Exercise Evaluation Using Time-of-Flight Image Data

Abstract

Many popular mass sporting events are based on the execution and evaluation of certain bodyweight or free weight exercises. The evaluation of these exercises is usually done by trainers or judges. Previous research has shown that movement classification is possible by observing the movements of body parts separately using UHF-RFID and a simple machine learning algorithm. This paper describes a possible way to automate this process using Time of Flight cameras. This is done in three steps, which are filtering, contour detection, and body part detection. The creation of an evaluation algorithm for the exercises was planned to be done with certified judges and coaches. However, this was not possible due to the COVID-19 pandemic. Still, preliminary results show that the positioning of the camera in relation to the athlete plays a key role in the evaluation accuracy. Also, automated repetition counting can be achieved in future versions of the proposed system.