Cognitive arbitration system

• US 20050256820A1
• Filed: 05/14/2004
• Published: 11/17/2005
• Est. Priority Date: 05/14/2004
• Status: Active Grant

• Current Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Original Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Inventors: Paradis, Rosemary D., Dugan, Peter, Tillotson, Dennis A., Lewis, Lori K.
• Granted Patent: US 7,340,443 B2
• Time in Patent Office: Days
• US Class Current: 706/52
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06N 7/02 using fuzzy logic computing arrangements based on biological models G06N3/00; computing arrangements using knowledge-based models G06N5/00

Abstract

A method and computer product is disclosed for arbitrating the outputs of a plurality of recognizers to select a hypothesis associated with a given input from a plurality of hypotheses. Fuzzy membership states are assigned to respective candidate outputs of the plurality of recognizers. At least one rule is selected according to the assigned fuzzy membership states. A selected rule provides a determined amount of support mass for an associated one of the plurality of hypotheses and a determined amount of uncertainty mass. A support value for each hypothesis and an overall uncertainty are determined from the provided mass values. A hypothesis having a highest support value while having an acceptable low level of uncertainty is selected.

Claims

1. A method of arbitrating the outputs of a plurality of recognizers to select a hypothesis associated with a given input from a plurality of hypotheses, comprising:

assigning fuzzy membership states to respective candidate outputs of the plurality of recognizers;

selecting at least one rule according to the assigned fuzzy membership states, a selected rule providing a determined amount of support mass for an associated one of the plurality of hypotheses and a determined amount of uncertainty mass;

determining a support value for each hypothesis and an overall uncertainty value from the provided mass values; and

selecting a hypothesis from the plurality of hypotheses having a highest support value.

2. A method as set forth in claim 1 wherein the step of determining a total support value for a given hypothesis includes multiplying the associated support masses for a plurality of selected rules associated with the hypothesis to form a support value and multiplying the associated uncertainty mass for the plurality of selected rules associated with the hypothesis to form an uncertainty value.

3. A method as set forth in claim 2 wherein computing the overall uncertainty value includes summing the uncertainty value associated with each hypothesis across the plurality of hypotheses.

4. A method as set forth in claim 2 wherein each rule has an associated priority value and the associated mass for the rule is weighted by the priority value.

5. A method as set forth in claim 4, wherein the associated priority values of the plurality of rules are configurable by a user to adjust the relative importance of each rule.

6. A method as set forth in claim 1 wherein the each of the plurality of recognizers produces a plurality of candidate outputs, each associated with one of the plurality of hypotheses and ranked by associated output scores, and the at least one rule is selected according to the assigned fuzzy membership states and the ranking of the candidate outputs.

7. A method as set forth in claim 1 wherein at least one of the plurality of recognizers includes a pattern recognition classifier, and a given hypothesis is associated with an output class of the at least one pattern recognition classifier.

8. A method as set forth in claim 1 further comprising the steps of:

comparing the support value for the selected hypothesis to a first threshold value;

comparing the overall uncertainty value to a second threshold value;

accepting the selected hypothesis if the support value meets the first threshold and the overall uncertainty value is below the second threshold value; and

outputting a negative result if the selected hypothesis is not accepted.

9. A method as set forth in claim 1 further comprising the step of normalizing the support values and the overall uncertainty to a desired scale.

10. A method as set forth in claim 9 further comprising the step of computing a plausibility range for each hypothesis from the normalized values.

11. A computer program product, fixed in a computer-readable medium and operative in a data processing system, that arbitrates the outputs of a plurality of recognizers to select a hypothesis for a given input from a plurality of hypotheses, comprising:

at least one fuzzy class assignor that assigns a fuzzy membership state to respective at least one candidate outputs of the plurality of recognizers;

a rule selector that evaluates a plurality of rules according to the assigned fuzzy membership states of the candidate outputs to determine which of the plurality of rules apply to each of the plurality of output hypotheses, an applicable rule providing a determined amount of support mass for an associated hypothesis and a determined amount of uncertainty mass; and

a data fusion module that determines a support value for each hypothesis and an overall uncertainty value from the provided mass values and selects a hypothesis having a highest support value.

12. A system as set forth in claim 11 wherein the plurality of hypotheses include output classes associated with at least one recognizer.

13. A system as set forth in claim 11, wherein the data fusion module normalizes the support values for the plurality of hypotheses to determine associated belief and plausibility values for each of the plurality of hypotheses.

14. A system as set forth in claim 13 wherein the fusion module outputs a negative hypothesis, indicating that an appropriate hypothesis cannot be selected for the input, if the determined belief for the associated hypothesis fails to meet a threshold value and the uncertainty level is high.

15. A system as set forth in claim 11 wherein a given recognizer produces a plurality of candidate outputs and ranks the candidate outputs according to associated output scores.

16. A system for classifying an input image into one of a plurality of output classes comprising:

a plurality of recognizers, a given recognizer selecting at least one of the plurality of output classes and assigning respective output scores to the selected at least one output class;

at least one fuzzy class assignor that assigns a fuzzy membership state to the selected output classes of the plurality of recognizers;

a rule selector that evaluates a plurality of rules according to the assigned fuzzy membership states to determine which of the plurality of rules apply to each of a plurality of output hypotheses, an applicable rule providing a determined amount of support mass for an associated hypothesis and a determined amount of uncertainty mass; and

a data fusion module that determines a support value for each hypothesis and an overall uncertainty value from the provided mass values and selects an output class associated with a hypothesis having a highest support value.

17. A system as set forth in claim 16 wherein the plurality of output class represent human faces.

18. A system as set forth in claim 16 wherein the plurality of recognizers includes a neural network classifier.

19. A system as set forth in claim 16 wherein the input image comprises a video input recording at least one entrant to a secured location.

20. A system as set forth in claim 19 wherein the classification system determines the identity of a given entrant from the video input.

21. A system as set forth in claim 20 wherein the input image is analyzed relative to an entrant profile associated with the identified entrant to determine if an abnormal condition exists.

22. A computer program product as set forth in claim 21, wherein the entrant profile includes the typical time of entry and place of entry by an entrant to the building.

23. A computer program product as set forth in claim 21, wherein the entrant profile includes items typically carried by the entrant.

24. A computer program product as set forth in claim 21, wherein the system includes an alarm that notifies a human operator when an abnormal condition exists.