Empowering Yoga Practice Through Pose Detection and Pose Accuracy Monitoring

This study introduces a method for evaluating yoga poses using pose detection technology to support self-learning. Leveraging the capabilities of pose detection technology, specifically the Mediapipe library in Python, the system accurately identifies and tracks key body landmarks in real-time. These landmarks serve as input features for Multinomial Logistic Regression, a statistical technique well-suited for modelling categorical outcomes such as different yoga poses. By utilizing Multinomial Logistic Regression on the coordinates of body landmarks, the system can effectively classify and assess the correctness of yoga postures. Moreover, the introduction of measurable metrics for pose accuracy allows for the development of interactive yoga games aimed at engaging youth and fostering enthusiasm. These metrics, derived from the probabilistic outputs of the logistic regression model, provide quantitative feedback on pose correctness, thereby facilitating self-assessment and motivating continued practice. Through the integration of pose detection technology and Multinomial Logistic Regression, this framework offers a promising solution for enhancing the effectiveness of yoga training and increasing engagement among practitioners.