Classification and analysis of EEG signals for imagined motor movements

Abstract

This paper presents a data driven approach to explore the variations in the electroencephalogram(EEG) signals when a person tries to imagine movements like moving his or her left hand, right hand, foot and tongue. The paper tries to find out the type of variations that occur in the EEG signals when such type of imagined movements are undertaken by a person and also the regions in the brain where the variations of EEG signals are the most pronounced. EEG data corresponding to the said actions was captured from three different persons using multiple electrodes placed over the head. Features based on auto regressive power spectral density and entropy measures have been used to analyze this data. This was followed by feature selection process to reveal the most prominent of the features. Analysis of the selected features revealed the positions of the electrodes which were picking up the variations in the EEG signals. This resulted in the identification of the regions in the head where the signal variations were most prominent. It was found that the positions were not fixed but varied from person to person. The findings have been backed up by time-frequency maps of the signals which describes the type of variations that happens in the EEG signals when different kinds of movements are imagined and how varied these variations are with respect to individual subjects as well as the types movements performed.