Research And Implementation Of The Somatosensory Aided Shape Interaction Mode Based On Kinect Action Recognition

Nowadays the science and technology is developing rapidly.In product design,the orientation of meeting the basic functional requirements of the users has gradually turned into meeting the perceptual demand-oriented requirements.In recent years,the huge development of virtual reality and somatosensory interaction technology also create a richer and varied way for computer-aided product design,greatly improving the efficiency of product design and human-computer interaction friendly.Based on the theory of human-computer interaction and computer graphics,this research has developed a somatosensory aided product design system based on Kinect platform,action recognition and virtual reality technology.The software engineering principles and requirements and survey methods are used to integrate somatosensory technology,especially the action recognition technology,into the product design system,which can solve the existing problems of computer aided product design interaction by a novel way of natural and efficient interaction,thus increasing the operability and availability of the computer aided product design system,which has a certain significance for the computer aided product design system interaction and overall product design.The basic structure of the SASI system is established,the functional requirements and non-functional requirements of the product design and human-computer interaction are analyzed to get the basic elements of the SASI system.The scene of the system is designed and the hierarchical logic between scenes is established.Each sub-scene and the interface between each pair of scenes are analyzed and designed,finally,the effect settings of each scene and the aiming function are also studied.Based on the principle of graph correlation,the target information at the intersection point obtained by ray tracing and collision detection is obtained.Through the coordinate transformation,the relevant useful information is finally processed.Their coordinates are got in the corresponding coordinate system and lastly interacted with the UI control.Aiming at the shortcomings of identifying high similarity motion capability of the existing motion recognition algorithms,a stable and efficient high similarity motion search and recognition algorithm is proposed.Based on BP neural network,the Kinect depth data and skeletal data are used to extract the relevant joint information and thus define the stable feature vector from them to describe the target action.Among them,the extraction of the feature vectors based on the joint information of the human action and BP neural network learning are the keys of this algorithm.The function of each human-computer interaction module of SASI system is studied,the function of the interface between Kinect hardware and SASI system is analyzed.The working principle and application method of two important data streams provided by Kinect API are researched,and the human body joint information is obtained by Kinect data stream to complete the human action recognition and gesture interaction with the system.Based on the research above,the interactive function of SASI system is realized.