.png){kind=icon}
Return
Spatio-temporal local graph convolutional network for skeleton-based action recognition
.png){kind=icon}
PRE
AI
DOI：10.1007/s13042-026-03270-7.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In recent years, Graph Convolutional Networks (GCNs) have been widely used for skeleton-based action recognition and have achieved remarkable performance. However, although existing GCNs can effectively capture global relationships between joints by spatial graph convolution, they still have limitations in capturing local and complex spatio-temporal relationships, and in fully utilizing semantic information for feature learning. To overcome these limitations, we propose a novel approach called Spatio-Temporal Local Graph Convolutional Network, which consists of a Partitioning Module (PM), a Spatio-Temporal Graph Convolution (ST-GC) module and a Joint Prototype Loss (JPL) function. Specifically, the PM partitions the global skeleton into multiple local skeletons, thereby facilitating the capture of local relationships between joints. Besides classical spatial graph convolution, the ST-GC also adopts temporal and spatio-temporal graph convolutions to effectively capture complex spatio-temporal dependencies. All three types of convolutions leverage their respective data-driven dynamic topologies for feature aggregation. In the JPL, prototypes with semantic information are enhanced by incorporating biological topology to introduce spatial structural information, thereby guiding feature learning to improve classification performance. We conduct extensive evaluations on the NTU RGB+D 60 and NTU RGB+D 120 datasets, demonstrating that our proposed method achieves competitive recognition performance.
Keywords:
Skeleton-based action recognition
Partitioning
Graph convolution
Prototype loss
Journal
IF:
2.7
Papers:
3.2K
Citations:
5.6K
Organization
Cited Papers
ShareSaveShareSaveShareSaveA Survey of the State of the Art in Monocular 3D Human Pose Estimation: Methods, Benchmarks, and Challenges.png)
SENSORS
OAAI
SENSORSIF3.5
OAAIShareSaveMulti-scale receptive fields: Graph attention neural network for hyperspectral image classification
PREAI
ShareSave
View more