Exercise Recognition and Analysis for Improved Fitness Tracking Using Machine Learning

The growing popularity of fitness tracking and wearable technologies has paved the way for more sophisticated methods of monitoring physical activities. This paper presents an exercise recognition framework designed to enhance fitness tracking by accurately identifying and classifying various physical exercises. The proposed framework uses wearable sensors, specif-ically smartphones equipped with accelerometers and gyroscopes, to collect motion data during different exercises. These exercises, including walking, walking up, walking down, running, and arm rotation, were selected to represent a variety of movement types to ensure robust classification. Several machine learning algorithms are explored, including random forest, and decision tree for the classification task. The Random Forest model outperformed other classifiers.