PCG denoising using AR-based Kalman Filter

Heart sound signal, known by Phonocardiogram (PCG) is one of the most important signals made by physical activity of heart and helps for diagnostic aims. PCG recording is not always in ideal conditions and noise may be added. One possible noise is white gaussian noise which exists in many real applications. There are different methods to denoise the signal. In this paper an AR based Kalman method is proposed. A Kalman filter based on an assumption that the signal is following an AR model, tries to find AR model parameters and then another Kalman filter, based on the found coefficients denoises the PCG signal. A Kalman smoother is finally used to improve the process. The results show signal-to-noise (SNR) ratio improvement of the signal for different input SNR values.