Open AccessAutomatic Segmentation for Virtual Human Motion
Author(s) -
Shi Ru Qu,
Tinxin Xu,
Liang Ma,
Jianxun Liu
Publication year - 2020
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/790/1/012161
Subject(s) - artificial intelligence , computer vision , motion (physics) , segmentation , computer science , motion capture , character (mathematics) , motion estimation , motion field , mathematics , geometry
A technique of automatic segmentation for motion data has been proposed, which describes high-dimensional motion with low-dimensional motion character and automatically segments the motion capture data by detecting changes of motion character. Gaussian Process Latent Variable Models has been used to reduce dimension for motion data. In this model, the motion data has been mapped from observation space to latent space. The character function of motion has been constructed in the latent space, and it has excellence including sensitive to all joints, simple construction, and so on. The motion data segmentation points can be detected to complete motion segmentation after analyzing character function by geometry character. The technique in this paper has well adaptation and high correct rate as shown in experiments.