Random Forest - Seated Posture Recognition in Trunk Orthosis

Abstract

This research investigates how combining sensor data with machine learning approaches can detect movements of seated individuals by analyzing information collected from inertial measurement units (IMUs) and electromyography (EMG) recordings. The IMU dataset contains tri-axial accelerometer data, timestamps, and user information, while the EMG dataset includes multi-channel muscle activity readings, both labeled with specific movement activities. To improve classification accuracy and robustness, we implement Random Forest as the primary machine learning model. By integrating IMU and EMG data through sensor fusion, our approach enhances movement detection reliability—critical for developing assistive technologies like trunk orthosis systems. The results demonstrate the effectiveness of Random Forest in accurately predicting seated movements, offering insights into its potential for biomechanical applications and rehabilitation technologies.