The Research On Human-computer Interaction Based On Kinect And Its Application In Mine Fire Escape Simulation System

With the development of virtual reality and Kinect human-computer interaction technology, Kinect is frequently applied to somatic games. This thesis mainly describes how operator escapes from the virtual burning coal mine using Kinect.Firstly, it describes the hardware and software configuration of the whole system.Secondly, it describes how to use the somatic device Kinect to capture the human depth image, and human bone data. Thirdly, it describes how to use human depth image to recognize the positive and negative gestures and how the use human bone data to identify the body. Finally, it describes how to build a virtual tunnel model with UE4. And how to control character in the coal tunnel walking in every direction,and how to control some devices like fire extinguisher and coal mining machine and so on. And identify the identity of the operator using gait recognition algorithm.When comes to the positive and negative gesture recognition, this thesis uses Kinect as the main camera. Firstly, it describes how to use library function in Kinect for windows SDK 2 to get human depth image, then, it describes how to divide the hand area and divide the hand into two layers using threshold valve. The first layer is the layer of vertical finger. It first uses contour detection algorithm to detect the hand contour, second it uses the convex hull detection algorithm to identify the number of the fingers. The second layer is the layer of frizzy fingers. It uses the contour detection algorithm to detect contour, and if the contour exists, it is the positive gesture, otherwise, it is the negative gestures. At last, the two layer identification results are combined to realize backhand recognition.As for gait recognition, firstly, it describes how to use library function in Kinect for windows SDK 2 to get human bone data, then, it describes how to analyze human walking posture before the recognition and extract the feature vector. The main characteristics of people's gait include speed, the pace of each step, the amplitude of each limb swing,this paper choose 10 body angles for analysis,and from the curve of the lime-varying angles, it will be found that those angles are periodic. Extract one period, and fit with polynomial function, and then choose the polynomial fitting parameters as a feature,and at last identify the identity of the person using KNN classification algorithm.