Abstract:
Velocity-based training (VBT) is gaining popularity among strength and conditioning coaches over traditional methods due to its ability to quantify training intensity using movement velocity (m/s) as a standard, which is useful for prescribing other training variables. Existing VBT systems vary in functionalities, setups, accuracy, and costs. This research aims to develop and validate a markerless computer vision algorithm that uses Pose Estimation Models and RGB-D images to accurately estimate movement velocity during weight training irrespective of image orientations. Initial results show that the developed algorithm has a Mean Absolute Percentage Error (MAPE) of 4.82% in estimating movement velocity non-intrusively, compared to standard systems. This suggests that the developed algorithm can be used to build complete VBT systems for athlete load management with real-time feedback and effective progress tracking in daily and long-term periodization, aiding in reducing training stress, predicting fatigue, and injuries of the athletes.