Skip to Main Content
A novel signal source separation method is introduced for analysis of intrinsic Optical Imaging (OI) and functional Magnetic Resonance Imaging (fMRI) data. This method is based on the fact that all real interesting signals are autocorrelated spatially as well as temporally. Many signal source separation algorithms which using autocorrelation or other structure information of the interesting signals, e.g. Canonical Correlation Analysis (CCA), have been proposed before. However, these traditional methods using only temporal structure in temporal signal separation and only spatial structure in spatial signal separation. In this article, we have given a new definition of the spatial distribution pattern of temporal signals, this make it possible to represent the spatial autocorrelations of temporal sources. Then an objective function which takes account of both temporal and spatial autocorrelations could be constructed. By maximizing such objective function, it is expected to find the real signal sources underlie the mixed data. The comparison experiment shows that the new method is of better performance in recovering the signal sources than the traditional methods.