Research Of Personalized 3D Human Motion Reconstruction And Detail Analysis Based On Kinect

With the rapid development of social economy and accelerating pace of life, most people tend to focus on the pursuit of material life more than on sports, spiritual life. This inevitably leads to a variety of physical and mental health problems. Numerous studies found that physical exercise is one of the effective ways to alleviate or even solve this problem. Therefore, in recent years, our research group made great efforts to develop a public sports teaching system to help common people doing exercises more professionally.There are three challenge problems in developing a sports teaching system. Firstly, the traditional motion capture systems are very expensive, imprecise and interferential. Secondly, the capture reconstruction system model is not life-like and real-time. Thirdly, there are no detailed motion comparison methods reported in the previous literature and also no standards for the qualitative and quantitative evaluation. The main contribution of this paper are presented as follows:First, balancing the precision and the price, we chose the Kinect sensor as the motion capture equipment. The Kinect has a moderate accuracy and it is inexpensive, non-invasive equipment which makes it has less impact on the user movement. The original depth information of human motions captured by the Kinect sensor are converted into the formation of skeleton data, which makes it convenient for the following motion reconstruction and comparison processes.Next, the skeleton motion data captured by the Kinect SDK are smoothed to reduce the motion jitters. Then, in order to solve the model distortion problem, realistic and personalized human models are created using MakeHuman software. Thirdly, we find a set of matching parameters to ease the problem of motion distortion and make the motion data and the model match better. Finally, we chose graphics rendering engine named OGRE to integrate the model and the motion data which reproduce a perfect reconstruction effect.Finally, an effective motion comparison method based on segmented multi-joint line graphs combined with the SIFT feature matching method is proposed. To begin with, the multi-joint 3D motion data are converted into a 2D line graph. Next, SIFT features of the 2D motion line graph are extracted. Finally, the line graphs are divided into several regions and then the comparison results can be calculated based on SIFT matching ratios between the tutor's local line graph and the trainee's local line graph. The experimental results show that the proposed method not only can easily deal with the several challenge problems in motion analysis, e.g., the problem of different rhythm of motions, the problem of a large amount of data, but also can provide detailed error correction cues.