Multiwavelet Feature Sets for ECG Beat Classification

Abstract

In this study heart beats are classified as normal right branch block left branch block and paced rhythm using electrocardiographic (ECG) signals obtained from the MIT-BIH cardiac arrhythmia database. Average standard deviation energy and entropy of discrete wavelet transform (DWT) coefficients are proposed as the features for the classification. The classification was performed by selecting the appropriate train function of artificial neural network. It has been observed that the combined use of two different feature sets from two wavelet families with the appropriate train function of neural network is better than use of individual wavelets features.