Dynamic clustering of nametags in an automated speech recognition system

• US 9,245,526 B2
• Filed: 04/25/2006
• Issued: 01/26/2016
• Est. Priority Date: 04/25/2006
• Status: Active Grant

• Current Assignee: General Motors LLC (General Motors Company)
• Original Assignee: General Motors LLC (General Motors Company)
• Inventors: Chengalvarayan, Rathinavelu
• Primary Examiner(s): Serrou, Abdelali
• Application Number:
  US11/380,149
• Publication Number:
  US 20070250320A1
• Time in Patent Office: 3,563 Days
• Field of Search: 704/245, 704/201, 704251-255, 704/275
• US Class Current: 1/1
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

A speech recognition method includes receiving a nametag utterance, decoding the nametag utterance to recognize constituent subwords of the nametag utterance, determining the number of subwords in the nametag utterance, and associating the nametag utterance with one or more of a plurality of different nametag clusters based on the number of subwords in the nametag utterance. According to preferred aspects of the method, a confusability check is performed on the nametag utterance within the cluster(s) associated with the nametag utterance, stored nametags are received from memory by decoding the nametag utterance within the cluster(s) associated with the nametag utterance, and the stored nametags are played back by cluster.

Claims

1. A speech recognition method comprising the steps of:

(a) receiving a nametag utterance via a microphone associated with an automatic speech recognition system;

(b) decoding the nametag utterance to recognize constituent subwords of the nametag utterance using at least one processor associated with the automatic speech recognition system;

(c) determining the number of subwords in the nametag utterance;

(d) associating the nametag utterance with one or more of a plurality of different nametag clusters based on the number of subwords in the nametag utterance;

(e) storing an associated nametag in memory, as representing the nametag utterance, among a plurality of stored nametags; and

(f) playing stored nametags back by cluster.

2. The method of claim 1 wherein the associating step (d) includes associating the nametag utterance with less than all of the plurality of different nametag clusters based on the number of subwords in the nametag utterance.

3. The method of claim 2 wherein the associating step (d) includes associating the nametag utterance with no more than two of the plurality of different nametag clusters based on the number of subwords in the nametag utterance.

4. The method of claim 3 wherein the associating step (d) includes associating the nametag utterance with only one of the plurality of different nametag clusters based on the number of subwords in the nametag utterance.

5. The method of claim 1 wherein step (b) comprises generating multiple nametag templates and wherein the determining step (c) includes determining the number of subwords by analyzing the multiple nametag templates.

6. The method of claim 5, wherein the determining step (c) includes calculating an average number of subwords in the multiple templates.

7. The method of claim 5, wherein the determining step (c) includes using the number of subwords from a highest ranking template of the multiple templates.

8. The method of claim 1, wherein the associating step (d) includes comparing the number of subwords to at least one predetermined threshold number and selecting one or more of the plurality of different nametag clusters based on the results of the comparison.

9. The method of claim 1, further comprising the step of (e) performing a confusability check on the nametag utterance within the cluster(s) associated with the nametag utterance.

10. The method of claim 9, further comprising the steps of (f) associating a nametag with the nametag utterance based on a best score from the step of performing a confusability check and storing the associated nametag in memory as representing the nametag utterance.

11. The method of claim 10 further comprising the step of (g) retrieving a stored nametag from memory by:

receiving a second nametag utterance;

decoding the second nametag utterance to recognize constituent subwords of the second nametag utterance;

determining the number of subwords in the second nametag utterance;

associating the second nametag utterance with one or more of the plurality of different nametag clusters based on the number of subwords in the second nametag utterance; and

further decoding the second nametag utterance within the one or more of the plurality of different nametag clusters of the step of associating the second nametag utterance to correctly match the second nametag utterance to its corresponding stored nametag.

13. The method of claim 1 further comprising the step of retrieving a stored nametag from memory by decoding the nametag utterance within the cluster(s) associated with the nametag utterance.

14. The method of claim 13, further comprising the step of calculating scores reflecting the likelihood that nametag templates match the nametag utterance within the cluster(s) associated with the nametag utterance to yield an N-best list ordered by likelihood scores.

15. The method of claim 14, further comprising the step of associating one of the nametag templates with the nametag utterance based on a best score from the likelihood scores.

16. The method of claim 14, further comprising the step of weighting the likelihood scores with the subword count to reorder the N-best list.

17. The method of claim 13, wherein the associating step (d) includes associating the nametag utterance with at least two of the plurality of different nametag clusters based on the number of subwords in the nametag utterance, and the decoding includes weighting the nametags within the clusters differently.

18. The method of claim 17, wherein the nametags are weighted by the average number of subwords within each of the clusters.

12. A speech recognition method comprising the steps of:

(a) receiving a nametag utterance via a microphone associated with an automatic speech recognition system;

(b) decoding the nametag utterance to recognize constituent subwords of the nametag utterance using at least one processor associated with the automatic speech recognition system;

(c) determining the number of subwords in the nametag utterance;

(d) associating the nametag utterance with one or more of a plurality of different nametag clusters based on the number of subwords in the nametag utterance;

(e) performing a confusability check on the nametag utterance within the cluster(s) associated with the nametag utterance;

(f) associating a nametag with the nametag utterance based on a best score from the step of performing a confusability check and storing the associated nametag in memory, as representing the nametag utterance, among a plurality of stored nametags;

(g) retrieving a stored nametag from memory by;

receiving a second nametag utterance via a microphone;

decoding the second nametag utterance to recognize constituent subwords of the second nametag utterance;

determining the number of subwords in the second nametag utterance;

associating the second nametag utterance with one or more of the plurality of different nametag clusters based on the number of subwords in the second nametag utterance; and

further decoding the second nametag utterance within the one or more of the plurality of different nametag clusters of the step of associating the second nametag utterance to correctly match the second nametag utterance to its corresponding stored nametag; and

(h) playing stored nametags back by cluster.

19. A speech recognition method for a telematics system that has a speech recognition system, the method comprising the steps of:

(a) receiving a nametag utterance via a microphone associated with an automatic speech recognition system;

(b) decoding the nametag utterance using a free subword recognizer to recognize constituent subwords of the nametag utterance using at least one processor associated with the automatic speech recognition system;

(c) determining the number of subwords in the nametag utterance;

(d) associating the nametag utterance with one or more of a plurality of different nametag clusters based on the number of subwords in the nametag utterance;

(e) performing a confusability check on the nametag utterance within the cluster(s) associated with the nametag utterance;

(f) associating the nametag utterance with one of a plurality of nametag templates based on a best score from the step of performing a confusability check and storing the one of the plurality of nametag templates in memory;

(g) retrieving stored nametags from memory by decoding the nametag utterance only within those cluster(s) associated with the nametag utterance; and

(h) playing the stored nametags back by cluster.

20. A speech recognition method for a vehicle having a telematics unit with an embedded speech recognition system, the method comprising the steps of:

(a) receiving a nametag utterance via a microphone associated with an automatic speech recognition system;

(b) decoding the nametag utterance using a free phoneme recognizer to recognize constituent phonemes of the nametag utterance using at least one processor associated with the automatic speech recognition system;

(c) determining the number of phonemes in the nametag utterance;

(d) associating the nametag utterance with one or more of a plurality of different nametag clusters based on the number of phonemes in the nametag utterance;

(e) performing a confusability check on the nametag utterance within the cluster(s) associated with the nametag utterance;

(f) associating the nametag utterance with one of a plurality of nametag templates based on a best score from the step of performing a confusability check and storing the one nametag template in memory;

(g) retrieving stored nametags from memory by decoding the nametag utterance only within those cluster(s) associated with the nametag utterance, comprising the substeps of;

receiving a second nametag utterance;

decoding the second nametag utterance to recognize constituent subwords of the second nametag utterance;

determining the number of subwords in the second nametag utterance;

associating the second nametag utterance with one or more of the plurality of different nametag clusters based on the number of subwords in the second nametag utterance; and

further decoding the second nametag utterance within the one or more of the plurality of different nametag clusters of the step of associating the second nametag utterance to correctly match the second nametag utterance to its corresponding stored nametag; and

(h) playing the stored nametags back by cluster.