Analysis and Synthesis of Human Motion Function Data Based on Decision Tree Classification

Abstract

At present, motion capture is widely used in computer animation, games, movies and robots, but it is still a difficult problem to synthesize stylized human motion. To solve this problem, a motion synthesis method based on decision tree classification and block principal component analysis is proposed. Block principal component analysis is carried out on the motion data grouped according to the characteristics of human skeleton structure, and low-dimensional subspace parameters with specific semantics are obtained. Triangular constraint is used to block the connection between moving frames which are far apart, thus ensuring the time sequence continuity of segmentation results; In the retrieval process, the similarity of key points is calculated according to different influence degrees in turn; Finally, an efficient motion retrieval simulation system is realized.