Encoding method for acquiring codes corresponding to prediction residuals, decoding method for decoding codes corresponding to noise or pulse sequence, encoder, decoder, program, and recording medium

• US 9,230,554 B2
• Filed: 02/08/2012
• Issued: 01/05/2016
• Est. Priority Date: 02/16/2011
• Status: Active Grant

• Current Assignee: Nippon Telegraph and Telephone Corporation
• Original Assignee: Nippon Telegraph and Telephone Corporation
• Inventors: Moriya, Takehiro, Harada, Noboru, Kamamoto, Yutaka, Hiwasaki, Yusuke, Fukui, Masahiro
• Primary Examiner(s): McFadden, Susan
• Application Number:
  US13/984,773
• Publication Number:
  US 20130317814A1
• Time in Patent Office: 1,427 Days
• Field of Search: 704/219
• US Class Current: 1/1
• 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 19/18 Vocoders using multiple modes
• CPC Class Codes: H03M 5/06 Code representation, e.g. transition, for a given bit cell depending only on the information in that bit cell

Abstract

In encoding, the number of bits to be assigned to codes corresponding to noise or a pulse sequence obtained according to prediction analysis applied to time series signals included in a predetermined time interval is switched according to whether an index that indicates a level of periodicity and/or stationarity of input time series signals satisfies a condition that indicates high periodicity and/or high stationarity or a condition that indicates low periodicity and/or low stationarity, to acquire the codes corresponding to the noise and the pulse sequence. In decoding, a decoding mode for codes corresponding to noise or a pulse sequence included in codes corresponding to a predetermined time interval is switched according to the same criterion as that described above to decode the codes corresponding to the noise or the pulse sequence to acquire noise or a pulse sequence corresponding to the predetermined time interval.

Claims

1. A computer-implemented encoding method executing on a processor, the method comprising acquiring codes corresponding to prediction residuals obtained according to prediction analysis applied to time series signals included in a predetermined time interval of input time series signals which are speech or acoustic signals, with the number of bits to be assigned to the codes corresponding to the prediction residuals being switched according to whether an index that indicates a level of periodicity and/or stationarity corresponding to the time series signals in the predetermined time interval or time series signals in an interval before the predetermined time interval of the input time series signals satisfies a condition that indicates high periodicity and/or high stationarity or a condition that indicates low periodicity and/or low stationarity,

wherein the number of bits of the codes corresponding to the prediction residuals, obtained when the index that indicates the level of periodicity and/or stationarity satisfies the condition that indicates high periodicity and/or high stationarity, is smaller than the number of bits of the codes corresponding to the prediction residuals, obtained when the index that indicates the level of periodicity and/or stationarity does not satisfy the condition that indicates high periodicity and/or high stationarity;

or

the number of bits of the codes corresponding to the prediction residuals, obtained when the index that indicates the level of periodicity and/or stationarity satisfies the condition that indicates low periodicity and/or low stationarity, is larger than the number of bits of the codes corresponding to the prediction residuals, obtained when the index that indicates the level of periodicity and/or stationarity does not satisfy the condition that indicates low periodicity and/or low stationarity,

wherein a predetermined total number of bits is assigned to codes corresponding to a time interval that includes the predetermined time interval and a second predetermined time interval after the predetermined time interval; and

the number of bits to be assigned to codes corresponding to second prediction residuals obtained according to prediction analysis applied to time series signals included in the second predetermined time interval is determined according to the number of unassigned bits determined from the total number of bits and the number of bits to be assigned to the codes corresponding to the prediction residuals obtained according to the prediction analysis applied to the time series signals included in the predetermined time interval, to acquire, with the determined number of bits, the codes corresponding to the second prediction residuals.

2. The encoding method according to claim 1,

wherein the index includes a quantized pitch gain or a value corresponding thereto;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto is larger than a predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto is smaller than the predetermined specified value.

3. The encoding method according to claim 1,

wherein the index includes a vector-quantized gain code corresponding to a combination of a quantized pitch gain or a value corresponding thereto, and a quantized fixed-codebook gain or a value corresponding thereto;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the vector-quantized gain code corresponds to a combination of a quantized pitch gain or a value corresponding thereto that is larger than a predetermined specified value, and a quantized fixed-codebook gain or a value corresponding thereto;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the vector-quantized gain code corresponds to a combination of a quantized pitch gain or a value corresponding thereto that is smaller than the predetermined specified value, and a quantized fixed-codebook gain or a value corresponding thereto.

4. The encoding method according to claim 1,

wherein the index includes a quantized pitch gain or a value corresponding thereto, and a quantized fixed-codebook gain or a value corresponding thereto;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that a ratio of the quantized pitch gain or the value corresponding thereto to the quantized fixed-codebook gain or the value corresponding thereto is larger than a predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that a ratio of the quantized pitch gain or the value corresponding thereto to the quantized fixed-codebook gain or the value corresponding thereto is smaller than a predetermined specified value.

5. The encoding method according to claim 1,

wherein the index includes a vector-quantized gain code corresponding to a combination of a quantized pitch gain or a value corresponding thereto, and a quantized fixed-codebook gain or a value corresponding thereto;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the vector-quantized gain code corresponds to a combination of a quantized pitch gain or a value corresponding thereto, and a quantized fixed-codebook gain or a value corresponding thereto, a ratio of the quantized pitch gain or the value corresponding thereto to the quantized fixed-codebook gain or the value corresponding thereto being larger than a predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the vector-quantized gain code corresponds to a combination of a quantized pitch gain or a value corresponding thereto, and a quantized fixed-codebook gain or a value corresponding thereto, a ratio of the quantized pitch gain or the value corresponding thereto to the quantized fixed-codebook gain or the value corresponding thereto being smaller than the predetermined specified value.

6. The encoding method according to claim 1,

wherein the index includes a quantized pitch gain or a value corresponding thereto, and a value corresponding to a quantized fixed-codebook gain;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto is larger than a first predetermined specified value and that the value corresponding to the quantized fixed-codebook gain is smaller than a second predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto is smaller than a first predetermined specified value and that the value corresponding to the quantized fixed-codebook gain is larger than a second predetermined specified value.

7. The encoding method according to claim 1,

wherein the index includes a vector-quantized gain code corresponding to a combination of a quantized pitch gain or a value corresponding thereto, and a value corresponding to a quantized fixed-codebook gain;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto corresponding to the vector-quantized gain code is larger than a first predetermined specified value and that the value corresponding to the quantized fixed-codebook gain corresponding to the vector-quantized gain code is smaller than a second predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto corresponding to the vector-quantized gain code is smaller than a first predetermined specified value and that the value corresponding to the quantized fixed-codebook gain corresponding to the vector-quantized gain code is larger than a second predetermined specified value.

8. The encoding method according to claim 1,

wherein the index includes a quantized PARCOR coefficient or a value corresponding thereto;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the quantized PARCOR coefficient or the value corresponding thereto is larger than a predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the quantized PARCOR coefficient or the value corresponding thereto is smaller than a predetermined specified value.

9. The encoding method according to claim 1,

wherein the index includes an estimated value of a prediction gain determined by a quantized PARCOR coefficient or a value corresponding thereto;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the estimated value of the prediction gain is larger than a predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the estimated value of the prediction gain is smaller than a predetermined specified value.

10. The encoding method according to claim 1,

wherein the index includes the magnitude of the difference between a value corresponding to a pitch period of the time series signals in a first time interval and a value corresponding to a pitch period of the time series signals in a second time interval having a predetermined positional relationship with the first time interval, or a value corresponding to the magnitude of the difference;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the magnitude of the difference or the value corresponding thereto is smaller than a predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the magnitude of the difference or the value corresponding thereto is larger than a predetermined specified value.

23. A non-transitory computer-readable recording medium having stored thereon a program for causing a computer to execute the process of the encoding method according to claim 1.

11. A computer-implemented decoding method executing on a processor, the method comprising decoding codes corresponding to noise or a pulse sequence included in inputted codes corresponding to a speech or acoustic signal sampled in a predetermined time interval to acquire noise or a pulse sequence corresponding to the predetermined time interval, with a decoding mode for the codes corresponding to the noise or the pulse sequence being switched according to whether an index that indicates a level of periodicity and/or stationarity corresponding to the predetermined time interval or an interval before the predetermined interval, the index being included in input codes or obtained from the codes, satisfies a condition that indicates high periodicity and/or high stationarity or a condition that indicates low periodicity and/or low stationarity,

wherein codes having a first number of bits, corresponding to the noise or the pulse sequence are decoded when the index that indicates the level of periodicity and/or stationarity satisfies the condition that indicates high periodicity and/or high stationarity, and codes having a second number of bits, corresponding to the noise or the pulse sequence are decoded when the index that indicates the level of periodicity and/or stationarity does not satisfy the condition that indicates high periodicity and/or high stationarity;

or

codes having the second number of bits, corresponding to the noise or the pulse sequence are decoded when the index that indicates the level of periodicity and/or stationarity satisfies the condition that indicates low periodicity and/or low stationarity, and codes having the first number of bits, corresponding to the noise or the pulse sequence are decoded when the index that indicates the level of periodicity and/or stationarity does not satisfy the condition that indicates low periodicity and/or low stationarity; and

the first number of bits is smaller than the second number of bits,

wherein a predetermined total number of bits is assigned to codes corresponding to a time interval that includes the predetermined time interval and a second predetermined time interval after the predetermined time interval; and

a decoding mode for codes corresponding to second noise or a second pulse sequence included in codes corresponding to the second predetermined time interval is switched according to the number of bits assigned to the second predetermined time interval, determined from the total number of bits and the number of bits of the codes corresponding to the noise or the pulse sequence included in the codes corresponding to the predetermined time interval, to decode the codes corresponding to the second noise or the second pulse sequence to acquire second noise or a second pulse sequence corresponding to the second predetermined time interval.

12. The decoding method according to claim 11,

wherein the index includes a quantized pitch gain or a value corresponding thereto, the quantized pitch gain being obtained from the input codes;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto is larger than a predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto is smaller than a predetermined specified value.

13. The decoding method according to claim 11,

wherein the index includes a vector-quantized gain code corresponding to a combination of a quantized pitch gain or a value corresponding thereto, and a quantized fixed-codebook gain or a value corresponding thereto, the vector-quantized gain code being included in the input codes;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the vector-quantized gain code corresponds to a combination of a quantized pitch gain or a value corresponding thereto that is larger than a predetermined specified value, and a quantized fixed-codebook gain or a value corresponding thereto;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the vector-quantized gain code corresponds to a combination of a quantized pitch gain or a value corresponding thereto that is smaller than a predetermined specified value, and a quantized fixed-codebook gain or a value corresponding thereto.

14. The decoding method according to claim 11,

wherein the index includes a quantized pitch gain or a value corresponding thereto, and a quantized fixed-codebook gain or a value corresponding thereto, the quantized pitch gain being obtained from the input codes and the quantized fixed-codebook gain being obtained from the input codes;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that a ratio of the quantized pitch gain or the value corresponding thereto to the quantized fixed-codebook gain or the value corresponding thereto is larger than a predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that a ratio of the quantized pitch gain or the value corresponding thereto to the quantized fixed-codebook gain or the value corresponding thereto is smaller than a predetermined specified value.

15. The decoding method according to claim 11,

wherein the index includes a vector-quantized gain code corresponding to a combination of a quantized pitch gain or a value corresponding thereto, and a quantized fixed-codebook gain or a value corresponding thereto, the vector-quantized gain code being included in the input codes;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the vector-quantized gain code corresponds to a combination of a quantized pitch gain or a value corresponding thereto, and a quantized fixed-codebook gain or a value corresponding thereto, a ratio of the quantized pitch gain or the value corresponding thereto to the quantized fixed-codebook gain or the value corresponding thereto being larger than a predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the vector-quantized gain code corresponds to a combination of a quantized pitch gain or a value corresponding thereto, and a quantized fixed-codebook gain or a value corresponding thereto, a ratio of the quantized pitch gain or the value corresponding thereto to the quantized fixed-codebook gain or the value corresponding thereto being smaller than a predetermined specified value.

16. The decoding method according to claim 11,

wherein the index includes a quantized pitch gain or a value corresponding thereto, and a value corresponding to a quantized fixed-codebook gain, the quantized pitch gain being obtained from the input codes and the quantized fixed-codebook gain being obtained from the input codes;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto is larger than a first predetermined specified value and that the value corresponding to the quantized fixed-codebook gain is smaller than a second predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto is smaller than a first predetermined specified value and that the value corresponding to the quantized fixed-codebook gain is larger than a second predetermined specified value.

17. The decoding method according to claim 11,

wherein the index includes a vector-quantized gain code corresponding to a combination of a quantized pitch gain or a value corresponding thereto, and a value corresponding to a quantized fixed-codebook gain, the vector-quantized gain code being included in the input codes;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto corresponding to the vector-quantized gain code is larger than a first predetermined specified value and that the value corresponding to the quantized fixed-codebook gain corresponding to the vector-quantized gain code is smaller than a second predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the quantized pitch gain or the value corresponding thereto corresponding to the vector-quantized gain code is smaller than a first predetermined specified value and that the value corresponding to the quantized fixed-codebook gain corresponding to the vector-quantized gain code is larger than a second predetermined specified value.

18. The decoding method according to claim 11,

wherein the index includes a quantized PARCOR coefficient or a value corresponding thereto, the quantized PARCOR coefficient being obtained from the input codes;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the quantized PARCOR coefficient or the value corresponding thereto is larger than a predetermined specified value; and

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the quantized PARCOR coefficient or the value corresponding thereto is smaller than the predetermined specified value.

19. The decoding method according to claim 11,

wherein the index includes an estimated value of a prediction gain determined by a quantized PARCOR coefficient or a value corresponding thereto, the quantized PARCOR coefficient being obtained from the input codes;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the estimated value of the prediction gain is larger than a predetermined specified value; and

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the estimated value of the prediction gain is smaller than the predetermined specified value.

20. The decoding method according to claim 11,

wherein the index includes the magnitude of the difference between a value corresponding to a pitch period in a first time interval, the pitch period being obtained from the input codes, and a value corresponding to a pitch period in a second time interval having a predetermined positional relationship with the first time interval, the pitch period being obtained from the input codes, or a value corresponding to the magnitude of the difference;

the condition that indicates high periodicity and/or high stationarity includes a condition that indicates that the magnitude of the difference or the value corresponding thereto is smaller than a first predetermined specified value;

or

the condition that indicates low periodicity and/or low stationarity includes a condition that indicates that the magnitude of the difference or the value corresponding thereto is larger than a second predetermined specified value.

24. A non-transitory computer-readable recording medium having stored thereon a program for causing a computer to execute the process of the decoding method according to claim 11.

21. A computer-implemented encoder for acquiring codes corresponding to prediction residuals obtained according to prediction analysis applied to time series signals included in a predetermined time interval of input time series signals which are speech or acoustic signals, with the number of bits to be assigned to the codes corresponding to the prediction residuals being switched according to whether an index that indicates a level of periodicity and/or stationarity corresponding to the time series signals in the predetermined time interval or time series signals in an interval before the predetermined time interval of the input time series signals satisfies a condition that indicates high periodicity and/or high stationarity or a condition that indicates low periodicity and/or low stationarity,

wherein the number of bits of the codes corresponding to the prediction residuals, obtained when the index that indicates the level of periodicity and/or stationarity satisfies the condition that indicates high periodicity and/or high stationarity, is smaller than the number of bits of the codes corresponding to the prediction residuals, obtained when the index that indicates the level of periodicity and/or stationarity does not satisfy the condition that indicates high periodicity and/or high stationarity;

or

the number of bits of the codes corresponding to the prediction residuals, obtained when the index that indicates the level of periodicity and/or stationarity satisfies the condition that indicates low periodicity and/or low stationarity, is larger than the number of bits of the codes corresponding to the prediction residuals, obtained when the index that indicates the level of periodicity and/or stationarity does not satisfy the condition that indicates low periodicity and/or low stationarity,

wherein a predetermined total number of bits is assigned to codes corresponding to a time interval that includes the predetermined time interval and a second predetermined time interval after the predetermined time interval; and

the number of bits to be assigned to codes corresponding to second prediction residuals obtained according to prediction analysis applied to time series signals included in the second predetermined time interval is determined according to the number of unassigned bits determined from the total number of bits and the number of bits to be assigned to the codes corresponding to the prediction residuals obtained according to the prediction analysis applied to the time series signals included in the predetermined time interval, to acquire the codes, with the determined number of bits, corresponding to the second prediction residuals.

22. A computer-implemented decoder for decoding codes corresponding to noise or a pulse sequence included in inputted codes corresponding to a speech or acoustic signal sampled in a predetermined time interval to acquire noise or a pulse sequence corresponding to the predetermined time interval, with a decoding mode for the codes corresponding to the noise or the pulse sequence being switched according to whether an index that indicates a level of periodicity and/or stationarity corresponding to the predetermined time interval or an interval before the predetermined interval, the index being included in input codes or obtained from the codes, satisfies a condition that indicates high periodicity and/or high stationarity or a condition that indicates low periodicity and/or low stationarity,

wherein codes having a first number of bits, corresponding to the noise or the pulse sequence are decoded when the index that indicates the level of periodicity and/or stationarity satisfies the condition that indicates high periodicity and/or high stationarity, and codes having a second number of bits, corresponding to the noise or the pulse sequence are decoded when the index that indicates the level of periodicity and/or stationarity does not satisfy the condition that indicates high periodicity and/or high stationarity;

or

codes having the second number of bits, corresponding to the noise or the pulse sequence are decoded when the index that indicates the level of periodicity and/or stationarity satisfies the condition that indicates low periodicity and/or low stationarity, and codes having the first number of bits, corresponding to the noise or the pulse sequence are decoded when the index that indicates the level of periodicity and/or stationarity does not satisfy the condition that indicates low periodicity and/or low stationarity; and

the first number of bits is smaller than the second number of bits,

wherein a predetermined total number of bits is assigned to codes corresponding to a time interval that includes the predetermined time interval and a second predetermined time interval after the predetermined time interval; and

a decoding mode for codes corresponding to second noise or a second pulse sequence included in codes corresponding to the second predetermined time interval is switched according to the number of bits assigned to the second predetermined time interval, determined from the total number of bits and the number of bits of the codes corresponding to the noise or the pulse sequence included in the codes corresponding to the predetermined time interval, to decode the codes corresponding to the second noise or the second pulse sequence to acquire second noise or a second pulse sequence corresponding to the second predetermined time interval.