Autoregressive Visual Decoding from EEG Signals
#EEG decoding #Autoregressive models #Brain-computer interface #Visual reconstruction #Machine learning #Transformer architecture #LaBraM #VQ-VAE
📌 Key Takeaways
- Researchers developed AVDE framework to decode visual information from EEG signals
- Current methods face challenges in bridging the modality gap between EEG and image data
- AVDE uses only 10% of parameters while outperforming previous state-of-the-art methods
- The framework reflects hierarchical nature of human visual perception in its generation process
📖 Full Retelling
Sicheng Dai and three fellow researchers introduced AVDE, a lightweight framework for decoding visual information from electroencephalogram (EEG) signals, in a paper submitted to arXiv on February 26, 2026, addressing significant challenges in bridging the gap between brain activity and visual data representation. The researchers developed this innovative approach as current methods for translating EEG signals into images face substantial difficulties in handling the fundamental differences between these data types, requiring complex multi-stage adaptation processes that often introduce inconsistencies and error accumulation. Furthermore, existing solutions based on large-scale diffusion models create computational barriers that limit their practical implementation in real-world brain-computer interface applications. AVDE represents a significant advancement by employing a more streamlined approach that maintains direct connections between input EEG signals and reconstructed images while dramatically reducing computational requirements. The framework first leverages LaBraM, a pre-trained EEG model, fine-tuned through contrastive learning to align EEG and image representations. It then implements an autoregressive generative framework using a 'next-scale prediction' strategy, where images are encoded into multi-scale token maps via a pre-trained VQ-VAE, and a transformer progressively predicts finer-scale tokens starting from EEG embeddings as the coarsest representation. This hierarchical generation process not only produces coherent visual results but also reflects the natural progression of human visual perception, offering both efficiency and interpretability for practical BCI applications.
🏷️ Themes
Machine Learning, Brain-Computer Interface, Neuroscience, Computer Vision
📚 Related People & Topics
Autoregressive model
Representation of a type of random process
In statistics, econometrics, and signal processing, an autoregressive (AR) model is a representation of a type of random process; as such, it can be used to describe certain time-varying processes in nature, economics, behavior, etc. The autoregressive model specifies that the output variable depend...
Machine learning
Study of algorithms that improve automatically through experience
Machine learning (ML) is a field of study in artificial intelligence concerned with the development and study of statistical algorithms that can learn from data and generalize to unseen data, and thus perform tasks without explicit instructions. Within a subdiscipline in machine learning, advances i...
Entity Intersection Graph
No entity connections available yet for this article.
Original Source
--> Computer Science > Machine Learning arXiv:2602.22555 [Submitted on 26 Feb 2026] Title: Autoregressive Visual Decoding from EEG Signals Authors: Sicheng Dai , Hongwang Xiao , Shan Yu , Qiwei Ye View a PDF of the paper titled Autoregressive Visual Decoding from EEG Signals, by Sicheng Dai and 3 other authors View PDF HTML Abstract: Electroencephalogram signals have become a popular medium for decoding visual information due to their cost-effectiveness and high temporal resolution. However, current approaches face significant challenges in bridging the modality gap between EEG and image data. These methods typically rely on complex adaptation processes involving multiple stages, making it hard to maintain consistency and manage compounding errors. Furthermore, the computational overhead imposed by large-scale diffusion models limit their practicality in real-world brain-computer interface applications. In this work, we present AVDE, a lightweight and efficient framework for visual decoding from EEG signals. First, we leverage LaBraM, a pre-trained EEG model, and fine-tune it via contrastive learning to align EEG and image representations. Second, we adopt an autoregressive generative framework based on a "next-scale prediction" strategy: images are encoded into multi-scale token maps using a pre-trained VQ-VAE, and a transformer is trained to autoregressively predict finer-scale tokens starting from EEG embeddings as the coarsest representation. This design enables coherent generation while preserving a direct connection between the input EEG signals and the reconstructed images. Experiments on two datasets show that AVDE outperforms previous state-of-the-art methods in both image retrieval and reconstruction tasks, while using only 10% of the parameters. In addition, visualization of intermediate outputs shows that the generative process of AVDE reflects the hierarchical nature of human visual perception. These results highlight the potential of autoregressive mod...
Read full article at source