Computer Vision in Assessment of Complexity of Exercises of Figure Skating

• Published in: Ural Symposium on Biomedical Engineering, Radioelectronics and Information Technology (Online) pp. 63 - 66
• Main Authors: Novikova, Irina, Romanenko, Sergey, Kulikova, Okxana, Sezganov, Vladislav, Veremchuk, Natalia, Sarychev, Vladislav
• Format: Conference Proceeding
• Language: English
• Published: IEEE 13.05.2024
• Subjects: Accuracy; Cameras; Computer vision; computer vision technology; data analysis; deep learning; human movements; Neural networks; Skeleton; Task analysis; Training
• ISSN: 2769-3635
• DOI: 10.1109/USBEREIT61901.2024.10584006

Managing the athlete's speed distribution strategy in complex-coordination sports using computer vision is one of the priority tasks in modern science. The research is aimed at developing an algorithm for calculating the speed performance of training exercises in figure skating using neural networks and RGB camera. Body-25 key-point format for human pose skeleton was used in the calculations. The algorithm developed by the authors includes the following steps: 1) video recording of the exercises is performed on an RGB camera; 2) 2D reconstruction of the coordinates of the point defining the general center of mass of the athlete's body using the Openpose library; 3) calculation of the first and second derivative in each beat of the exercise time; 4) construction of reference curves of the skater's velocity change during the exercises and determination of the phases corresponding to the specified process; 5) calculation of indicators characterizing the strategies of the exercise velocity change in figure skating. The programming language Python 3.10 was used. The validation of the calculation results was carried out using the Kinovea video analysis program. Using the author's algorithm, two main strategies for changing speed during the exercise "basic stroking" and one strategy for the exercise "sleds" were identified. The application of computer vision systems allows to calculate with sufficient accuracy the indicators of speed in figure skating and to optimize the training process in complex-coordination sports.