Building Linkage Models of Human Arms

在这个项目中，我们开发了一种系统的方法，通过人体上臂的运动测量来建立人体上臂的个人链接模型。采用A3D位置传感器和柔性关节角度传感器测量上臂运动和手腕运动。首先，我们在上臂的连杆模型中引入了一个二连杆一自由度机构来描述人体肩部的运动。需要对链接模型中涉及的个人模型参数进行估计，使模型能够应用于个人。因此，我们建立了一种通过预先进行的简单运动测量来识别模型参数的方法。通过测量可以识别出所引入的二连杆一自由度机构中肩关节中心位置、两连杆长度及两个关节角度关系、肘关节中心位置、前臂长度等模型参数。传统的1-连杆1-DOF模型不能同时描述人手的位置和方向，而该模型能够以简单的方式描述人手的位置和方向。此外，个体之间的差异可以很容易地考虑到链路模型中，并且通过模型参数的识别可以消除传感器位置的差异。其次，我们开发了柔性关节角度传感器来测量人类手腕的运动。由于腕部运动范围较窄，我们发现3D传感器不适合测量腕部运动。因此，我们有使用应变片的角度传感器原型，并发现只要手腕角度在50度以下，它们就有足够的精度。如上所述，我们发现所提出的方法可以有效地建立人体上臂的个人模型。建立手腕模型是一个悬而未决的问题。

In this project, we have developed a systematic method to build a personal link model of human upper arms through their motion measurements. A3D position sensor and flexible joint angle sensors are utilized in the measurements of upper arm motion and wrist motion.First, we have introduced a 2-link 1-DOF mechanism into the link model of a upper arm to describe human shoulder motion. It is required to estimate personal model parameters involved in the link model so that the model can be applied into individual persons. Thus, we have established a method to identify model parameters through the measurements of simple motions performed in advance. Model parameters including the location of shoulder joint center, length of two links and the relationship between two joint angles in the introduced 2-link 1-DOF mechanism, the location of elbow joint center, and length of forearm can be identified through the measurements. Traditional 1-link 1-DOF models cannot describe the position and the orientation of human hand simultaneously while the proposed mode is capable of describing them in a simple manner. In addition, difference among individual persons can be taken into the link model easily as well as the difference of sensor locations can be eliminated through the identification of model parameters. Second, we have developed flexible joint angle sensors to measure human wrist motion. Since the range of wrist motion is narrow, we have found that the 3D sensor is inappropriate to the measurements of wrist motion. Thus, we have prototyped angle sensors using strain gauges and have found that they have sufficient accuracy as long as the wrist angle is under 50 degrees. As mentioned above, we have found that the proposed method is effective to build a personal model of human upper arms. Building wrist models is an open issue.