BCIwiki (bciwiki.com) — Researchers report that ATCNet-CIAM, a framework for multi-session motor imagery EEG signal classification, achieves 86.32% and 87.96% accuracy on BCI IV-2a and BCI IV-2b respectively under the standard protocol, according to a preprint posted to arXiv on July 26, 2026 that has not yet undergone peer review.
Motor imagery-based EEG is widely used in non-invasive brain-computer interfaces, but robust decoding across sessions and subjects remains a core challenge. ATCNet-CIAM combines temporal convolution, channel attention, and a CIAM module for multi-session motor imagery EEG classification. On the WBCIC-MI dataset, the within-session two-class and three-class tasks reach 89.46% and 83.64% respectively. The manuscript has been accepted for publication at the International Conference on Intelligence Systems and Robotics for Sustainable Development (ISRSD) 2026.