基于计算机视觉的居家康复训练评估算法  

Home Rehabilitation Training Evaluation Algorithm Based on Computer Vision

作　　者：郑奇 郭立泉 陈静[1,2] 杨朝[3] 王晓军 熊大曦 ZHENG Qi;GUO Li-quan;CHEN Jing;YANG Zhao;WANG Xiao-jun;XIONG Da-xi(Division of Life Sciences and Medicine,School of Biomedical Engineering(Suzhou),University of Science and Technology of China,Hefei 230026,China;Suzhou Institute of Biomedical Engineering and Technology,Chinese Academy of Sciences,Suzhou 215163,China;Department of Respiratory Medicine,Suzhou Sci-Tech City Hospital,Nanjing Medical University,Suzhou 215153,China;Department of Neurosurgery,Suzhou Xiangcheng People′s Hospital,Suzhou 215131,China)

机构地区：[1]中国科学技术大学生物医学工程学院(苏州)生命科学与医学部,合肥230026 [2]中国科学院苏州生物医学工程技术研究所,江苏苏州215163 [3]南京医科大学附属苏州科技城医院呼吸内科,江苏苏州215153 [4]苏州市相城人民医院神经外科,江苏苏州215131

出　　处：《小型微型计算机系统》2022年第11期2336-2341,共6页Journal of Chinese Computer Systems

基　　金：国家重点研发计划子课题项目(2018YFC1313602)资助。

摘　　要：针对老年人肌肉力量、平衡和活动能力等运动功能受损问题,如何有效开展居家的肢体康复训练和评估,是目前临床的痛点和研究热点.对此提出一种基于计算机视觉和支持向量机(Support Vector Machine,SVM)的居家康复训练评估算法.通过MobileNetV3结构改进了姿态估计算法,在普通RGB摄像头和家用级电脑上实现了约30FPS的实时人体姿态估计,降低了使用成本;用创新性的归一化方法和动态时间规整算法(Dynamic Time Warping,DTW),解决了人体体型差异和老年人动作滞后的问题;通过支持向量机算法建立了与临床评估结果相映射的模型.48名受试者参与了居家康复训练评估实验,结果表明,算法准确度为93.2%,与康复专家评分呈强正线性关系(r=0.967).It is a clinical pain point and research hotspot that how to carry out effective body rehabilitation training and evaluation at home for the elderly with impaired motor function such as muscle strength,balance and mobility.Therefore,a home rehabilitation training evaluation algorithm based on computer vision and Support Vector Machine(SVM)was proposed.The MobileNetv3 structure is used to improve the pose estimation algorithm,and the real-time human pose estimation of about 30 FPS is realized on common RGB camera and home computer,which reduces the cost of use.An innovative normalization method and Dynamic Time Warping(DTW)algorithm were used to solve the problems of the difference of human body size and the movement lag of the elderly.Support vector machine(SVM)algorithm was used to build a mapping model with clinical evaluation results.A total of 48 subjects participated in the home rehabilitation training evaluation experiment,and the results showed that the accuracy of the algorithm was 93.2%,which showed a strong positive relationship with the rehabilitation expert′s score(r=0.967).

关 键 词：计算机视觉 康复评估 支持向量机 姿态估计 动态时间规整 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]