Brain Computer Interface Study

Abstract

Brain Computer Interface (BCI) is a system which offers a user a direct communication link between his or her brain and an external device such as a computer so that the device can be commanded solely by means of brain activations. Electroencephalography (EEG) is the most commonly used brain activation technique in the latest BCI-related researches at present. Motor imagery, which is a process of mental imagination of movements, has been widely employed as an experimental paradigm in the study of BCI as the EEG signals measured during motor imagery can be converted into commands. In this project, a multiclass EEG motor imagery-based BCI system model, which is a combination of prepossessing, feature extraction, feature selection and classification techniques and some new methods, is proposed and implemented to classify motor imagery tasks. MATLAB 2012a and BioSig toolbox are used to program and implement the proposed model. The data set used in testing the proposed model is obtained from BCI Competition IV. The performance of the proposed model can be clearly seen by evaluation of the classification results of the data set in terms of accuracy rate and Kappa coefficient.