Abstract
EEG is a noninvasive method used to study the neural activity of the brain. This method has already proved its application in emotion recognition systems. EEG-based emotion recognition is preferred over facial-image-based emotion recognition in many cases like burned or paralyzed faces. Most of the research in this area is on subject-dependent emotion recognition where the same subject’s EEG data has used for training and testing. However, in case of lack of labeled training data of a subject, there is a need for a subject-independent emotion recognition system, which could capture emotion-based EEG features common to subjects. This study aims to develop an EEG-based emotion recognition system with subject-independent approach using the publically available database DREAMER. In this research, a convolutional neural network (CNN) model, which takes raw EEG as input and classifies emotions in Valence–Arousal space is proposed for subject independent emotion recognition. The results of the proposed model have compared with the other existing studies and observed remarkable improvement on the DREAMER database. The result of the proposed CNN model shows 75.93% accuracy for valence and 81.48% for arousal classification on the DREAMER database for subject independent emotion classification.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call