Sensor data-based workout classification model

Strength training is a vital part of a complete fitness regimen, but automated tracking systems for free weight exercises have not been fully developed. This study was conducted to explore the potential of context-aware applications in strength training by analyzing wristband accelerometer and gyroscope data collected from five participants performing major barbell exercises. Various supervised machine learning models were built and evaluated to replicate personal trainer tasks such as exercise classification, repetition counting, and explored preliminary detection of improper form. Various supervised machine learning models, including Naïve Bayes, Neural Networks, and Random Forest, were built and evaluated to replicate personal trainer tasks such as exercise classification, repetition counting, and improper form detection. Among these, Random Forest achieved the highest classification accuracy of 99.07%, demonstrating strong potential to overcome existing challenges in automatic exercise tracking. These findings suggest that machine learning can enable the development of smart, wearable-based digital personal trainers, offering real-time feedback, improving workout safety, and supporting personalized fitness programs.