AUTOMATED DISTORTION CLASSIFICATION

• US 20110125500A1
• Filed: 11/25/2009
• Published: 05/26/2011
• Est. Priority Date: 11/25/2009
• Status: Active Grant

• Current Assignee: General Motors LLC (General Motors Company)
• Original Assignee: General Motors LLC (General Motors Company)
• Inventors: Chengalvarayan, Rathinavelu, Talwar, Gaurav
• Granted Patent: US 8,438,030 B2
• Time in Patent Office: Days
• US Class Current: 704/251
• 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 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices
• CPC Class Codes: G10L 21/0208 Noise filtering

Abstract

A method of and system for automated distortion classification. The method includes steps of (a) receiving audio including a user speech signal and at least some distortion associated with the signal; (b) pre-processing the received audio to generate acoustic feature vectors; (c) decoding the generated acoustic feature vectors to produce a plurality of hypotheses for the distortion; and (d) post-processing the plurality of hypotheses to identify at least one distortion hypothesis of the plurality of hypotheses as the received distortion. The system can include one or more distortion models including distortion-related acoustic features representative of various types of distortion and used by a decoder to compare the acoustic feature vectors with the distortion-related acoustic features to produce the plurality of hypotheses for the distortion.

Claims

1. A method of automated distortion classification, comprising the steps of:

(a) receiving audio including a user speech signal and at least some distortion associated with the signal;

(b) pre-processing the received audio to generate acoustic feature vectors;

(c) decoding the generated acoustic feature vectors to produce a plurality of hypotheses for the distortion; and

(d) post-processing the plurality of hypotheses to identify at least one distortion hypothesis of the plurality of hypotheses as the received distortion.

2. The method of claim 1, further comprising the step of:

(e) using the identified at least one distortion hypothesis to improve speech signal processing.

3. The method of claim 2 wherein step (e) includes emphasizing acoustic features that correlate to the user speech signal and deemphasizing acoustic features that correlate to the distortion.

4. The method of claim 2 wherein step (e) includes omitting acoustic features that correlate to the distortion.

5. The method of claim 1, wherein steps (a)-(d) are carried out by an automatic speech recognition system installed in a vehicle.

6. A method of automated distortion classification using an automated speech recognition system including an acoustic interface, a pre-processor, a decoder using acoustic models, and a post-processor, the method comprising the steps of:

(a) receiving audio including a user speech signal and at least some distortion associated with the signal;

(b) pre-processing the received audio to generate acoustic feature vectors;

(c) decoding the generated acoustic feature vectors with the decoder to produce a plurality of hypotheses for the distortion; and

(d) post-processing the plurality of hypotheses produced from the decoder to identify at least one distortion hypothesis of the plurality of hypotheses as the received distortion.

7. The method of claim 6, further comprising the step of:

(e) using the identified at least one distortion hypothesis from the post-processor to improve speech recognition.

8. The method of claim 7 wherein step (e) includes using the identified at least one distortion hypothesis from the post-processor to modify at least one of the acoustic interface, the pre-processor, or the acoustic models during a current speech recognition session.

9. The method of claim 8 wherein step (e) includes applying, or adjusting parameters of, electronic signal conditioners, electronic signal filters, noise suppression algorithms, or adaptive filters.

10. The method of claim 8 wherein step (e) includes emphasizing acoustic features that correlate to the user speech signal and deemphasizing acoustic features that correlate to the distortion.

11. The method of claim 8 wherein step (e) includes at least one of emphasizing acoustic features that correlate to the user speech signal or deemphasizing acoustic features that correlate to the distortion.

12. The method of claim 11 wherein step (e) is applied in the pre-processor.

13. The method of claim 11 wherein step (e) is applied in the acoustic models.

14. The method of claim 8 wherein step (e) includes omitting acoustic features that correlate to the distortion.

15. The method of claim 14 wherein step (e) is applied in the pre-processor.

16. The method of claim 14 wherein step (e) is applied in the acoustic models.

17. An automated distortion classification system comprising:

an acoustic interface to receive audio including a user speech signal and at least some distortion associated with the signal;

a pre-processor to pre-process the received audio to generate acoustic feature vectors;

a decoder to decode the generated acoustic feature vectors;

at least one distortion model including a plurality of distortion-related acoustic features representative of various types of distortion and used by the decoder to compare the acoustic feature vectors with the distortion-related acoustic features to produce a plurality of hypotheses for the distortion; and

a post-processor to post-process the plurality of hypotheses produced from the decoder to identify at least one distortion hypothesis of the plurality of hypotheses as the received distortion.

18. The system of claim 17 wherein the identified at least one distortion hypothesis from the post-processor is used to modify at least one of the acoustic interface, the pre-processor, or the acoustic models during a current speech recognition session.

19. The system of claim 18 wherein various parameters of at least one of electronic signal conditioners, electronic signal filters, noise suppression algorithms, or adaptive filters are applied or adjusted in response to the identified at least one distortion hypothesis.

20. The system of claim 18 wherein acoustic features that correlate to the user speech signal are emphasized and acoustic features that correlate to the distortion are deemphasized.