Method for conditioning a digital speech signal

• US 6,775,650 B1
• Filed: 06/02/2000
• Issued: 08/10/2004
• Est. Priority Date: 09/18/1997
• Status: Expired due to Term

• Current Assignee: Apple Inc.
• Original Assignee: Nortel Networks Corporation
• Inventors: Lockwood, Philip, Lubiarz, Stephane
• Primary Examiner(s): Dorvil, Richemond
• Assistant Examiner(s): Lerner, Martin
• Application Number:
  US09/509,146
• Time in Patent Office: 1,530 Days
• Field of Search: 704/203, 704/205, 704/207, 704/208, 704/216, 704/217, 704/222, 704/267, 704/268, 704/220, 704/265
• US Class Current: 704/205
• CPC Class Codes: G10L 19/0212 using orthogonal transformation
• CPC Class Codes: G10L 21/0232 Processing in the frequency domain
• CPC Class Codes: G10L 21/0364 for improving intelligibility
• CPC Class Codes: G10L 25/78 Detection of presence or absence of voice signals switching of direction of transmission by voice frequency in two-way loud-speaking telephone systems H04M9/10
• CPC Class Codes: G10L 25/93 Discriminating between voiced and unvoiced parts of speech signals G10L25/90 takes precedence

Abstract

The invention concerns a method for conditioning a digital speech signal(s) processed by successive frames, which consists carrying out a harmonic analysis to estimate the pitch on each frame where it has a speech activity, and in oversampling at an oversampling frequency (f_e) which is a multiple of the estimated pitch.

Claims

1. Method of conditioning a digital speech signal processed by successive frames, comprising a harmonic analysis of the speech signal to estimate a pitch frequency of the speech signal over each frame in which the speech signal features vocal activity, and, after estimating the pitch frequency of the speech signal over one frame, conditioning the speech signal of said one frame by oversampling the speech signal in the time domain at an oversampling frequency which is an integer multiple of the estimated pitch frequency.

2. Method according to claim 1, wherein spectral components of the speech signal are computed by distributing the conditioned signal into blocks of N samples transformed into the frequency domain, N being a predetermined integer, and wherein the ratio between the oversampling frequency and the estimated pitch frequency is a factor of the number N.

3. Method according to claim 2, wherein the number N is a power of 2.

4. Method according to claim 2, wherein a degree of voicing of the speech signal is estimated over the frame from an entropy of an autocorrelation of spectral components computed on the basis of the conditioned signal.

5. Method according to claim 4, wherein the degree of voicing is measured on the basis of a normalised entropy H of the form:

$H=\frac{\underset{k=0}{\overset{N/2-1}{\sum <br><br>}}<br><br>A<br><br>\left(k\right)·<br><br>\log <br><br>\left[A<br><br>\left(k\right)\right]}{\log <br><br>\left(N/2\right)}$

where A(k) is the normalised autocorrelation defined by;

$A<br><br>\left(k\right)=\frac{\underset{f=0}{\overset{N/2-1}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>S_{n,f}^2·<br><br>S_{n,f+k}^2}{\underset{f=0}{\overset{N/2-1}{\sum <br><br>}}<br><br>\underset{f^{\mathrm{\prime <br><br>}}=0}{\overset{N/2-1}{\sum <br><br>}}<br><br>S_{n,f}^2·<br><br>S_{n,f+f^{\mathrm{\prime <br><br>}}}^2}$

S_n,f² designating said spectral component of rank f computed on the basis of the oversampled signal.

6. Method according to claim 1, wherein, after processing each conditioned signal frame, a number of signal samples supplied by such processing is retained which is equal to an integer multiple of the ratio between an initial sampling frequency and the estimated pitch frequency.

7. Method according to claim 1, wherein the estimation of the pitch frequency of the speech signal over a frame includes the steps of:

estimating time intervals between two consecutive breaks of the signal which can be attributed to glottal closures of speaker occurring during the frame, the estimated pitch frequency being inversely proportional to said time intervals;

interpolating the speech signal in said time intervals, so that the conditioned signal resulting from such interpolation has a constant time interval between two consecutive breaks.

8. Method according to claim 7, wherein, after processing each frame, a number of samples of the speech signal supplied by such processing is retained which corresponds to an integer number of estimated time intervals.

9. Device for conditioning a digital speech signal processed by successive frames, comprising harmonic analysis means to estimate a pitch frequency of the speech signal over each frame in which the speech signal features vocal activity, and conditioning means for conditioning the speech signal of said frame by oversampling the speech signal in the time domain at an oversampling frequency which is an integer multiple of the estimated pitch frequency.

10. Device according to claim 9, distributing the conditioned signal into blocks of N samples, N being a predetermined integer, and means for computing spectral components of the speech signal by transforming said blocks into the frequency domain, and wherein the ratio between the oversampling frequency and the estimated pitch frequency is a factor of the number N.

11. Device according to claim 10, wherein the number N is a power of 2.

12. Device according to claim 10, further comprising means for estimating a degree of voicing of the speech signal over each frame from an entropy of an autocorrelation of spectral components computed on the basis of the conditioned signal.

13. Device according to claim 12, wherein the degree of voicing is measured on the basis of a normalised entropy H of the form:

$H=\frac{\underset{k=0}{\overset{N/2-1}{\sum <br><br>}}<br><br>A<br><br>\left(k\right)·<br><br>\log <br><br>\left[A<br><br>\left(k\right)\right]}{\log <br><br>\left(N/2\right)}$

where A(k) is the normalised autocorrelation defined by;

$A<br><br>\left(k\right)=\frac{\underset{f=0}{\overset{N/2-1}{\sum <br><br>}}<br><br>\text{\hspace{1em}<br><br>}<br><br>S_{n,f}^2·<br><br>S_{n,f+k}^2}{\underset{f=0}{\overset{N/2-1}{\sum <br><br>}}<br><br>\underset{f^{\mathrm{\prime <br><br>}}=0}{\overset{N/2-1}{\sum <br><br>}}<br><br>S_{n,f}^2·<br><br>S_{n,f+f^{\mathrm{\prime <br><br>}}}^2}$

S_n,f² designating said spectral component of rank f computed on the basis of the oversampled signal.

14. Device according to claim 9, wherein, after processing each conditioned signal frame, a number of signal samples supplied by such processing is retained which is equal to an integer multiple of the ratio between an initial sampling frequency and the estimated pitch frequency.

15. Device according to claim 9, wherein the harmonic analysis means include:

means for estimating time intervals between two consecutive breaks of the signal which can be attributed to glottal closures of a speaker occurring during a frame, the estimated pitch frequency being inversely proportional to said time intervals;

means for interpolating the speech signal in said time intervals, so that the conditioned signal resulting from such interpolation has a constant time interval between two consecutive breaks.

16. Device according to claim 15, wherein, after processing each frame, a number of samples of the speech signal supplied by such processing is retained which corresponds to an integer number of estimated time intervals.