METHOD AND DEVICE FOR RECOGNIZING VOICE

• US 20170011736A1
• Filed: 12/18/2014
• Published: 01/12/2017
• Est. Priority Date: 04/01/2014
• Status: Active Grant

• Current Assignee: Baidu Online Network Technology (Beijing) Co., Ltd (Baidu Incorporated)
• Original Assignee: Baidu Online Network Technology (Beijing) Co., Ltd (Baidu Incorporated)
• Inventors: WANG, Zhiqian, HU, Na, MU, Xiangyu, LI, Bo, JIA, Lei, WEI, Wei
• Granted Patent: US 9,805,712 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G10L 15/01 Assessment or evaluation of speech recognition systems
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 19/0208 Subband vocoders
• CPC Class Codes: G10L 19/022 Blocking, i.e. grouping of samples in time; Choice of analysis windows; Overlap factoring

Abstract

A method for recognizing a voice and a device for recognizing a voice are provided. The method includes: collecting voice information input by a user; extracting characteristics from the voice information to obtain characteristic information; decoding the characteristic information according to an acoustic model and a language model obtained in advance to obtain recognized voice information, wherein the acoustic model is obtained by data compression in advance.

Claims

1. A method for recognizing a voice, comprising:

collecting voice information input by a user;

extracting characteristics from the voice information to obtain characteristic information;

decoding the characteristic information according to an acoustic model and a language model obtained in advance to obtain recognized voice information, wherein the acoustic model is obtained by data compression in advance.

2. The method according to claim 1, wherein after obtaining characteristic information, the method further comprises:

filtering the characteristic information to obtain filtered characteristic information, so as to decode the filtered characteristic information.

3. The method according to claim 2, wherein filtering the characteristic information comprises:

performing an extraction of frame skipping on the characteristic information.

4. The method according to claim 1, wherein decoding the characteristic information according to an acoustic model and a language model obtained in advance to obtain recognized voice information comprises:

performing a data compression on the characteristic information to obtain compressed characteristic information, and calculating the compressed characteristic information according to the acoustic model that is obtained by the data compression in advance to obtain a score of acoustic model;

calculating data after acoustic model scoring according to the language model to obtain a score of language model;

obtaining the recognized voice information according to the score of acoustic model and the score of language model.

5. The method according to claim 4, wherein calculating the compressed characteristic information comprises:

performing a parallel computation on the compressed characteristic information.

6. The method according to claim 5, wherein the parallel computation comprises at least one of data parallel computation, instruction parallel computation and thread parallel computation.

14. The method according to claim 2, wherein decoding the characteristic information according to an acoustic model and a language model obtained in advance to obtain recognized voice information comprises:

performing a data compression on the characteristic information to obtain compressed characteristic information, and calculating the compressed characteristic information according to the acoustic model that is obtained by the data compression in advance to obtain a score of acoustic model;

calculating data after acoustic model scoring according to the language model to obtain a score of language model;

obtaining the recognized voice information according to the score of acoustic model and the score of language model.

15. The method according to claim 3, wherein decoding the characteristic information according to an acoustic model and a language model obtained in advance to obtain recognized voice information comprises:

performing a data compression on the characteristic information to obtain compressed characteristic information, and calculating the compressed characteristic information according to the acoustic model that is obtained by the data compression in advance to obtain a score of acoustic model;

calculating data after acoustic model scoring according to the language model to obtain a score of language model;

obtaining the recognized voice information according to the score of acoustic model and the score of language model.

7. A device for recognizing a voice, comprising:

a collecting module, configured to collect voice information input by a user;

an extracting module, configured to extract characteristics from the voice information to obtain characteristic information;

a decoding module, configured to decode the characteristic information according to an acoustic model and a language model obtained in advance to obtain recognized voice information, wherein the acoustic model is obtained by data compression in advance.

8. The device according to claim 7, further comprising:

a filtering module, configured to filter the characteristic information to obtain filtered characteristic information, so as to decode the filtered characteristic information.

9. The device according to claim 8, wherein the filtering module is specifically configured to perform an extraction of frame skipping on the characteristic information.

10. The device according to claim 7, wherein the decoding module is specifically configured to:

perform a data compression on the characteristic information to obtain compressed characteristic information, and calculate the compressed characteristic information according to the acoustic model that is obtained by the data compression in advance to obtain a score of acoustic model;

calculate data after acoustic model scoring according to the language model to obtain a score of language model;

obtain the recognized voice information according to the score of acoustic model and the score of language model.

11. The device according to claim 10, wherein the decoding module calculates the compressed characteristic information by:

performing a parallel computation on the compressed characteristic information.

12. The device according to claim 11, wherein the parallel computation comprises at least one of data parallel computation, instruction parallel computation and thread parallel computation.

16. The device according to claim 8, wherein the decoding module is specifically configured to:

perform a data compression on the characteristic information to obtain compressed characteristic information, and calculate the compressed characteristic information according to the acoustic model that is obtained by the data compression in advance to obtain a score of acoustic model;

calculate data after acoustic model scoring according to the language model to obtain a score of language model;

obtain the recognized voice information according to the score of acoustic model and the score of language model.

17. The device according to claim 9, wherein the decoding module is specifically configured to:

perform a data compression on the characteristic information to obtain compressed characteristic information, and calculate the compressed characteristic information according to the acoustic model that is obtained by the data compression in advance to obtain a score of acoustic model;

calculate data after acoustic model scoring according to the language model to obtain a score of language model;

obtain the recognized voice information according to the score of acoustic model and the score of language model.

13. A mobile device, comprising:

one or more processors;

a memory;

one or more programs, wherein the one or more programs are stored in the memory, and when executed by the one or more processors, perform following operations;

collecting voice information input by a user;

extracting characteristics from the voice information to obtain characteristic information;

decoding the characteristic information according to an acoustic model and a language model obtained in advance to obtain recognized voice information, wherein the acoustic model is obtained by data compression in advance.