Classification of Motor Imagery Based EEG Signals

Abstract

Brain Computer Interface (BCI) enable the user to interact with system only through brain activity, usually measured by Electroencephalography (EEG). BCI systems additionally offers analysis of Motor Imagery EEG, which may be appeared, is a novel way of communication for the patients who are physically disabled. Motor Imagery based EEG data (left hand, right hand, or foot) movements supplied by BCI Competition IV dataset1. The data signals were band-pass filtered between 0.05 and 200Hz and sampled at 100Hz. The features extracted from the raw data with respect to time and frequency domain of required channels. Motor Imagery based EEG (left hand, right hand or foot) data classified using machine learning algorithm namely Support Vector Machine (SVM) and Back Propagation Neural Network (BPNN) for four normal human subjects (a, b, f, g). Analysis of motor imagery-based EEG data was studied using EEGLAB toolbox. Selected data are presented from raw data in channel data (scroll), representation of channel location in 2D and 3D form, channel spectra and maps and channel properties.