Research on Action Posture Feature Extraction of Volleyball Players Based on AR Technology

This study presents an innovative approach for extracting posture features of volleyball players' hitting actions using Augmented Reality (AR) technology. The research introduces a multi-modal system combining AR devices, depth cameras, accelerometers, and gyroscopes to capture real-time motion data, including 3D skeletal coordinates, dynamic sensor data, and video sequences. A core focus of the study is the development of an optimized algorithm for feature extraction, integrating Convolutional Neural Networks (CNN), Graph Convolutional Networks (GCN), and Long Short-Term Memory (LSTM) networks to improve accuracy in capturing key motion points and dynamic relationships during volleyball hitting. The methodology involves a two-phase approach: first, real-time data collection of athletes' motions during key action phases (preparation, jump, hit, and landing); second, a detailed preprocessing pipeline that includes Kalman filtering for noise reduction, time-series alignment, and data normalization. These steps ensure data consistency and accuracy across multiple data sources. Following this, the feature extraction process is enhanced through the integration of GCN for modeling skeletal joint relationships, CNN for extracting static features, and LSTM for temporal sequence analysis, improving both feature precision and the algorithm's ability to handle dynamic motion changes. Experimentation involved 50 volleyball players, each performing 30 hitting actions, yielding a comprehensive dataset used for training and validating the proposed algorithm. The results demonstrate significant improvements in keypoint localization accuracy, dynamic feature extraction, and phase-based action analysis. The optimized algorithm outperformed traditional approaches, showing reduced computation time and enhanced robustness in complex motion scenarios. These findings contribute valuable insights into the application of AR in sports motion analysis and provide a scalable solution for real-time athlete performance evaluation and feedback. This study not only enhances the technical capabilities for analyzing volleyball hitting actions but also sets the groundwork for applying AR technology to broader sports training and analysis contexts.