Method and apparatus for adjusting a spectrum shape of a speech signal

• US 5,864,798 A
• Filed: 09/17/1996
• Issued: 01/26/1999
• Est. Priority Date: 09/18/1995
• Status: Expired due to Term

• Current Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Original Assignee: Kabushiki Kaisha Toshiba (Toshiba Corporation)
• Inventors: Amada, Tadashi, Oshikiri, Masahiro, Miseki, Kimio, Akamine, Masami, Yamashita, Akinobu
• Primary Examiner(s): Knepper, David D.
• Assistant Examiner(s): Wieland, Susan
• Application Number:
  US08/714,260
• Time in Patent Office: 861 Days
• Field of Search: 704/225, 704/224, 704/217, 704/219
• US Class Current: 704/225
• CPC Class Codes: G10L 19/26 Pre-filtering or post-filtering
• CPC Class Codes: G10L 25/12 the extracted parameters being prediction coefficients

Abstract

Adjusting the shape of a spectrum of a speech signal includes the steps of using a first filter with pole-zero transfer function A(z)/B(z) for subjecting a speech signal to a spectrum envelop emphasis and a second filter cascade-connected with the first filter, for compensating for a spectral tilt due to the first filter, independently deriving two filter coefficients used in the second filter for compensating for the spectral tilt from the pole-zero transfer function, and compensating for the spectral tilt corresponding to the pole-zero transfer function according to the derived filter coefficients.

Claims

1. A method for adjusting a spectrum shape of an input speech signal, comprising the steps of:

cascade-connecting a first filter having a first pole-zero transfer function for subjecting said input speech signal to a spectrum envelop emphasis and a second filter having a second pole-zero transfer function for compensating a spectral tilt of the spectrum shape of the input speech signal caused by the first filter;

independently deriving two filter coefficients used in the second filter from the first pole-zero transfer function of said first filter; and

compensating the spectral tilt using the derived filter coefficients,

wherein the second pole-zero transfer function in a z transform domain comprises at least a first-order pole-zero transfer function expressed by (1-μ

_z Z^-1)/(1-μ

_p Z^-1), where μ

_z and μ

_p are filter coefficients whose absolute values are smaller than 1 and which are independent from each other, and said step of deriving the filter coefficients derives said μ

_z from a zero transfer function of the first filter and derives said μ

_z from a pole transfer function of the first filter.

2. The method according to claim 1, wherein said step of deriving the filter coefficients includes a step of extracting pole and zero filter coefficients corresponding to the two filter coefficients from the first filter and inputting the pole and zero filter coefficients to the second filter.

3. The method according to claim 1, further comprising a step of subjecting the input speech signal to pitch emphasis and inputting the pitch-emphasized signal to the first filter to be subjected to the spectrum envelop emphasis by the first filter.

4. The method according to claim 1, wherein said step of deriving the filter coefficients includes a step of using weighting factors set in a relation of C1<

C3<

C0, deriving said μ

_p from a value obtained by weighting a first autocorrelation coefficient derived from the filter coefficient of the zero transfer function by the weighting factor C0 when the first autocorrelation is smaller than a threshold value which is approximately zero and weighting the first autocorrelation coefficient by the weighting factor C1 when the first autocorrelation coefficient is larger than the threshold value, and deriving said μ

_z from a value obtained by weighting a second autocorrelation coefficient derived from the filter coefficient of the pole transfer function by the weighting factor C3.

5. The method according to claim 1, further comprising a step of determining a gain needed to set a power of a speech signal whose spectral tilt is compensated to equal a power of the input speech signal.

6. The method according to claim 5, wherein said step of determining the gain includes the steps of:

determining a sign of the gain to be multiplied by the speech signal whose spectral tilt is compensated; and

replacing the gain by a predetermined positive value if the gain is determined to be negative.

7. The method according to claim 5, wherein said step of determining the gain includes the steps of:

determining a sign of the gain to be multiplied by the speech signal whose spectral tilt is compensated; and

replacing the gain by a value greater than or equal to zero and less than one if the gain is determined to be negative.

8. An apparatus for adjusting a spectrum shape of an input speech signal, comprising:

a first filter having a pole-zero transfer function which subjects said input speech signal to a spectrum envelop emphasis; and

a second filter which compensates a spectral tilt of the spectrum shape of the input speech signal caused by said first filter, the second filter including;

a calculator which independently derives two filter coefficients from the pole-zero transfer function of said first filter; and

a filter section which subjects a speech signal output from said first filter to a filtering process using the derived filter coefficients and which compensates the spectral tilt caused by the first filter,

wherein said calculator calculates a first parameter corresponding to a first-order partial autocorrelation coefficient which is approximated to a spectrum envelop of a zero transfer function of said first filter and a second parameter corresponding to a first-order partial autocorrelation coefficient which is approximated to a spectrum envelop of a pole transfer function of said first filter, said calculator inputs the first parameter and the second parameter to said filter section, and said filter section includes a transfer function which uses the first parameter and the second parameter to compensate the spectral tilt caused by the first filter.

9. The apparatus according to claim 8, further comprising a pitch harmonics emphasis filter which subjects the input speech signal to a pitch emphasis and which inputs the pitch-emphasized signal to said first filter to be subjected to the spectrum envelop emphasis by said first filter.

10. The apparatus according to claim 8, further comprising a gain controller which sets a power of a speech signal whose spectral tilt is compensated to equal a power of the input speech signal.

11. An apparatus for adjusting a spectrum shape of an input speech signal, comprising:

a first filter having a pole-zero transfer function which subjects said input speech signal to a spectrum envelop emphasis; and

a second filter which compensates a spectral tilt of the spectrum shape of the input speech signal caused by said first filter, the second filter including;

a calculator which independently derives two filter coefficients from the pole-zero transfer function of said first filter; and

a filter section which subjects a speech signal output from said first filter to a filtering process using the derived filter coefficients and which compensates said spectral tilt caused by the first filter,

wherein said calculator calculates a first parameter corresponding to multiple-order partial autocorrelation coefficients which are approximated to a spectrum envelop of a zero transfer function of said first filter and a second parameter corresponding to multiple-order partial autocorrelation coefficients which are approximated to a spectrum envelop of a pole transfer function of said first filter, said calculator inputs the first parameter and the second parameter to said filter section, and said filter section includes a transfer function which uses the first parameter and the second parameter to compensate the spectral tilt caused by said first filter.

12. An apparatus for adjusting a spectrum shape of an input speech signal, comprising:

a synthesis filter which analyzes said input speech signal to output synthesis filter data;

a calculator which calculates weighting filter data and a pole-zero transfer function using the synthesis filter data output from the synthesis filter; and

a weighting filter which filters the input speech signal using the calculated weighting filter data and the calculated pole-zero transfer function, the weighting filter including a first filter having a first pole-zero transfer function and a second filter having a second pole-zero transfer function, said second filter compensates a spectral tilt of the spectrum shape of the input speech signal caused by the first filter,

wherein the second filter has a function of a first-order zero filter having a z domain transfer function expressed by 1-μ

_z Z^-1 and a function of a first-order pole filter having a z domain transfer function expressed by 1/(1-μ

_p z^-1), where an absolute value of μ

_p is smaller than 1.

13. The apparatus according to claim 12, wherein the weighting filter derives parameters of the second filter from the pole-zero transfer function of the first filter individually and sets a characteristic of the second filter by combining the parameters thereof.

14. An apparatus for adjusting a spectrum shape of an input speech signal, comprising:

a first filter having a pole-zero transfer function represented by transfer functions A(z)/B(z);

a second filter cascade-connected to the first filter and having a first parameter and a second parameter, said second filter compensates characteristics of said first filter; and

parameter deriving means for individually deriving the first parameter and the second parameter from the transfer functions A(z) and B(z),

wherein the parameter deriving means includes a first parameter output section for predicting characteristics of at least one of

     1) the transfer function A(z) and

     2) an inverse transfer function 1/A(z) to derive a first predictive coefficient and to output the first predictive coefficient as the first parameter; and

a second parameter output section for predicting characteristics of at least one of

     1) the transfer function B(z) and

     2) an inverse transfer function 1/B(z) to derive a second predictive coefficient and to output the second predictive coefficient as the second parameter.

15. A method for adjusting a spectrum shape of an input speech signal, comprising the steps of:

preparing a first filter having a pole-zero transfer function represented by A(z)/B(z) and a second filter for compensating characteristics of the first filter, the second filter having a first-order transfer function represented by (1-μ

_z Z^-1)/(1-μ

_p Z^-1), where μ

_z and μ

_p are respective filter coefficients whose absolute values are smaller than 1; and

filtering the speech signal by means of the first and second filters.

16. The method according to claim 15, wherein the step of deriving includes a step of deriving μ

_p from the transfer function A(z) and μ

_z from the transfer function B(z).

17. The method according to claim 16, wherein said step of deriving includes a step of using weighting factors set in a relation of C1<

C3<

C0, deriving said μ

_p from a value obtained by weighting a first autocorrelation coefficient derived from a filter coefficient of the transfer function A(z) by the weighting factor C0 when the first autocorrelation coefficient is smaller than a threshold value which is approximately zero and weighting the first autocorrelation coefficient by the weighting factor C1 when the first autocorrelation coefficient is larger than the threshold value, and deriving said μ

_z from a value obtained by weighting a second autocorrelation coefficient derived from a filter coefficient of the transfer function B(z) by the weighting factor C3.

18. The method according to claim 15, further comprising the steps of:

determining a gain needed to set a power of a speech signal whose spectral tilt is compensated to equal a power of the input speech signal;

determining the sign of the gain to be multiplied by the speech signal whose spectral tilt is compensated; and

replacing the gain by a predetermined positive value if the gain is determined to be negative.

19. The method according to claim 15, further comprising the steps of:

determining a gain needed to set a power of a speech signal whose spectral tilt is compensated to equal a power of the input speech signal;

determining the sign of the gain to be multiplied by the speech signal whose spectral tilt is compensated; and

replacing the gain by a predetermined value which is greater than or equal to zero and less than one if the gain is determined to be negative.

20. A method for adjusting a spectrum shape of an input speech signal, comprising the steps of:

preparing a first filter having a pole-zero transfer function represented by transfer functions A(z)/B(z) and a second filter for compensating characteristics of the first filter, the second filter having a first-order transfer function represented by (1-μ

_z Z^-1)/(1-μ

_p Z^-1), where μ

_z and μ

_p are respective filter coefficients whose absolute values are smaller than 1;

deriving two parameters used in the second filter from the transfer functions A(z) and B(z) individually; and

filtering the speech signal by means of the first and second filters.

21. The method according to claim 20, further comprising the steps of:

determining a gain needed to set a power of a speech signal whose spectral tilt is compensated to equal a power of the input speech signal;

determining the sign of the gain to be multiplied by the speech signal whose spectral tilt is compensated; and

replacing the gain by a predetermined positive value if the gain is determined to be negative.

22. The method according to claim 20, further comprising the steps of:

determining a gain needed to set a power of a speech signal whose spectral tilt is compensated to equal a power of the input speech signal;

determining the sign of the gain to be multiplied by the speech signal whose spectral tilt is compensated; and

replacing the gain by a predetermined value which is greater than or equal to zero and less than one if the gain is determined to be negative.