EEG-based classification of the intensity of emotional responses

Considering the fact that emotional experiences are stored in the brain, classifying emotion from brain activity measured using electroencephalography has become a trend. In most of the previous studies, the user's emotions were classified based on a stimulus. In this paper, we present a model that can classify the emotion intensity by the participants' self-report. Two machine learning classifiers are considered: support vector machine (SVM) and convolutional neural networks (CNN). Results demonstrated that both SVM and CNN models perform well with four classes of emotions (positive/negative valence high/low arousal combination) where SVM achieved an accuracy of 85% whereas CNN achieved 81%. Considering 12 classes of emotional responses (low, medium, and high intensity for positive/negative valence high/low arousal combination) by the participants' self report resulted in an accuracy of 70% for SVM and 69% for CNN. The proposed model excels in classifying emotional intensity and provides superior performance compared to the state-of-the-art emotion classification systems.