Compact EEG Features for Person Identification

In the present work, we consider the opportunity for person identification from multichannel EEG signals. Specifically, we propose a method for designing compressed characteristic features representing the aggregated EEG activity for a selected period of time. The features are obtained as the averaged values of the relations of activity between specific EEG electrodes, using different sizes of the aggregation window. Twelve sets of these features, designed through the proposed method, were evaluated in a common experimental protocol based on the well-known DEAP database to assess the practical applicability when combined with typical classifiers. The experimental results validate the practical significance of the proposed method for developing intelligent HMIs.