Personalized Monitoring in Home Healthcare: An Assistive System for Post Hip Replacement Rehabilitation

The rehabilitation process for hip replacement surgery relies on supervised exercises recommended by medical authorities. However, limitations in therapist availability, budget constraints, and evaluation inconsistencies have prompted the need for a more accessible and user-friendly solution. In this paper, we propose a scalable, user-friendly, and cost-effective vision-based human action recognition system utilizing machine learning (ML) and 2D cameras. By providing personalized monitoring, our solution aims to address the limitations of traditional rehabilitation methods and support productive home-based healthcare. A key component of our work involves the use of deep learning (DL) method to align time-series exercise data, which ensures accurate analysis and assessment. Additionally, we introduce the concept of a Golden Feature, which plays a critical role in the framework by providing valuable insights into exercise execution and contributing to overall system accuracy. Furthermore, our framework goes beyond predicting exercise scores and focuses on predicting comments for partially successful cases using a multi-label ML model. This allows for a deeper understanding of the clinical reasons behind partial success, such as the patient's physical condition and their execution of the exercise. By identifying and analyzing these factors, our framework provides meaningful feedback and guidance to support effective rehabilitation. When evaluated on multiple exercises, the system achieved an accuracy level of 80% or higher on predicting execution score, and 72% on predicting the execution feedback.