Affiliations: Faculty of Computer Science and Engineering, Xi'an University of Technology, Xi'an, Shaanxi, China
Correspondence:
[*]
Corresponding author: Minghua Zhao, Faculty of Computer Science and Engineering, Xi'an University of Technology, No. 5, South Jinhua Road, Xi'an 710048, Shaanxi, China. E-mail:zhaominghua@xaut.edu.cn
Abstract: A novel approach to the construction of motion graphs
based on a new similarity measurement method is proposed in this paper. The
main contributions of this paper are: i) to improve the accuracy of motion
similarity computing, a new similarity measurement method is presented which
combines frame distance based on dynamic time warping (DTW) with uniform
warping window and motion distance based on motion features; ii)to avoid
illogical jump of standard motion graph, a new automatic construction
strategy of motion graphs based on similarity of motion clips is proposed,
transition points of which are matched based on the path with the minimum
cost and motions are synthesized based on motion transition points.
Experimental results based on the CMU MoCap human motion database show that
the proposed method can confirm the logical correctness and naturalness in
the junction of two human locomotion sequences and its efficiency is two or
three times faster than the standard motion graph.
