Speaker identification with user-selected password phrases

• US 5,913,192 A
• Filed: 08/22/1997
• Issued: 06/15/1999
• Est. Priority Date: 08/22/1997
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: AT&T Corporation (AT&T, Inc.)
• Inventors: Parthasarathy, Sarangarajan, Rosenberg, Aaron Edward
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): SAX, ROBERT L
• Application Number:
  US08/916,662
• Time in Patent Office: 662 Days
• Field of Search: 704/256, 704/250, 704/244
• US Class Current: 704/256.1
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 17/24 the user being prompted to utter a password or a predefined phrase
• CPC Class Codes: G10L 2015/085 Methods for reducing search complexity, pruning

Abstract

A speaker identification system includes a speaker-independent phrase recognizer. The speaker-independent phrase recognizer scores a password utterance against all the sets of phonetic transcriptions in a lexicon database to determine the N best speaker-independent scores, determines the N best sets of phonetic transcriptions based on the N best speaker-independent scores, and determines the N best possible identities. A speaker-dependent phrase recognizer retrieves the hidden Markov model corresponding to each of the N best possible identities, and scores the password utterance against each of the N hidden Markov models to generate a speaker-dependent score for each of the N best possible identities. A score processor coupled to the outputs of the speaker-independent phrase recognizer and the speaker-dependent phrase recognizer determines a putative identity. A verifier coupled to the score processor authenticates the determined putative identity.

Claims

1. A speaker identification system, comprising:

a lexicon database storing for each of a group of enrolled users a set of one or more phonetic transcriptions of a password utterance associated with the enrolled user;

an HNM database storing for each of the group of enrolled users a hidden Markov model corresponding to said password utterance;

a speaker-independent phrase recognizer, which (i) selects N best matching password utterances based on an unknown utterance, and (ii) determines a speaker-independent score for each of the N best matching password utterances;

a speaker-dependent phrase recognizer, which determines a speaker-dependent score for each of the N best matching password utterances; and

a score processor, which (i) for each of the N best matching password utterances, sums the speaker-independent score and the speaker-dependent score to generate a combined score, and (ii) determines a putative identity based on the highest combined score.

2. A system as defined in claim 1, further comprising:

a verifier which determines a verification score and compares the verification score to a verification threshold.

3. A system as defined in claim 2, wherein:

the verification score reflects a difference between the speaker-dependent score for said putative identity and the speaker-independent score for said putative identity.

4. A system as defined in claim 1, wherein:

the speaker-independent phrase recognizer (i) scores said unknown utterance against all the sets of phonetic transcriptions in the lexicon database, (ii) determines the N best sets of phonetic transcriptions, and (iii) selects the N best matching password utterances based on the N best sets of phonetic transcriptions.

5. A system as defined in claim 1, wherein:

the speaker-dependent phrase recognizer (i) retrieves the hidden Markov model corresponding to each of the N best matching password utterances, and (ii) scores said unknown utterance against each of the N hidden Markov models to generate the speaker-dependent score.

6. A method for identifying a speaker, comprising the steps of:

speaking a password utterance;

determining a putative identity from an enrolled user group based on said password utterance by

(i) selecting N best matching password utterances,

(ii) determining a speaker-independent score for each of the N best matching password utterances,

(iii) determining a speaker-dependent score for each of the N best matching password utterances,

(iv) for each of the N best matching password utterances, summing the speaker-independent score and the speaker-dependent score to generate a combined score, and

(v) determining the putative identity based on the highest combined score; and

verifying the determined putative identity using said password utterance.

7. A method as defined in claim 6, further comprising the steps of:

determining a verification score, and

comparing the verification score to a verification threshold.

8. A method as defined in claim 7, wherein:

the verification score reflects a difference between the speaker-dependent score for said putative identity and the speaker-independent score for said putative identity.

9. A method as defined in claim 6, wherein:

a sequence of digits is concatenated with the password utterance.

14. A method as defined in claim 6, further comprising the step of:

extracting features characterizing the password utterance.

10. A speaker identification system, comprising:

means for selecting a putative identity from an enrolled user group based on a spoken password utterance, wherein the means for selecting includes a means for summing speaker-independent and speaker-dependent scores to obtain a combined score; and

means for verifying the selected putative identity using said spoken password utterance.

11. A system as defined in claim 10, further comprising:

means for transducing a spoken password utterance into an electrical speech signal.

12. A system as defined in claim 10, further comprising:

means for extracting features characterizing the spoken password utterance.

13. A system as defined in claim 10, wherein:

a sequence of digits is concatenated with the spoken password utterance.