Application of second and higher order subspace tracking in multichannel data analysis

The problem of blind source separation (BSS) and tracking from time-varying mixtures is an open-problem of biomedical signal processing research. In this study we present a framework for decomposing and tracking instantaneous separation matrices of independent component analysis (ICA) solutions of BSS. The decomposition is based on the tracking of the second order statistics (SOS) and higher order statistic (HOS) stages of ICA. We investigate the variations of data subspaces by means of tracking the principal angle and Givens rotation angles of the instantaneous mixture. The application of this technique is illustrated for electrocardiogram signals. We shown how the SOS and HOS variations of time-varying mixtures can be decoupled and may correspond to the second and higher order properties of the data. This new approach is believed to have various advantages for online subspace tracking in blind and semi-blind scenarios and the better examination of statistic characteristics of multichannel data, especially for biosignal processing.