Adaptive Compression of Human Animation Data

Abstract

In this paper we introduce an adaptive motion compression technique using the discrete wavelet transform. Based on the analysis of human animation data, and its frequency contents, the wavelet analysis is utilised to achieve motion compression. It has been found that even with high compression ratios (up to 86%), the compressed animation is visually very close to the original animation. This property is potentially useful for many areas of animation such as motion interpolation/blending and networked virtual environments. One of our motivations is that motion blending/interpolation between motion capture samples is one of the successful techniques for synthesis of novel realistic human animation. One of the major concerns of that method is the increase of the database size with the large number of samples required to synthesise a wide range of movements. Previous research has addressed this issue by trying to reduce the number of samples required for interpolation. The approach that we introduce in this paper is to reduce the individual sample’s size using compression. Integration of both these approaches promises to allow a realistic animation with reduced database size. In networked virtual environments and on-line games, compressing the animation data can reduce the transmission load and help in achieving real-time performance and realism with reduced cost.