Fusion of Multiscale Features Via Centralized Sparse-attention Network for EEG Decoding

TL;DR

The paper tackles EEG decoding by addressing spatiotemporal heterogeneity with a multi-branch architecture in which each temporal scale has its own spatial extractor. It introduces EEG-CSANet, a main-auxiliary fusion framework that uses multiscale self-attention in the main branch and sparse cross-attention in auxiliary branches to robustly fuse features. Across five public EEG datasets, EEG-CSANet achieves state-of-the-art accuracies and demonstrates strong robustness through extensive ablations and visualizations. The authors also release the code on GitHub to promote reproducibility and baseline development in EEG decoding.

Abstract

Electroencephalography (EEG) signal decoding is a key technology that translates brain activity into executable commands, laying the foundation for direct brain-machine interfacing and intelligent interaction. To address the inherent spatiotemporal heterogeneity of EEG signals, this paper proposes a multi-branch parallel architecture, where each temporal scale is equipped with an independent spatial feature extraction module. To further enhance multi-branch feature fusion, we propose a Fusion of Multiscale Features via Centralized Sparse-attention Network (EEG-CSANet), a centralized sparse-attention network. It employs a main-auxiliary branch architecture, where the main branch models core spatiotemporal patterns via multiscale self-attention, and the auxiliary branch facilitates efficient local interactions through sparse cross-attention. Experimental results show that EEG-CSANet achieves state-of-the-art (SOTA) performance across five public datasets (BCIC-IV-2A, BCIC-IV-2B, HGD, SEED, and SEED-VIG), with accuracies of 88.54%, 91.09%, 99.43%, 96.03%, and 90.56%, respectively. Such performance demonstrates its strong adaptability and robustness across various EEG decoding tasks. Moreover, extensive ablation studies are conducted to enhance the interpretability of EEG-CSANet. In the future, we hope that EEG-CSANet could serve as a promising baseline model in the field of EEG signal decoding. The source code is publicly available at: https://github.com/Xiangrui-Cai/EEG-CSANet

Figures (9)

• Figure 1: The overall architecture of EEG-CSANet, including data augmentation, multi-branch convolution, feature fusion architecture, and temporal convolutional.
• Figure 2: Traditional multiscale feature extraction method and the proposed multibranch feature extraction method. (a) Traditional multiscale feature extraction method. (b) Proposed multibranch feature extraction method.
• Figure 3: Ablation experiments results on three EEG datasets (BCIC-IV-2A, SEED, and SEED-VIG).
• Figure 4: Performance comparison of individual and fused branches in EEG-CSANet on BCIC-IV-2A
• Figure 5: The UMAP Visualization of EEG-CSANet Features Before and After Training. (a) BCIC-IV-2A for subject 1. (b) BCIC-IV-2B for subject 7. (c) HGD for subject 9. (d) SEED for subject 11. (e) SEED-VIG for subject 18.