Real-Time Yoga Pose Recognition and Correction Feedback System Using Deep Learning and Computer Vision

Yoga pose detection is a rapidly evolving field that integrates advanced technology with wellness, addressing the need for accurate and adaptable systems for diverse user groups. This paper proposes a novel framework that combines RGB images, depth maps, and skeletal joint data to enhance multi-modal feature extraction and improve spatial-temporal analysis for tasks such as action recognition and human activity modeling. The framework overcomes challenges such as pose variations and occlusions through techniques like data augmentation, model optimization, and robust performance evaluation. Experimental results demonstrate the system’s high accuracy of 91.7%, successfully detecting six distinct yoga pose sequences in real time and providing self-correction feedback to users, thereby improving performance. The practical applications of this approach are significant, with potential use cases in yoga studios, fitness centers, and home environments, promoting scalable, precise, and globally accessible yoga practices.