Wavelet feature extraction for ECG beat classification

Electrocardiography (ECG) signal is a bioelectrical signal which depicts the cardiac activity of the heart. It is a technique used primarily as a diagnostic tool for various cardiac diseases. ECG provides necessary information on the electrophysiology and changes that may occur in the heart. Due to the increase in mortality rate associated with cardiac diseases worldwide despite recent technological advancement, early detection of these diseases is of paramount importance. This paper has proposed a robust ECG feature extraction technique suitable for mobile devices by extracting only 200 samples between R-R intervals as equivalent R-T interval using Pan Tompkins algorithm at preprocessing stage. The discrete wavelet transform (DWT) of R-T interval samples are calculated and the statistical parameters of wavelet coefficients such as mean, median, standard deviation, maximum, minimum, energy and entropy are used as a time-frequency domain feature. The proposed hybrid technique has been tested by classifying three ECG beats as normal, right bundle branch block (Rbbb) and paced beat using the signals from Massachusetts Institute of Technology Beth Israel Hospital (MIT-BIH) arrhythmia database and processed using Matlab 2013 environment. Classification has been performed using neural network backpropagation algorithm because of its simplicity. While equivalent R-T interval features gives average accuracy of 98.22%, the proposed hybrid method gives a promising result with average accuracy of 99.84% with reduced classifier computational complexity.