基于EEG信号特征的脑力疲劳快速检测方法

张朋; 周前祥; 于洪强; 王川

doi:10.13700/j.bh.1001-5965.2021.0211

基于EEG信号特征的脑力疲劳快速检测方法

doi: 10.13700/j.bh.1001-5965.2021.0211

张朋¹,
周前祥¹,
于洪强²,
王川^3, ,

1.
北京航空航天大学生物与医学工程学院，北京 100191
2.
航天员科研训练中心，北京 100194
3.
中国人民解放军海军特色医学中心，上海 200433

基金项目: 载人航天空间医学试验项目(HYZHXM03003)；武器装备军内科研项目基金(20AZ0702)

详细信息

作者简介:
张朋

周前祥

王川等：基于EEG信号特征的脑力疲劳快速检测方法研究 5

通讯作者:
E-mail： hg04381@163.com

中图分类号: V7
计量
- 文章访问数: 892
- HTML全文浏览量: 128
- PDF下载量: 115
- 被引次数: 0
出版历程
- 收稿日期: 2021-04-23
- 录用日期: 2021-07-18
- 网络出版日期: 2021-07-24
- 整期出版日期: 2023-01-30

Fast detection method of mental fatigue based on EEG signal characteristics

1.
School of Biological Science and Medical Engineering，Beihang University，Beijing 100191，China
2.
Astronaut Research and Training Center，Beijing 100194，China
3.
Naval Medical Center of PLA，Shanghai 200433，China

Funds: Manned Space Medical Experiment Project (HYZHXM03003); Military Scientific Research Project Fund of Weapons and Equipment of China (20AZ0702)

More Information

Corresponding author: E-mail：hg04381@163.com

摘要

摘要:
空间站飞行过程中航天员容易产生脑力疲劳，其是影响作业效率和引起失误的主要因素。为此，研究人体脑力疲劳的快速检测方法，将有利于保障在轨运行安全。脑电波(EEG)的特征变化能够反映出大脑疲劳状态，但现有EEG方法分析脑力疲劳时需要多个导联的信号，这严重限制了其在空间站环境中的实际应用。通过地基实验，采用36 h睡眠剥夺的方式成功诱发出45名受试者的多种脑力疲劳状态。针对EEG信号的非平稳性，设计的8层db4小波变换结构，有效分解出了δ、θ、α和β脑节律波。先使用方差分析( ANOVA)和Logistic回归筛选出脑力疲劳敏感特征，再依据脑力疲劳敏感特征数量进一步筛选出脑力疲劳敏感导联，应用6个敏感导联的特征分别构建了随机森林回归模型。加权融合6个导联处的回归模型，形成脑力疲劳快速检测模型，其平均精确率高达85.25%。
- 脑力疲劳 /
- 脑电波 /
- 脑电特征 /
- 随机森林 /
- 检测模型
Abstract:
During the flight in space station, astronauts are prone to mental fatigue, which is the main factor that affects the efficiency of operations and causes errors. For this reason, studying rapid detection methods for human mental fatigue will help ensure the safety of on-orbit operations. The characteristic changes of the electroencephalogram (EEG) can reflect the fatigue state of the brain. Still, the existing EEG method requires multiple lead signals when analyzing mental fatigue, which seriously limits its practical application in the space station environment. This study successfully induced various mental fatigue states in 45 subjects through a foundation experiment using 36 hours of sleep deprivation. Aiming at the non-stationarity of EEG signals, the designed 8-layer db4 wavelet transform structure effectively decomposes δ, θ, α, and β brain rhythm waves. First, screen out the mental fatigue sensitivity characteristics using analysis of variance (ANOVA) and Logistic regression. Secondly, according to the number of sensitive features of mental fatigue, the sharp leads of mental fatigue were further screened out. Finally, the characteristics of 6 keen leaders were used to construct random forest regression models. Finally, the weighted fusion of the regression models at 6 leads to a rapid detection model of mental fatigue, with an average accuracy rate of up to 85.25%.
- mental fatigue /
- electroencephalogram /
- electroencephalogram characteristics /
- random forest /
- detection model

HTML全文

图 1 脑电信号采集场景

Figure 1. EEG acquisition scene

下载: 全尺寸图片幻灯片

图 2 小波变换结构

Figure 2. Wavelet transform structure

下载: 全尺寸图片幻灯片

图 3 小波变换提取的脑节律波

Figure 3. Brain rhythm wave extracted by wavelet transform

下载: 全尺寸图片幻灯片

图 4 脑力疲劳状态统计结果

Figure 4. Statistical results of mental fatigue

下载: 全尺寸图片幻灯片

图 5 前4个敏感特征的分布差异

Figure 5. Distribution differences of the first 4 sensitive features

下载: 全尺寸图片幻灯片

图 6 脑力疲劳敏感导联的特征统计

Figure 6. Feature statistics of sensitive leads of mental fatigue

下载: 全尺寸图片幻灯片

图 7 脑力疲劳检测模型的ROC曲线

Figure 7. ROC curve of mental fatigue detection model

下载: 全尺寸图片幻灯片

表 1 脑力疲劳主观评分及等级划分

Table 1. Subjective score and grade division of mental fatigue

脑力疲劳分级	主观评分
无疲劳感	s<2
轻微疲劳	2≤s<4
中度疲劳	4≤s<6
较严重疲劳	6≤s<8
很严重疲劳	s≥8

下载: 导出CSV

表 2 脑力疲劳的敏感特征统计

Table 2. Statistics of sensitive characteristics of mental fatigue

敏感特征	导联位置	P	AUC
α波绝对能量	F₃	0.000 4	0.695±0.054
θ/β相对能量	F_Z	0.000 6	0.682±0.049
Hjorth复杂性	O_Z	0.000 8	0.669±0.057
Renyi熵	F_Z	0.000 8	0.655±0.062
α/(δ+θ+α+β)相对能量	O₁	0.001 2	0.647±0.053
δ波绝对能量	O₂	0.001 8	0.649±0.042
Tsallis熵	F₃	0.003 7	0.603±0.057
Shannon熵	F₄	0.005 1	0.612±0.046
δ/(δ+θ+α+β)相对能量	FC₁	0.008 3	0.623±0.052
Hjorth活动性	F₄	0.012 9	0.598±0.051
(α+θ)/(α+β)相对能量	P_Z	0.014 6	0.587±0.045
β波绝对能量	O₁	0.017 3	0.591±0.050
Teager平均能量	F₄	0.024 0	0.584±0.049
对数能量熵	O_Z	0.026 9	0.575±0.054
(α+θ)/β相对能量	F₃	0.030 5	0.572±0.043
θ波绝对能量	O₁	0.034 9	0.565±0.038
Hjorth移动性	C_Z	0.038 5	0.560±0.047
α/β相对能量	F_Z	0.042 7	0.552±0.039
β/(δ+θ+α+β)相对能量	F_Z	0.048 3	0.554±0.042

下载: 导出CSV

表 3 脑力疲劳敏感导联的重要特征

Table 3. Important characteristics of mental fatigue sensitive leads

F_Z	O₁	F₄	O₂	O_Z	F₃
θ/β相对能量	α/(δ+θ+α+β)相对能量	Hjorth活动性	δ波绝对能量	Hjorth复杂性	α波绝对能量
Renyi熵	β波绝对能量	Shannon熵	α/(δ+θ+α+β)相对能量	Teager平均能量	(α+θ)/β相对能量
α/β相对能量	θ波绝对能量	θ/β相对能量	Hjorth复杂性	对数能量熵	Tsallis熵
β/(δ+θ+α+β)相对能量	Hjorth活动性	δ波绝对能量	θ/β相对能量	α/β相对能量	θ/β相对能量
δ波绝对能量	Teager平均能量	δ/(δ+θ+α+β)相对能量	(α+θ)/(α+β)相对能量	β/(δ+θ+α+β)相对能量	α/(δ+θ+α+β)相对能量
Teager平均能量	δ波绝对能量	(α+θ)/(α+β)相对能量	Teager平均能量	Shannon熵	δ波绝对能量
Hjorth活动性	θ/β相对能量	Hjorth移动性	对数能量熵	Hjorth移动性	Hjorth活动性
Renyi熵	Shannon熵	β波绝对能量	Hjorth移动性	β波绝对能量	α/β相对能量
β波绝对能量	Hjorth移动性	Renyi熵	α/β相对能量	(α+θ)/β相对能量	Renyi熵
(α+θ)/(α+β)相对能量	α/β相对能量	Teager平均能量	θ波绝对能量	θ/β相对能量	Teager平均能量

下载: 导出CSV

表 4 敏感导联回归模型的AUC值

Table 4. AUC value of sensitive lead regression model

敏感导联	无疲劳感	轻度疲劳	中度疲劳	较重疲劳	严重疲劳
F_Z	0.7951	0.7056	0.7854	0.7056	0.7469
O₁	0.7458	0.6891	0.7121	0.7905	0.7854
F₄	0.7956	0.7236	0.7541	0.7125	0.7965
O₂	0.7521	0.7532	0.6957	0.7558	0.7236
O_Z	0.8023	0.7456	0.7554	0.7415	0.7451
F₃	0.7685	0.6785	0.7028	0.7307	0.8218

下载: 导出CSV

表 5 脑力疲劳融合检测模型的比较

Table 5. Comparison of fusion detection models for mental fatigue

模型	Logistic	ANN	RF
SVM	0.4456↓	0.3895↑	0.0245↑
Logistic		0.2453↑	0.0187↑
ANN			0.0236↑
RF
注：↑指横向单元格中的方法优于纵向单元格中的方法， ↓指横向单元格中的方法差于纵向单元格中的方法。

下载: 导出CSV

表 6 脑力疲劳检测模型的分类结果

Table 6. Classification results of mental fatigue detection model

检测模型	精确率/%	召回率/%	F₁/%
无疲劳感	87.71	20.82	33.65
轻度疲劳	81.04	31.20	45.05
中度疲劳	82.71	29.86	43.88
较重疲劳	86.08	19.64	31.98
严重疲劳	88.73	18.29	30.33
平均值	85.25	23.96	36.98

下载: 导出CSV

参考文献(24)

[1]	田芸, 于赛克, 周前祥, 等. 眼动指标在脑力疲劳研究中的应用分析[J]. 人类工效学, 2015, 21(4): 69-73. doi: 10.13837/j.issn.1006-8309.2015.04.0014 TIAN Y, YU S K, ZHOU Q X, et al. Analysis of the application of eye movement index in the study of mental fatigue[J]. Chinese Journal of Ergonomics, 2015, 21(4): 69-73(in Chinese). doi: 10.13837/j.issn.1006-8309.2015.04.0014
[2]	牛国庆, 李师. 脑力疲劳与非疲劳状态眼动指标的判别[J]. 安全与环境学报, 2019, 19(1): 88-93. NIU G Q, LI S. Identification and determination of the mental fatigue status through the eyelid movement frequencies[J]. Journal of Safety and Environment, 2019, 19(1): 88-93(in Chinese).
[3]	FOONG R, ANG K K, QUEK C, et al. Assessment of the efficacy of EEG-based MI-BCI with visual feedback and EEG correlates of mental fatigue for upper-limb stroke rehabilitation[J]. IEEE Transactions on Bio-Medical Engineering, 2020, 67(3): 786-795. doi: 10.1109/TBME.2019.2921198
[4]	KAR G, HEDGE A. Effects of a sit-stand-walk intervention on musculoskeletal discomfort, productivity, and perceived physical and mental fatigue, for computer-based work[J]. International Journal of Industrial Ergonomics, 2020, 78(4): 102983.
[5]	MONTEIRO T G, SKOURUP C, ZHANG H X. Using EEG for mental fatigue assessment: A comprehensive look into the current state of the art[J]. IEEE Transactions on Human-Machine Systems, 2019, 49(6): 599-610. doi: 10.1109/THMS.2019.2938156
[6]	MARTIN K, THOMPSON K G, KEEGAN R, et al. Mental fatigue does not affect maximal anaerobic exercise performance[J]. European Journal of Applied Physiology, 2015, 115(4): 715-725. doi: 10.1007/s00421-014-3052-1
[7]	HOLMES G P, KAPLAN J E, GANTZ N M, et al. Chronic fatigue syndrome: A working case definition[J]. Annals of Internal Medicine, 1988, 108(3): 387-389. doi: 10.7326/0003-4819-108-3-387
[8]	MCCORMICK F, KADZIELSKI J, LANDRIGAN C P, et al. Surgeon fatigue: A prospective analysis of the incidence, risk, and intervals of predicted fatigue-related impairment in residents[J]. Archives of Surgery, 2012, 147(5): 430-435.
[9]	邱健, 赵显超, 程金湘, 等. 健康青年男性脑力疲劳模型构建以及基于节律类型的分析[J]. 中风与神经疾病杂志, 2019, 36(11): 1008-1012. QIU J, ZHAO X C, CHENG J X, et al. Construction of mental fatigue model of healthy young men and analysis based on different chronotype[J]. Journal of Apoplexy and Nervous Diseases, 2019, 36(11): 1008-1012(in Chinese).
[10]	WASCHER E, RASCH B, SÄNGER J, et al. Frontal theta activity reflects distinct aspects of mental fatigue[J]. Biological Psychology, 2014, 96: 57-65. doi: 10.1016/j.biopsycho.2013.11.010
[11]	MÖCKEL T, BESTE C, WASCHER E. The effects of time on task in response selection - An ERP study of mental fatigue[J]. Scientific Reports, 2015, 5: 10113. doi: 10.1038/srep10113
[12]	BROWNSBERGER J, EDWARDS A, CROWTHER R, et al. Impact of mental fatigue on self-paced exercise[J]. International Journal of Sports Medicine, 2013, 34(12): 1029-1036. doi: 10.1055/s-0033-1343402
[13]	HOPSTAKEN J F, Van Der LINDEN D, BAKKER A B, et al. A multifaceted investigation of the link between mental fatigue and task disengagement[J]. Psychophysiology, 2015, 52(3): 305-315. doi: 10.1111/psyp.12339
[14]	TELEŃCZUK B, BAKER S N, KEMPTER R, et al. Correlates of a single cortical action potential in the epidural EEG[J]. NeuroImage, 2015, 109: 357-367. doi: 10.1016/j.neuroimage.2014.12.057
[15]	ROELANDS B, DE PAUW K, MEEUSEN R. Neurophysiological effects of exercise in the heat[J]. Scandinavian Journal of Medicine & Science in Sports, 2015, 25(S1): 65-78.
[16]	王群, 程佳, 刘志文. 一种新的脑电信号睡眠分期方法[J]. 航天医学与医学工程, 2015, 28(1): 22-27. doi: 10.16289/j.cnki.1002-0837.2015.01.004 WANG Q, CHENG J, LIU Z W. A novel sleep staging method of EEG signals[J]. Space Medicine & Medical Engineering, 2015, 28(1): 22-27(in Chinese). doi: 10.16289/j.cnki.1002-0837.2015.01.004
[17]	罗志增, 鲁先举, 周莹. 基于脑功能网络和样本熵的脑电信号特征提取[J]. 电子与信息学报, 2021, 43(2): 412-418. doi: 10.11999/JEIT191015 LUO Z Z, LU X J, ZHOU Y. EEG feature extraction based on brain function network and sample entropy[J]. Journal of Electronics & Information Technology, 2021, 43(2): 412-418(in Chinese). doi: 10.11999/JEIT191015
[18]	张丽平, 詹长安. 心算任务复杂度对脑电theta, alpha和beta波的影响[J]. 航天医学与医学工程, 2019, 32(3): 235-242. ZHANG L P, ZHAN C A. Impact of mental arithmetic complexity on theta, alpha and beta power of EEG[J]. Space Medicine & Medical Engineering, 2019, 32(3): 235-242(in Chinese).
[19]	TREJO L J, KNUTH K, PRADO R, et al. EEG-based estimation of mental fatigue: Convergent evidence for a three-state model[C]// Foundations of Augmented Cognition. Berlin : Springer, 2007: 201-211.
[20]	JAP B T, LAL S, FISCHER P, et al. Using EEG spectral components to assess algorithms for detecting fatigue[J]. Expert Systems With Applications, 2009, 36(2): 2352-2359. doi: 10.1016/j.eswa.2007.12.043
[21]	KAR S, BHAGAT M, ROUTRAY A. EEG signal analysis for the assessment and quantification of driver’s fatigue[J]. Transportation Research Part F:Traffic Psychology and Behaviour, 2010, 13(5): 297-306. doi: 10.1016/j.trf.2010.06.006
[22]	LIU J P, ZHANG C, ZHENG C X. EEG-based estimation of mental fatigue by using KPCA-HMM and complexity parameters[J]. Biomedical Signal Processing and Control, 2010, 5(2): 124-130. doi: 10.1016/j.bspc.2010.01.001
[23]	ZHANG C, WANG H, FU R R. Automated detection of driver fatigue based on entropy and complexity measures[J]. IEEE Transactions on Intelligent Transportation Systems, 2014, 15(1): 168-177. doi: 10.1109/TITS.2013.2275192
[24]	张振祥, 张璟. 关于日本《疲劳症状自评量表》(2002)[J]. 人类工效学, 2003, 9(3): 60-62. doi: 10.13837/j.issn.1006-8309.2003.03.018 ZHANG Z X, ZHANG J. About Japan’s self rating scale of fatigue symptoms 2002[J]. Chinese Ergonomics, 2003, 9(3): 60-62(in Chinese). doi: 10.13837/j.issn.1006-8309.2003.03.018