Human Gait Tracking And Recognition Technology Based On Kinect

With the continuous development and progress of science and technology, people pay more and more attention to virtual reality technology, which is widely used in industrial simulation, virtual roaming, games, medical, military, aerospace and other fields.Especially in the field of virtual roaming, it’s very popular with people because of its strong visual impact and rich interactivity. However, the way of virtual roaming is usually through virtual glasses, helmets, gloves and other equipment to meet people’s needs. There is less interaction with people’s walking.This paper proposes a method of walking device made of pedals based on reading a large number of literature.Through the video device to capture the movement of human gait and transmit the data to walking device and model in the virtual world, which makes them move at the same time. The main content of this paper is researching how to capture and recognize the movement of human gait, how to transmit the data to virtual model and how to make the user and model move simultaneously.First of all, this paper makes an investigation and research on capture way of human gait motion, proposes a method of human gait tracking and recognition based on Kinect, analyzes the functions of Kinect and the architecture of Kinect for Windows SDK. Get the depth image of human gait motion by Kinect and use the method of background difference to separate the object and background, which gets the single depth image of human body.Then this paper designs a method of gait tracking based on skeleton model matching according to the principle of particle filter algorithm. Determine each joint’s initial position in the skeleton model according to the matching of the initial frame, combining the movement of human gait and using particle filtering algorithm to predict the location of bone joint, so as to realize the real-time matching of skeleton model and image. Meanwhile, this paper also proposes a gait tracking method based on template matching algorithm, using the normal camera to record, setting two color markers on human feet, to realize the human gait tracking by capturing the movement of two markers.This paper suggests that use the function of the angle of hip, knee, ankle about time variation as the feature of human gait motion according to skeleton model. Use the method of optimized DTW algorithm to recognize the gait posture of turning left, turning right, up and down the stairs.Then, analyze and verify the data captured by Kinect.Finally, use 3ds Max to establish a skeleton model which is equal to virtual model and design the scripts of virtual model motion in Unity 3d, which can make the skeleton joints of virtual model get the real-time data captured by Kinect, so as to realize the movement of user and virtual model at the same time.