A Case Study for the Automatic Supervision of Body-Weight Exercises: The Squat
Paolo Aliprandi, Letizia Girardi, Giulia Martinelli, Francesco G. B. De Natale, Niccolò Bisagno, Nicola Conci · 2023
Evaluating body-weight exercises is typically per-formed by expert evaluators who assess the correctness of movements. In this study, we propose a novel method that enables automated assessment of exercise quality, with a specific focus on squats. The assessment is conducted by automatically collecting and measuring the body's position relative to a target sample provided by an expert. Our system comprises a consumer-grade depth camera and a costly but precise motion capture system, allowing for the retrieval of 3D joint positions. During the execution of the exercise, we capture the trajectories of each joint, which are then synthesized into a score. The key indicators to score a repetition of the exercise are the positions of the head and pelvis, as well as the angle of the spine. We show how the consumer-grade camera can approximate the measurement results of the motion capture system, thus indicating the possibility of building a good supervision system with limited resources. This system can be used both for training to evaluate exercise quality and prevent injuries, as well as for rehabilitation, by quantitatively measuring a patient's progress.