Selection of Optimal Frequency Bands of the Electroencephalogram Signal in Brain-Computer Interface

This article proposes a new method to increase the performance of brain-computer interface (BCI) taking into account the individual characteristics of users. The idea of the method consists in the automatic selection of the most informative frequency bands of the electroencep halogram (EEG) signal. As a measure of information content we use the accuracy of the imagery movement classes’ separation. The first part of the article explores differences in sensorimotor rhythms of the EEG signal between users. The second part provides a mathematical formulation of the optimal frequency bands selection problem, which is considered as a one-criterion optimization task. Boundaries of the frequency bands are considered as the variable parameters while the assessment of the classification accuracy acts as an objective function. In the following sections we propose to find a solution of the optimization task using a genetic algorithm. In the last section we compare the efficiency of the described method with other ones, including the algorithm based on the estimation of the EEG signal energy in the classical frequency bands. As a test data we use EEG recordings submitted to BCI Competition IV. In conclusion the main results and future lines of research are discussed