Yoga is widely recognized for its physical, mental, and therapeutic benefits, particularly among individuals seeking stress relief, improved flexibility, and mind–body balance. However, traditional yoga instruction may be challenging for neurodivergent individuals or those with sensory sensitivities, who can find in-person settings overwhelming. This study introduces YogaVR, an AI-integrated virtual reality (VR) system designed to deliver real-time posture correction and guided yoga sessions in a calm and immersive environment. Built in Unity for the Meta Quest 3S, the system uses PoseNet for pose estimation via an external camera mounted on a Raspberry Pi 4-powered four-wheel drive (4WD) robot that autonomously tracks the user to enable multi-angle monitoring. An AI voice assistant, powered by OpenAI models, offers corrective feedback and personalized recommendations on posture, diet, and exercises. WebXR and WebRTC are employed to stream external camera feeds for real-time analysis. The system provides a supportive alternative to conventional yoga classes, helping users build confidence and consistency in practice without social pressure. A focus group evaluation with 30 participants reported a high average user acceptance score of 8.1 out of 10 with a Standard Deviation of 1.9, indicating strong usability and positive reception.
Kumar et al. (Tue,) studied this question.