Independent Component Analysis of temporally extended EEG

This project deals with the development and application of new tools which will allow the analysis of functional brain signals recorded during temporally extended cognition. The focus will be on cognitive activities of long duration instead of short "pulse``-evoked cognition as in the response averaging paradigm. Therefore, classical averaging necessary to enhance the signal-to-noise ratio will no longer be possible because of the great temporal variability of signals. Independent Component Analysis (ICA) has already been proven to enable analysis of single trials of functional brain signals and was applied to the analysis of short duration cognitive events. This project will extend the method of ICA to enable analysis of longer segments (several seconds) of single trial EEG recordings with variable length. This will shift the focus from independent components accounting for stimulus and response locked events to components accounting for non-phase locked and oscillatory EEG activity. If components are no longer time- locked to either stimulus or response, new ways of finding similarities across single trials of independent components must be found. Our research will concentrate on algorithms for sequence alignment like dynamic time warping or Hidden Markov models (HMM), but also on already existing extensions of ICA. For example, ICA mixture models and Hidden Markov-ICA both allow for automatic context switching in time series analysis. The project can be seen as part of the new emerging field of "cognitive event-related brain dynamics`` which carries the potential to allow for an analysis of truly cognitive behavior instead of pseudo "pulse``-evoked cognition.