Method and computer program product for identifying and correcting systematic noise in a pattern recognition system

• US 20040032981A1
• Filed: 08/13/2002
• Published: 02/19/2004
• Est. Priority Date: 08/13/2002
• Status: Active Grant

• Current Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Original Assignee: Lockheed Martin Corporation (Martin Marietta Corporation)
• Inventors: Ii, David L., Reitz, Elliott D. II, Tillotson, Dennis A.
• Granted Patent: US 7,305,122 B2
• Time in Patent Office: Days
• US Class Current: 382/159
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting

Abstract

The present invention recites a method and computer program product for identifying and correcting systematic noise in a pattern recognition classifier. A plurality of input patterns that have been determined not to be associated with any of a set of at least one represented output class are rejected by the pattern recognition classifier. A subset of pattern samples is selected from the rejected input patterns based upon the similarity of each pattern to one of the represented output classes. The selected pattern samples are subjected to an independent review to determine if they were correctly rejected. The classifier is retrained based upon this independent review of the selected pattern samples.

Claims

1. A method for identifying and correcting systematic noise in a pattern recognition classifier, comprising:

rejecting a plurality of input patterns determined not to be associated with any of a set of at least one represented output class by the pattern recognition classifier;

selecting a subset of pattern samples from the rejected input patterns based upon the similarity of each pattern to one of the represented output classes;

subjecting the selected pattern samples to an independent review to determine if they were correctly rejected; and

retraining the classifier based upon the independent review of the selected pattern samples.

2. A method as set forth in claim 1, wherein the step of subjecting the pattern samples to independent review includes subjecting the clustered pattern samples to a computerized verification process.

3. A method as set forth in claim 1, wherein the step of subjecting the pattern samples to independent review includes outputting them to a human operator.

4. A method as set forth in claim 1, wherein the input patterns include scanned images.

5. A method as set forth in claim 4, wherein at least one of the represented output classes represents a variety of postal indicia.

6. A method as set forth in claim 4, wherein at least one of the represented output classes represents an alphanumeric character.

7. A computer program product, operative in a data processing system, for identifying and correcting systematic noise in a pattern recognition classifier, comprising:

a classification portion that rejects a plurality of input patterns determined not to be associated with any of a set of at least one represented output class;

a selection portion that selects a subset of pattern samples from the rejected input patterns based upon the similarity of each pattern to one of the represented output classes;

a review portion that subjects the selected pattern samples to an independent review to determine if they were correctly rejected; and

a training portion that retrains the classifier based upon the independent review of the selected pattern samples.

8. A computer program product as set forth in claim 7, wherein the review portion submits the pattern samples to a computerized verification process.

9. A computer program product as set forth in claim 7, wherein the review portion outputs the selected pattern samples to a human operator.

10. A computer program product as set forth in claim 7, wherein the input patterns include scanned images.

11. A computer program product as set forth in claim 10, wherein at least one of the represented output classes represents a variety of postal indicia.

12. A computer program product as set forth in claim 10, wherein at least one of the represented output classes represents an alphanumeric character.