Automatic Arrhythmia Detection Using Support Vector Machine Based on Discrete Wavelet Transform

Arrhythmia is abnormal electrical activity in the heart bringing about less effective pumping. An abnormally fast electrical signal initiates two problems: (1) the heart pumps too quick; and (2) ventricles are filled with an inadequate amount of blood. On the other hand, an abnormally slow electrical signal pumps a sufficient amount of blood out of the heart but too slow. Arrhythmia is classified by both its location of origin and rate. Some arrhythmias are life-threatening and eventually result in cardiac arrest. Hence, the purpose of this study is to present a robust implementation algorithm to discriminate between normal sinus rhythm and three types of arrhythmia: atrial fibrillation (AF), ventricular fibrillation (VF), and supra ventricular tachycardia (SVT) that were collected from physionet database. This is attained by capturing the main features that contain both frequency and location information of the signal through discrete wavelet transform, followed by principal component analysis on each decomposed level. Features were reduced through statistical analysis as an input to support vector machine with optimized parameters that resulted in overall accuracy of 96.89%.