Artificial intelligence and markerless motion capture in kinesiology: a systematic review

Artificial intelligence and computer vision have made significant progress in recent decades, profoundly impacting many scientific and professional disciplines, kinesiology included. The development of markerless motion capture technologies has enabled precise tracking of kinematic and dynamic parameters of the human body movements without the need for physical markers or complex equipment. These technologies use advanced computer vision and deep learning algorithms to analyze human movements in real time, allowing the quantification of biomechanical parameters such as joint angles, movement speed, stride length, gait asymmetries, and complex movements such as jumping or running. Markerless technologies reduce preparation and recording time and allow movement analysis in natural conditions, making them useful not only in laboratory but also in clinical, sports, rehabilitation, and everyday settings. The use of smartphone video recordings further facilitates the availability and implementation of these systems. However, the application of markerless technologies to complex three-dimensional movements, such as trunk rotations or upper limb activities, remains a challenge. The accuracy of these systems depends on various factors, including movement type, number of cameras, recording quality, and lighting conditions. Advances in deep learning and computer vision allow continuous improvement in reliability, making these systems more competitive with the traditional marker-based methods. Markerless technologies have significant potential in rehabilitation and sports performance optimization, but further development is needed regarding validation standardization and algorithmic robustness. This paper aims to show how markerless technologies enable new approaches in the analysis of human movement, exploring their advantages, challenges, and potential for further development in kinesiology.