Method and apparatus for transcoding between different speech encoding/decoding systems

• US 7,231,345 B2
• Filed: 01/24/2005
• Issued: 06/12/2007
• Est. Priority Date: 07/24/2002
• Status: Expired due to Fees

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Murashima, Atsushi
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US11/039,969
• Publication Number:
  US 20050131682A1
• Time in Patent Office: 869 Days
• Field of Search: None
• US Class Current: 704/219
• CPC Class Codes: G10L 19/083 the excitation function being an excitation gain G10L25/90 takes precedence
• 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 19/173 Transcoding, i.e. converting between two coded representations avoiding cascaded coding-decoding

Abstract

A code converting apparatus for converting a first code sequence conforming to a first system to a second code sequence conforming to a second system, in which a speech decoding circuit acquires a first linear prediction coefficient and the information on an excitation signal from the first code sequence, and actuates a filter having the aforementioned first linear prediction coefficient with the excitation signal obtained from the information on the excitation signal, to generate a first speech signal. A gain code generating circuit calculates a gain minimizing the distance between a second speech signal, generated from the information, obtained from the second code sequence, and the first speech signal (optimum gain), and corrects the optimum gain and the gain code generating circuit then finds the gain information in the second code sequence, based on the optimum gain as corrected (optimum gain corrected), the above optimum gain and a gain read out from a gain codebook of the second system. The gain is found at this time, in a non-speech segment, based on a speech decision value, using an evaluation function which will reduce time variations of the gain of the second system.

Claims

1. A code converting method for converting a first code sequence conforming to a first system to a second code sequence conforming to a second system, comprising the steps of:

acquiring a first linear prediction coefficient and first information on an excitation signal from said first code sequence and actuating a filter having said first linear prediction coefficient with the excitation signal obtained from said first information to generate a first speech signal;

deriving an optimum gain based on a second speech signal generated by second information obtained from said second code sequence, and based on said first speech signal;

correcting said optimum gain to generate corrected optimum gain; and

finding gain information in said second code sequence based on said corrected optimum gain, said optimum gain and on a gain read out from a gain codebook for said second system.

2. The code converting method according to claim 1, further comprising the steps of:

determining a speech decision value, discriminating a speech segment/non-speech segment, based on said first linear prediction coefficient; and

finding the gain information in said second code sequence, using an evaluation function which will decrease a temporal variation of the gain in said second code sequence, when said speech decision value indicates the non-speech segment.

3. The code converting method according to claim 1, further comprising the steps of:

calculating a first square error from the corrected optimum gain and from the gain read out from said gain codebook;

calculating a second square error from said optimum gain and from the gain read out from said gain codebook; and

selecting a gain minimizing an evaluation function which is based on said first square error and said second square error from said gain codebook to find the gain information in said second code sequence.

4. The code converting method according to claim 3, wherein said evaluation function is composed of said first square error, said second square error and weighting coefficients.

5. The code converting method according to claim 3, further comprising the steps of:

determining a speech decision value, discriminating a speech segment/non-speech segment, based on said first linear prediction coefficient;

said evaluation function being found by taking a weighted average value of said first and second square errors by weighting coefficients; and

setting said weighting coefficients to respective preset values, based on said speech decision value, depending on the speech segment and the non-speech segment, to calculate said evaluation function.

6. The code converting method according to claim 1, wherein said corrected optimum gain is based on a long-term average value of said optimum gain.

7. The code converting method according to claim 1, wherein the gain minimizing the distance between the second speech signal generated by the information obtained from said second code sequence and said first speech signal is found as said optimum gain.

8. A code converting apparatus for converting a first code sequence conforming to a first system to a second code sequence conforming to a second system, said apparatus comprising:

a speech decoding circuit for acquiring a first linear prediction coefficient and first information on an excitation signal from said first code sequence and actuating a filter having said first linear prediction coefficient with the excitation signal obtained from said first information to generate a first speech signal;

an optimum gain calculating circuit for calculating an optimum gain based on a second speech signal generated by second information obtained from said second code sequence, and based on said first speech signal;

an optimum gain correcting circuit for correcting said optimum gain to generate corrected optimum gain; and

a gain encoding circuit for finding the gain information in said second code sequence based on said corrected optimum gain, said optimum gain and on a gain read out from a gain codebook for said second system.

9. The code converting apparatus according to claim 8, wherein said gain encoding circuit includes a unit for finding the gain information in said second code sequence by calculating a first square error from the corrected optimum gain and from the gain read out from said gain codebook;

calculating a second square error from said optimum gain and from the gain read out from said gain codebook; and

selecting a gain minimizing an evaluation function which is based on said first square error and said second square error from said gain codebook to find the gain information in said second code sequence.

10. The code converting apparatus according to claim 9, wherein said evaluation function is composed of said first square error, said second square error and weighting coefficients.

11. The code converting apparatus according to claim 9, further comprising;

a speech/non-speech discriminating circuit for outputting a speech decision value, discriminating a speech segment/non-speech segment, based on said first linear prediction coefficient;

said gain encoding circuit finding said evaluation function by taking a weighted average value of said first and second square errors by weighting coefficients; and

setting said weighting coefficients to respective preset values, based on said speech decision value, depending on the speech segment and the non-speech segment, to calculate said evaluation function.

12. The code converting apparatus according to claim 8, wherein said corrected optimum gain is based on the long-term average value of said optimum gain.

13. The code converting apparatus according to claim 8, wherein the gain minimizing the distance between the second speech signal generated by the information obtained from said second code sequence and said first speech signal is found as said optimum gain.

14. The code converting apparatus according to claim 8, further comprising:

a speech/non-speech discriminating circuit for discriminating a speech decision value, discriminating a speech segment/non-speech segment, based on said first linear prediction coefficient;

said gain encoding circuit finding the gain information in said second code sequence, using an evaluation function which will decrease temporal variations of the gain in said second code sequence, when said speech decision value indicates the non-speech segment.

15. A computer program product in a medium used by a computer that composes a code converting apparatus for converting a first code sequence, conforming to a first system, into a second code sequence conforming to a second system, comprising a program for causing said computer to execute:

(a) the processing of acquiring first linear prediction coefficient and first information on an excitation signal from said first code sequence and actuating a filter having said first linear prediction coefficient with the excitation signal obtained from said first information to generate a first speech signal;

(b) the processing of calculating an optimum gain based on a second speech signal generated by second information obtained from a second code sequence, and based on said first speech signal;

(c) the processing of correcting said optimum gain to generate corrected optimum gain; and

(d) the processing of finding the gain information in said second code sequence based on said corrected optimum gain, said optimum gain and on a gain read out from a gain codebook for said second system.

16. The computer program product according to claim 15, further comprising a program to cause said computer to execute the processing of finding the gain information in said second code sequence by calculating a first square error from the corrected optimum gain and from the gain read out from said gain codebook:

calculating a second square error from said optimum gain and from the gain read out from said gain codebook; and

selecting a gain minimizing an evaluation function which is based on said first square error and said second square error from said gain codebook to find the gain information in said second code sequence.

17. The computer program product according to claim 16, wherein said evaluation function is composed of said first square error, said second square error and weighting coefficients.

18. The computer program product according to claim 16, further comprising a program to cause the computer to execute:

the processing of outputting a speech decision value, discriminating the speech segment/non-speech segment, based on said first linear prediction coefficient; and

the processing of finding said evaluation function by taking a weighted average value of said first and second square errors by weighting coefficients, and setting said weighting coefficients to respective preset values, based on said speech decision value, depending on the speech segment and the non-speech segment, to calculate said evaluation function.

19. The computer program product according to claim 15, wherein said corrected optimum gain is based on a long-term average value of said optimum gain.

20. The computer program product according to claim 15, wherein the gain minimizing the distance between the second speech signal generated by the information obtained from said second code sequence and said first speech signal is found as said optimum gain.

21. The computer program product according to claim 15, further comprising a program to cause the computer to execute:

the processing of outputting a speech decision value, discriminating speech segment/non-speech segment, based on said first linear prediction coefficient; and

the processing of finding the gain information in said second code sequence, using an evaluation function which will decrease temporal variations of the gain in said second code sequence, when said speech decision value indicates the non-speech segment.

22. A recording medium that may be read out by a computer, said recording medium having recorded thereon said program as defined in claim 15.