Research On Human Posture Recognition And Interaction Technology In Virtual Environment

As a cutting-edge research discipline,virtual reality technology can simulate reality and create a highly realistic digital virtual environment,so it can bring deep visual experience and rich cognitive information to people.However,the interaction methods in the traditional virtual environment have great limitations and reduce the practical value of the virtual environment.Therefore,consider the organic combination of somatosensory interaction technology and virtual environment to solve it.Posture is an important part of body interaction,and its recognition accuracy has an important influence on the sense of interaction experience.In order to further improve the accuracy of recognition,research on human body posture recognition on the basis of human skeleton diagrams.First,analyze the currently commonly used pose estimation algorithms,use openpose to complete the extraction of human body key point,then further study the feature extraction and recognition methods on the basis of constructing the human skeleton map,construct the deep neural network model for the spatiotemporal feature information of the human skeleton Extraction,using graph convolution to learn spatial features combined with multi-scale time convolution(Multi scale TCN)captured skeleton action sequence information in the time dimension,so that the network model can learn higher-level features,and the network model is solved by adding a residual network for the degradation problem.Proposed a Multi-scale ST-GCN method to model the dynamic information of bones,and the attention model is used to learn the importance of corresponding key points in different poses,through a comparative experiment designed to evaluate and test the model with two types of CS and CV data sets on the data set NTU-RGBD,The effect of the proposed network model is verified.The proposed algorithm model for human posture recognition is used to improve the accuracy to 80.1% and 89.8%,Design a posture training scene for virtual crane,using human posture to interact,the accuracy reach to 96%.This method of interaction in the virtual environment becomes more natural and improve the user's feeling.Figure 51;Table 13;Reference 62...