工程科学学报，第41卷，第6期：817-823,2019年6月 Chinese Journal of Engineering,Vol.41,No.6:817-823,June 2019 D0:10.13374/j.issn2095-9389.2019.06.014;http:/journals..usth.edu.cn 基于深度学习的人体低氧状态识别 于露，金龙哲区，王梦飞，徐明伟 北京科技大学土木与资源工程学院，北京100083 ☒通信作者，E-mail:in@ustb.edu.cn 摘要通过低氧实验提出一种快速识别人体低氧状态的方法.通过搭建深层神经网络训练实验数据识别氧气体积分数 (16%-21%)与人体可耐受极端低氧气体积分数(15.5%-16%)条件下光电容积脉搏波(photoplethysmography,PPG)信号， 获得人体生理状态的模式识别网络.经测试该网络的识别正确率可达92.8%.利用混淆矩阵及接受者操作性能(receiver op- erating characteristic,R0C)曲线分析，混淆矩阵的训练集、验证集、测试集、全集识别正确率分别达到97.9%、94.8%、92.8%和 96.3%,AUC(area under curve)值接近1，认为该网络分类性能优良，并且可在4s内完成整个识别过程. 关键词有限空间：低氧伤害：光电容积脉搏波：深度学习：状态识别 分类号X912 Recognition of human hypoxic state based on deep learning YU Lu,JIN Long-zhe,WANG Meng fei,XU Ming-wei School of Civil and Resource Engineering,University of Science and Technology Beijing,Beijing 100083,China Corresponding author,E-mail:lzjinz@ustb.edu.cn ABSTRACT Due to the development of industrialization,low-oxygen environment has become common in the confined spaces of con- struction industries,chemical industries,military,urban underground spaces,and poorly ventilated crowed areas and caused a large number of hypoxic injuries.The traditional method of preventing hypoxic injuries is to monitor the oxygen concentration in the environ- ment without considering the difference in oxygen tolerance limits when the human body is in different physiological states.Photople- thysmography (PPG)can comprehensively reflect physiological information,including heart rate,blood pressure,blood oxygen satura- tion,cardiovascular blood flow parameters,and respiratory rate.When the human body enters a hypoxic environment,the physiological parameters change rapidly,resulting in a change in the PPG signal.By measuring the PPG signal of the human body,the physiological state is considered to determine whether the human body reaches the oxygen tolerance limit.This study proposed a method for quickly identifying the hypoxic state of the human body using hypoxia experiment.According to the latest research on aviation medicine,moun- tain medicine and naval submarine medicine,15.5%oxygen volume fraction can guarantee the basic life safety of personnel.Through the training experimental data of a constructed deep neural network,the PPG signal of a human in normal oxygen volume fraction (16%-21%)and extremely low-oxygen volume fraction (15.5%-16%)was determined to obtain the pattern recognition network of human physiological state.After testing,the recognition accuracy of the network could reach 92.8%.Using the confusion matrix and receiver operating characteristic curve analysis,the accuracy rate of training set,verification set,test set,and ensemble recognition of the confusion matrix reached 97.9%,94.8%,92.8%,and 96.3%,respectively.The area under the curve value is close to 1,the network classification performance is excellent,and the entire identification process could be completed within 4 s. KEY WORDS confined space;hypoxic injury:photoplethysmography:deep learning:state recognition 收稿日期：20190306 基金项目：国家“十三五”重点科技支撑资助项目(2016YFC0801700)工程科学学报，第 41 卷，第 6 期: 817--823，2019 年 6 月 Chinese Journal of Engineering，Vol． 41，No． 6: 817--823，June 2019 DOI: 10． 13374 /j． issn2095--9389． 2019． 06． 014; http: / /journals． ustb． edu． cn 基于深度学习的人体低氧状态识别 于 露，金龙哲，王梦飞，徐明伟 北京科技大学土木与资源工程学院，北京 100083 通信作者，E-mail: lzjin@ ustb． edu． cn 摘 要 通过低氧实验提出一种快速识别人体低氧状态的方法． 通过搭建深层神经网络训练实验数据识别氧气体积分数 ( 16% ～ 21% ) 与人体可耐受极端低氧气体积分数( 15. 5% ～ 16% ) 条件下光电容积脉搏波( photoplethysmography，PPG) 信号， 获得人体生理状态的模式识别网络． 经测试该网络的识别正确率可达 92. 8% ． 利用混淆矩阵及接受者操作性能( receiver op￾erating characteristic，ＲOC) 曲线分析，混淆矩阵的训练集、验证集、测试集、全集识别正确率分别达到 97. 9% 、94. 8% 、92. 8% 和 96. 3% ，AUC ( area under curve) 值接近 1，认为该网络分类性能优良，并且可在 4 s 内完成整个识别过程． 关键词 有限空间; 低氧伤害; 光电容积脉搏波; 深度学习; 状态识别 分类号 X912 收稿日期: 2019--03--06 基金项目: 国家“十三五”重点科技支撑资助项目( 2016YFC0801700) Ｒecognition of human hypoxic state based on deep learning YU Lu，JIN Long-zhe ，WANG Meng-fei，XU Ming-wei School of Civil and Ｒesource Engineering，University of Science and Technology Beijing，Beijing 100083，China Corresponding author，E-mail: lzjinz@ ustb． edu． cn ABSTＲACT Due to the development of industrialization，low-oxygen environment has become common in the confined spaces of con￾struction industries，chemical industries，military，urban underground spaces，and poorly ventilated crowed areas and caused a large number of hypoxic injuries． The traditional method of preventing hypoxic injuries is to monitor the oxygen concentration in the environ￾ment without considering the difference in oxygen tolerance limits when the human body is in different physiological states． Photople￾thysmography ( PPG) can comprehensively reflect physiological information，including heart rate，blood pressure，blood oxygen satura￾tion，cardiovascular blood flow parameters，and respiratory rate． When the human body enters a hypoxic environment，the physiological parameters change rapidly，resulting in a change in the PPG signal． By measuring the PPG signal of the human body，the physiological state is considered to determine whether the human body reaches the oxygen tolerance limit． This study proposed a method for quickly identifying the hypoxic state of the human body using hypoxia experiment． According to the latest research on aviation medicine，moun￾tain medicine and naval submarine medicine，15. 5% oxygen volume fraction can guarantee the basic life safety of personnel． Through the training experimental data of a constructed deep neural network，the PPG signal of a human in normal oxygen volume fraction ( 16% --21% ) and extremely low-oxygen volume fraction ( 15. 5% --16% ) was determined to obtain the pattern recognition network of human physiological state． After testing，the recognition accuracy of the network could reach 92. 8% ． Using the confusion matrix and receiver operating characteristic curve analysis，the accuracy rate of training set，verification set，test set，and ensemble recognition of the confusion matrix reached 97. 9% ，94. 8% ，92. 8% ，and 96. 3% ，respectively． The area under the curve value is close to 1，the network classification performance is excellent，and the entire identification process could be completed within 4 s． KEY WOＲDS confined space; hypoxic injury; photoplethysmography; deep learning; state recognition