Speech recognition system which returns recognition results as a reconstructed language model with attached data values

• US 5,689,617 A
• Filed: 03/14/1995
• Issued: 11/18/1997
• Est. Priority Date: 03/14/1995
• Status: Expired due to Term

• Current Assignee: Apple Computer Incorporated (Apple Inc.)
• Original Assignee: Apple Computer Incorporated (Apple Inc.)
• Inventors: Reeves, Arthur Arlo, Rodarmer, Kurt W., Pallakoff, Matthew G.
• Primary Examiner(s): TSANG, FAN S
• Application Number:
  US08/404,472
• Time in Patent Office: 980 Days
• Field of Search: 395/2.52-2.65, 395/2.66
• US Class Current: 704/255
• CPC Class Codes: G10L 15/193 Formal grammars, e.g. finite state automata, context free grammars or word networks

Abstract

A speech recognition system operating on a computer system, where the speech recognition system uses a language model with embedded structure and attached data values, and the speech recognition system returns recognition results as a reconstructed language model with embedded structure and attached data values. An application program can receive and traverse the embedded structure of the recognition results and use the attached data values to improve the speed and accuracy of interpretation of the speech recognition results.

Claims

1. A method of speech recognition for a speech recognition system operating on a computer system, comprising the steps of:

the speech recognition system generating a language model with embedded structure and attached data values;

the speech recognition system receiving a sound signal;

the speech recognition system processing the sound signal to extract speech features;

the speech recognition system comparing and matching the speech features to the language model for generating a recognition result having the embedded structure and attached data values of the language model for those portions of the language model matched to the speech features.

2. A method of speech recognition as in claim 1, further comprising the steps of an application program receiving the recognition result and traversing the embedded structure and attached data values of the recognition result.

3. A method of speech recognition for a speech recognition system operating on a computer system which receives a sound signal and produces a recognition result, comprising the steps of:

the speech recognition system generating a language model having an embedded structure in which language model elements are held, and having data values attached to the elements of the language model;

the speech recognition system receiving the sound signal;

the speech recognition system processing the sound signal to extract speech features;

the speech recognition system comparing the speech features to the language model to identify portions of the language model matched by the speech features; and

the speech recognition system generating the recognition result, by including within the recognition result the embedded structure and attached data values of the language model for the portions of the language model matched by the speech features.

4. A method of speech recognition as in claim 3, further comprising the steps of an application program receiving the recognition result and traversing the embedded structure and attached data values of the recognition result.

5. A method of speech recognition for a speech recognition system operating on a computer system, comprising the steps of:

an application program generating and embedding a structure and attached data values to elements of a language model;

a speech recognizer receiving and storing the language model;

the speech recognition system receiving a sound signal;

the speech recognition system processing the sound signal to extract speech features;

the speech recognizer receiving the speech features and comparing the speech features to the language model to identify portions of the language model which match the speech features;

the speech recognizer generating a recognition result by placing the embedded structure and attached data values of the portions of the language model which match the speech features into the recognition result; and

returning the recognition result from the speech recognizer to the application program.

6. A method of speech recognition as in claim 5, further comprising the steps of the application program receiving the recognition result and traversing the embedded structure and attached data values of the recognition result to determine a next action of the application program.

7. A speech recognition system operating on a computer system, comprising:

a speech feature extractor adapted for receiving a digitized sound signal and extracting speech features;

a language model with embedded structure and attached data values;

a speech recognizer for comparing the speech features to the language model to identify a matched portion of the language model, the speech recognizer further adapted to generate a recognition result comprising the embedded structure and attached data values of the matched portion of the language model; and

an application program for receiving the recognition result.

8. A speech recognition system operating on a computer system for receiving a sound signal and producing a recognition result, comprising:

a speech feature extractor adapted for receiving the sound signal and generating speech features;

a language model builder for generating a language model having language model elements and attached data values;

a speech recognizer for comparing the speech features to the language model to identify matched language model elements and attached data values;

the speech recognizer generating the recognition result of the matched language model elements and attached data values.

9. A speech recognition system as in claim 8, further comprising an application program for receiving the recognition result.