Object Shape Recognition from EEG Signals during Tactile and Visual Exploration

Abstract

Humans understand the world around us by visual and tactile exploration of the objects. The objective of this paper is to recognize the object-shapes from EEG signals while the subjects are exploring the same by visual and tactile means. The various object shapes are classified from electroencephalogram (EEG) signals that are stimulated by only tactile, only visual and by both means. EEG signals were acquired and analyzed from six electrodes, namely F3,F4,FC5,FC6,O1 and O2, where each pair of electrodes are located on frontal, somato-sensory and occipital region of the brain responsible for cognitive processing, tactile and visual perception. Mu-desynchronization in alpha and beta bands is used as the EEG modality for this purpose. Power spectral density (PSD) features are extracted and classified using support vector machine (SVM) classifiers in their corresponding object-shape classes. The results showed that object-shapes are best classified from EEG signals during only tactile exploration. The object shapes classified from EEG signals during only tactile exploration yielded highest mean classification accuracy of 88.34%. The average classification accuracy over all three object exploration modalities is 83.89%.KeywordsTactile perceptionvisual perceptionobject-shape recognitionelectroencephalogrampower spectral densitysupport vector machine