User adaptable speech recognition system

• US 5,684,924 A
• Filed: 05/19/1995
• Issued: 11/04/1997
• Est. Priority Date: 05/19/1995
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: Kurzweil Applied Intelligence, Inc. (Intel Corporation)
• Inventors: Teel, Mary-Marshall, Avery, Susan Rousmaniere, Stanley, Barbara Ann, Sejnoha, Vladimir
• Primary Examiner(s): Zele, Krista M.
• Assistant Examiner(s): Wolinsky, Scott
• Application Number:
  US08/444,391
• Time in Patent Office: 900 Days
• Field of Search: 395/2.6, 395/2.49, 395/2.47, 395/2.61, 395/2.39, 395/2.45, 395/2.64, 395/2.5, 395/2.31, 395/2.52, 395/2.53, 395/2.42
• US Class Current: 704/233
• CPC Class Codes: G10L 15/06 Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice G10L15/14 takes precedence
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 2015/0631 Creating reference templates; Clustering
• CPC Class Codes: G10L 2015/226 using non-speech characteristics

Abstract

A speech recognition system is disclosed which comprises a core speech recognition program and a plurality of utility program modules for adjusting various recognition parameters such as gain, sensitivity and acceptance threshold and for improving the training of word models. The system further provides a decision tree and utility controlling program module which can be invoked by a user confronting problems during the running of the core program. The system utilizes user input to traverse the decision tree and to initiate appropriate ones of the utility program modules thereby to alter the on-going behavior of the core recognition program.

Claims

1. A speech recognition system comprising:

input means for receiving an input utterance from a user and receiving also background noise;

a core speech recognizer which compares an input utterance with a multiplicity of trained models representing words and generates scores for each of at least some of said models, which scores represent the degree of match of the respective model with the utterance,

said core speech recognizer having a plurality of adjustable parameters including a sensitivity parameter which controls a discrimination level defining sound levels accepted as an utterance rather than background noise;

a first utility program module which responds to user input to change said sensitivity parameter;

a training utility program module which responds to user input to further train a selected word model on the basis of a particular user'"'"'s voice characteristics;

a discrimination improving utility program module which responds to user input identifying a desired word and an intruder word to improve the differentiation in match between the respective models representing the desired and intruder words; and

a decision tree and utility controlling program module which can be invoked by a user confronting problems during running of said core speech recognizer and which utilizes input provided by the user to traverse said decision tree and to initiate corresponding ones of said utility program modules and thereby alter the on-going operating behavior of said core speech recognizer.

2. A speech recognition system as set forth in claim 1 wherein said discrimination improving utility program module deletes the model representing the intruder word which is closest to the models representing the desired word.

3. A speech recognition system as set forth in claim 2 wherein said discrimination improving module creates a new model of the intruder word if the only remaining pre-existing model for that word is deleted.

4. A speech recognition system as set forth in claim 2 wherein said discrimination improving utility program module creates additional models of the desired word from user generated samples.

5. A speech recognition system comprising:

input means for receiving an input utterance from a user and receiving also background noise;

a core speech recognizer which compares an input utterance with a multiplicity of trained word models and generates scores for each of at least a portion of said models, which scores represent the degree of match of the respective model with the utterance,

said core speech recognizer having a plurality of adjustable parameters including;

a sensitivity parameter which controls a discrimination level between background noise and sound levels accepted as an utterance, and

a threshold parameter which functions to exclude models having scores worse than the threshold parameter;

a plurality of primary utility program modules which respond to user input to change system parameters including

said sensitivity parameter; and

said threshold parameter;

a training utility program module which responds to user input to train a selected word model on the basis of a particular user'"'"'s voice characteristics; and

a decision tree and utility controlling program module which can be invoked by a user confronting problems during running of said core speech recognizer and which utilizes user input to traverse said decision tree and to initiate a corresponding one of said utility program modules and thereby alter the on-going operating behavior of said core speech recognizer.

6. A speech recognition system comprising:

input means, including means for providing an adjustable amplitude gain, for receiving an input utterance from a user and receiving also background noise;

a core speech recognizer which compares an input utterance with a multiplicity of trained models representing words and generates scores for each of at least some of said models, which scores represent the degree of match of the respective model with the utterance,

said core speech recognizer having a plurality of adjustable parameters including;

a sensitivity parameter which controls a discrimination level between background noise and sound levels accepted as an utterance, and

a threshold parameter for scores which functions to exclude presumably unlikely models;

a plurality of primary utility program modules which respond to user input to change system parameters including

said adjustable gain;

said sensitivity parameter; and

said threshold parameter;

a training utility program module which responds to user input to train a selected word model on the basis of a particular user'"'"'s voice characteristics;

a discrimination improving utility program module which responds to user input identifying a desired word and an intruder word to improve the separation between the respective models representing the desired and intruder words; and

a decision tree and utility controlling program module which can be invoked by a user confronting problems during running of said core speech recognizer and which utilizes input provided by the user to traverse said decision tree and to initiate a corresponding one of said utility program modules and thereby alter the on-going operating behavior of said core speech recognizer.

7. A speech recognition system as set forth in claim 6 wherein said discrimination improving utility program module deletes the model representing the intruder word which is closest to the models representing the desired word.

8. A speech recognition system as set forth in claim 7 wherein said discrimination improving module creates a new model of the intruder word if the only remaining pre-existing model for that word is deleted.