Voice recognition system and voice recognition device

• US 9,761,228 B2
• Filed: 11/20/2013
• Issued: 09/12/2017
• Est. Priority Date: 02/25/2013
• Status: Active Grant

• Current Assignee: Mitsubishi Electric Corporation
• Original Assignee: Mitsubishi Electric Corporation
• Inventors: Ogawa, Isamu, Hanazawa, Toshiyuki, Narita, Tomohiro
• Primary Examiner(s): Yen, Eric
• Application Number:
  US14/649,128
• Publication Number:
  US 20160275950A1
• Time in Patent Office: 1,392 Days
• CPC Class Codes: G10L 15/10 using distance or distortion measures between unknown speech and reference templates
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems
• CPC Class Codes: G10L 15/34 Adaptation of a single recogniser for parallel processing, e.g. by use of multiple processors or cloud computing
• CPC Class Codes: G10L 25/51 for comparison or discrimination

Abstract

There are provided a recognition result candidate comparator 205 that compares a plurality of server-side voice recognition result candidates received by a receiver 204, to detect texts having a difference, and a recognition result integrator 206 that integrates a client-side voice recognition result candidate and a server-side voice recognition result candidate on the basis of the client-side voice recognition result candidate, the server-side voice recognition result candidate, and a detection result provided by the recognition result candidate comparator 205, to decide a voice recognition result.

Claims

1. A voice recognition system including a server device and a voice recognition device on a client side, which is connected to said server device, wherein

said server device comprises:

a server-side receiver that receives voice data from said voice recognition device;

a server-side voice recognizer that performs voice recognition on said voice data received by said server-side receiver, and generates a plurality of server-side voice recognition result candidates; and

a server-side transmitter that transmits said plurality of server-side voice recognition result candidates generated by said server-side voice recognizer to said voice recognition device, and wherein

said voice recognition device comprises;

a voice inputter that converts a voice into said voice data;

a client-side voice recognizer that performs voice recognition on said voice data converted by said voice inputter, and generates a client-side voice recognition result candidate;

a client-side transmitter that transmits said voice data converted by said voice inputter to said server device;

a client-side receiver that receives said plurality of server-side voice recognition result candidates transmitted by said server-side transmitter,

a recognition result candidate comparator that compares said plurality of server-side voice recognition result candidates received by said client-side receiver, to detect texts having a difference;

a recognition result integrator that integrates said client-side voice recognition result candidate and said plurality of server-side voice recognition result candidates by replacing a portion of one of said plurality of server-side voice recognition result candidates with a portion of said client-side voice recognition result candidate based on said client-side voice recognition result candidate, said plurality of server-side voice recognition result candidates, and a detection result provided by said recognition result candidate comparator, to decide a voice recognition result; and

an outputter that outputs said voice recognition result decided by said recognition result integrator.

2. The voice recognition system according to claim 1, wherein said voice recognition device includes an input rule determinator that compares said client-side voice recognition result candidate with utterance rule patterns in each of which a predetermined key word is brought into correspondence with an utterance rule of said predetermined key word, and determines an utterance rule of said voice data, and wherein said recognition result integrator integrates said client-side voice recognition result candidate and said plurality of server-side voice recognition result candidates based on said client-side voice recognition result candidate, said plurality of server-side voice recognition result candidates, the detection result provided by said recognition result candidate comparator, and the utterance rule determined by said input rule determinator.

3. The voice recognition system according to claim 2, wherein said voice recognition device includes an input voice/recognition result storage that stores said voice data converted by said voice inputter and the voice recognition result decided by said recognition result integrator while bringing them into correspondence with each other, and a recognition result candidate corrector that acquires said plurality of server-side voice recognition result candidates corresponding to said voice data stored in said input voice/recognition result storage and generates a database when said voice recognition device is started, and that also compares the generated database with said plurality of server-side voice recognition result candidates received by said client-side receiver, and wherein said recognition result integrator integrates said client-side voice recognition result candidate and said plurality of server-side voice recognition result candidates based on a comparison result provided by said recognition result candidate corrector.

4. The voice recognition system according to claim 2, wherein said recognition result candidate comparator compares said plurality of said server-side voice recognition result candidates received by said client-side receiver to detect a plurality of texts having a difference, and determines whether the plurality of detected texts show same contents, and said recognition result integrator replaces a detected text with a proper noun based on said plurality of server-side voice recognition result candidates when said recognition result candidate comparator determines that the plurality of detected texts detected show the same contents.

5. The voice recognition system according to claim 1, wherein said recognition result comparator compares said plurality of server-side voice recognition result candidates when said recognition result comparator determines that said plurality of server-side voice recognition result candidates are received from said server device.

6. The voice recognition system according to claim 1, wherein said client-side transmitter transmits said voice data to said server device before said client-side voice recognizer generates said client-side voice recognition result candidate.

7. The voice recognition system according to claim 1, wherein said recognition result integrator decides the voice recognition result without calculating a corresponding numerical score for said client-side voice recognition result candidate and said plurality of server-side voice recognition result candidates.

8. The voice recognition system according to claim 1, wherein said recognition result integrator decides the voice recognition result to be the one of said plurality of server-side voice recognition result candidates when said recognition result candidate comparator does not detect any difference between said plurality of server-side voice recognition result candidates.

9. A voice recognition device on a client side which is connected to a server device having a voice recognition function, said voice recognition device comprising:

a voice inputter that converts a voice into voice data;

a client-side voice recognizer that performs voice recognition on said voice data converted by said voice inputter, and generates a client-side voice recognition result candidate;

a client-side transmitter that transmits said voice data converted by said voice inputter to said server device;

a client-side receiver that receives a plurality of server-side voice recognition result candidates generated by said server device based on said voice data transmitted by said client-side transmitter;

a recognition result candidate comparator that compares said plurality of server-side voice recognition result candidates received by said client-side receiver, to detect texts having a difference;

a recognition result integrator that integrates said client-side voice recognition result candidate and said plurality of server-side voice recognition result candidates by replacing a portion of one of said plurality of server-side voice recognition result candidates with a portion of said client-side voice recognition result candidate based on said client-side voice recognition result candidate, said plurality of server-side voice recognition result candidates, and a detection result provided by said recognition result candidate comparator, to decide a voice recognition result; and

an outputter that outputs the voice recognition result decided by said recognition result integrator.

10. A voice recognition system including a server device and a voice recognition device on a client side, which is connected to said server device, wherein

said server device comprises:

a server-side receiver that receives voice data from said voice recognition device;

a server-side voice recognizer that performs voice recognition on said voice data received by said server-side receiver, and generates a plurality of server-side voice recognition result candidates; and

a server-side transmitter that transmits said plurality of server-side voice recognition result candidates generated by said server-side voice recognizer to said voice recognition device, and wherein

said voice recognition device comprises;

a voice inputter that converts a voice into said voice data;

a client-side voice recognizer that performs voice recognition on said voice data converted by said voice inputter, and generates a client-side voice recognition result candidate;

a client-side transmitter that transmits said voice data converted by said voice inputter to said server device;

a client-side receiver that receives said plurality of server-side voice recognition result candidates transmitted by said server-side transmitter;

a recognition result candidate comparator that compares said plurality of server-side voice recognition result candidates received by said client-side receiver, to detect texts having a difference;

a recognition result integrator that integrates said client-side voice recognition result candidate and said plurality of server-side voice recognition result candidates, said plurality of server-side voice recognition result candidates, and a detection result provided by said recognition result candidate comparator, to decide a voice recognition result; and

an outputter that outputs said voice recognition result decided by said recognition result integrator,

wherein said voice recognition device includes an input rule determinator that compares said client-side voice recognition result candidate with utterance rule patterns in each of which a predetermined key word is brought into correspondence with an utterance rule of said predetermined key word, and determines an utterance rule of said voice data, and wherein said recognition result integrator integrates said client-side voice recognition result candidate and said plurality of server-side voice recognition result candidates based on said client-side voice recognition result candidate, said plurality of server-side voice recognition result candidates, the detection result provided by said recognition result candidate comparator, and the utterance rule determined by said input rule determinator.