Hierarchical adaptive multi-scale hypergraph attention convolution network for skeleton-based action recognition

Open Webpage

Honghong Yang, Sai Wang, Lu Jiang, Yuping Su, Yumei Zhang

Published: 2025, Last Modified: 07 Nov 2025Appl. Soft Comput. 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Highlights•A HAM-HGNet is designed for skeleton-based action recognition. It is composed by HACM module, MHAC module andTSAM module.•A hierarchical adaptive clustering partition module (HACM) is formulated to construct dynamic graph topology.•A multi-scale hypergraph attention convolution module (MHAC) is formed to capture the inherent spatial characteristic.•A temporal segmentation attention constrained encoding module (TSAM) is designed to model the relationship between joints.