Electrocardiogram Signal Classification for the Detection of Abnormalities Using Discrete Wavelet Transform and Artificial Neural Network Back Propagation Algorithm

Abstract

Analysis of electrocardiogram signals for detecting the cardiac abnormalities by utilizing the discrete wavelet transform (DWT) and the artificial neural network with (BPA) back propagation algorithm is being preceded in this study. The proposed algorithm is utilized for detecting the abnormality condition in the electrocardiogram signal sample and perform the classification into two various classes such as normal type and abnormal type of classifying the ECG signal. The ECG sample data have been acquired from the MIT-BIH arrhythmia database. Among the 48 files in MIT-BIH cardiac arrhythmia database, 45 files of each single minute of recording have been chosen for acquisition. And among the 45 files, 25 number of files are being determined as the normal category of ECG signal, and the remaining 20 number of files are being determined as the abnormal category of ECG signal. Once the ECG is being acquired from the database, preprocessing of the signal is being undergone with which the noise is removed. After the process of denoising, extraction of features is being carried out under two different sections, one is the morphological features of electrocardiogram signal, and the other one is the features selected on the basis of discrete wavelet transform which are combinedly given as an input parameter to the classifier section. Artificial neural network is being depicted as the classifier of ECG signals which utilizes the back propagation algorithm. The analysis over its classification performance metrics is being made over calculating the sensitivity, specificity, positive predictivity, and accuracy percentage. The classification over normal and the abnormal category has been resulted with the accuracy of 98.21% using the artificial neural network back propagation algorithm.