Fitness Activity Recognition Using Machine Learning Techniques
Mohamed Maher Ben Ismail, Yasser Al-Ali, Ouiem Bchir · Advances in systems analysis, software engineering, and high performance computing book series · 2024
In today's world of technological advancement, smartphones have become akin to a cybernetic implant for most people. Those phones that we carry everywhere are capable of doing things from taking pictures to tracking the number of steps we take. Fitness activities can be recognized using the smartphone's inertial sensors. Moreover, machine learning can be exploited to design predictive or descriptive models able to discover and/or predict fitness activity patterns with better accuracy. As different fitness activities exhibit different patterns and characteristics, classifying them would allow the users to better track their performance, endurance and calories burned. This chapter introduces a ML-based system that can accurately recognize the user's fitness activity. Specifically, deep learning models were investigated to automatically map the signals captured by the sensors of the user's smart device into some pre-defined classes of fitness activities. The designed models were implemented, validated, and tested using standard benchmark datasets and appropriate performance measures.