Robust emotion classification using neural network models

In the field of Brain Computer Interface, Emotion recognition plays an increasingly crucial role. As psychological understanding of emotions progresses, feature extraction along with classification of electroencephalogram (EEG) representation of these emotions becomes a more important challenge. In this work, Neural Networks as a type of high accuracy robust statistical learning model was employed in order to classify human emotions from the DEAP [7] dataset containing the measured EEG signals for Emotion Classification research. We take advantage of two Neural Network based models the first one of which is the Deep Neural Network and the other one is the Convolutional Neural Network in order to classify Valence, Arousal, Dominance and liking into two categories of Yes or No (High and Low) and to classify Valence and Arousal into three categories of (High, Normal, Low), The achieved accuracy surpasses those achieved in other papers indicating that these models carry the ability to be used as a high achieving classifier for BCI signals.