System and method for categorical analysis of time dependent dynamic processes

• US 20040243328A1
• Filed: 05/26/2004
• Published: 12/02/2004
• Est. Priority Date: 05/28/2003
• Status: Active Grant

• Current Assignee: Henry M Jackson Foundation For The Advancement of Military Medicine Incorporated
• Original Assignee: Christopher Joseph Cellucci, Paul Ernest Rapp, Tanya Schmah
• Inventors: Rapp, Paul Ernest, Schmah, Tanya, Cellucci, Christopher Joseph
• Granted Patent: US 7,117,108 B2
• Time in Patent Office: Days
• US Class Current: 702/71
• CPC Class Codes: A61B 5/349 Detecting specific parameters of the electrocardiograph cycle
• CPC Class Codes: A61B 5/374 Detecting the frequency distribution of signals, e.g. detecting delta, theta, alpha, beta or gamma waves
• CPC Class Codes: A61B 5/377 using evoked responses
• CPC Class Codes: A61B 5/4094 Diagnosing or monitoring seizure diseases, e.g. epilepsy
• CPC Class Codes: A61B 5/7264 Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems neural networks per se G06N3/00; expert systems per se G06N5/00
• CPC Class Codes: A61B 5/7267 involving training the classification device
• CPC Class Codes: G06F 2218/00 Aspects of pattern recognition specially adapted for signal processing
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: Y02A 90/10 Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

A system and method for performing a categorical analysis on one or more time dependent dynamic processes is provided. A reference library of data pertaining to multiple characteristics of time series reflective of the dynamic process is created and used to define selected categories for performing the categorical analysis.

Claims

1. A method for performing a categorical analysis on one or more time dependent dynamic processes, the method comprising:

capturing a time series associated with a selected type of dynamic process;

based on the selected type of dynamic process, selecting a reference library of data of different quantifiable dynamic characteristics of reference time series, wherein the data is segmented into at least one classification of groups where each classification'"'"'s segmentation into groups represents a correlation of quantified dynamic data of a predefined set of a plurality of quantifiable dynamic classification characteristics from among the different quantifiable dynamic characteristics;

processing quantified dynamic data of the captured time series for the predefined set of quantifiable dynamic classification characteristics of a selected classification of groups within the reference library; and

classifying the captured time series with respect to the groups of the selected classification based on the processed quantified dynamic data of the captured time series.

2. The method of claim 1 where the time series are bio-metric signals and a first set of quantifiable dynamic classification characteristics for a first classification of groups includes at least one characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion wherein the processing of quantified dynamic data of the captured time series is done for the first set of quantifiable dynamic classification characteristics of the first classification of groups within the reference library and the classifying of the captured time series with respect to the groups of the first classification.

3. The method of claim 2 where a second set of quantifiable dynamic classification characteristics for a second classification of groups includes at least two characteristics from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion wherein the processing of quantified dynamic data of the captured time series is also done for the second set of quantifiable dynamic classification characteristics of the second classification of groups within the reference library and the classifying of the captured time series is also done with respect to the groups of the second classification.

4. The method of claim 2 wherein:

the classifying of the captured time series with respect to the groups of the selected classification based on the processed quantified dynamic data of the captured time series includes calculating the probability of the processed quantified dynamic data of the captured time series is a member of each group of the first classification and classifying the captured time series in the group for which a highest probability of membership is calculated.

5. The method of claim 4 wherein the calculating of the probability of the processed quantified dynamic data of the captured time series is a member of each group of the selected classification includes calculating probabilities using at least one criteria from the group consisting of minimum Mahalanobis distance or maximum Bayesian likelihood.

6. The method of claim 1 where a first set of quantifiable dynamic classification characteristics for a first classification of groups includes at least two characteristics from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion wherein the processing of quantified dynamic data of the captured time series is done for the first set of quantifiable dynamic classification characteristics of the first classification of groups within the reference library and the classifying of the captured time series with respect to the groups of the first classification.

7. The method of claim 1 wherein a first set of quantifiable dynamic classification characteristics for a first classification of groups includes fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion wherein the processing of quantified dynamic data of the captured time series is done for the first set of quantifiable dynamic classification characteristics of the first classification of groups within the reference library and the classifying of the captured time series with respect to the groups of the first classification.

8. The method of claim 1 wherein the classifying of the captured time series with respect to the groups of the selected classification based on the processed quantified dynamic data of the captured time series includes calculating the probability of the processed quantified dynamic data of the captured time series is a member of each group of the selected classification and classifying the captured time series in the group for which a highest probability of membership is calculated.

9. The method of claim 1 wherein the calculating of the probability of the processed quantified dynamic data of the captured time series is a member of each group of the selected classification includes calculating probabilities using at least one criteria from the group consisting of minimum Mahalanobis distance or maximum Bayesian likelihood.

10. The method of claim 1 wherein the classifying of the captured time series with respect to the groups of the selected classification based on the processed quantified dynamic data of the captured time series includes calculating the probability of the processed quantified dynamic data of the captured time series is a member of each group of the selected classification and classifying the captured time series as a member of each group for which a calculated probability of membership exceeds a membership threshold.

11. The method of claim 1 wherein the classifying of the captured time series with respect to the groups of the selected classification based on the processed quantified dynamic data of the captured time series includes calculating the probability of the processed quantified dynamic data of the captured time series is a member of each group of the selected classification and classifying the captured time series as not a member of each group for which a calculated probability of membership does not exceed a membership threshold.

12. The method of claim 1 where the dynamic process is locomotive behavior and the time series are movement trajectories and a first set of quantifiable dynamic classification characteristics for a first classification of groups includes at least one characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion wherein the processing of quantified dynamic data of the captured time series is done for the first set of quantifiable dynamic classification characteristics of the first classification of groups within the reference library and the classifying of the captured time series with respect to the groups of the first classification.

13. The method of claim 1 where the dynamic process is brain wave activity and the time series are electroencephalograms (EEGs) and a first set of quantifiable dynamic classification characteristics for a first classification of groups includes at least one characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion wherein the processing of quantified dynamic data of the captured time series is done for the first set of quantifiable dynamic classification characteristics of the first classification of groups within the reference library and the classifying of the captured time series with respect to the groups of the first classification.

14. The method of claim 1 where the dynamic process is heart activity and the time series are electrocardiograms (ECGs) and a first set of quantifiable dynamic classification characteristics for a first classification of groups includes at least one characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion wherein the processing of quantified dynamic data of the captured time series is done for the first set of quantifiable dynamic classification characteristics of the first classification of groups within the reference library and the classifying of the captured time series with respect to the groups of the first classification.

15. A system for categorical analysis of time dependent dynamic processes, the system comprising:

a processor having an input for receiving data of quantified dynamic characteristics of a captured time series associated with a selected type of dynamic process;

a memory unit coupled with the processor;

the memory unit including a reference library of data of different quantifiable dynamic characteristics of reference time series associated with the selected type of dynamic process wherein the data is segmented into at least one classification of groups where each classification'"'"'s segmentation into groups represents a correlation of quantified dynamic data of a predefined set of a plurality of quantifiable dynamic classification characteristics from among the different quantifiable dynamic characteristics; and

the processor being configured to process quantified dynamic characteristic data of the captured time series for the set of classification characteristics of a selected classification of groups within the reference library to classify the captured time series with respect to the groups of the selected classification based on the processed quantified dynamic characteristic data.

16. The invention of claim 15 where the time series are bio-metric signals and a first set of quantifiable dynamic classification characteristics for a first classification of groups includes at least one characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion wherein the processor is configured to process quantified dynamic data of the captured time series for the first set of quantifiable dynamic classification characteristics of the first classification of groups within the reference library to classify the captured time series with respect to the groups of the first classification by calculating the probability of the processed quantified dynamic data of the captured time series is a member of each group of the first classification and classifying the captured time series in the group for which a highest probability of membership is calculated.

17. The invention of claim 16 wherein the processor is configured to calculate the probability of the processed quantified dynamic data of the captured time series is a member of each group of the selected classification by using at least one criteria from the group consisting of minimum Mahalanobis distance or maximum Bayesian likelihood.

18. The invention of claim 15 where the dynamic process is locomotive behavior, the time series are movement trajectories and a first set of quantifiable dynamic classification characteristics for a first classification of groups includes at least one characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion wherein the processor is configured to process quantified dynamic data of the captured time series for the first set of quantifiable dynamic classification characteristics of the first classification of groups within the reference library to classify the captured time series with respect to the groups of the first classification.

19. The method of claim 15 where the dynamic process is brain wave activity, the time series are electroencephalograms (EEGs) and a first set of quantifiable dynamic classification characteristics for a first classification of groups includes at least one characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion wherein the processor is configured to process quantified dynamic data of the captured time series for the first set of quantifiable dynamic classification characteristics of the first classification of groups within the reference library to classify the captured time series with respect to the groups of the first classification.

20. The method of claim 15 where the dynamic process is heart activity, the time series are electrocardiograms (ECGs) and a first set of quantifiable dynamic classification characteristics for a first classification of groups includes at least one characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion wherein the processor is configured to process quantified dynamic data of the captured time series for the first set of quantifiable dynamic classification characteristics of the first classification of groups within the reference library to classify the captured time series with respect to the groups of the first classification.

21. A method for creating a reference library of data for use in performing categorical analysis of time dependent dynamic processes, the method comprising:

capturing reference time series associated with a selected type of dynamic process;

deriving data of selected quantifiable dynamic characteristics for each reference time series; and

segmenting the data into at least one classification of groups where each classification'"'"'s segmentation into groups represents a correlation of data of a predefined set of quantified dynamic classification characteristics from among the selected quantified dynamic characteristics whereby a subject captured time series of the selected type of dynamic process can be classified into a group of the selected classification based on correlating quantified dynamic data derived from the subject captured time series corresponding to the predefined set of quantified dynamic classification characteristics of the selected classification of groups.

22. The method of claim 21 further comprising defining a first set of quantifiable dynamic classification characteristics for a first classification of groups by:

deriving data for at least three quantifiable dynamic characteristics for at least a selected minimum number of reference time series from each group of the first classification of groups;

calculating effectiveness with respect to each quantifiable dynamic characteristic based on the respective data derived for the reference time series of each respective group; and

selecting at least two quantifiable dynamic characteristics that have the highest calculated effectiveness for inclusion in the first set of quantifiable dynamic classification characteristics for the first classification of groups.

23. The method of claim 22 where the time series are bio-metric signals and wherein data is derived for at least one quantifiable dynamic characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion.

24. The method of claim 22 wherein data is derived for quantifiable dynamic characteristics including fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion.

25. The method of claim 22 further comprising defining a second set of quantifiable dynamic classification characteristics for a second classification of groups by:

deriving data for at least three quantifiable dynamic characteristics for at least a selected minimum number of reference time series from each group of the second classification of groups;

calculating effectiveness with respect to each quantifiable dynamic characteristic based on the respective data derived for the reference time series of each respective group of the second classification of groups; and

selecting at least two quantifiable dynamic characteristics that have highest calculated effectiveness for inclusion in the second set of quantifiable dynamic classification characteristics for the second classification of groups.

26. The method of claim 25 where the time series are bio-metric signals and wherein data is derived for defining the first set of quantifiable dynamic classification characteristics is also used for defining the second set of quantifiable dynamic classification characteristics where a first reference time series is included in a group of the first classification of groups and also a group of the second classification of groups.

27. The method of claim 26 wherein data is derived for at least one quantifiable dynamic characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion.

28. The method of claim 26 wherein data is derived for quantifiable dynamic characteristics including fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion.

29. The method of claim 22 where the dynamic process is locomotive behavior and the time series are movement trajectories and wherein data is derived for at least one quantifiable dynamic characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion.

30. The method of claim 22 where the dynamic process is brain wave activity and the time series are electroencephalograms (EEGs) and wherein data is derived for at least one quantifiable dynamic characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion.

31. The method of claim 22 where the dynamic process is heart activity and the time series are electrocardiograms (ECGs) and wherein data is derived for at least one quantifiable dynamic characteristic from the group consisting of fractal dimension, Richardson dimension, Lempel-Ziv complexity, Hurst exponent and relative dispersion.

32. The method of claim 22 wherein the calculating effectiveness with respect to each quantifiable dynamic characteristic based on the respective data derived for the reference time series of each respective group includes calculating partial F-values with respect to each quantifiable dynamic characteristic based on the respective data derived for the reference time series of each respective group.