VR Cybersickness Classification Using Machine Learning Models on Open-Access EEG Datasets from the Human Vestibular Network

Aug 4, 2026

Upload by Quinn

This paper investigates EEG-based VR cybersickness classification using machine learning models, focusing on four key bottlenecks: the limited availability of open-access EEG datasets, imbalanced label distributions, insufficient evaluation of generalizability with data from unrelated control groups, and limited analysis of personalized EEG data beyond individual baselines. Using an open-access database collected during VR cybersickness induction, the work prioritizes generalizable models rather than gender-dependent approaches, addressing reproducibility and robustness for VR neurofeedback applications.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages6

Reading Time15 min

AccessFree to download

Introduction

Machine learning and EEG-based cybersickness classification

Motivation and unresolved challenges

Related Work

Open-access EEG database

Imbalanced data and ground-truth labels

Citation for published version: Li, G, Yonsei University, Yonsei University & University of Glasgow 2025, 'VR Cybersickness Classification Using Machine Learning Models on Open-Access EEG Datasets from the Human Vestibular Network', Paper presented at the 7th IEEE International Conference on Artificial Intelligence & eXtended and Virtual Reality, Lisbon, Portugal, 27/01/25-29/01/25 pp. 171-175. [https://doi.org/10.1109/AIxVR63409.2025.00033](https://doi.org/10.1109/AIxVR63409.2025.00033) DOI: 10.1109/AIxVR63409.2025.00033 Publication date: 2025 Document Version Peer reviewed version Link to publication University of Bath Alternative formats If you require this document in an alternative format, please contact: [openaccess@bath.ac.uk](openaccess@bath.ac.uk) General rights Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights. Take down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim. Download date: 17. Sep. 2025 VR Cybersickness Classification Using Machine Learning Models on Open-Access EEG Datasets from the Human Vestibular Network Gang Li Department of Computer Science University of Bath Bath, UK [gl755@bath.ac.uk](gl755@bath.ac.uk) Hae-Rin Byeon Department of Computer Science and Engineering Yonsei University Seoul, South Korea [byeonn@yonsei.ac.kr](byeonn@yonsei.ac.kr) Frank Pollick School of Pscyhology and Neuroscience University of Glasgow Glasgow, UK [Frank.Pollick@glasgow.ac.uk](Frank.Pollick@glasgow.ac.uk) Sung-Bae Cho Department of Computer Science and Engineering Yonsei University Seoul, South Korea [sbcho@yonsei.ac.kr](sbcho@yonsei.ac.kr) Abstract—This paper addresses key challenges in EEGbased cybersickness classification using machine learning (ML) models. Despite significant research in this area four critical issues remain unresolved: 1) the availability of open-access EEG datasets; 2) imbalanced data distribution; 3) limited generalizability testing, and 4) insufficient exploration of personalized EEG data. Keywords—VR cybersickness, EEG, machine learning I. INTRODUCTION Machine learning (ML)-driven brain-computer interfaces (BCI), integrated with VR headsets, represent a significant technical advancement in closed-loop cybersickness reduction. Because this approach has the potential to directly enhance VR users’ resistance to cybersickness by neurofeedback [1] or neuromodulation [2], [3], [4], [5] without the need to redesign VR contents. As the first phase in this approach, numerous studies have explored electroencephalogram (EEG)-based cybersickness classification using ML models. However (see Related Work below), none have utilized open EEG datasets, resulting in limited data accessibility and reproducibility. More importantly, the ground truth for cybersickness is highly imbalanced due to nauseogenic cybersickness experiments and the pre-defined ethical threshold on ground truth scores. Despite this, none of the previous EEG studies have clearly described the data distribution of the ground truth labels, not mention to address the challenge of imbalanced ground truth labels. Furthermore, none of the previous studies tested the generalizability of the ML models using data from a control group unrelated to cybersickness. Also, the role of personalized EEG data corrected from individual baseline in cybersickness classification is underexplored. This study investigated the above three issues based on an open-access EEG database. This paper focuses on the development of generalizable ML models rather than gender-dependent models. Because our recent study did not find that gender is a significant predictor for VR

FAQs

What unresolved challenges does the paper target in EEG-based cybersickness classification?

The paper targets four issues: lack of open-access EEG datasets, imbalanced data distributions, limited generalizability testing, and insufficient exploration of personalized EEG data.

Why is class imbalance a problem for cybersickness ground truth?

The ground truth labels are highly imbalanced due to the nature of nauseogenic induction experiments and ethical thresholds applied to ground-truth scores.

What dataset does the study rely on to address these issues?

The study uses an open-access EEG database collected during VR cybersickness induction experiments, referenced as available via DOI 10.5281/zenodo.6373681.

Aug 4, 2026

Upload by Quinn