A convolutional transformer model for EEG-driven cybersickness detection in VR experience

Abstract

Cybersickness (CS) is a condition that negatively affects user comfort during virtual reality (VR) experiences, manifesting as symptoms such as dizziness, nausea, and eye strain. In this study, a hybrid model combining Convolutional Neural Networks (CNN) and Transformer architectures is proposed to detect CS using EEG-based physiological signals. To overcome data limitations and enhance generalization capability, data augmentation and K-Means-based clustering techniques were applied. The model’s performance was comparatively evaluated across different data scenarios, with the best results obtained using the clustered and time-reversed augmented dataset. The findings demonstrate that the proposed approach provides an effective solution for CS detection.