An improved approach to the efficient construction of and search operations in motion graphs

Motion graphs have been widely used as an effective technique in the synthesis of human motions.To synthesize well-formed motions,motion graphs require smooth transitions and good connectivity.However,the time and costs grow quadratically as the graph size increases.In this paper,we present a method to set up a hierarchical structure,called a structure table.The coarse level of this graph models different motion categories and their transitions,and the fine level mainly captures different motion styles.An interpolation space method is used to improve the graph connectivity at the fine level.This structure guides the establishment of a faster and more realistic graphical object.After the building process,this structure also helps with searching the motion graph,by performing a traverse search and placing the findings into a direct look-up table or local search depository.Moreover,when there is no direct link between the different motions,a navigation map is introduced to conduct an appropriate search path via a series of transition motion clips.The experimental results indicate that our approach could lower the computation time in construction and searching,while simultaneously providing a richer variety of motion styles.