EEG Features Extraction and Classification Methods in Motor Imagery Based Brain Computer Interface

Abstract

Brain-computer interfaces (BCIs) are systems which enable users to control devices using only their brain’s neural activity. Its main goal is to allow for non-muscular communication with the external world, which may be the only way for patients in a locked-in state, in addition to a wide range of other applications such as rehabilitation for stroke patients, entertainment purposes, etc. There are various neurophysiological mechanisms that can be used to create a BCI. In this paper, we used a Motor Imagery (MI) approach, by recording and analyzing electroencephalogram (EEG) signals from 12 subjects. We compared 9 different types of features that are basically related either to the power or the phase of the signals. We tried 3 different classifiers: Linear Support Vector Machine (SVM), quadratic SVM, and Linear Discriminant Analysis (LDA). We also demonstrated the importance of having some sort of feedback during the training sessions. The best classification results that we achieved were up to 88.8% and averaged 76.1%.