MotionFusion: A Robust Ensemble Learning Framework for Accurate Sensor-Based Human Activity Recognition

Authors

Hussein K. Almulla, Electronic Computer Center, University of Anbar, Anbar, IraqFollow
Hussam J. Mohammed, Department of Computer Science - College of Computer Science and Information Technology, University of Anbar, Anbar, IraqFollow
Alaa S. Al-Waisy, Electronic Computer Center, University of Anbar, Anbar, IraqFollow
Shumoos Al-Fahdawi, Department of Computer Science - College of Computer Science and Information Technology, University of Anbar, Anbar, IraqFollow
Ahmed Adnan Had, Artificial Intelligence Sciences Department, College of Sciences, Al-Mustaqbal University, 51001, Babil, IraqFollow
Bourair AL-Attar, College of Medicine, University of Al-Ameed, Karbala, IraqFollow

Abstract

Human activity Recognition (HAR) has emerged as an important research area due to its potential applications in health, sport, and recreation. The widespread availability of smartphone sensors has facilitated data collection for HAR systems. Although machine learning and deep learning models have proven to be effective in detecting human activity from sensor data, their performance may be limited, this study proposes MotionFusion which is an ensemble learning model to increase HAR accuracy utilizing accelerometer and gyroscope data from a smartphone. By combining Histogram-Based Gradient Boosting, Random Forest, and Extra Trees models with a Support Vector Machine classifier and using feature selection techniques, we aim to classify six common functions accurately: sit, lie down, walk, stand, walk up and/down. A comprehensive evaluation was conducted using a well-established benchmark dataset, the UCI-HAR dataset that was collected from two sensors (accelerometer and gyroscope). The experimental results show that the MotionFusion model achieves more than 99% classification accuracy when using data from both sensors. Moreover, it obtained high accuracy rates of 99% and 98% when using accelerometer and gyroscope sensor data independently.

Recommended Citation

Almulla, Hussein K.; Mohammed, Hussam J.; Al-Waisy, Alaa S.; Al-Fahdawi, Shumoos; Had, Ahmed Adnan; and AL-Attar, Bourair (2025) "MotionFusion: A Robust Ensemble Learning Framework for Accurate Sensor-Based Human Activity Recognition," Iraqi Journal for Computer Science and Mathematics: Vol. 6: Iss. 3, Article 14.
DOI: https://doi.org/10.52866/2788-7421.1289
Available at: https://ijcsm.researchcommons.org/ijcsm/vol6/iss3/14