Autoregressive Modeling based ECG Cardiac Arrhythmias’ Database System

Abstract

This article proposes an ECG (electrocardiography) database system based on linear filtering, wavelet transform, PSD analysis, and adaptive AR modeling technologies to distinguish 19 ECG beat types for classification. This paper uses the Savitzky-Golay filter and wavelet transform for noise reduction, and wavelet analysis and AR modeling techniques for feature extraction to design a database system of AR coefficients describing the ECG signals with different arrhythmia types. In the experimental part of this work, the proposed algorithm performance is evaluated using an ECG dataset containing 19 different types including normal sinus rhythm, atrial premature contraction, ventricular premature contraction, ventricular tachycardia, ventricular fibrillation, supraventricular tachycardia, and other types from the MIT-BIH Arrhythmia Database. The simulation is performed in a MATLAB environment.