The Research Of Motion Analysis Techniques For Computer Aided Training System

Motion analysis technology is a very broad field, which involves the computer vision,artificial intelligence, pattern recognition and so on. It is a cross strong discipline. Due to itshigh application value, so the motion analysis technology has always been the hot researching.In recent years, it has spawned a hot research topic from the motion analysis technology-humanbehavior recognition. Human behavior recognition technology has extremely extensiveapplication prospect and economic value such as in sports analysis, intelligent video monitoring,virtual reality and human-computer interaction and so on. So it got the extensive attention of theresearchers. Due to the complexity of the human body movement and the diversity of theexternal environment, human behavior recognition becomes a challenging subject. The study ofhuman behavior identification is still in a primary stage, it remains to be for further explorationand research.Take table tennis as the research background, this article made the following job:1. First of all, this paper introduced the research significance and background of themotion analysis technology and the present development situation. In this part mainlyintroduces the more mainstream method and the core technology in the field of motion analysis.It is concluded that the emphasis and difficulty of this topic research. This indicates thedirection of the research in this paper.2. We had a research on image preprocessing and moving target detection techniques. Firstwe introduced image enhancement technology, which is a necessary step in practicalapplication. Then we also analyzed target segmentation technology, and the human skindetection method based on YCbCr color space is proposed in this paper, we used this method tosplit table tennis athletes’ arms from the background, which is a preparation for the follow-upresearch.3. For the question of K-L transform algorithm can only applied to static image recognition and can not be used to identify the image sequence, the profile synthesis method forimage sequence was put forward in this paper, and we also analyzed the reliability of thismethod. Then we can use K-L transform for trajectory recognition.4. We made a research on motion recognition technology based on trajectorycharacteristics. First we analyzed feature extraction method from geometric and shape feature,and used Hu matrix to extract the invariant feature of trajectory, then made K-L transform forthe matrix feature. The experimental results show that Hu matrix can be as effectivecharacteristics of the trajectory, and the trajectory recognition effect used Hu matrix made K-Ltransformation is much better than the effect only used K-L transform.