SYSTEM, METHOD AND PROGRAM FOR SPEECH PROCESSING

• US 20090210224A1
• Filed: 08/28/2008
• Published: 08/20/2009
• Est. Priority Date: 08/31/2007
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Ichikawa, Osamu, Fukuda, Takashi, Nishimura, Masafumi
• Granted Patent: US 8,812,312 B2
• Time in Patent Office: Days
• US Class Current: 704/233
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/20 Speech recognition techniques specially adapted for robustness in adverse environments, e.g. in noise, of stress induced speech G10L21/02 takes precedence
• CPC Class Codes: G10L 25/24 the extracted parameters being the cepstrum

Abstract

The present invention relates to a system, method and program for speech recognition. In an embodiment of the invention a method for processing a speech signal consists of receiving a power spectrum of a speech signal and generating a log power spectrum signal of the power spectrum. The method further consists of performing discrete cosine transformation on the log power spectrum signal and cutting off cepstrum upper and lower terms of the discrete cosine transformed signal. The method further consists of performing inverse discrete cosine transformation on the signal from which the cepstrum upper and lower terms are cut off. The method further consists of converting the inverse discrete cosine transformed signal so as to bring the signal back to a power spectrum domain and filtering the power spectrum of the speech signal by using, as a filter, the signal which is brought back to the power spectrum domain.

Claims

1. A speech processing system for processing a speech signal using a computer, comprising:

a first means for receiving a power spectrum of a speech signal and generating a log power spectrum of the power spectrum;

a second means for performing discrete cosine transformation on an output from the second means;

a third means for receiving an output from the second means to cut off cepstrum upper and lower terms of the output;

a fourth means for receiving an output from the third means to perform inverse discrete cosine transformation on the output;

a fifth means for converting an output from the fourth means so as to bring the output back to a power spectrum domain; and

a sixth means for filtering the power spectrum of the speech signal by using, as a filter, the output which is brought back to the power spectrum domain.

2. The speech processing system according to claim 1, wherein the fifth means further includes means for normalizing the output from the fourth means.

3. The speech processing system according to claim 1, wherein the third means performs cutting off to obtain terms in a domain corresponding to a harmonic structure in a possible range as a human speech.

4. A speech processing method for processing a speech signal using a computer, comprising:

receiving a power spectrum of a speech signal and generating a log power spectrum signal of the power spectrum;

performing discrete cosine transformation on the log power spectrum signal;

cutting off cepstrum upper and lower terms of the discrete cosine transformed signal;

performing inverse discrete cosine transformation on the signal from which the cepstrum upper and lower terms are cut off;

converting the inverse discrete cosine transformed signal so as to bring the signal back to a power spectrum domain; and

filtering the power spectrum of the speech signal by using, as a filter, the signal which is brought back to the power spectrum domain.

5. The method of claim 4, further comprising:

normalizing the signal which is brought back to the power spectrum domain.

6. The method of claim 4, wherein the cutting off the cepstrum upper and lower terms are performed to obtain terms in a domain corresponding to a harmonic structure in a possible range as a human speech.

7. A computer program product comprising a computer useable medium including a computer readable program, wherein the speech processing program when executed on a computer causes the computer to perform the method steps for operating a computer to process a speech signal for speech recognition. The method comprising the steps of:

receiving a power spectrum of a speech signal and generating a log power spectrum signal of the power spectrum;

performing discrete cosine transformation on the log power spectrum signal;

cutting off cepstrum upper and lower terms of the discrete cosine transformed signal;

performing inverse discrete cosine transformation on the signal from which the cepstrum upper and lower terms are cut off;

converting the inverse discrete cosine transformed signal so as to bring the signal back to a power spectrum domain; and

filtering the power spectrum of the speech signal by using, as a filter, the signal which is brought back to the power spectrum domain.

8. The computer program product of claim 7, further comprising:

normalizing the signal which is brought back to the power spectrum domain.

9. The computer program product of claim 7, wherein the cutting off the cepstrum upper and lower terms are performed to obtain terms in a domain corresponding to a harmonic structure in a possible range as a human speech.

10. A speech recognition system for performing speech recognition using a computer, comprising:

a first means for receiving a power spectrum of a speech signal and generating a log power spectrum of the power spectrum;

a second means for performing discrete cosine transformation on an output of the log power spectrum;

a third means for receiving an output from the second means to cut off cepstrum upper and lower terms of the output;

a fourth means for receiving an output from the third means to perform inverse discrete cosine transformation on the output;

a fifth means for converting an output from the fourth means so as to bring the output back to a power spectrum domain; and

a sixth means for filtering the power spectrum of the speech signal by using, as a filter, the output which is brought back to the power spectrum domain, wherein speech recognition processing is performed by using the filtered power spectrum.

11. The speech recognition system according to claim 10, wherein the fifth means further includes means for normalizing the output from the fourth means.

12. A speech output system for outputting a speech captured with a microphone using a computer, comprising:

a first means for performing A/D conversion on the speech captured with the microphone to output a digital speech signal;

a second means for performing discrete Fourier transformation on the digital speech signal to output a power spectrum of the speech signal;

a third means for receiving the power spectrum of the speech signal to generate a log power spectrum of the power spectrum;

a fourth means for performing discrete cosine transformation on an output of the log power spectrum;

a fifth means for receiving an output from the fourth means to cut off cepstrum upper and lower terms of the output;

a sixth means for receiving an output from the fifth means to perform inverse discrete cosine transformation on the output;

a seventh means for converting an output from the sixth means so as to bring the output back to a power spectrum domain;

an eighth means for filtering the power spectrum of the speech signal by using, as a filter, the output which is brought back to the power spectrum domain; and

a ninth means for performing D/A conversion on the filtered power spectrum to output an analog speech signal.

13. A speech processing system for processing a speech signal using a computer, comprising:

a first means for receiving a power spectrum of a speech signal and generating a log power spectrum of the power spectrum;

a second means for performing discrete cosine transformation on an output from an output of the log power spectrum;

a third means for receiving an output from the second means to cut off cepstrum upper and lower terms of the output;

a fourth means for receiving an output from the third means to perform inverse discrete cosine transformation on the output;

a fifth means for converting an output from the fourth means so as to bring the output back to a power spectrum domain; and

a sixth means for filtering the power spectrum of the speech signal by using, as a filter, the output which is brought back to the power spectrum domain.