Toward intelligent construction safety: A systematic review of IMU-based fall-from-height detection and a data-driven framework for proactive risk management

Abstract

Falls from height (FFH) remain the leading cause of fatalities in the construction industry globally, accounting for 36% of workplace deaths in the Australian construction sector between 2009 and 2019. Despite the growing adoption of safety technologies, existing reviews remain fragmented, frequently conflating FFH with the same level of falls and failing to examine inertial measurement unit (IMU) based systems within a structured, construction-specific framework. This study addresses these gaps through a systematic review guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) of IMU-based research on fall-from-height detection in construction environments. From 892 records retrieved across four academic databases (Web of Science, ScienceDirect, ASCE Library, Emerald) spanning 2015–2025, 20 peer-reviewed studies were included following rigorous screening and QUADAS-2 quality assessment. A bibliometric analysis identified construction safety, deep learning, and risk assessment as the dominant research themes across 10 geographic regions. Three critical gaps were identified: the absence of standardised pre-fall detection mechanisms, a lack of benchmarking frameworks for hybrid sensor systems, and a shortage of field-deployable solutions tailored to the complexity of real construction environments. To address these gaps, a data-driven modular framework is proposed, structured into three interconnected blocks: real-time posture monitoring via tri-axial IMU sensing; threshold-based fall event confirmation using empirically calibrated acceleration and angular velocity parameters; and automated, severity-graded emergency notification via IoT-enabled communication. The framework introduces a three-tier severity classification based on quantified sensor parameters, enabling proportionate emergency response rather than binary fall/no-fall detection. This review advances the shift toward intelligent, proactive, data-driven construction safety management. It establishes a foundation for future adaptive machine learning integration, construction-specific FFH dataset development, longitudinal field validation and Building Information Modeling (BIM)-enabled digital twin deployment.