Stationary Wavelet Transform and Entropy-Based Features for ECG Beat Classification

Abstract

In this study, heartbeats are classified as normal, right bundle branch block (Rbbb), paced beat, and left bundle branch block (Lbbb), using the electrocardiography (ECG) signals from the MIT-BIH arrhythmia database. The statistical parameters and entropy of stationary wavelet transform (SWT) coefficients are proposed as the features for the classification. The classification was performed using artificial neural networks. It was observed that both statistical parameters and the entropy features perform better than the time domain, and the discrete time wavelet transform coefficient features. The classification performance of the different wavelet families is also considered.