摘要
競速直排輪運動要求學員在訓練中保持正確姿勢，以提升運動表現並減少傷害風險。然而，膝蓋與腳踝的常見姿勢問題對初學者的學習造成挑戰。隨著影像處理和人體姿態偵測技術的發展，將這些技術應用於運動姿勢分析為量化學員的姿勢表現提供了新方法，能輔助教學與學習過程。本研究採用OpenPose作為人體姿態偵測工具，結合影像處理與數據分析技術，量化學員在直排輪練習過程中的膝蓋與腳踝角度特徵，並針對特定的姿勢問題進行分類與評估。實驗設計分為四個階段：第一階段開發步伐週期自動偵測系統，透過分析關鍵點座標自動切割完整步伐；第二階段驗證姿勢辨識的準確度，將系統判定結果與教練提供的標準（Ground Truth）進行比對評估系統表現；第三階段分析學員的姿勢表現，通過數據生成針對性的回饋評論；第四階段探討髖部位移與重心偏移的相關性，並利用動態閾值方法檢測滑行過程中的異常幀，分析異常對關節點偵測與姿勢分析的影響。研究結果顯示，自動步伐偵測系統展現出75.37%的F1分數，能有效識別大部分的步伐週期，為未來全自動化姿勢分析奠定基礎。在後續的姿勢分析實驗中，為確保判斷的準確性，採用人工標記的步伐進行評估。結合影像處理與姿態偵測技術，可以有效量化學員的滑行姿勢表現，對膝蓋與腳踝等關鍵部位的姿勢特徵進行深入分析，並提供清晰的數據回饋，幫助學員理解與改善不良姿勢。本研究為競速直排輪運動中的姿勢分析提供了系統化解決方案，未來可進一步拓展至其他運動項目與即時偵測應用。
並列摘要
Speed skating requires maintaining proper posture during training to enhance performance and reduce injury risk. Common posture issues in the knees and ankles present challenges for beginners. With advancements in image processing and human pose estimation, these technologies provide a new approach to quantifying posture performance, supporting both teaching and learning.This study uses OpenPose as a pose estimation tool, integrating image processing and data analysis to evaluate knee and ankle angles during skating practice. The research is divided into four stages: the first develops an automated stride cycle detection system by analyzing key point coordinates; the second validates the system's posture recognition accuracy against coach-provided Ground Truth; the third generates detailed feedback on posture performance; and the fourth explores the relationship between hip displacement and center-of-gravity shifts, using dynamic thresholds to detect abnormalities and assess their impact on joint detection and posture analysis.Results show that the automated stride detection system achieves an F1 score of 71.13%, effectively identifying most stride cycles. Combining image processing and pose estimation methods can quantify skating posture, analyze key features, and provide actionable feedback to help learners improve. This study offers a systematic approach to posture analysis for speed skating, with potential applications in other sports and real-time systems.
參考文獻
[1] Allinger, T., & van den Bogert, A. J. (1997). Skating technique for the straights, based on the optimization of a simulation model. Medicine and Science in Sports and Exercise, 29(2), 279-286.
[2] Bazarevsky, V., Grishchenko, I., Raveendran, K., Zhu, T., Zhang, F., & Grundmann, M. (2020). BlazePose: On-device real-time body pose tracking. arXiv preprint arXiv:2006.10204.
[3] Cambridge, E. D. J., Sidorkewicz, N., Ikeda, D. M., & McGill, S. M. (2012). Progressive hip rehabilitation: The effects of resistance band placement on gluteal activation during two common exercises. Journal of Orthopaedic & Sports Physical Therapy, 42(9), 731-740.
[4] Cao, Z., Simon, T., Wei, S. E., & Sheikh, Y. (2016). Realtime multi-person 2D pose estimation using part affinity fields. arXiv preprint arXiv:1611.08050.
[5] Cao, Z., Simon, T., Wei, S. E., & Sheikh, Y. (2017). Realtime multi-person 2d pose estimation using part affinity fields. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (pp. 7291-7299).