ECG Signals Denoising in State Space using UFIR Filtering for Features Extraction

Features extraction from the electrocardiography (ECG) signal measurements are required for medical needs to detect heart abnormalities and different kinds of diseases. One of the standard techniques providing denoising and extracting features of ECG signals employs one-step prediction. We show that better accuracy can be obtained using UFIR filtering and smoothing. We develop the UFIR-based fast algorithms in discrete-time state-space with adaptive optimal averaging horizon, which is required to minimize the mean square error. In this paper, we exploit ECG measurements and artificial data related to normal heartbeats. Higher accuracy of the developed state-space UFIR smoothing algorithm against the traditional prediction-based one is demonstrated experimentally and by simulation.