Data-driven character animation synthesis

Abstract

In this article, we describe about data-driven character motion synthesis for use mainly on a full-body skeleton structure. Due to the simplicity of capturing motion nowadays, the main issue for animating characters is how to reduce the cost of applying such motion to the characters and how to recycle the motion for interactive motion synthesis. An additional topic of interest is how to convert the style of the movements while preserving the context of the motion. In this article, we primarily cover machine learning techniques that can be useful for such purposes.