Posture Correction using Human Pose Estimation

Abstract: Health and fitness have become a priority to a vast majority of the population ever since the COVID-19 pandemic has struck. People have resolved to eat healthy and exercise to ensure good health. The increase in the interest to exercise amidst public movement restriction led to the boom of the remote fitness industry. Various applications such Home Workouts, Nike Fit and Cult.fit prescribe a set of exercises to perform but do not assess how the exercise is being performed. Performing exercises using wrong postures might lead to injuries. Human pose estimation is a technique that uses Artificial Intelligence to detect the pose of a human and calculate angles between different parts of the body. Using human pose detection libraries like OpenPose, computer vision tools like OpenCV and geometry heuristics, appropriate posture correction suggestions can be given. This work aims to develop a solution that bridges the mentioned shortcoming of the remote fitness industry.

Keywords: Human Pose Estimation, Geometry Heuristics, Posture Correction, Remote Fitness Applications, Artificial intelligence, Computer Vision.