Adaptive speaker verification apparatus and method including alternative access control

• US 6,119,084 A
• Filed: 12/29/1997
• Issued: 09/12/2000
• Est. Priority Date: 12/29/1997
• Status: Expired due to Term

• Current Assignee: Avaya Incorporated
• Original Assignee: Nortel Networks Corporation
• Inventors: Roberts, Tracy, Will, Craig A.
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US08/999,086
• Time in Patent Office: 988 Days
• Field of Search: 704/243, 704/246, 704/273, 704/231, 704/244, 704/251, 379/88.02
• US Class Current: 704/246
• CPC Class Codes: G10L 17/04 Training, enrolment or model building
• CPC Class Codes: G10L 17/20 Pattern transformations or operations aimed at increasing system robustness, e.g. against channel noise or different working conditions

Abstract

A useful, novel, and nonobvious speech verification apparatus, method and media incorporating the same, as well as a secure interactive voice response ("IVR") system. The speaker verification apparatus and method involves a determination whether one or more speech samples captured from a speaker substantially matches a speaker template in a registration database. If no substantial match is found, alternate access control is utilized to allow the speaker to verify herself, without the use of speaker verification techniques. If alternative access control is able to verify the speaker, the captured speech samples are used to modify the corresponding speaker template. The so-modified speaker template may account for acoustic idiosyncrasies or anomalies perceived in the speech samples to improve verification rates in subsequent acoustic attempts.

Claims

1. A method for verifying a speaker of a captured speech sample, comprising the steps of:

determining whether the captured speech sample substantially matches a speaker template in a registration database corresponding to a registered speaker; and

performing the following steps if it is determined that the captured speech sample fails to substantially match any of the speaker templates in the template database, the steps comprising;

selectively identifying the speaker of the captured speech sample as one of the registered speakers using an alternative access process; and

modifying a selected one of the speaker templates in the template database corresponding to the registered speaker identified in said identifying step based on the captured speech sample.

2. The method of claim 1, wherein the alternative access process is a non-speech-based access process.

3. The method of claim 2, wherein the non-speech access process comprises an interactive, keypad-based process.

4. The method of claim 3, wherein

unique access information is defined for each registered speaker;

and

wherein the interactive, keypad-based process comprises the steps of;

issuing the speaker of the captured speech sample a prompt for access information;

receiving keypad data;

comparing the received keypad data against the access information for any of the registered speakers; and

selectively identifying the speaker of the captured voice sample as one of the registered speakers based on comparison results obtained in said comparing step.

5. The method of claim 4, wherein the access information defined for at least a first one of the registered speakers is phonemically related to the speaker template corresponding thereto.

6. The method of claim 4, further comprising repeating said issuing, receiving, comparing and identifying steps of the interactive, keypad-based process until the first of:

a predetermined limit of prompts have been issued to the speaker of the captured voice sample; and

the speaker of captured voice sample has been identified in said identifying step based on the comparison results obtained in said comparing step.

7. The method of claim 1, wherein the captured speech sample comprises a plurality of speech samples.

8. The method of claim 1, wherein said determining step determines whether the captured speech sample substantially matches one of the speaker templates using a speech recognition process.

9. The method of claim 8, wherein said determining step determines whether the captured speech sample substantially matches one of the speaker templates using an acoustic speaker verification process in tandem with the speech recognition process.

10. An adaptive speaker verification method, comprising the steps of:

capturing a speech sample;

determining whether the captured speech sample substantially matches a speaker template in a registration database corresponding to a registered speaker; and

performing the following steps if it is determined that the captured speech sample fails to substantially match any of the speaker templates in the template database, the steps comprising;

selectively identifying the speaker of the captured speech sample as one of the registered speakers using an alternative access process; and

modifying a selected one of the speaker templates in the template database corresponding to the registered speaker identified in said identifying step based on the captured speech sample.

11. The method of claim 10, wherein the alternative access process is a non-speech-based access process.

12. The method of claim 11, wherein the non-speech access process comprises an interactive, keypad-based process.

13. The method of claim 12, wherein

unique access information is defined for each registered speaker;

and

wherein the interactive, keypad-based process comprises the steps of;

issuing the speaker of the captured speech sample a prompt for access information;

receiving keypad data;

comparing the received keypad data against the access information for any of the registered speakers; and

selectively identifying the speaker of the captured voice sample as one of the registered speakers based on comparison results obtained in said comparing step.

14. The method of claim 13, wherein the access information defined for at least a first one of the registered speakers is phonemically related to the speaker template corresponding thereto.

15. The method of claim 13, further comprising repeating said issuing, receiving, comparing and identifying steps of the interactive, keypad-based process until the first of:

a predetermined limit of prompts have been issued to the speaker of the captured voice sample; and

the speaker of captured voice sample has been identified in said identifying step based on the comparison results obtained in said comparing step.

16. The method of claim 10, wherein the captured speech sample comprises a plurality of speech samples.

17. The method of claim 10, wherein said determining step determines whether the captured speech sample substantially matches one of the speaker templates using a speaker dependent speech recognition process.

18. The method of claim 17, wherein said determining step determines whether the captured speech sample substantially matches one of the speaker templates using an acoustic speaker verification process in tandem with the speech recognition process.

19. An article of manufacture for use in a computer system capable of recognizing a speaker of a captured speech sample, the article of manufacture comprising a computer usable medium including:

computer readable program code embodied in the computer useable medium for causing the computer to determine whether a captured speech sample substantially matches a speaker template in a registration database corresponding to a registered speaker;

computer readable program code embodied in the computer useable medium for causing the computer to selectively identify the speaker of the captured speech sample as one of the registered speakers using an alternative access process if it is determined that the captured speech sample fails to substantially match any of the speaker templates in the template database; and

computer readable program code embodied in the computer useable medium for causing the computer to modify a selected one of the speaker templates in the template database corresponding to the identified registered speaker based on the captured speech sample if it is determined that the captured speech sample fails to substantially match any of the speaker templates in the template database.

20. The article of manufacture of claim 19, wherein the alternative access process is a non-speech-based access process.

21. The article of manufacture of claim 20, wherein the non-speech access process comprises an interactive, keypad-based process.

22. The article of claim 19, wherein

unique access information is defined for each registered speaker;

and

wherein the access information defined for at least a first one of the registered speakers is contextually-related to the speaker template corresponding thereto.

23. The article of manufacture of claim 19, wherein the captured speech sample comprises a plurality of speech samples.

24. The article of manufacture of claim 19, further comprising computer readable program code embodied in the computer useable medium for causing the computer to determine whether the captured speech sample substantially matches one of the speaker templates using a speech recognition process.

25. The article of manufacture of claim 19, further comprising computer readable program code embodied in the computer useable medium for causing the computer to determine whether the captured speech sample substantially matches one of the speaker templates using an acoustic speaker verification process in tandem with the speech recognition process.

26. The article of manufacture of claim 19, further comprising computer readable program code embodied in the computer usable medium for causing the computer to perceive and capture a speech sample from the speaker.

27. A speaker verification apparatus, comprising:

a processor; and

a memory in communication with said processor, said memory comprising;

computer readable program code for causing the processor to determine whether a captured speech sample substantially matches a speaker template in a registration database corresponding to a registered speaker;

computer readable program code for causing the processor to selectively identify a speaker of the captured speech sample as one of the registered speakers using an alternative access process if it is determined that the captured speech sample fails to substantially match any of the speaker templates in the template database; and

computer readable program code for causing the processor to modify a selected one of the speaker templates in the template database corresponding to the identified registered speaker based on the captured speech sample if it is determined that the captured speech sample fails to substantially match any of the speaker templates in the template database.

28. The speaker verification apparatus of claim 27, further comprising a speech capture unit coupled to said processor for perceiving and capturing speech.

29. A speaker verification apparatus, comprising:

a memory including a registration database capable of storing speaker template data for a plurality of registered speakers;

a determination unit in communication with said database to determine whether a captured speech sample substantially matches a speaker template in the registration database;

a speaker identification unit in communication with said database and responsive to said determination unit to selectively identify a speaker of the captured speech sample as one of the registered speakers using an alternative access process if said determination unit determines that the captured speech sample fails to substantially match any of the speaker templates in said database; and

a template modification unit in communication with said database and responsive to said determination and speaker identification units, said template modification unit modifying a selected one of the speaker templates in said database corresponding to the registered speaker identified by said speaker identification unit based on the captured speech sample if said determination unit determines that the captured speech sample fails to substantially match any of the speaker templates in said database.

30. A secure interactive voice response system, comprising:

a speech capture unit for perceiving and capturing speech; and

a speaker verification apparatus, comprising;

a memory including a registration database capable of storing speaker template data for a plurality of registered speakers;

a determination unit in communication with said database to determine whether a captured speech sample substantially matches a speaker template in the registration database;

a speaker identification unit in communication with said database and responsive to said determination unit to selectively identify a speaker of the captured speech sample as one of the registered speakers using an alternative access process if said determination unit determines that the captured speech sample fails to substantially match any of the speaker templates in said database; and

a template modification unit in communication with said database and responsive to said determination and speaker identification units, said template modification unit modifying a selected one of the speaker templates in said database corresponding to the registered speaker identified by said speaker identification unit based on the captured speech sample if said determination unit determines that the captured speech sample fails to substantially match any of the speaker templates in said database.