Recognizing Physiotherapy Exercises using Machine Learning
Jayneel Kanungo, Dhanashree Mane, Pratik Mahajan, Saumya Salian · 2023
Physiotherapy, which is also referred to as physical therapy (PT) is a healthcare profession that is primarily concerned with correcting or physiotherapy aims to improve mobility and functionality while addressing impairments through the use of mechanical force and movements, manual therapy, exercise therapy, and electrotherapy. However, the effectiveness of the treatment relies on the patient's adherence to the recommended exercises. Failure to comply with the prescribed exercises may lead to unsuccessful outcomes. The examination of human posture finds various uses in sports and medical science, such as monitoring patients, analyzing lifestyle, and caring for the elderly. However, computer vision techniques, which have been predominantly utilized in this area, have their limitations when it comes to delivering real-time solutions. Thus, Machine Learning based solution are being planned and used for the human posture recognition and detection. For individuals who are new to specialized exercises, it can be challenging to identifying the areas where they are making mistakes without any external help or guidance. To address this, we propose a Physiotherapy exercise assessment system that utilizes pose detection to support self-learning of yoga. The model has an accuracy rate of 96% for pose detection. The system first detects a Physiotherapy exercise using multi-part detection solely through a PC camera. It then calculates the difference in specific body angles between the user's pose and that of the instructor. If the discrepancy surpasses a set threshold, the system differentiates between the feedback of the instructor and the user and offers a corrective suggestion.