Enrollment in speech recognition

• US 6,212,498 B1
• Filed: 03/28/1997
• Issued: 04/03/2001
• Est. Priority Date: 03/28/1997
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: Dragon Systems, Inc. (Microsoft Corporation)
• Inventors: Gould, Joel, Wald, David E., Graham, Noah M., Scattone, Francesco, Albina, Toffee A., Sherwood, Stefan
• Primary Examiner(s): Knepper, David D.
• Application Number:
  US08/825,536
• Time in Patent Office: 1,467 Days
• Field of Search: 704/200, 704/201, 704/231, 704/243-245, 704/235, 704/260
• US Class Current: 704/244
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 2015/0638 Interactive procedures

Abstract

Enrolling a user in a speech recognition system includes analyzing acoustic content of a user utterance and determining whether the utterance matches a portion of an enrollment text. If the user utterance matches a portion of the enrollment text, the acoustic content is used to update acoustic models corresponding to the portion of the enrollment text.

Claims

1. A computer-implemented method for enrolling a user in a continuous-speech recognition system, comprising:

analyzing acoustic content of a user utterance including multiple words;

determining, based on the analysis, whether the multiple words of the user utterance substantially match a multiple-word portion of an enrollment text; and

if the multiple words of the user utterance substantially match a multiple-word portion of the enrollment text, using the acoustic content of the user utterance to update acoustic models corresponding to the multiple-word portion of the enrollment text.

2. The method of claim 1, wherein determining whether the multiple words of the user utterance substantially match a multiple-word portion of the enrollment text comprises using an enrollment grammar corresponding to the enrollment text.

3. The method of claim 2, wherein determining whether the multiple words of the user utterance substantially match a multiple-word portion of the enrollment text further comprises using a rejection grammar.

4. The method of claim 3, wherein the rejection grammar is a phoneme grammar.

5. The method of claim 4, wherein the rejection grammar models an utterance using a set of phonemes that is smaller than a set of phonemes used by the enrollment grammar.

6. The method of claim 1, further comprising determining an enrollment position within the enrollment text, wherein determining whether the multiple words of the user utterance substantially match a multiple-word portion of the enrollment text comprises determining whether the multiple words of the user utterance match a multiple-word portion of the enrollment text beginning at the enrollment position.

7. The method of claim 6, further comprising displaying the enrollment text and the enrollment position.

8. The method of claim 6, further comprising advancing the enrollment position in the enrollment text past the multiple-word portion of the enrollment text.

9. The method of claim 1, further comprising:

determining an enrollment position within the enrollment text;

determining, using an enrollment grammar, a portion of the enrollment text beginning at the enrollment position that best matches the user utterance;

determining, using the enrollment grammar, an enrollment score for the best matching portion of the enrollment text;

determining, using a rejection grammar, a recognition candidate that best matches the user utterance; and

determining, using a rejection grammar, a rejection score for the best matching recognition candidate,

wherein determining whether the user utterance matches a portion of the enrollment text is based on the enrollment score and the rejection score.

10. The method of claim 1, further comprising selecting the enrollment text from a plurality of enrollment texts.

11. The method of claim 10, wherein each of the plurality of enrollment texts has a corresponding enrollment grammar, and determining whether the user utterance matches a portion of an enrollment text further comprises using the enrollment grammar corresponding to the selected enrollment text.

12. The method of claim 1, further comprising receiving an enrollment text for use as the enrollment text.

13. The method of claim 12, further comprising generating an enrollment grammar corresponding to the received enrollment text, wherein determining whether the multiple words of the user utterance substantially match a multiple-word portion of the enrollment text uses the generated enrollment grammar.

14. The method of claim 1, further comprising ignoring the user utterance if the multiple words of the user utterance do not substantially match a multiple-word portion of the enrollment text.

15. A computer-implemented method for enrolling a user into a continuous-speech recognition system, comprising:

displaying an enrollment text;

displaying an enrollment position within the enrollment text;

receiving a user utterance including multiple words;

determining whether a match exists between the multiple words of the user utterance and a multiple-word portion of the enrollment text beginning at the enrollment position;

updating the enrollment position if a match exists; and

displaying the updated enrollment position.

16. The method of claim 15, wherein displaying the enrollment position and the updated enrollment position further comprises displaying a marker at the enrollment position.

17. The method of claim 15, wherein displaying the enrollment position and the updated enrollment position further comprises highlighting the enrollment text at the enrollment position.

18. The method of claim 15, wherein displaying the enrollment position and the updated enrollment position further comprises displaying a cursor in the enrollment text at the enrollment position.

19. Computer software, residing on a computer-readable storage medium, comprising instructions for causing a computer to:

analyze acoustic content of a user utterance including multiple words;

determine, based on the analysis, whether the multiple words of the user utterance substantially match a multiple-word portion of an enrollment text; and

use the acoustic content of the user utterance to update acoustic models corresponding to the multiple-word portion of the enrollment text if the multiple words of the user utterance substantially match a multiple-word portion of the enrollment text.

20. A speech recognition system for enrolling a user, comprising:

a display for displaying an enrollment text to a user;

an input device for receiving speech signals;

a processor for;

determining a user utterance including multiple words from a received speech signal;

analyzing acoustic content of the user utterance;

determining, based on the acoustic analysis, whether the multiple words of the user utterance substantially match a multiple-word portion of an enrollment text; and

using the user utterance to update acoustic models corresponding to the multiple-word portion of the enrollment text if multiple words of the user utterance substantially match a multiple-words portion of the enrollment text.