Research On Multi-kinect Human-computer Interaction Model

The Natural Interaction technology is an important part of human-computer Interaction,playing a huge impact in VR(Virtual Reality),and then Kinect was produced with the concept of human-computer interaction.At present,most of somatosensory applications were developed only using a single device,but one device cannot meet the more site environment due to the limitation device perspective,leading to more narrow bone data capture range.At the same time,the target body is obscured and cannot be pinpointed when body moves under one single device that results in the decrease of the system performance.In this paper,the use of many somatosensory devices for human-computer interaction model and technical research,solves the narrow angle problem under one single device,and makes up the human activity space,and the solution is provided for human-computer interaction system applications design under larger space and environment.The research background and meaning,and the development of human-computer interaction technology,and the Kinect hardware and software of characteristics,and the Unity engine are introduced respectively,and,the user skeleton model under the device,and the target user body tracking and identification principle,are also introduced and analyzed,and the main research contents in this paper are as follows:(1)Camera calibration work:Firstly,the human bone data acquisition system was established and introduced,and then,the human body skeleton point data and the camera calibration technology under different Kinects are also analyzed and introduced.And the space coordinate calibration model under multi-Kinect was built,and the space coordinates calibration algorithm were gave the proposal and improvement measure to complete the camera calibration.(2)The fusion work of human bone joint data:Firstly,the data fusion technology and the related principle of the traditional Kalman filter algorithm are introduced,and then the bone joint point motion model is also established based on the movement characteristics of limb bones point,and the distributed fusion process of body skeleton data is completed by using the Kalman filter algorithm.Finally,the experimental analysis is gave with the limb joint data under the adjacent Kinect,and in this paper,the availability and effectiveness of the model and algorithm is also verified,and the skeleton joint data fusion work is completed.(3)The target body gesture recognition work:The human body skeleton model is built after completing the work of camera calibration and data fusion,and the proposal is gave which human body action is identified and classified by using the different bone joint point space coordinates,through the analysis of characteristics of the human body joints under Unity engine.And then,the human body skeleton model and the DTW algorithm are analyzed and introduced,and the body gestures is described by the established vector Angle model of human body skeleton space,and the work is completed that is simple body action and the design of the human-computer interaction system combining skeleton point coordinates with the DTW algorithm.