基于EEG信号的脑疲劳检测方法

Method of brain fatigue detection based on EEG signal

摘要

脑疲劳是指大脑因长时间高度集中和过度记忆思考,导致脑部神经元活性降低、能耗增加的现象。其危害巨大,不仅降低人的注意力和反应力,长期还可能引发健康问题。为预防大脑疲劳对人们所带来的身心危害,提出一种基于多特征融合和改进LSTM神经网络的脑疲劳检测方法。该方法以脑电信号为分析对象,综合运用小波包变换、小波包熵以及样本熵进行多维度特征提取。将不同方法提取的特征采用单一和融合的方法,分别输入分类器进行识别。实验结果表明,所选特征组合对大脑清醒与疲劳状态具有良好区分度,各类分类模型均能实现有效识别。其中改进的LSTM网络分类方法识别率最高,能够达到95.83%。多组对照实验验证,该方法在识别精度、算法稳定性上均优于常规方法。

Abstract

Brain fatigue refers to a phenomenon where the activity of brain neurons decreases and energy consumption increases due to prolonged high concentration, excessive memory and thinking. It is extremely harmful, as it not only reduces peoples′ attention and responsiveness, but may also lead to health problems in the long run. To prevent the physical and mental hazards caused by brain fatigue to people, this paper proposes a brain fatigue detection method based on multi-feature fusion and improved LSTM neural network. Taking electroencephalographic (EEG) signals as the analysis object, this method comprehensively uses wavelet packet transform, wavelet packet entropy and sample entropy for multi-dimensional feature extraction. The features extracted by different methods are input into the classifier for recognition in the form of single feature and feature fusion respectively. Experimental results show that the selected feature combination has a good distinguishability between the awake and fatigued states of the brain, and all types of classification models can achieve effective recognition. Among them, the improved LSTM network classification method has the highest recognition rate, reaching 95.83%. Multiple comparative experiments verify that this method outperforms conventional methods in terms of recognition accuracy and algorithm stability.

关键词

疲劳检测 / 脑电 / 多特征融合 / 特征分类 / 神经网络

Key words

fatigue detection / EEG / multi-feature fusion / feature classification / neural network

[1] 俞康, 陶庆, 殷润生, 等. 用于脑控系统的一种去噪脑电疲劳检测方法[J]. 现代电子技术, 2023, 46(2): 176-180.
[2] 张勉, 易灿南, 蒋复量, 等. 疲劳对操纵员监视行为影响机制的脑电试验[J]. 中国安全科学学报, 2024, 34(9): 50-58.
[3] 许子明, 牛一帆, 温旭云, 等. 基于脑电信号的认知负荷评估综述[J]. 航天医学与医学工程, 2021, 34(4): 339-348.
[4] Jin H, Xiao M, Liu L, et al. Relationship between physical fatigue and mental fatigue based on multimodal measurement under different load levels[J]. Ergonomics, 2024, 67(11): 11-16.
[5] 管凯捷, 姚康, 任谊文, 等. 基于头动与眼动的脑疲劳检测方法研究[J]. 航天医学与医学工程, 2020, 33(3): 214-220.
[6] Wen Q, Zha F, Shan L, et al. Electroacupuncture attenuates middle cerebral artery occlusion-induced learning and memory impairment by regulating microglial polarization in hippocampus[J]. The International Journal of Neuroscience, 2024, 12(5): 11-16.
[7] 王洪涛, 殷浩钧, 陈创泉, 等. 基于脑电信号的驾驶疲劳检测综述[J]. 华中科技大学学报(自然科学版), 2022, 50(11): 54-65,78.
[8] Li Y, Wang D, Liu F. The auto-correlation function aided sparse support matrix machine for eeg-based fatigue detection[J]. IEEE Transactions on Circuits and Systems II: Express Briefs, 2022, 70(2): 836-840.
[9] 王晓璐, 高翔, 许敏鹏, 等. 基于睡眠剥夺模型的脑力疲劳检测方法研究[J]. 生物医学工程学杂志, 2015, 32(3): 497-502.
[10] Albawardi H, Almoaibed A, Abbas N A, et al. Design of low-cost steady state visually evoked potential-based brain computer interface using openBCI and neuromore[C]// 2021 4th International Conference on Bio-Engineering for Smart Technologies (BioSMART). Paris: IEEE, 2021: 1-4.
[11] 郑赟, 马玉良, 孙明旭, 等. 一种疲劳驾驶检测中的脑电信号通道选择方法[J]. 中国生物医学工程学报, 2022, 41(4): 402-411.
[12] 尹瑾, 刘爱萍, 李畅, 等. 基于并行CNN和Transformer的脑电降噪网络[J]. 信号处理, 2023, 39(8):1419-1432.
[13] 冯凯, 董秀成, 刘栋博. 基于经验模态分解-小波包变换的表面肌电信号手势识别[J]. 中国医学物理学杂志, 2021, 38(4):461-467.
[14] 张崇, 郑崇勋, 张连毅, 等. 基于多导脑电特征的生理性精神疲劳分析[J]. 西安交通大学学报, 2007(2): 250-254.
[15] 沈晓燕, 王雪梅, 王燕. 基于样本熵和模式识别的脑电信号识别算法研究[J]. 计算机工程与科学, 2020, 42(8):1482-1488.
[16] 沈振乾, 李文强, 任甜甜, 等. 基于CNN-NLSTM的脑电信号注意力状态分类方法[J]. 中文信息学报, 2024, 38(4): 38-49.
[17] 张莉, 李承桓, 秦美翠, 等. 基于LSTM网络优化的电力客服语音情感识别系统[J]. 自动化技术与应用, 2025, 44(10):66-69.
[18] 翟政中, 黄永涛, 刘华云. 基于LSTM的电气化铁路牵引供电故障识别研究[J]. 自动化技术与应用, 2025, 44(6):95-98, 179.

基金资助

国家自然科学基金项目(61801197);江苏省自然科学基金项目(BK20181004);江苏高校青蓝工程;徐州市科技计划项目资助(KC22290);江苏师范大学研究生科研与实践创新计划资助项目(2024XKT0248)