Cardiac Arrhythmia Detection using

Cardiac Arrhythmia is a growing health problem. According to American Heart Association by 2030 there will be at least 9 million cases of Atrial Fibrillation in the United States. The recent advances in telemedicine industry gives us an opportunity to address this concern. There have been many algorithms proposed to detect Cardiac Arrhythmia, and most of them are solely based on heartrate variability. The 2017 PhysioNet/CinC Challenge entitled “AF Classification from a short single lead ECG recording” has been designed to find an optimal solution for this problem. In this thesis we suggest a new adaptive Electrocardiogram modeling to extract the features of the signal. Using extracted features from the competition’s training dataset using this method and applying gaussian SVM with 5-fold cross-validation, we achieved 72% accuracy rate.