Systems and methods for access filtering employing relaxed recognition constraints

• US 6,107,935 A
• Filed: 02/11/1998
• Issued: 08/22/2000
• Est. Priority Date: 02/11/1998
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Maes, Stephane Herman, Comerford, Liam David
• Primary Examiner(s): Zimmerman, Brian
• Assistant Examiner(s): Dalencourt, Yves
• Application Number:
  US09/021,755
• Time in Patent Office: 923 Days
• Field of Search: 340/825.31, 340/825.34, 704/246, 704/250, 704/273, 704/245, 367/197, 382/115, 382/116, 382/117, 382/118
• US Class Current: 340/5.52
• CPC Class Codes: G07C 9/37 using biometric data, e.g. fingerprints, iris scans or voice recognition
• CPC Class Codes: G10L 17/06 Decision making techniques; Pattern matching strategies

Abstract

A speaker recognition system for selectively permitting access by a requesting speaker to one of a service and facility include an acoustic front-end for computing at least one feature vector from a speech utterance provided by the requesting speaker; a speaker dependent codebook store for pre-storing sets of acoustic features, in the form of codebooks, respectively corresponding to a pool of previously enrolled speakers; a speaker identifier/verifier module operatively coupled to the acoustic front-end, wherein: the speaker identifier/verifier module identifies, from identifying indicia provided by the requesting speaker, a previously enrolled speaker as a claimed speaker; further, the speaker identifier/verifier module associates, with the claimed speaker, first and second groups of previously enrolled speakers, the first group being defined as speakers whose codebooks are respectively acoustically similar to the claimed speaker (i.e., cohort set) and the second group being defined as speakers whose codebooks are acoustically similar to the claimed speaker but not as acoustically similar as the codebooks of the speakers in the first group (i.e., legion set); and still further, the speaker identifier/verifier module verifies the requesting speaker by comparing the at least one feature vector of the requesting speaker to the codebooks of the previously enrolled speakers in the second group and, in response to such comparison, generates an indicator indicating that the requesting speaker is one of verified and not verified for access to one of the service and facility.

Claims

1. A user recognition system for selectively permitting access by a requesting user to one of a service and facility, the system comprising:

means for extracting at least one biometric attribute from the requesting user;

means for storing previously extracted biometric attributes respectively corresponding to a pool of previously enrolled users;

means for identifying, from identifying indicia provided by the requesting user, a previously enrolled user as a claimed user;

means for associating, with the claimed user, first and second groups of previously enrolled users, the first group being defined as users who are respectively biometrically similar to the claimed user with respect to the extracted biometric attribute and the second group being defined as users who are biometrically similar to the claimed user but not as biometrically similar as the users in the first group; and

means for verifying the requesting user by comparing the extracted biometric attribute of the requesting user to the pre-stored biometric attributes of the previously enrolled users in the second group and, in response to such comparison, generating an indicator indicating that the requesting user is one of verified and not verified for access to one of the service and facility.

2. The system of claim 1, wherein the biometric attributes are acoustic characteristics associated with the requesting user and the previously enrolled users.

3. A speaker recognition system for selectively permitting access by a requesting speaker to one of a service and facility, the system comprising:

means for computing at least one feature vector from a speech utterance provided by the requesting speaker;

means for pre-storing sets of acoustic features respectively corresponding to a pool of previously enrolled speakers;

means for identifying, from identifying indicia provided by the requesting speaker, a previously enrolled speaker as a claimed speaker;

means for associating, with the claimed speaker, first and second groups of previously enrolled speakers, the first group being defined as speakers who are respectively acoustically similar to the claimed speaker and the second group being defined as speakers who are acoustically similar to the claimed speaker but not as acoustically similar as the speakers in the first group; and

means for verifying the requesting speaker by comparing the at least one feature vector of the requesting speaker to the pre-stored acoustic features of the previously enrolled speakers in the second group and, in response to such comparison, generating an indicator indicating that the requesting speaker is one of verified and not verified for access to one of the service and facility.

4. The system of claim 3, further comprising means for combining at least one of a password, a hardware key, an additional biometric measurement to generate an indicator indicating that the requesting speaker is one of verified and not verified for access to one of the service and facility.

5. The system of claim 3, further comprising means for decoding the speech utterance provided by the requesting speaker to generate a decoded request.

6. The system of claim 5, further comprising conditional access control means, responsive to the decoded request and the speaker verification indicator, for passing through the decoded request to another system responsive to the request, if the requesting speaker is verified.

7. The system of claim 5, further comprising conditional access control means, responsive to the decoded request and the speaker verification indicator, for executing the decoded request if the requesting speaker is verified.

8. The system of claim 3, wherein the pre-stored sets of acoustic features respectively corresponding to the pool of previously enrolled speakers are respectively stored in codebooks.

9. The system of claim 8, wherein the codebooks are respectively characterized by codewords which are characterized by means, variances and scores.

10. The system of claim 9, wherein acoustic similarity between speakers is determined by a distance measure between codewords of the codebooks respectively associated with the speakers.

11. The system of claim 3, wherein the identifying means further comprises means for inputting an identification code of the requesting speaker and comparing the code to identification codes of the previously enrolled speakers to identify the claimed speaker.

12. The system of claim 3, wherein the identifying means further comprises means for comparing the at least one feature vector of the requesting speaker to the pre-stored acoustic features of the previously enrolled speakers to identify the claimed speaker.

13. The system of claim 3, wherein the acoustic features are used to build acoustic models.

14. The system of claim 3, wherein the verification means comparison is relaxed during an enrollment phase so that a portion of the speakers in the first group are accepted and the speakers in the second group are rejected.

15. The system of claim 3, further comprising means for building background models from the pre-stored acoustic features of the previously enrolled speakers in the pool with the exception of the first group such that said background models are employed by the verifying means for comparison to the at least one feature vector associated with the requesting speaker.

16. The system of claim 3, wherein one of the service and facility to which access is selectively permitted includes a voice dialing system.

17. The system of claim 3, wherein one of the service and facility to which access is selectively permitted includes a computer software installation program.

18. The system of claim 3, wherein one of the service and facility to which access is selectively permitted includes execution of a computer program.

19. The system of claim 3, wherein one of the service and facility to which access is selectively permitted includes a computer file encryption program.

20. The system of claim 3, wherein one of the service and facility to which access is selectively permitted includes a computer display screen lock program.

21. The system of claim 3, wherein the first and second groups are formed substantially concurrent with the speaker requesting access to one of the service and facility.

22. A method for recognizing a speaker in order to selectively permit access by a requesting speaker to one of a service and facility, the method comprising the steps of:

(a) computing at least one feature vector from a speech utterance provided by the requesting speaker;

(b) pre-storing sets of acoustic features respectively corresponding to a pool of previously enrolled speakers;

(c) identifying, from identifying indicia provided by the requesting speaker, a previously enrolled speaker as a claimed speaker;

(d) associating, with the claimed speaker, first and second groups of previously enrolled speakers, the first group being defined as speakers who are respectively acoustically similar to the claimed speaker and the second group being defined as speakers who are acoustically similar to the claimed speaker but not as acoustically similar as the speakers in the first group; and

(e) verifying the requesting speaker by comparing the at least one feature vector of the requesting speaker to the pre-stored acoustic features of the previously enrolled speakers in the second group and, in response to such comparison, generating an indicator indicating that the requesting speaker is one of verified and not verified for access to one of the service and facility.

23. The method of claim 22, further comprising the step of decoding the speech utterance provided by the requesting speaker to generate a decoded request.

24. The method of claim 23, further comprising the step of passing through the decoded request to another system responsive to the request, if the speaker is verified.

25. The method of claim 23, further comprising the step of executing the decoded request if the speaker is verified.

26. The method of claim 22, wherein the pre-stored sets of acoustic features respectively corresponding to the pool of previously enrolled speakers are respectively stored in codebooks.

27. The method of claim 26, wherein the codebooks are respectively characterized by codewords which are characterized by means, variances and scores.

28. The method of claim 27, wherein acoustic similarity between speakers is determined by a distance measure between codewords of the codebooks respectively associated with the speakers.

29. The method of claim 22, wherein the identifying step further comprises the step of inputting an identification code of the requesting speaker and comparing the code to identification codes of the previously enrolled speakers to identify the claimed speaker.

30. The method of claim 22, wherein the identifying step further comprises the step of comparing the at least one feature vector of the requesting speaker to the pre-stored acoustic features of the previously enrolled speakers to identify the claimed speaker.

31. The method of claim 22, wherein the acoustic features are used to build acoustic models.

32. The method of claim 22, wherein the verification comparison is relaxed during an enrollment phase so that a portion of the speakers in the first group are accepted and the speakers in the second group are rejected.

33. The method of claim 22, further comprising the step of building background models from the pre-stored acoustic features of the previously enrolled speakers in the pool with the exception of the first group such that said background models are employed in the verifying step for comparison to the at least one feature vector associated with the requesting speaker.

34. The method of claim 22, wherein the first and second groups are formed substantially concurrent with the speaker requesting access to one of the service and facility.

35. The method of claim 22, wherein one of the service and facility to which access is selectively permitted includes voice dialing system.

36. The method of claim 22, wherein one of the service and facility to which access is selectively permitted includes a computer software installation program.

37. The method of claim 22, wherein one of the service and facility to which access is selectively permitted includes a computer file encryption program.

38. The method of claim 22, wherein one of the service and facility to which access is selectively permitted includes a computer display screen lock program.

39. A speaker recognition system for selectively permitting access by a requesting speaker to one of a service and facility comprises:

an acoustic front-end for computing at least one feature vector from a speech utterance provided by the requesting speaker;

a speaker dependent codebook store for pre-storing sets of acoustic features, in the form of codebooks, respectively corresponding to a pool of previously enrolled speakers;

a speaker identifier/verifier module operatively coupled to the acoustic front-end, wherein;

the speaker identifier/verifier module identifies, from identifying indicia provided by the requesting speaker, a previously enrolled speaker as a claimed speaker;

the speaker identifier/verifier module associates, with the claimed speaker, first and second groups of previously enrolled speakers, the first group being defined as speakers whose codebooks are respectively acoustically similar to the claimed speaker and the second group being defined as speakers whose codebooks are acoustically similar to the claimed speaker but not as acoustically similar as the codebooks of the speakers in the first group; and

the speaker identifier/verifier module verifies the requesting speaker by comparing the at least one feature vector of the requesting speaker to the codebooks of the previously enrolled speakers in the second group and, in response to such comparison, generates an indicator indicating that the requesting speaker is one of verified and not verified for access to one of the service and facility.