Automatic ECG Image Classification Using HOG and RPC Features by Template Matching

Cardiac disease is the most dangerous killer all over the world. Electrocardiogram plays a significant role for cardiac disease diagnosis. In this work with the advent of image processing technology, a confirmative tool is developed for heart disease diagnosis. The proposed work demonstrates an automatic classification system of ECG images using Histogram of Oriented Gradients (HOG) and Row Pixel Count (RPC) features. The intention of this work is to classify three major types of cardiac diseases namely Arrhythmia, Myocardial Infarction, and Conduction Blocks by template matching. The experiments were conducted on the Physiobank dataset of both normal and abnormal patients. A comparison is made for the experimental results obtained using HOG and RPC, and the performance is studied. The HOG gives a better performance of 94.0 % accuracy.