Method and apparatus for recognizing speech information based on prediction

• US 6,598,017 B1
• Filed: 07/20/1999
• Issued: 07/22/2003
• Est. Priority Date: 07/27/1998
• Status: Expired due to Term

• Current Assignee: Canon Ayutthaya Limited (Canon Inc.)
• Original Assignee: Canon Ayutthaya Limited (Canon Inc.)
• Inventors: Yamamoto, Naoko, Wakai, Masanori, Suda, Aruna Rohra
• Primary Examiner(s): ABEBE, DANIEL DEMELASH
• Application Number:
  US09/356,691
• Time in Patent Office: 1,463 Days
• Field of Search: 704/235, 704/240, 704/255, 704/241, 704/243, 704/251, 704/256, 704/257, 704/259
• US Class Current: 704/251
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning

Abstract

An apparatus for recognizing sound information includes a sound recognition unit for recognizing sound information. A knowledge base stores knowledge concerning a type of data represented by the sound information. A prediction unit predicts the type of data represented by sound information to be subsequently recognized by checking the knowledge stored in the knowledge base. The sound recognition unit recognizes sound information to be subsequently recognized based on a prediction result obtained by the prediction unit. With this arrangement, the recognition accuracy is enhanced.

Claims

1. An information processing apparatus comprising:

storage means for storing prediction information predicting a next phoneme, syllable, or word to be input based upon the meaning of one or more previously input words;

first determining means for determining whether inputted sound information is produced by a human;

means for acquiring information concerning specific language or non-language human sounds;

second determining means for determining whether the inputted sound information is the specific language or non-language human sounds;

third determining means for determining that the inputted sound information is language information in the event that said first determining means determines that the inputted sound information is produced by a human and said second determining means determines that the inputted sound information is not the specific language or non-language human sounds;

recognition means for recognizing the inputted sound information as a specific word, syllable, or phoneme, based on the prediction information predicting the current input word, syllable, or phoneme as a next word, syllable, or phoneme to be input based upon the meaning of one or more previously input words in the event that said third determining means determines that the inputted sound information is language information;

a knowledge base for storing knowledge concerning a type of data represented by the sound information and specific language or non-language human sounds;

prediction means for predicting a next word, syllable or phoneme to be input based upon the meaning of one or more previously input words by referring to the knowledge stored in said knowledge base; and

updating means for updating the prediction information stored in said storage means based on a prediction result obtained by said prediction means.

2. An information processing apparatus according to claim 1, wherein said prediction means predicts the information to be subsequently recognized by said recognition means based on a previous recognition result recognized by said recognition means.

3. An information processing apparatus according to claim 1, wherein said prediction means predicts a type of data represented by information to be recognized by said recognition means.

4. An information processing apparatus according to claim 1, further comprising analysis means for analyzing information obtained as a recognition result by said recognition means, wherein said prediction means performs a predicting operation based on an analysis result obtained by said analysis means.

5. An information processing apparatus according to claim 1, wherein said recognition means outputs a plurality of items of information as a recognition result.

6. An information processing apparatus according to claim 5, wherein said prediction means predicts the information to be subsequently recognized in response to each of the plurality of items of information.

7. An information processing apparatus according to claim 1, wherein said prediction means predicts items of information which are likely to be subsequently input.

8. An information processing apparatus according to claim 7, wherein said recognition means selects a recognition result from among the items of information predicted by said prediction means.

9. An information processing apparatus according to claim 1, wherein said prediction means predicts an item of information which is less likely to be subsequently input.

10. An information processing apparatus according to claim 9, wherein said recognition means rejects the item of information which is less likely to be subsequently recognized.

11. An information processing apparatus according to claim 1, further comprising determination means for determining the correctness of information obtained as a recognition result by said recognition means.

12. An information processing apparatus according to claim 11, wherein said prediction means performs a predicting operation based on a determination result obtained by said determination means.

13. An information processing apparatus according to claim 11, further comprising correction means for correcting the information obtained as a recognition result by said recognition means based on a determination result obtained by said determination means.

14. An information processing apparatus according to claim 13, wherein said correction means corrects information obtained as a new recognition result by checking the knowledge stored in said knowledge base based on a previous recognition result.

15. An information processing apparatus according to claim 13, wherein said prediction means performs a predicting operation based on the recognition result obtained by said recognition means and performs a re-predicting operation based on a correction result obtained by correcting the recognition result by said correction means.

16. An information processing apparatus according to claim 1, wherein said first determining means distinguishes human speech from a mechanical sound based on a frequency difference.

17. An information processing apparatus according to claim 1, further comprising correction means for correcting information obtained as a recognition result by said recognition means, wherein said prediction means performs a predicting operation based on a correction result obtained by said correction means.

18. An information processing apparatus according to claim 1, wherein said recognition means recognizes speech information of a natural language in units of one of words, syllables, and phonemes.

19. An information processing apparatus according to claim 18, wherein said prediction means predicts a word including one of a syllable and a phoneme to be subsequently recognized when said recognition means uses one of the syllables and the phonemes, respectively, as a unit of recognition, and said prediction means predicts one of the syllable or the phoneme to be subsequently recognized based on the predicted word.

20. An information processing apparatus according to claim 18, further comprising selection means for selecting the unit for recognition.

21. An information processing apparatus according to claim 20, wherein said selection means selects the unit for recognition according to whether a previous recognition result was successfully obtained.

22. An information processing apparatus according to claim 1, wherein said recognition means recognizes the information by performing a matching operation by checking the knowledge stored in said knowledge base.

23. An information processing apparatus according to claim 22, wherein the matching operation comprises a dynamic programming matching.

24. An information processing apparatus according to claim 1, wherein said prediction means predicts the sound information to be subsequently recognized based on a previous operation.

25. An information processing apparatus according to claim 24, wherein said prediction means predicts a subsequent operation based on the previous operation and predicts the sound information to be subsequently recognized based on the predicted subsequent operation.

26. An information processing apparatus according to claim 25, herein said prediction means predicts a specific operation as the subsequent operation when it is unable to predict the subsequent operation based on the previous operation.

27. An information processing apparatus according to claim 24, wherein the sound information to be subsequently recognized comprises sound information to be initially recognized.

28. An information processing apparatus according to claim 1, wherein said knowledge base comprises language knowledge.

29. An information processing apparatus according to claim 28, wherein the language knowledge comprises a descriptive rule of natural-language information, and said prediction means predicts the sound information to be subsequently recognized based on the rule.

30. An information processing apparatus according to claim 1, wherein said knowledge base comprises domain knowledge.

31. An information processing apparatus according to claim 30, wherein said prediction means predicts, based on the domain knowledge, information belonging to a specific domain as the sound information to be subsequently recognized.

32. An information processing apparatus according to claim 1, wherein said knowledge base comprises general knowledge.

33. An information processing apparatus according to claim 1, wherein said prediction means predicts the sound information to be initially recognized.

34. An information processing apparatus according to claim 1, further comprising processing means for executing processing corresponding to a processing command when the processing command is recognized as a recognition result by said recognition means.

35. An information processing apparatus according to claim 34, further comprising reporting means for reporting a processing result obtained by said processing means.

36. An information processing apparatus according to claim 35, wherein said reporting means reports the processing result by sound.

74. An information processing apparatus according to claim 1, wherein said third determining means determines that the inputted sound information is to be processed as non-language information in the event that said second determining means determines that the inputted sound information is the specific language or non-language human sounds, and said recognition means recognizes the inputted sound information as non-language information in the event that said third determining means determines that the inputted sound information is to be processed as non-language information.

76. An information processing apparatus according to claim 1, wherein said third determining means determines that the inputted sound information is to be processed as non-language information in the event that said first determining means determines that the inputted sound information is not produced by the human, and said recognition means recognizes the inputted sound information as non-language information in the event that said third determining means determines that the inputted sound information is to be processed as non-language information.

37. An information processing method comprising:

a first determining step for determining whether inputted sound information is produced by a human;

an acquiring step for acquiring information concerning specific language or non-language human sounds;

a second determining step for determining whether the inputted sound information is the specific language or non-language human sounds;

third determining step for determining that the inputted sound information is language information in the event that said first determining step determines that the inputted sound information is produced by a human and said second determining step determines that the inputted sound information is not the specific language or non-language human sounds;

a recognition step of recognizing the inputted sound information to be a specific word, syllable, or phoneme based on prediction information predicting the current input word, syllable, or phoneme as a next word, syllable, or phoneme to be input based upon the meaning of one or more previously input words in the event that inputted sound information is determined by said third determining step to be language information;

a prediction step of predicting a next word, syllable, or phoneme to be input based upon the meaning of one or more previously input words by checking knowledge stored in a knowledge base for storing knowledge concerning a type of data represented by sound information and for storing the specific language or non-language human sounds; and

an updating step of updating the prediction information based on a prediction result obtained in said prediction step.

38. An information processing method according to claim 37, wherein said prediction step predicts the information to be subsequently recognized in said recognition step based on a previous recognition result obtained in said recognition step.

39. An information processing method according to claim 37, wherein said prediction step predicts a type of data represented by information to be subsequently recognized in said recognition step.

40. An information processing method according to claim 37, further comprising an analysis step of analyzing information obtained as a recognition result in said recognition step, wherein said prediction step performs a predicting operation based on an analysis result obtained in said analysis step.

41. An information processing method according to claim 37, wherein said recognition step outputs a plurality of items of information as a recognition result.

42. An information processing method according to claim 41, wherein said prediction step predicts the information to be subsequently recognized in response to each of the items of information.

43. An information processing method according to claim 37, wherein said prediction step predicts items of information which are likely to be subsequently recognized.

44. An information processing method according to claim 43, wherein said recognition step selects a recognition result from among the items of information which are likely to be subsequently recognized.

45. An information processing method according to claim 37, wherein said prediction step predicts an item of information which is less likely to be subsequently recognized.

46. An information processing method according to claim 45, wherein said recognition step rejects the item of information which is less likely to be subsequently recognized.

47. An information processing method according to claim 37, further comprising a determination step of determining the correctness of information obtained as a recognition result in said recognition step.

48. An information processing method according to claim 47, wherein said prediction step performs a predicting operation based on a determination result obtained in said determination step.

49. An information processing method according to claim 47, further comprising a correction step of correcting the information obtained as a recognition result in said recognition step based on a determination result obtained in said determination step.

50. An information processing method according to claim 49, wherein said correction step corrects information obtained as a new recognition result by checking the knowledge stored in said knowledge base based on a previous recognition result.

51. An information processing method according to claim 49, wherein said prediction step performs a predicting operation based on a recognition result obtained in said recognition step and performs a re-predicting operation based on a correction result obtained by correcting the recognition result in said correction step.

52. An information processing method according to claim 37, wherein said first determining step distinguishes human speech from a mechanical sound based on a frequency difference.

53. An information processing method according to claim 37, further comprising a correction step of correcting information obtained as a recognition result in said recognition step, wherein said prediction step performs a predicting operation based on a correction result obtained in said correction step.

54. An information processing method according to claim 37, wherein said recognition step recognizes sound information of a natural language in units of one of words, syllables, and phonemes.

55. An information processing method according to claim 54, wherein said prediction step predicts a word including one of a syllable and a phoneme to be subsequently recognized when said recognition step uses one of the syllables and the phonemes, respectively as a unit for recognition, and said prediction step predicts one of the syllable and the phoneme to be subsequently recognized based on the predicted word.

56. An information processing method according to claim 47, further comprising a selection step of selecting the unit for recognition.

57. An information processing method according to claim 56, wherein said selection step selects the unit for recognition according to whether a previous recognition result was successfully obtained.

58. An information processing method according to claim 37, wherein said recognition step recognizes the information by performing a matching operation by checking the knowledge stored in said knowledge base.

59. An information processing method according to claim 58, wherein the matching operation comprises a dynamic programming matching operation.

60. An information processing method according to claim 37, wherein said prediction step predicts the sound information to be subsequently recognized based on a previous operation.

61. An information processing method according to claim 60, wherein said prediction step predicts a subsequent operation based on the previous operation, and predicts the sound information to be subsequently recognized based on the predicted subsequent operation.

62. An information processing method according to claim 61, wherein said prediction step predicts a specific operation as the subsequent operation when it is unable to predict the subsequent operation based on the previous operation.

63. An information processing method according to claim 60, wherein the sound information to be subsequently recognized comprises sound information to be initially recognized.

64. An information processing method according to claim 37, wherein said knowledge base comprises language knowledge.

65. An information processing method according to claim 64, wherein the language knowledge comprises a descriptive rule of natural-language information, and said prediction step predicts the sound information to be subsequently recognized based on the rule.

66. An information processing method according to claim 37, wherein said knowledge base comprises domain knowledge.

67. An information processing method according to claim 66, wherein said prediction step predicts, based on the domain knowledge, information belonging to a specific domain as the sound information to be subsequently recognized.

68. An information processing method according to claim 37, wherein said knowledge base comprises general knowledge.

69. An information processing method according to claim 37, wherein said prediction step predicts sound information to be initially recognized.

70. An information processing method according to claim 37, further comprising a processing step of executing processing corresponding to a processing command recognized as a recognition result in said recognition step.

71. An information processing method according to claim 70, further comprising a reporting step of reporting a processing result obtained in said processing step.

72. An information processing method according to claim 70, wherein said reporting step reports the processing result by sound.

75. An information processing method according to claim 37, wherein said third determining means determines that the inputted sound information is to be processed as non-language information in the event that said second determining means determines that the inputted sound information is the specific language or non-language human sounds, and said recognition step recognizes the inputted sound information as non-language information in the event that said third determining step determines that the inputted sound information is to be processed as non-language information.

77. An information processing method according to claim 37, wherein said third determining step determines that the inputted sound information is to be processed as non-language information in the event that said first determining step determines that the inputted sound information is not produced by the human, and said recognition step recognizes the inputted sound information as non-language information in the event that said third determining step determines that the inputted sound information is to be processed as non-language information.

73. A computer-readable storage medium storing a response program for controlling a computer to perform speech recognition, said program comprising codes for causing the computer to perform:

a first determining step for determining whether inputted sound information is produced by a human;

an acquiring step for acquiring information concerning specific language or non-language human sounds;

a second determining means for determining whether the inputted sound information is the specific language or non-language human sounds;

third determining step for determining that the inputted sound information is language information in the event that said first determining step determines that the inputted sound information is produced by a human and said second determining step determines that the inputted sound information is not the specific language or non-language human sounds;

a recognition step of recognizing the inputted sound information as a specific word, syllable, or phoneme based on prediction information predicting the current input word, syllable, or phoneme as a next input word, syllable, or phoneme to be input based upon the meaning of one or more previously input words in the event that inputted sound information is determined by said third determining step to be language information;

a prediction step of predicting a next word, syllable, or phoneme to be input based upon the meaning of one or more previously input words by checking knowledge stored in a knowledge base for storing knowledge concerning a type of data represented by sound information and for storing the specific language or non-language human sounds; and

an updating step of updating the prediction information based on a prediction result obtained in said prediction step.