Signal Separation for Ballistocardiography via Locally Projective Noise Reduction

The work presented in this paper focuses on the separation of the respiratory and cardiac components of a ballistocardiogram (BCG) recorded by means of a bed-mounted BCG system. The aim is to preserve the original waveform of both components and thus to enable the calculation of more sophisticated medical markers, e.g. markers based on the phasecoupling between respiration and heartbeat. For this purpose, an iterative method based on locally projective noise reduction is developed. The key advantage of the proposed method is that it distinguishes signals not by their frequency components but by the typical time and amplitude scale of deflections in the signals.