非接触式生理心理检测在长期模拟失重效应分析中的应用研究  

Research on the application of non-contact physiological and psychological detection in the analysis of long-term simulated weightlessness effects

作　　者：丁帅[1,2] 许梓[3,4] 戎千 刘书娟[3,4] 刘子豪 武元[3] 俞尧 李志利[3,4] 宋程 曲丽娜[3,4] 王浩[1,2] 李莹辉[3,4] Ding Shuai;Xu Zi;Rong Qian;Liu Shujuan;Liu Zihao;Wu Yuan;Yu Yao;Li Zhili;Song Cheng;Qu Lina;Wang Hao;Li Yinghui(School of Management,Hefei University of Technology,Hefei 230009,China;Key Laboratory of Process Optimization and Intelligent Decision-making of Ministry of Education,Hefei 230009,China;China Astronaut Research and Training Center,Beijing 100094,China;State Key Laboratory of Space Medicine,China Astronaut Research and Training Center,Beijing 100094,China)

机构地区：[1]合肥工业大学管理学院,合肥230009 [2]过程优化与智能决策教育部重点实验室,合肥230009 [3]中国航天员科研训练中心,北京100094 [4]中国航天员科研训练中心航天医学全国重点实验室,北京100094

出　　处：《航天医学与医学工程》2024年第2期78-83,98,共7页Space Medicine & Medical Engineering

基　　金：国家自然科学基金项目(72293581);中国载人航天预先研究项目(地星二号—NO.0040)。

摘　　要：目的在模拟失重生理效应场景下,构建基于面部视频的非接触式生理心理检测模型,探索长期模拟失重生理效应分析的非接触式心率和负性心境状态检测新方法。方法搭建可见光和热红外视频融合分析的非接触式生理心理数据采集系统,采集“地星二号”90 d头低位卧床实验中受试者生理心理相关数据,构建基于图卷积网络面部多区域特征融合的非接触式心率检测模型和考虑数据可靠性的非接触式负性心境状态检测模型,并以指夹式心率和POMS-SF量表为标签,验证模型的有效性。结果研究结果显示,非接触式心率检测模型的Bland-Altman图差值平均数为-1.26 bpm,且96.3%的误差值检测数据处于95%一致性区间内,模型检测的心率与指夹式心率有较高的一致性;非接触式负性心境状态检测模型对于紧张、压抑、愤怒、疲劳的检测准确率均超过0.85,且心率、AU06、眼部视线、头部姿态对心境状态检测的影响显著。结论非接触式生理心理检测方法不仅能够用于长期模拟失重的生理效应分析,还可为未来长期空间飞行的航天员在轨健康保障提供新途径。Objective Explore a non-contact physiological and psychological detection model based on facial video in simulations of weightlessness effects,research new methods for non-contact heart rate and negative mood state detection in long-term simulations of weightlessness effect analysis.Methods Construct a non-contact physiological and psychological data collection system for fusion analysis of visible light and thermal infrared videos.Collect physiological and psychological data of volunteers in the“Earth Star-Ⅱ”90-day head-down bed rest experiment.A noncontact heart rate detection model based on GCN facial multi-region feature fusion and a non-contact negative mood state detection model considering data reliability were constructed,and the effectiveness of the models were validated with finger clip heart rate and POMS-SF scale as labels.Results The experimental results show that the average difference in the Bland-Altman plot of the non-contact heart rate detection model is−1.26 bpm,and 96.3%of value error detection data falls within the 95%confidence interval,indicating a high consistency between the model detected heart rate and the finger clip heart rate.The non-contact negative mood state detection model achieves an accuracy of>0.85 for detecting tension,depression,anger,and fatigue.Features such as heart rate,AU06,eye gaze,and head pose were observed to be important to mood state detection.Conclusion Non-contact physiological and psychological detection methods not only can be utilized for long-term physiological analysis in simulations of weightlessness effects,but also provide a novel technical approach for on-orbit astronauts health assurance during long-term space flight in the future.

关 键 词：模拟失重效应 非接触检测 心率检测 心境状态分析 

分 类 号：R853[医药卫生—航空、航天与航海医学] R852.22[医药卫生—临床医学]