ViEEG: Hierarchical Visual Neural Representation for EEG Brain Decoding

  • 2025-09-02 09:03:26
  • Minxu Liu, Donghai Guan, Chuhang Zheng, Chunwei Tian, Jie Wen, Qi Zhu
  • 0

Abstract

Understanding and decoding brain activity into visual representations is afundamental challenge at the intersection of neuroscience and artificialintelligence. While EEG visual decoding has shown promise due to itsnon-invasive, and low-cost nature, existing methods suffer from HierarchicalNeural Encoding Neglect (HNEN)-a critical limitation where flat neuralrepresentations fail to model the brain's hierarchical visual processinghierarchy. Inspired by the hierarchical organization of visual cortex, wepropose ViEEG, a neuro-We further adopt hierarchical contrastive learning forEEG-CLIP representation alignment, enabling zero-shot object recognition.Extensive experiments on the THINGS-EEG dataset demonstrate that ViEEGsignificantly outperforms previous methods by a large margin in bothsubject-dependent and subject-independent settings. Results on the THINGS-MEGdataset further confirm ViEEG's generalization to different neural modalities.Our framework not only advances the performance frontier but also sets a newparadigm for EEG brain decoding. inspired framework that addresses HNEN. ViEEGdecomposes each visual stimulus into three biologically alignedcomponents-contour, foreground object, and contextual scene-serving as anchorsfor a three-stream EEG encoder. These EEG features are progressively integratedvia cross-attention routing, simulating cortical information flow fromlow-level to high-level vision.

 

Quick Read (beta)

loading the full paper ...