Word competition models in voice recognition

• US 7,624,013 B2
• Filed: 09/10/2004
• Issued: 11/24/2009
• Est. Priority Date: 09/10/2004
• Status: Active Grant

• Current Assignee: Scientific Learning Corp. (Carnegie Learning, Inc.)
• Original Assignee: Scientific Learning Corp. (Carnegie Learning, Inc.)
• Inventors: Barker, Simon
• Primary Examiner(s): Abebe; Daniel D
• Application Number:
  US10/938,749
• Publication Number:
  US 20060058996A1
• Time in Patent Office: 1,901 Days
• Field of Search: 704/251, 704/257, 704/255
• US Class Current: 704/251
• CPC Class Codes: G09B 19/04 Speaking with audible presentation of the material to be studied G09B5/04
• CPC Class Codes: G09B 5/04 with audible presentation of the material to be studied sound recording or reproducing G11B
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

A system and method relate to voice recognition software, and more particularly to voice recognition tutoring software to assist in reading development.

Claims

1. A computer implemented method for recognizing speech patterns, the method comprising:

segmenting a word into a string of consecutive phonemes;

storing a plurality of sequences of the phonemes, at least one of the sequences of phonemes being associated with a mispronunciation of the word;

associating a correctness indication with at least some of the sequences of the phonemes;

providing a plurality of levels, at least some of the levels having multiple sequences of the phonemes associated with the level and at least some of the sequences of phonemes being associated with multiple ones of the plurality of levels with the correctness indication varying based on the level;

comparing, by a computer system, an utterance with the plurality of sequences of phonemes;

determining, by the computer system, if a match exists between the utterance and a particular one of the sequences of phonemes; and

determining, by the computer system, an accuracy of the utterance based on the determined match, the level, and the correctness indication.

2. The method of claim 1 wherein storing the plurality of sequences of the phonemes includes storing a sequence including all the phonemes for the word.

3. The method of claim 1 wherein storing the plurality of sequences of the phonemes includes storing a truncated sequence of phonemes for the word.

4. The method of claim 1 wherein the plurality of levels includes at least one of loose, medium, and strict.

5. The method of claim 1, wherein the sequence of phonemes associated with the mispronunciation of the word comprises a pronunciation that is not recognized as an utterance corresponding to the word.

6. The method of claim 1, wherein the at least one of the sequences associated with the mispronunciation of the word comprises a competition model for the word.

7. The method of claim 1, further comprising:

associating a probabilistic weighting with each of the plurality of sequences of the phonemes.

8. The method of claim 7, wherein:

at least one of the sequences of phonemes is associated with a correct pronunciation of the word, and

associating the probabilistic weighting with each of the plurality of sequences of the phonemes comprises associating a greater probabilistic weighting to at least one of the sequences of phonemes associated with a correct pronunciation of the word than to at least one of the sequences of phonemes associated with the mispronunciation of the word.

9. The method of claim 1, wherein storing a plurality of sequences of the phonemes comprises storing at least one sequence of phonemes omitting at least one phoneme that is preceding a last one of the phonemes in the string and succeeding a first one of the phonemes in the string.

10. A computer implemented method comprising:

segmenting, by a computer system, a word into a string of consecutive phonemes;

generating, by the computer system, sequences of phonemes having at least one omitted phoneme; and

associating a first set of multiple ones of the sequences of the phonemes with a first pronunciation level for the word, the first set of the sequences of the phonemes being associated with mispronunciations of the word for the first pronunciation level;

associating a second set of multiple ones of the sequences of the phonemes with a second pronunciation level for the word, the second set of the sequences of the phonemes being associated with mispronunciations of the word for the second pronunciation level and being different from the first set of sequences of phonemes associated with the first level.

11. The method of claim 10 further comprising:

associating a third set of multiple ones of the sequences of the phonemes with a third pronunciation level for the word, the third set of the sequences of the phonemes being associated with mispronunciations of the word for the third pronunciation level and being different from the first and second sets of sequences of phonemes associated with the first level and the second level;

wherein the first pronunciation level comprises a loose pronunciation level,

the second pronunciation level comprises a medium pronunciation level, and

the third pronunciation level comprises a strict pronunciation level.

12. The method of claim 10 wherein the first and second pronunciation levels are based on pronunciation of the words.

13. The method of claim 10 wherein the sequences of phonemes includes a truncated sequence of phonemes for the word.

14. The method of claim 10 wherein the words include at least 3 phonemes.

15. The method of claim 10, wherein the sequence of phonemes associated with the mispronunciation comprises a pronunciation not recognized as an utterance corresponding to recognition of the word.

16. The method of claim 10, wherein the at least one of the sequences associated with the mispronunciation of the word comprises a competition model for the word.

17. The method of claim 16, further comprising storing a complete sequence of the phonemes, the complete sequence being associated with a correct pronunciation of the word.

18. The method of claim 10, further comprising:

associating a probabilistic weighting with each of the plurality of sequences of the phonemes.

19. The method of claim 18, wherein:

at least one of the sequences of phonemes is associated with a correct pronunciation of the word, and

associating the probabilistic weighting with each of the plurality of sequences of the phonemes comprises associating a greater probabilistic weighting to at least one of the sequences of phonemes associated with a correct pronunciation of the word than to at least one of the sequences of phonemes associated with the mispronunciation of the word.

20. A speech-recognizing device configured to:

segment a word into a string of consecutive phonemes;

store a plurality of sequences of the phonemes, at least one of the sequences being associated with a mispronunciation of the word and omitting at least one phoneme that is preceding a last one of the phonemes in the string and succeeding a first one of the phonemes in the string;

associate a correctness indication with at least some of the sequences of the phonemes;

provide a plurality of levels wherein the correctness indication varies based on the level, at least some of the levels having multiple sequences of the phonemes associated with the level and at least some of the sequences of phonemes being associated with multiple ones of the plurality of levels with the correctness indication varying based on the level;

compare an utterance with the plurality of sequences of phonemes;

determine if a match exists between the utterance and a particular one of the sequences of phonemes; and

determine an accuracy of the utterance based on the determined match, the level, and the correctness indication.

21. The speech-recognizing device of claim 20 further configured to store a sequence of all the phonemes for the word.

22. The speech-recognizing device of claim 20 further configured to store a truncated sequence of phonemes for the word.

23. A device configured to:

segment a word into a string of consecutive phonemes;

generate sequences of phonemes having at least one omitted phoneme; and

associate a first set of multiple ones of the sequences of the phonemes with a first pronunciation level for the word, the first set of the sequences of the phonemes being associated with mispronunciations of the word for the first pronunciation level;

associate a second set of multiple ones of the sequences of the phonemes with a second pronunciation level for the word, the second set of the sequences of the phonemes being associated with mispronunciations of the word for the second pronunciation level and being different from the first set of sequences of phonemes associated with the first level.

24. The device of claim 23 wherein the first and second pronunciation levels are based on pronunciation of the words.

25. The device of claim 23 wherein the sequences of phonemes includes a complete sequence of phonemes for the word.

26. The device of claim 23 wherein the sequences of phonemes includes truncated sequence of phonemes for the word.

27. A computer program product, tangibly embodied in an information carrier, for executing instructions on a processor, the computer program product being operable to cause a machine to:

segment a word into a string of consecutive phonemes;

store a plurality of sequences of the phonemes, at least one of the sequences being associated with a mispronunciation of the word;

associate a correctness indication with at least some of the sequences of the phonemes;

provide a plurality of levels, at least some of the levels having multiple sequences of the phonemes associated with the level and at least some of the sequences of phonemes being associated with multiple ones of the plurality of levels with the correctness indication varying based on the level;

compare an utterance with the plurality of sequences of phonemes;

determine if a match exists between the utterance and a particular one of the sequences of phonemes; and

determine an accuracy of the utterance based on the determined match, the level, and the correctness indication.

28. The computer program product of claim 27 further comprising instructions to cause a machine to store a sequence of all the phonemes for the word.

29. The computer program product of claim 27 further comprising instructions to cause a machine to store a truncated sequence of phonemes for the word.

30. The computer program product of claim 27 further comprising instructions to cause a machine to store a sequence of phonemes omitting at least one phoneme that is preceding a last one of the phonemes in the string and succeeding a first one of the phonemes in the string.

31. A computer program product, tangibly embodied in an information carrier, for executing instructions on a processor, the computer program product being operable to cause a machine to:

segment a word into a string of consecutive phonemes;

generate sequences of phonemes having at least one omitted phoneme; and

associate a first set of multiple ones of the sequences of the phonemes with a first pronunciation level for the word, the first set of the sequences of the phonemes being associated with mispronunciations of the word for the first pronunciation level;

associate a second set of multiple ones of the sequences of the phonemes with a second pronunciation level for the word, the second set of the sequences of the phonemes being associated with mispronunciations of the word for the second pronunciation level and being different from the first set of sequences of phonemes associated with the first level.

32. The computer program product of claim 31 wherein the first and second levels are based on pronunciation of the words.

33. The computer program product of claim 31 wherein the sequences of phonemes includes truncated sequence of phonemes for the word.