Human Motion Simulation Based On Bayesian Program Learning

The simulation of human motion is an important research direction in the field of virtual reality.With the continuous development and maturity of human motion simulation technology,it has been widely used in many fields,such as education,film and television,military and so on.Due to the use of image processing technology to obtain human motion data from the video are human feature extraction and tracking difficult,easy to be interfered by the video capture device,environment and other factors,in order to better the human motion simulation,motion simulation methods mostly used motion capture data access.However,due to the limitations of the human body,many actions can not achieve the desired results,at this time the need to capture the data editing.However,due to the large amount of motion simulation data,it is very hard to edit the motion data,which is time-consuming and inefficient.In order to get rid of the artificial drag and drop editing mode of action,improve the naturalness of human motion simulation,motion simulation to achieve the desired effect,improve work efficiency,this paper presents a simulation method of Bayesian learning program based on the movement of HMBPL(Human motion simulation based on Bayesian program learning)method.The HMBPL method is divided into three parts: HMBPL motion learning,HMBPL motion synthesis and HMBPL motion simulation.(1)HMBPL motor learning: each frame motion data analytic motion capture BVH format,which can be decomposed into simpler components,are combined according to the relationship between the part and the root node,generating the concept,the concept of learning from the movement,and then add the concept of motion variables for generating motion data new.(2)HMBPL motion synthesis: due to the motion learning is only generated motion data,in order to connect the motion data frame to form motion data,motion synthesis.HMBPL motion synthesis according to the action data generation algorithmknowledge,know in advance and to generate the motion of probability calculation,a frame data to select the optimum,generate motion sequences,obtain new motion data.(3)HMBPL motion simulation: after getting the new BVH format motion data,the motion data is mapped to the virtual human in the MATLAB environment,and the virtual human is simulated.HMBPL method requires only a small number of samples to learn the characteristics of the sample data,the realization of the virtual man,learn an action,can naturally learn other actions.The experiments show that the HMBPL algorithm can improve the simulation results,reduce the workload,reduce the time cost and improve the work efficiency.