Behavioral Tracking and Inference Algorithm Using Skeleton Vector Information Analysis

This paper proposed a behavioral tracking and inference algorithm by skeleton vector information analysis. The proposed algorithm uses only RGB+D information to track and inference using CCTV images without additional equipment. First, the articulated points of action are extracted from the extracted RGB+D image. And the continuous vector data is then calculated using the relation vector expression at the extracted action point. Then, continuous vector data is applied to nonlinear inference curve expressions to inference angular velocity and state of shaking. Apply the learned model to various datasets to verify the accuracy of behavioral tracking and inference. As a result, simulations of 3D action information have demonstrated significant performance differences.