Analyzing Training Exercises with Mathematical Algorithms and Convolutional Neural Network
Szabolcs Róbert Bakos, Miklós Sipos · 2023
This paper describes an application, which aims to help people who are doing training exercises at home. The system takes a video as input and provides feedback about the mistakes that were made during the execution of the exercise by using Convolutional Neural Network (CNN) to determine the limb positions and mathematical calculations to find the angles between them. From this information the system can decide with high accuracy, if the range of motion was correct during the following exercises: shoulder press, single arm biceps curl, single arm bent over row, lunge non-alternating version, single arm triceps extension.