Suppression of Artifacts from Seismocardiogram Signal using Two-Stage Kalman Filtering Model

In this paper, a unified denoising framework is proposed to suppress the noises and artifacts from a seismo-cardiogram (SCG) signal. The proposed method consists of a unique set of two Kalman filter models in a cascaded fashion. Each of the Kalman filters is modelled separately to serve two different purposes. First stage Kalman filter (KF1) is modelled to reduce the irregularity and the intermittency of the noisy signal. It also predicts the displacement and the velocity of the chest wall, produced due to cardiac mechanics. The low frequency artifact is predicted in the second stage Kalman filter (KF2) and then it is removed from the KF1 resulted SCG signal. The performance of our proposed framework is tested using SCG signals from CEBS database available at Physionet archive. The proposed method achieves an average normalized cross correlation (NCC) of 94% for signals having contaminations of AWGN noise with 15 dB input signal-to-noise ratio (SNR). The qualitative analysis of experimental results and comparison with existing methods clearly show that the proposed method produces promising noise suppression results without distorting the clinical features.