Method and apparatus to improve accuracy of mobile speech-enabled services

• US 7,174,298 B2
• Filed: 06/24/2002
• Issued: 02/06/2007
• Est. Priority Date: 06/24/2002
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Sharma, Sangita R
• Primary Examiner(s): Chawan; Vijay B.
• Application Number:
  US10/178,126
• Publication Number:
  US 20030236665A1
• Time in Patent Office: 1,688 Days
• Field of Search: 704/231, 704/251, 704/260, 704/235, 704/275, 704/240, 704/256, 704/270.1, 704/243, 704/244, 704/258, 379/88.03, 379/406.06, 379/406.08, 455/564, 455/556.1
• US Class Current: 704/270.1
• CPC Class Codes: G10L 15/065 Adaptation
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications

Abstract

A speech recognition system includes a user profile to store acoustic data and a corresponding text transcript. A speech recognition (“SR”) server downloads the acoustic data and the corresponding text transcript that are stored in the user profile. A speech recognition engine is included to adapt an acoustic model based on the acoustic data.

Claims

1. A speech recognition system, comprising:

a mobile communication device;

a user profile in the mobile communication device, the user profile to store acoustic data and a corresponding text transcript;

a speech recognition (“

SR”

) server to download the acoustic data and the corresponding text transcript that are stored in the user profile; and

a speech recognition engine to adapt an acoustic model based on the acoustic data.

2. The speech recognition system according to claim 1, wherein the acoustic data is a digitized sample of a user'"'"'s voice.

3. The speech recognition system according to claim 1, wherein the speech recognition server records a background environment received from a mobile communication device.

4. The speech recognition system according to claim 3, wherein the speech recognition server combines the background environment and the acoustic data, and the speech recognition engine adapts the acoustic model based on the acoustic data and the background environment.

5. The speech recognition system according to claim 1, wherein the acoustic model is user-independent.

6. A method of adapting an acoustic model to improve recognition accuracy, comprising:

loading accoustic data and a corresponding text transcript from a communication device to a speech recognition (SR) server, the acoustic data obtained from speech by a person; and

adapting an acoustic model for the SR server based at least in part on the acoustic data from the communication device.

7. The method according to claim 6, wherein the acoustic data comprises a digitized sample of the person'"'"'s voice.

8. The method according to claim 6, further comprising:

recording, at the SR server, a background environment received from the mobile communication device; and

adapting the acoustic model, based at least in part on the acoustic data and the recorded background environment.

9. The method according to claim 6, further comprising:

receiving the acoustic model from a user profile within the mobile communication device.

10. The method according to claim 6, wherein the acoustic model for the SR server comprises a user-independent acoustic model.

11. The method according to claim 6, further comprising:

using the revised acoustic model to process subsequent voice queries from the person.

12. An article comprising:

a storage medium having stored thereon instructions that when executed by a machine result in the following;

loading acoustic data and a corresponding text transcript from a user profile in a communication device, and

adapting an acoustic model for a speech recognition (SR) engine, based at least in part on the acoustic data.

13. The article according to claim 12, wherein the acoustic data is a digitized sample of the user'"'"'s voice.

14. The article according to claim 12, wherein a SR server records a background environment received from the mobile communication device.

15. The article according to claim 14, wherein the SR server combines the background environment and the acoustic data, and the SR engine adapts the acoustic model based on the acoustic data and the background environment.

16. The article according to claim 12, wherein a SR server performs downloading the acoustic data and the corresponding text transcript.

17. The article according to claim 12, wherein a SR engine performs adapting the acoustic model based on the acoustic data.

18. The article according to claim 12, wherein the acoustic model is user-independent.

19. The article according to claim 12, wherein execution of the instructions further results in using a revised acoustic model to process subsequent voice queries of the user.

20. A processing system, comprising:

a user profile to store user-dependent speech data for adapting a speech recognition model, the user-dependent speech data comprising a text transcript and corresponding acoustic data obtained from speech by a person;

a transmitter to transmit the user-dependent speech data from the user profile in the processing system to a speech recognition (“

SR”

) server, to allow the SR server to adapt a speaker-independent acoustic model into a revised acoustic model, based at least in part on the acoustic data from the user-dependent speech data; and

the transmitter to transmit voice queries from the person to the SR server, to be translated into text using the revised acoustic model.

21. A processing system according to claim 20, wherein the processing system comprises a mobile communication device.

22. A processing system according to claim 20, wherein the acoustic data comprises a digitized sample of the person'"'"'s voice.