Symbolism and Directivity of Joint Keypoints in Temporal and Spatial Dimensions in Human Pose Prediction With GCN-Based Model

作     者：Li, Jinhui Huang, Jianying Kang, Hoon 

作者机构：Chung Ang Univ Sch Elect & Elect Engn Seoul 06974 South Korea 

出 版 物：《IEEE ACCESS》 (IEEE Access)

年 卷 期：2023年第11卷

页      面：146090-146102页

核心收录：

基　　金：Chung-Ang University Research Scholarship Grant  in 2022 

主　　题：Graph convolutional networks (GCN) spatial temporal graph convolutional networks (STGCN) 3D datasets human pose prediction directional & symbolic method human joints key points 

摘      要：A wide variety of methods have been developed to predict the posture of the human body at a given point in time based on data on previous movements. More recently, prediction models based on deep learning have become a topic of active research and development. In this study, we adopt the strategy of separating spatial and temporal information based on an existing STGCN model to extract features effectively in both space and time, and we analyzed the effects of signed or unsigned and directed or undirected forecasts of the positions of human joints with this approach. We propose a method using an encoder based on a modified graph adjacency matrix in a graph convolutional network model and focus especially on the terms of the signs and directions of data on the locations of the joints in space and time. We also introduce a global residual block. The results of an experimental evaluation of our proposed method showed that we obtained better performance by applying the signed and directed features independently to the spatial and temporal adjacency matrices. The proposed model exhibited noticeable improvements in several aspects. In future research, we expect these features of the modified adjacency matrix to help learning models understand the correlation between symbols and directions for various actions and poses.