Electrocardiogram features detection using stationary wavelet transform

Abstract

The main objective of this paper is to provide a novel stationary wavelet transform (SWT) based method for electrocardiogram (ECG) feature detection. The proposed technique uses the detail coefficients of the ECG signal decomposition by SWT and the selection of the appropriate coefficient to detect a specific wave of the signal. Indeed, the temporal and frequency analysis of these coefficients allowed us to choose detail coefficient of level 2 (Cd2) to detect the R peaks. In contrast, the coefficient of level 3 (Cd3) is determined to extract the Q, S, P, and T waves from the ECG. The proposed method was tested on recordings from the apnea and Massachusetts Institute of Technology–Beth Israel hospital (MIT-BIH) databases. The performances obtained are excellent. Indeed, the technique presents a sensitivity of 99.83%, a predictivity of 99.72%, and an error rate of 0.44%. A further important advantage of the method is its ability to detect different waves even in the presence of baseline wander (BLW) of the ECG signal. This property makes it possible to bypass the filtering operation of BLW.