Enhanced EEG-Based Emotion Detection Technique using Deep Belief Network and Wavelet Transform

Today's, the role of emotion in communication , brain-computer interface, brain diseases and mental states, car driver monitoring and recommendation systems is proven. Therefore, automatic emotions detection has become one of the most challenging issue. Until now, numerous studies have been addressed different technique on improving automatic emotion detection.In this study, to achieve bether validation in classification of emotion by EEG signals, we combined wavelet transform with deep belief network. For, non-stationary and time-varying are the most important properties of EEG signals, we decided to use discrete wavelet transform (sym8) for extracting features such as power, then applied deep belief network as a classifier to classify emotions according to two-dimensional arousal-valence model. To examine the effectiveness of the method, we used DEAP database and mapped different emotions on two different classes of valence and arousal. Final results show an acceptable enhancement with the accuracy of 75.52% and 81.03% for valence and arousal, respectively.