EEG Signals Based Motor Imagery and Movement Classification for BCI Applications

Abstract

The Brain-Computer Interface (BCI) is a system that uses the neural activity data of the brain to control the devices in the outside world, in other words, to communicate. BCI studies of wearable sensor EEG sensor technology have gained momentum. In this study, in order to enable the use of electroencephalogram (EEG) patterns in BCI applications, the extraction of statistical-based features, the selection of the most effective features with the NCA method, and the determination of the type of motion request with classification algorithms were carried out. The PhysioNet EEG Motor Movement/Imagery dataset was used. For six different types of motion and imaging, 30 statistical features were calculated (960 in total) for each channel of the EEG signals received from the 48-channel EEG sensor head, and the most effective 120 features were selected with NCA. The selected feature set is given as input to the LD, NB, SVM classification algorithms. The test accuracy success of the models is 91.18%, 95.41%, and 99.51%, respectively. These results show that the proposed method will give successful results in BCI applications.