Method and apparatus for automated speaker parameters adaptation in a deployed speaker verification system

• US 9,865,266 B2
• Filed: 02/25/2013
• Issued: 01/09/2018
• Est. Priority Date: 02/25/2013
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Colibro, Daniele Ernesto, Vair, Claudio, Farrell, Kevin R.
• Primary Examiner(s): Leland, III, Edwin S
• Application Number:
  US13/776,502
• Publication Number:
  US 20140244257A1
• Time in Patent Office: 1,779 Days
• Field of Search: None
• CPC Class Codes: G10L 17/04 Training, enrolment or model building
• CPC Class Codes: G10L 17/06 Decision making techniques; Pattern matching strategies
• CPC Class Codes: G10L 17/12 Score normalisation
• CPC Class Codes: G10L 17/20 Pattern transformations or operations aimed at increasing system robustness, e.g. against channel noise or different working conditions
• CPC Class Codes: G10L 17/22 Interactive procedures; Man-machine interfaces

Abstract

Typical speaker verification systems usually employ speakers'"'"' audio data collected during an enrollment phase when users enroll with the system and provide respective voice samples. Due to technical, business, or other constraints, the enrollment data may not be large enough or rich enough to encompass different inter-speaker and intra-speaker variations. According to at least one embodiment, a method and apparatus employing classifier adaptation based on field data in a deployed voice-based interactive system comprise: collecting representations of voice characteristics, in association with corresponding speakers, the representations being generated by the deployed voice-based interactive system; updating parameters of the classifier, used in speaker recognition, based on the representations collected; and employing the classifier, with the corresponding parameters updated, in performing speaker recognition.

Claims

1. A method of performing speaker recognition, the method comprising:

collecting representations of voice characteristics, in association with corresponding speakers, the representations being generated by a deployed voice-based interactive system;

updating system-level parameters of a classifier used in speaker recognition, based on the representations collected, the representations collected being indicative of inter-speaker and intra-speaker variations in voice characteristics, the system-level parameters updated to improve distinguishability between different speakers by maximizing inter-speaker variations while accommodating intra-speaker variability, wherein updating the system-level parameters of the classifier includes updating the system-level parameters each time a given number of successful identifications of at least one speaker is achieved; and

employing the classifier, with the corresponding system-level parameters updated, in performing speaker recognition.

2. A method according to claim 1 further comprising updating at least one score normalization employed by the classifier.

3. A method according to claim 1 further comprising:

collecting a representation of a speaker identity in association with at least one representation of voice characteristics of a corresponding speaker; and

comparing the representation of the speaker identity collected to a speaker identification provided by the classifier.

4. A method according to claim 1 further comprising computing the representations of voice characteristics based on field data received by the deployed voice-based interactive system.

5. A method according to claim 1 further comprising generating the representations of voice characteristics.

6. A method according to claim 1, wherein updating the system-level parameters of the classifier includes updating the system-level parameters based on a subset of the representations of the voice characteristics collected.

7. A method according to claim 6 further comprising selecting the subset of the representations based on representations associated with successful identification of the at least one speaker by the deployed voice-based interactive system.

8. A method according to claim 7 further comprising selecting the subset of the representations based on representations exhibiting large variations with respect to corresponding representations previously maintained by the deployed voice-based interactive system.

9. A method according to claim 1, wherein updating the system-level parameters of the classifier includes updating one or more representations of voice characteristics associated with the at least one speaker.

10. An apparatus of performing speaker recognition, the apparatus comprising:

at least one processor; and

at least one memory with computer code instructions stored thereon,

the at least one processor and the at least one memory, with the computer code instructions, being configured to cause the apparatus to;

collect representations of voice characteristics, in association with corresponding speakers, the representations being generated by a deployed voice-based interactive system;

update system-level parameters of a classifier, used in speaker recognition, based on the representations collected, the representations collected being indicative of inter-speaker and intra-speaker variations in voice characteristics, the system-level parameters updated to improve distinguishability between different speakers by maximizing inter-speaker variations while accommodating intra-speaker variability, wherein in updating the system-level parameters of the classifier, the at least one processor and the at least one memory, with the computer code instructions, being configured to cause the apparatus to update the system-level parameters each time a given number of successful identifications of at least one speaker is achieved; and

employ the classifier, with the corresponding system-level parameters updated, in performing speaker recognition.

11. An apparatus according to claim 10, wherein the at least one processor and the at least one memory, with the computer code instructions, being configured to cause the apparatus to further update at least one score normalization employed by the classifier.

12. An apparatus according to claim 10, wherein the at least one processor and the at least one memory, with the computer code instructions, being configured to cause the apparatus to further:

collect a representation of a speaker identity in association with at least one representation of voice characteristics of a corresponding speaker; and

compare the representation of the speaker identity collected to a speaker identification provided by the classifier.

13. An apparatus according to claim 10, wherein the at least one processor and the at least one memory, with the computer code instructions, being configured to cause the apparatus to further compute the representations of voice characteristics based on field data received by the deployed voice-based interactive system.

14. An apparatus according to claim 10, wherein the at least one processor and the at least one memory, with the computer code instructions, being configured to cause the apparatus to further generate the representations of voice characteristics.

15. An apparatus according to claim 10, wherein in updating the system-level parameters of the classifier, the at least one processor and the at least one memory, with the computer code instructions, being configured to cause the apparatus to update the system-level parameters based on a subset of the representations of the voice characteristics collected.

16. An apparatus according to claim 15, wherein the at least one processor and the at least one memory, with the computer code instructions, being configured to cause the apparatus to further select the subset of the representations based on representations associated with successful identification of the at least one speaker by the deployed voice-based interactive system.

17. An apparatus according to claim 16, wherein the at least one processor and the at least one memory, with the computer code instructions, being configured to cause the apparatus to further select the subset of the representations based on representations exhibiting large variations with respect to corresponding representations previously maintained by the deployed voice-based interactive system.

18. A non-transitory computer-readable medium with computer software instructions stored thereon, the computer software instructions when executed by a processor cause an apparatus to perform:

collecting representations of voice characteristics, in association with corresponding speakers;

updating system-level parameters of a classifier, used in speaker recognition, based on the representations collected, the representations collected being indicative of inter-speaker and intra-speaker variations in voice characteristics, the system-level parameters updated to improve distinguishability between different speakers by maximizing inter-speaker variations while accommodating intra-speaker variability, wherein updating the system-level parameters of the classifier includes updating the system-level parameters each time a given number of successful identifications of at least one speaker is achieved; and

employing the classifier, with the corresponding system-level parameters updated, in performing speaker recognition.