The Research Of Human Posture Recognition Based On Kinect Sensor

In the past two decades the rapid development of Internet and communication technology laid the cornerstone of the development of the current foundation,with it smart phones,smart wear smart devices,smart cars,virtual reality equipment has been gradually integrated into people's work and life in all aspects.The traditional way of input,like keyboard,mouse input and multi-touch technology has not fully meet the needs of people,human-computer interaction is confronted with a new challenge once again.This article using Kinect's skeleton tracking technology,design a platform of human posture recognition based on Kinect,and study the somatosensory technology of human-computer interaction.First of all,after learning Kinect for windows SDK,the skeletal information is retrieved by using Kinect.Processing the skeletal information,the skeletal vector is established.And the three-dimensional angle component of the skeletal vector is stored in the form of matrix to establish the mathematic model of human posture template.Secondly,the standard library of human pose was established,which included 12 common human body posture templates.After analyzing the skeletal information of the standard posture,in order to reduce the coupling between the skeletal vectors and the computational complexity,we choose to convert the human posture recognition into the limb posture recognition,and extract the eigenvalues of each limb posture.Then,based on the eigenvalue of limb posture,the posture classifier is designed in accordance with the template matching algorithm based on the vector eigenvalue of the skeleton,so as to realize the human posture recognition.Finally,programming by Visual Studio 2010,Kinect based human posture recognition platform is completed.The test of the platform include error debugging and functional testing.Human posture recognition is the key test items.The template matching algorithm based on skeletal vector eigenvalue and the skeletal tracking technique of Kinect are combined to realize the human body position identification,and the recognition rate is above ninety-eight percent.Which effectively solves the problem that the human posture recognition algorithm has high complexity and is not easy to be realized.