Automatic Heterogeneous Motion Retargeting Using Semantic Inter-mediate Skeleton

Fully automatic motion retargeting between heterogeneous skeletons is not feasible by existing methods. To solve the problem, we proposed a hierarchical skeleton corresponding method based on semantic inter-mediate skeleton and multi-parameter optimization for biped and quadruped characters. The human and quadruped skeletons, which are widely used in the animation industry, are introduced as inter-mediate skeletons. Series of geometry and topology parameters are analyzed, and then skeleton nodes are classified into three layers according to their significant level and dipartite degree. According to the selected parameters for each layer, and semantic information from inter-mediate skeleton, skeleton nodes are corresponded layer by layer. At last, motion data is transmitted under constraints of gaits and end trajectories. The results from experiments among several typical character skeletons with various features proved that our method is feasible for fully automatic motion retargeting of mass heterogeneous characters with nonstandard body proportions and locally complex structures.