Object shape recognition from EEG signals with tactile, visuo-tactile and audio-tactile stimuli

The aim of the present work is to analyze the basis of object shape recognition from EEG signals through different sensory stimuli. EEG signals corresponding to object shape perception for tactile stimulus, visuo-tactile stimulus, as well as audio-tactile stimulus were acquired separately from the concerned regions of the human brain. Adaptive auto-regressive parameters with different model orders and power spectral density features were extracted from the acquired EEG signals for classification. Classification of each of the feature spaces was carried out in one-against-one approach using Support Vector Machine classifier with Radial Basis Function kernel to recognize ten different object shapes and its performance in terms of classification accuracy, sensitivity, specificity and computation times were noted in each case. Highest average recognition rate of 88.02% over all features was obtained for classifying the ten object shapes from audio-tactile stimulated EEG signals.