Detecting Mental Workload in Virtual Reality Using EEG Spectral Data: A Deep Learning Approach

Abstract

Mental workload, which denotes the amount of mental effort a task requires users to exert, is a critical consideration in various human-computer interaction scenarios, including virtual reality (VR) interactions. Automatic detection of mental workload as users are completing their tasks in interactive systems is crucial in terms of avoiding the possibility of overwhelming users and negatively affecting their task performance. To this end, the current study investigated the possibility of classifying mental workload levels in VR from electroencephalogram (EEG) signals through the application of deep learning models. The performances of baseline machine learning models and more advanced deep learning models were compared. The results obtained using the limited data available provide preliminary evidence that EEG signals alone might not be sufficient to reliably detect mental workload levels in VR and highlight the need for constructing larger datasets and potentially including other physiological data.