DenoFlow: Flow Matching for SSVEP Denoising under Real Physiological Artifacts
DenoFlow proposes a novel method for cleaning EEG signals from SSVEP systems by framing it as a transport problem. Instead of learning a direct mapping, it regresses the velocity of the path between noisy and clean data points. This learned field is then integrated forward to produce denoised trials, supervised by a classifier trained jointly with the network.