Method and system for pattern recognition in a signal using morphology aware symbolic representation

Patent metadata

Field	Value
Publication number	US-11263450-B2
Application number	US-201916264786-A
Country	US
Kind code	B2
Filing date	Feb 1, 2019
Priority date	Mar 7, 2018
Publication date	Mar 1, 2022
Grant date	Mar 1, 2022

Abstract

The present disclosure addresses the technical problem of information loss while representing a physiological signal in the form of symbols and for recognizing patterns inside the signal. Thus making it difficult to retain or extract any relevant information which can be used to detect anomalies in the signal. A system and method for anomaly detection and discovering pattern in a signal using morphology aware symbolic representation has been provided. The system discovers pattern atoms based on the strictly increasing and strictly decreasing characteristics of the time series physiological signal, and generate symbolic representation in terms of these pattern atoms. Additionally the method possess more generalization capability in terms of granularity. This detects discord/abnormal phenomena with consistency.

First claim

What is claimed is:

1. A method for detecting anomaly and discovering pattern in a signal using symbolic representation of the signal, the method comprising a processor implemented steps of:
sensing the signal from a person using a sensor, wherein the signal is a time series signal with a plurality of time points ( 202 );
finding a plurality of maxima points and a plurality of minima points in the signal, wherein the plurality of maxima points and the plurality of minima points are following morphology of the signal ( 204 );
deriving a plurality of features using the plurality of maxima points and the plurality of minima points which are adjacent to each other, wherein the plurality of features comprises amplitude difference and number of sampling points between minima to maxima points which are adjacent to each other and amplitude difference and number of sampling points between maxima to minima points which are adjacent to each other ( 206 );
performing a proximity based clustering on the plurality of features to get cluster centroid values corresponding to each of the plurality of features ( 208 );
representing the cluster centroid values in a form of a plurality of symbols in descending order in terms of the amplitude difference and the number of sampling points corresponding to minima to maxima points and maxima to minima points ( 210 );
representing the signal into the symbolic form using the plurality of symbols, wherein the symbolic form comprising one or more strings of symbols ( 212 );
deriving a dissimilarity metric, wherein the dissimilarity metric is a function of: regularity of the strings of symbols, and
a distance measure between the string of symbols for detecting anomalous patterns ( 214 ); and
detecting the anomalous pattern in the signal using the dissimilarity metric ( 216 ).

2. The method of claim 1 , further comprising: representing the string of symbols using minima to maxima points and maxima to minima points encoded by their corresponding occurrences of cluster centroid values in their consecutive order of occurrences.

3. The method of claim 2 , wherein the encoding of occurrence of consecutive minima to maxima points and consecutive maxima to minima points are based on a pair of symbols known as pattern atoms.

4. The method of claim 3 , wherein the string of symbols further comprising pattern atom with highest regularity in their occurrences.

5. The method of claim 4 , further comprising measuring regularity in the interval of the frequency of occurrences of the pattern atoms.

6. The method of claim 1 , wherein the step of representing the signal further comprising segmenting the string of symbols using consistent pattern atom.

7. The method of claim 1 , wherein the step of representing the signal further comprising merging of string of symbols with a corresponding plurality of segments.

8. The method of claim 1 , wherein the plurality of features further comprising number of points from adjacent minima to maxima and a number of points from adjacent maxima to minima.

9. The method of claim 1 , wherein the plurality of maxima points and the plurality of minima points are exploiting strictly rising and falling edges of the signal.

10. The method of claim 1 , further comprising the step of preprocessing the sensed signal, wherein the sensed signal is a physiological signal captured from a physiological sensor.

11. The method of claim 1 , wherein the step of performing the proximity based clustering further comprising:
receiving the plurality of features derived from the signal;
setting a value for maximum number of clusters formed for the plurality of derived features;
performing at least one step based on the distribution of plurality of features:
if the plurality of features has Gaussian distribution, then
dividing the plurality of features into K equi-probable regions of Gaussian distribution and finding the cluster centroids of the derived features,

else,
arranging the plurality of features in the ascending order and finding the first difference of data stored in dataset of derived features;
breaking the first difference of data stored in dataset of derived features into two clusters where maximum difference occurs; and
checking a condition for the number of cluster.

12. A system for detecting anomaly and pattern discovery in a signal using symbolic representation of the signal, the system comprising:
a sensor ( 102 ) for sensing the signal from a person, wherein the signal is a time series signal with a plurality of time points;
a memory ( 106 ); and
a processor ( 108 ) in communication with the memory ( 106 ), the processor ( 108 ) is configured to:
find a plurality of maxima points and a plurality of minima points in the signal, wherein the plurality of maxima points and the plurality of minima points are following morphology of the signal;
derive a plurality of features using the plurality of maxima points and the plurality of minima points which are adjacent to each other, wherein the plurality of features comprises an amplitude difference and a number of sampling points between minima to maxima points which are adjacent to each other and the amplitude difference and a number of sampling points between maxima to minima points which are adjacent to each other;
perform a proximity based clustering on the plurality of features to get a cluster centroid values corresponding to each of the plurality of features;
represent the cluster centroid values in a form of a plurality of symbols in descending order in terms of the amplitude difference and the number of sampling points corresponding to minima to maxima points and maxima to minima points, and represent the signal into the symbolic form using the plurality of symbols, wherein the symbolic form comprising one or more string of symbols;
derive a dissimilarity metric, wherein the dissimilarity metric is a function of regularity of the strings of symbols, and
a distance measured between the string of symbols for detecting anomalous patterns; and
detect the anomalous pattern in the signal using the dissimilarity metric.

13. A computer program product comprising a non-transitory computer readable medium having a computer readable program embodied therein, wherein the computer readable program, when executed on a computing device, causes the computing device to:
sense the signal from a person using a sensor, wherein the signal is a time series signal with a plurality of time points ( 202 );
find a plurality of maxima points and a plurality of minima points in the signal, wherein the plurality of maxima points and the plurality of minima points are following morphology of the signal ( 204 );
derive a plurality of features using the plurality of maxima points and the plurality of minima points which are adjacent to each other, wherein the plurality of features comprises amplitude difference and number of sampling points between minima to maxima points which are adjacent to each other and amplitude difference and number of sampling points between maxima to minima points which are adjacent to each other ( 206 );
perform a proximity based clustering on the plurality of features to get cluster centroid values corresponding to each of the plurality of features ( 208 );
represent the cluster centroid values in a form of a plurality of symbols in descending order in terms of the amplitude difference and the number of sampling points corresponding to minima to maxima points and maxima to minima points ( 210 );
represent the signal into the symbolic form using the plurality of symbols, wherein the symbolic form comprising one or more strings of symbols (

Assignees

• Tata Consultancy Services Ltd

Inventors

• Bandyopadhyay Soma
• Ukil Arijit
• Puri Chetanya
• Singh Rituraj
• Pal Arpan
• Murthy C A

Classifications

• G06F2218/10

  by analysing the shape of a waveform, e.g. extracting parameters relating to peaks · CPC title

• G06F2218/12

  Classification; Matching · CPC title

• G06F18/23211 Primary

  with adaptive number of clusters · CPC title

• G06F18/23213

  with fixed number of clusters, e.g. K-means clustering · CPC title

• G06K9/00536

  Physics · mapped topic