基于IMU和Kinect的人体下肢动作捕捉算法的研究  

Research on Human Lower Limb Motion Capture Algorithm Based on IMU and Kinect

作　　者：郭欣 孙侨 张燕 GUO Xin;SUN Qiao;ZHANG Yan(Artificial Intelligence and Data Science,Hebei University of Technology College,Tianjin 300130,China;Qinhuangdao Research Institute of National Research Center for Rehabilitation Technical Aids,Qinhuangdao 066000,China)

机构地区：[1]河北工业大学人工智能与数据科学学院,天津300130 [2]国家康复辅具研究中心秦皇岛研究院,河北秦皇岛066000

出　　处：《控制工程》2023年第1期169-176,共8页Control Engineering of China

基　　金：国家自然科学基金资助项目(61703135,61773151)。

摘　　要：Kinect传感器在进行人体下肢关节定位时,因数据噪声多、深度数据不匹配、环境复杂等因素影响,导致出现局部位置信息丢失的情况,为此,提出了一种基于四元数的改进算法,通过使用低成本的惯性测量单元(IMU)来对Kinect传感器的异常数据进行判别与修正。使用改进后的四元数法对髋关节、膝关节和踝关节进行位置解算,以进一步提高解算结果的准确度,简化繁琐的实验流程。通过实验数据的对比发现:改进后的四元数算法可以让复杂的实验过程得以简化,关节角度的均方根误差(RMSE)相较于改进前的算法减小了约0.175;在IMU与Kinect传感器数据相结合后,能够很好地对其异常数据进行修正。When Kinect sensor is applied to locate the joint of lower-limb,the local location information usually missed as the exist of data noises,unmatched depth data and complex circumstance.The improved quaternion algorithm is mentioned to judge and rectify the abnormal data with low costing inertia measurement unit(IMU).Moreover,it can improve the calculation precision and simplify the experimental process,as it is used to calculate the hips location data,knees location date and ankles location data.The comparison experimental result is following:the root-mean-square error(RMSE)have reduced 0.175 compared with previous algorithm,the abnormal data can be rectified by coupling IMU and Kinect sensor data.

关 键 词：人体下肢关节定位 惯性测量单元 四元数 欧拉角 KINECT 

分 类 号：TP272[自动化与计算机技术—检测技术与自动化装置]