Deep Learning-based Simulator Sickness Estimation from 3D Motion

J. Zhao , K. T. P. Tran , A. Chalmers , W. K. Hoh , R. Yao , A. Dey , J. Wilmott , J. Lin , M. Billinghurst , R. W. Lindeman , T. Rhee

ISMAR

How sick a headset experience makes someone depends heavily on its motion. In a VR study that let users score their sickness in real time with an instant dial, the authors found that translation and rotation affect simulator sickness differently and that users' demographics and self-rated susceptibility carry signal too. Guided by these findings, they built a deep-learning model that estimates sickness from decomposed 3D motion features plus user-profile information, and it proved more accurate than approaches based on optical flow from recorded video.

@inproceedings{zhao2023,
  author    = {J. Zhao and K. T. P. Tran and A. Chalmers and W. K. Hoh and R. Yao and A. Dey and J. Wilmott and J. Lin and M. Billinghurst and R. W. Lindeman and T. Rhee},
  title     = {Deep Learning-based Simulator Sickness Estimation from 3D Motion},
  booktitle = {ISMAR},
  pages     = {39-48},
  doi       = {10.1109/ISMAR59233.2023.00018},
  year      = {2023}
}