摘要
实时睡姿监测可用于睡姿改善和疾病预防。针对睡姿监测的实时性问题,设计了睡姿实时监测系统。基于自行研发的床单式柔性压力传感器进行数据采集,使用YOLOv5对预处理后得到的数据集进行训练和测试,通过LabVIEW和Python通信接口,搭建实时监测系统,调用训练好的神经网络模型,并将睡姿识别结果显示在LabVIEW前端页面中。实验结果表明:该实时监测系统测试集识别准确率达到了99.7%,整晚睡眠实验准确率为96.2%,从开始采样到识别结果显示到前端页面的时间不超过70 ms,综合高准确率和高处理速度两大优点,且在睡姿监测的基础上增加了对于体动的监测,为建立功能完善的智能无人护理系统奠定基础。
Real-time sleeping posture monitoring can be used for sleeping posture improvement and disease prevention.Aiming at the real-time problem of sleeping posture monitoring,a real-time sleeping posture monitoring system is designed.Based on the self-developed sheet flexible pressure sensor for data collection,the preprocessed dataset are trained and tested using YOLOv5.Through LabVIEW and Python communication interface,the real-time monitoring system is built.The trained neural network model can be called,and the sleeping posture recognition results can be displayed on the front-end page of LabVIEW.The experimental results show that the recognition accuracy rate of the test set of the real-time monitoring system reaches 99.7%,and the accuracy rate of the whole night sleep experiment is 96.2%.The time from the start of sampling to the display of the recognition results to the front-end page does not exceed 70 ms,and the two advantages of high accuracy rate and high processing speed are combined.On the basis of sleeping posture monitoring,the monitoring of body movement is added.This system lays the foundation for the establishment of an intelligent unmanned nursing system with complete functions.
作者
赵海文
赵悦焜
齐德瑄
李磊
王佳阔
郭士杰
ZHAO Haiwen;ZHAO Yuekun;QI Dexuan;LI Lei;WANG Jiakuo;GUO Shijie(School of Mechanical Engineering,Hebei University of Technology,Tianjin 300401,China;Hebei Key Laboratory of Robot Sensing and Human-Robot Interaction,Tianjin 300401,China)
出处
《传感器与微系统》
北大核心
2025年第3期120-124,共5页
Transducer and Microsystem Technologies
基金
国家重点研发计划资助项目(2021YFC0122700)
河北省高等学校科学技术研究项目(ZD2020313)。
关键词
睡姿
实时监测系统
无束缚
柔性压力传感器
体动
sleeping posture
real-time monitoring system
unconstrained
flexible pressure sensor
body move-ment
