Motion: A Mobile Application for Yoga Pose Accuracy and Consistency
Adrian Paul Reyes, Mel Jefferson Gabutan, Ada Pauline Villacarlos, Cherry Lyn C. Sta. Romana, Chris Jordan G. Aliac · 2023
Pre-recorded yoga tutorials have become more prevalent for the benefits it offers. The drawback of learning yoga using these tutorials is that practitioners do not receive feedback on their execution, which could lead to severe injuries. To eliminate this problem, software applications identify and correct improper executions; however, they do not provide a criteria-based evaluation. To address this issue, this study presents Motion, a mobile application that evaluates a yoga practitioner’s pose accuracy and consistency in real-time and provides a performance history of previous executions. The application utilizes MediaPipe Pose and ML Kit to perform Human Pose Estimation and Machine Learning to track and identify yoga execution. Angle data of essential body parts for each frame of a practitioner’s execution are stored and used to evaluate pose accuracy and consistency. The evaluation results are then stored in an SQLite database and displayed in the performance history. The application’s pose classifier achieved a 99.33% accuracy, demonstrating its effectiveness in the real-time classification of yoga poses, and achieved an average modified PSSUQ score of 1.99, indicating high usability. With these implications and Motion’s ability to provide real-time evaluation and progress tracking, the application can be used as a valuable tool in the fields of yoga, health, and education.