NeuroHaptic World is a prototype study investigating whether individualized EEG motor-imagery classification outputs can serve as explainable control inputs for virtual scene transitions and software-simulated haptic parameter mapping. Using the PhysioNet EEG Motor Movement/Imagery Dataset, this study focuses on left-hand versus right-hand motor imagery tasks from runs R04, R08, and R12. The study first evaluates individualized EEG decoding performance across multiple subjects. Subject S007 is then selected post hoc as a best-case individualized prototype demonstration. For S007, a CSP + Linear SVM model achieved 95.56% accuracy and a 95.55% F1-score under five-fold stratified within-subject cross-validation using out-of-fold predictions. Based on the S007 out-of-fold predictions and a confidence threshold of 0.65, the offline scene-control mapping evaluation produced 91.11% command coverage, 100.00% command accuracy among issued commands, a 0.00% false trigger rate, and an 8.89% abstention rate. These results demonstrate the feasibility of mapping individualized EEG prediction outputs into interpretable virtual scene-control parameters under the tested prototype conditions. The current version implements software-simulated haptic parameter mapping only. At present, no nanoscale haptic suit is available for direct experimental validation in this study. Therefore, this work does not report real haptic hardware performance or real tactile experience outcomes. This study does not claim general performance across users, real-time deployment performance, mind-reading capability, medical diagnostic use, validated user immersion improvement, or a completed virtual reality system. The S007 result is presented only as a best-case individualized prototype demonstration and should not be interpreted as equivalent performance across all users. The English manuscript is provided as the primary paper, together with a Chinese translation.
孟泽 吴 (Sat,) studied this question.