Hybridization Method Based ECG Signals Classification

Abstract

The ECG (Electrocardiogram) is the common reliable and easiest to utilize tool for diagnosis of cardiac arrhythmias. Manually diagnosing the arrhythmia beats is very hectic, as the ECG signals are non-linear and produce long records for analysis. It is very difficult for specialists to evaluate time domain features of minute variations, such as lines, intensity & intervals of ECG Signals in pure human judgments. This manuscript discusses an automatic approach to machine learning and the outcome of the initial algorithm identification of five separate heart rhythms. Support Vector Networks (SVN) is utilized to remove the features and besides that Independent Component Analysis ( ICA) is the technique utilized to provide reduction of dimensionality. The kernel support vector machine function works for the tenfold classification and the ECG Signal cross validation. The concept of variance analysis is utilized to select significant features, and accuracy reliability is measured by the assist of Cohen's kappa statistics. The publicly available MIT-BIH database on arrhythmias is utilized to analyze various types of arrhythmias. This is a massive ECG data collection of various types of records and it includes five separate groups of classification arrhythmia such as SupraVEB, Non-ectopic, VEB, Unknown Beat(Ubeat) and Fusion beat(Fbeat). This methodology will produce an efficient tool to check a person's cardiac health which will produce a smart, automated technology for the specialist and paramedics to deal with heart arrhythmia