A robust Brain Computer Interface system for classifying multi motor imagery tasks over daily sessions

In Brain Computer Interface (BCI), the thoughts of a subject is read to provide an appropriate way of communication using only brain signals. The Information of electroencephalogram (EEG) signals defer between subjects depending on their thoughts according to research. In this paper, a comparison between different types of features tested by several classifiers is done to propose a model for classifying multi motor imagery (MI) tasks through offline analysis for a single subject performing one session daily for a week. Several classifiers and feature extraction techniques were used. Common Spatial Pattern (CSP) as a feature classified by Linear Discriminant Analysis (LDA) was found to outperform all other combinations with an average classification accuracy above 88%.