| Abstract [eng] |
This study investigates the application of artificial intelligence methods for evaluating the performance of upper limb rehabilitation exercises based on visual data. The relevance of the research is driven by the increasing demand for automated solutions in the healthcare domain, aiming to objectively and consistently assess patient movements. The study analyzes video recordings capturing upper limb movements during rehabilitation exercises. For human body skeletal detection and extraction of key joint points, the MediaPipe algorithm was employed, enabling the transformation of visual data into structured motion information. Based on the extracted skeletal data, motion parameters such as joint angles, movement trajectories, and stability metrics were calculated. During the research, machine learning models were developed and applied to evaluate the quality of exercise performance. The study examines the effectiveness of different models by comparing their ability to distinguish between correct and incorrect movements. The evaluation is performed using features extracted from video data. The obtained results demonstrate that artificial intelligence methods, combined with skeletal detection algorithms, can be effectively applied for automated assessment of rehabilitation exercises. The models are capable of identifying inaccurate or improperly executed movements and provide a basis for objective evaluation. It was determined that such solutions can enhance evaluation consistency and reduce the impact of subjectivity. The results of the study confirm the potential of artificial intelligence in the field of rehabilitation and provide a foundation for future research focused on more advanced motion analysis methods and real- time assessment systems. |
