Binary particle swarm optimization (BPSO) based channel selection in the EEG signals and its application to speller systems

Social participation of people with disabilities is tried to be increased with state-supported projects recently. However, even in neuromuscular diseases such as Motor Neurone Disease (MND), Full Sliding Status (TSD), even the communication skills of individuals are interrupted. Brain-Computer Interfaces (BBA), which have a few decades of history and an increasing number of studies with exponential momentum, are being developed to enable individuals with such disorders to communicate with their environment. Spelling systems are BBA systems that detect the letters that the person focuses on the matrix of letters and numbers on a screen and convert them into text through the application. In this context, with the random flashing of the letters on the screen, it aims to detect the electrical changes occurring in the brain as a result of the stimulus given to the person. Research reveals that the stimulus that the individual encounters cause an amplitude in the EEG signal called P300, between 250 and 500 ms. Brain-computer interfaces are used through EEG signals to provide environmental interactions for individuals with restricted movements due to stroke or neurodegenerative diseases. The multi-channel structure of EEG signals both increases system cost and reduces processing speed. For this reason, reducing the system cost by detecting more active electrodes during the process increases the accessibility of people. In this context, the use of optimization techniques in electrode selection is used to determine the most effective channels by a random selection method. In the study, particle herd optimization algorithm, one of the herd-based optimization techniques, was used with two classifiers, SVM and Boosted Tree, and the eight most frequently selected channels were determined to improve system performance in terms of speed and accuracy.