Yoga Trainer for Beginners Via Machine Learning
Omar Tarek, Omar Magdy, Ayman Atia · 2021
Yoga is the practice for both mind and body as it has been proven scientifically on various occasions. Due to the modern advancements of technology, remote Yoga practice sessions have been increasing in popularity following the increase of demand for professional Yoga instructors. In order to tackle this problem, we proposed a system that uses machine learning techniques utilizing an ANN (Artificial Neural Network) model and a human pose tracking model to classify Yoga Hatha movements and detect Incorrect Yoga poses while providing real-time constructive feedback for practitioners to get them to maintain the correct posture for a specific Yoga Hatha pose. This system aims to enhance the learning experience and reduce the practice time for beginners while still retaining a versatile environment. In this paper, we managed to achieve a testing accuracy of 82.2% for our proposed model and were successfully able to reduce the average practice time by an average of 6.4 seconds when tested on 20 participants of different body features.