TOWARDS PREDICTION AND IMPROVEMENT OF EEG-BASED MI-BCI PERFORMANCE.

Abstract

Motor imagery (MI) is one of the mental activities to control a brain-computer interface (BCI) system. The main goal of this thesis is to improve the MI-BCI performance of the subjects through addressing two of the main challenges in MI-BCI systems: BCI deficiency and EEG nonstationarity. Therefore, novel neurophysiological predictors are proposed to identify poor performance subjects, and new neurofeedback training is suggested to improve subjects? BCI performances. Subsequently, two adaptive algorithms are proposed to address inter-session non-stationarity and to enhance the performance of the subjects. The results showed that low attention level and low alpha band power in resting state resulted in poor BCI performance. Moreover, the MI-BCI performance of the subjects improved from neurofeedback training using the proposed predictor and by using the proposed adaptive algorithms. In conclusion, the purpose of the proposed methods was to come up with a more practical BCI system for real-world applications.