Research on Intelligent Bodybuilding System Based on Machine Learning

Abstract

In recent years, people’s health is facing many challenges as their workload is increasing and their lives are becoming more and more stressful. In this context, healthy living has become a topic of concern and more and more people are choosing to promote their bodies through fitness. To address these existing problems in action recognition research, this paper designs and implements a machine learning-based intelligent fitness system to monitor three important parameters in physical activity: the type of action, the number of actions, and the period of action. Through the action recognition algorithm and the period calculation method, the three important parameters of action type, number of actions, and action period are calculated to generate a more comprehensive description of the limb actions. Experiments are conducted to show that the proposed deep neural network learns well on small datasets, achieving 97.61% action recognition accuracy and SVM achieving over 96% recognition accuracy.