Bio-phonetic multi-phrase speaker identity verification

• US 20040162726A1
• Filed: 02/13/2003
• Published: 08/19/2004
• Est. Priority Date: 02/13/2003
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Chang, Hisao M.
• Granted Patent: US 7,222,072 B2
• Time in Patent Office: Days
• US Class Current: 704/247
• CPC Class Codes: G10L 17/04 Training, enrolment or model building
• CPC Class Codes: G10L 17/14 Use of phonemic categorisation or speech recognition prior to speaker recognition or verification
• CPC Class Codes: G10L 17/22 Interactive procedures; Man-machine interfaces
• CPC Class Codes: G10L 17/24 the user being prompted to utter a password or a predefined phrase
• CPC Class Codes: G10L 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices

Abstract

A speaker identity claim (SIC) utterance is received and recognized. The SIC utterance is compared with a voice profile registered under the SIC, and a first verification decision is based thereon. A first dynamic phrase (FDP) is generated, and a user is prompted to speak same. An FDP utterance is received, and compared with the voice profile registered under the SIC to make a second verification decision. If the second verification decision indicates a high or low confidence level, the speaker identity claim is accepted or rejected, respectively. If the verification decision indicates a medium confidence level, a second dynamic phrase (SDP) is generated, and the user is prompted to speak same. An SDP utterance is received, and compared with the voice profile registered under the SIC to make a third verification decision. The speaker identity claim is accepted or rejected based on the third verification decision.

Claims

1. A method comprising:

receiving a speaker identity claim (SIC) utterance;

recognizing the SIC utterance;

extracting at least one voice characteristic from the SIC utterance and comparing same with a voice profile registered under the SIC in a voice profile database to generate a first score;

making a first verification decision based on the first score;

generating and prompting a user to speak a first dynamic phrase (FDP);

receiving an FDP utterance;

extracting at least one voice characteristic from the FDP utterance and comparing same with the voice profile registered under the SIC in the voice profile database to generate a second score;

making a second verification decision based on the second score;

if the second verification decision indicates a low confidence level, rejecting the speaker identity claim;

if the second verification decision indicates a high confidence level, accepting the speaker identity claim; and

if the verification decision indicates a medium confidence level;

generating and prompting the user to speak a second dynamic phrase (SDP);

receiving an SDP utterance;

extracting at least one voice characteristic from the SDP utterance and comparing same with the voice profile registered under the SIC in the voice profile database to generate a third score; and

making a third verification decision based on the third score.

2. The method of claim 1 further comprising:

if the third verification decision indicates a low confidence level, rejecting the speaker identity claim; and

if the third verification decision indicates a high confidence level, accepting the speaker identity claim.

3. The method of claim 1 further comprising:

if the second verification decision indicates the low confidence level, tagging the FDP as being attempted by an imposter.

4. The method of claim 1 further comprising:

modifying a threshold that defines the high confidence level for the second verification decision.

5. The method of claim 4 wherein said modifying the threshold is performed in response to an undesirably high level of imposter attempts.

6. The method of claim 4 wherein said modifying the threshold is based on a result of at least one of the second verification decision and the third verification decision.

7. The method of claim 1 wherein the FDP is selected from biographical, geographical and time-sensitive transactional information associated with the SIC.

8. The method of claim 1 wherein the SDP is selected from biographical, geographical and time-sensitive transactional information associated with the SIC.

9. The method of claim 1 wherein the second verification decision is further based on the first score.

10. The method of claim 1 wherein the third verification decision is further based on the first score and the second score.

11. The method of claim 1 further comprising:

if the second verification decision indicates a high confidence level, updating the voice profile associated with the SIC based on at least one of the SIC utterance and the FDP utterance.

12. The method of claim 1 further comprising:

if the third verification decision indicates a high confidence level, updating the voice profile associated with the SIC based on at least one of the SIC utterance, the FDP utterance and the SDP utterance.

13. The method of claim 1 further comprising, prior to receiving the SIC utterance:

generating a registration document for the user, the registration document comprising text based on personal information associated with the user;

securely distribute the registration document and a time-sensitive private key (TSPK) to the user;

receiving a voice call from the user;

receiving the TSPK from the user in the voice call;

prompting the user to speak the text from the registration document; and

recording the user speaking the text in the voice call and storing same in the voice profile database.

14. The method of claim 13 further comprising, prior to receiving the SIC utterance:

prompting the user to speak additional text not included in the registration document; and

recording the user speaking the additional text in the voice call and storing same in the voice profile database.

15. A system comprising:

a voice profile database having a plurality of voice profiles each registered under an associated speaker identity claim (SIC); and

a verification component and a speaker identity verification engine which cooperate to;

receive an SIC utterance;

recognize the SIC utterance;

extract at least one voice characteristic from the SIC utterance and compare same with a voice profile registered under the SIC in the voice profile database to generate a first score;

make a first verification decision based on the first score;

generate and prompt a user to speak a first dynamic phrase (FDP);

receive an FDP utterance;

extract at least one voice characteristic from the FDP utterance and compare same with the voice profile registered under the SIC in the voice profile database to generate a second score;

make a second verification decision based on the second score;

if the second verification decision indicates a low confidence level, reject the speaker identity claim;

if the second verification decision indicates a high confidence level, accept the speaker identity claim; and

if the verification decision indicates a medium confidence level;

generate and prompt the user to speak a second dynamic phrase (SDP);

receive an SDP utterance;

extract at least one voice characteristic from the SDP utterance and compare same with the voice profile registered under the SIC in the voice profile database to generate a third score; and

make a third verification decision based on the third score.

16. The system of claim 15 wherein the verification component is to reject the speaker identity claim if the third verification decision indicates a low confidence level, and to accept the speaker identity claim if the third verification decision indicates a high confidence level.

17. The system of claim 15 wherein the verification component is to tag the FDP as being attempted by an imposter if the second verification decision indicates the low confidence level.

18. The system of claim 15 wherein the verification component is to modify a threshold that defines the high confidence level for the second verification decision.

19. The system of claim 18 wherein the threshold is modified in response to an undesirably high level of imposter attempts.

20. The system of claim 18 wherein the threshold is modified based on a result of at least one of the second verification decision and the third verification decision.

21. The system of claim 15 wherein the FDP is selected from biographical, geographical and time-sensitive transactional information associated with the SIC.

22. The system of claim 15 wherein the SDP is selected from biographical, geographical and time-sensitive transactional information associated with the SIC.

23. The system of claim 15 wherein the second verification decision is further based on the first score.

24. The system of claim 15 wherein the third verification decision is further based on the first score and the second score.

25. The system of claim 15 wherein, if the second verification decision indicates a high confidence level, the verification component is to update the voice profile associated with the SIC based on at least one of the SIC utterance and the FDP utterance.

26. The system of claim 15 wherein if the third verification decision indicates a high confidence level, the verification component is to update the voice profile associated with the SIC based on at least one of the SIC utterance, the FDP utterance and the SDP utterance.

27. The system of claim 15 further comprising:

a document creation component to generate a registration document and a time-sensitive private key (TSPK) for secure distribution to the user, the registration document comprising text based on personal information associated with the user; and

a registration component to;

receive a voice call from the user;

receive the TSPK from the user in the voice call;

prompt the user to speak the text from the registration document; and

record the user speaking the text in the voice call and storing same in the voice profile database.

28. The system of claim 27 wherein the registration component is to:

prompt the user to speak additional text not included in the registration document; and

record the user speaking the additional text in the voice call and storing same in the voice profile database.