首页|A spatiotemporal separable graph convolutional network for oddball paradigm classification under different cognitive-load scenarios
A spatiotemporal separable graph convolutional network for oddball paradigm classification under different cognitive-load scenarios
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
Elsevier
The application of flight automation systems has increased the demand for detecting the cognitive load of pilots. Event-related potentials (ERPs) based on electroencephalogram (EEG) signals contain crucial information regarding the human cognitive load. Accurate analysis of ERP signals are essential for the detection of cognitive load. However, existing ERP analysis methods typically rely on manual feature extraction or simple convolutional filters, whereas the spatiotemporal dependencies in EEG signals are disregarded. Herein, we propose a spatiotemporal separable graph convolutional network (STSGCN) to automatically extract spatiotemporal features in EEG signals. Utilising temporal-gate unit for temporal features and graph convolutional networks for spatial features, STSGCN merges temporal and spatial features using separable convolution. We validate the reliability of the STSGCN in classifying P300 ERPs under different cognitive-load scenarios. The results show that the STSGCN outperforms conventional convolutional neural networks in terms of accuracy and robustness, thus providing algorithmic support for the application of ERPs in actual-flight cognitive-load detection.
Graph convolutional networksMultivariate time seriesSpatiotemporal featuresBrain-computer interfaceElectroencephalographyP300 event-related potential
Yuangan Li、Ke Li、Shaofan Wang、Haopeng Wu、Pengjiao Li
展开 >
School of Aeronautic Science and Engineering, Beihang University, Beijing, China
NETL
NSTL
Elsevier
