EEG Processing to Discriminate Transitive-Intransitive Motor Imagery Tasks: Preliminary Evidences using Support Vector Machines

It is known that brain dynamics significantly changes during motor imagery tasks of upper limb involving different kind of interactions with an object. Nevertheless, an automatic discrimination of transitive (i.e., actions involving an object) and intransitive (i.e., meaningful gestures that do not include the use of objects) imaginary actions using EEG dynamics has not been performed yet. In this study we exploit measures of EEG spectra to automatically discern between imaginary transitive and intransitive movements of the upper limb. To this end, nonlinear support vector machine algorithms are used to properly combine EEG-derived features, while a recursive feature elimination procedure highlights the most discriminant cortical regions and associated EEG frequency oscillations. Results show the significance of $\gamma ( 30 -45$ Hz) oscillations over the fronto-occipital and ipsilateral-parietal areas for the automatic classification of transitive-intransitive imaginary upper limb movements with a satisfactory accuracy of 70.97%.