基于功能性脑网络和图卷积网络的驾驶疲劳检测

doi:10.3969/j.issn.1674-8484.2025.02.005

摘要/Abstract

摘要：

为了解决在疲劳检测中构建功能性脑网络（FBN）时，设置阈值标准较为模糊的问题，该文提出设置固定阈值，采用图卷积网络（GCN）来优化学习脑网络图特征。文中在构建FBN时设置阈值为0.5，提取脑网络的度和聚类系数特征，并输入GCN模型，模型对图特征进行学习优化，实现检测分类。结果表明：该模型检测的准确率可以达到88.90%；利用度中心性发现脑网络中的14个重要电极，其中基于7个重要电极构建的GCN模型检测的准确率为87.2%，检测速度更快，综合性能优于基于30导的检测模型。

关键词: 图卷积网络（GCN）, 功能性脑网络（FBN）, 简化通道, 驾驶疲劳

Abstract:

To address the issue of ambiguous threshold criteria in constructing functional brain networks (FBN) for fatigue detection, this paper proposed to set a fixed threshold and employing graph convolutional networks (GCN) to optimize the learning of brain network graph features. A threshold of 0.5 was set for building the FBN, and the degree and clustering coefficient features of the network were extracted. These features were then input into the GCN, which learned and optimized the graph features for detection classification. The results show that the preposed model's detection accuracy has reached 88.90%. Furthermore, degree centrality identifies 14 significant electrodes within the brain network. Among them, the GCN model built on 7 key electrodes achieves an 87.2% detection accuracy, with faster detection speed and superior overall performance compared to the detection model based on 30 leads.

Key words: graph convolutional networks (GCN), functional brain networks (FBN), simplified channels, driver fatigue

中图分类号:

U491.31

徐军莉. 基于功能性脑网络和图卷积网络的驾驶疲劳检测[J]. 汽车安全与节能学报, 2025, 16(2): 226-233.

XU Junli. Driver fatigue detection based on functional brain networks and graph convolutional networks[J]. Journal of Automotive Safety and Energy, 2025, 16(2): 226-233.

图/表 13

参考文献 18

[1]	常文文, 闫光辉, 杨志飞, 等. 基于脑电熵值特征和功能连接的不同线型道路下驾驶状态检测[J]. 电子学报, 2023, 51(10): 2874-2883. doi: 10.12263/DZXB.20210885
	CHANG Wenwen, YAN Guanghui, YANG Zhifei, et al. Detection of driving status under different linear road types based on electroencephalogram entropy features and functional connectivity[J]. Acta Electron Sini, 2023, 51(10): 2874-2883. (in Chinese)
[2]	王家曜, 张震, 宋光乐, 等. 一种基于脑电信号的疲劳驾驶检测方法[J]. 控制工程, 2024, 31(6): 1091-1098.
	WANG Jiayao, ZHANG Zhen, SONG Guangle, et al. A method for fatigue driving detection based on electroencephalogram signals[J]. Contr Engi, 2024, 31(6): 1091-1098. (in Chinese)
[3]	张蓉, 张艺竞. 基于频谱分析的自动和手动驾驶疲劳的脑电特征比较[J]. 人类工效学, 2023, 29(6): 14-21.
	ZHANG Rong, ZHANG YiJing. Comparison of electroencephalogram features for fatigue in automatic and manual driving based on spectral analysis[J]. Huma Fact Ergon, 2023, 29(6): 14-21. (in Chinese)
[4]	康小刚. 基于脑电信号的驾驶疲劳状态检测及缓解方法研究[D]. 沈阳: 东北电力大学, 2022.
	KANG Xiaogang. Research on driving fatigue state detection and alleviation methods based on electroencephalogram signals[D]. Shenyang: North Elect Power Univ, 2022. (in Chinese)
[5]	商晓锋, 罗志增, 史红斐. 基于脑肌耦合导联选择和最小生成树网络的脑电特征提取[J]. 仪器仪表学报, 2022, 43(7): 191-198.
	SHANG Xiaofeng, LUO Zhenzeng, SHI Hongfei. EEG feature extraction based on brain-muscle coupling lead selection and minimum spanning tree network[J]. China J Sci Instrum, 2022, 43(7): 191-198. (in Chinese)
[6]	KONG Wangzeng, LIN Weicheng, Fabio B, et al. Investigating driver fatigue versus alertness using the granger causality network[J]. Sensors (Basel), 2015, 15(8): 19181-19198.
[7]	ZHANG Lingyun, QIU Taorong, LIN Zhiqiang, et al. Construction and application of functional brain network based on entropy[J]. Entropy. 2020, 22(11):1-19.
[8]	穆振东. 性别差异在功能性脑网络的驾驶疲劳检测中的影响[J]. 海峡科技与产业, 2019(2): 83-85.
	MU Zhendong. The impact of gender differences on driving fatigue detection in functional brain networks[J]. Strait Sci Ind, 2019 (2): 83-85. (in Chinese)
[9]	Wijk V, Bernadette C M, Stam C J, et al. Comparing brain networks of different size and connectivity density using graph theory[J]. PLoS One, 2010, 5(10): 1-13.
[10]	WANG Huwang, XU Qing, FU Rongrong. Study on the effect of man-machine response mode to relieve driving fatigue based on EEG and EOG[J]. Sensors, 2019, 19(22): 4883-4900.
[11]	GUO Shengnan, LIN Youfang, FENG Ning, et al. Attention based spatial-temporal graph convolutional networks for traffic flow forecasting[J]. Proc AAAI Conf Artif Intel, 2019, 33(1): 922-929.
[12]	CAI Yujun, GE Liuhao, LIU Jun, et al. Exploiting spatial-temporal relationships for 3D pose estimation via graph convolutional networks[C]// IEEE Int’l Conf Comput Visi. Seoul, Kerea. 2019: 2272-2281.
[13]	LIU Hanjie, ZHANG Jinren, LIU Qingshan, et al. Minimum spanning tree based graph neural network for emotion classification using EEG[J]. Neural Network. 2022, 145: 308-318.
[14]	LIAN Qi, OI Yu, PAN Gangwang, et al. Learning graph in graph convolutional neural networks for robust seizure prediction[J]. J Neural Engi, 2020, 17(3):1-14.
[15]	王晶晶, 李艳军, 曹愈远, 等. 基于卷积神经网络的疲劳驾驶识别[J]. 航空计算技术, 2022, 52(5): 60-63,68.
	WANG Jingjing, LL Yanjun, CAO Yuyuan, et al. Fatigue driving recognition based on convolutional neural networks[J]. Aviat Comput Tech, 2022, 52(5): 60-63, 68. (in Chinese)
[16]	GAO Dongrui, TANG Xue, WAN Manqing, et al. EEG driving fatigue detection based on log-Mel spectrogram and convolutional recurrent neural networks[J]. Front Neurosci, 2023, 17: 1-14.
[17]	范晓丽, 牛海燕, 周前祥, 等. 基于EEG的脑力疲劳特征研究[J]. 北京航空航天大学学报, 2016, 42(7): 1406-1413.
	FAN Xiaoli, NIU Haiyan, ZHOU Qianxiang, et al. Research on EEG-based mental fatigue features[J]. J Beijing Univ Aeron Astron, 2016, 42(7): 1406-1413. (in Chinese)
[18]	田野, 何陆宁, 刘天娇, 等. 趣味性听觉材料对驾驶疲劳的作用:来自EEG的证据[J]. 心理与行为研究, 2020, 18(4): 474-481.
	TIAN Ye, HE Lunning, LIU Tianjiao, et al. The effect of entertaining auditory materials on driving fatigue: evidence from EEG[J]. Res Psych Behav, 2020, 18(4): 474-481. (in Chinese)

编号	最高模型检测率/ %	对应阈值
1	90.83	0.75
2	70.00	0.1
3	90.00	0.3
4	65.83	0.5
5	70.83	0.45
6	77.50	0.70
7	92.50	0.60
8	75.83	0.85
9	78.33	0.75
10	68.33	0.5
11	70.83	0.8
12	68.33	0.90
13	82.50	0.55
14	67.50	0.8
15	87.50	0.8
16	75.00	0.55

编号	最高模型检测率/ %	对应阈值
1	90.83	0.75
2	70.00	0.1
3	90.00	0.3
4	65.83	0.5
5	70.83	0.45
6	77.50	0.70
7	92.50	0.60
8	75.83	0.85
9	78.33	0.75
10	68.33	0.5
11	70.83	0.8
12	68.33	0.90
13	82.50	0.55
14	67.50	0.8
15	87.50	0.8
16	75.00	0.55

隐藏层大小	16
学习率	0.001
dropout	0.3
图卷积层	2
优化器	Adam
早停法的阈值参数	10
迭代次数epoch	100

隐藏层大小	16
学习率	0.001
dropout	0.3
图卷积层	2
优化器	Adam
早停法的阈值参数	10
迭代次数epoch	100

电极	度中心值
电极	C3	CZ	C4	TP7	F3	CP3	FCZ	FC4	FT8	FC3	T3	T4	FT7	FZ
1	0.72	0.72	0.72	0.71	/	/	0.71	0.71	0.72	0.71	0.71	0.71	/	/
2	0.72	0.71	0.72	0.71	/	0.71	0.71	/	0.71	0.70	0.72	0.71	/	/
3	0.91	0.90	0.90	0.90	0.89	/	0.89	/	/	0.90	0.90	0.89	/	0.89
4	0.75	0.74	0.74	/	0.73	/	0.74	0.72	0.72	0.74	0.75	/	0.74	/
5	0.82	0.81	0.82	0.80	/	0.79	0.79	/	/	0.80	0.81	0.81	0.79	/
6	0.74	0.74	0.74	/	0.73	/	0.74	0.73	0.74	0.74	0.75	0.74	/	/
7	0.83	/	0.80	0.79	/	/	0.81	/	0.79	0.83	0.80	0.79	0.82	0.81
8	0.85	0.85	0.85	0.83	0.84	/	0.85	0.85	/	0.86	0.86	0.84	/	0.82
9	0.80	0.79	0.79	0.78	/	/	0.79	0.77	/	0.78	0.78	0.79	0.77	0.75
10	0.85	0.84	0.84	0.84	/	0.83	0.84	0.83	0.84	/	0.85	0.83	/	0.81
11	0.85	0.84	0.85	0.83	/	/	0.84	0.83	0.84	0.85	0.84	0.84	/	0.83
12	0.84	0.83	0.83	0.82	0.81	/	/	/	0.83	0.83	0.84	0.83	0.83	0.81
13	0.83	0.82	0.83	/	0.82	/	0.82	0.32	0.81	0.82	0.82	0.80	/	0.79
14	0.96	0.95	0.95	0.95	0.95	0.95	0.94	0.94	0.94	0.94	/	/	/	0.92
15	0.89	0.89	0.88	0.88	0.89	/	0.89	/	/	0.89	0.89	0.89	0.88	0.88
16	0.82	0.81	/	/	0.81	/	0.81	0.81	0.81	0.81	0.82	0.81	0.81	0.79