KEYWORD MODEL GENERATION FOR DETECTING USER-DEFINED KEYWORD

• US 20150302847A1
• Filed: 08/22/2014
• Published: 10/22/2015
• Est. Priority Date: 04/17/2014
• Status: Active Grant

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: Yun, Sungrack, Kim, Taesu
• Granted Patent: US 9,953,632 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/07 to the speaker
• CPC Class Codes: G10L 2015/027 Syllables being the recognition units
• CPC Class Codes: G10L 2015/088 Word spotting
• CPC Class Codes: G10L 2015/226 using non-speech characteristics

Abstract

According to an aspect of the present disclosure, a method for generating a keyword model of a user-defined keyword in an electronic device is disclosed. The method includes receiving at least one input indicative of the user-defined keyword, determining a sequence of subwords from the at least one input, generating the keyword model associated with the user-defined keyword based on the sequence of subwords and a subword model of the subwords, wherein the subword model is configured to model a plurality of acoustic features of the subwords based on a speech database, and providing the keyword model associated with the user-defined keyword to a voice activation unit configured with a keyword model associated with a predetermined keyword.

Claims

1. A method, performed in an electronic device, for generating a keyword model of a user-defined keyword, the method comprising:

receiving at least one input indicative of the user-defined keyword;

determining a sequence of subwords from the at least one input;

generating the keyword model associated with the user-defined keyword based on the sequence of subwords and a subword model of the subwords, wherein the subword model is configured to model a plurality of acoustic features of the subwords based on a speech database; and

providing the keyword model associated with the user-defined keyword to a voice activation unit configured with a keyword model associated with a predetermined keyword.

2. The method of claim 1, further comprising:

detecting, by the voice activation unit, the user-defined keyword or the predetermined keyword in an input sound based on the keyword model associated with the user-defined keyword and the keyword model associated with the predetermined keyword.

3. The method of claim 2, further comprising:

performing a function associated with the user-defined keyword or the predetermined keyword.

4. The method of claim 1, wherein the subword model is adjusted based on an input sound for at least one sentence.

5. The method of claim 1, wherein generating the keyword model associated with the user-defined keyword comprises determining a threshold score of the keyword model associated with the user-defined keyword.

6. The method of claim 5, wherein the keyword model associated with the user-defined keyword is adjusted by:

receiving a test input sound indicative of the user-defined keyword;

determining a matching score of the test input sound based on the keyword model associated with the user-defined keyword; and

adjusting the threshold score of the keyword model associated with the user-defined keyword based on the matching score.

7. The method of claim 1, wherein the at least one input includes text for the user-defined keyword.

8. The method of claim 1, wherein the at least one input includes at least one sample sound indicative of the user-defined keyword.

9. The method of claim 8, wherein determining the sequence of subwords comprises:

generating at least one sequence of subwords from the at least one sample sound based on the subword model; and

determining the sequence of subwords based on the at least one sequence of subwords.

10. The method of claim 9, wherein the subwords in the at least one sequence of subwords include at least one of phones, phonemes, triphones, and syllables.

11. The method of claim 8, further comprising:

generating at least one mixed sample sound by mixing the at least one sample sound with at least one type of noise.

12. The method of claim 11, wherein determining the sequence of subwords comprises:

generating at least two sequences of subwords from the at least one sample sound and the at least one mixed sample sound based on the subword model; and

determining the sequence of subwords based on the at least two sequences of subwords.

13. The method of claim 9, wherein determining the sequence of subwords based on the at least one sequence of subwords comprises selecting one of the at least one sequence of subwords having a longest length as the sequence of subwords.

14. The method of claim 11, wherein generating the keyword model associated with the user-defined keyword comprises:

generating the keyword model associated with the user-defined keyword based on the sequence of subwords, the subword model, the at least one sample sound, and the at least one mixed sample sound.

15. The method of claim 1, further comprising:

adding a silence portion as a subword unit to a beginning or an end of the sequence of subwords to generate the sequence of subwords.

16. The method of claim 1, wherein the at least one input includes text for the user-defined keyword and at least one sample sound indicative of the user-defined keyword.

17. An electronic device for generating a keyword model of a user-defined keyword, comprising:

an input unit configured to receive at least one input indicative of the user-defined keyword;

a voice activation unit configured with a keyword model associated with a predetermined keyword; and

a user-defined keyword model generation unit configured to determine a sequence of subwords from the at least one input, generate the keyword model associated with the user-defined keyword based on the sequence of subwords and a subword model of the subwords, and provide the keyword model associated with the user-defined keyword to the voice activation unit,

wherein the subword model is configured to model a plurality of acoustic features of the subwords based on a speech database.

18. The electronic device of claim 17, wherein the at least one input includes text for the user-defined keyword.

19. The electronic device of claim 17, wherein the input unit includes a sound sensor configured to receive at least one sample sound indicative of the user-defined keyword as the at least one input.

20. The electronic device of claim 19, further comprising a subword recognition unit configured to generate at least one sequence of subwords from the at least one sample sound based on the subword model.

21. The electronic device of claim 19, further comprising a mixed sound generation unit configured to generate at least one mixed sample sound by mixing the at least one sample sound with at least one type of noise.

22. The electronic device of claim 17, wherein the voice activation unit is configured to detect the user-defined keyword or the predetermined keyword in an input sound based on the keyword model associated with the user-defined keyword and the keyword model associated with the predetermined keyword.

23. The electronic device of claim 17, wherein the voice activation unit is configured to perform a function associated with the user-defined keyword or the predetermined keyword.

24. The electronic device of claim 19, further comprising a pre-adaptation unit configured to adjust the subword model based on an input sound for at least one sentence that is received by the sound sensor.

25. The electronic device of claim 17, wherein the user-defined keyword model generation unit is configured to determine a threshold score of the keyword model associated with the user-defined keyword.

26. The electronic device of claim 25, wherein the user-defined keyword model generation unit is configured to:

determine a matching score of a test input sound indicative of the user-defined keyword that is received by the sound sensor based on the keyword model associated with the user-defined keyword; and

adjust the threshold score of the keyword model associated with the user-defined keyword based on the matching score.

27. A non-transitory computer-readable storage medium storing instructions for generating a keyword model of a user-defined keyword in an electronic device, the instructions causing a processor to perform operations of:

receiving at least one input indicative of the user-defined keyword;

determining a sequence of subwords from the at least one input;

generating the keyword model associated with the user-defined keyword based on the sequence of subwords and a subword model of the subwords, wherein the subword model is configured to model a plurality of acoustic features of the subwords based on a speech database; and

providing the keyword model associated with the user-defined keyword to a voice activation unit configured with a keyword model associated with a predetermined keyword.

28. An electronic device for generating a keyword model of a user-defined keyword, comprising:

means for receiving at least one input indicative of the user-defined keyword;

means for determining a sequence of subwords from the at least one input;

means for generating the keyword model associated with the user-defined keyword based on the sequence of subwords and a subword model of the subwords, wherein the subword model is configured to model a plurality of acoustic features of the subwords based on a speech database; and

means for providing the keyword model associated with the user-defined keyword to a voice activation unit configured with a keyword model associated with a predetermined keyword.

29. The electronic device of claim 28, wherein the at least one input includes at least one sample sound indicative of the user-defined keyword.

30. The electronic device of claim 29, further comprising:

means for generating at least one mixed sample sound by mixing the at least one sample sound with at least one type of noise.