ELAI-SGCN: An explainable lightweight adaptive information-perceiving spiking graph convolutional network for EEG-based emotion recognition

Jingxin Liu; Zikai Song; Xihang Qiu; Ran Cai; Jian Zhang; Lixian Zhu; Fuze Tian; Bin Hu

doi:10.1016/j.neunet.2025.108413

ELAI-SGCN: An explainable lightweight adaptive information-perceiving spiking graph convolutional network for EEG-based emotion recognition

Jingxin Liu
, Zikai Song
, Xihang Qiu
, Ran Cai
, Jian Zhang
, Lixian Zhu^*
, Fuze Tian
, Bin Hu

^*此作品的通讯作者

Ministry of Education in China
Beijing Institute of Technology
Lanzhou University

科研成果: 期刊稿件 › 文章 › 同行评审

摘要

Background:Emotion recognition is increasingly essential for diagnosing mental disorders like depression and anxiety. Electroencephalography (EEG) is widely adopted for this purpose due to its high temporal resolution and non-invasive nature. However, existing EEG-based models often neglect the brain's dynamic neural connectivity, inadequately modeling spatial topology, and rely on extensive redundant data, increasing computational complexity and limiting performance. Methods: To address these limitations, we propose ELAI-SGCN, a lightweight and explainable framework for EEG analysis. ELAI-SGCN employs a trainable spiking encoder to transform raw EEG signals into sparse spike-based representations, preserving critical temporal dynamics in an event-driven manner to provide critical information for the subsequent Spiking Neural Network model. Simultaneously, the graph convolution module adaptively models inter-regional connectivity through efficient spike-based operations, enabling interpretable and resource-efficient EEG analysis. Results:We validated ELAI-SGCN on the DEAP and SEED emotion recognition datasets. On the DEAP dataset, the model achieved classification accuracies of 87.08 % for valence and 89.96 % for arousal, with only 60.48 K parameters and a computational cost of 0.36 M FLOPs. On the SEED dataset, it reached 94.63 % accuracy for three-class emotion recognition with 99.8 % fewer parameters and a more than 250-fold decrease in FLOPs compared to Dynamic Graph Convolutional Neural Networks. Conclusion:ELAI-SGCN introduces a novel spiking-based dynamic graph convolutional framework that enables efficient and interpretable modeling of EEG spatiotemporal dynamics. ELAI-SGCN outperformed most existing approaches in both accuracy and computational efficiency, demonstrating its suitability for lightweight, real-time EEG-based emotion recognition. Its lightweight design supports deployment on bedside clinical devices, offering a promising solution for emotion recognition and laying a foundation for next-generation intelligent psychological assessment systems with broad clinical potential.

源语言	英语
文章编号	108413
期刊	Neural Networks
卷	197
DOI	https://doi.org/10.1016/j.neunet.2025.108413
出版状态	已出版 - 5月 2026
已对外发布	是

访问文件

10.1016/j.neunet.2025.108413

其它文件与链接

链接到 Scopus 的出版物

引用此

@article{9b3d498312eb441ea6b7b76a18e67eca,

title = "ELAI-SGCN: An explainable lightweight adaptive information-perceiving spiking graph convolutional network for EEG-based emotion recognition",

abstract = "Background:Emotion recognition is increasingly essential for diagnosing mental disorders like depression and anxiety. Electroencephalography (EEG) is widely adopted for this purpose due to its high temporal resolution and non-invasive nature. However, existing EEG-based models often neglect the brain's dynamic neural connectivity, inadequately modeling spatial topology, and rely on extensive redundant data, increasing computational complexity and limiting performance. Methods: To address these limitations, we propose ELAI-SGCN, a lightweight and explainable framework for EEG analysis. ELAI-SGCN employs a trainable spiking encoder to transform raw EEG signals into sparse spike-based representations, preserving critical temporal dynamics in an event-driven manner to provide critical information for the subsequent Spiking Neural Network model. Simultaneously, the graph convolution module adaptively models inter-regional connectivity through efficient spike-based operations, enabling interpretable and resource-efficient EEG analysis. Results:We validated ELAI-SGCN on the DEAP and SEED emotion recognition datasets. On the DEAP dataset, the model achieved classification accuracies of 87.08 \% for valence and 89.96 \% for arousal, with only 60.48 K parameters and a computational cost of 0.36 M FLOPs. On the SEED dataset, it reached 94.63 \% accuracy for three-class emotion recognition with 99.8 \% fewer parameters and a more than 250-fold decrease in FLOPs compared to Dynamic Graph Convolutional Neural Networks. Conclusion:ELAI-SGCN introduces a novel spiking-based dynamic graph convolutional framework that enables efficient and interpretable modeling of EEG spatiotemporal dynamics. ELAI-SGCN outperformed most existing approaches in both accuracy and computational efficiency, demonstrating its suitability for lightweight, real-time EEG-based emotion recognition. Its lightweight design supports deployment on bedside clinical devices, offering a promising solution for emotion recognition and laying a foundation for next-generation intelligent psychological assessment systems with broad clinical potential.",

keywords = "Electroencephalography, Emotion recognition, Graph convolution, Spiking neural networks",

author = "Jingxin Liu and Zikai Song and Xihang Qiu and Ran Cai and Jian Zhang and Lixian Zhu and Fuze Tian and Bin Hu",

note = "Publisher Copyright: {\textcopyright} 2025",

year = "2026",

month = may,

doi = "10.1016/j.neunet.2025.108413",

language = "English",

volume = "197",

journal = "Neural Networks",

issn = "0893-6080",

publisher = "Elsevier Ltd.",

}

TY - JOUR

T1 - ELAI-SGCN

T2 - An explainable lightweight adaptive information-perceiving spiking graph convolutional network for EEG-based emotion recognition

AU - Liu, Jingxin

AU - Song, Zikai

AU - Qiu, Xihang

AU - Cai, Ran

AU - Zhang, Jian

AU - Zhu, Lixian

AU - Tian, Fuze

AU - Hu, Bin

PY - 2026/5

Y1 - 2026/5

N2 - Background:Emotion recognition is increasingly essential for diagnosing mental disorders like depression and anxiety. Electroencephalography (EEG) is widely adopted for this purpose due to its high temporal resolution and non-invasive nature. However, existing EEG-based models often neglect the brain's dynamic neural connectivity, inadequately modeling spatial topology, and rely on extensive redundant data, increasing computational complexity and limiting performance. Methods: To address these limitations, we propose ELAI-SGCN, a lightweight and explainable framework for EEG analysis. ELAI-SGCN employs a trainable spiking encoder to transform raw EEG signals into sparse spike-based representations, preserving critical temporal dynamics in an event-driven manner to provide critical information for the subsequent Spiking Neural Network model. Simultaneously, the graph convolution module adaptively models inter-regional connectivity through efficient spike-based operations, enabling interpretable and resource-efficient EEG analysis. Results:We validated ELAI-SGCN on the DEAP and SEED emotion recognition datasets. On the DEAP dataset, the model achieved classification accuracies of 87.08 % for valence and 89.96 % for arousal, with only 60.48 K parameters and a computational cost of 0.36 M FLOPs. On the SEED dataset, it reached 94.63 % accuracy for three-class emotion recognition with 99.8 % fewer parameters and a more than 250-fold decrease in FLOPs compared to Dynamic Graph Convolutional Neural Networks. Conclusion:ELAI-SGCN introduces a novel spiking-based dynamic graph convolutional framework that enables efficient and interpretable modeling of EEG spatiotemporal dynamics. ELAI-SGCN outperformed most existing approaches in both accuracy and computational efficiency, demonstrating its suitability for lightweight, real-time EEG-based emotion recognition. Its lightweight design supports deployment on bedside clinical devices, offering a promising solution for emotion recognition and laying a foundation for next-generation intelligent psychological assessment systems with broad clinical potential.

AB - Background:Emotion recognition is increasingly essential for diagnosing mental disorders like depression and anxiety. Electroencephalography (EEG) is widely adopted for this purpose due to its high temporal resolution and non-invasive nature. However, existing EEG-based models often neglect the brain's dynamic neural connectivity, inadequately modeling spatial topology, and rely on extensive redundant data, increasing computational complexity and limiting performance. Methods: To address these limitations, we propose ELAI-SGCN, a lightweight and explainable framework for EEG analysis. ELAI-SGCN employs a trainable spiking encoder to transform raw EEG signals into sparse spike-based representations, preserving critical temporal dynamics in an event-driven manner to provide critical information for the subsequent Spiking Neural Network model. Simultaneously, the graph convolution module adaptively models inter-regional connectivity through efficient spike-based operations, enabling interpretable and resource-efficient EEG analysis. Results:We validated ELAI-SGCN on the DEAP and SEED emotion recognition datasets. On the DEAP dataset, the model achieved classification accuracies of 87.08 % for valence and 89.96 % for arousal, with only 60.48 K parameters and a computational cost of 0.36 M FLOPs. On the SEED dataset, it reached 94.63 % accuracy for three-class emotion recognition with 99.8 % fewer parameters and a more than 250-fold decrease in FLOPs compared to Dynamic Graph Convolutional Neural Networks. Conclusion:ELAI-SGCN introduces a novel spiking-based dynamic graph convolutional framework that enables efficient and interpretable modeling of EEG spatiotemporal dynamics. ELAI-SGCN outperformed most existing approaches in both accuracy and computational efficiency, demonstrating its suitability for lightweight, real-time EEG-based emotion recognition. Its lightweight design supports deployment on bedside clinical devices, offering a promising solution for emotion recognition and laying a foundation for next-generation intelligent psychological assessment systems with broad clinical potential.

KW - Electroencephalography

KW - Emotion recognition

KW - Graph convolution

KW - Spiking neural networks

UR - https://www.scopus.com/pages/publications/105025111753

U2 - 10.1016/j.neunet.2025.108413

DO - 10.1016/j.neunet.2025.108413

M3 - Article

C2 - 41420938

AN - SCOPUS:105025111753

SN - 0893-6080

VL - 197

JO - Neural Networks

JF - Neural Networks

M1 - 108413

ER -

ELAI-SGCN: An explainable lightweight adaptive information-perceiving spiking graph convolutional network for EEG-based emotion recognition

摘要

访问文件

其它文件与链接

学术指纹

引用此