Genetic Algorithm-Based Optimization of High-Density EEG Channels for Eye State Classification with a Refined Subset Approach

In brain-computer interface (BCI) technologies, neuronal communication occurs via electrical impulses, manifesting itself as undulating patterns in electroencephalography (EEG) recordings. This work investigates the cerebral activity associated with the opening and closing of the eyes, with the aim of identifying the EEG channels that are most crucial to analyzing this particular activity. Six unique characteristics were derived from the EEG data of 64 channels. The genetic algorithm has been applied to determine the optimal set of channels that maintain a high classification accuracy in the detection of eye states. After optimizing the channels, a carefully selected subset was utilized for the final assessment using six classifiers. The comparative study revealed that multilayer perception (MLP) reached a maximum accuracy of 93.12%, with channels refined by the suggested method. This technique facilitates maintaining an ideal number of channels, thus minimizing costs associated with preprocessing large data sets for BCI studies.