Method and apparatus for transcribing speech when a plurality of speakers are participating

• US 20030125940A1
• Filed: 01/02/2002
• Published: 07/03/2003
• Est. Priority Date: 01/02/2002
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Basson, Sara H., Faisman, Alexander, Kanevsky, Dimitri, Fairweather, Peter Gustav, Sorensen, Jeffery Scott
• Granted Patent: US 6,996,526 B2
• Time in Patent Office: Days
• US Class Current: 704/231
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems
• CPC Class Codes: G10L 17/00 Speaker identification or verification techniques

Abstract

A method and apparatus are disclosed for transcribing speech when a number of speakers are participating. A number of different speech recognition systems, each with a different speaker model, are executed in parallel. When the identity of all of the participating speakers is known and a speaker model is available for each participant, each speech recognition system employs a different speaker model suitable for a corresponding participant. Each speech recognition system decodes the speech and generates a corresponding confidence score. The decoded output having the highest confidence score is selected for presentation to a user. When all participating speakers are not known, or when there are too many participants to implement a unique speaker model for each participant, a speaker independent speech recognition system is employed together with a speaker specific speech recognition system. A controller selects between the decoded outputs of the speaker independent speech recognition system and the speaker specific speech recognition system based on information received from a speaker identification system and a speaker change detector.

Claims

1. A method for transcribing speech of a plurality of speakers, comprising:

providing said speech to a plurality of speech decoders, each of said decoders using a speaker model for one of said speakers and generating a confidence score for each decoded output; and

selecting a decoded output based on said confidence score.

2. The method of claim 1, further comprising the step of aligning each of said decoded outputs in time.

3. The method of claim 1, wherein one or more of said speech decoders are on a remote server.

4. The method of claim 1, further comprising the step of presenting said selected decoded output to a user.

5. The method of claim 1, further comprising the step of manually selecting an alternate decoded output if said assigned output is incorrect.

6. The method of claim 5, further comprising the step of adapting said selecting step based on said manual selection.

7. The method of claim 1, further comprising the step of presenting several decoded outputs to a user with an indication of said corresponding confidence score.

8. The method of claim 1, further comprising the step of presenting said decoded output as a string of words if said corresponding confidence score exceeds a certain threshold and as a string of phones if said corresponding confidence score is below a certain threshold.

9. The method of claim 1, further comprising the step of presenting said decoded output as a string of words for the decoded output having the highest confidence score and as phones or syllables for all other decoded outputs.

10. The method of claim 1, wherein said selecting step further comprises the step of determining if a decoded output includes an isolated word from a second speaker in a string of words from a first speaker.

11. A method for transcribing speech of a plurality of speakers, comprising:

providing said speech to a speaker independent speech recognition system and a speaker specific speech recognition system; and

decoding said speech using said speaker independent speech recognition system whenever the identity of the current speaker is unknown.

12. The method of claim 11, wherein said decoding step continues until a speaker identification system identifies an unknown speaker.

13. The method of claim 11, wherein one or more of said speaker independent speech recognition system and said speaker specific speech recognition system are on a remote server.

14. The method of claim 11, further comprising the step of presenting said selected decoded output to a user.

15. A method for transcribing speech of a plurality of speakers, comprising:

providing said speech to a speaker independent speech recognition system and a speaker specific speech recognition system; and

decoding said speech using said speaker specific speech recognition system with a speaker model for an identified speaker until there is a speaker change.

16. The method of claim 15, further comprising the step of decoding said speech using a speaker independent speech recognition system until the identity of a speaker is determined and the appropriate speaker model is loaded.

17. The method of claim 15, wherein one or more of said speaker independent speech recognition system and said speaker specific speech recognition system are on a remote server.

18. The method of claim 15, further comprising the step of presenting said selected decoded output to a user.

19. A system for transcribing speech of a plurality of speakers, comprising:

a memory that stores computer-readable code; and

a processor operatively coupled to said memory, said processor configured to implement said computer-readable code, said computer-readable code configured to;

provide said speech to a plurality of speech decoders, each of said decoders using a speaker model for one of said speakers and generating a confidence score for each decoded output; and

select a decoded output having a highest confidence score.

20. The system of claim 19, wherein said processor is further configured to align each of said decoded outputs in time.

21. The system of claim 19, wherein one or more of said speech decoders are on a remote server.

22. The system of claim 19, wherein said processor is further configured to present said selected decoded output to a user.

23. A system for transcribing speech of a plurality of speakers, comprising:

a memory that stores computer-readable code; and

a processor operatively coupled to said memory, said processor configured to implement said computer-readable code, said computer-readable code configured to;

provide said speech to a speaker independent speech recognition system and a speaker specific speech recognition system; and

decode said speech using said speaker independent speech recognition system whenever the identity of the current speaker is unknown.

24. The system of claim 23, wherein said processor performs said decoding until a speaker identification system identifies an unknown speaker.

25. The system of claim 23, wherein one or more of said speaker independent speech recognition system and said speaker specific speech recognition system are on a remote server.

26. The system of claim 23, wherein said processor is further configured to present said selected decoded output to a user.

27. A system for transcribing speech of a plurality of speakers, comprising:

a memory that stores computer-readable code; and

a processor operatively coupled to said memory, said processor configured to implement said computer-readable code, said computer-readable code configured to;

provide said speech to a speaker independent speech recognition system and a speaker specific speech recognition system; and

decode said speech using said speaker specific speech recognition system with a speaker model for an identified speaker until there is a speaker change.

28. The system of claim 27, wherein said processor is further configured to decode said speech using a speaker independent speech recognition system until the identity of a speaker is determined and the appropriate speaker model is loaded.

29. The system of claim 27, wherein one or more of said speaker independent speech recognition system and said speaker specific speech recognition system are on a remote server.

30. The system of claim 27, wherein said processor is further configured to present said selected decoded output to a user.

31. An article of manufacture for transcribing speech of a plurality of speakers, comprising:

a computer readable medium having computer readable code means embodied thereon, said computer readable program code means comprising;

a step to provide said speech to a plurality of speech decoders, each of said decoders using a speaker model for one of said speakers and generating a confidence score for each decoded output; and

a step to select a decoded output having a highest confidence score.

32. An article of manufacture for transcribing speech of a plurality of speakers, comprising:

a computer readable medium having computer readable code means embodied thereon, said computer readable program code means comprising;

a step to provide said speech to a speaker independent speech recognition system and a speaker specific speech recognition system; and

a step to decode said speech using said speaker independent speech recognition system whenever the identity of the current speaker is unknown.

33. An article of manufacture for transcribing speech of a plurality of speakers, comprising:

a computer readable medium having computer readable code means embodied thereon, said computer readable program code means comprising;

a step to provide said speech to a speaker independent speech recognition system and a speaker specific speech recognition system; and

a step to decode said speech using said speaker specific speech recognition system with a speaker model for an identified speaker until there is a speaker change.