New OSPDIM framework improves EEG-based BCI adaptation with class imbalance

Researchers have developed OSPDIM, a novel online source-free domain adaptation framework designed to address label shifts in electroencephalography (EEG) based Brain-Computer Interfaces (BCIs). This method corrects geometric misalignment on the Riemannian manifold by introducing a manifold-constrained bias parameter that is optimized on-the-fly. Extensive experiments on motor imagery datasets demonstrate that OSPDIM significantly outperforms standard Riemannian baselines, particularly in online adaptation scenarios with severe class imbalance, offering a robust solution for practical BCI systems. AI

IMPACT Improves robustness of EEG-based BCIs in real-world scenarios with dynamic label shifts.

RANK_REASON The cluster contains a research paper detailing a new method for improving EEG-based BCIs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Brain-Computer Interfaces
• electroencephalography
• motor imagery
• OSPDIM
• Riemannian alignment
• Riemannian Centering Transformation

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →