Emotion Recognition Based on DEAP Database using EEG Time-Frequency Features and Machine Learning Methods

In recent years, research of the human emotional state is becoming importance, especially in its application for patient monitoring and in the treatment management system of that patient. In this paper, an EEG based emotion recognition system is developed that consists of a feature extraction subsystem and a classifier subsystem. As better performance of the feature extraction subsystem may produce higher recognition accuracy, nine features derived from the time and frequency domain from the EEG signal is used and analyzed. We have utilized support vector machine and Random Forest methods for classifying the emotional state of the subject, and compare its results with other machine learning methods. Using two-fold data validation model, the experiment result shows that the highest recognition accuracy is produced by using Random Forest method, i.e., 62.58%.