Analysis and synthesis of human dance motions

Abstract

This paper presents the method for synthesizing stylistic human motions through visual observations. The human motion data is acquired from a motion capture system. The whole motion sequence is divided into motion elements and clustered into groups according to the correlation of end-effectors' trajectories. We call these segments 'motion primitives'. Concatenating these motion primitives, we can generate new dance motions. We also think that a motion primitive consists of a basic motion and a motion style. The basic motion is common to all dancers, and the style represents their characteristics. We extracted these two components through further analysis steps. The experiment results show the validity of our approach.