• Marina D. Korableva, Saint Petersburg Electrotechnical University "LETI", (Saint Petersburg, Russia)
• Yana A. Bekeneva, Saint Petersburg Electrotechnical University "LETI", (Saint Petersburg, Russia)

In recent years, applications providing users with physical activity programs have become very popular. This trend is driven by the accelerating pace of life, the lack of time to visit fitness centers, and the growing need to maintain a healthy level of physical activity. Digital platforms offer users a variety of training programs: specialized workouts for various muscle groups, cardio programs, and customized training plans tailored to the user's fitness level, goals, and limitations. The main advantage of such apps is the ability to train anywhere and anytime. However, a significant drawback of such apps is the lack of professional supervision to ensure proper exercise performance. Incorrect technique can not only reduce the effectiveness of workouts but also lead to injuries, such as strains, joint damage, and muscle damage. This is especially critical for beginners who have not yet mastered basic exercises and safety precautions. The goal of this work is to create an intelligent system for assessing the quality of exercise performance. The system is based on analyzing the video stream from the user's device camera, comparing the user's technique with a reference model, and automatically recognizing the trainer's reference technique. The training process is accompanied by visual feedback: indication of correct body position, detection of technique errors, and display of key body points (joints, limbs, etc.) in real time. This approach allows users to receive high-quality feedback immediately, significantly increasing the effectiveness of training and reducing the risk of injury due to improper technique. The system can provide recommendations for adjusting posture, tempo, and range of motion, making independent exercise safer and more effective. The study demonstrated the feasibility of using the developed system for static exercises, such as yoga. Further research and development focuses on dynamic exercises and significantly expanding the system's functionality and scope of application.

Fitness software, exercise, computer vision, image processing, pose detection, key points

2026-03-05