Yimeng Zhang

EVA-Net proposes a two-stage framework that uses action videos as semantic priors to achieve strong subject-independent EEG motor decoding, significantly outperforming text-based methods.