SPEECH RECOGNITION APPARATUS AND METHOD

• US 20080162137A1
• Filed: 11/13/2007
• Published: 07/03/2008
• Est. Priority Date: 12/28/2006
• Status: Active Grant

• Current Assignee: Nissan Motor Co., Ltd.
• Original Assignee: Nissan Motor Co., Ltd.
• Inventors: Ono, Takeshi, Katsuragawa, Keiko, Tomikashi, Minoru, Tozuka, Eiji, Saitoh, Daisuke
• Granted Patent: US 7,949,524 B2
• Time in Patent Office: Days
• US Class Current: 704/251
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/088 Word spotting

Abstract

At least one recognized keyword is presented to a user, and the keyword is corrected appropriately upon receipt of a correction for the presented result. A standby-word dictionary used for recognition of uttered speech is generated according to a result of the correction to recognize the uttered speech. Therefore, even if recognized keywords contain an error, the error can be corrected and uttered speech can be accurately recognized.

Claims

1. A speech recognition apparatus comprising:

a speech input unit configured to receive an input of a speech utterance;

a keyword recognition unit configured to recognize at least one keyword included in the speech utterance as a recognition result;

a presentation unit configured to present the recognition result;

a correction input unit configured to receive a correction input for the recognition result;

a correction unit configured to correct the recognition result based on the correction input to create a correction result;

a dictionary generation unit configured to generate a standby-word dictionary for recognizing the speech utterance using the correction result; and

a speech utterance recognition unit configured to recognize the speech utterance using the standby-word dictionary.

2. The speech recognition apparatus according to claim 1 wherein the correction input unit is configured to receive the correction input by speech via the speech input unit.

3. The speech recognition apparatus according to claim 1, further comprising:

an operation unit operated by an operator wherein the correction input unit is configured to receive the correction input by operation of the operation unit.

4. The speech recognition apparatus according to claim 3 wherein the operation unit includes a touch panel.

5. The speech recognition apparatus according to claim 1 wherein the presentation unit is configured to select a keyword having a recognition score within a predetermined range from among the recognition result and to present the keyword.

6. The speech recognition apparatus according to claim 1 wherein the presentation unit is configured to select at least two keywords based on a similarity of acoustic features from among the recognition result and to present the selected keywords.

7. The speech recognition apparatus according to claim 1 wherein the presentation unit is configured to select a keyword based on a data size of the standby-word dictionary from among the recognition result and to present the selected keyword.

8. The speech recognition apparatus according to claim 1 wherein the presentation unit is configured to select keywords based on a probability of appearing within a recognition-target vocabulary element from among the recognition result and to present the selected keywords.

9. The speech recognition apparatus according to claim 1 wherein the presentation unit, the correction input unit, and the correction unit are configured to operate in parallel with the dictionary generation unit and the speech utterance unit.

10. The speech recognition apparatus according to claim 1 wherein the keyword recognition unit is configured to recognize a plurality of keywords, the dictionary generation unit is configured to generate the standby-word dictionary as a union of recognition-target vocabulary elements including the plurality of keywords corrected by the correction unit.

11. The speech recognition apparatus according to claim 1 wherein the correction input is at least one of a deletion from and an addition to the recognition result.

12. The speech recognition apparatus according to claim 1 wherein the dictionary generation unit is further configured to specify an appearance position within the speech utterance of each of the at least one keyword and to generate the standby-word dictionary based on the appearance position.

13. The speech recognition apparatus according to claim 1 wherein the dictionary generation unit is further configured to generate a search query according to the correction result for searching the standby-word dictionary and to generate the standby-word dictionary so as to include a recognition-target vocabulary element retrieved using the search query.

14. A speech recognition apparatus comprising:

means for receiving a speech utterance;

means for recognizing at least one keyword included in the speech utterance as a recognition result;

means for presenting the recognition result;

means for receiving a correction to the recognition result;

means for correcting the recognition result based on the correction to generate a correction result;

means for generating a standby-word dictionary using the correction result; and

means for recognizing the speech utterance using the standby-word dictionary.

15. A speech recognition method comprising:

receiving a speech utterance;

presenting at least one keyword included in the speech utterance;

correcting the at least one keyword;

generating a standby-word dictionary according to a result of correcting the at least one keyword; and

recognizing the speech utterance using the standby-word dictionary.

16. The speech recognition method of claim 15, further comprising:

selecting from among the at least one keyword a selected keyword having a recognition score within a predetermined range; and

presenting the selected keyword.

17. The speech recognition method of claim 15, further comprising:

selecting from among the at least one keyword a selected keyword based on a standby-word dictionary data size of the selected keyword; and

presenting the selected keyword.

18. The speech recognition method of claim 17, further comprising:

selecting from among the at least one keyword a selected keyword having the largest standby-dictionary data size; and

presenting the selected keyword.

19. The speech recognition method of claim 15 wherein presenting the at least one keyword further includes recognizing a plurality of keywords, the method further comprising:

selecting from among the plurality of keywords at least two selected keywords based on a conflict in meaning between the selected keywords; and

presenting the selected keywords.

20. The speech recognition method of claim 15 wherein presenting the at least one keyword further includes recognizing a plurality of keywords, the method farther comprising:

selecting from among the plurality of keywords at least two selected keywords based on an interphoneme distance between the selected keywords; and

presenting the selected keywords.