ABSTRACT
Atrial fibrillation (AF) is the most common arrhythmia in clinical practice. The paper introduces a new method for ventricular activity cancellation in AF from surface electrocardiogram (ECG) signals. The proposed method is based on AF signal extraction using adaptive echo state neural network (ESN). Adaptive ESN estimates a time-varying, nonlinear transfer function between two ECG leads and separates ventricular activity from atrial activity. The method was compared with conventional pre-whitened recursive least squares (RLS) based linear adaptive filter. Both algorithms were applied to surrogate ECG data with known component of AF signal. The results revealed that the ESN based nonlinear filter extracts f-waves more accurately than the conventional pre-whitened RLS based linear algorithm, especially in lower amplitude (< 0.05 mV) AF signals.
