摘要
针对起立意图难以预测及黑箱模型缺乏生物力学解释的问题,提出一种考虑动态平衡的起立意图预测方法.根据人体起立运动遵循的动态平衡规律,借鉴一种考虑起立运动速度的质心轨迹非线性处理手段,即外推质心,将质心轨迹直角坐标平面变换到动态平衡基准平面,降低特征数据对人体运动速度的依赖程度,提高起立意图判断准确性.分别使用支持向量机和神经网络对所提方法进行验证,并与传统数据处理手段进行对比.结果表明:模型预测的准确率均有提升,支持向量机的预测准确率由73.3%~74.1%提升到79.9%,起立时刻与预测时刻的时间偏差从96~101 ms降低到70 ms,神经网络的预测准确率从81.2%提升到86.2%,时间偏差从87 ms降低到59 ms.
Aiming at the problem that the sit-to-stand intention is difficult to predict and the black box model is lack of biomechanical explanation,a sit-to-stand intention prediction method considering dynamic equilibrium was proposed.According to the law of dynamic equilibrium followed by the human body's sit-to-stand movement,a nonlinear processing method of the centroid trajectory considering the speed of the sit-to-stand movement was proposed,i.e.,extrapolated center of mass(CoM),transforming the Cartesian coordinate plane of the CoM to the dynamic equilibrium reference plane,reducing the dependence of feature data on the human body's motion speed,and improving the accuracy of the judgment of the sit-to-stand intention.Support vector machines(SVM)and neural networks(NN)were used to validate the proposed method,and the proposed method was compared with traditional data processing methods.Results show that the accuracy of model prediction is improved.The prediction accuracy of SVM is improved from 73.3%~74.1%to 79.9%,and the time deviation between stand up time and assist time is reduced from 96~101 ms to 70 ms.The prediction accuracy of the NN increases from 81.2%to 86.2%,and the time deviation decreases from 87 ms to 59 ms.
作者
庞牧野
魏东盛
罗晶
魏巍
PANG Muye;WEI Dongsheng;LUO Jing;WEI Wei(School of Automation,Wuhan University of Technology,Wuhan 430070,China;School of Electrical and Information Engineering,Suzhou University of Science and Technology,Suzhou 215009,Jiangsu China)
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2024年第5期50-55,共6页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(62203341).
关键词
起立运动
动态平衡
外推质心
意图预测
支持向量机
神经网络
sit-to-stand movement
dynamic equilibrium
extrapolation center of mass
intention prediction
support vector machines
neural network
