Pso based classification of EEG signals recorded during imagery of hand grasp movement

One of the important objective of the Brain Computer Interface (BCI) systems is to search innovative solutions like rehabilitation scenario for disabled or patient subjects. People who have stroke or have an accident still can provide accurately some imagery movements. Automated decoding of these imagery movements from brain signals will be very helpful for rehabilitation and the development of robot-assisted technologies based on BCI systems. Then, work on the patient‟s data instead of using healthy subject‟s data can be more meaningful for these interfaces. In this thesis work, a dataset that of EEG brain imaging data for 10 stroke patients having hand functional disability was used. This current data was also used in Clinical BCI Challenge WCCI 2020 competition. With proposed method, the effective electrodes and features were selected for high classification accuracy purpose. In feature selection stage, the Particle Swarm Optimization (PSO) algorithm was used. Through selected effective parameters, discrimination of imagery of right and left hand movement was done with 84.32%, 80.25%, 77.25% and 83.08% accuracy rate by using respectively k-nearest neighbors, linear discriminant analysis, support vector machines and bagging decision tree algorithms.