Research On Human Motion Recognition Based On Kinect Human Skeleton Model

Human action recognition is a hot research question in computer vision and pattern recognition. Human action recognition is widely used in the intelligent monitoring, film actions production, human-computer interaction, motion analysis, medical rehabilitation and other related fields,and creates a lot of social and economic benefits. This paper analyzes the research status of human action recognition and related action recognition algorithm, based on the kinect human skeleton model,expanding the study of human action recognition. The main work in this paper is as follows:(1) First, constructing of the motion recognition feature vector. This paper extracts the feature information of the spatial position from the human skeleton motion video via the kinect equipment, on this basis, Combines the static and dynamic properties of human action to construct recognition feature vectors.(2) Secondly, proposed a new model of frame selection.in the process of identification of motion video, Not all of the frame images are favorable for recognition, Some frames may even confuse the action category, affecting the final recognition results, proposed the model of the frame selection algorithm based on accumulated Motion Energy. This model will greatly improve the recognition efficiency, reduce the amount of computation and shorten the running time.(3) Then, the paper constructed a human action recognition system based on fuzzy SVM, The feature vectors obtained were normalized firstly, Secondly, using principal component analysis to reducing dimension, and finally training modeling and classifying based on fuzzy support vector machine algorithm.(4) The last part of the paper gives the construction of the experimental platform and analysis of the experimental results, detail analysis the impact of the number of the experimental data and the experimental objects, compared the constructed action recognition system and the current advanced recognition algorithms to identify the pros and cons, finally gives the detailed experimental results data and showing in tabular.