B. Kulahcioglu, S. Ozdemir, and B. Kumova (Turkey)
Symbolic Time Series Analysis (STA), Piecewise Aggregate Approximation (PAA), Symbolic Aggregate Approximation (SAX), ECG, Coarse Graining
Symbolic Time Series Analysis (STA) is an emerging methodology that involves coarse graining of the signals. Repeating segments of the time series are associated with symbols, thereby reducing the complexity of the series. It facilitates data mining tasks to be performed easily such as indexing, clustering, classification, summarization, and anomaly detection. This study involves symbolization through Symbolic Aggregate Approximation (SAX) with Piecewise Aggregate Approximation (PAA). The same ECG series is symbolized first by PLA and then PAA. Coarsing the series by PLA proved to be more problematic than PAA. At coarser scales, details are lost in noise with PLA, whereas local features become clearer with PAA. However during the analyses of ECGs of various subjects, it is understood that PAA fails when the series is not perfectly periodic as in rotating machinery. This fact is contrasted with the synthetic ECG which is manipulated to be perfectly periodic to juxtapose the results of the two trials. It is deduced that PAA delivers better pattern detection when signals are truly periodic.
Important Links:
Go Back
