3D Motion Analysis And Recognition Based On Measured Data

In recent years,3D motion analysis and identification becomes a research hotspot in the field of computer animation. How to describe the character of motion capture data, and to query motion, and how to improve the identification rate of the query motion, which would play an important role for solving the problems about3D animation.A new algorithm of motion analysis and identification is proposed in this paper, which is based on some approaches studied in previous research. The basic principle of motion capture data is briefly introduced firstly, and the data parameterization and similarity measure algorithm are discussed in this paper too. The retrieval result is analyzed according to the proposed method. The feedback learning frame is constructed to improve the motion analysis, which will be satisfied result with the customer. The works is divided into some parts in this paper, which can be described as bellows:(1) Based on CMU motion capture database, the testing motion database can be obtained in term of clustering analysis and other methods for further motion analysis.(2) Based on obtained motion database, several common motion features are described and compared, based on the advantages and disadvantages of every method, a suitable motion description can be found for motion capture data.(3) Due to the parameterized motion data, it is difficult for the computer to handle the high-dimension motion data. By comparing the several common dimension reduction methods, a suitable motion clustering algorithm is found to meet the motion descriptor dimensional reduction. (4) A human motion capture data retrieval method based on Quaternion and EMD is presented, which is a novel motion retrieval approach. The several motion retrieval methods are discussed. Experimental results show that proposed method has higher precision and higher stability compared to other existed methods.(5) A motion capture data retrieval method based on feedback learning is proposed in this paper. Many tests and the results about the algorithm have been analyzed further. Experimental results show that proposed retrieval system has higher recognition rate and higher feasibility.