Unsupervised incremental adaptation using maximum likelihood spectral transformation

• US 20060009972A1
• Filed: 08/30/2005
• Published: 01/12/2006
• Est. Priority Date: 11/16/2000
• Status: Active Grant

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Lubensky, David M., Yuk, Dongsuk
• Granted Patent: US 7,269,555 B2
• Time in Patent Office: Days
• US Class Current: 704/234
• CPC Class Codes: G10L 15/065 Adaptation
• 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 21/0216 characterised by the method used for estimating noise

Abstract

In a speech recognition system, a method of transforming speech feature vectors associated with speech data provided to the speech recognition system includes the steps of receiving likelihood of utterance information corresponding to a previous feature vector transformation, estimating one or more transformation parameters based, at least in part, on the likelihood of utterance information corresponding to a previous feature vector transformation, and transforming a current feature vector based on maximum likelihood criteria and/or the estimated transformation parameters, the transformation being performed in a linear spectral domain. The step of estimating the one or more transformation parameters includes the step of estimating convolutional noise N_i^α and additive noise N_i^β for each ith component of a speech vector corresponding to the speech data provided to the speech recognition system.

Claims

1. In a speech recognition system, a method of transforming speech feature vectors associated with speech data provided to the speech recognition system, the method comprising the steps of:

receiving likelihood of utterance information corresponding to a previous feature vector transformation;

estimating one or more transformation parameters based, at least in part, on the likelihood of utterance information corresponding to a previous feature vector transformation; and

transforming a current feature vector based on at least one of maximum likelihood criteria and the estimated transformation parameters, the transformation being performed in a linear spectral domain;

wherein the step of estimating the one or more transformation parameters comprises the step of estimating convolutional noise N_i^α

and additive noise N_i^β

for each ith component of a speech vector corresponding to the speech data provided to the speech recognition system.

2. The method of claim 1, wherein the step of transforming the current feature vector is performed in feature space.

3. The method of claim 1, wherein the step of transforming the current feature vector is performed in model space.

4. The method of claim 1, wherein the maximum likelihood criteria is a maximum likelihood spectral transformation (MLST).

5. The method of claim 1, wherein the step of estimating the one or more transformation parameters further comprises the step of defining a diagonal matrix A with A_ii=1/N_i^α

, and defining b_i=−

N_i^β

/N_i^α

.

6. The method of claim 5, further comprising the steps of:

determining A_ii in accordance with an expression

$A_{\mathrm{ii}}=\frac{T<br><br>\text{\hspace{1em}<br><br>}<br><br>{\mathrm{\Sigma <br><br>}}_t<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}<br><br>m_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}-{\mathrm{\Sigma <br><br>}}_t<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}<br><br>{\mathrm{\Sigma <br><br>}}_t<br><br>m_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}}{T<br><br>\text{\hspace{1em}<br><br>}<br><br>{\mathrm{\Sigma <br><br>}}_t<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)<br><br>2}-{\mathrm{\Sigma <br><br>}}_t<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}<br><br>{\mathrm{\Sigma <br><br>}}_t<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}};<br><br>$

determining b_i in accordance with an expression

$b_i=\frac{-A_{\mathrm{ii}}<br><br>{\mathrm{\Sigma <br><br>}}_t<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}+{\mathrm{\Sigma <br><br>}}_t<br><br>m_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}}{T};<br><br>$

where x_t,i<sup>(ε

)</sup> and m_t,i<sup>(ε

)</sup> are sub-linear spectral values of a feature vector and corresponding mean vector, respectively, for each ith component of the speech vector.

7. The method of claim 1, wherein the step of transforming the current feature vector further comprises the step of determining

8. An article of manufacture for transforming speech feature vectors associated with speech data provided to the speech recognition system, comprising a machine readable medium containing one or more programs which when executed implement the steps of:

receiving likelihood of utterance information corresponding to a previous feature vector transformation;

estimating one or more transformation parameters based, at least in part, on the likelihood of utterance information corresponding to a previous feature vector transformation; and

transforming a current feature vector based on at least one of maximum likelihood criteria and the estimated transformation parameters, the transformation being performed in a linear spectral domain;

wherein the step of estimating the one or more transformation parameters comprises the step of estimating convolutional noise N_i^α

and additive noise N_i^β

for each ith component of a speech vector corresponding to the speech data provided to the speech recognition system.

9. The article of claim 8, wherein the step of transforming the current feature vector is performed in a feature space.

10. The article of claim 8, wherein the step of transforming the current feature vector is performed in a model space.

11. The article of claim 8, wherein the maximum likelihood criteria is a maximum likelihood spectral transformation (MLST).

12. The article of claim 8, wherein the step of estimating the one or more transformation parameters further includes the step of defining a diagonal matrix A with A_ii=1/N_i^α

, and defining b_i=−

N_i^β

/N_i^α

.

13. The article of claim 12, wherein the step of estimating the one or more transformation parameters further comprises the steps of:

determining A_ii in accordance with an expression

$A_{\mathrm{ii}}=\frac{T<br><br>\underset{t}{\sum <br><br>}<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}<br><br>m_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}-\underset{t}{\sum <br><br>}<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}<br><br>\underset{t}{\sum <br><br>}<br><br>m_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}}{T<br><br>\underset{t}{\sum <br><br>}<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)<br><br>2}-\underset{t}{\sum <br><br>}<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}<br><br>\underset{t}{\sum <br><br>}<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}};<br><br>$

and

determining b_i in accordance with an expression

$b_i=\frac{-A_{\mathrm{ii}}<br><br>\underset{t}{\sum <br><br>}<br><br>x_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}<br><br>\underset{t}{\sum <br><br>}<br><br>m_{t,i}^{\left(\mathrm{\varepsilon <br><br>}\right)}}{T};<br><br>$

where x_t,i<sup>(ε

)</sup> and m<sup>t,i(ε

)</sup> are sub-linear spectral values of a feature vector and corresponding mean vector, respectively, for each ith component of the speech vector.

14. The article of claim 8, wherein the step of transforming the current feature vector includes the step of determining

15. Apparatus for transforming speech feature vectors associated with speech data provided to a speech recognition system, the apparatus comprising:

at least one processing device operative;

(i) to receive likelihood of utterance information corresponding to a previous feature vector transformation;

(ii) to estimate one or more transformation parameters based, at least in part, on the likelihood of utterance information corresponding to a previous feature vector transformation;

(iii) to transform a current feature vector based on at least one of maximum likelihood criteria and the estimated transformation parameters, the transformation being performed in a linear spectral domain; and

(iv) to estimate convolutional noise N_i^α

and additive noise N_i^β

for each ith component of a speech vector corresponding to the speech data provided to the speech recognition system.