A Sparse Semantic Parametric Model for Interactive Motion Synthesis

Since the parameters of existing parametric motion synthesis approaches are structure-inconsistent and have less intuitive meanings,we propose a sparse semantic parametric model.Our method extracts several sparse bases vectors based on Sparse Principal Component Analysis and Group Lasso,so that the original motion is represented as the linear combination of these bases and their weights in the combination corresponds to the motion parameters.Our experiments demonstrate the extracted bases have intuitive meanings,and each motion parameter controls motions in a specific aspect.Taking the walking motion as an example,our method can easily adjust stride,swinging ranges of arms,etc.by modifying the parameters.