Automated decision making using time-varying stream reliability prediction

• US 7,228,279 B2
• Filed: 03/26/2003
• Issued: 06/05/2007
• Est. Priority Date: 03/26/2003
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Ramaswamy, Ganesh N., Potamianos, Gerasimos, Chaudhari, Upendra V., Neti, Chalapathy
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US10/397,762
• Publication Number:
  US 20040193415A1
• Time in Patent Office: 1,532 Days
• Field of Search: None
• US Class Current: 704/273
• CPC Class Codes: G10L 17/06 Decision making techniques; Pattern matching strategies
• CPC Class Codes: G10L 17/20 Pattern transformations or operations aimed at increasing system robustness, e.g. against channel noise or different working conditions

Abstract

Automated decision making techniques are provided. For example, a technique for generating a decision associated with an individual or an entity includes the following steps. First, two or more data streams associated with the individual or the entity are captured. Then, at least one time-varying measure is computed in accordance with the two or more data streams. Lastly, a decision is computed based on the at least one time-varying measure. One form of the time-varying measure may include a measure of the coverage of a model associated with previously-obtained training data by at least a portion of the captured data. Another form of the time-varying measure may include a measure of the stability of at least a portion of the captured data. While either measure may be employed alone to compute a decision, preferably both the coverage and stability measures are employed. The technique may be used to authenticate a speaker.

Claims

1. A method of generating a decision associated with one of an individual and an entity, the method comprising the steps of:

inputting two or more data streams associated with one of the individual and the entity;

computing at least one time-varying measure in accordance with the two or more data streams;

computing a decision output value based on the at least one time-varying measure; and

authenticating one of the individual and the entity based on the decision output value;

wherein the at least one time-varying measure comprises a measure of the coverage of a model associated with previously-obtained training data by at least a portion of the captured data, further wherein the coverage measure is determined in accordance with an inverse likelihood computation, the inverse likelihood computation comprising modeling, for a time t, a neighborhood of a test vector associated with the captured data to generate a test data model, and measuring the likelihood of one or more parameters of the training data model with respect to the test data model.

2. The method of claim 1, wherein the decision is not computed until the at least one time-varying measure is one of greater than and equal to a given threshold value.

3. The method of claim 1, wherein the decision is a speaker authentication decision.

4. The method of claim 1, wherein the two or more data streams comprise two or more of audio data, video data, and other data relating to one of the individual and the entity.

5. The method of claim 1, wherein the feature space associated with the test data model is transformed into the feature space associated with the training data model.

6. The method of claim 1, wherein the two or more data streams comprise data derived from the data associated with one of the individual and the entity.

7. The method of claim 1, wherein the measure computation step further comprises computing a second time-varying measure associated with the two or more data streams.

8. The method of claim 7, wherein the at least one time-varying measure and the second time-varying measure comprise a coverage measure and a stability measure.

9. The method of claim 1, wherein the at least one time-varying measure is normalized.

10. The method of claim 9, wherein the normalization induces a context dependence.

11. A method of generating a decision associated with one of an individual and an entity, the method comprising the steps of:

inputting two or more data streams associated with one of the individual and the entity;

computing at least one time-varying measure in accordance with the two or more data streams;

computing a decision output value based on the at least one time-varying measure; and

authenticating one of the individual and the entity based on the decision output value;

wherein the at least one time-varying measure comprises a measure of the stability of at least a portion of the captured data.

12. The method of claim 11, wherein the stability measure is determined in accordance with a deviation computation.

13. The method of claim 12, wherein the deviation computation comprises computing a deviation of a score at time t from a point estimate of the score at time t based on a neighborhood of test vectors associated wit the inputted data.

14. A method of generating a decision associated with one of an individual and an entity, the method comprising the steps of:

inputting two or more data streams associated with one of the individual and the entity;

computing at least one time-varying measure in accordance with the two or more data streams;

computing a decision output value based on the at least one time-varying measure; and

authenticating one of the individual and the entity based on the decision output value;

wherein the measure computation step further comprises computing a second time-varying measure associated with the two or more data streams, and the at least one time-varying measure and the second time-varying measure comprise a coverage measure and a stability measure;

further wherein the measure computation step further comprises the steps of;

computing the coverage measure and the stability measure for each data stream;

normalizing the coverage measure and the stability measure over the two or more data streams; and

forming discriminants by combining scores obtained from the captured data based on the normalized measures, the discriminants being used to compute the decision.

15. Apparatus for generating a decision associated with one of an individual and an entity, the apparatus comprising:

a memory; and

at least one processor coupled to the memory and operative to;

(i) input two or more data streams associated with one of the individual and the entity;

(ii) compute time-varying measures in accordance with the two or more data streams;

(iii) compute a decision output value based on the time-varying measures; and

(iv) authenticate one of the individual and the entity based on the decision output value;

wherein the measure computation operation further comprises;

computing a coverage measure and a stability measure for each data stream;

normalizing the coverage measure and the stability measure over the two or more data streams; and

forming discriminants by combining scores obtained from the captured data based on the normalized measures, the discriminants being used to compute the decision.

16. The apparatus of claim 15, wherein the decision output value is a speaker authentication decision.

17. The apparatus of claim 15, wherein the two or more data streams comprise two or more of audio data, video data, and other data relating to one of the individual and the entity.

18. The apparatus of claim 15, wherein the two or more data streams comprise data derived from the data associated with one of the individual and the entity.

19. An article of manufacture for generating a decision associated with one of an individual and an entity, comprising a machine readable medium containing one or more programs which when executed implement the steps of:

inputting two or more data streams associated with one of the individual and the entity;

computing time-varying measures in accordance with the two or more data streams;

computing a decision output value based on the time-varying measures; and

authenticating one of the individual and the entity based on the decision output value;

wherein the measure computation step further comprises;

computing a coverage measure and a stability measure for each data stream;

normalizing the coverage measure and the stability measure over the two or more data streams; and

forming discriminants by combining scores obtained from the captured data based on the normalized measures, the discriminants being used to compute the decision.