Efficient Algorithms For Human Motion Capture Data Recovery

With the rise,development and improvement of motion capture technology,the rapid and effective acquisition of high-precision motion data is favored by the public.As a new type of multimedia data,motion capture data is widely used in many fields such as computer animation,film and television special effects,medical rehabilitation.However,even the first-hand data collected using professional motion capture device systems(such as the Motion Analysis system)will inevitably contain problems such as noise,missing values and outliers.Therefore,how to efficiently and quickly restore existing motion capture data to true and accurate motion data has become a hot issue in current motion capture technology.In this thesis,a fast and effective optimization algorithm is proposed for human motion capture data recovery.The specific contents are as follows:In the first chapter,we first briefly introduce the research background,significance,current situation and application of motion capture.Then the problems of motion capture technology and the representation,characteristics and recovery model of motion data are discussed.In addition,the related mathematical theory to be used later is briefly introduced.Finally,the main work and content arrangement of this article are explained.In the second chapter of this thesis,we propose an effective proximity fixed piont(PFP)algorithm for human motion capture data recovery with low rank and noise sparsity.We give the derivation and construction process of the PFP algorithm,and obtain the convergence results of the algorithm.Second,through two sets of experiments,we illustrate the effectiveness of the PFP algorithm.Compared with the Inexact Augmented Lagrange Multipliers(IALM),which solves the RPCA problem,the PFP algorithm has some improvement in running time and iteration steps.In addition,the simulation results based on human motion data recovery problem show that the PFP algorithm is better than the Singular Value Thresholding(SVT)algorithm in experimental results and running time.In Chapter 3,we propose a proximity fixed point temporal stability(PFP-TS)algorithm for solving human motion capture data recovery problems with low rank,noise sparsity and temporaling stability.Among the existing methods for solving this problem,the method with good effect is based on the Augmented Lagrange Multipliers(ALM)idea of the temporal stable and noise robust martix completion model(TSMC)optimization algorithm.However,the limitation of the TSMC optimization algorithm for solving this motion data recovery problem is that the method cannot theoretically guarantee the convergence of the algorithm.In order to overcome this shortcoming,this paper proposes a PFP-TS algorithm with convergence guarantee,and gives the proof of strict convergence of the algorithm.Simulation results show that the PFP-TS algorithm is effective and more competitive than the TSMC optimization algorithm and the PFP algorithm.