Speech processing apparatus, speech processing method, program, and recording medium

• US 7,657,430 B2
• Filed: 07/20/2005
• Issued: 02/02/2010
• Est. Priority Date: 07/22/2004
• Status: Expired due to Fees

• Current Assignee: Sony Corporation (Sony Group Corp.)
• Original Assignee: Sony Corporation (Sony Group Corp.)
• Inventors: Ogawa, Hiroaki
• Primary Examiner(s): Dorvil; Richemond
• Assistant Examiner(s): YEN, ERIC L
• Application Number:
  US11/185,182
• Publication Number:
  US 20060020461A1
• Time in Patent Office: 1,658 Days
• Field of Search: 704251-252, 704/255, 704/257, 704243-244
• US Class Current: 704/243
• CPC Class Codes: G10L 15/04 Segmentation; Word boundary detection
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering

Abstract

An apparatus inputs an utterance and performs speech recognition on the input utterance. The speech processing apparatus determines whether the recognition result contains an unknown word. If it is determined that the recognition result contains an unknown word, it is then determined whether the recognition result is rejected or not. If it is determined that the recognition result is not rejected, a word corresponding to the unknown word contained in the recognition result is required. The apparatus can be used as a speech processing apparatus.

Claims

1. A speech processing apparatus for processing an input utterance and registering an unknown word contained in the input utterance into a dictionary database on the basis of the processing result, comprising:

recognition means for recognizing the input utterance;

unknown word determination means for determining whether the recognition result of the input utterance obtained by the recognition means contains an unknown word on the basis of an acoustic model representing acoustic features of individual phonemes and syllables of a language;

recognition result rejection means for determining whether the recognition result determined by the unknown word determination means to contain an unknown word is rejected or not for acquisition and for registering into the dictionary database; and

word extracting means for registering and for acquiring pronunciation of a word corresponding to the unknown word contained in the recognition result determined not to be rejected by the recognition result rejection means, wherein the acquired pronunciation of the word is used for subsequent speech recognition,

wherein the recognition result rejection means determines whether the recognition result is rejected or not on the basis of a confidence measure for a known word immediately before the unknown word and a confidence measure for a known word immediately after the unknown word contained in the recognition result.

2. The speech processing apparatus according to claim 1, wherein the recognition result rejection means determines whether the recognition result is rejected or not on the basis of an unknown-word language model for a sentence containing an unknown word.

3. The speech processing apparatus according to claim 2, wherein the unknown-word language model is a grammar model written with a finite state automaton.

4. The speech processing apparatus according to claim 2, wherein the unknown-word language model is a statistical language model.

5. The speech processing apparatus according to claim 1, wherein the recognition means recognizing the input utterance based on at least feature parameters which include spectrum, power linear predictive coefficients, spectrum coefficients and a line spectrum pair.

6. A speech processing method for processing an input utterance and registering an unknown word contained in the input utterance into a dictionary database on the basis of the processing result, comprising the steps of:

(a) recognizing, utilizing a recognition unit, the input utterance;

(b) determining, utilizing an unknown word determination unit, whether the recognition result of the input utterance obtained by step (a) contains an unknown word on the basis of an acoustic model representing acoustic features of individual phonemes and syllables of a language;

(c) determining, utilizing a recognition result rejection unit, whether the recognition result determined at step (b) to contain an unknown word is rejected or not for acquisition and for registering into the dictionary database; and

(d) registering and acquiring, utilizing a word extracting unit, pronunciation of a word corresponding to the unknown word contained in the recognition result determined not to be rejected by step (c),

wherein the acquired pronunciation of the word is used for subsequent speech recognition, and

wherein the recognition result rejection unit determines whether the recognition result is rejected or not on the basis of a confidence measure for a known word immediately before the unknown word and a confidence measure for a known word immediately after the unknown word contained in the recognition result.

7. A computer-readable recording medium storing a program, the program processing an input utterance and registering an unknown word contained in the input utterance into a dictionary database on the basis of the processing result, the program including the steps of:

(a) recognizing the input utterance;

(b) determining whether the recognition result of the input utterance obtained by step (a) contains an unknown word on the basis of an acoustic model representing acoustic features of individual phonemes and syllables of a language;

(c) determining whether the recognition result determined at step (b) to contain an unknown word is rejected or not for acquisition and registering into the dictionary database; and

(d) registering and acquiring pronunciation of a word corresponding to the unknown word contained in the recognition result determined not to be rejected by step (c),

wherein the acquired pronunciation of the word is used for subsequent speech recognition, and

wherein the recognition result is determined to be rejected or not on the basis of a confidence measure for a known word immediately before the unknown word and a confidence measure for a known word immediately after the unknown word contained in the recognition result.

8. A speech processing apparatus for processing input utterance and registering an unknown word contained in the input utterance into a database on the basis of the processing result, comprising:

a recognition unit configured to recognize the input utterance;

an unknown word determination unit configured to determine whether the recognition result of the input utterance obtained by the recognition unit contains an unknown word on the basis of an acoustic model representing acoustic features of individual phonemes and syllables of a language;

a recognition result rejection unit configured to determine whether the recognition result determined by the unknown word determination unit to contain an unknown word is rejected or not for acquisition and for registering into the database; and

a word extracting unit configured to register and to acquire pronunciation of a word corresponding to the unknown word contained in the recognition result determined not to be rejected by the recognition result rejection unit,

wherein the acquired pronunciation of the word is used for subsequent speech recognition, and

wherein the recognition result rejection unit determines whether the recognition result is rejected or not on the basis of a confidence measure for a known word immediately before the unknown word and a confidence measure for a known word immediately after the unknown word contained in the recognition result.

9. A speech processing apparatus for processing an input utterance and registering an unknown word contained in the input utterance into a database on the basis of the processing result, comprising:

recognition means for recognizing the input utterance based on both an acoustic score matching a duration of the input utterance for a known word and an acoustic score for a duration of the utterance for an unknown word;

unknown word determination means for determining whether the recognition result of the input utterance obtained by the recognition means contains the unknown word on the basis of an acoustic model representing acoustic features of individual phonemes and syllables of a language;

recognition result rejection means for determining whether the recognition result determined by the unknown word determination means to contain an unknown word is rejected or not for acquisition and for registering into the database; and

word extracting means for storing pronunciation of a word corresponding to the unknown word contained in the recognition result determined not to be rejected by the recognition result rejection means,

wherein the acquired pronunciation of the word is used for subsequent speech recognition, and

wherein the recognition result rejection means determines whether the recognition result is rejected or not on the basis of a confidence measure for a known word immediately before the unknown word and for a known word immediately after the unknown word contained in the recognition result.

10. The speech processing apparatus according to claim 9, wherein the recognition result rejection means determines whether the recognition result is rejected or not on the basis of an unknown-word language model for a sentence containing an unknown word.

11. The speech processing apparatus according to claim 10, wherein the unknown-word language model is a grammar model written with a finite state automaton.

12. The speech processing apparatus according to claim 10, wherein the unknown-word language model is a statistical language model.

13. The speech processing apparatus according to claim 9, wherein matching against known words is performed word by word and matching for the unknown word is performed syllable by syllable.

14. A speech processing method for processing an input utterance and registering an unknown word contained in the input utterance into a database on the basis of the processing result, comprising the steps of:

(a) recognizing, utilizing a recognition unit, the input utterance based on an acoustic score matching a duration of the input utterance for a known word and an acoustic score for a duration of the utterance for an unknown word;

(b) determining, utilizing an unknown word determination unit, whether the recognition result of the input utterance obtained by step (a) contains an unknown word on the basis of an acoustic model representing acoustic features of individual phonemes and syllables of a language;

(c) determining, utilizing a recognition result rejection unit, whether the recognition result determined at step (b) to contain an unknown word is rejected or not for acquisition and for registering into the database; and

(d) storing, utilizing a word extracting unit, pronunciation of a word corresponding to the unknown word contained in the recognition result determined not to be rejected by step (c),

wherein the acquired pronunciation of the word is used for subsequent speech recognition, and

wherein the recognition result rejection unit determines whether the recognition result is rejected or not on the basis of a confidence measure for a known word immediately before the unknown word and a confidence measure for a known word immediately after the unknown word contained in the recognition result.

15. A computer-readable recording medium storing a program, the program processing an input utterance and registering an unknown word contained in the input utterance into a database on the basis of the processing result, the program including the steps of:

(a) recognizing the input utterance based on an acoustic score matching a duration of the input utterance for a known word and an acoustic score for a duration of the utterance for an unknown word;

(b) determining whether the recognition result of the input utterance obtained by step (a) contains an unknown word on the basis of an acoustic model representing acoustic features of individual phonemes and syllables of a language;

(c) determining whether the recognition result determined at step (b) to contain an unknown word is rejected or not for acquisition and for registering into the database; and

(d) storing pronunciation of a word corresponding to the unknown word contained in the recognition result determined not to be rejected by step (c),

wherein the acquired pronunciation of the word is used for subsequent speech recognition, and

wherein the recognition result is determined to be rejected or not on the basis of a confidence measure for a known word immediately before the unknown word and a confidence measure for a known word immediately after the unknown word contained in the recognition result.

16. A speech processing apparatus for processing input utterance and registering an unknown word contained in the input utterance into a database on the basis of the processing result, comprising:

a recognition unit configured to recognize the input utterance based on an acoustic score matching a duration of the input utterance for a known word and an acoustic score for a duration of the utterance for an unknown word on the basis of an acoustic model representing acoustic features of individual phonemes and syllables of a language;

an unknown word determination unit configured to determine whether the recognition result of the input utterance obtained by the recognition unit contains an unknown word;

a recognition result rejection unit configured to determine whether the recognition result determined by the unknown word determination unit to contain an unknown word is rejected or not for acquisition and for registering into the database; and

a word extracting unit configured to store pronunciation of a word corresponding to the unknown word contained in the recognition result determined not to be rejected by the recognition result rejection unit,

wherein the acquired pronunciation of the word is used for subsequent speech recognition, and

wherein the recognition result rejection unit determines whether the recognition result is rejected or not on the basis of a confidence measure for a known word immediately before the unknown word and a confidence measure for a known word immediately after the unknown word contained in the recognition result.