System or method for selecting classifier attribute types

• US 7,197,180 B2
• Filed: 02/28/2003
• Issued: 03/27/2007
• Est. Priority Date: 05/30/2001
• Status: Expired due to Fees

• Current Assignee: Eaton Corp. (Wisconsin) (Eaton Corporation PLC.)
• Original Assignee: Eaton Corp. (Wisconsin) (Eaton Corporation PLC.)
• Inventors: Farmer, Michael Edward
• Primary Examiner(s): Johns; Andrew W.
• Application Number:
  US10/375,946
• Publication Number:
  US 20030234519A1
• Time in Patent Office: 1,488 Days
• Field of Search: 382/159, 382/160, 382/224, 382/228, 382/260, 382/261
• US Class Current: 382/160
• CPC Class Codes: B60R 16/0231 Circuits relating to the driving or the functioning of the vehicle monitoring tyres B60C23/00; indicating overspeed B60K31/00; for dash boards B60K37/00, B60Q3/10; for indicating emergencies B60Q1/52; brake control systems B60T; registering or indicating the working of vehicles G07C5/00; measuring distance G01C, e.g. combinations of speed and distance G01C23/00; engine indicators G01L; measuring speed or acceleration G01P
• CPC Class Codes: B60R 2021/0044 Chest
• CPC Class Codes: B60R 2021/01315 monitoring occupant displacement
• CPC Class Codes: B60R 21/013 including means for detecting collisions, impending collisions or roll-over
• CPC Class Codes: B60R 21/0134 responsive to imminent contact with an obstacle , e.g. using radar systems
• CPC Class Codes: B60R 21/01538 for image processing, e.g. cameras or sensor arrays
• CPC Class Codes: B60R 21/01542 detecting passenger motion
• CPC Class Codes: B60R 21/01552 detecting position of specific human body parts, e.g. face, eyes or hands
• CPC Class Codes: B60R 21/01556 Child-seat detection systems
• CPC Class Codes: B60R 21/01558 monitoring crash strength
• CPC Class Codes: G06F 18/211 Selection of the most significant subset of features
• CPC Class Codes: G06T 7/215 Motion-based segmentation
• CPC Class Codes: G06T 7/254 involving subtraction of images
• CPC Class Codes: G06T 7/277 involving stochastic approaches, e.g. using Kalman filters
• CPC Class Codes: G06T 7/70 Determining position or orientation of objects or cameras camera calibration G06T7/80
• CPC Class Codes: G06V 10/771 Feature selection, e.g. selecting representative features from a multi-dimensional feature space
• CPC Class Codes: G06V 20/593 Recognising seat occupancy
• CPC Class Codes: G06V 40/10 Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands

Abstract

A system or method (collectively “selection system”) is disclosed for selecting attributes for a classifier in a sensor system. The selection system selects attribute types using statistical distributions of the attribute values associated with those attribute types. Attribute types not within the selectively identified subset of attribute types can be filtered out before such data is sent to a classifier. The system can use a test data subsystem for storing and accessing actual sensor data. A distribution analysis subsystem can perform statistical analyses on the test data to identify underlying distributions, and to compare individual attribute types to such distributions. An attribute selection subsystem, wherein said attribute selection subsystem selectively identifies a subset of attribute types from said subset of attribute types.

Claims

1. A method for configuring a filter for use with a classifier, comprising:

selectively identifying a subset of attribute types from a plurality of attribute types with a distribution statistic;

converting an attribute value associated with each attribute type into a scaled value; and

modifying the filter so that only the subset of attribute types are accessible by the classifier.

2. The method of claim 1, further comprising:

generating the distribution statistic with a Mann-Whitney heuristic.

3. The method of claim 1, further comprising:

generating the distribution statistic from a null hypothesis set of statistics.

4. The method of claim 1, further comprising:

generating the distribution statistic from a number of distribution classes.

5. The method of claim 1, further comprising:

comparing the distribution statistic to a predetermined distribution statistic threshold.

6. The method of claim 1, further comprising:

ranking the distribution statistic; and

eliminating attribute types associated with a ranking that is below a predetermined rank.

7. The method of claim 1, further comprising:

generating a combined class statistic;

ranking the combined class statistic; and

eliminating attribute types from the subset of attribute types that are associated with a ranking that is below a predetermined rank.

8. The method of claim 7, wherein the generating of a combined class statistic includes multiplying the combined class statistic by an importance-based multiple associated with a class.

9. The method of claim 1, further comprising:

identifying a redundant test sample of attribute types by performing a k-newest neighbor heuristic; and

eliminating the attribute types in the redundant test sample from the subset of attribute types.

10. The method of claim 9, wherein the redundant test sample is associated with a confidence factor of 100 by the k-newest neighbor heuristic.

11. The method of claim 1, further comprising:

generating a covariance coefficient from a covariance heuristic;

comparing the covariance coefficient to a predetermined covariance threshold; and

removing attribute types from the subset of attribute types where the attribute type is affiliated with a covariance coefficient tat exceeds the predetermined covariance threshold.

12. The method of claim 1, wherein the filter and classifier are used by an airbag deployment mechanism.

13. The method of claim 12, wherein the attribute types are associated wit a visual image of a person capable of being captured by a video camera.

14. A system for selecting attribute types for inclusion in a classifier, comprising:

a test data subsystem, including a data sample comprising a plurality of attribute types;

wherein said test data subsystem further includes a normalization module, said normalization module including a plurality of scaled values, wherein said normalization module generates said plurality of scaled values, and associates said plurality of scaled values with said plurality of attribute types;

a distribution analysis subsystem, including a plurality of distribution statistics, wherein said distribution analysis subsystem generates said plurality of distribution statistics from said plurality of attribute types; and

an attribute selection subsystem, wherein said attribute selection subsystem selectively identifies a subset of attribute types from said plurality of attribute types.

15. The system of claim 14, wherein said attribute selection subsystem includes a predetermined threshold value, and wherein said subset of attribute types are selectively identified by comparing said distribution statistics with the predetermined threshold value.

16. The system of claim 14, wherein said distribution analysis subsystem further includes a plurality of combined class statistic, wherein said distribution analysis subsystem generates said combined class statistics from said distribution statistics, and wherein said attribute selection subsystem ranks the attribute types by ranking said combined class statistics.

17. The system of claim 14, wherein said attribute selection subsystem selectively identifies a predefined number of attribute types.

18. The system of claim 14, wherein said data sample includes a subset of redundant data, wherein said test data subsystem includes a redundancy heuristic, and wherein said test data subsystem invokes said redundancy heuristic to remove said subset of redundant data.

19. The system of claim 18, wherein said redundancy heuristic is a k-nearest neighbor heuristic.

20. The system of claim 14, wherein said distribution statistics relate to a plurality of classes.

21. The system of claim 14, wherein said distribution statistics are created with a Mann-Whitney heuristic.

22. The system of claim 14, wherein said subset of selectively identified attribute types are embedded in an airbag deployment mechanism.

23. The system of claim 22, wherein each attribute type in said subset includes an attribute value, and wherein said airbag deployment mechanism includes a video camera for capturing said plurality of attribute values relating to said attribute types.