Atrial Activity Signal Reconstruction Based on Two-sided AR model

With the rapid development of mobile communication techniques, various mobile health applications are springing up, where the single-lead atrial activity (AA) signal extraction method is urgently required. However, due to QRS residuals, the extracted AA signals by most existing methods are heavily distorted. In this paper, started from signal reconstruction perspective, a novel autoregressive (AR) model based method is presented. Two-sided AA samples around QRS segment are first modeled by AR model, and then the samples in QRS segment are reconstructed using the extrapolation formula. The experimental results on both simulated and real electrocardiograms with atrial fibrillation show that, after using the proposed method, QRS residuals were significantly reduced.