Arrhythmia Recognition and Classification using Modified ECG Morphology and Signal Feature Extraction Analysis

Abstract

Arrhythmia appearing with the presence of abnormal heart electrical activity is efficiently recognized and classified. ECG data set used isderived from the MITBIH in which ECG signals are divided into the three classes: normal beat (N), Tachycardia (fast heartbeat), Bradycardia (slow heartbeat). Our proposed method can distinguish these with an accuracy of 97.80%.The morphological features are
extracted along with basic signal features. The features are supplied to Generalized Regression Neural Network (GRNN) for classification. The sensitivities for the classes are 99.27%, 87.47%, 94.71% and the positive predictivities are 98.48%, 95.25%, 95.22% respectively. The detection sensitivity of the has a better performance by combining proposed features than by using the ECG morphology. The proposed method is compared with four selected peer algorithms and delivers solid results.