System and method for verifying origin of input through spoken language analysis

• US 8,489,399 B2
• Filed: 06/15/2009
• Issued: 07/16/2013
• Est. Priority Date: 06/23/2008
• Status: Active Grant

• Current Assignee: Knapp Investment Company Limited
• Original Assignee: John Nicholas and Kristin Gross Trust
• Inventors: Gross, John Nicholas
• Primary Examiner(s): COLUCCI, MICHAEL C
• Application Number:
  US12/484,837
• Publication Number:
  US 20090319274A1
• Time in Patent Office: 1,492 Days
• Field of Search: 704/246, 705/18
• US Class Current: 704/260
• 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 a computing resource by using a test to identify an origin of a voice. The test is based on analyzing a spoken utterance to determine if it was articulated by an unauthorized human or a text to speech (TTS) system.

Claims

1. A method of controlling access to a computing system comprising:

selecting first test text data to be articulated as a first speech utterance by a first entity providing input to the computing system based on measurable differences in articulation between a human speaker and a machine for the first test text data exceeding a target threshold;

storing a voice print for the first entity at the computing system based on said first speech utterance being converted into recognized speech data;

wherein said first entity can include either a human or a computer using a synthesized voice;

receiving a second speech utterance by a second entity;

processing said second recognized speech data with said computing system to determine whether said second speech utterance also originated from said first entity; and

controlling whether said second entity is allowed to access an account and/or data based on comparing said voice print to said second recognized speech data.

2. The method of claim 1, wherein said access is used for one or more of the following:

a) establishing an online account; and

/or

b) accessing an online account; and

/or

c) establishing a universal online ID; and

/or

d) accessing a universal online ID; and

/or

e) sending email; and

/or

f) accessing email; and

/or

g) posting on a message board; and

/or

h) posting on a web log; and

/or

i) posting on a social network site page;

j) buying or selling on an auction site; and

/or

k) posting a recommendation for an item/service; and

/or

l) selecting an electronic ad.

3. A method of identifying a source of data input to a computing system using prosodic elements of speech comprising:

a) presenting a challenge item to an entity, which challenge item is associated with a reference set of words and associated reference prosodic scores;

b) receiving speech utterance from an entity related to said challenge item including an input set of words;

c) processing said speech utterance with said computing system to compute input prosodic scores of said input set of words;

d) comparing said input prosodic scores and said reference prosodic scores; and

e) generating a determination of whether said speech utterance originated from a machine or a human based on step (d);

wherein said challenge item is supplemented with visual cues, said visual cues being adapted to induce said reference prosodic scores.

4. The method of claim 3 further including the step:

recognizing said input set of words to compute an additional prosodic score based on an identity of said input set of words, and comparing said additional prosodic words to a second reference prosodic score related to a content of said reference set of words.

5. The method of claim 3, wherein said visual cues are selected from a database of visual cues determined by reference to a database of human vocalizations to most likely result in said reference prosodic scores.

6. A method of identifying a source of data input to a computing system using prosodic elements of speech comprising:

a) presenting a challenge item to an entity, which challenge item is associated with a reference set of words and associated prosodic characteristics;

b) receiving speech utterance from an entity related to said challenge item;

wherein said reference set of words represents a selected set of one or more contiguous words which when vocalized have a measurable difference in prosodic characteristics between a reference human voice and a reference computer synthesized voice that exceeds a target threshold;

c) processing said speech utterance with said computing system to compute first prosodic characteristics of said entity;

d) generating a determination of whether said speech utterance originated from a machine or a human based on step (c).

7. The method of claim 6 further including the steps:

estimating a first computer synthesized voice that best correlates to said entity; and

selecting said challenge item based on an identity of said first entity so as to maximize a difference in prosodic characteristics.

8. The method of claim 6, further including steps:

soliciting utterances from a plurality of separate computing machines to determine their respective prosodic characteristics; and

storing said plurality of associated prosodic characteristics in a database of known computing entities.

9. The method of claim 8, wherein multiple samples of individual challenge sentences are collected.

10. The method of claim 6 wherein visual cues are added to induce said entity to vocalize said reference set of words using said reference human voice.

11. A method of implementing a CAPTCHA (Completely Automatic Public Turing Test To Tell Humans And Computers Apart) to identify a source of data input to a computing system comprising:

a) using a set of human test subjects to identify whether a reference challenge item was vocalized by a human or a computer;

b) training the computing system with samples of human voices and computer synthesized voices articulating a set of the reference challenge items;

c) receiving a speech utterance from an entity related to one of said set of reference challenge items;

d) determining with the trained computer system whether said speech utterance was vocalized by a machine or a human.

12. The method of claim 11, wherein said reference challenge items are ranked and sorted according to a score provided by said human test subjects, and further including a step:

presenting said one of said set of reference challenge items based on said score.

13. A method of implementing a CAPTCHA (Completely Automatic Public Turing Test To Tell Humans And Computers Apart) to identify a source of data input to a computing system comprising:

a) training the computing system with samples of human voices articulating a set of reference challenge items;

b) receiving a speech utterance from an entity related to one of said set of reference challenge items;

c) determining with the trained computer system whether said speech utterance was vocalized by a machine or a human;

wherein said computing system uses one or more speech models that are optimized for identifying humans using said set of reference challenge items; and

wherein said set of reference challenge items represent a selected set of one more contiguous words which when articulated have a difference in acoustical characteristics between a reference human voice and a reference computer synthesized voice that exceeds a target threshold as measured by a reference group of human listeners, and at least some of said acoustical characteristics are used to train said one or more speech models.

14. The method of claim 13, including a step:

using a set of human test subjects to identify whether a reference challenge item was vocalized by a human or a computer prior to using it in the training of the computing system.

15. A challenge system for identifying a source of data input to a computing system comprising:

one or more software routines implemented in a computer readable medium and

adapted to cause the challenge system to;

select first test text data to be articulated as a first speech utterance by a first entity providing input to the computing system, based on measurable differences in articulation between a human speaker and a machine for the first test text data exceeding a target threshold;

store a voice print for the first entity at the computing system based on said first speech utterance being converted into recognized speech data;

wherein said first entity can include either a human or a computer using a synthesized voice;

receive a second speech utterance by a second entity;

process said second recognized speech data with said computing system to determine whether said second speech utterance also originated from said first entity;

control whether said second entity is allowed to access an account and/or data based on comparing said voice print to said second recognized speech data.

16. A challenge system for identifying a source of data input to a computing system using prosodic elements of speech comprising:

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

a) present a challenge item to an entity, which challenge item is associated with a reference set of words and associated reference prosodic scores;

b) receive speech utterance from an entity related to said challenge item including an input set of words;

c) process said speech utterance with said computing system to compute input prosodic scores of said input set of words;

d) compare said input prosodic scores and said reference prosodic scores; and

e) generate a determination of whether said speech utterance originated from a machine or a human based on step (d);

wherein said challenge item is supplemented with visual cues, said visual cues being adapted to induce said reference prosodic scores.

17. A challenge system for identifying a source of data input to a computing system using prosodic elements of speech comprising:

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

a) present a challenge item to an entity, which challenge item is associated with a reference set of words and associated prosodic characteristics;

b) receive speech utterance from an entity related to said challenge item;

wherein said reference set of words represents a selected set of one or more contiguous words which when vocalized have a measurable difference in prosodic characteristics between a reference human voice and a reference computer synthesized voice that exceeds a target threshold;

c) process said speech utterance with said computing system to compute first prosodic characteristics of said entity;

d) generate a determination of whether said speech utterance originated from a machine or a human based on step (c).

18. A system for implementing a CAPTCHA (Completely Automatic Public Turing Test To Tell Humans And Computers Apart) to identify a source of data input to a computing system comprising:

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

a) use a set of human test subjects to identify whether a reference challenge item was vocalized by a human or a computer;

b) train the computing system with samples of human voices and computer synthesized voices articulating a set of the reference challenge items;

c) receive a speech utterance from an entity related to one of said set of reference challenge items;

d) determine with the trained computer system whether said speech utterance was vocalized by a machine or a human.

19. A system for implementing a CAPTCHA (Completely Automatic Public Turing Test To Tell Humans And Computers Apart) to identify a source of data input to a computing system comprising:

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

a) train the computing system with samples of human voices articulating a set of reference challenge items;

b) receive a speech utterance from an entity related to one of said set of reference challenge items;

c) determine with the trained computer system whether said speech utterance was vocalized by a machine or a human;

wherein said computing system uses one or more speech models that are optimized for identifying humans using said set of reference challenge items wherein said set of reference challenge items represent a selected set of one more contiguous words which when articulated have a difference in acoustical characteristics between a reference human voice and a reference computer synthesized voice that exceeds a target threshold as measured by a reference group of human listeners, and at least some of said acoustical characteristics are used to train said one or more speech models.