Human Emotion Recognition using EEG Signal in Music Listening

Electroencephalogram (EEG) signal provides information about the emotion state of an individual and music can be used as a stimulus to evoke specific kinds of emotions in the human brain. The proposed work focuses on discrete emotion recognition to classify the EEG signals into nine different emotion states. A novel hybrid feature extraction model based on time, frequency, and time-frequency domain is proposed to extract important features from EEG signal. Various machine learning models (k-nearest neighbor (k-NN), Random Forest, and XGBoost) and a deep learning algorithm (Convolution Neural Network (CNN)) are used to classify the emotions with promising results on DREAMER dataset. It is found that kNN, random forest, XGBoost, and CNN provide classification accuracy of 94.49%, 99.94%, 99.39%, and 99.90% respectively. The proposed work is compared with state-of-the-art techniques and the efficiency of the hybrid feature extraction model improves classification accuracy.