Method and device for generating an adapted reference for automatic speech recognition

• US 20020069053A1
• Filed: 11/06/2001
• Published: 06/06/2002
• Est. Priority Date: 11/07/2000
• Status: Active Grant

• Current Assignee: Optis Wireless Technology, LLC (Brevet Capital)
• Original Assignee: Telefonaktiebolaget LM Ericsson
• Inventors: Dobler, Stefan, Bruckner, Raymond, Kiessling, Andreas, Schleifer, Ralph
• Granted Patent: US 6,961,702 B2
• Time in Patent Office: Days
• US Class Current: 704/231
• CPC Class Codes: G10L 15/07 to the speaker
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering
• CPC Class Codes: G10L 2015/0635 updating or merging of old and new templates; Mean values; Weighting
• CPC Class Codes: G10L 2015/0636 Threshold criteria for the updating

Abstract

The invention relates to a method for generating an adapted reference for automatic speech recognition. In a first step, recognition is performed based on a spoken utterance and a recognition result which corresponds to a currently valid reference is obtained. In a second step, the currently valid reference is adapted in accordance with the utterance in order to create an adapted reference. In a third step, the adapted reference is assessed and it is decided if the adapted reference is used for further recognition.

Claims

1. A method for generating adapted references for automatic speech recognition, comprising:

performing recognition based on a spoken utterance and obtaining a recognition result which corresponds to a currently valid reference;

adapting the currently valid reference in accordance with the spoken utterance to generate an adapted reference; and

assessing the adapted reference and deciding if the adapted reference is to be used for further recognition.

2. The method of claim 1, wherein the adapted reference is assessed by determining a distance between the adapted reference and the currently valid reference.

3. The method of claim 2, wherein the adapted reference is assessed also based on a user behaviour.

4. The method of claim 2, wherein the adapted reference is assessed by further determining distances between the adapted reference and currently valid references which do not correspond to the recognition result.

5. The method of claim 2, further comprising analyzing a histogram of previously determined distances in order to obtain one or more parameters for deciding if the adapted reference is used for further recognition.

6. The method of claim 2, wherein the distance is determined by dynamic programming.

7. The method claim 1, wherein the adapted reference is assessed based on a user behaviour.

8. The method of claim 1, further comprising substituting the currently valid reference by the adapted reference.

9. The method of claim 1, further comprising storing the adapted reference in addition to the currently valid reference.

10. The method claim 1, wherein the adapted reference is created only when a user behaviour indicates that the recognition result is correct.

11. A method for generating adapted references to be used for automatic speech recognition, comprising:

receiving a spoken utterance and matching a pattern of the spoken utterance with corresponding patterns of a plurality of currently valid references to obtain the currently valid reference which corresponds to the spoken utterance;

adapting the pattern of the currently valid reference corresponding to the spoken utterance by shifting the pattern of the currently valid reference towards the pattern of the spoken utterance to generate an adapted reference;

assessing the adapted reference by determining the distance between the adapted reference and at least one of the currently valid references; and

deciding on the basis of the one or more distances if the adapted reference is to be used for recognizing a subsequent spoken utterance.

13. The computer program product with program code means according to claim 12 stored on a computer-readable recording medium.

15. The device of claim 14, further comprising a first memory and a second memory for storing a first and a second set of references, respectively.

16. The device of claim 15, further comprising a selection unit which sets a pointer to either the first or the second memory depending on whether the first or the second set of references are to be used for recognition.

17. The device of claim 15, further comprising a third memory for storing a set of mother references.

18. The device of claim 14, further comprising a selection unit for setting pointers which allow to determine all references currently valid for recognition of spoken utterances

12. A computer program product with program code means for performing the steps of:

performing recognition based on a spoken utterance and obtaining a recognition result which corresponds to a currently valid reference;

adapting the currently valid reference in accordance with the spoken utterance to generate an adapted reference; and

assessing the adapted reference and deciding if the adapted reference is to be used for further recognition when the product is executed in a computing unit.

14. A device for generating adapted references for automatic speech recognition, comprising:

a speech recognizer for performing recognition based on a spoken utterance and for obtaining a recognition result which corresponds to a currently valid reference;

an adaption unit which adapts the currently valid reference in accordance with the spoken utterance to generate an adapted reference; and

an assessing unit which assesses the adapted reference and which decides if the adapted reference is used for further recognition.

19. A device for generating adapted references for automatic speech recognition, comprising:

a speech recognizer which matches a pattern of a spoken utterance with patterns of a plurality of currently valid references to identify a currently valid reference which corresponds to the spoken utterance;

an adaption unit which shifts the pattern of the currently valid reference which corresponds to the spoken utterance towards the pattern of the spoken utterance to generate an adapted reference; and

an assessing unit which assesses the adapted reference and which decides if the adapted reference is to be used by the speech recognizer for recognizing a subsequent spoken utterance.