| In our daily life,learning action is a stage that we have to go through to do many things,such as learning dancing,martial arts,boxing,shooting,badminton swing,golf swing and so on.We need to learn slowly from the basic movements,compare the standard postures,analyze the shortcomings of our own movements and correct them,so that we can practice a set of standard movements in a cycle.While learning standard movements,we can learn them one-on-one,study in the mirror by ourselves,or watch videos or pictures to practice.However,these traditional learning methods not only take a long time,but also are not easy to find and correct minor erroneous actions,and it is difficult to improve learning efficiency,and there is no unified quantitative standard for the standard degree of action.Virtual systems based on virtual reality technology can assist people to train various skills,while greatly reducing the cost of learning these skills and operational risks,such as aircraft driving training system,medical surgery simulation system,virtual environment soldier training system and so on.In this paper,under the environment of virtual reality technology,using the second generation Kinect device launched by Microsoft,we can track the characteristics of user skeleton data in real time,and develop an assistant system for action learning.The main work of this paper is as follows:(1)The skeleton skin technology is used to construct the virtual 3D human model.When calculating the skin deformation during the movement of the model,the binary quaternion linear skin algorithm is used instead of the default linear skin algorithm to make the three-dimensional model more realistic.(2)Using GPU instead of CPU to update skin vertex and render three-dimensional character model,rendering and demonstrating three-dimensional character animation based on Unity platform,which reduces CPU load and makes more rational use of hardware resources;(3)The human motion detection algorithm is improved to optimize the user data collected by Kinect device,which significantly reduces the calculation error and delay,and improves the real-time performance of the algorithm and the accuracy of user action tracking.(4)A complete action learning assistant system is designed and implemented,which enables users to learn and train the actions they need at any time,and compares the real-time action data with the standard action data of users,feeds back the dislocation information to users in real time,and guides users to correct errors.The experimental results show that the system can better assist users in action learning. |
