DISTINGUISHING OUT-OF-VOCABULARY SPEECH FROM IN-VOCABULARY SPEECH

• US 20070265849A1
• Filed: 05/11/2006
• Published: 11/15/2007
• Est. Priority Date: 05/11/2006
• Status: Active Grant

• Current Assignee: General Motors LLC (General Motors Company)
• Original Assignee: General Motors Corporation (Motors Liquidation Co. GUC Trust)
• Inventors: Chengalvarayan, Rathinavelu, Grost, Timothy
• Granted Patent: US 8,688,451 B2
• Time in Patent Office: Days
• US Class Current: 704/257
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems

Abstract

A speech recognition method includes receiving input speech from a user, processing the input speech using a first grammar to obtain parameter values of a first N-best list of vocabulary, comparing a parameter value of a top result of the first N-best list to a threshold value, and if the compared parameter value is below the threshold value, then additionally processing the input speech using a second grammar to obtain parameter values of a second N-best list of vocabulary. Other preferred steps include: determining the input speech to be in-vocabulary if any of the results of the first N-best list is also present within the second N-best list, but out-of-vocabulary if none of the results of the first N-best list is within the second N-best list; and providing audible feedback to the user if the input speech is determined to be out-of-vocabulary.

Claims

1. A speech recognition method comprising the steps of:

(a) receiving input speech from a user;

(b) processing the input speech using a first grammar to obtain parameter values of a first N-best list of vocabulary;

(c) comparing at least one parameter value of a top result of the first N-best list to at least one predetermined threshold value; and

(d) subsequently processing the input speech using a second grammar to obtain parameter values of a second N-best list of vocabulary, if the compared at least one parameter value is below the at least one predetermined threshold value.

2. The method of claim 1, wherein the at least one parameter value is at least one of a confidence score, segment duration, likelihood score, or signal-to-noise ratio.

3. The method of claim 1, wherein step (d) is disabled if frequency of out-of-vocabulary utterances falls below a predetermined threshold.

4. The method of claim 3, wherein the top result from step (c) is determined to be the vocabulary contained in the user'"'"'s speech.

5. The method of claim 1, wherein step (d) is enabled depending on a characteristic of the user.

6. The method of claim 1, wherein step (d) is enabled for predetermined types of grammar.

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

(e) determining the input speech to be in-vocabulary, if any of the results of the first N-best list is also present within the second N-best list, but out-of-vocabulary if any of the results of the first N-best list is not within the second N-best list.

8. The method of claim 7, further comprising the step of creating the second grammar by generating a limited set of vocabulary that includes in-vocabulary within the first grammar.

9. The method of claim 7, further comprising the step of creating the second grammar by combining two or more other grammars used by a speech recognition system used to implement the method.

10. The method of claim 7, further comprising the step of creating the second grammar by generating a limited set of vocabulary that does not include in-vocabulary of the first grammar.

11. The method of claim 7, further comprising the step of creating the second grammar by generating a limited set of vocabulary that does not include triphones or pentaphones of words of the in-vocabulary first grammar.

12. The method of claim 8, wherein step (e) includes determining the input speech to be in-vocabulary, if one of the top results of the first N-best list is also present within the second N-best list, but out-of-vocabulary if none of the top results of the first N-best list is within the second N-best list.

13. The method of claim 12, wherein step (e) includes determining the input speech to be in-vocabulary, if the top result of the first N-best list is also present within the second N-best list, but out-of-vocabulary if the top result of the first N-best list is not within the second N-best list.

14. The method of claim 12, wherein step (e) includes determining the input speech to be in-vocabulary, if the top result of the first N-best list is also present within the top results of the second N-best list, but out-of-vocabulary if the top result of the first N-best list is not within the top results of the second N-best list.

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

(f) providing audible feedback to the user if the input speech is determined to be out-of-vocabulary.

16. The method of claim 15, wherein the step (f) includes instructing the user to use vocabulary within the first grammar.

17. The method of claim 16, wherein the step (f) further includes providing a list of acceptable vocabulary within the first grammar.

18. The method of claim 1, further comprising determining whether latency between steps (b) and (d) falls outside a defined period and, if so, skipping step (d) and redefining the first grammar to also include the second grammar.

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

(a) receiving input speech from a user to be processed into recognized speech;

(b) processing the input speech using a first grammar to obtain parameter values of a first N-best list of probable vocabulary suggested by a decoder;

(c) comparing at least one parameter value of a top result of the first N-best list to at least one predetermined threshold value;

(d) subsequently processing the input speech using a second grammar to obtain parameter values of a second N-best list of probable vocabulary suggested by the decoder, if the compared at least one parameter value is below the at least one predetermined threshold value; and

(e) determining the input speech to be in-vocabulary, if any of the results of the first N-best list is also present within the second N-best list, but out-of-vocabulary if none of the results of the first N-best list is within the second N-best list.

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 input speech from a user to be processed into recognized speech;

(b) processing the input speech using a first grammar to obtain parameter values of a first N-best list of probable vocabulary suggested by a decoder;

(c) comparing at least one parameter value of a top result of the first N-best list to at least one predetermined threshold value, wherein the at least one parameter value is at least one of a confidence score, segment duration, likelihood score, or signal-to-noise ratio;

(d) subsequently processing the input speech using a second grammar to obtain parameter values of a second N-best list of probable vocabulary suggested by the decoder, if the compared at least one parameter value is below the at least one predetermined threshold value;

(e) determining the input speech to be in-vocabulary, if the top result of the first N-best list is also present within the second N-best list, but out-of-vocabulary if the top result of the first N-best list is not within the second N-best list; and

(f) providing audible feedback to the user if the input speech is determined to be out-of-vocabulary.