The researchers report that under strict leave-one-subject-out conditions on the WAY-EEG-GAL dataset, the framework achieved an R2 of 0.817 in offline settings and 0.793 in simulated real-time evaluation with latency suitable for real-time deployment, demonstrating strong cross-subject generalization. This study is a preprint and has not been peer reviewed.
Preprint: EEGForceFusion for Subject-Independent Grasp Force Decoding
Summary
A preprint proposes EEGForceFusion, a hybrid EEG decoding framework that jointly models continuous and tokenized representations for grasp-force decoding, where continuous decoding is limited by complex temporal dynamics, high inter-subject variability and poor generalization. The framework combines convolutional-recurrent representation learning, quantized tokenization and Transformer-based temporal modeling in a unified fusion regression architecture to capture both fine-grained neural structure and long-range temporal dependencies. Under strict leave-one-subject-out cross-validation on the WAY-EEG-GAL dataset, it reached an offline R² of 0.817 and a simulated real-time R² of 0.793, with latency suited to real-time deployment. The study has not been peer reviewed.
Why it matters
The leave-one-subject-out R² of 0.817 offline and 0.793 in simulated real time is significant because it demonstrates cross-subject generalization — decoding grasp force without per-user calibration, the practical hurdle keeping force decoding out of real-time, subject-independent BCI deployments.
Compiled by BCIwiki from public sources
Sources · 1
arxiv.org 2026-07-27