Analysis of Motor Action and Motor Imagery Signals for BCI Applications

Abstract

Brain Computer Interface ( BCI) is a computerized system that acquires brain signals, extracts and classifies features during different mental activities, and converts them into correct control signals, and transfers them to external devices. BCI helps people with motor disabilities. Real-time application of a BCI system needs an efficient classification of motor tasks. Motor Imagery task identification based on EEG signals is still challenging for researchers. Extraction of robust, mutual information and discriminative features which can be converted into device commands is the biggest challenge in Motor Imagery BCI system. This study aims to analyse the effectiveness of motor and motor imagery classification for left hand and right-hand movements. The motor and motor imagery of left and right-hand movements is defined using statistical features of a higher order that are fed to classifier SVM and Random Forest Classifier. Using SVM classifier, for motor action the classification accuracy of 62.5% was reached and for motor imagery classification accuracy of 45.83% was reached. Using random forest classifier, for motor action the classification accuracy of 80.2% was reached and for motor imagery classification accuracy of 64.58% was reached.