DISTRIBUTED VOICE RECOGNITION SYSTEM AND METHOD

• US 20070011010A1
• Filed: 06/28/2006
• Published: 01/11/2007
• Est. Priority Date: 07/06/2005
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Lawrence, Stephen, Dow, Barry, Pickering, John
• Granted Patent: US 7,716,051 B2
• Time in Patent Office: Days
• US Class Current: 704/270.100
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications

Abstract

A distributed voice recognition system (500) and method employs principles of bottom-up (i.e., raw input) and top-down (i.e., prediction based on past experience) processing to perform client-side and server-side processing by (i) at the client-side, replacing application data by a phonotactic table (504); (ii) at the server-side, tracking separate confidence scores for matches against an acoustic model and comparison to a grammar; and (iii) at the server-side using a contention resolver (514) to weight the client-side and server-side results to establish a single output which represents the collaboration between client-side processing and server-side processing.

Claims

1. A voice recognition system distributed between a client-side and a server-side, comprising:

client-side means for processing received voice information to produce a possible match between the voice information and predetermined data;

server-side means for processing received voice information to produce a possible match between the voice information and predetermined data; and

server-side means for collaborating between the possible matches at the client-side and the server-side to produce a resultant match.

2. The system of claim 1 wherein the client-side means comprises an acoustic front end and a decoder, and the server-side means comprises an acoustic front end and a decoder.

3. The system of claim 2 wherein the decoder of the server-side means comprises means for producing a list of a plurality of best candidate matches based on confidence score.

4. The system of claim 1, wherein the client-side means comprises a phonotactic table holding the predetermined data.

5. The system of claim 1 wherein the server-side means comprises means for producing confidence scores for matches against an acoustic model and comparison to a grammar.

6. The system of claim 1 wherein the means for collaborating comprises contention resolution means for weighting the possible matches at the client-side and the server-side to establish a single output.

7. The system of claim 6 wherein the contention resolution means is arranged to validate the server-side possible match against the client-side possible match if the grammar confidence score falls below a rejection threshold.

8. The system of claim 7 wherein the contention resolution means is arranged further to suppress completely the server-side possible match if a confidence score for a given sound combination is also low for a client-side possible match.

9. A voice recognition method distributed between a client-side and a server-side, the method comprising:

performing client-side processing of received voice information to produce a possible match between the voice information and predetermined data;

performing server-side processing of received voice information to produce a possible match between the voice information and predetermined data; and

performing server-side collaboration between the possible matches at the client-side and the server-side to produce a resultant match.

10. The method of claim 9 wherein the step of performing client-side processing comprises performing acoustic front end processing and performing decoding, and the step of performing server-side processing comprises performing acoustic front end processing and performing decoding.

11. The method of claim 10 wherein the step of server-side decoding comprises producing a list of a plurality of best candidate matches based on confidence score.

12. The method of claim 9, wherein the step of performing client-side processing comprises using a phonotactic table holding the predetermined data.

13. The method of claim 9 wherein the step of performing server-side processing comprises producing confidence scores for matches against an acoustic model and comparison to a grammar.

14. The method of claim 9 wherein the step of performing server-side collaboration comprises weighting the possible matches at the client-side and the server-side to establish a single output.

15. The method of claim 14 wherein the step of performing server-side collaboration comprises validating the server-side possible match against the client-side possible match the grammar confidence score falls below a rejection threshold.

16. The method of claim 15 wherein the step of performing server-side collaboration further comprises suppressing completely the server-side possible match if a confidence score for a given sound combination is also low for a client-side possible match.

17. A computer program element stored on a data carrier and comprising computer program means for instructing the computer to perform substantially the method of any one of claims 9-16.