Publication

Abstract : With the increasing adoption of Augmented Reality (AR) and Virtual Reality (VR) across different fields, such as training, health care, and immersive entertainment, issues of prolonged use and how it affects users’ mental well-being have been raised. Extended exposure to AR/VR environments can lead to cognitive overload, stress, and mental fatigue, negatively affecting performance and overall user experience. This study introduces an advanced EEG-based system designed to monitor users’ cognitive states in real time, ensuring a safer and more sustainable AR/VR experience. The system detects cognitive stress and mental fatigue using electroencephalography (EEG) signals and relies on strong preprocessing methods like wavelet transformation, Z-score normalization, and statistical feature extraction to improve data quality. To classify cognitive states most accurately, different machine learning models were used, such as deep learning-based classifiers like TabNet and one-dimensional convolutional neural networks (1D CNN), along with traditional models like Random Forest, Support Vector Machine (SVM), and k-Nearest Neighbors (KNN). TabNet produced the best classification accuracy of 91.5 %, proving to be effective in identifying cognitive stress patterns. Comparative performance study of all models was performed to analyze their effectiveness in identifying stress and fatigue states. The suggested system allows for real-time intervention, which minimizes mental load and improves the safety of users during immersion. Future research directions are multi-sensor fusion with physiological signals such as electrocardiography (ECG) and eye-tracking for enhanced stress detection, and real-time optimization methods for quicker feedback and adaptive management of user experience in AR/VR applications.