Citation: Lee, S.; Koo, B.; Yang, S.;
Kim, J.; Nam, Y.; Kim, Y.
Fall-from-Height Detection Using
Deep Learning Based on IMU Sensor
Data for Accident Prevention at
Construction Sites. Sensors 2022, 22,
6107. https://doi.org/10.3390/
s22166107
Academic Editors: Pietro Picerno,
Andrea Mannini and Clive D’Souza
Received: 15 July 2022
Accepted: 8 August 2022
Published: 16 August 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Article
Fall-from-Height Detection Using Deep Learning Based on
IMU Sensor Data for Accident Prevention at Construction Sites
Seunghee Lee
†
, Bummo Koo
†
, Sumin Yang, Jongman Kim , Yejin Nam and Youngho Kim *
Department of Biomedical Engineering, Yonsei University, Wonju 26493, Korea
* Correspondence: younghokim@yonsei.ac.kr; Tel.: +82-33-760-2859
† These authors contributed equally to this work.
Abstract:
Workers at construction sites are prone to fall-from-height (FFH) accidents. The severity
of injury can be represented by the acceleration peak value. In the study, a risk prediction against
FFH was made using IMU sensor data for accident prevention at construction sites. Fifteen general
working movements (NF: non-fall), five low-hazard-fall movements, (LF), and five high-hazard-
FFH movements (HF) were performed by twenty male subjects and a dummy. An IMU sensor
was attached to the T7 position of the subject to measure the three-axis acceleration and angular
velocity. The peak acceleration value, calculated from the IMU data, was 4 g or less in general work
movements and 9 g or more in FFHs. Regression analysis was performed by applying various deep
learning models, including 1D-CNN, 2D-CNN, LSTM, and Conv-LSTM, to the risk prediction, and
then comparing them in terms of their mean absolute error (MAE) and mean squared error (MSE).
The FFH risk level was estimated based on the predicted peak acceleration. The Conv-LSTM model
trained by MAE showed the smallest error (MAE: 1.36 g), and the classification with the predicted
peak acceleration showed the best accuracy (97.6%). This study successfully predicted the FFH risk
levels and could be helpful to reduce fatal injuries at construction sites.
Keywords: fall-from-height; IMU sensor; deep learning; risk prediction
1. Introduction
Fall-from-height (FFH) accidents account for an extremely high proportion of accidents
at construction sites with a fairly high mortality rate. Choi et al. [
1
] conducted a comparative
analysis of accidents that occurred between 2011 and 2015 in three countries: the United
States, Korea, and China. Accidents were found to occur frequently at construction sites,
with the U.S. showing a 26% increase (from 781 to 985), while China and Korea showed a
28% decrease (2634 to 1891) and a 21% decrease (from 621 to 493), respectively. The average
mortality rate was the highest in Korea (17.9 persons), followed by the U.S. and China
(9.4 and 5.3 persons, respectively).
The Occupational Safety and Health Administration (OSHA) requires implement-
ing physical safety measures to reduce such accidents at construction sites [
2
]. Primary
protection measures include implementing guardrails, covers, safety nets, and physical
safety devices, while secondary protection measures include the use of a personal fall arrest
system (PFAS) whereby the impact of an FFH accident can be minimized [
3
]. A PFAS com-
prises a connector, full-body harness, lanyard, and rescue line, and may prevent a person
from falling when properly configured [
4
]. A PFAS cannot prevent FFHs but can effectively
avoid fatalities from FFHs [
5
]. Yang et al. [
2
] reported that fatalities due to losing balance
can be avoided when the PFAS is properly used; however, if a worker is suspended from a
PFAS for a prolonged time, there is a risk of suspension trauma, orthostatic intolerance,
or other serious injuries, and workers still sustain injuries due to the incompleteness of
a PFAS [
6
]. Furthermore, the effect of a PFAS is insignificant when a person falls from a
height below 15 ft, and accidents occur because workers do not properly wear PFAS due to
their inconvenience during work [7].
Sensors 2022, 22, 6107. https://doi.org/10.3390/s22166107 https://www.mdpi.com/journal/sensors
