Multi-modal classification of cognitive load in a VR-based training system

S. S. Bhat , C. Dobbins , A. Dey , O. Sharma

ISMAR

Effective training must challenge the learner without overwhelming them, which means reading cognitive load objectively rather than by observation. In this study, 30 participants built models across VR levels designed to induce low-to-high load while EEG and electrodermal activity were recorded. After feature selection, an XGBoost classifier best distinguished load levels (F1 around 0.83, accuracy around 0.81), and analysis showed the strongest contributions came from frontal and occipital brain regions and from tonic skin-conductance features.

@inproceedings{bhat2023,
  author    = {S. S. Bhat and C. Dobbins and A. Dey and O. Sharma},
  title     = {Multi-modal classification of cognitive load in a VR-based training system},
  booktitle = {ISMAR},
  pages     = {503-512},
  doi       = {10.1109/ISMAR59233.2023.00065},
  year      = {2023}
}