CAIA: Cognitive-Guided Adaptive Blurring for EEG Visual Decoding
Researchers propose CAIA, a framework for EEG-based visual decoding that addresses information granularity mismatch and low signal-to-noise ratio. On the visual side, it simulates selective attention via adaptive blurring; on the EEG side, it uses neural oscillation priors and information bottleneck to enhance SNR. The method dynamically integrates center-biased and saliency-guided visual cues through cross-modal attention.
Key facts
- CAIA stands for Cognitive-guided Adaptive blurring with Information-Constrained Alignment.
- The framework targets EEG-based visual decoding.
- It addresses severe information granularity mismatch and low SNR of EEG signals.
- Existing approaches treat static visual features, ignoring dynamic selectivity and frequency specificity.
- Visual side simulates selective attention to reduce redundancy.
- EEG side leverages neural oscillation priors and information bottleneck mechanism.
- A cognitive-dynamics-based adaptive blurring mechanism is devised.
- It integrates center-biased and saliency-guided visual cues via cross-modal attention.
Entities
Institutions
- arXiv