Feature classification for time series data

• US 6,735,550 B1
• Filed: 01/15/2002
• Issued: 05/11/2004
• Est. Priority Date: 01/16/2001
• Status: Expired due to Term

• Current Assignee: University Corporation For Atmospheric Research
• Original Assignee: University Corporation For Atmospheric Research
• Inventors: Goodrich, Robert K., Weekley, Richard A., Cornman, Lawrence B.
• Primary Examiner(s): Barlow, John
• Assistant Examiner(s): Walling, Meagan S
• Application Number:
  US10/047,311
• Time in Patent Office: 847 Days
• Field of Search: 702/182, 702/189, 702/190, 702/167, 707/6, 707/7, 707/10, 707/104.1, 707/200, 707/202, 706/900
• US Class Current: 702/182
• CPC Class Codes: G06F 18/21 Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
• CPC Class Codes: G06F 18/2433 Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
• CPC Class Codes: Y10S 707/99936 Pattern matching access
• CPC Class Codes: Y10S 707/99937 Sorting

Abstract

A processing system uses image processing technology to classify features in time series data. In atmospheric applications, the features can be used to detect outliers and/or failure modes in time series data from atmospheric measurement systems. The processing system is configured to: 1) process the time series data with a plurality of membership functions to generate a plurality of hypersurfaces, 2) process the hypersurfaces to generate a composite hypersurface, 3) process the composite hypersurface to identify clusters, and 4) process the clusters to classify the features.

Claims

1. A method of operating a processing system to classify a feature in time series data, the method comprising:

processing the time series data to generate a plurality of hypersurfaces;

processing the hypersurfaces with a plurality of membership functions to identify a plurality of clusters and to classify the clusters into a plurality of cluster tapes;

processing the clusters from one of the cluster types to construct the feature; and

classifying the feature.

2. The method of claim 1 wherein processing the hypersurfaces to identify the clusters comprises contouring the hypersurfaces to form the clusters.

3. The method of claim 1 wherein the cluster types comprise an atmospheric cluster type.

4. The method of claim 1 wherein the cluster types comprise a failure mode cluster type.

5. The method of claim 1 further comprising calculating feature membership values for the time series data based on the classified feature.

6. The method of claim 5 further comprising detecting outliers in the time series data based the feature membership values.

7. The method of claim 5 wherein the feature membership values indicate confidence values for the time series data.

8. The method of claim 1 wherein the hypersurfaces have a height scale from zero to one.

9. The method of claim 1 wherein processing the hypersurfaces with the membership functions to identify the clusters comprises:

processing the hypersurfaces to generate a composite hypersurface; and

processing the composite hypersurface to identify the clusters.

10. A system to classify a feature in time series data, the system comprising:

an interface configured to receive the time series data; and

a processing system configured to process the time series data to generate a plurality of hypersurfaces, process the hypersurfaces with a plurality of membership functions to identify a plurality of clusters and to classify the clusters into a plurality of cluster types, process the clusters from one of the cluster types to construct the feature, and classify the feature.

11. The system of claim 10 wherein the processing system is configured to contour the hypersurfaces to form the clusters.

12. The system of claim 10 wherein the cluster types comprise an atmospheric cluster type.

13. The system of claim 10 wherein the cluster types comprise a failure mode cluster type.

14. The system of claim 10 wherein the processing system is configured to calculate feature membership values for the time series data based on the classified feature.

15. The system of claim 14 wherein the processing system is configured to detect outliers in the time series data based the feature membership values.

16. The system of claim 14 wherein the feature membership values indicate confidence values for the time series data.

17. The system of claim 10 wherein the hypersurfaces have a height scale from zero to one.

18. The system of claim 10 wherein:

the processing system is configured to process the hypersurfaces to generate a composite hypersurface, and process the composite hypersurface to identify the clusters.

19. A software product to classify a feature in time series data, the software produce comprising:

application software configured to direct a processing system to process the time series data to generate a plurality of hypersurfaces, process the hypersurfaces with a plurality of membership functions to identify a plurality of clusters and to classify the clusters into a plurality of cluster types, process the clusters from one of the cluster types to construct the feature, and classify the feature; and

a storage system that stores the application software.

20. The software product of claim 19 wherein the application software is configured to direct the processing system to contour the hypersurfaces to form the clusters.

21. The software product of claim 19 wherein the cluster types comprise an atmospheric cluster type.

22. The software product of claim 19 wherein the cluster types comprise a failure mode cluster type.

23. The software product of claim 19 wherein the application software is configured to direct the processing system to calculate feature membership values for the time series data based on the classified feature.

24. The software product of claim 23 wherein the application software is configured to direct the processing system to detect outliers in the time series data based the feature membership values.

25. The software product of claim 23 wherein the feature membership values indicate confidence values for the time series data.

26. The software product of claim 19 wherein the hypersurfaces have a height scale from zero to one.

27. The software product of claim 19 wherein:

the application software is configured to direct the processing system to process the hypersurfaces to generate a composite hypersurface, and process the composite hypersurface to identify the clusters.