NONLINEAR SET TO SET PATTERN RECOGNITION

• US 20080256130A1
• Filed: 02/22/2008
• Published: 10/16/2008
• Est. Priority Date: 02/22/2007
• Status: Active Grant

• Current Assignee: Colorado State University Research Foundation
• Original Assignee: Colorado State University Research Foundation
• Inventors: Kirby, Michael Joseph, Peterson, Christopher Scott
• Granted Patent: US 7,917,540 B2
• Time in Patent Office: Days
• US Class Current: 707/104.1
• CPC Class Codes: G06F 18/21375 involving differential geometry, e.g. embedding of pattern manifold
• CPC Class Codes: G06V 10/7715 Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods
• CPC Class Codes: G06V 40/172 Classification, e.g. identification

Abstract

Variations in the states of patterns can be exploited for their discriminatory information and should not be discarded as noise. A pattern recognition system compares a data set of unlabeled patterns having variations of state in a set-by-set comparison with labeled arrays of individual data sets of multiple patterns also having variations of state. The individual data sets are each mapped to a point on a parameter space, and the points of each labeled array define a subset of the parameter space. If the point associated with the data set of unlabeled patterns satisfies a similarity criterion on the parameter space subset of a labeled array, the data set of unlabeled patterns is assigned to the class attributed to that labeled array.

Claims

1. A method of classifying a data set of related unlabeled patterns, the method comprising:

encoding a collection of data sets of patterns onto a parameter space, each data set being encoded to at least one point on the parameter space, each data set further being allocated to a labeled array of data sets, each labeled array being designated to a class;

defining a mapping operator for each class that maps the data sets of the labeled array onto the parameter space in satisfaction of a similarity criterion based on the encoded points of that labeled array on the parameter space;

encoding the data set of related unlabeled patterns to at least one point on the parameter space;

generating a similarity measurement for each class based on the encoded point of the data set of related unlabeled patterns by mapping the data set of related unlabeled patterns on the parameter space using the mapping operator for the class;

labeling the data set of related unlabeled patterns as a member of a class, if the similarity measurement associated with the mapping operator of the class satisfies a classification criterion.

2. The method of claim 1 wherein a similarity measurement associated with the mapping operator of the class satisfies the classification criterion if the similarity measurement is the minimum of all similarity measurements associated with the mapping operators of the classes.

3. The method of claim 1 wherein the operation of encoding a collection of data sets and the operation of encoding the data set of related unlabeled patterns comprise mapping all of the data sets on the parameter space using the same encoding algorithm.

4. The method of claim 1 wherein the similarity criterion is satisfied if the data sets of the labeled array are mapped using the mapping operator of the labeled array to the same location, within a defined threshold, on the parameter space as the encoded points of that labeled array on the parameter space.

5. The method of claim 1 wherein a similarity measurement represents a difference between a point encoded on the parameter space from the data set of related unlabeled patterns and a point mapped from the data set of related unlabeled patterns by the mapping operator of the class.

6. The method of claim 1 wherein the labeling operation comprises:

labeling the data set of related unlabeled patterns as a member of the class, if the similarity measurement associated with the mapping operator of the class satisfies the classification criterion and the similarity measurement satisfies a similarity threshold.

7. The method of claim 1 further comprising:

indicating an unsuccessful classification, if no similarity measurement satisfies a similarity threshold.

8. One or more computer readable storage media storing computer executable instructions for executing a computer process classifying a data set of related unlabeled patterns on a computing system, the computer process comprising:

encoding a collection of data sets of patterns onto a parameter space, each data set being encoded to at least one point on the parameter space, each data set further being allocated to a labeled array of data sets, each labeled array being designated to a class;

defining a mapping operator for each class that maps the data sets of the labeled array onto the parameter space in satisfaction of a similarity criterion based on the encoded points of that labeled array on the parameter space;

encoding the data set of related unlabeled patterns to at least one point on the parameter space;

generating a similarity measurement for each class based on the encoded point of the data set of related unlabeled patterns by mapping the data set of related unlabeled patterns on the parameter space using the mapping operator for the class;

labeling the data set of related unlabeled patterns as a member of a class, if the similarity measurement associated with the mapping operator of the labeled array satisfies a classification criterion.

9. The one or more computer readable media of claim 8 wherein a similarity measurement associated with the mapping operator of the class satisfies the classification criterion if the similarity measurement is the minimum of all similarity measurements associated with the mapping operators of the classes.

10. The one or more computer readable media of claim 8 wherein the operation of encoding a collection of data sets and the operation of encoding the data set of related unlabeled patterns comprise mapping all of the data sets on the parameter space using the same encoding algorithm.

11. The one or more computer readable media of claim 8 wherein the similarity criterion is satisfied if the data sets of the labeled array are mapped using the mapping operator of the labeled array to the same location, within a defined threshold, on the parameter space as the encoded points of that labeled array on the parameter space.

12. The one or more computer readable media of claim 8 wherein a similarity measurement represents a difference between a point encoded on the parameter space from the data set of related unlabeled patterns and a point mapped from the data set of related unlabeled patterns by the mapping operator of the class.

13. The one or more computer readable media of claim 8 wherein the labeling operation comprises:

labeling the data set of related unlabeled patterns as a member of the class, if the similarity measurement associated with the mapping operator of the class satisfies the classification criterion and the similarity measurement satisfies a similarity threshold.

14. The one or more computer readable media of claim 8 wherein the computer process further comprises:

indicating an unsuccessful classification, if no similarity measurement satisfies a similarity threshold.

15. A method of classifying a data set of related unlabeled patterns, the method comprising:

encoding a collection of data sets of patterns to points on a nonlinear parameter space, wherein data sets allocated to a labeled array are grouped in a region of the nonlinear parameter space, each labeled array being designated to a class;

defining a mapping operator for each class that maps the data sets of the labeled array onto the nonlinear parameter space in satisfaction of a similarity criterion based on the encoded points of that labeled array on the nonlinear parameter space;

generating a similarity measurement for each class based on the data set of related unlabeled patterns by mapping the data set of related unlabeled patterns on the nonlinear parameter space using the mapping operator for the class;

labeling the data set of related unlabeled patterns as a member of a class, if the similarity measurement associated with the mapping operator of the class satisfies a classification criterion.

16. The method of claim 15 wherein a similarity measurement associated with the mapping operator of the class satisfies the classification criterion if the similarity measurement is the minimum of all similarity measurements associated with the mapping operators of the classes.

17. The method of claim 15 wherein the operation of encoding a collection of data sets and the operation of encoding the data set of related unlabeled patterns comprise mapping all of the data sets on the nonlinear parameter space using the same encoding algorithm.

18. The method of claim 15 wherein the similarity criterion is satisfied if the data sets of the labeled array are mapped using the mapping operator of the labeled array to the same location, within a defined threshold, on the nonlinear parameter space as the encoded points of that labeled array on the nonlinear parameter space.

19. The method of claim 15 wherein a similarity measurement represents a difference between a point encoded on the nonlinear parameter space from the data set of related unlabeled patterns and a point mapped from the data set of related unlabeled patterns by the mapping operator of the class.

20. The method of claim 15 wherein the labeling operation comprises:

labeling the data set of related unlabeled patterns as a member of a class, if the similarity measurement associated with the mapping operator of the class satisfies the classification criterion and the similarity measurement satisfies a similarity threshold.