ECG-based Arrhythmia Detection using Average Energy and Zero-crossing Features with Support Vector Machine

Arrhythmia, a serious heart disease, is one of the main causes of sudden cardiac death around the globe. Cardiac arrhythmias can occur to anyone and at any age. The electrocardiogram (ECG) arrhythmia is a series of the abnormal electrical pulses recorded from heart by using electrocardiograph. In this study, we present a method for arrhythmia detection from the ECG signal based on the average energy and zero-crossing quantities that are extracted from the ECG records. The proposed system is composed of ECG signal preprocessing, feature extraction, and arrhythmia detection. The ECG signals used in this study were obtained from MIT-BIH arrythmia, MIT-BIH normal sinus rhythm and QT databases. To distinguish between the normal and abnormal (arrhythmia) ECG signal, we implement a support vector machine (SVM) as a classifier. K-fold cross validation technique is employed to evaluate and validate the performance of the proposed system. We measure the performance of the system via accuracy, sensitivity, specificity, and precision. The best detection results based on the average energy feature with 10-fold cross validation presents are 96.67 % average accuracy, 93.33% sensitivity, 100% specificity and 100% precision.