Granted Patent B2

US 7,881,929 · App. 11/828,093 · Granted Feb 1, 2011

Abstract

Claims (56)

1. A method of ambient noise injection for use with speech recognition in a production vehicle, the method comprising the steps of:

monitoring audio including user speech;

receiving an utterance from the user speech;

retrieving vehicle-specific ambient noise;

prepending the vehicle-specific ambient noise to the utterance before pre-processing and decoding the utterance;

pre-processing the utterance including the vehicle-specific ambient noise; and

decoding the pre-processed utterance including the vehicle-specific ambient noise.

2. The method set forth in claim 1 , wherein the vehicle-specific ambient noise was generated in at least one test vehicle of the same make and model as the production vehicle.

3. The method set forth in claim 1 , wherein the vehicle-specific ambient noise was generated in at least one test vehicle of the same type as the production vehicle.

4. The method set forth in claim 1 , wherein the vehicle-specific ambient noise was generated in at least one test vehicle of the same segment as the production vehicle.

5. The method set forth in claim 1 , wherein the vehicle-specific ambient noise was generated according to a plurality of different vehicle speeds.

6. The method set forth in claim 1 , wherein the vehicle-specific ambient noise was generated according to a plurality of different HVAC settings.

7. The method set forth in claim 1 , wherein the vehicle-specific ambient noise was generated according to a plurality of different vehicle parameters.

8. The method set forth in claim 1 , wherein the vehicle-specific ambient noise is from at least one trailing frame associated with a user utterance preceding the received utterance.

9. The method set forth in claim 1 , wherein the vehicle-specific ambient noise was generated in at least one test vehicle corresponding to the production vehicle and is selected as a proxy for at least one trailing frame associated with a user utterance preceding the received utterance.

10. The method set forth in claim 1 , further comprising the steps of:

identifying at least one speechless frame trailing the utterance;

storing the speechless frame(s);

receiving a subsequent utterance from the user speech;

retrieving the stored speechless frame(s); and

prepending the speechless frame(s) to the subsequent utterance.

11. The method set forth in claim 1 , further comprising:

identifying no speechless frames trailing the utterance;

estimating speech energies in at least one frame trailing the utterance;

comparing the estimated speech energies with speech energies of frames having the vehicle-specific ambient noise, and which are pre-recorded in at least one test vehicle corresponding to the production vehicle;

identifying at least one of the pre-recorded vehicle-specific ambient noise frames having speech energies corresponding to the estimated speech energies;

storing the corresponding pre-recorded vehicle-specific ambient noise frame(s);

receiving a subsequent utterance from the user speech;

retrieving the stored pre-recorded vehicle-specific ambient noise frame(s); and

prepending the retrieved pre-recorded vehicle-specific ambient noise frame(s) to the subsequent utterance.

12. A method of ambient noise injection for speech recognition in a production vehicle, the method comprising the steps of:

initiating a current speech recognition session;

monitoring audio including user speech;

receiving an utterance from the user speech;

determining that the received utterance is a first utterance following the initiation of the current speech recognition session;

retrieving at least one vehicle-specific ambient noise frame from ambient noise frames pre-recorded in a test vehicle corresponding to the production vehicle;

prepending the vehicle-specific ambient noise frame(s) to the utterance(s);

pre-processing the utterance including the ambient noise frame(s); and

decoding the pre-processed utterance including the ambient noise frame(s).

13. The method set forth in claim 12 , further comprising the steps of:

identifying at least one speechless frame trailing the utterance; and

storing the speechless frame(s).

14. The method set forth in claim 13 , further comprising the steps of:

receiving a subsequent utterance from the user speech;

retrieving the stored speechless frame(s); and

prepending the speechless frame(s) to the subsequent utterance.

15. The method set forth in claim 12 , further comprising the steps of:

identifying no speechless frames trailing the utterance;

estimating speech energies in at least one frame trailing the utterance;

comparing the estimated speech energies with speech energies of the pre-recorded ambient noise frames;

identifying at least one of the pre-recorded ambient noise frames having speech energies corresponding to the estimated speech energies; and

storing the corresponding pre-recorded ambient noise frame(s).

16. The method set forth in claim 15 , further comprising the steps of:

receiving a subsequent utterance from the user speech;

retrieving the stored pre-recorded ambient noise frame(s); and

prepending the retrieved pre-recorded ambient noise frame(s) to the subsequent utterance.

Assignments (14)