Providing informative feedback in a low-cost rehabilitation system using machine learning

Abstract

Rehabilitation is a core process in helping people recover from a wide range of health issues, including injuries and diseases. Although advancements in technology and the use of artificial intelligence have facilitated the development of tools to aid in rehabilitation processes, there is a lack of low-cost solutions that patients, without requiring advanced care, can use at home. In this work, we propose a low-cost intelligent system for lower limb rehabilitation that uses machine learning to provide informative feedback to users. Compared to existing solutions, our system offers the advantage of real-time feedback, informing patients whether they are performing exercises correctly. It also suggests posture corrections to prevent injuries and accelerate the recovery process. Moreover, our system can be used at home on a smartphone, tablet, or personal computer, and does not require patients to purchase additional devices, which is a significant benefit. The system includes four exercises: Squat, Romanian Deadlift, Glute Bridge, and Donkey Kick. Validation tests with end-users reinforced the usability of this system and confirmed the importance of real-time feedback. The results were also useful for identifying areas for improvement, particularly with the Squat exercise, which is among the more challenging exercises to perform correctly.