Joint Angle Data Representation for Data Driven Human Motion Synthesis

Abstract For ergonomic assessment of manual assembly tasks, digital simulation has received increasing attention due to its efficiency compared to physical prototypes. One of the crucial parts of digital simulation is an accurate animation of the digital human model (DHM). Current digital simulation tools such as Delmia V5 require interactive manual editing to produce animations, which is time consuming and can look unnatural. On the other hand, data-driven motion synthesis that is based on motion capture data can produce natural motions with little user involvement. The practical difficulty lies in processing motion data into a parameterized motion model. A common approach is decomposing motions and categorizing them into finite short motion primitives. For each motion primitive, motion data is represented as a numerical vector, on which functional principal component analysis (FPCA) is applied to reduce dimensionality. In this work, different ways of representing joint angles from motion capture data are explored: Euler angle, quaternion and exponential map. The data representations are evaluated for their reconstruction error with FPCA. In the tests, quaternion representation shows best performance for motion data representation, which contradicts a preference in literature for exponential map representation. Therefore, quaternion representation is considered appealing for statistically modelling motion data.