Improving Motion Synthesis by Semantic Control

A suitable motion similarity measurement plays an important role in motion analysis for motion graph based human motion synthesis. Most existing methods of motion synthesis are based only on numerical similarity measure of motions, which cannot identify the semantically similar motions in the presence of significant spatial and temporal variations. This paper focuses on introducing semantic control into motion graph based motion synthesis. Relational features are implemented for describing and specifying motions at a high semantic level to cope with spatial variations within a class of semantically similar motions. Motion templates are then automatically derived based on a self-learning procedure from the training motions for capturing the spatio-temporal characteristics of an entire given class of semantically related motions. Automatic semantic annotation is performed on the unknown motion data document by identifying the presence of certain motion class from matching their respective motion class templates. Finally, the semantic control is introduced into motion graph based human motion synthesis, which provides user the higher level of intuitive semantically controls on motion synthesis. Experiments of motion synthesis demonstrate the effectiveness of the approach which achieves high quality output of human motion synthesis from motion capture database.