Electroencephalography-based recognition of six basic emotions in virtual reality environments

Abstract

Emotion recognition (ER) in virtual reality (VR) environments has attracted increasing attention in the field of affective computing and human–computer interaction (HCI). Electroencephalography (EEG) is an effective tool for capturing human emotions and mental states, and the demand for collecting and analyzing EEG signals in VR environments has been on the increase. In this study, we built an EEG dataset for six basic emotions in VR environments—that is, the VR six discrete emotional EEG data (VRSDEED) dataset—and developed a machine-learning framework for classifying emotional states using it. Fifty participants were shown six VR emotion-inducing videos while their EEG signals were recorded. To determine the optimal features and classification methods, we extracted four frequency-domain features and applied five different machine-learning algorithms. To enhance prediction performance, we fused different features and used recursive feature elimination with cross-validation for feature selection. Classification performance was assessed using accuracy and F1-scores. Our findings demonstrated that (1) the light gradient-boosting machine algorithm achieved the highest average classification accuracy of 89.94%; (2) isolated features achieved the highest accuracy of 89.50% in the γ frequency band and feature fusion enhanced recognition performance; (3) the θ and γ frequency bands were more significant in emotion processing than other bands. The proposed dataset and framework should serve as valuable resources and provide insights for research on ER in VR environments.