东北大学学报(自然科学版) ›› 2024, Vol. 45 ›› Issue (4): 600-608.DOI: 10.12068/j.issn.1005-3026.2024.04.018

• 资源与土木工程 • 上一篇    

基于生理信号的危险作业人员心理负荷识别研究

郝锐, 郑欣, 李怡霖   

  1. 东北大学 资源与土木工程学院,辽宁 沈阳 110819
  • 收稿日期:2022-11-11 出版日期:2024-04-15 发布日期:2024-06-26
  • 作者简介:郝 锐(1999-),女,陕西渭南人,东北大学硕士研究生
    郑 欣(1978-),女,辽宁本溪人,东北大学副教授,博士生导师.
  • 基金资助:
    国家重点研发计划项目(2021YFC3001303)

Research on Identifying the Psychological Load of Operators in Hazardous Operations Based on Physiological Signals

Rui HAO, Xin ZHENG, Yi-lin LI   

  1. School of Resources & Civil Engineering,Northeastern University,Shenyang 110819,China. Corresponding author: ZHENG Xin,E-mail: zhengxin@mail. neu. edu. cn
  • Received:2022-11-11 Online:2024-04-15 Published:2024-06-26

摘要:

为识别危险作业岗位作业人员的心理负荷,提高人机系统可靠性,以含能材料起爆作业诱导被试人员心理负荷,采集30名被试人员在静息状态和心理负荷下的心率、脑电图和眼动信号进行心理负荷识别研究.首先,采用配对t检验与秩和检验对采集的心率、脑电图和眼动信号进行统计分析,8种脑电、3种眼动及9种心率特征在静息状态和心理负荷下具有显著变化;其次,对初选获得的生理指标分别采用Pearson 相关分析、最大相关最小冗余(MRMR)算法和主成分分析(PCA)进行特征降维;最后,基于上述3种方法降维处理后得到生理指标采用Logistic Regression,KNN,SVM,XG-Boost,Decision Tree和Random Forest机器学习方法进行心理负荷识别.结果表明,基于MRMR的心理负荷特征选择结果,采用Random Forest机器学习方法具有更好的识别性能(ACC=0.917,SN=1.0,SP=0.857,F1=0.909,AUC=0.971).本研究为有效识别危险作业人员心理负荷提供了理论依据.

关键词: 危险作业, 心理负荷识别, 生理信号, 机器学习

Abstract:

To identify the psychological load of operators in hazardous operations and improve the reliability of man‐machine systems, the psychological load was induced by the detonation of energy?containing materials, and the heart rate, EEG (electroencephalogram), and eye movement signals of 30 subjects were collected for psychological load identification under the resting state and psychological load. Firstly, the paired t‐test and rank sum test were used to statistically analyze the collected heart rate, EEG and eye movement signals. Eight EEG, three eye movement, and nine heart rate features were significantly changed under the resting state and psychological load. Secondly, Pearson correlation analysis, maximum relevance minimum redundancy (MRMR) algorithm and principal component analysis (PCA) were applied to reduce dimension of the physiological indexes obtained from the preliminary selection. Finally, the physiological indicators obtained after dimensionality reduction based on the above three methods were used for psychological load identification by Logistic Regression, KNN, SVM, XG‐Boost, Decision Tree, and Random Forest machine learning methods. The results showed that the Random Forest machine learning method has better identification performance (ACC=0.917, SN=1.0, SP=0.857, F1=0.909, AUC=0.971) based on MRMR’s psychological load feature selection results. The current research provides a theoretical basis for the effective identification of the psychological load of operators in hazardous operations.

Key words: hazardous operation, psychological load identification, physiological signal, machine learning

中图分类号: