HFGCN:Hypergraph Fusion Graph Convolutional Networks for Skeleton-Based Action Recognition
Dong, Pengcheng Guo, Xin Xu, Kun Li, Shuai Chang, Xiaojun Li, Zhihui Li, Jing Sun, Jiande
Dong, Pengcheng
Guo, Xin
Xu, Kun
Li, Shuai
Chang, Xiaojun
Li, Zhihui
Li, Jing
Sun, Jiande
Supervisor
Department
Computer Vision
Embargo End Date
Type
Journal article
Date
License
Language
English
Collections
Research Projects
Organizational Units
Journal Issue
Abstract
In recent years, action recognition has received much attention and wide application due to its important role in video understanding. Most of the research on action recognition methods focused on improving the modeling capabilities between skeleton point pairs to improve performance. The topological modeling between skeleton points and body parts was seldom considered. Although the topology of the skeleton points is classified via a data-driven approach in some studies, the nature of the skeleton point in terms of kinematics has not been taken into consideration as far as we know. Therefore, in this paper, we draw on the theory of kinematics to adapt the topological relations of the skeleton point and propose a topological relation group based on body parts and distance from the core of the body. We propose a Hypergraph Fusion Graph Convolutional Network (HFGCN) to synthesize these topological relations for action recognition. In particular, the proposed model is able to focus on the human skeleton points and the different body parts simultaneously, and thus construct the topology, which improves the recognition accuracy. We use a hypergraph to represent the categorical relationships of these skeleton points and incorporate the hypergraph into a graph convolution network to model the higher-order relationships among the skeleton points and enhance the feature representation of the network. In addition, a hypergraph attention module and a hypergraph graph convolution module are designed to optimize topology modeling in temporal and channel dimensions respectively and enhance the feature representation of the network. We conducted extensive experiments on three widely used datasets, that is, NTU RGB + D, NTU RGB + D 120, and NW-UCLA. The results validate that our proposed method can achieve the best performance when compared with the state-of-the-art skeleton-based methods.
Citation
P. Dong, X. Guo, K. Xu, S. Li, X. Chang, Z. Li , et al., "HFGCN:Hypergraph Fusion Graph Convolutional Networks for Skeleton-Based Action Recognition," IEEE Transactions on Multimedia, vol. PP, no. 99, pp. 1-13, 2026, https://doi.org/10.1109/tmm.2026.3688407.
Source
IEEE Transactions on Multimedia
Conference
Keywords
46 Information and Computing Sciences, 4602 Artificial Intelligence
Subjects
Source
Publisher
IEEE