Classification of ECG Signals Related to Paroxysmal Atrial Fibrillation

Paroxysmal atrial fibrillation is a life threatening arrhythmia which leads to sudden cardiac death. Cardiac professionals are always looking to obtain a maximum accuracy in identifying and treating heart disorders. The new method of automatic feature extraction and classification of paroxysmal atrial fibrillation is proposed in this paper. The first step toward classifying paroxysmal disorder is to decompose the ECG signals (healthy and unhealthy) using wavelet transformation techniques. Corresponding to these decomposed levels, the values of ECG signals are computed on the basis of entropy by using the method of cross recurrence quantification analysis. The classification was implemented by probabilistic neural network (PNN) concept. Overall gained accuracy by using PNN classifier is 86.6%. The purpose of this work is to develop a smart method for the proper classification of paroxysmal AF arrhythmias. Long-Term AF Database (Itafdb) and MIT-BIH Fantasia Database (fantasia) have been chosen from Physio Bank ATM for carrying out this work.