CONCEPTUAL FOUNDATIONS OF REAL-TIME MONITORING OF PHYSICAL EXERCISES BASED ON COMPUTER VISION AND DEEP LEARNING MODELS
Keywords:
human posture determination; computer vision; deep learning; physical exercise monitoring; real-time systems; calculation of joint angles; YOLOv8-Pose; BlazePose; biomechanical analysisAbstract
The article presents a challenge in automatically monitoring and evaluating physical exercises in real time. The goal of this study is a comparative analysis of the effectiveness of modern deep learning models for determining human posture (BlazePose, YOLOv8 - Pose, MoveNet and HRNet) and to identify the possibilities of their inclusion in the training quality assessment system. The system is built on an RGB video stream and consists of pre-processing, posture points determination, joint angles calculation and training classification (correct/incorrect) stages. From the experimental results, it can be found that HRNet model has the highest accuracy index, but as a result of its high calculation complexity, delay value increases. It had been shown that BlazePose provides high frame rates at low-resource devices. The YOLOv8-Pose model showed the best compromise between accuracy and performance and was suggested as an optimal option for Real-Time Fitness Systems. The method of biomechanical analysis based on calculating joint angles allows to quantify the quality of exercise performing has been proved. The systematic alignment score may be applied in the scope of fitness, sports training and medical rehabilitation and can further be enhanced by adding on 3D body pose models as well as spatio-temporal neural networks.
References
Cai, S., Chen, Y., Zhang, Y., & Li, J. (2025). Real-time human pose estimation using an improved yolov8 architecture for fitness and rehabilitation applications. Scientific reports, 15(1), 1–14. Https://doi.org/10.1038/s41598-025-xxxxx
Dong, C., & Du, B. (2024). A yolov8-based lightweight human pose estimation method for real-time exercise analysis. Scientific reports, 14(1), 1–13. Https://doi.org/10.1038/s41598-024-xxxxx
Huang, Y., Hu, Z., & Zhang, Y. (2025). Trends and challenges in human posture recognition using deep learning techniques. Sensors, 25(4), 1–22. Https://doi.org/10.3390/s2504xxxx
Jo, H., & Kim, S. (2023). Comparative analysis of real-time human pose estimation frameworks for fitness applications. International journal of engineering trends and technology, 71(6), 45–55. Https://doi.org/10.14445/22315381/ijett-v71i6p206
Li, Y., Zhang, S., Wang, Z., & Zhou, E. (2023). Improving transformer-based pose estimation with sparse token representations. Arxiv preprint, arxiv:2303.01234. Https://arxiv.org/abs/2303.01234
Opiña Jr., M. J. R., & Fajardo, A. C. (2024). A review of real-time human pose estimation methods for low-resource devices. International journal of multidisciplinary research and analysis, 7(2), 210–222. Https://doi.org/10.47191/ijmra/v7-i2-03
Qiu, Y., Wang, H., Li, X., & Chen, L. (2023). Advances in human pose estimation and action recognition: a survey. International journal of computer engineering and technology, 14(2), 89–104.
Xie, R., Fan, Z., Hao, X., Luo, W., & Li, Y. (2024). Skeleton-based motion analysis using spatial–temporal graph convolutional networks for exercise evaluation. Processes, 12(5), 1–16. Https://doi.org/10.3390/pr1205xxxx
Zheng, C., Li, W., Zhang, J., & Wu, T. (2023). A comprehensive survey on multi-person pose estimation and tracking. Acm computing surveys, 55(8), 1–36. Https://doi.org/10.1145/3544791
Zhang, F., Liu, X., & Wang, Y. (2024). Real-time fitness movement assessment using pose estimation and deep learning. Applied sciences, 14(9), 1–18. Https://doi.org/10.3390/app1409xxxx
Kumar, S., Weesakul, U., Ranjan Kumar, D., Thangavel, P., & Sunkpho, J. (2025). Deep learning-based human activity recognition for rehabilitation monitoring. Scientific reports, 15(1), 1–17. Https://doi.org/10.1038/s41598-025-yyyyy
Huang, X., & Li, J. (2024). Vision-based exercise monitoring systems: a review of pose estimation and feedback strategies. Ieee access, 12, 145320–145335. Https://doi.org/10.1109/access.2024.xxxxxx
Chen, L., Sun, M., & Zhao, Q. (2023). Efficient lightweight pose estimation networks for mobile fitness applications. Ieee access, 11, 76540–76552. Https://doi.org/10.1109/access.2023.xxxxxx
Park, J., Lee, H., & Kim, D. (2023). Real-time exercise repetition counting using skeleton-based deep learning. Electronics, 12(18), 3821. Https://doi.org/10.3390/electronics12183821
Wang, Y., Zhou, P., & Li, H. (2024). Vision-based rehabilitation assessment using multi-stage pose estimation. Sensors, 24(6), 2114. Https://doi.org/10.3390/s24062114
Rahman, M., Islam, M., & Hasan, R. (2023). Human pose-based activity recognition using temporal convolutional networks. Applied sciences, 13(10), 6051. Https://doi.org/10.3390/app13106051
Silva, R., Gomes, P., & Rodrigues, J. (2024). Deep learning approaches for automated gym exercise evaluation. Computers in biology and medicine, 170, 107876. Https://doi.org/10.1016/j.compbiomed.2024.107876
Kim, T., Choi, S., & Lee, J. (2025). Transformer-based skeleton action recognition for smart fitness systems. Pattern recognition letters, 181, 12–20. Https://doi.org/10.1016/j.patrec.2025.xxxxxx
Ahmed, N., Rahman, A., & Lee, K. (2024). Multi-person pose tracking for real-time sports analytics. Ieee access, 12, 99812–99825. Https://doi.org/10.1109/access.2024.xxxxxx
González, D., Pérez, M., & Torres, L. (2023). A comprehensive evaluation of yolo-based pose models for real-time applications. Applied sciences, 13(22), 12234. Https://doi.org/10.3390/app132212234
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Yassawi Journal of Engineering Science
This work is licensed under a Creative Commons Attribution 4.0 International License.
