Recognition system for determining whether speech is confusing or inconsistent

• US 5,987,411 A
• Filed: 12/17/1997
• Issued: 11/16/1999
• Est. Priority Date: 12/17/1997
• Status: Expired due to Term

• Current Assignee: Avaya Incorporated
• Original Assignee: Northern Telecom Limited (Nortel Networks Corporation)
• Inventors: Ma, Hung S., Petroni, Marco
• Primary Examiner(s): Wieland, Susan
• Application Number:
  US08/992,260
• Time in Patent Office: 699 Days
• Field of Search: 704/255, 704/251, 704/239, 704/240, 704/241, 704/238
• US Class Current: 704/255
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 2015/0638 Interactive procedures

Abstract

Methods and systems consistent with the present invention enroll a candidate phrase uttered by a user in a dictionary having at least one previously enrolled phrase. The system receives utterances of the candidate phrase and determines whether the first utterance is confusingly similar to a previously enrolled phrase and whether they are consistent with each other. The system then enrolls the candidate phrase in the dictionary according to these determinations.

Claims

1. A system for enrolling candidate phrases uttered by a user in a dictionary having at least one previously enrolled phrase stored therein, the system comprising:

means for receiving a first utterance of the candidate phrase uttered by the user;

first means for determining whether the first utterance is confusingly similar to a previously enrolled phrase;

means for receiving a second utterance of the candidate phrase uttered by the user;

second means for determining whether the second utterance is confusingly similar to a previously enrolled phrase;

third means for determining whether the first utterance is inconsistent with the second utterance; and

means for enrolling the candidate phrase in the dictionary according to the determinations made by the first, second and third means for determining.

2. The system of claim 1, wherein the first and second means for determining further include:

means for forming a recognition value representing a degree to which an utterance of the candidate phrase matches a previously enrolled phrase; and

means for comparing the recognition value to a threshold value to determine whether the utterance is confusingly similar a previously enrolled phrase.

3. The system of claim 2, wherein the means for comparing further includes:

means for comparing the recognition value to the threshold value having a value between 0.85 and 0.97.

4. The system of claim 2, wherein the first and second means for determining further include:

means for concluding that the utterance is confusing when the recognition value is above the threshold value.

5. The system of claim 1, wherein the third means for determining further includes:

means for converting each utterance into a feature vector;

means for forming a distance value between the feature vectors representing each utterance; and

means for comparing the distance value to a threshold value.

6. The system of claim 5, wherein the means for comparing further includes:

means for comparing the distance value to the threshold value having a value equal to 17.

7. The system of claim 5, wherein the third means for determining further includes:

means for concluding that the utterance is inconsistent when the distance value exceeds the threshold value.

8. The system of claim 5, wherein the means for forming the distance value further includes:

means for finding a dynamic time warping distance between the feature vectors representing each utterance.

9. The system of claim 1, further including:

means for receiving a third utterance of the candidate phrase said by the user according to determinations made by the first, second and third means for determining.

10. The system of claim 9, further including:

fourth means for determining whether the third utterance is confusingly similar to a previously enrolled phrase.

11. The system of claim 10, further including:

fifth means for determining whether the third utterance is inconsistent with the first utterance and the second utterance.

12. The system of claim 11, wherein the means for enrolling the candidate phrase further includes:

means for enrolling the candidate phrase in the dictionary according to the determinations made by the fourth and fifth means for determining.

13. The system of claim 1, further including:

means for indicating to the user why the candidate phrase was not enrolled in the dictionary when the means for enrolling the candidate phrase determines that the candidate phrase should not be enrolled in the dictionary.

14. The system of claim 1, further including:

means for indicating to the user why the candidate phrase was enrolled in the dictionary when the means for enrolling the candidate phrase determines that the candidate phrase should be enrolled in the dictionary.

15. A method for enrolling candidate phrases uttered by a user in a dictionary having at least one previously enrolled phrase stored therein, the method comprising:

receiving a first utterance of the candidate phrase uttered by the user;

determining whether the first utterance is confusingly similar to a previously enrolled phrase;

receiving a second utterance of the candidate phrase uttered by the user;

determining whether the second utterance is confusingly similar to a previously enrolled phrase;

determining whether the first utterance is inconsistent with the second utterance; and

enrolling the candidate phrase in the dictionary according to the determination of whether the first and second utterances are confusing and whether the first and second utterances are inconsistent.

16. The method of claim 15, wherein the step for determining whether the utterances are confusing further includes the substeps of:

forming a recognition value representing a degree to which an utterance of the candidate phrase matches a previously enrolled phrase; and

comparing the recognition value to a threshold value to determine whether the utterance is confusingly similar to a previously enrolled phrase.

17. The method of claim 16, wherein the substep of comparing further includes the substep of:

comparing the recognition value to the threshold value having a value between 0.85 and 0.97.

18. The method of claim 16, wherein the step of determining whether the utterances are confusing further includes the substep of:

concluding that the utterance is confusing when the recognition value exceeds the threshold value.

19. The method of claim 15, wherein the step of determining whether the utterances are inconsistent further includes the substeps of:

converting each utterance into a feature vector;

forming a distance value between the feature vectors representing each utterance; and

comparing the distance value to a threshold value to determine whether the utterances are inconsistent.

20. The method of claim 19, wherein the substep of comparing the distance value further includes the substep of:

comparing the distance value to the threshold value having a value equal to 17.

21. The method of claim 19, wherein the step of determining whether the utterances are inconsistent further includes the substep of:

concluding that the utterance is inconsistent when the distance value exceeds the threshold value.

22. The method of claim 19, wherein the step of determining the distance value further includes the substep of:

finding the distance value by determining a dynamic time warping distance between the feature vectors representing each utterance.

23. The method of claim 15, further including the step of:

receiving a third utterance of the candidate phrase said by the user according to the determination of whether the first and the second utterances are confusing and whether the first and second utterances are inconsistent.

24. The method of claim 23, further including the step of:

determining whether the third utterance is confusingly similar to a previously enrolled phrase.

25. The method of claim 24, further including the step of:

determining whether the third utterance is inconsistent with the first utterance and the second utterance.

26. The method of claim 25, wherein the step of enrolling the candidate phrase further includes the substep of:

enrolling the candidate phrase in the dictionary according to the determination of whether the third utterance is confusing and whether the third utterance is inconsistent with the first and second utterances.

27. The method of claim 15, further including the step of:

indicating to the user why the candidate phrase was not enrolled in the dictionary when the candidate phrase is not enrolled in the dictionary.

28. The method of claim 15, further including the step of:

indicating to the user why the candidate phrase was enrolled in the dictionary when the candidate phrase is enrolled in the dictionary.