System and Method for Discriminating Between Speakers for Authentication

• US 20160328547A1
• Filed: 07/19/2016
• Published: 11/10/2016
• Est. Priority Date: 06/23/2008
• Status: Active Application

• Current Assignee: Knapp Investment Company Limited
• Original Assignee: John Nicholas and Kristin Gross Trust U/A/D April 13 2010
• Inventors: Gross, John Nicholas
• Granted Patent: US 10,276,152 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 21/32 using biometric data, e.g. fingerprints, iris scans or voiceprints
• CPC Class Codes: G10L 13/027 Concept to speech synthesisers; Generation of natural phrases from machine-based concepts generation of parameters for speech synthesis out of text G10L13/08
• CPC Class Codes: G10L 13/08 Text analysis or generation of parameters for speech synthesis out of text, e.g. grapheme to phoneme translation, prosody generation or stress or intonation determination
• CPC Class Codes: G10L 15/02 Feature extraction for speech recognition; Selection of recognition unit
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 17/00 Speaker identification or verification techniques
• CPC Class Codes: G10L 17/02 Preprocessing operations, e.g. segment selection; Pattern representation or modelling, e.g. based on linear discriminant analysis [LDA] or principal components; Feature selection or extraction
• 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 17/22 Interactive procedures; Man-machine interfaces
• CPC Class Codes: G10L 17/26 Recognition of special voice characteristics, e.g. for use in lie detectors; Recognition of animal voices
• CPC Class Codes: H04L 63/102 Entity profiles
• CPC Class Codes: H04L 63/123 received data contents, e.g. message integrity
• CPC Class Codes: H04M 2203/2027 Live party detection

Abstract

An audible based electronic challenge system is used to control access to resources by using a test to identify an origin of a voice. The test is based on optimized text sentences selected for their discrimination capability in identifying different speakers.

Claims

1. A computer-implemented method of controlling an entity'"'"'s access to an account comprising:

processing a set of candidate spoken text based challenge sentences with a computing system to identify one or more optimized spoken challenge sentences for controlling an entity'"'"'s access to the account;

wherein said one or more optimized spoken challenge sentences are selected by the computing system based on their acoustical characteristics exceeding a target discrimination capability threshold of a speech recognition engine used for distinguishing different entities uttering such challenge sentences;

selecting a first optimized spoken challenge sentence with the computing system to be articulated as a first speech utterance by a first entity accessing the account;

storing a first voice print for the first entity at the computing system for the optimized spoken challenge sentence as articulated by the first entity;

receiving a second speech utterance by a second entity during an account access attempt;

processing said second speech utterance with said computing system to determine speech characteristics of said second entity; and

performing a comparison with the computing system of said speech characteristics of said second entity to said first voice print to determine whether said second speech utterance also originated from said first entity; and

controlling whether said second entity is allowed to access the account based on said comparison.

2. The method of claim 1 including a step:

selecting said optimized spoken challenge sentences by measuring and ranking a discrimination capability by said speech recognition engine for distinguishing different entities uttering said set of candidate spoken text based challenge sentences.

3. The method of claim 1 wherein said recognition of words within said second speech utterance are performed to identify content of an answer in such utterance and confirm that it matches an expected answer.

4. The method of claim 3 wherein an accuracy of said answer is measured by comparing it to an answer determined from collaborative filtering.

5. The method of claim 1 wherein an interactive dialog of multiple questions and associated answers is performed with said second entity to solicit multiple second speech utterances as part of said comparison step.

6. The method of claim 1 wherein said first optimized spoken challenge sentence is selected randomly from said one or more optimized spoken challenge sentences.

7. The method of claim 1 wherein said first voice print is obtained during an enrollment process and stored in a profile for said first entity.

8. The method of claim 7 wherein said profile includes voice characteristics of said first entity, including speech acoustic characteristic and speech prosodic characteristics.

9. The method of claim 7 wherein phoneme content is selected for discrimination of said first entity during an enrollment process based on demographics and/or an identity of said first entity.

10. The method of claim 1 wherein a number of unique speech sounds sufficient to distinguish between entities are selected for and presented in said first optimized spoken challenge sentence.

11. The method of claim 1 wherein words are selected from a corpus of content for inclusion in said first optimized spoken challenge sentence based on their phonetic distinctiveness.

12. The method of claim 1 wherein said one or more spoken challenge sentences are selected based on an expected human processing and articulation time.

13. The method of claim 1 wherein an articulation time required by such second entity to provide said second speech utterance is also considered to determine if such utterance originated from said first entity.

14. The method of claim 1 wherein processing of said second speech utterance and controlling of access to said account is performed by multiple different verification servers.

15. The method of claim 1 wherein said first optimized spoken challenge sentence differs from a second spoken challenge sentence presented to said second entity

16. The method of claim 1 wherein demographic information for the first entity is used to generate a random challenge sentence according to a challenge grammar to maximize speaker discriminatory ability of the random challenge sentence by the speech recognizer.

17. The method of claim 1 further including steps:

i) computing a first prosodic score for said first optimized spoken challenge sentence based on target prosodic characteristics; and

     2) computing a second prosodic score for said second speech utterance based on measured prosodic characteristics; and

     3) comparing said first prosodic score with said second prosodic score as part of said step of determining whether said second entity is allowed access to the account.

18. The method of claim 17 further including a step:

presenting said first optimized spoken challenge sentence so as to increase prosodic elements in said second speech utterance.

19. The method of claim 1 wherein said optimized spoken challenge sentence is selected based on an identity of said first entity so as to maximize a difference in prosodic characteristics to other entities articulating said optimized spoken challenge sentence.

20. The method of claim 1 wherein said one or more optimized spoken challenge sentences are ranked and sorted according to a score provided by human test subjects identifying whether a challenge item was vocalized by a human or a computer.

21. A challenge system for controlling access to an electronic account on a

computing system comprising:

one or more software routines implemented in a computer readable medium and adapted to cause the challenge system to;

process a set of candidate spoken text based challenge sentences to identify one or more optimized spoken challenge sentences for controlling an entity'"'"'s access to the account;

wherein said one or more optimized spoken challenge sentences have acoustical characteristics exceeding a target discrimination capability threshold of a speech recognition engine used for distinguishing different entities uttering such challenge sentences;

select a first optimized spoken challenge sentence to be articulated as a first speech utterance by a first entity accessing the account;

store a first voice print for the first entity at the computing system for the optimized spoken challenge sentence as articulated by the first entity;

receive a second speech utterance by a second entity during an account access attempt;

process said second speech utterance to determine speech characteristics of said second entity; and

performing a comparison of said speech characteristics of said second entity to said first voice print to determine whether said second speech utterance also originated from said first entity; and

control whether said second entity is allowed to access the account based on said comparison.