System for using spoken utterances to provide access to authorized humans and automated agents

• US 9,075,977 B2
• Filed: 10/17/2014
• Issued: 07/07/2015
• Est. Priority Date: 06/23/2008
• Status: Active Grant

• Current Assignee: Knapp Investment Company Limited
• Original Assignee: John Nicholas and Kristin Gross Trust U/A/D April 13 2010
• Inventors: Gross, John Nicholas
• Primary Examiner(s): COLUCCI, MICHAEL C
• Application Number:
  US14/517,453
• Publication Number:
  US 20150039315A1
• Time in Patent Office: 263 Days
• Field of Search: 704/246, 704/243, 704/244, 704/256.4, 704/273, 705/18, 705/14.69, 379/168, 455/410, 709/225, 726/2, 726/23, 726/7
• 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

Voice verification systems for controlling access to electronic accounts using spoken utterances are disclosed. These systems include a database of words and a sentence synthesis engine that uses the dictionary to construct challenge sentences. A challenge logic routine selects a challenge sentence and processes a spoken response from an entity to determine if the entity is authorized to access an electronic account. Both human and machine spoken utterances can be processed as responses, and in some embodiments, the entity may be an automated agent authorized by a human to conduct transactions in said electronic account.

Claims

1. A voice verification system, comprising:

a. a challenge database of words comprising a challenge dictionary;

b. a sentence synthesis engine implemented on a computing system and configured to;

i. retrieve words from the challenge dictionary; and

ii. construct challenge sentences from selected ones of said words;

wherein said challenge sentences are constructed to be suitable as spoken verification inputs by an entity for accessing an electronic account;

c. a challenge logic routine implemented on a computing system and configured to;

i. select a first challenge sentence from said challenge sentences for a first entity; and

ii. determine if such first entity is authorized to access said electronic account based on processing a first spoken response presented in response to said first challenge;

wherein both human and machine articulations can be processed as responses, and said first entity is an automated agent authorized by a human to conduct transactions in said electronic account.

2. The system of claim 1, wherein the challenge sentences include words selected to maximize a difference in acoustical characteristics between two different utterances of the same challenge sentence articulated and originating from two different entities.

3. The system of claim 1, wherein the challenge sentences are comprised of a sequence of words, such that each word in the sequence is randomly selected.

4. The system of claim 1, wherein the challenge sentences are constructed without regard to correct syntax or arrangement of words.

5. The system of claim 1, wherein said automated agent is required to present a separate authorization code to confirm authority for such transactions.

6. The system of claim 1, wherein both speech recognition and natural language processing are performed to identify both the words and the meaning of the first spoken response to determine if it is an appropriate answer to the challenge sentence.

7. The system of claim 1, wherein the challenge logic routine and challenge synthesis routine are stored and execute on separate computing systems.

8. The system of claim 1, wherein a scenario builder routine is configured to construct a sequence of challenge questions and corresponding answers for a first topic, such that the first entity can be engaged and subjected to a sequence of challenges in an interactive dialog session concerning said first topic as part of determining whether to grant access.

9. The system of claim 8, wherein the scenario builder is configured to use visual or text cues to elicit spoken responses.

10. The system of claim 9, wherein the scenario builder is configured to use collaborative filtering to determine appropriate answers for a challenge question.

11. The system of claim 8, wherein the first entity is prompted to select said first topic to be used for said interactive dialog session challenge.

12. The system of claim 1, wherein said first spoken response can be used to authorize a transaction.

13. The system of claim 1, wherein said first challenge sentence is presented in spoken form to the first entity as part of an electronic account access request.

14. The system of claim 1, wherein a first determination is made to identify if such first entity is on a white list of authorized users, and a second determination is made to identify is such firs entity is on a black list.

15. The system of claim 1, wherein the challenge is configured to perform both:

a first audio analysis to identify physical characteristics of a speaker of said first spoken response; and

a second behavior analysis to identify speed, accent and pronunciation of said first spoken response by said speaker;

wherein both analyses are performed during an interactive dialog session to determine if access should be granted to said speaker.

16. The system of claim 1, further wherein the challenge logic routine is configured to:

cause selected ones of the challenge sentences to be articulated by at least one machine text-to-speech (TTS) system as candidate articulations;

rank the candidate articulations based on a score identifying a likelihood that such candidate articulation originated from a machine; and

train a statistical language model recognizer to recognize machine TTS articulations based on selecting candidate articulations according to said ranking, such that a subset of said plurality of candidate challenge items identified as originating from a machine are used as a seed set;

wherein both speech recognition and natural language processing are performed to identify both the words and an answer presented in said first spoken response.

17. A voice verification system, comprising:

(a) a challenge database of words comprising a challenge dictionary;

(b) a sentence synthesis engine implemented on a computing system and configured to;

i. retrieve words from the challenge dictionary; and

ii. construct challenge sentences from selected ones of said words;

wherein said challenge sentences are constructed to be suitable as spoken verification inputs by an entity for accessing an electronic account;

(c) a challenge logic routine implemented on a computing system and configured to;

i. select a first challenge sentence from said challenge sentences for a first entity;

ii. prompt said first entity to articulate first answers related to said first challenge sentence;

iii. process a first spoken response presented in said first answers to said first challenge sentence;

iv. repeat (c)(i) to (c)(iii) for a plurality of second challenge sentences, such that a plurality of second spoken responses are obtained for a plurality of second answers during an interactive question and answer dialog session; and

v. determine if such first entity is authorized to access said electronic account and/or conduct a transaction based on the results of steps (c)(i) to (c)(iv);

wherein both human and machine articulations can be processed as responses, and said first entity is an automated agent authorized by a human to conduct transactions in said electronic account.

18. The system of claim 17, wherein said challenge sentences are constructed so as to maximize a differential in speech distinction information between two different entities of any particular sentence.