Real‐Time Motion Capture and AI ‐Assisted Training Strategy Using Edge Computing and Smartphone Sensors

ABSTRACT Traditional cloud‐centric architectures face significant challenges when processing high‐frequency multimodal data from massive smart devices and smartphone sensors used during motion capture and training. Current approaches also struggle with the low‐latency computational demands of various sensors. Edge computing has appeared as a novel way to reduce the processing latency. Hence, to address these challenges, a real‐time motion capture method is proposed under the edge computing framework for an AI‐assisted sports training strategy. First, an edge computing‐based framework is designed to leverage distributed edge resources and lightweight AI models for sensor data capture and training guidance on smartphones. Besides, based on the lightweight Long Short‐Term Memory (LSTM) model, we propose a real‐time motion capture and AI‐assisted sport training strategy. The LSTM is integrated to analyze temporal dependencies in sequential data, which is ideal for capturing motion patterns, while the feedback mechanism is applied to optimize the sports training process iteratively. By comparing the proposed method with recent state‐of‐the‐art approaches, the experimental results demonstrate that our method shows better performance in latency, accuracy, and F 1‐score.