Detecting evoked potentials with SVD- and ICA-based statistical models

In the functional diagnostics of human sensor systems, the analysis of stimulus responses embedded in an electroencephalogram (EEG), e.g. evoked potentials (EPs), is of high relevance for an objective electrophysiological assessment. The aim of this work is to detect weak EPs from highly contaminated signal traces. In principle this can be done using methods of spatiotemporal signal processing, which simultaneously increase the weak SNR (signal-to-noise ratio). However, methods based on any a priori knowledge of spatial or temporal properties as well as the propagation speed and direction are not applicable. Models with adjustable signal properties similar to real cortical activity are necessary for the development and evaluation of new methods of spatiotemporal signal processing. A model is needed which can be used in forward- and inverse-projection calculations. This study aims to develop a signal generator of the background EEG activity with embedded EPs of fully adjustable signal parameters. The study also compares the results of modeled signal analysis by known methods for signal decomposition, SVD (singular value decomposition) and ICA (independent component analysis).