Explore the Feasibility of Integrating EEG Neurofeedback in VR Exercise Gaming for Perceived Physical Exertion Study and Classification using Machine Learning

This study investigated the effectiveness of using electroencephalogram (EEG) data to predict exercise intensity during virtual reality (VR) gameplay, specifically with the game Beat Saber on the Meta Quest 2 headset. Participants first completed a Six-Minute Walk Test (6MWT), followed by playing Beat Saber at three difficulty levels (easy, medium, and hard). During each activity, EEG, accelerometer, and gyroscope data were collected. Various feature selection techniques and feature sets were employed to train and evaluate the performance of five machine learning (ML) models in predicting four levels of exercise intensity. The effectiveness of using EEG, accelerometer, and gyroscope data was also compared. The results showed that incorporating EEG features significantly improved the accuracy of the exercise intensity prediction model. Among the five ML models, Gradient Boosting achieved the highest performance, with an F1-score of 0.825. Additionally, EEG data proved to be more reliable than accelerometer and gyroscope data for detecting exercise intensity. This study demonstrated the technical feasibility of data collection in a mixed-reality exercise context, effectively aligning VR gameplay with EEG data. These findings could open new avenues for future research in designing personalized health plans and conducting health monitoring studies.