基于MA和5DT手套的上肢在线动作捕捉系统

发布时间：2018-01-04 19:38

本文关键词：基于MA和5DT手套的上肢在线动作捕捉系统　出处：《浙江大学》2017年硕士论文　论文类型：学位论文

【摘要】：人体上肢动作捕捉技术在人机交互、疾病康复、脑机接口等诸多领域都有广泛应用。随着相关技术的进步,对于运动数据的实时性、稳定性和精确性有着越来越高的要求。本文利用MA(Motion Analysis)光学动作捕捉系统在线捕捉手臂的运动信息,并且对丢失标记点进行在线预测;同时利用5DT数据手套来获取手部运动信息。最终将手臂运动信息与手部运动信息整合起来,驱动虚拟上肢模型,实时重建人体上肢运动。本文主要研究内容如下:基于光学动作捕捉系统计算手臂关节运动信息。将手臂运动简化为一个7自由度的模型,利用相应的标记点模块坐标来计算关节空间位置。针对不同的标记点丢失情况,分别设定相应的预测方法,实时预测丢失标记点的空间坐标;利用贴放在皮肤表面的标记点与相应关节的固定位置关系,建立损失函数,使用最小二乘法求解标记点的旋转中心,该旋转中心即为相应关节的空间位置。基于数据手套计算各个手指关节的空间位置。本文采用了 26自由度的手部运动模型,将手指各个自由度的旋转角度与数据手套传感器数值对应起来。利用传感器数据计算手指关节自由度的旋转角度。最后通过前向运动学方法计算手指关节位置。统一手臂与手部坐标并实时绘制上肢运动。利用手部和前臂的标记点计算得到手腕坐标。利用手部的三个标记点计算得到手的方向。根据手腕坐标和手的方向将手的运动映射到手臂上,得到完整的人体上肢骨骼关节运动信息。根据这些信息,利用OpenGL实时绘制一个可以与真实人体上肢同步运动的虚拟上肢。通过计算,MA丢失标记点的预测误差基本集中在0.5mm以内,关节位置与标记点距离变化在1mm左右,在10分钟的长时间运行过程中系统可以一直稳定在线运行,证明该系统可以实时在线的得到上肢运动数据。
[Abstract]:Human upper limb motion capture technology has been widely used in human-computer interaction, disease rehabilitation, brain-computer interface and many other fields. With the progress of related technology, real-time motion data. There are more and more requirements for stability and accuracy. This paper uses the MA(Motion Analysis) optical motion capture system to capture the motion information of the arm online. And the lost mark points are predicted on line. At the same time, 5DT data gloves are used to obtain the hand movement information. Finally, the arm motion information and the hand motion information are integrated to drive the virtual upper limb model. The main contents of this paper are as follows: based on the optical motion capture system, the motion information of the arm joint is calculated, and the arm motion is simplified into a 7 degree of freedom model. The joint space position is calculated by using the corresponding marker point module coordinates, and the corresponding prediction methods are set up to predict the spatial coordinates of the missing mark points in real time. The loss function was established by using the fixed position relationship between the labeled points attached to the skin surface and the corresponding joints, and the rotation center of the mark points was solved by the least square method. The center of rotation is the space position of the corresponding joint. The spatial position of each finger joint is calculated based on the data glove. In this paper, the hand motion model with 26 degrees of freedom is adopted. The rotation angle of each degree of freedom of the finger is matched with the data glove sensor. The rotation angle of the degree of freedom of the finger joint is calculated by the sensor data. Finally, the position of the joint of the finger is calculated by the method of forward kinematics. The coordinate of arm and hand is unified and the movement of upper limb is plotted in real time. The coordinate of wrist is calculated by using the marking point of hand and forearm. The direction of hand is calculated by using the three marking points of hand. According to the direction of wrist and hand, The movement of the hand maps to the arm. According to these information, a virtual upper limb which can move synchronously with real human upper limb can be drawn by OpenGL in real time. The prediction error of MA missing mark point is mainly within 0.5 mm, and the distance between joint position and mark point is about 1 mm. It is proved that the system can get the upper limb motion data online in real time.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP212.9;R49

【参考文献】