Development of an Automated Sports Teaching Assistance Tool Based on Image Recognition

In the context of deep integration between nationwide fitness initiatives and education informatization, traditional sports teaching faces challenges such as low efficiency in manual instruction and imprecise movement assessment.The advancement of image recognition technology provides a solid foundation for the intelligent transformation of sports education.However, existing studies still face bottlenecks in terms of motion recognition accuracy and the practical utility of teaching assistance tools.Traditional machine learning approaches rely heavily on handcrafted features, making it difficult to capture the spatiotemporal complexity of sports movements.While deep learning models have shown promise, they often overlook higher-order correlations among human body joints and the temporal dependencies of action sequences, resulting in suboptimal performance in recognizing dynamic and interactive movements.Moreover, current tools generally lack intuitive and effective modules for visualizing movement information.To address these issues, this study focuses on the development of an automated sports teaching assistance tool based on image recognition.The main contributions include: (1) proposing an enhanced hypergraph convolutional network that models higher-order joint correlations and incorporates temporal feature learning to improve the recognition accuracy of complex sports movements; and (2) designing a multidimensional motion information visualization scheme, enabling dynamic motion trajectory display and key joint deviation analysis to provide intuitive feedback for both teaching and learning.The research outcomes are expected to break through the spatial and temporal limitations of traditional instruction and establish a precise, personalized support system for sports education, offering both theoretical and technical support for its digital transformation.