code book means for storing a drive unvoiced sound source signal within a predetermined past time period;

second noise code book means for storing in advance a noise signal; and

second generating means for generating said unvoiced sound synthesized speech signal on the basis of an output of said second adaptive code book means and an output of said second noise code book means.

14. A speech codec according to claim 13, wherein said second generating means includes second mixing means for mixing said output of said second adaptive code book means and said output of said second noise code book means with multiplying the same by predetermined gains, respectively;

and second synthesizing filter means for receiving an output of said second mixing means.

15. A speech codec according to claim 9, wherein said pitch extracting means includes predictive residual signal outputting means for evaluating a predictive residual signal on the basis of the signal of said inputted speech, and extracts an interval between a preceding pitch pulse position of said drive sound source signal from a succeeding pitch pulse position at which an amplitude value of said predictive residual signal becomes larger than a predetermined value within a predetermined range in the vicinity of a position separated from the preceding pitch pulse position by said pitch-scale.

16. A speech codec according to claim 15, wherein said pitch extracting means includes predictive residual signal outputting means for evaluating a predictive residual signal on the basis of the signal of said inputted speech, and phase equalization means for phase-equalizing said predictive residual signal, and extracts an interval between a preceding pitch pulse position of said drive sound source signal from a peak position of said phase-equalized predictive residual signal as said pitch-scale.

17. A speech codec according to claim 9, wherein said first speech synthesizing means includes pulse pattern generating means for generating a pulse pattern signal in response to said pitch-scale;

first adaptive code book means for storing a drive voiced sound source signal within a predetermined past time period;

first noise code book means for storing in advance a noise signal; and

first generating means for generating said voiced sound synthesized speech signal on the basis of said pulse pattern signal, an output of said first adaptive code book means and an output of said first noise code book means.

18. A speech codec according to claim 17, wherein the drive voiced sound source data stored in said first adaptive code book is copied into said second adaptive code book when said voiced sound synthesized speech signal is selected by said selecting means, and the drive unvoiced sound source data stored in said second adaptive code book is copied into said first adaptive code book when said unvoiced sound synthesized speech signal is selected by said selecting means.