Removal of baseline wander in ECG using the statistical properties of fractional Brownian motion

In this paper, we propose a novel method of removal of baseline wander noise from a noisy ECG signal. We have shown that the original (clean) ECG signal can be modeled as a 2^nd order fractional Brownian motion (fBm) process, while the baseline wander noise can be modeled as a 1^st order fBm process. We use the eigenvectors of the auto-covariance matrix of the 2^nd order fBm process characterizing the original ECG signal to design an M-channel uniformly decimated filterbank. Noisy ECG signal is fed to this filterbank and the subband signals are processed for removal of baseline wander noise. Performance comparison with other baseline wander removal methods shows the efficacy of the proposed method.