Method and system for dynamic speech recognition using free-phone scoring

• US 6,243,678 B1
• Filed: 04/07/1998
• Issued: 06/05/2001
• Est. Priority Date: 04/07/1998
• Status: Expired due to Term

• Current Assignee: Avaya Incorporated
• Original Assignee: Lucent Technologies, Inc. (Nokia Corporation)
• Inventors: Erhart, George W., Hartung, Ronald L.
• Primary Examiner(s): Zele, Krista
• Assistant Examiner(s): Opsasnick, Michael N.
• Application Number:
  US09/056,346
• Time in Patent Office: 1,155 Days
• Field of Search: 704/231, 704/249, 704/251, 704/254, 704/255
• US Class Current: 704/249
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units

Abstract

A system performing a speech recognition process requests and receives a claim of identity. The system accesses a customer database and generates a subword spelling of a stored text string where the text string includes predetermined information. The system accesses a subword model database to construct a subword model comprising a series of phonemes. The system requests and receives a speech utterance. A free-phone model of the speech utterance is generated independently of the information stored in the customer database. The system generates a free-phone score. The system generates a word score. The system determines whether the speech utterance matches the stored text string based on the free-phone score and the word score.

Claims

1. A method for recognizing a speech utterance as a predetermined unit of speech, the method comprising:

requesting and receiving a claim of identity;

accessing a customer database in response to receiving said claim of identity;

wherein a predetermined unit of speech is stored in said customer database;

generating a subword spelling of said predetermined unit of speech;

accessing a subword model database that stores a plurality of phonemes, each phoneme representing a sound, to construct a subword model comprising a series of phonemes;

requesting and receiving a speech utterance;

generating a free-phone model of the speech utterance without accessing said customer database;

calculating a free-phone score, said free-phone score representing a likelihood that said free-phone model accurately represents the speech utterance;

calculating a word score, said word score representing a likelihood said subword spelling accurately represents said speech utterance;

determining, based upon said free-phone score and said word score, whether the speech utterance matches the predetermined unit of speech by;

calculating a confidence score based upon said word score and said free-phone score by (i) comparing said word score to said free-phone score, and (ii) if said word score is not better than said free-phone score, comparing said word score to a fraction of said free-phone score.

2. The method of claim 1, further comprising the step of:

comparing said confidence score to a threshold confidence score.

3. The method of claim 2 wherein said step of calculating a confidence score includes calculating a weighted average of said word score and said free-phone score.

4. The method of claim 2 wherein said step of determining includes determining a difference between said word score and said free-phone score.

5. An article of manufacture comprising:

a computer readable medium having computer readable program code means embodied thereon, said computer readable program code means comprising means for generating a free-phone score based upon an utterance, and means for determining, based upon said free-phone score, whether said utterance matches a unit of speech;

wherein said computer readable program code means further includes means for generating a word score based upon said unit of speech and wherein said means for determining determines whether said utterance matches said unit of speech based upon said free-phone score and said word score.

6. A device comprising:

a memory device having a representation of a unit of speech stored therein;

a processor coupled to said memory device, said processor configured to generate a free-phone score based upon a speech utterance and further configured to determine whether said speech utterance is said unit of speech based upon said free-phone score;

an input device for receiving said speech utterance, said input device coupled to said processor;

wherein said processor further generates a word score based upon said unit of speech and determines whether said speech utterance is said unit of speech based upon said free-phone score and said word score;

wherein said processor is configured to compare said word score to said free-phone score, and determine whether said word score is better than said free-phone score;

wherein said processor is further configured to compare said word score to a fraction of said free-phone score.

7. The device of claim 6 wherein said processor is further configured to compare said word score to a fraction of said free-phone score.

8. The device of claim 7 wherein said fraction is one-half.

9. The device of claim 6 wherein said representation of said unit of speech is a subword spelling of said unit of speech.

10. A speech recognition system for recognizing a speech utterance as a predetermined unit of speech, the system comprising:

means for requesting and receiving a claim of identity;

means for accessing a customer database in response to receiving said claim of identity;

wherein a predetermined unit of speech is stored in said customer database;

means for generating a subword spelling of said predetermined unit of speech;

means for accessing a subword model database that stores a plurality of phonemes, each phoneme representing a sound, to construct a subword model comprising a series of phonemes;

means for requesting and receiving a speech utterance;

means for generating a free-phone model of said speech utterance without accessing said customer database;

means for calculating a free-phone score, said free-phone score representing a likelihood that said free-phone model accurately represents said speech utterance;

means for calculating a word score, said word score representing a likelihood said subword spelling accurately represents said speech utterance;

means for determining, based upon said free-phone score and said word score, whether said speech utterance matches said predetermined unit of speech by;

calculating a confidence score based upon said word score and said free-phone score by (i) comparing said word score to said free-phone score, and (ii) if said word score is not better than said free-phone score, comparing said word score to a fraction of said free-phone score.