Kinect在肢残者运动姿态识别中的应用及有效性研究

发布时间：2018-02-16 02:52

本文关键词： Kinect V VICON 姿态静态动态　出处：《中国康复医学杂志》2017年02期 　论文类型：期刊论文

【摘要】：目的:比较Kinect是否能替代传统运动捕捉设备用于肢残者运动姿态研究。方法:将30例受试者分成两组分别在Kinect与VICON运动捕捉系统下进行实验,并对各组数据进行预处理,利用皮尔逊系数相关法,验证两组受试者各个关节角之间的相关性,对各关节角相关性强度进行评定。结果:关联性较高的数据为矢状面数据,髋、膝、背曲、腰部屈伸等关节V-K(VICON与Kinect)数据相关性最高(r0.7)。假肢者的姿态识别中V-K相关性系数低于健体者(r2r1)。假肢者的髋、膝、背曲等关节相关性系数差异不大,健体者与假肢者的肘关节数据相关性系数差异性较大。结论:Kinect替代VICON需从采集的关节角的映射面、任务及被试者三方面考虑,在人体矢状面、任务过程中有较少自遮挡关节点、被试者能够自由完成规定任务动作情况下,Kinect可以作为有效替代工具研究人体姿态识别。
[Abstract]:Objective: to compare whether Kinect can replace the traditional motion capture equipment for the study of motion posture of limb disabled patients. Methods: 30 subjects were divided into two groups to carry out the experiment under the Kinect and VICON motion capture system, and the data of each group were preprocessed. Pearson coefficient correlation method was used to verify the correlation between each joint angle of the two groups and evaluate the correlation strength of each joint angle. Results: the higher correlation data were sagittal plane data, hip, knee, dorsal curvature, The correlation coefficient of V-K correlation coefficient in posture recognition of prosthetic limb was lower than that in healthy person. The correlation coefficient of hip, knee and dorsal flexion of prosthesis was not different from that of the prosthetic joint, such as hip, knee, dorsal curvature, and so on, the correlation coefficient of V-K and Kinect was the highest in the lumbar flexion and extension joints, and the correlation coefficient of V-K was lower than that of the healthy person. The correlation coefficient of elbow joint data between the healthy and the prosthetic is quite different. Conclusion the mapping surface of the joint angle, the task and the subjects need to be considered in terms of the mapping surface of the joint angle, the task and the subjects. There are fewer self-occlusion nodes in the sagittal plane of the human body and in the course of the task. Kinect can be used as an effective alternative to human posture recognition.
【作者单位】：华中科技大学机械科学与工程学院;
【基金】：国家自然科学基金项目(71301057) 上海航天科技创新基金项目(SAST201409)
【分类号】：C913.69;R496

【相似文献】