Adaptation of symbols

• US 7,676,366 B2
• Filed: 01/13/2003
• Issued: 03/09/2010
• Est. Priority Date: 01/13/2003
• Status: Expired due to Fees

• Current Assignee: ART Advanced Recognition Technologies, Inc. (Microsoft Corporation)
• Original Assignee: ART Advanced Recognition Technologies, Inc. (Microsoft Corporation)
• Inventors: Mochary, Ran, El-Hay, Tal, Yadid, Tal, Solomon, Sasi, Bartur, Itamar
• Primary Examiner(s): Armstrong; Angela A.
• Application Number:
  US10/340,841
• Publication Number:
  US 20040138893A1
• Time in Patent Office: 2,612 Days
• Field of Search: 704/244, 704/277, 704/235, 704/251, 704/275
• US Class Current: 704/251
• CPC Class Codes: G10L 15/04 Segmentation; Word boundary detection

Abstract

A speaker adaptation system and method for speech models of symbols displays a multi-word symbol to be spoken as a symbol. The supervised adaptation system and method has unsupervised adaptation for multi-word symbols, limited to the set of words associated with each multi-word symbol.

Claims

1. A speaker adaptation system comprising:

a storage unit which stores at least reference speech models of symbols; and

a network generator in communication with said storage unit which generates a network of models for a speaker adaptation session corresponding to a series of symbols presentable to a user during said session, said network of models comprising;

multiple models for at least one multi-word symbol and models from more than one language; and

a model adapter which adapts at least a portion of said reference speech models from at least a portion of speech from said user.

2. A system according to claim 1 and also comprising a second model adapter which adapts models not adapted by said first model adapter using at least a portion of otherwise available models.

3. A speaker adaptation system comprising:

a symbol string generator which generates a series of symbols presentable to a user during a speaker adaptation session, where a multi-word symbol is presented as a symbol; and

an adaptation network generator which generates a network of models corresponding to said series of symbols, said network of models comprising models from more than one language.

4. A system according to claim 3 and also comprising a recognizer which matches a speech signal of a user saying said series of symbols with a portion of models in said network and a model adapter which adapts at least said portion of models using at least a portion of said speech signal.

5. A system according to claim 4 and also comprising a second model adapter which adapts models not adapted by said first model adapter using at least a portion of otherwise available models.

6. A speaker adaptation system comprising:

means to display symbols to a user during a speaker adaptation session, wherein a multi-word symbol to be spoken is displayable as a symbol; and

an adaptation network generator which generates a network of models corresponding to a series of symbols to be said, said network of models comprising models from more than one language.

7. A system according to claim 6 and also comprising a recognizer which matches a speech signal of a user saying said series of symbols with a portion of models in said network and a model adapter which adapts at least said portion of models using at least a portion of said speech signal.

8. A system according to claim 7 and also comprising a second model adapter which adapts models not adapted by said first model adapter using at least a portion of otherwise available models.

9. A speaker adaptation system for speech models of a set of symbols presentable to a user during a speaker adaptation session, the system comprising:

an unsupervised adaptation functionality for multi-word symbols of said set of symbols, limited to a set of words associated with each multi-word symbol;

wherein said set of words includes words from more than one language;

a supervised adaptation functionality for single word symbols of said set of symbols; and

a model adapter which adapts at least a portion of said speech models from at least a portion of speech from said user.

10. A system according to claim 9 and comprising a symbol string generator which generates a series of symbols to be said, where a multi-word symbol is presented as a symbol.

11. A system according to claim 9 and comprising an adaptation network generator which generates a network of models corresponding to a series of symbols to be said, said network of models comprising multiple models for at least one multi-word symbol.

12. A system according to claim 9 and also comprising a model adapter which adapts models not adapted by said functionalities using at least a portion of otherwise available models.

13. A computer-implemented method in which a computer system initiates execution of software instructions stored in memory for speaker adaptation, the computer-implemented method comprising:

generating a network of reference speech models corresponding to a series of symbols presentable to a user during a speaker adaptation session, said network of models comprising;

multiple models for at least one multi-word symbol and models from more than one language;

storing, in a storage unit, said network of reference speech models corresponding to said series of symbols; and

adapting at least a portion of said reference speech models from at least a portion of speech from said user.

14. A computer-implemented method according to claim 13 and also comprising adapting models not adapted in said first step of adapting using at least a portion of otherwise available models.

15. A computer-implemented method in which a computer system initiates execution of software instructions stored in memory for speaker adaptation, the computer-implemented method comprising:

generating a series of symbols to be presentable to a user during a speaker adaptation session, where a multi-word symbol is presented as a symbol;

generating a network of models corresponding to said series of symbols, said network of models comprising models from more than one language; and

storing, in a storage unit, said network of models corresponding to said series of symbols.

16. A computer-implemented method according to claim 15 and also comprising matching a speech signal of a user saying said series of symbols with a portion of models in said network and adapting at least said portion of models using at least a portion of said speech signal.

17. A computer-implemented method according to claim 16 and also comprising adapting models not adapted in said first step of adapting using at least a portion of otherwise available models.