Abstract:

Brain Computer Interface (BBA) is based on the classification of a person’s intention or intentional thought without any physical ability. In BBA studies, the data is obtained using the electroencephalography (EEG) method while exposed to some surrounding (network noise, etc.) or internal (eight movements, ECG, etc.) noise. One of the noises that can significantly affect the accurate classification of EEG data is electro-oculographic (EOG) noises. In this study, the 2a data set of the BCI-IV competition was used by the University of Graz. The data set of 4 classes is reduced to 2 classes, only for the right and left imitation movements to be classified offline. EEG data from 22 channels and EOG data from 3 electroths that were simultaneously installed around the eye were obtained. In the data set used, traditional band-passing filters and recursive least square (RLS) adaptive filters were used to clean EOG noises from EEG signals. The engine movement imagination has been applied to the input of the specificity vectors extracted by the common spatial patterns (CSP) method, linear discriminant analysis (LDA), support vector machines (DVM), naive bayes (NB) and k-NN classification algorithms. The combination of Chebyshev Type 2 and DVM provided the highest classification performance with an average accuracy of 72%. The classification results obtained by the RLS adaptive filter are also compared to the results of the band-transmitted filters. The RLS algorithm and the LDA combination showed lower performance than the results obtained by traditional band-passing filters with a 64% accuracy rate. The findings obtained in the study are supporting literature. The findings of the study will also be able to keep light on the work to be done in the future.

Keywords: