基于CART决策树的管制员疲劳检测研究  

Study on Air Traffic Controllers’ Fatigue Detection Based on CART Decision Tree

作　　者：靳慧斌[1] 朱国蕾 吕川 Jin Huibin;Zhu Guolei;Lu Chuan(Civil Aviation University of China,Tianjin 300300,China)

机构地区：[1]中国民航大学

出　　处：《航天医学与医学工程》2018年第6期601-606,共6页Space Medicine & Medical Engineering

基　　金：国家自然科学基金(U1333112);中国民航局安全能力专项资金(TMSA2017-246-1/2);中央高校基本科研业务费专项资金(3122014B007)

摘　　要：目的为保障民航安全,开发实时检测管制员疲劳的技术方法,积极进行疲劳预警。方法通过模拟管制实验采集正常与剥夺睡眠状态下受试者的脑电、心电、呼吸、体温和眼动等多源数据并计算疲劳特征指标,同时记录受试者的主观疲劳程度和操作绩效,在此基础上建立识别管制疲劳状态的CART决策树模型。结果RR间期、LF/HF、快波/慢波、PERCLOS和扫视速度均与管制员疲劳呈较强相关,利用决策树融合这五项指标构建疲劳检测模型,其识别受试者正常与疲劳状态的准确率为94.4%,对5级疲劳度的预测准确度为77.5%。结论该模型可以为管制员疲劳监测和预警提供技术手段。Objective To detect air traffic controllers’ fatigue effectively and make warning early. Methods An experimental platform based on simulated air traffic control software was constructed. EEG,ECG,respiration,temperature and eye movements data of normal and sleep deprived subjects were collected by MP150 physiological polygraph and eye tracker and the fatigue characteristic index was calculated,and the degree of subjective fatigue and performance were recorded. On this basis,a CART decision tree model was established to identify the air traffic controllers’ fatigue. Results The RR interval,the ratio of LF to HF,the ratio of fast and slow wave,PERCLOS and saccadic velocity were all significantly correlated with the air traffic controllers’ fatigue. The CART decision tree was used to fuse these five indicators to build air traffic controllers’ fatigue detection model. The accuracy of judging the normal and sleep deprived group was 94. 4% and the accuracy of recognizing the five levels fatigue was 77. 5%. Conclusion It can provide detection method for air traffic controllers’ fatigue.

关 键 词：空中交通管制 疲劳检测 生理参数 眼动 决策树 

分 类 号：R857.1[医药卫生—航空、航天与航海医学]