EEG classification model for virtual reality motion sickness based on multi-scale CNN feature correlation.

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Virtual Reality Motion Sickness (VRMS) poses a significant challenge to the widespread adoption of virtual reality (VR) technology.
• Accurate detection of VRMS is crucial for developing effective countermeasures and improving user experience.

Purpose of the Study:

• To propose a novel Convolutional Neural Network (CNN) model for detecting VRMS using electroencephalogram (EEG) data.
• To leverage multi-scale feature correlation within EEG signals for enhanced VRMS detection.

Main Methods:

• Utilized multi-scale 1D convolutional layers to extract temporal features from multi-lead EEG data.
• Converted time-domain features into correlation-based brain network features via feature adjacent matrices.
• Fused multi-scale correlation features and employed a channel attention module for classification.

Main Results:

• The proposed CNN model achieved high performance metrics: 98.66% accuracy, 98.65% precision, 98.68% recall, and 98.66% F1-score.
• Demonstrated superior performance compared to existing classic and advanced EEG recognition models.
• Validated the model using resting-state EEG data collected from subjects experiencing virtual roller coaster scenarios.

Conclusions:

• The developed CNN model effectively recognizes VRMS using resting-state EEG.
• The findings indicate the model's potential for real-world VR applications to mitigate motion sickness.