摘要
目的:设计一种基于家用摄像头实时视频序列的老年人健康智能监测提醒算法,以监测老年人居家活动中由于身体不适引发的面部疼痛表情并提醒相关人员。方法:首先,利用多任务级联卷积网络对视频序列进行人脸检测并确定人脸区域,排除无关信息的干扰;然后,构建VGG16网络模型,再使用预训练的VGGface模型参数对模型进行初始化,并使用UNBC-McMaster肩部疼痛数据库进行模型的微调,从而检测和识别面部的疼痛表情;最后,采用开源的wechatpy项目调用PC端微信,向相关人员发送提醒消息。采用留一交叉验证来评估该算法的疼痛识别性能,并使用实际环境中所获得的视频来验证该算法的泛化性。结果:该算法的疼痛识别率为94.1%,AUC值为0.969;该算法能够在实验对象出现疼痛时进行报警提醒,并将该提醒反馈给相关人员。结论:该算法能够基于家用摄像头视频序列精确地检测到人脸和识别出疼痛表情,并即时发送提醒消息,为“互联网+照护服务”提供了一个可行的技术实现思路。
Objective To design an intelligent health monitoring reminder algorithm for the elderly based on real-time video sequences from the home camera to alert relevant personnel after detecting the facial pain expressions of the elderly caused by physical discomfort during home activities.Methods Firstly a multi-task cascaded convolutional network was used to detect faces in the video sequence and determine the face region after eliminating the interference of irrelevant information;a VGG16 network model was constructed and then initialized using the pre-trained VGGface model parameters and fine-tuned with the UNBC-McMaster Shoulder Pain Database,so as to detect and recognize pain expressions on the face;finally we-chatpy on the PC was called with the open-source wechatpy project to send reminder messages to relevant personnel.The algorithm had its pain recognition performance evaluated by leave-one-out cross validation,and its generalizability verified with the videos obtained in real-world environments.Results It's proved that the algorithm had the pain recognition rate being 94.1%and AUC value being 0.969,which was able to provide an alarm reminder when the experimental subject was in pain and sent this reminder to the relevant personnel.Conclusion The algorithm can accurately detect the face and recognize the pain expression based on the video from the home camera and send reminder messages immediately,which provides a feasible technical idea for"Internet+care services".
作者
冯跃飞
李湘
付元媛
余孟儒
刘洋
黄东
FENG Yue-fei;LI Xiang;FU Yuan-yuan;YU Meng-ru;LIU Yang;HUANG Dong(Department of Military Medical Information Technology of Military Biomedical Engineering School,Air Force Medical University,Xi'an 710032,China;The Forth Cadet Regiment of Basic Medical Science Academy,Air Force Medical University,Xi'an 710032,China)
出处
《医疗卫生装备》
CAS
2023年第8期29-33,共5页
Chinese Medical Equipment Journal
基金
国家自然科学基金面上项目(81871424)
陕西省自然科学基础研究计划一般项目(2023-JC-QN-0704)。
作者简介
冯跃飞(1997-),男,硕士研究生,助教,研究方向为医学影像人工智能,E-mail:deer_von@163.com;通信作者:黄东,E-mail:hdong137@163.com。
