Karelin Alexey, Brazhenko Dmitry, Kliukovkin Georgii, Chernenko Yehor
Independent Researcher, 30-392 Krakow, Poland.
Independent Researcher, Seattle, WA 98107, USA.
Sensors (Basel). 2025 Aug 10;25(16):4943. doi: 10.3390/s25164943.
This study presents a real-time hand tracking and collision detection system for immersive mixed-reality boxing training on Apple Vision Pro (Apple Inc., Cupertino, CA, USA). Leveraging the device's advanced spatial computing capabilities, this research addresses the limitations of traditional fitness applications that lack precision for technique-based sports like boxing with visual-only hand tracking. The system is designed to provide objective feedback by recognizing boxing-specific gestures with sub-centimeter accuracy and validating biomechanical correctness during punch execution. A three-stage pipeline consisting of geometric filtering, biomechanical validation, and punch technique assessment rejects accidental or improper motions. Experimental evaluation involving 12 participants demonstrated a gesture recognition accuracy of 96.3% and a technique validation accuracy of 88.5%. The system consistently operated at 60 FPS with low latency and high robustness across diverse lighting conditions. These results indicate the potential of Apple Vision Pro as a platform for precision sports training and highlight the educational impact of mixed reality in democratizing access to high-quality boxing instruction. The proposed framework is extensible to other skill-based sports requiring fine motor control and real-time feedback.
本研究展示了一种用于苹果Vision Pro（美国加利福尼亚州库比蒂诺苹果公司）沉浸式混合现实拳击训练的实时手部跟踪与碰撞检测系统。利用该设备先进的空间计算能力，本研究解决了传统健身应用的局限性，这些应用仅通过视觉手部跟踪对拳击等基于技术的运动缺乏精度。该系统旨在通过以亚厘米级精度识别特定拳击手势并在出拳执行期间验证生物力学正确性来提供客观反馈。由几何滤波、生物力学验证和出拳技术评估组成的三阶段管道可排除意外或不当动作。涉及12名参与者的实验评估显示手势识别准确率为96.3%，技术验证准确率为88.5%。该系统在各种光照条件下始终以60帧每秒的速度运行，具有低延迟和高鲁棒性。这些结果表明苹果Vision Pro作为精密运动训练平台的潜力，并突出了混合现实在普及高质量拳击教学方面的教育影响。所提出的框架可扩展到其他需要精细运动控制和实时反馈的基于技能的运动。
