AI-Based Yoga Pose Detection And Wellness System: A Systematic Review

Abstract

Yoga has transitioned from general fitness to clinical rehabilitation, including post-surgical recovery and chronic disease management. This shift increases the necessity for high-precision monitoring, as incorrect pose execution in a clinical context poses significant physical risks. While current automated pose recognition systems leveraging deep learning models achieve high accuracy rates, a systematic review of 45 recent studies reveals that existing research consistently lacks the integrated medical judgment required for safe clinical application. To address this "wellness integration gap," this research proposes an Artificial Intelligence (AI) based yoga pose detection and wellness system designed to bridge the gap between technical engineering and clinical safety. The methodology utilizes real-time computer vision, specifically employing YOLOv8 and MediaPipe frameworks, to evaluate practitioner alignment against validated safety standards. Preliminary results indicate that the system effectively identifies postural deviations that could lead to injury, providing a scalable solution for remote therapy. This study contributes a framework for integrating physiological safety parameters into deep learning models, ensuring that automated systems provide clinically sound guidance for users in a home or clinical setting.