Chatbot for Remote Physical Knee Rehabilitation with Exercise Monitoring based on Machine Learning

A machine learning-based chatbot has been developed for remote knee rehabilitation, incorporating exercise monitoring and feedback through video analysis. The MediaPipe framework is utilized for recognizing poses and movements of patients. Forty features have been constructed for machine learning algorithms to determine exercise characteristics. The model was trained using multi-class logistic regression and achieved an accuracy of 98% on test data. Video tests evaluated by the machine learning model were used for objective progress monitoring. The chatbot provides patients with feedback and objective progress indicators, facilitating controlled rehabilitation.