VRSA Net: VR Sickness Assessment Considering Exceptional Motion for 360° VR Video

Area of Science:

• Virtual Reality (VR) and Human-Computer Interaction
• Computational Neuroscience and Perception
• Biomedical Engineering

Background:

• Virtual reality (VR) content viewing safety is a significant concern, with VR sickness being a primary issue.
• Objective assessment of VR sickness is crucial for enhancing user experience and safety in immersive environments.
• Exceptional motion patterns in VR videos are identified as critical factors inducing VR sickness.

Purpose of the Study:

• To propose a novel objective VR sickness assessment (VRSA) network for automatically predicting VR sickness scores.
• To develop a deep generative model that analyzes motion patterns to quantify VR sickness.
• To establish a reliable method for assessing VR sickness beyond subjective reporting.

Main Methods:

• A VRSA network comprising a VR video generator and a VR sickness score predictor was developed.
• The VR video generator was trained on common videos to learn human motion perception tolerance.
• The difference between original and generated videos highlighted exceptional motion, which was then projected to predict the VR sickness score.

Main Results:

• A new dataset including 360° videos, physiological signals, and subjective questionnaires was created for evaluation.
• Experimental results showed the proposed VRSA network achieved a high correlation with human perceptual scores for VR sickness.
• The VRSA network effectively quantifies VR sickness by analyzing motion-induced perceptual differences.

Conclusions:

• The proposed VRSA network offers an objective and automated method for assessing VR sickness.
• This approach holds potential for improving the safety and comfort of VR content consumption.
• Further research can build upon this method for more sophisticated VR experience design and safety protocols.