Synthesising character animation using statistical models, perception, and databases of motion

Abstract

Computer generated virtual humaus are an important part of many 3D graphical applications, with a range of purposes from entertainment to education. Character animators generated characters with varying degrees in both appearance and motion from realistic to highly stylised. Plausible, appealing, and engaging characters contribute to high levels of immersion and enjoyment among human users and are important to ensure audience engagement. However, creating high-quality motion is a very laborious task, and there is a lack of automatic tools for motion creation. The goal of this thesis is to synthesise character animation to speed-up this process, using motion capture, perceptual metrics, and statistical modelling from both artist-created and real human movement.