There are observations indicating that the central nervous system (CNS) decomposes a movement into several successive sub-movements as an effective strategy to control the motor task. In this study, we propose an algorithm in which, Arm Reaching Movement (ARM) in 3D space is decomposed into several successive phases using zero joint angle jerk features of the arm kinematic data. The presented decomposition algorithm for 3D motions is, in fact, an improved and generalized version of the decomposition method proposed earlier by Emadi and Bahrami in 2012 for 2D movements. They assumed that the motion is coordinated by minimum jerk characteristics in joint angles space in each phase. However, at the first glance, it seems that in 3D ARM joint angles are not coordinated based on the minimum jerk features. Therefore, we defined a resultant variable in the joint space and showed that one can use its jerk properties together with those of the elbow joint in movement decomposition. We showed that phase borders determined with the proposed algorithm in 3D ARM, are defined with jerk characteristics of ARM’s performance variable. We observed the same results in the Sit-to-Stand (STS) movement, too. Thus, based on our results, we suggested that any 3D motion can be decomposed into several phases, such that in each phase a set of principal patterns (PPs) extracted by Principal Component Analysis (PCA) method are linearly recruited to regenerate angle trajectories of each joint. Our results also suggest that the CNS, as the primary policy, may simplify the control of the ARMs by reducing the dimension of the control space. This dimension reduction might be accomplished by decomposing the movement into successive phases in which the movement satisfies the minimum joint angle jerk constraint. Then, in each phase, a set of PPs are recruited in the joint space to regenerate angle trajectory of each joint. Then, the dimension of the control space will be the number of the recruitment coefficients.

中文翻译：

---

3D 人体手臂在连续阶段达到具有主要模式的运动规划。

有观察表明，中枢神经系统 (CNS) 将一个运动分解为几个连续的子运动，作为控制运动任务的有效策略。在这项研究中，我们提出了一种算法，其中使用手臂运动数据的零关节角急动特征将 3D 空间中的手臂伸手运动 (ARM) 分解为几个连续的阶段。所提出的 3D 运动分解算法实际上是 Emadi 和 Bahrami 早在 2012 年针对 2D 运动提出的分解方法的改进和推广版本。他们假设运动是由每个阶段的关节角度空间中的最小冲击特性协调的。然而，乍一看，似乎在 3D ARM 中，关节角度不是基于最小加加速度特征来协调的。所以，我们在关节空间中定义了一个结果变量，并表明可以在运动分解中将其急动特性与肘关节的特性一起使用。我们表明，在 3D ARM 中使用所提出的算法确定的相位边界是用 ARM 性能变量的急动特性定义的。我们在坐到站 (STS) 运动中也观察到了相同的结果。因此，基于我们的结果，我们建议任何 3D 运动都可以分解为几个阶段，这样在每个阶段中，通过主成分分析 (PCA) 方法提取的一组主模式 (PP) 被线性招募以重新生成角轨迹每个关节。我们的结果还表明，作为主要策略的 CNS 可以通过减少控制空间的维度来简化对 ARM 的控制。这种降维可以通过将运动分解成连续的阶段来实现，在这些阶段中，运动满足最小关节角冲击约束。然后，在每个阶段，在关节空间中招募一组 PP 以重新生成每个关节的角度轨迹。那么，控制空间的维度将是招聘系数的数量。