Dynamic barge-in in a speech-responsive system

• US 7,162,421 B1
• Filed: 05/06/2002
• Issued: 01/09/2007
• Est. Priority Date: 05/06/2002
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Wu, Su-Lin, Strope, Brian, Shahshahani, Ben, Zeppenfeld, Torsten
• Primary Examiner(s): Azad; Abul K.
• Application Number:
  US10/140,526
• Time in Patent Office: 1,709 Days
• Field of Search: None
• US Class Current: 704/233
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

A method and system for barge-in acknowledgement are disclosed. A prompt is attenuated upon detection of speech. The speech is accepted and the prompt is terminated if the speech corresponds to an allowable response.

Claims

1. A method comprising:

playing a prompt at a first level to a user of a speech recognition system;

attenuating the prompt without terminating the prompt, in response to detecting a speech signal during playing of the prompt;

extracting an utterance from the speech signal;

determining whether the utterance corresponds to an allowable response by verifying the user by comparing the utterance to a plurality of voice samples, determining a level of acoustic confidence, and determining whether a meaning of the utterance correlates to the prompt;

if the utterance corresponds to the allowable response, then accepting the utterance and terminating the prompt; and

if the utterance does not correspond to the allowable response, then rejecting the utterance and intensifying the prompt.

2. The method of claim 1 wherein the determining the level of acoustic confidence comprises matching the utterance to acoustic samples.

3. The method of claim 1 wherein the determining whether the meaning of the utterance correlates to the prompt comprises matching the utterance to natural language samples.

4. The method of claim 1, wherein:

said attenuating the prompt comprises attenuating the prompt to a second level in response to detecting first speech during playing of the prompt; and

the method further comprises;

attenuating the prompt to a third level in response to detecting second speech during playing of the prompt, wherein the second level is higher than the first level.

5. The method of claim 4, further comprising:

determining whether a previous utterance was rejected;

wherein said attenuating the prompt to a third level is in response to determining that the previous utterance was rejected.

6. A method comprising:

playing a prompt at a default level to a user of a speech recognition system;

attenuating the prompt without terminating the prompt, in response to detecting a speech signal occurring during playing of the prompt;

extracting an utterance from the speech signal;

determining whether the utterance corresponds to an allowable response by performing a single operation from a set of operations;

if the utterance corresponds to the allowable response, then accepting the utterance and terminating the prompt; and

if the utterance does not correspond to the allowable response, then rejecting the utterance and intensifying the prompt.

7. The method of claim 6 wherein the set of operations comprises verifying the user by comparing the utterance to a plurality of voice samples operation, determining a level of acoustic confidence operation and determining whether a meaning of the utterance correlates to the prompt operation.

8. The method of claim 7 wherein the determining the level of acoustic confidence operation comprises matching the utterance to acoustic samples.

9. The method of claim 7 wherein the determining whether the meaning of the utterance correlates to the prompt operation comprises matching the utterance to natural language samples.

10. The method of claim 9 wherein the matching the utterance to the natural language samples comprises determining a request corresponding to matching natural language samples.

11. A method comprising:

attenuating a prompt to a first level without terminating the prompt, in response to detecting a first speech signal during playing of the prompt at a default level;

intensifying the prompt in response to determining that an utterance extracted from the first speech signal does not correspond to an allowed response; and

attenuating the prompt to a second level in response to detecting a second speech signal during playing of the prompt at the default level, wherein the second level is higher than the first level.

12. The method of claim 11 wherein the utterance does not correspond to the allowed response if a sample of the utterance does not match a voice sample from a plurality of voice samples.

13. The method of claim 11 wherein the utterance does not correspond to the allowed responses if the utterance does not match acoustic models.

14. The method of claim 11 wherein the utterance does not correspond to the allowed response if a meaning of the utterance does not correlate to a meaning of the prompt.

15. The method of claim 14 wherein the meaning of the utterance does not correlate to the meaning of the prompt if the utterance does not match natural language models.

16. The method of claim 11, wherein said attenuating the prompt to a second level is in response to determining that the previous utterance was rejected.

17. A speech recognition system comprising:

means for attenuating a prompt to a first level without terminating the prompt, in response to detecting a first speech signal during playing of the prompt at a default level;

means for intensifying the prompt to the default level in response to determining that an utterance extracted from the first speech signal does not correspond to an allowed response; and

means for attenuating the prompt to a second level in response to detecting a second speech signal during playing of the prompt at the default level, the second level is higher than the first level.

18. The speech recognition system of claim 17 wherein the utterance does not correspond to the allowed response if a sample of the utterance does not match a voice sample from a plurality of voice samples.

19. The speech recognition system of claim 17 wherein the utterance does not correspond to the allowed response if the utterance does not match acoustic models.

20. The speech recognition system of claim 17 wherein the utterance does not correspond to the allowed response if a meaning of the utterance does not correlate to a meaning of the prompt.