Detection of Atrial Fibrillation using Autoregressive modeling

Abstract

A‎atrial fibrillation (AF) is the arrhythmia that commonly causes death in the adults. We measured AR coefficients using Burg’s method for each 15 second segment of ECG. These features are classified using the different statistical classifiers: kernel SVM and KNN classifier. The performance of the algorithm was evaluated on signals from MIT Physionet database.. The effect of AR model order and data length was tested on the classification results. This method shows better results can be used for practical use in the clinics.‏