Research on Virtual Reality Motion Sickness Recognition Model Based on 1D-CNN Using EEG Signals

Abstract

Virtual Reality technology holds broad application prospects across various fields, but "Virtual Reality Motion Sickness " has emerged as a critical bottleneck restricting its development. This study aims to deeply explore the neural mechanisms of VRMS through Brain-Computer Interface technology. A high-precision Electroencephalogram (EEG) system was employed to collect EEG signals and subjective motion sickness scores from 23 participants before and after exposure to VR scenarios under strictly controlled experimental conditions. Multidimensional EEG features were extracted, and a lightweight One-Dimensional Convolutional Neural Network (1D-CNN) model specifically designed for one-dimensional time-series signals was constructed to automatically identify and classify VRMS states. Experimental results demonstrated that the proposed model achieved an accuracy of 90.07%, an AUC value of 96.88%, a precision of 92.58%, and a sensitivity of 96.60% in five-fold cross-validation. Its comprehensive performance significantly outperformed baseline models such as traditional Support Vector Machines, Artificial Neural Networks, and EEGNet. This research not only contributes to uncovering the neural basis of VRMS but also provides a novel technical pathway for developing personalized, non-pharmaceutical VRMS intervention strategies, thereby bearing significant implications for promoting the healthy development of VR technology.