Method of using microphone characteristics to optimize speech recognition performance

• US 8,600,741 B2
• Filed: 08/20/2008
• Issued: 12/03/2013
• Est. Priority Date: 08/20/2008
• Status: Active Grant

• Current Assignee: General Motors LLC (General Motors Company)
• Original Assignee: General Motors LLC (General Motors Company)
• Inventors: Talwar, Gaurav, Chengalvarayan, Rathinavelu, Gratke, Jesse T., Gullapalli, Subhash B., Fecher, Dana B.
• Primary Examiner(s): Harper, Vincent P
• Application Number:
  US12/194,605
• Publication Number:
  US 20100049516A1
• Time in Patent Office: 1,931 Days
• Field of Search: 704/231, 704/234, 704/243, 704/244
• US Class Current: 704/231
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/07 to the speaker

Abstract

A system and method for tuning a speech recognition engine to an individual microphone using a database containing acoustical models for a plurality of microphones. Microphone performance characteristics are obtained from a microphone at a speech recognition engine, the database is searched for an acoustical model that matches the characteristics, and the speech recognition engine is then modified based on the matching acoustical model.

Claims

1. A method of tuning a speech recognition engine to an individual microphone, the method comprising:

(a) providing a database containing acoustical models for a plurality of microphones;

(b) receiving microphone performance characteristics from a microphone at a speech recognition engine;

(c) searching the database for an acoustical model that matches the characteristics;

(d) modifying the speech recognition engine based on the matching acoustical model; and

(e) if no matching acoustical model is found, making adjustments to a feature extraction phase of speech recognition and using a default acoustic model modified by the adjustments to the feature extraction phase, wherein the adjustments include reducing cepstral parameters up to a high cut-off frequency such that only frequencies below the high cut-off frequency will be used for speech extraction.

2. The method of claim 1, wherein the microphone performance characteristics comprise at least one of frequency response, sensitivity, or signal bandwidth.

3. The method of claim 1, wherein the microphone performance characteristics are determined by testing the microphone and the determined results are recorded on the microphone.

4. The method of claim 1, wherein the microphone is carried by a vehicle.

5. The method of claim 1, wherein the database containing acoustical models is located at a call center.

6. The method of claim 1, wherein the microphone stores the microphone performance characteristics at the microphone.

7. The method of claim 5, wherein the microphone performance characteristics are stored on a microphone using a Transducer Electronic Data Sheet (TEDS).

8. The method of claim 1, wherein step (d) comprises limiting a high-frequency encoding (HFE) algorithm using the microphone performance characteristics.

9. A method of tuning a speech recognition engine to an individual microphone, the method comprising:

(a) receiving microphone performance characteristics that are stored at a microphone;

(b) searching a database for an acoustical model that matches the microphone performance characteristics;

(c) if an acoustical model matches the microphone performance characteristics, uploading the matching acoustical model and applying the model to the speech recognition engine; and

(d) if an acoustical model does not match the microphone performance characteristics, making adjustments to a feature extraction phase of speech recognition and using a default acoustic model modified by the adjustments to the feature extraction phase wherein the adjustments include reducing cepstral parameters up to a high cut-off frequency such that only frequencies below the high cut-off frequency will be used for speech extraction.

10. The method of claim 9, wherein the microphone performance characteristics comprise at least one of frequency response, sensitivity, or signal bandwidth.

11. The method of claim 9, wherein the microphone performance characteristics comprise data indicating the frequency response of the microphone.

12. The method of claim 9, wherein the microphone is carried by a vehicle.

13. The method of claim 9, further comprising providing a database containing acoustical models that is located at a call center.

14. The method of claim 9, wherein the microphone is a digital microphone.

15. The method of claim 9, wherein the microphone performance characteristics are stored on a microphone using a Transducer Electronic Data Sheet (TEDS).

16. The method of claim 9, wherein steps (c) and (d) comprise limiting a high-frequency encoding (HFE) algorithm using the data parameters.

17. A method of tuning a speech recognition engine to an individual microphone, the method comprising:

(a) providing a database containing acoustical models for a plurality of microphones;

(b) receiving data parameters from a digital microphone at a speech recognition engine, wherein the data parameters indicates the performance characteristics of the microphone;

(c) searching the database for an acoustical model that matches the data parameters;

(d) if an acoustical model is not matched to the data parameters, sending new data parameters to the digital microphone;

(e) reconfiguring the digital microphone to perform based on the new data parameters; and

(f) saving the new data parameters at the microphone.

18. The method of claim 17, wherein the digital microphone includes a bandpass filter or a digital filter.