Motion Capture Compression

Speakers: Okan Arikan

Track: Programming

Format: Lecture

Experience Level: All

Description: Classical methods for motion compression include subsampling, spline fitting, Principal Component Analysis and Wavelet compression. Most successful compression algorithms rely on features of human perception or on knowing the nature of the data that they compress. Unfortunately, none of these baseline methods has been designed especially for motion data. This presentation will introduce a novel compression algorithm especially designed for high quality motion capture data. People are very sensitive to high frequency errors (jitters) and environmental contacts (such as foot-ground contact) in motion. The presentation will include demonstrations of why this is true and how we can design a compression method to pay special attention to such features of animation.

Idea Takeaway: The audience will learn why compression is important and how to tailor an effective compression technique by considering the features of human perception and the nature of the data. The presentation will give a very effective algorithm for compressing high quality motion capture data.

Intended Audience: The talk is intended for general audience as the growing amount of game assets per title is a pervasive problem for the entire game development pipeline. The presentation will include some math, but it is not beyond the general audience of GDC.