Murai Akihiko, Yamane Katsu, Nakamura Yoshihiko
Department of Mechano-Informatics, the University of Tokyo, 7-3-1, Hongo, Bunkyo-ku, Tokyo, 113-8656, Japan.
Annu Int Conf IEEE Eng Med Biol Soc. 2009;2009:6542-5. doi: 10.1109/IEMBS.2009.5334508.
In this paper, we propose two methods to quantitatively analyze the motor skill in sports. The first method is the dimensionality reduction using the principal component analysis (PCA). The motion data, e.g. the joint angles (143-dimensional vector) or the muscle tensions (989-dimensional vector), are projected to a lower dimensional space that well represents the characteristics of original data. The similarities and differences become clear by observing the data in the low-dimensional space. The second method utilizes the joint stiffness obtained from joint kinematics and a biological muscle model. Though muscle tension data contain richer information than joint angle data, the dimension is so high that simply applying PCA does not give useful insights. Here we calculate the joint stiffness using the muscle tension data and a biological muscle model. This information represents the muscle usage skill which can not be observed only from motion data, and reflects the redundancy of the muscle tensions. We demonstrate the two methods by analyzing skilled performers' motions.
在本文中，我们提出了两种定量分析运动中运动技能的方法。第一种方法是使用主成分分析（PCA）进行降维。运动数据，例如关节角度（143维向量）或肌肉张力（989维向量），被投影到一个能很好地表示原始数据特征的低维空间。通过观察低维空间中的数据，异同点变得清晰。第二种方法利用从关节运动学和生物肌肉模型获得的关节刚度。虽然肌肉张力数据比关节角度数据包含更丰富的信息，但维度如此之高，以至于简单地应用PCA并不能给出有用的见解。在这里，我们使用肌肉张力数据和生物肌肉模型来计算关节刚度。此信息表示仅从运动数据中无法观察到的肌肉使用技能，并反映了肌肉张力的冗余性。我们通过分析熟练表演者的动作来演示这两种方法。
