iSmartYog: A Real Time Yoga Pose Recognition and Correction Feedback Model Using Deep Learning for Smart Healthcare

This paper presents the iSmartYog system model, a deep learning-based approach for yoga pose recognition and correction feedback generation. The system utilizes a hybrid deep learning model consisting of a convolutional neural network (CNN) and a Gated Recurrent Unit (GRU) to recognize 24 different yoga asanas from videos. A new video dataset of yoga pose dataset with participation of 31 individuals to consider the yoga pose as an action. Body keypoints detection from the practitioner's pose was carried out using the Mediapipe library to extract features of the poses. The CNN layer was used to extract spatial features from the keypoints of each frame, and the GRU was used to give temporal predictions. The iSmartYog system offers a unique approach to hybridizing CNN-GRU for yoga pose recognition, providing new insights into the field of smart healthcare and human action recognition. By accurately recognizing yoga poses, the system generates feedback to aid individuals in correcting their poses, thereby reducing the risk of injury and enhancing the benefits of yoga.