Diffusion Autoencoded foundation model for reconstructing and enhacing EEG signals. Thinking about using in combination with the BrainAccess HALO EEG headset I have been looking at. A big issue for autistic hobbyist like myself is affordable EEG hardware, this kind of denoising model would be a lifesaver.
A generative model for reconstructing and enhancing EEG signals. 380M Parameters Diffusion Autoencoder
Montage-Agnostic Scaling. Works across any EEG channel configuration. Hardware-efficient gains. Improves signal quality and downstream performance. Cross-dataset generalization. Generalizes across acquisition systems. Designed for noisy, real-world EEG data.