Method for updating voiceprint feature model and terminal

• US 9,685,161 B2
• Filed: 12/30/2014
• Issued: 06/20/2017
• Est. Priority Date: 07/09/2012
• Status: Active Grant

• Current Assignee: Huawei Device Company Limited (Huawei Investment & Holding Co., Ltd.)
• Original Assignee: Huawei Device Company Limited (Huawei Investment & Holding Co., Ltd.)
• Inventors: Lu, Ting
• Primary Examiner(s): Chawan, Vijay B
• Application Number:
  US14/585,486
• Publication Number:
  US 20150112680A1
• Time in Patent Office: 903 Days
• Field of Search: 704246, 704273, 704245, 704250, 704244, 379 8802, 379 8801, 379 8804
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/07 to the speaker
• CPC Class Codes: G10L 15/142 Hidden Markov Models [HMMs]
• CPC Class Codes: G10L 17/02 Preprocessing operations, e.g. segment selection; Pattern representation or modelling, e.g. based on linear discriminant analysis [LDA] or principal components; Feature selection or extraction
• CPC Class Codes: G10L 17/04 Training, enrolment or model building
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering

Abstract

A method for updating a voiceprint feature model and a terminal are provided that are applicable to the field of voice recognition technologies. The method includes: obtaining an original audio stream including at least one speaker; obtaining a respective audio stream of each speaker of the at least one speaker in the original audio stream according to a preset speaker segmentation and clustering algorithm; separately matching the respective audio stream of each speaker of the at least one speaker with an original voiceprint feature model, to obtain a successfully matched audio stream; and using the successfully matched audio stream as an additional audio stream training sample for generating the original voiceprint feature model, and updating the original voiceprint feature model.

Claims

1. A method for updating a voiceprint feature model, comprising:

obtaining an original audio stream comprising at least one speaker;

obtaining a respective audio stream of each speaker of the at least one speaker in the original audio stream according to a preset speaker segmentation and clustering algorithm;

separately matching the respective audio stream of each speaker of the at least one speaker with an original voiceprint feature model to obtain a successfully matched audio stream;

using the successfully matched audio stream as an additional audio stream training sample for generating the original voiceprint feature model; and

updating the original voiceprint feature model to improve a voice recognition capability of a computing device that uses the original voiceprint feature model to identify the at least one speaker.

2. The method according to claim 1, wherein before obtaining the original audio stream comprising the at least one speaker, the method further comprises establishing the original voiceprint feature model according to a preset audio stream training sample.

3. The method according to claim 2, wherein obtaining the respective audio stream of each speaker of the at least one speaker in the original audio stream according to the preset speaker segmentation and clustering algorithm comprises:

segmenting the original audio stream into a plurality of audio clips according to a preset speaker segmentation algorithm, wherein each audio clip of the plurality of audio clips comprises only audio information of a same speaker of the at least one speaker; and

clustering, according to a preset speaker clustering algorithm, the audio clips that comprise only the same speaker of the at least one speaker, to generate an audio stream that comprises only the audio information of the same speaker of the at least one speaker.

4. The method according to claim 3, wherein separately matching the respective audio stream of each speaker of the at least one speaker with the original voiceprint feature model to obtain the successfully matched audio stream comprises:

obtaining a matching degree between the audio stream of each speaker of the at least one speaker and the original voiceprint feature model according to the audio stream of each speaker of the at least one speaker and the original voiceprint feature model; and

selecting an audio stream corresponding to a matching degree that is the highest and is greater than a preset matching threshold as the successfully matched audio stream.

5. The method according to claim 2, wherein separately matching the respective audio stream of each speaker of the at least one speaker with the original voiceprint feature model to obtain the successfully matched audio stream comprises:

obtaining a matching degree between the audio stream of each speaker of the at least one speaker and the original voiceprint feature model according to the audio stream of each speaker of the at least one speaker and the original voiceprint feature model; and

selecting an audio stream corresponding to a matching degree that is the highest and is greater than a preset matching threshold as the successfully matched audio stream.

6. The method according to claim 1, wherein obtaining the respective audio stream of each speaker of the at least one speaker in the original audio stream according to the preset speaker segmentation and clustering algorithm comprises:

segmenting the original audio stream into a plurality of audio clips according to a preset speaker segmentation algorithm, wherein each audio clip of the plurality of audio clips comprises only audio information of a same speaker of the at least one speaker; and

clustering, according to a preset speaker clustering algorithm, the audio clips that comprise only the same speaker of the at least one speaker to generate an audio stream that comprises only the audio information of the same speaker of the at least one speaker.

7. The method according to claim 6, wherein separately matching the respective audio stream of each speaker of the at least one speaker with the original voiceprint feature model to obtain the successfully matched audio stream comprises:

obtaining a matching degree between the audio stream of each speaker of the at least one speaker and the original voiceprint feature model according to the audio stream of each speaker of the at least one speaker and the original voiceprint feature model; and

selecting an audio stream corresponding to a matching degree that is the highest and is greater than a preset matching threshold as the successfully matched audio stream.

8. The method according to claim 1, wherein separately matching the respective audio stream of each speaker of the at least one speaker with the original voiceprint feature model to obtain a successfully matched audio stream comprises:

obtaining a matching degree between the audio stream of each speaker of the at least one speaker and the original voiceprint feature model according to the audio stream of each speaker of the at least one speaker and the original voiceprint feature model; and

selecting an audio stream corresponding to a matching degree that is the highest and is greater than a preset matching threshold as the successfully matched audio stream.

9. The method according to claim 1, wherein using the successfully matched audio stream as the additional audio stream training sample for generating the original voiceprint feature model and updating the original voiceprint feature model comprises:

generating a corrected voiceprint feature model according to the successfully matched audio stream and the preset audio stream training sample, wherein the preset audio stream training sample is an audio stream for generating the original voiceprint feature model; and

updating the original voiceprint feature model to the corrected voiceprint feature model.

10. The method according to claim 1, further comprising unlocking a screen of a mobile phone based upon matching the original voiceprint feature model.

11. A terminal, comprising:

a non-transitory computer readable medium having instructions stored thereon; and

a computer processor coupled to the non-transitory computer readable medium and configured to execute the instructions to;

obtain an original audio stream comprising at least one speaker;

obtain a respective audio stream of each speaker of the at least one speaker in the original audio stream according to a preset speaker segmentation and clustering algorithm;

separately match the respective audio stream of each speaker of the at least one speaker with an original voiceprint feature model, to obtain a successfully matched audio stream;

use the successfully matched audio stream as an additional audio stream training sample for generating the original voiceprint feature model; and

update the original voiceprint feature model to improve a voice recognition capability of a computing device that uses the original voiceprint feature model to identify the at least one speaker.

12. The terminal according to claim 11, wherein the computer processor is further configured to execute the instructions to:

obtain a preset audio stream training sample; and

establish the original voiceprint feature model according to the preset audio stream training sample.

13. The terminal according to claim 12, wherein the computer processor is further configured to execute the instructions to:

segment the original audio stream into a plurality of audio clips according to a preset speaker segmentation algorithm, wherein each audio clip of the plurality of audio clips comprises only audio information of a same speaker of the at least one speaker; and

cluster, according to a preset speaker clustering algorithm, the audio clips that comprise only the same speaker of the at least one speaker, to generate an audio stream that comprises only the audio information of the same speaker of the at least one speaker.

14. The terminal according to claim 13, wherein the computer processor is further configured to execute the instructions to:

obtain a matching degree between the audio stream of each speaker of the at least one speaker and the original voiceprint feature model according to the audio stream of each speaker of the at least one speaker and the original voiceprint feature model; and

select an audio stream corresponding to a matching degree that is the highest and is greater than a preset matching threshold as the successfully matched audio stream.

15. The terminal according to claim 11, wherein the computer processor is further configured to execute the instructions to:

segment the original audio stream into a plurality of audio clips according to a preset speaker segmentation algorithm, wherein each audio clip of the plurality of audio clips comprises only audio information of a same speaker of the at least one speaker; and

cluster, according to a preset speaker clustering algorithm, the audio clips that comprise only the same speaker of the at least one speaker to generate an audio stream that comprises only the audio information of the same speaker of the at least one speaker.

16. The terminal according to claim 15, wherein the computer processor is further configured to execute the instructions to:

obtain a matching degree between the audio stream of each speaker of the at feast one speaker and the original voiceprint feature model according to the audio stream of each speaker of the at least one speaker and the original voiceprint feature model; and

select an audio stream corresponding to a matching degree that is the highest and is greater than a preset matching threshold as the successfully matched audio stream.

17. The terminal according to claim 11, wherein the computer processor is further configured to execute the instructions to:

obtain a matching degree between the audio stream of each speaker of the at least one speaker and the original voiceprint feature model according to the audio stream of each speaker of the at least one speaker and the original voiceprint feature model; and

select an audio stream corresponding to a matching degree that is the highest and is greater than a preset matching threshold as the successfully matched audio stream.

18. The terminal according to claim 12, wherein the computer processor is further configured to execute the instructions to:

obtain a matching degree between the audio stream of each speaker of the at least one speaker and the original voiceprint feature model according to the audio stream of each speaker of the at least one speaker and the original voiceprint feature model; and

select an audio stream corresponding to a matching degree that is the highest and is greater than a preset matching threshold as the successfully matched audio stream.

19. The terminal according to claim 11, wherein the computer processor is further configured to execute the instructions to:

generate a corrected voiceprint feature model according to the successfully matched audio stream and the preset audio stream training sample; and

update the original voiceprint feature model to the corrected voiceprint feature model.

20. The terminal according to claim 11, wherein the computer processor is further configured to execute the instructions to unlock a screen of a mobile phone based upon matching the original voiceprint feature model.