Character Motion Synthesis Based on Deep Learning: A Survey

Abstract

Character motion synthesis can be more cost-effective, flexible, and time-efficient compared to motion capture or traditional animation. As character motion synthesis regards multiple major industries, along with the development in deep learning techniques, character motion synthesis based on deep learning conspicuously receives substantial attention, resulting in numerous related studies that need to be analyzed and summarized. This paper presents an overview of character motion synthesis based on deep learning. Firstly, it epitomizes methods incorporating different types of neural networks, then encapsulates methods that did not utilize neural networks but simply deep learning, such as deep reinforcement learning, and lastly summarizes and evaluates the advantages and limitations of different deep learning methods on character motion synthesis.