作者: Yang Yang , Lanling Zeng , Howard Leung
关键词:
摘要: In this paper, we propose a novel method to extract keyframes from motion capture data for people better visualize and understand the content of motion. It first applies Butterworth filter remove noise in data, then carries out principal component analysis (PCA) reduce dimension. By detecting zero-crossing points velocity components, initial set can be derived. The are further pruned group similar postures together redundancies. An experiment based on CMU database is carried out, result suggests that extracted with our could support visualization understanding data.