Mind Mat: Personalized Yoga Recommendation System for Wellbeing and Wellness Using Machine Learning Techniques

Yoga is a truly effective practice in the realms of physical, flexibility, and mental wellness. Unfortunately, traditional yoga instruction tends to lack feedback designed for students, thus it is difficult for practitioners to achieve good form which can lead to ineffective practice or injury. In this paper, we present MindMat: an AI-based personalized yoga recommendation system aimed at providing real-time custom posture correction and dynamic feedback. MindMat utilizes a user-specific health profile, real-time pose estimation score, and the pose database, organized with information on contraindications, to generate personalized yoga practice sessions. Utilizing MediaPipe BlazePose for accurate skeletal tracking and reinforcement learning to create personalized recommendations, MindMat can continue to improve its accuracy and effectiveness as it is used. The system tracks as user practices, evaluates their performance, detects mis- alignments, and provides instant feedback to improve user performance and ensure the practice is safe. Experimental results demonstrated MindMat's ability to improve pose accuracy and flexibility and decrease risk of injury. With its intelligent and adaptive feedback process, MindMat offers a novel solution for personalized wellness in a way that would allow for all practitioners of yoga to be effective and injury-free. Overall MindMat is a promising personalized yoga practice system for both novice and experienced practitioners.