.png){kind=icon}
返回
Two-stream attentive spatial-temporal graph convolutional network for P300 detection in brain-computer interface
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2025.130874.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
P300 potential detection is a critical technology in the field of brain-computer interfaces (BCIs), with significant implications for rehabilitation and neuroscience research. Due to the distinctive spatiotemporal characteristics of P300 potentials, accurately extracting and modeling spatiotemporal features is essential for effective P300 detection. However, existing multi-channel electroencephalogram (EEG) signals-based P300 detection methods face two major challenges in capturing discriminative spatiotemporal features: (1) the inability to effectively model the non-Euclidean connectivity relationships among EEG electrode channels, and (2) the failure to simultaneously focus on P300-relevant critical electrode channels and time slices. To address these issues, we propose a novel Two-Stream Attentive Spatial-Temporal Graph Convolutional Network (TASTGCN) for P300 detection. Specifically, we design a two-stream spatial-temporal graph convolution module (TSTGCM): the spatial GCN stream utilizes an adjacency matrix adaptively learned from electrode features to flexibly capture dynamic brain connectivity patterns, while the temporal GCN stream extracts temporal dependencies in parallel. Furthermore, a spatiotemporal attention refinement module (STARM) adaptively weighs different electrodes and time points to perceive the most salient P300-related information. Finally, an attentive feature fusion module (AFFM) based on a cross-attention mechanism is employed to integrate the refined features and generate the final discriminative representations. Extensive experiments on public datasets demonstrate that TASTGCN significantly outperforms state-of-the-art methods, validating its ability to capture the discriminative spatiotemporal features of P300 potentials.
Keyword:
P300 detection
Brain-computer interfaces (BCIs)
Graph neural network (GNN)
Electroencephalogram (EEG)
期刊
IF:
7.5
论文数:
3.0W
被引数:
10.2W
机构
引用论文
分享收藏American Electroencephalographic Society Guidelines for Standard Electrode Position Nomenclature
PREAI
分享收藏
Combining complexity measures of EEG data: multiplying measures reveal previously hidden information.png)
IF0
PREAI
分享收藏
查看更多内容