Comparison of neural network, ANFIS, and SVM classifiers for PVC arrhythmia detection

Premature ventricular contraction (PVC) beats are of great importance in evaluating and predicting life threatening ventricular arrhythmias. The aim of this study is to improve the diagnosis level of detection of PVC arrhythmia from ECG signals. This improvement is based on an appropriate choice of features for the selected task. We extracted fourteen features including, temporal, morphological features from MIT/BIH ECG signals database and then applying LDA algorithm, we reduced them into nine features. Finally we use a Neural Network, an ANFIS, and a SVM as classifiers. Satisfactory result obtained with accuracy rates of 99.8% for Neural Network classifier, 94.8673% for ANFIS classifier, and 97.57 for SVM classifier.