Method and system for collaborative speech recognition for small-area network

• US 20020143532A1
• Filed: 04/02/2001
• Published: 10/03/2002
• Est. Priority Date: 04/02/2001
• Status: Active Grant

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Winarski, Daniel James, Wong, Tin-Lup, McLean, James Gordon
• Granted Patent: US 6,885,989 B2
• Time in Patent Office: Days
• US Class Current: 704/235
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications

Abstract

The present invention provides a method and system for collaborative speech recognition in a network. The method includes: capturing speech as at least one audio stream by at least one capturing device; producing a plurality of text streams from the at least one audio stream by at least one recognition device; and determining a best recognized text stream from the plurality of text streams. The present invention allows multiple computing devices connecting to a network, such as a Small Area Network (SAN), to collaborate on a speech recognition task. The devices are able to share or exchange audio data and determine the best quality audio. The devices are also able to share text results from the speech recognition task and the best result from the speech recognition task. This increases the efficiency of the speech recognition process and the quality of the final text stream.

Claims

1. A method for collaborative speech recognition in a network, comprising the steps of:

(a) capturing speech as at least one audio stream by at least one capturing device;

(b) producing a plurality of text streams from the at least one audio stream by at least one recognition device; and

(c) determining a best recognized text stream from the plurality of text streams.

2. The method of claim 1, wherein the capturing step (a) further comprises:

(a1) determining a best quality audio stream from a plurality of audio streams.

3. The method of claim 2, further comprising:

(a2) routing the best quality audio stream to a plurality of recognition devices.

4. The method of claim 1, wherein the producing step (b) comprises:

(b1) producing the plurality of text streams from a best quality audio stream of a plurality of audio streams.

5. The method of claim 1, where in the determining step (c) comprises:

(c1) assessing agreement between the plurality of text streams;

(c2) determining an interim best recognized text stream; and

(c3) correcting the interim best recognized text stream to obtain the best recognized text stream.

6. The method of claim 1, further comprising:

(d) storing the best recognized text stream in a repository.

7. The method of claim 1, wherein the at least one capturing device and the at least one recognition device is the same device.

8. The method of claim 1, wherein a capturing device is a device which comprises speech recognition technology.

9. The method of claim 1, wherein a recognition device is a device which comprises speech recognition technology.

10. A method for collaborative speech recognition in a net work, comprising the steps of:

(a) capturing speech as a plurality of audio streams by a plurality of capturing devices;

(b) determining a best quality audio stream from the plurality of audio streams;

(c) producing a plurality of text streams from the best quality audio stream by at least one recognition device; and

(d) determining a best recognized text stream from the plurality of text streams.

11. The method of claim 10, wherein the determining step (b) further comprises:

(b1) routing the best quality audio stream to the at least one recognition device.

12. The method of claim 10, wherein the determining step (d) comprises:

(d1) assessing agreement between the plurality of text streams;

(d2) determining an interim best recognized text stream; and

(d3) correcting the interim best recognized text stream to obtain the best recognized text stream.

13. The method of claim 10, further comprising:

(e) storing the best recognized text stream in a repository.

14. The method of claim 10, wherein the plurality of capturing devices and the at least one recognition device are the same devices.

15. The method of claim 10, wherein a capturing device is a device which comprises speech recognition technology.

16. The method of claim 10, wherein a recognition device is a device which comprises speech recognition technology.

18. The medium of claim 17, wherein the capturing instruction (a) further comprises instructions for:

(a1) determining a best quality audio stream from a plurality of audio streams.

19. The medium of claim 18, further comprising instructions for:

(a2) routing the best quality audio stream to a plurality of recognition devices.

20. The medium of claim 17, wherein the producing instruction (b) comprises instructions for:

(b1) producing the plurality of text streams from a best quality audio stream of a plurality of audio streams.

21. The medium of claim 17, wherein the determining instruction (c) comprises instructions for:

(c1) assessing agreement between the plurality of text streams;

(c2) determining an interim best recognized text stream; and

(c3) correcting the interim best recognized text stream to obtain the best recognized text stream.

22. The medium of claim 17, further comprising instructions for:

(d) storing the best recognized text stream in a repository.

23. The medium of claim 17, wherein the at least one capturing device and the at least one recognition device is the same device.

24. The medium of claim 17, wherein a capturing device is a device which comprises speech recognition technology.

25. The medium of claim 17, wherein a recognition device is a device which comprises speech recognition technology.

26. The medium of claim 17, wherein a capturing device is a device which comprises speech recognition technology.

27. The medium of claim 17, wherein a recognition device is a device which comprises speech recognition technology.

29. The medium of claim 28, wherein the determining instruction (b) further comprises instructions for:

(b1) routing the best quality audio stream to the at least one recognition device.

30. The medium of claim 28, wherein the determining instruction (d) comprises instructions for:

(d1) assessing agreement between the plurality of text streams;

(d2) determining an interim best recognized text stream; and

(d3) correcting the interim best recognized text stream to obtain the best recognized text stream.

31. The medium of claim 28, further comprising instructions for:

(e) storing the best recognized text stream in a repository.

32. The medium of claim 28, wherein the plurality of capturing devices and the at least one recognition device are the same devices.

33. The medium of claim 28, wherein a capturing device is a device which comprises speech recognition technology.

34. The medium of claim 28, wherein a recognition device is a device which comprises speech recognition technology.

36. The system of claim 35, further comprising:

a plurality of participating devices, wherein each of the plurality of participating devices is capable of applying a correction to an interim best recognized text stream to obtain the best recognized text stream.

37. The system of claim 35, further comprising:

a repository for storing the best recognized text stream.

17. A computer readable medium with program instructions for providing collaborative speech recognition in a network, the instructions for:

(a) capturing speech as at least one audio stream by at least one capturing device;

(b) producing a plurality of text streams from the at least one audio stream by at least one recognition device; and

(c) determining a best recognized text stream from the plurality of text streams.

28. A computer readable medium with program instructions for providing collaborative speech recognition in a network, the instructions for:

(a) capturing speech as a plurality of audio streams by a plurality of capturing devices;

(b) determining a best quality audio stream from the plurality of audio streams;

(c) producing a plurality of text streams from the best quality audio stream by at least one recognition device; and

(d) determining a best recognized text stream from the plurality of text streams.

35. A system, comprising:

at least one capturing device, wherein the at least one capturing device comprises speech capture technology, wherein the at least one capturing device is capable of capturing at least one audio stream;

at least one recognition device, wherein the at least one recognition device comprises speech recognition technology, wherein the at least one recognition device is capable of producing a plurality of text streams from the at least one audio stream; and

a designated arbitration device, wherein the designated arbitration device is capable of determining a best recognized text stream from the plurality of text streams.