Telling Minds Apart: Classification of EEG Signals Based on Comparison of Brain Activity Maps

Abstract

The need to assign a particular human subject to a certain group arises in many tasks related to measurement of cognitive abilities or their application in interaction tasks. Analysis of frequencies in electroencephalograms is one of the useful approaches for the differentiation, but there is no agreed-upon method due to different frequency bands associated with various cognitive functions and personality traits. In a pilot study described in the paper, two obviously different groups of EEG signals for 26 subjects are employed: recorded with the subjects’ eyes open and the eyes closed. Brain activity maps in WinEEG are produced and 3 alternative algorithms are used to calculate pairwise image similarities for the maps per three groups: EO-EO, EC-EC, and EC-EO. The differences between all the groups are statistically significant, and the proposed “coarsening” approach towards EEG classification can easily yield accuracy of 81.25%. Its potential benefits include no need for advanced brain electric activity registration equipment and no reliance on sophisticated analysis methods that are not entirely resilient to noise in the EEG signals.