Multivariate autoregressive modeling for cardiac arrhythmia classification using multilayer perceptron neural networks

This paper presents an effective electrocardiogram (ECG) arrhythmia classification scheme including three modules. In the first module we investigate the application of a finite impulse response (FIR) least squares filter for noise reduction of the ECG signals. The features extraction module explores the ability of multivariate autoregressive (MVAR) modeling to extract relevant features from two-lead electrocardiogram signals in order to classify certain cardiac arrhythmias. Then a number of multilayer perceptron (MLP) neural networks with different number of layers and seven training algorithms are designed. The performances of the networks for speed of convergence and accuracy classifications are evaluated for various ECG data types including normal sinus rhythm, atrial premature contraction, premature ventricular contraction, ventricular tachycardia and supraventricular tachycardia obtained from the MIT-BIH database. The overall classification accuracy of the proposed scheme is 99.7%. Our experimental results have successfully validated that the integration of the proposed features extraction method with the MLP classifier can achieve satisfactory classification accuracy.