Removal of Baseline Wander in ECG Signals Using Singular Spectrum Analysis

A technique of removal baseline wander (BW) from electrocardiogram (ECG) signals based on singular spectrum analysis (SSA) is presented. With SSA applied, the ECG signal can be decomposed into trends, oscillations or noise components based on the singular value decomposition (SVD). By applying frequency analysis on the factor vectors which are produced by SVD, the decomposition components that may be interpretable as basic trend is selected to reconstruct the BW signal and then removal it from the ECG signal. The proposed method is compared with the state-of-the-art BW removal methods utilizing Hilbert vibration decomposition and multivariate empirical mode decomposition. The simulations were performed using real BW and clinical ECG signals and the results show that the proposed method performs better in terms of correlation coefficient and signal-to-noise ratio.