Study on Complexity of EEG Time Series Under Different Mental States

In this paper. two kinds of surrogate data are designed according to electroencephalography (EEG) data. They are a binomial random series with the same source entropy as that of EEG and an AR series that is the output of AR model constructed from EEG data and driven by white noise. Significant differences among the complexity of these three series show that BEG is far from being random and has certain pattern in it which can be partially modeled by AR. The complexity of EEG series recorded during three different mental states is computed and the result of two-way analysis of variance shows that the complexity can discriminate the three mental states significantly.