Physiotherapeutic Aid using ML and AI

Abstract

In the recent year 2020 due to corona many people were not able to have appointments with their doctors and they were locked in their house. The younger generation can possibly maintain their health even if they do not go outside for days, but the older generation have to do regular exercise to maintain their health and well-being; prescribed by the doctors. It is not just because of corona that we have experienced this but we have also come to a situation where older people cannot visit their doctors. So, as a solution for these problems/challenges, we came up with an idea so that they can do their exercise and maintain their health without visiting their doctors or physiotherapist. Artificial Intelligence is technically defined as the development of technology which is used to perform technology operations which require involvement of human intelligence. Machine learning is one of the key components of artificial intelligence and it provides us with the ability of both supervised and unsupervised learning for training our model. AI technology today can be in different forms such as software programs as well as hardware interfaces to develop a system which is capable of learning from their own datasets. In our project AI with machine learning can be used for posture detection and then assessment of patients. We provided physiotherapy using AI and ML. We used normal running feed and we got good frame rates. These older generations have to do live exercise in front of the camera, our software will detect their pose/position whether they are doing their exercises correctly or not. By comparing the poses obtained from the live feed to the images or the videos obtained from the dataset. If the pose is not matched to the dataset, it is terminated or denied and deemed as wrong exercise. Our software is going to tell them where they went wrong. Now we have to decide a perfect algorithm/method to detect/estimate the pose with much higher accuracy. We are going to compare all the algorithms present till date regarding pose estimation and select the algorithms which give best accuracy.