George, Joselin (2025) Development of Machine Learning Algorithms for the Automated Recognition of Turning Events during Walking using Wearable Sensors. PhD thesis, Victoria University.

Abstract

Turning is a common and biomechanically demanding component of everyday locomotion, yet it remains one of the least studied aspects of gait despite its strong association with fall risk. Directional changes require coordinated modulation of speed, body orientation, and centre of mass (CoM) trajectory, and these demands pose particular challenges for older adults and individuals with balance impairments. Reliable, prospective prediction of turning and accurate estimation of CoM-related stability measures are therefore essential for assistive technologies aimed at fall prevention and balance support. However, existing threshold-based algorithms perform inconsistently in real-world settings and are highly sensitive to variability in turning angle, speed, and individual movement strategies. This thesis investigates whether machine learning (ML) models applied to wearable inertial measurement units (IMUs) can (1) prospectively predict the onset of turning and (2) estimate short-term pelvis position as a practical proxy for CoM movement.

Search Google Scholar

Repository staff login