Machine Learning Algorithm to Detect EEG based Emotion states using Virtual-Video stimuli

The interpretation and detection of human emotions are crucial functions of the center-nervous system. Physiological signals are used widely to develop emotion recognition systems in recent years. Detection of emotions using electroencephalogram is a challenging task since they exhibit nonlinear nature. In this paper, three popular machine-learning models are implemented on an acquired database using a Novel approach. The EEG signals are acquired by showing Virtual Reality Video clips for eight discrete emotions. Thirty-four features are extracted in time and frequency domains. The frequency bands decomposition is carried out using 4-level Discrete-Wavelet-Transforms. The generated feature vectors are used on KNN, SVM, and ANN algorithm-based classifiers to classify the data into four, three, and two emotion states. The overall accuracy using ANN, SVM, and KNN classifiers is 85.50%, 73.50%, and 66.75% respectively for 4 classes. The performance of ANN classifiers gave a better accuracy compared to other models.