Automated Parameter Estimation for a Nonlinear Signal Separation Scheme

Abstract

The current trend towards telemedicine has sparked a renewed interest in Ballistocardiography (BCG), a method for measuring mechanical cardiac activity. Due to its unobtrusive nature it is considered as a promising approach for home monitoring systems. However, BCG recordings are often contaminated by respiratory baseline variations, which have to be separated from the cardiac component before analysis. Performing this separation with linear filters can be difficult, since the spectra of the cardiac and respiratory components overlap. A recently proposed nonlinear separation method circumvents the problem of spectral overlap by performing the signal separation in delay space. However, finding the correct parameters for the new method is difficult and had to be performed manually. In this paper an automated parameter estimation algorithm is proposed, which is able to estimate suitable parameters for the nonlinear signal separation method.