Code excited linear prediction speech coding system

• US 5,261,027 A
• Filed: 12/28/1992
• Issued: 11/09/1993
• Est. Priority Date: 06/28/1989
• Status: Expired due to Term

• Current Assignee: Fujitsu Limited
• Original Assignee: Fujitsu Limited
• Inventors: Tanaka, Yoshinori, Taniguchi, Tomohiko, Amano, Fumio, Ohta, Yasuji, Unagami, Shigeyuki, Sasama, Akira
• Primary Examiner(s): Fleming, Michael R.
• Assistant Examiner(s): Doerrler, Michelle
• Application Number:
  US07/997,667
• Time in Patent Office: 316 Days
• Field of Search: 381/29-40, 381/46-47, 381/51, 395/2
• US Class Current: 704/200
• 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 2019/0003 Backward prediction of gain
• CPC Class Codes: G10L 2019/0005 Multi-stage vector quantisation
• CPC Class Codes: G10L 2019/0011 Long term prediction filters, i.e. pitch estimation
• CPC Class Codes: G10L 25/93 Discriminating between voiced and unvoiced parts of speech signals G10L25/90 takes precedence

Abstract

A code excited linear prediction (CELP) type speech signal coding system is provided, a code vector obtained by applying linear prediction to a vector of a residual speech signal of white noise is stored in a code book. A pitch prediction vector obtained by applying linear prediction to a residual signal of a preceding frame is given a delay corresponding to a pitch frequency and added to the code vector. Use is made of an impulse vector obtained by applying linear prediction to a residual signal vector of impulses having a predetermined relationship with the vectors of the white noise code book. Variable gains are given to at least the above code vector and impulse vector, a reproduced signal is produced, and this reproduced signal is used for identification of the input speech signal. Thus, a pulse series corresponding to the sound source of voiced speech sounds is created.

Claims

1. A method of encoding and transmitting an input speech signal by code excited linear prediction type encoding to provide a decodable signal, said method comprising the steps of:

(a) providing a residual signal vector from a white noise code book, based on an error signal so as to reduce the error signal,

(b) applying linear prediction to the white noise residual signal vector to obtain a code vector and a first coefficient,

(c) applying linear prediction to a residual signal of a previous speech signal delayed by a pitch frequency to obtain a pitch prediction vector and a second coefficient,

(d) providing an impulse residual signal vector having a predetermined relationship with the residual signal vector from the white noise code book,

(e) applying linear prediction to the impulse residual signal vector provided in step (d) to obtain an impulse vector and a third coefficient,

(f) applying variable gains to at least the code vector obtained by said step (b) and the impulse vector obtained by said step (e),

(g) adding the code, pitch prediction and impulse vectors after applying the variable gains in step (f) to form a reproduced signal,

(h) evaluating a difference between the reproduced signal formed by said step (g) and the input speech signal to provide the error signal for said step (a), and

(i) transmitting a decodable signal based on at least the first, second and third coefficients.

2. A method according to claim 1, wherein respective impulse residual signal vectors provided in said step (d) correspond to the residual signal vectors of the white noise code book.

3. A method according to claim 2, wherein the impulse residual signal vector provided in step (d) corresponds to predetermined pulse positions in the residual signal vectors of the white noise code book.

4. A method according to claim 2, wherein the impulse residual signal vectors provided in step (d) correspond to pulse positions of a maximum amplitude in the white noise residual signal vectors of the code book.

5. A method according to claim 4, wherein the impulse residual signal vectors provided in said step (d) and the pulse positions of the maximum amplitude are stored in a separately provided code book.

6. A method according to claim 2, wherein the impulse residual signal vectors provided in said step (d) and pulse positions of a maximum amplitude are stored in a separately provided code book.

7. A method according to claim 1, wherein the impulse residual signal vectors provided in said step (d) having a predetermined relationship with the code vector of the code book are main element impulses in the white noise residual signal vectors of the code book.

8. A method according to claim 1, further comprising the step of:

(j) adjusting the white noise residual signal vector and the impulse residual signal vector by a predetermined coefficient derived from a vector of the input speech signal and the pitch prediction vector obtained by said applying linear prediction to a residual signal of a preceding frame.

9. A method according to claim 8, further comprising the step of:

(k) weighting the white noise residual signal vector and the impulse residual signal vector by a predetermined coefficient derived from the vector of the input speech signal and the pitch prediction vector obtained by said applying linear prediction to a residual signal of a preceding frame.

10. A method according to claim 9, further comprising the steps of:

(l) adding the white noise residual signal vector and the impulse residual signal vector in a ratio according to an intensity of a pitch correlation obtained by applying linear prediction to the vector of the input speech signal and the pitch prediction vector obtained by said applying linear prediction to a residual signal of a preceding frame.

11. A method according to claim 10, wherein the pitch correlation in said step (l) is a function of angle.

12. A method according to claim 1, wherein the impulse residual signal vector is separated from the white noise residual signal vector.

13. An apparatus for encoding and transmitting an input speech signal, comprising:

a pitch frequency delay circuit to delay a residual signal of a previous speech signal by a pitch frequency,

a code book to store a plurality of white noise residual signal vectors,

an impulse generating circuit to generate an impulse having a predetermined relationship with the white noise residual signal vectors stored in said code book,

a linear prediction circuit operatively connected to said pitch frequency delay circuit, said code book, and said impulse generating circuit to output vectors and a coefficient,

a variable gain circuit operatively connected to said linear prediction circuit to apply a variable gain to at least one of the output vectors of said linear prediction circuit,

a first addition circuit operatively connected to said variable gain circuit to produce a reproduced composite vector,

a second addition circuit operatively connected to said first addition circuit to add the reproduced composite vector and a vector of the input speech signal to output an error signal,

an evaluating circuit operatively connected to said second addition circuit and said code book to identify a white noise residual signal vector stored in said code book in response to the error signal, and

an output transmitter operatively connected to at least said linear prediction circuit to transmit a decodable signal based on at least the coefficient.

14. An apparatus according to claim 13,

wherein said linear prediction circuit comprises a first linear prediction unit operatively connected to said pitch frequency delay circuit to provide a pitch prediction vector, a second linear prediction unit operatively connected to said code book to provide a white noise prediction vector and a third linear prediction unit operatively connected to said impulse generating circuit to provide an impulse prediction vector;

wherein said first addition circuit includes;

a first adder operatively connected to said first and second linear prediction units to add the pitch and white noise prediction vectors to produce a sum vector, and

a second adder operatively connected to said third linear prediction unit and said first adder to add the impulse prediction vector and the sum vector to produce the reproduced composite vector.

15. An apparatus according to claim 13,

wherein said linear prediction circuit comprises a first linear prediction unit operatively connected to said pitch frequency delay circuit to provide a pitch prediction vector, a second linear prediction unit operatively connected to said code book to provide a white noise prediction vector and a third linear prediction unit operatively connected to said impulse generating circuit to provide an impulse prediction vector;

and

wherein said apparatus further comprises a main element pulse position detection circuit operatively connected to said impulse generating circuit and said second linear prediction unit to drive said impulse generating circuit in response to the white noise prediction vector output from said second linear prediction unit.

16. An apparatus according to claim 15, wherein said main element pulse position detection circuit determines a pulse position allowing a smallest phase error between the white noise prediction vector and the impulse prediction vector, the impulse prediction vector obtained by applying linear prediction in said third linear prediction unit to one pulse from said impulse generating circuit which is corresponding to sample times of residual signal vector stored in said code book.

17. An apparatus according to claim 13, wherein said impulse generating circuit comprises another code book to store a plurality of impulses corresponding to the white noise residual signal vectors stored in said code book.

18. An apparatus according to claim 17, wherein said another code book stores the impulses in an order representative of maximum pulses in the white noise residual signal vectors stored in said code book.

19. An apparatus according to claim 17, wherein said impulse generating circuit includes an impulse separating circuit which separates the impulses from the vectors of white noise residual signal vectors stored in said code book.

20. An apparatus according to claim 13,

wherein said linear prediction circuit comprises a first linear prediction unit operatively connected to said pitch frequency delay circuit to provide a pitch prediction vector, a second linear prediction unit operatively connected to said code book to provide a white noise prediction vector and a third linear prediction unit operatively connected to said impulse generating circuit to provide an impulse prediction vector;

wherein said variable gain circuit comprises a first variable gain unit operatively connected to said second linear prediction unit to apply a first variable gain to the white noise prediction vector and a second variable gain unit operatively connected to said third linear prediction unit to apply a second variable gain to the impulse prediction vector; and

wherein said apparatus further comprises

a weighting circuit operatively connected to said first and second variable gain units to control said first and second variable gain units, and

a pitch correlation calculating circuit operatively connected to said weighting circuit and at least said first linear prediction unit to receive the pitch prediction vector from said first linear prediction unit and to control said first and second variable gain units.

21. An apparatus for encoding and transmitting an input speech signal to provide a decodable signal, comprising:

first code book means for storing first data and generating a white noise signal based on the stored first data and an index;

second code book means for storing second data and generating an impulse signal based on the stored second data and the index;

linear prediction means for applying linear prediction to the white noise and impulse signals and generating a coefficient;

processing means for comparing the white noise and impulse signals with the input speech signal to provide an error signal;

evaluating means for generating the index based on the error signal; and

transmitting means for transmitting a decodable signal based on at least the coefficient.

22. An apparatus according to claim 21, wherein said processing means comprises:

adding means for adding the white noise and impulse signals after said linear prediction means applies linear prediction to the white noise and impulse signals; and

comparing means for comparing the white noise and impulse signals after said adding means adds the white noise and impulse signals.

23. An apparatus according to claim 22,

wherein said apparatus further comprises a pitch frequency delay unit operatively connected to provide a residual signal of a previous speech signal to said linear prediction means;

wherein said linear prediction means comprises means for outputting a pitch prediction vector based on the residual signal of a previous speech signal; and

wherein said adding means comprises means for further adding the pitch prediction vector, the white noise and the impulse signals.

24. An apparatus according to claim 23,

wherein one of the first and second code book means is a table and another of the first and second code book means is a code book;

and

wherein said apparatus further comprises an impulse separating circuit for receiving outputs of the table and the code book and generating the white noise and impulse signals.

25. An apparatus according to claim 24, further comprising:

hysteresis means for storing a previous speech signal; and

subtractor means for subtracting the previous speech signal from a present speech signal to provide the input speech signal to said processing means.

26. An apparatus according to claim 23, further comprising:

hysteresis means for storing a previous speech signal; and

subtractor means for subtracting the previous speech signal from a present speech signal to provide the input speech signal to said processing means.

27. An apparatus according to claim 26,

wherein said apparatus further comprises a pitch correlation calculation unit operatively connected to said linear prediction unit and said subtractor to output weights;

and

wherein said linear prediction means includes multipliers operatively connected to said pitch correlation calculation unit to weight the white noise and impulse signals by the weights.

28. An apparatus according to claim 21, wherein one of the first and second code book means is a table and another is a code book;

and

wherein said apparatus further comprises an impulse separating circuit operatively connected to receive outputs of the table and the code book to generate the white noise and impulse signals.

29. An apparatus for encoding an input speech signal, comprising:

code book means for storing white noise data and generating a white noise signal based on the stored white noise data and an index;

impulse means for generating an impulse signal having a predetermined relationship with the white noise data stored in said code book means based on the index;

linear prediction means for applying linear prediction to the white noise and impulse signals and generating a coefficient;

processing means for comparing the white noise and impulse signals with the input speech signal to provide an error signal;

evaluating means for generating the index based on the error signal; and

transmitting means for transmitting a decodable signal based on at least the coefficient.

30. An apparatus according to claim 29,

wherein said apparatus further comprises pitch prediction means for applying pitch prediction to the white noise and impulse signals and generating another coefficient;

and

wherein said transmitting means comprises means for transmitting the decodable signal based on at least the coefficient, the another coefficient and the index.

31. An apparatus according to claim 30, wherein said processing means comprises:

adding means for adding the white noise and impulse signals before said pitch prediction means applies pitch prediction and said linear prediction means applies linear prediction; and

comparing means for comparing the white noise and impulse signals after said pitch prediction means applies pitch prediction and said linear prediction means applies linear prediction.

32. A method of encoding and transmitting an input speech signal to provide a decodable signal, comprising the steps of:

(a) generating a first signal based on stored first data and an index;

(b) generating a second signal based on stored second data and the index;

(c) applying linear prediction to the first and second signals and generating third and fourth signals and a coefficient;

(d) adding the third and fourth signals to generate a fifth signal;

(e) comparing the fifth signal with the input speech signal to generate an error signal;

(f) generating the index based on the error signal; and

(g) transmitting a decodable signal based on at least the coefficient.

33. A method according to claim 32, wherein the first signal is a white noise signal and the second signal is an impulse signal.

34. A method according to claim 33, further comprising the steps of:

(h) storing a previous speech signal; and

(i) subtracting the previous speech signal stored in said step (h) from a present speech signal to provide the input speech signal for said comparing in said step (e).

35. An apparatus for receiving and decoding a decodable signal to reproduce a speech signal, comprising:

receiving means for receiving and demultiplexing the decodable signal to generate at least an index signal and a coefficient;

first code book means for storing first data and generating a white noise signal based on the stored first data and the index signal from the receiving means;

second code book means for storing second data and generating an impulse signal based on the stored second data and the index signal from the receiving means;

linear prediction means for applying linear prediction to the white noise and impulse signals based on the coefficient from said receiving means to reproduce the speech signal.

36. An apparatus for receiving and decoding a decodable signal to reproduce a speech signal, comprising:

receiving means for receiving and demultiplexing the decodable signal to generate at least an index signal, a coefficient and a phase signal;

code book means for storing a plurality of white noise residual signal vectors and outputting a white noise residual signal vector based on the index signal from said receiving means;

impulse generating means for generating an impulse signal having a predetermined relationship with the white noise residual signal vectors stored in said code book based on the phase signal from said receiving means; and

linear prediction means for applying linear prediction to the white noise residual signal vectors and the impulse signal based on the coefficient from the receiving means to reproduce the speech signal.