Decoding of Brain Signals to Detect Perceived Color-Stimuli using Convolutional Neural Network

The paper aims at determining the active brain regions responsible for perceiving and understanding the sense of three basic color stimuli: red, green and blue. This is achieved in two main steps. In the first step, we take EEG response to color stimuli from the scalp using the standard 10-20 electrode system. Experiments undertaken using Exact Low Resolution Electromagnetic Topographic (eLORETA) software reveal that there exist long term (around 1 second) correlations between activated brain regions and the perceptual process of specific color stimulus. For instance, the parietal and the occipital lobe activations have long duration correlations with the blue color stimuli; whereas the prefrontal and the occipital lobe activations have correlations with the red color, while the temporal and the occipital lobe activations have correlations with the green color. In the second step, we classify the perceived color of the brain signals acquired from the selected brain regions. A one dimensional based Convolutional Neural Network (1DCNN) classifier has been designed to perform the classification process by utilizing the brain signals from the activated lobes. The present classifier model has also been compared with other primitive classifiers. Performance analysis followed by statistical tests undertaken reveals that the 1D CNN classifier outperforms its traditional counterparts by a wide margin. The proposed technique is expected to have interesting applications to explain the malfunctioning in recognition of colored stimuli due to damage in certain brain lobes like occipital, temporal etc.