Accelerating Quasi Periodic Dynamic Time Warping For Similarity Search in ECG Time Series

Imen, Boulnemour; Bachir, Boucheham; Abdelmadjid, Lahreche

المجلة الجزائرية للعلوم
Volume 7, Numéro 14, Pages 24-29
2021-06-03

Accelerating Quasi Periodic Dynamic Time Warping For Similarity Search In Ecg Time Series

Authors : Imen Boulnemour . Bachir Boucheham . Abdelmadjid Lahreche .

Abstract

Several methods have been proposed for similarity search in time series (TS). Our previouslypublishedmethodQP-DTW(QuasiPeriodicDynamicTimeWarping)has beenproposedasarivalmethodtoDTWforquasiperiodicTSalignment.However the method is expensive in terms of time complexity. Consequently it is not appropriateforsimilaritysearchinTS,wherethemethodswithlesstimecomplexity areneeded.InthispaperweproposeanacceleratedversionoftheQP-DTWmethod and we use it to solve the problem of searching similar ECG (Electrocardiogram) time series in a data base. We have focused on ECG TS for its utility in various medical domains, including Diagnostic of cardiac disease and biometry. The proposed Accelerated QP-DTW (A-QPDTW) used Haar Wavelet Transform as a techniquetoreducethedimensionalityoftheTSandtheKNN(Knearestneighbor) approach to search for TS that are similar to a given query. Experimental results show that the proposed method is more efficient than QPDTW method with comparativeaccuracy.AllresultshavebeenobtainedusingtheECGTSoftheUCR (University of California at Riverside) time series Dataset, universally admittedas the first Benchmark in time series analyses.