Marie Dominique Schmidt; Ioannis Iossifidis
Decoding motor intention early and reliably is central to responsive assistive devices and adaptive rehabilitation. In this study, we investigate how precisely movement direction and target location can be inferred from multichannel EMG signals, and how early this information becomes available relative to movement onset.
We present a pipeline combining data-driven temporal segmentation with classical and deep learning models. In a delayed reaching task with 25 spatial targets (separated by 14° in azimuth/altitude), we report strong multi-target prediction performance—e.g., about 80% accuracy with Random Forest and 75% with a CNN.
Beyond accuracy, we systematically analyze how performance changes with fewer channels, different feature sets, and different temporal windows. The results suggest that reliable intention decoding can be achieved even with substantially reduced data, supporting practical, real-time deployment.
Reference
arXiv: 2603.05418, 2026.
Links
BibTeX
@article{schmidt2026spatialtemporal,
title={The Spatial and Temporal Resolution of Motor Intention in Multi-Target Prediction},
author={Schmidt, Marie Dominique and Iossifidis, Ioannis},
journal={arXiv preprint arXiv:2603.05418},
year={2026}
}