Analysis of Cybersickness through Biosignals: an approach with Symbolic Machine Learning
Wilson Ricardo Cardoso Silva, Thiago Malheiros Porcino, Carla Denise Castanho, Ricardo Pezzuol Jacobi · 2024
Cybersickness represents one of the main obstacles to the use of Virtual Reality, often triggered by the use of Head-mounted Display devices. The symptoms associated with cybersickness can vary among individuals and include nausea, dizziness, eye strain, and headache, which may persist for minutes or even hours after exposure to Virtual Reality. According to the literature, cybersickness has a considerable impact on physiological signals such as delta waves in the Electroencephalogram; Heart Rate and Heart Rate Variability, derived from the Electrocardiogram; Electrodermal Activity; and Electrogastrography, all of which show a significant correlation with this condition. In this study, we investigated the use of biosignals to identify the possible causes associated with cybersickness in Virtual Reality. Our main hypothesis is that the combination of quantitative and subjective assessments, combined with Symbolic Machine Learning techniques, is effective in creating a ranking of the main causative/indicative factors of this condition. The results of this study highlight significant contributions to the understanding of factors influencing cybersickness symptoms. Statistical analyses confirmed the relationship between physiological changes and cybersickness symptoms. By including biosignals in our model, we achieved a significant gain, with an AUC of 0.95. The rankings of the main factors, both for the model without and with the inclusion of biosignals, confirmed previous research described in the literature. To the best of our knowledge, this is the first work to employ Symbolic Machine Learning models combining data from user profiles, game, and biosignals to detect the causes of cybersickness and generate a ranking of the most relevant factors.