Body Movement Recognition Based On Kinect

Human action recognition is a research hotspot in the fields of computer vision and pattern recognition.Human motion recognition technology has a wide range of applications in many fields,and has created huge social value and considerable economic value for the society.How to improve the accuracy and stability of body motion recognition,so that the machine can understand the human motion more quickly.Meeting people's needs and understanding people's expressions is the focus of current research.In this paper,the Kinect sensor was used to collect the bone point data of the body movements,and related algorithms were studied to classify and recognize the body movements.The recognized body movements were combined with mobile robots and applied to real life.(1)This article analyzes the principle of the Kinect human sensor for skeletal tracking and skeleton extraction,and uses the sensor with the Visual Studio 2015 development platform to collect and process body motion bone data.Eight kinds of static limb action samples and four kinds of dynamic limb action samples are defined,and the static limb action database and dynamic limb action database are established;(2)The study uses deep learning neural network algorithms to train and recognize the established static body movement database data,and compares and analyzes the recognition effects of the template matching algorithm and the K-NN algorithm.It is concluded that the static body movement recognition rates of the three algorithms are all reach more than 90%;The research uses deep learning neural network algorithm and support vector machine algorithm to train and recognize the established dynamic body movement database data,and compares the classification effect of the two,and concludes that the recognition rate of the two algorithms is more than 80%.This verifies the effectiveness of the deep learning neural network algorithm in body movement recognition;(3)This paper designs a robot remote control system based on body action recognition,builds a verification platform of man in the loop,combines body recognition with mobile robot,and the experimenter monitors the external environment through the network camera,and uses the body action to control the remote robot.