Back to AI News
arxiv

SPSC improves zero-shot cross-subject EEG-to-image reconstruction

Researchers introduce Semantic Preservation and Subspace Calibration (SPSC), a framework for reconstructing images from EEG signals when the target subject provides no labels or image embeddings. SPSC combines Object Mask Contrastive Regularization during training with a Subspace Candidate Gate that performs label-free semantic correction at inference. In leave-one-subject-out experiments across ten subjects on THINGS-EEG, the method improves cross-subject reconstruction from 0.552 to 0.713 on CLIP and from 0.612 to 0.745 on Alex.5. The work targets semantic shift and low signal-to-noise ratio in EEG-based brain-to-image generation.

AI Brief

It reports a concrete zero-shot cross-subject EEG-to-image method with quantitative gains, useful for researchers working on brain-computer interfaces and generative models.

arXiv:2611.09380v1 Announce Type: cross Abstract: Pretrained vision-language models and diffusion generators have driven rapid progress in brain-to-image generation, while EEG-based reconstruction is still constrained by low signal-to-noise ratio and cross-subject semantic shifts. Zero-shot cross-subject reconstruction poses even greater challenges: EEG semantic embeddings learned from source subjects deviate from the manifold of image embeddings of the unseen target subject, particularly when neither target labels nor target image embeddings are available. To address this issue, we introduce the Semantic Preservation and Subspace Calibration (SPSC) framework, which first uses Object Mask Contrastive Regularization (OMCR) with object identity for backend training, and then constructs Subspace Candidate Gate (SCG) to achieve label-free semantic correction for target subjects during inference. Specifically, OMCR regularizes the conditioning space with object-level positive samples to preserve object-level feature neighborhoods within the reconstruction prior and low-level branch, while SCG exploits training image prototypes and batch statistics of unlabeled target EEG to select candidates for each task within subspaces of image prototypes. Under the leave-one-subject-out (LOSO) setting across ten subjects on the THINGS-EEG dataset, SPSC improves the cross-subject reconstruction from 0.552 to 0.713 on CLIP and from 0.612 to 0.745 on Alex.5. These significant gains indicate that the semantically calibrated conditions can better support generation for the target subjects under label-free inference.
Be the first to rate this article.

Comments & Questions (0)

Captcha
Please be respectful — let's keep the conversation friendly.

No comments yet

Be the first to comment!