Speech recognition device, vehicle having the same, and speech recognition method

Patent metadata

Field	Value
Publication number	US-9812125-B2
Application number	US-201414562243-A
Country	US
Kind code	B2
Filing date	Dec 5, 2014
Priority date	Jul 28, 2014
Publication date	Nov 7, 2017
Grant date	Nov 7, 2017

Abstract

A speech recognition device is configured to increase usability by retrying speech recognition without returning to a previous operation or a re-input of speech when a user's speech is misrecognized. The speech recognition device is further configured increase accuracy of recognition by changing a search environment when the user's speech is misrecognized or when re-recognition is performed since the recognized speech is rejected due to a low confidence. A vehicle includes a speech input device configured to receive speech; and a speech recognition device configured to recognize the received speech and output a recognition result of the received speech. The speech recognition device resets a recognition environment applied to speech recognition and re-recognizes the received speech when a re-recognition instruction is input by a user, and resets the reset recognition environment to an initial value when the re-recognition is completed.

First claim

What is claimed is:

1. A vehicle comprising:
a speech input device configured to receive speech;
a speech recognition device configured to recognize the received speech and output a recognition result of the received speech including a plurality of instruction candidates corresponding to the received speech; and
at least one of a display configured to display the plurality of instruction candidates or a speaker configured to output the plurality of instruction candidates,
wherein the speech recognition device resets a recognition environment applied to speech recognition and re-recognizes the received speech when a re-recognition instruction is input by a user, excludes the recognition result of the received speech previously output when the re-recognition is performed, and resets the reset recognition environment to an initial value when the re-recognition is completed,
wherein the recognition environment includes at least one of an accuracy parameter related to accuracy of speech recognition, a threshold value of a confidence score, or a search range, and
wherein the accuracy parameter represents information on the number of search nodes, and accuracy of speech recognition increases as the number of search nodes increases.

2. The vehicle according to claim 1 , wherein the speech recognition device searches for the plurality of instruction candidates corresponding to the received speech, and outputs an instruction having the confidence score greater than a predetermined threshold value among the found instructions as the recognition result.

3. The vehicle according to claim 1 , wherein the speech recognition device resets the accuracy parameter to be higher.

4. The vehicle according to claim 1 , wherein the speech recognition device resets the threshold value of the confidence score to be smaller.

5. The vehicle according to claim 1 , wherein the speech recognition device converts the received speech into speech data, detects end point information from the speech data to determine a speech section, and extracts a feature vector from the speech section,
and further includes a memory configured to store the detected end point information and the extracted feature vector.

6. The vehicle according to claim 5 , wherein, when the re-recognition instruction is input by the user, the speech recognition device re-recognizes the received speech using the end point information and feature vector stored in the memory.

7. The vehicle according to claim 2 , wherein, when there is no instruction having the confidence score greater than the predetermined threshold value among the found instructions, the speech recognition device resets the recognition environment applied to speech recognition and recognizes speech re-input by the user according to the reset recognition environment.

8. The vehicle according to claim 7 , wherein the recognition environment includes the accuracy parameter related to accuracy of speech recognition and the threshold value of the confidence score.

9. A speech recognition device comprising:
a memory configured to store information on input speech; and
a speech recognition device configured to recognize the input speech and output a recognition result of the input speech, including a plurality of instruction candidates corresponding to a received input speech, to a display or a speaker,
wherein the speech recognition device resets a recognition environment applied to speech recognition and re-recognizes the received input speech when a re-recognition instruction is input by a user, excludes the recognition result of the input speech previously output when the re-recognition is performed, and resets the reset recognition environment to an initial value when the re-recognition is completed,
wherein the recognition environment includes at least one of an accuracy parameter related to accuracy of speech recognition, a threshold value of a confidence score, or a search range, and
wherein the accuracy parameter represents information on the number of search nodes, and accuracy of speech recognition increases as the number of search nodes increases.

10. The device according to claim 9 ,
wherein the speech recognition device searches for the plurality of instruction candidates corresponding to the input speech, and outputs an instruction having the confidence score greater than a predetermined threshold value among the found instructions as the recognition result.

11. The device according to claim 9 ,
wherein the speech recognition device resets the accuracy parameter to be higher.

12. The device according to claim 10 ,
wherein the speech recognition device resets the threshold value of the confidence score to be smaller.

13. The device according to claim 9 ,
wherein the speech recognition device detects end point infomation from speech data to determine a speech section, and extracts a feature vector from the speech section, and
information on the input speech includes the detected end point information and the extracted feature vector.

14. The device according to claim 13 ,
wherein, when the re-recognition instruction is input by the user, the speech recognition device re-recognizes the input speech using the end point information and feature vector stored in the memory.

15. A speech recognition method comprising:
recognizing input speech, using a speech recognition device, when speech is input;
outputting a recognition result of the input speech, including a plurality of instruction candidates corresponding to a received input speech, using a display or a speaker;
when a re-recognition instruction is input by a user, resetting a recognition environment including at least one of an accuracy parameter related to accuracy of speech recognition, a threshold value of a confidence score, or a search range, using a speech recognition device;
excluding the recognition result of the input speech output before the re-recognition instruction is input from a search range when the re-recognition is performed, using the speech recognition device;
recognizing the received input speech again by applying the reset recognition environment using the speech recognition device; and
outputting a re-recognition result, excluding the recognition result of the input speech when the re-recognition is performed,
wherein the accuracy parameter represents information on the number of search nodes, and accuracy of speech recognition increases as the number of search nodes increases.

16. The method according to claim 15 , further comprising
resetting the reset recognition environment to an initial value again when the re-recognition is completed.

17. The method according to claim 15 ,
wherein the resetting of the recognition environment includes resetting the accuracy parameter to be higher.

18. The method according to claim 15 ,
wherein the resetting of the recognition environment includes resetting the threshold value of the confidence score to be smaller.

19. A vehicle comprising:
a speech input device configured to receive speech;
a speech recognition device configured to recognize the received speech and output a recognition result of the received speech including a plurality of instruction candidates corresponding to the received speech; and
at least one of a display configured to display the plurality of instruction candidates or a speaker configured to output the plurality of instruction candidates,
wherein the speech recognition device resets a recognition environment applied to speech recognition and re-reco

Assignees

• Hyundai Motor Co Ltd

Inventors

• Lee Chang-Heon
• Bang Kyuseop

Classifications

• G10L15/02

  Feature extraction for speech recognition; Selection of recognition unit · CPC title

• G10L15/08

  Speech classification or search · CPC title

• G10L15/26

  Speech to text systems (G10L15/08 takes precedence) · CPC title

• G10L2015/228

  of application context · CPC title

• G10L15/265

  Physics · mapped topic