Complexity Measures of Electroencephalographic Data

Abstract

The intention of this project is to examine whether it is possible to measure alertness using EEG recorded on awake subjects. This thesis tries to examine if a measure of attentiveness could be derived from measures of complexity of EEG data. Tn order to avoid averaging required by Fourier analysis, a dynamical embedding is realised. As well, it allows us to reconstruct and to characterise the underlying generator of the data. On the results of this embedding, a signal filtering is proposed and different measures of complexity are derived. An analysis of these measures is performed by comparing them with information available about the attentiveness of the subject during the experiments. Finally, a feature extraction is performed. On the basis of this feature extraction, a filtering of the signal is realised. It may allows us to improve the accuracy of the different measures proposed.