Speech recognition method, apparatus and storage medium

• US 20010032075A1
• Filed: 03/27/2001
• Published: 10/18/2001
• Est. Priority Date: 03/31/2000
• Status: Active Grant

• Current Assignee: Canon Ayutthaya Limited (Canon Inc.)
• Original Assignee: Canon Ayutthaya Limited (Canon Inc.)
• Inventors: Yamamoto, Hiroki
• Granted Patent: US 6,801,892 B2
• Time in Patent Office: Days
• US Class Current: 704/256
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/142 Hidden Markov Models [HMMs]

Abstract

Disclosed is a speech recognition method in a speech recognition apparatus to applying speech recognition to a voice signal applied thereto. The input voice signal is converted from an analog to a digital signal and sequences of feature vectors are extracted based upon the digital signal (S12). A search space is defined by the sequences of feature vectors and an HMM (16) prepared beforehand for each unit of speech. The search space allows a transition between HMMs only in specific feature-vector sequences. A search is conducted in this space to find an optimum path for which the largest acoustic likelihood regarding the voice signal is obtained to find the result of recognition (S14), and this result is output (S15).

Claims

1. A speech recognition method comprising the steps of:

extracting sequences of feature vectors from an entered voice signal; and

subjecting the voice signal to speech recognition using a search space in which an HMM-to-HMM transition is not allowed in specific feature-vector sequences.

2. The method according to

claim 1

, wherein the search space is a space defined by two axes, one axis of which comprises HMM state sequences corresponding to one or more words to be recognized and the other axis of which comprises the feature-vector sequences.

3. The method according to

claim 1

, wherein the search space allows an HMM-to-HMM transition only every N-number of feature-vector sequences.

4. The method according to

claim 3

, wherein the value of N falls within the range 2≦

N≦

4.

5. The method according to

claim 1

, wherein the HMM corresponds to any of a phoneme, syllable, demi-syllable, word or diphone.

6. A program executed by a computer, said program comprising:

program code of a step of extracting sequences of feature vectors from an entered voice signal; and

program code of a step of subjecting the voice signal to speech recognition using a search space in which an HMM-to-HMM transition is not allowed in specific feature-vector sequences.

7. A storage medium storing a program executed by a computer, said storage medium comprising:

program code of a step of extracting sequences of feature vectors from an entered voice signal; and

program code of a step of subjecting the voice signal to speech recognition using a search space in which an HMM-to-HMM transition is not allowed in specific feature-vector sequences.

8. A speech recognition apparatus comprising:

extraction means for extracting sequences of feature vectors from an entered voice signal; and

recognition means for subjecting the voice signal to speech recognition using a search space in which an HMM-to-HMM transition is not allowed in specific feature-vector sequences.

9. The apparatus according to

claim 8

, wherein the search space is a space defined by two axes, one axis of which comprises HMM state sequences corresponding to one or more words to be recognized and the other axis of which comprises the feature-vector sequences.

10. The apparatus according to

claim 8

, wherein the search space allows an HMM-to-HMM transition only every N-number of feature-vector sequences.

11. The apparatus according to

claim 10

, wherein the value of N falls within the range 2≦

N≦

4.

12. The apparatus according to

claim 8

, wherein the HMM corresponds to any of a phoneme, syllable, demi-syllable, word or diphone.