IRMA-International.org: Creator of Knowledge
Information Resources Management Association
Advancing the Concepts & Practices of Information Resources Management in Modern Organizations

Dynamic Convolutional Network for Enhancing Effectiveness of Action Recognition Under Deep Learning Technology

Dynamic Convolutional Network for Enhancing Effectiveness of Action Recognition Under Deep Learning Technology
View Sample PDF
Author(s): Siyu Wang (Department of Physical Education, Guangzhou City University of Technology, China), Xing Liu (Department of Physical Education, Guangzhou City University of Technology, China), Dong Lu (School of Competitive Sports, Guangdong Vocational Institute of Sport, China), Xiaoyi Yang (Auckland Bioengineering Institute, University of Auckland, New Zealand)and Jing Chang (College of Physical Education, Guangdong Baiyun University, China)
Copyright: 2026
Volume: 19
Issue: 1
Pages: 16
Source title: International Journal of Information Technologies and Systems Approach (IJITSA)
Editor(s)-in-Chief: Sangbing (Jason) Tsai (International Engineering and Technology Institute (IETI), Hong Kong)and Wei Liu (Qingdao University, China)
DOI: 10.4018/IJITSA.396697

Purchase


Abstract

With the rapid advancement of artificial intelligence and deep learning, leveraging intelligent algorithms to enhance physical education has become a prominent research focus. Yoga, which integrates fitness, rehabilitation, and psychological regulation, demands high accuracy and standardization in movements. Traditional teaching methods, however, often exhibit low instructional efficiency and limited individualization, making them insufficient for large-scale, multilevel learners. To address this challenge, the authors propose a yoga instruction method based on a deep learning dynamic convolutional network (DCN). Keypoint detection and data augmentation were applied to a publicly available yoga action dataset to build libraries of standard and incorrect movement samples. The model integrates dynamic convolution kernels with channel and self-attention mechanisms, enabling adaptive adjustment of kernel weights on the basis of input features and enhancing the network's ability to capture subtle movement details.

Related Content

Siyu Wang, Xing Liu, Dong Lu, Xiaoyi Yang, Jing Chang. © 2026. 16 pages.
Wenjing Nian, Ye Gao, Juntang Yang, Xiaoyi Yang, Luguang Wen. © 2026. 19 pages.
Yang Li, Yang Yang, Zihan Wang. © 2026. 18 pages.
He Danni. © 2026. 21 pages.
Yiyu Dai, Junzheng Lu, Zesen Li, Jiawei Li, Yunxi Lu. © 2026. 18 pages.
Tao Shi, Qiang Wan. © 2026. 17 pages.
Weijian Zhu, Jingting Lin. © 2026. 19 pages.
Body Bottom