Neural encoding of rigid object-shape perception from visually stimulated EEG signals

Object shape recognition in the human brain in response to visual stimulus takes place by synchronous activities of the occipital and parietal regions. The same process due to tactile stimulus is related to the parietal region only. For complete perceptual understanding of objects both visual and tactile information is essential. Therefore, in case of damage of the parietal region object recognition by both vision as well as touch is hampered. The present work aims to predict the parietal EEG features from available occipital EEG in response to visual stimulus of different object shapes. Prediction is done by regression analysis using a back-propagation neural network with Levenberg-Marquardt optimization for weight adaptation and by means of least squares polynomial fitting as well. The parietal EEG features are predicted from occipital EEG features with Mean Squared Error (MSE) of 0.15 in 0.3 seconds and 0.0634 in 0.3 seconds using neural network and polynomial fitting respectively, on an average, over a number of subjects. Thus in the absence of parietal EEG signals, both the techniques can be implemented for efficient prediction of parietal features from occipital EEG in real time.