关键词: Kinect, 人体动作识别, 划分区域, Multi-HMM（多隐马尔可夫）模型, 耦合KNN（耦合K最邻近）

Abstract:

In order to solve the problem that the feature extraction was easy to be affected by external factors and the computational complexity was high.Using a more effective solution, that was the use of depth data for human motion recognition. Using the joint data collected by Kinect, the human joint was divided into five regions. The vector angle of each region was discretized to describe the different states, and then the Baum-Welch algorithm was used to study the multi- HMM(multi-Hidden Markov Model)model, and used the forward algorithm to establish the generation region and action class probability matrix. On this basis, the regional and action categories were internally coupled and coupled to analyze, thus expressing the interaction between the joints. Finally, the KNN algorithm based on coupling was used to complete the motion recognition. Through the experimental test, the recognition rate of the five actions was more than 90%, and compared with the methods such as 3D Trajectories. The comprehensive recognition rate obtained by the experiment was higher than the contrast methods, and has obvious advantages.

Key words: