Muscle artifact cancellation in ECG signal using a dynamical model and particle filter

Muscle artifact cancellation from Electrocardiogram (ECG) signal is an important task in the field of signal processing Because of its non-Gaussian non-stationary nature. Hence, in this paper a nonlinear Bayesian filtering Framework for ECG Denoising is proposed. This algorithm has the advantage of handling non-Gaussian non-stationary situations over other conventional Bayesian frameworks such as Exteneded Kalman Filter (EKF) and Extended Kalman Smoother (EKS). A particle weighting strategy is also proposed to automatically control the reliance of the particle filter to the acquired measurements even in low input SNRs in which neither the ECG model nor the obtained measurements are trustworthy. We evaluated the proposed filter on several normal ECG signals selected from MIT-BIH normal sinus rhythm database by artificially adding non-stationary muscle artifact (MA) noise over a range of low SNRs from 10 to -5dB. Owing to its nonlinear framework and particle weighting scheme, our algorithm attained better results in all input SNRs in comparison to EKF and EKS frameworks.