Method and system for endpoint automatic detection of audio record

• US 9,330,667 B2
• Filed: 10/29/2010
• Issued: 05/03/2016
• Est. Priority Date: 10/29/2010
• Status: Active Grant

• Current Assignee: Iflytek Co., Ltd.
• Original Assignee: Iflytek Co., Ltd.
• Inventors: Wei, Si, Hu, Guoping, Hu, Yu, Liu, Qingfeng
• Primary Examiner(s): Adesanya, Olujimi
• Application Number:
  US13/878,818
• Publication Number:
  US 20130197911A1
• Time in Patent Office: 2,013 Days
• Field of Search: 704/210, 704/215, 704/248, 704/253
• US Class Current: 1/1
• CPC Class Codes: G10L 15/083 Recognition networks G10L15/142, G10L15/16 take precedence
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 25/87 Detection of discrete points within a voice signal

Abstract

A method and system for endpoint automatic detection of audio record is provided. The method comprises the following steps: acquiring a audio record text and affirming the text endpoint acoustic model for the audio record text; starting acquiring the audio record data of each frame in turn from the audio record start frame in the audio record data; affirming the characteristics acoustic model of the decoding optimal path for the acquired current frame of the audio record data; comparing the characteristics acoustic model of the decoding optimal path acquired from the current frame of the audio record data with the endpoint acoustic model to determine if they are the same; if yes, updating a mute duration threshold with a second time threshold, wherein the second time threshold is less than a first time threshold. This method can improve the recognizing efficiency of the audio record endpoint.

Claims

1. A method for detecting an endpoint of an audio record, comprising presetting a mute duration threshold as a first time threshold;

wherein the method further comprises;

obtaining an audio record text;

determining an acoustic model for a text endpoint of the audio record text; and

obtaining each frame of audio record data in turn starting from an audio record start frame of the audio record data;

determining a characteristics acoustic model of a decoding optimal path for an obtained current frame of the audio record data; and

determining that the characteristics acoustic model of the decoding optimal path for the current frame of the audio record data is the same as the acoustic model for the endpoint;

updating the mute duration threshold to a second time threshold, wherein the second time threshold is smaller than the first time threshold.

2. The method according to claim 1, wherein the determining an acoustic model for a text endpoint comprises:

generating a decoding network corresponding to the text according to the audio record text, and determining a last acoustic model of the decoding network as the acoustic model for the text endpoint.

3. The method according to claim 2, wherein the determining a characteristics acoustic model of a decoding optimal path for a current frame of the audio record data comprises:

extracting an MFCC characteristic corresponding to a preset acoustic model from the current frame of the audio record data to obtain the decoding optimal path for the current frame of the audio record data; and

determining a last acoustic model of the decoding optimal path for the current frame of the audio record data as the characteristics acoustic model of the decoding optimal path.

4. The method according to claim 1, further comprising:

retaining the mute duration threshold as the first time threshold if it is determined that the characteristics acoustic model of the decoding optimal path for the current frame of the audio record data is different from the acoustic model for the endpoint.

5. The method according to claim 1, wherein after one frame of audio record data is obtained, the method further comprises:

ending the audio record if it is determined that the obtained current frame of the audio record data is mute data and a current mute duration is larger than a current mute duration threshold.

6. The method according to claim 1, wherein before the obtaining each frame of audio record data, the method further comprises:

receiving the audio record data and determining the audio record start frame of the audio record data.

7. The method according to claim 6, wherein the determining the audio record start frame of the audio record data comprises:

determining in turn whether each frame of the audio record data is the mute data or non-mute data, and using a first frame of the non-mute data as the audio record start frame.

8. A system for detecting an endpoint of an audio record, wherein a mute duration threshold is preset as a first time threshold;

and the system further comprises;

a first determining unit adapted to obtain an audio record text and determine an acoustic model for a text endpoint of the audio record text;

a first obtaining unit adapted to obtain each frame of audio record data in turn starting from an audio record start frame of the audio record data;

a second determining unit adapted to determine a characteristics acoustic model of a decoding optimal path for an obtained current frame of the audio record data; and

a threshold determining unit adapted to update the mute duration threshold to the second time threshold if it is determined that the characteristics acoustic model of the decoding optimal path for the current frame of the audio record data is the same as an acoustic model for the endpoint, wherein the second time threshold is smaller than the first time threshold.

9. The system according to claim 8, wherein the first determining unit comprises:

an obtaining subunit adapted to obtain the audio record text;

a network establishing subunit adapted to establish a decoding network corresponding to the text according to the audio record text; and

a first characteristic determining subunit adapted to determine a last acoustic model of the decoding network as the acoustic model for the text endpoint.

10. The system according to claim 9, wherein the second determining unit comprises:

an extracting subunit adapted to extract an MFCC characteristic corresponding to a preset acoustic model from the current frame of the audio record data to obtain the decoding optimal path for the current frame of the audio record data; and

a second characteristic determining subunit adapted to determine a last acoustic model of the decoding optimal path for the current frame of the audio record data as the characteristics acoustic model of the decoding optimal path.

11. The system according to claim 8, wherein the threshold determining unit is further adapted to retain the mute duration threshold as the first time threshold if it is determined that the characteristics acoustic model of the decoding optimal path for the current frame of the audio record data is different from the acoustic model for the endpoint.

12. The system according to claim 8, further comprising:

an audio record control unit adapted to end the audio record if it is determined that the obtained current frame of the audio record data is mute data and a current mute duration is larger than a current mute duration threshold.

13. The system according to claim 8, further comprising:

a receiving unit adapted to receive the audio record data and determine the audio record start frame of the audio record data.

14. The system according to claim 13, wherein the receiving unit comprises:

a receiving subunit adapted to receive the audio record data; and

a start frame determining subunit adapted to determine in turn whether each frame of the audio record data is the mute data or non-mute data, and use a first frame of the non-mute data as the audio record start frame.