Sports behavior analysis technology based on GCN and domain knowledge graph
Abstract To improve the performance of sports behavior recognition, the spatial temporal graph convolutional network is introduced to analyze the spatial temporal features of sports behavior, achieving accurate action recognition. In the experimental results, the proposed graph convolutional network...
Uloženo v:
| Vydáno v: | Discover Computing Ročník 28; číslo 1; s. 1 - 22 |
|---|---|
| Hlavní autoři: | , |
| Médium: | Journal Article |
| Jazyk: | angličtina |
| Vydáno: |
Springer
17.11.2025
|
| Témata: | |
| ISSN: | 2948-2992 |
| On-line přístup: | Získat plný text |
| Tagy: |
Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
|
| Shrnutí: | Abstract To improve the performance of sports behavior recognition, the spatial temporal graph convolutional network is introduced to analyze the spatial temporal features of sports behavior, achieving accurate action recognition. In the experimental results, the proposed graph convolutional network algorithm achieved accuracy of 95.20%, 93.80%, and 92.50% in single-person, multi-person, and complex scenes, respectively. The spatial temporal graph convolutional network achieved a recognition accuracy of 95.3% in sports behavior analysis, with a Top-5 accuracy of 98.4%, an average recognition time of 12.5 ms/frame, and a parameter count of only 12.5 million, demonstrating its advantages in real-time performance and model complexity. The results indicate that the framework combining the proposed graph convolutional network algorithm, improved Openpose estimation, and spatial temporal graph convolutional network can effectively capture the spatio-temporal features of sports behavior and significantly improve recognition performance. |
|---|---|
| ISSN: | 2948-2992 |
| DOI: | 10.1007/s10791-025-09794-w |