A Novel Fitness Tracker Model Using Machine Learning Algorithms

In today’s era of fitness tracking, modern devices use metrics like distance covered and heart rate during runs or walks, they fall short when it comes to accurately counting repetitions during weight exercises, particularly with barbells. This project aims to bridge this gap by introducing a cutting-edge machine-learning model tailored specifically for tracking barbell exercises such as Bench Press, Squats, Overhead Presses, Deadlifts, and Rows. This project aims to optimize the accuracy of exercise classification and repetition counting through comprehensive data processing, feature engineering, and predictive modelling. Various machine learning classifiers, including Neural Networks, Random Forests, K-Nearest Neighbors, Decision Trees, and Naive Bayes, are employed for this purpose. Notably, the Decision Tree algorithm achieved the highest accuracy score among the classifiers tested. Thus, leveraging its power, the final model is built upon the Decision Tree algorithm. The ultimate objectives are to enhance the accuracy of exercise identification and accurately count the number of repetitions performed using peak counting algorithms and low-pass filtering techniques applied to accelerometer data. This research contributes to the advancement of personalized fitness tracking, empowering individuals to make informed decisions based on their exercise data.