Graduate Studies


First Advisor

Mehmet C. Vuran

Date of this Version

Fall 12-4-2015


Aria, S. S. (2011) Near-Miss Fall Detection for Ironworkers Using Threshold-Based Features


A THESIS Presented to the Faculty of The Graduate College at the University of Nebraska In Partial Fulfillment of Requirements For the Degree of Master of Science, Major: Computer Science, Under the Supervision of Professor Mehmet C. Vuran. Lincoln, Nebraska: July, 2015

Copyright (c) 2015 Sepideh S. Aria


Computers are part of our everyday life. Today, more than ever, we have more processing power available to us. Human activity recognition is changing its face. Now desktop computers are no longer required in order to receive data and process them. Sensors, communication modules and processing power are available many handheld devices. One of those devices is smartphone. They have reasonable processing power and power management, and due to increasing application of smartphones in activity recognition, they are equipped with many sensors. In this research, a threshold-based techniques to detect iron-workers’ near-miss falls is proposed. A near-miss fall is a loss of balance, without any danger by itself but can be considered as an indicator. Construction industry and more specifically Iron-workers are responsible for the most fatal injuries due to falls. This raises the need to reduce number of fatalities and injuries due to fall in construction industry by introducing safety features, fall detection systems to provide medical assistance and if possible avoiding a fall in the first place. Near-miss falls are considered accident precursors. Detecting a near-miss fall can eliminate or reduce the risk of an actual fall. To achieve this, an algorithm to detect near-miss falls using a threshold-based techniques have been proposed. Threshold-based techniques are lightweight and computationally inexpensive, therefore they can easily be implemented on smartphones and moreover, they can perform as a real time near-miss fall detection algorithm. Compared to existing methods, this research uses a set of features to achieve best performance. These features are driven have been used before in detecting human activity. A combination of these features provided higher level of performance in the detection algorithm.

Advisor: Mehmet C. Vuran