Scaled Indexing of General Shapes for complicated 3D motion recognition

Motion recognition based on trajectory is important for motion analysis. Complicated motion recognition is still a challenge in various applications of robot and automation. In this paper, we propose a novel framework with a new model, Scaled Indexing of General Shapes (S-IGS), for complicated motion recognition. The Scaled IGS is a quantified hierarchical model, representing 3D motion trajectories with mixed-parameterized primitives. The mixed parameters include not only general shape classes, but also their reference values. The reference value is a particular parameter of primitive which is effective to distinguish the primitives of the same general shape class. Based on this model, we explore the motion recognition with both primitive alignment and inner-parameter matching. The conducted experimental results verified the accuracy and efficiency of this approach.