Sound sample verification for generating sound detection model

Patent metadata

Field	Value
Publication number	US-9837068-B2
Application number	US-201514682009-A
Country	US
Kind code	B2
Filing date	Apr 8, 2015
Priority date	Oct 22, 2014
Publication date	Dec 5, 2017
Grant date	Dec 5, 2017

Abstract

A method for verifying at least one sound sample to be used in generating a sound detection model in an electronic device includes receiving a first sound sample; extracting a first acoustic feature from the first sound sample; receiving a second sound sample; extracting a second acoustic feature from the second sound sample; and determining whether the second acoustic feature is similar to the first acoustic feature.

First claim

What is claimed:

1. A method performed in an electronic device, the method comprising:
receiving a first sound sample;
extracting a first acoustic feature from the first sound sample;
receiving a second sound sample;
extracting a second acoustic feature from the second sound sample;
determining whether the second acoustic feature is similar to the first acoustic feature; and
in response to determining that the second acoustic feature is not similar to the first acoustic feature, outputting a determination result indicating whether the second sound sample is to be used in generating a sound detection model, the determination result including a request for a replacement sound sample.

2. The method of claim 1 , wherein the determination result indicates to a user that the second sound sample will not be used in generating the sound detection model.

3. The method of claim 1 , further comprising:
determining a first signal-to-noise ratio of the first sound sample;
determining a second signal-to-noise ratio of the second sound sample; and
generating the sound detection model based on at least one of the first signal-to-noise ratio or the second signal-to-noise ratio.

4. The method of claim 1 , further comprising:
receiving a third sound sample;
extracting a third acoustic feature from the third sound sample;
based on determining that the third acoustic feature is similar to the first acoustic feature, determining a combined acoustic feature based on the first acoustic feature and the third acoustic feature;
receiving a fourth sound sample;
extracting a fourth acoustic feature from the fourth sound sample; and
determining whether the fourth acoustic feature is similar to the combined acoustic feature.

5. The method of claim 1 , wherein each of the first acoustic feature and the second acoustic feature includes at least one of spectral features or time domain features, and wherein the electronic device comprises a mobile communication device.

6. The method of claim 1 , wherein each of the first acoustic feature and the second acoustic feature includes a sequence of subwords, and wherein the subwords include at least one of phones, phonemes, triphones, or syllables.

7. The method of claim 1 , further comprising generating the sound detection model at a device that comprises a multimedia device.

8. The method of claim 1 , further comprising:
receiving an additional sound sample in response to determining that the second acoustic feature is not similar to the first acoustic feature;
extracting a particular acoustic feature from the additional sound sample; and
determining whether the particular acoustic feature is similar to the first acoustic feature.

9. The method of claim 8 , further comprising generating the sound detection model based on at least one of the first sound sample and the additional sound sample in response to determining that the particular acoustic feature is similar to the first acoustic feature.

10. The method of claim 9 , wherein generating the sound detection model comprises:
determining a threshold value of the sound detection model based on at least one of the first sound sample or the additional sound sample; and
adjusting the threshold value based on the second sound sample.

11. The method of claim 8 , further comprising:
displaying an indication that the additional sound sample is similar to the first sound sample in response to determining that the particular acoustic feature is similar to the first acoustic feature.

12. The method of claim 1 , wherein each of the first sound sample and the second sound sample is indicative of at least one of a command to activate the electronic device or a command to control an application or a function in the electronic device.

13. The method of claim 9 , further comprising:
receiving input data associated with a speech input; and
recognizing, from the input data, at least one of a keyword or a user based on the sound detection model.

14. An electronic device comprising:
a sound sensor configured to generate a plurality of sound samples including a first sound sample and a second sound sample;
a sound sample verification unit configured to extract a first acoustic feature from the first sound sample and to extract a second acoustic feature from the second sound sample; and
an output unit configured to output a determination result in response to a determination by the sound sample verification unit that the second acoustic feature is not similar to the first acoustic feature, the determination result indicating whether the second sound sample is to be used in generating a sound detection model, the determination result including a request for a replacement sound sample.

15. The electronic device of claim 14 , further comprising:
a sound detection model generation unit configured to generate the sound detection model based on at least one of the plurality of sound samples in response to a determination by the sound sample verification unit that acoustic features of two or more of the plurality of sound samples are similar to one another; and
an input unit configured to receive an input indicative of at least one of the plurality of sound samples,
wherein the sound sample verification unit is further configured to:
determine a combined acoustic feature based on the two or more of the plurality of sound samples in response to a determination by the sound sample verification unit that the acoustic features of the two or more of the plurality of sound samples are similar to one another;
extract another acoustic feature from another sound sample received by the sound sensor; and
determine whether the other acoustic feature is similar to the combined acoustic feature.

16. A non-transitory computer-readable storage medium comprising instructions causing at least one processor of an electronic device to perform operations of:
receiving a first sound sample;
extracting a first acoustic feature from the first sound sample;
receiving a second sound sample;
extracting a second acoustic feature from the second sound sample; and
determining whether the second acoustic feature is similar to the first acoustic feature; and
in response to determining that the second acoustic feature is not similar to the first acoustic feature;
outputting a determination result indicating whether the second sound sample is to be used in generating a sound detection model;
receiving an additional sound sample in response to determining that the second acoustic feature is not similar to the first acoustic feature;
extracting a particular acoustic feature from the additional sound sample;
determining whether the particular acoustic feature is similar to the first acoustic feature; and
displaying an indication that the additional sound sample is similar to the first sound sample in response to determining that the particular acoustic feature is similar to the first acoustic feature.

17. The non-transitory computer-readable storage medium of claim 16 , wherein the sound detection model is generated without the second acoustic feature based on determining that the second acoustic feature is not similar to the first acoustic feature.

18. The non-transitory computer-readable storage medium of claim 16 , wherein the operations further comprise:
receiving a third sound sample;
extracting a third acoustic feature from the third sound sample; and
determining a combined acoustic feature based on the first acoustic feature and the third acoustic feature in response to determining

Assignees

• Qualcomm Inc

Inventors

• Moon Sunkuk
• Jin Minho
• Xia Haiying
• Huang Hesu
• Dale Warren Frederick

Classifications

• G10L15/02 Primary

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

• G10L2015/022

  Demisyllables, biphones or triphones being the recognition units · CPC title

• G10L15/22

  Procedures used during a speech recognition process, e.g. man-machine dialogue · CPC title

• G10L15/063 Primary

  Training · CPC title

• G10L2015/027

  Syllables being the recognition units · CPC title