Device, system, and method of liveness detection utilizing voice biometrics

• US 8,874,442 B2
• Filed: 04/17/2013
• Issued: 10/28/2014
• Est. Priority Date: 11/26/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Aley-Raz, Almog, Krause, Nir Moshe, Salmon, Michael Itzhak, Gazit, Ran Yehoshua
• Primary Examiner(s): Vo, Huyen X.
• Application Number:
  US13/864,343
• Publication Number:
  US 20130226582A1
• Time in Patent Office: 559 Days
• Field of Search: 704/250, 704/273, 704/275, 704/235, 704246-251, 704/231, 704/270, 704/243
• US Class Current: 704/247
• CPC Class Codes: G10L 17/04 Training, enrolment or model building
• CPC Class Codes: G10L 17/10 Multimodal systems, i.e. based on the integration of multiple recognition engines or fusion of expert systems
• CPC Class Codes: G10L 17/24 the user being prompted to utter a password or a predefined phrase

Abstract

Device, system, and method of liveness detection using voice biometrics. For example, a method comprises: generating a first matching score based on a comparison between: (a) a voice-print from a first text-dependent audio sample received at an enrollment stage, and (b) a second text-dependent audio sample received at an authentication stage; generating a second matching score based on a text-independent audio sample; and generating a liveness score by taking into account at least the first matching score and the second matching score.

Claims

1. A method comprising:

generating, with a processor, a first matching score based on a comparison between;

(a) a voice-print from a first text-dependent audio sample received at an enrollment stage, and (b) a second text-dependent audio sample received at an authentication stage;

subsequent to the comparison between the voice-print from the first text-dependent audio sample and the second text-dependent audio sample, generating, with a processor, a second matching score based on a text-independent audio sample;

generating, with a processor, an utterance validation score based on said text-independent audio sample; and

generating, with a processor, a liveness score by taking into account at least the first matching score, the second matching score, and the utterance validation score.

2. The method of claim 1, wherein generating the second matching score comprises:

generating the second matching score based on a comparison between;

(i) a voice-print from said second text-dependent audio sample received at said authentication stage, and (ii) said text-independent audio sample received at a liveness detection stage.

3. The method of claim 2, wherein generating the second matching score comprises:

generating a first interim score based on a comparison between;

(1) the text-independent audio sample received at said authentication stage, and (2) a text-independent voice-print of said text-independent audio sample received at said liveness detection stage;

generating a second interim score based on a comparison between;

(A) the text-independent audio sample received at said liveness detection stage, and (B) a text-independent voice-print of said text-independent audio sample received at said authentication stage; and

generating said second matching score based on at least said first interim score and said second interim score.

4. The method of claim 1, wherein generating the utterance validation score comprises:

applying an Automatic Speech Recognition algorithm to generate said utterance validation score.

5. The method of claim 1, wherein the first text-dependent audio sample comprises speech of one or more phrases, and wherein the second text-dependent audio sample comprises speech of at least one of the one or more phrases.

6. The method of claim 1, wherein at least one of said enrollment stage, said authentication stage or said liveness detection stage is performed over a telephonic communication channel.

7. The method of claim 1, wherein generating the second matching score based on the text-independent audio sample comprises:

pseudo-randomly selecting a phrase from a plurality of pre-defined phrases; and

prompting a user to say said phrase.

8. At least one non-transitory machine-readable medium storing instructions that, when executed by at least one processor, cause the at least one processor to perform a method comprising:

generating a first matching score based on a comparison between;

(a) a voice-print from a first text-dependent audio sample received at an enrollment stage, and (b) a second text-dependent audio sample received at an authentication stage;

subsequent to the comparison between the voice-print from the first text-dependent audio sample and the second text-dependent audio sample, generating a second matching score based on a text-independent audio sample;

generating an utterance validation score based on said text-independent audio sample; and

generating a liveness score by taking into account at least the first matching score, the second matching score, and the utterance validation score.

9. The at least one machine-readable medium of claim 8, wherein generating the second matching score comprises:

generating the second matching score based on a comparison between;

(i) a voice-print from said second text-dependent audio sample received at said authentication stage, and (ii) said text-independent audio sample received at a liveness detection stage.

10. The at least one machine-readable medium of claim 9, wherein generating the second matching score comprises:

generating a first interim score based on a comparison between;

(1) the text-independent audio sample received at said authentication stage, and (2) a text-independent voice-print of said text-independent audio sample received at said liveness detection stage;

generating a second interim score based on a comparison between;

(A) the text-independent audio sample received at said liveness detection stage, and (B) a text-independent voice-print of said text-independent audio sample received at said authentication stage; and

generating said second matching score based on at least said first interim score and said second interim score.

11. The at least one machine-readable medium of claim 8, wherein generating the utterance validation score comprises applying an Automatic Speech Recognition algorithm to generate said utterance validation score.

12. The at least one machine-readable medium of claim 8, wherein at least one of the first matching score and the second matching score comprises a binary value.

13. The at least one machine-readable medium of claim 8, wherein at least one of said enrollment stage, said authentication stage or said liveness detection stage is performed over a telephonic communication channel.

14. The at least one machine-readable medium of claim 8, wherein generating the second matching score based on the text-independent audio sample comprises:

pseudo-randomly selecting a phrase from a plurality of pre-defined phrases, and

prompting a user to say said phrase.

15. An apparatus comprising:

at least one processor; and

at least one machine-readable medium configured to store instructions, which, when executed by the at least one processor, cause the apparatus to implement a speaker recognition module comprising;

a text-dependent enrollment module to capture a text-dependent audio sample during an enrollment stage, and to generate from the text-dependent audio sample a text-dependent voice signature during said enrollment stage;

a text-dependent audio acquisition module to acquire a text-dependent audio sample during an authentication stage;

a text-dependent authentication module to generate a first score based on a comparison between;

(a) the text-dependent voice signature generated during said enrollment stage, and (b) the text-dependent audio sample acquired during said authentication stage;

a text-independent audio acquisition module to acquire a text-independent audio sample during a liveness detection stage by prompting a user to say a phrase selected pseudo-randomly from a pool of pre-defined phrases;

a text-independent authentication module to generate, subsequent to the comparison between the text-dependent voice signature and the text-dependent audio sample, a second score based at least on a comparison between;

(1) said text-independent audio sample acquired during said liveness detection stage, and (2) a text independent voice signature generated from said text-dependent audio sample acquired during said authentication stage;

an utterance validation module to generate a validation score corresponding to validity of said text-independent audio sample acquired during said liveness detection stage; and

a fusion module to generate a liveness detection score based on at least said first score, said second score, and said validation score.

16. The apparatus of claim 15, wherein the text-independent authentication module comprises:

a first text-independent enrollment module to generate a first text-independent voice print based on a text-dependent audio sample acquired during said authentication stage;

a second text-independent enrollment module to generate a second text-independent voice print based on the text-independent audio sample acquired during the liveness detection stage;

a first text-independent authentication unit to generate a first interim matching score based on a comparison between;

(i) the first text-independent voice print, and (ii) the text-independent audio sample acquired during the liveness detection stage;

a second text-independent authentication unit to generate a second interim matching score based on a comparison between;

(i) the second text-independent voice print, and (ii) the text-independent audio sample acquired during the authentication stage; and

a combination module to generate said second score based on the first interim matching score and the second interim matching score.