Egocentric camera-based fall detection system using rotation, motion, HOG and LBP

Bouakaz, Hichem (2020) Egocentric camera-based fall detection system using rotation, motion, HOG and LBP. Master's Thesis / Essay, Computing Science.

Preview	Text mCS_2020_BouakazH.pdf Download (3MB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (125kB)

Abstract

Elderly people are the fastest-growing segment of the population in the Netherlands and the world. According to the World Health Organization (WHO), falls - after road accidents- are the second leading cause of unintentional injury deaths worldwide. This work presents a fall detection system, based on egocentric cameras, to assist the living of the elderly both in indoor and outdoor environments. During this research, a dataset containing 1459 pre-recorded video sequences of falls and non-fall activities was used. The videos were recorded using one camera mounted on the waist and one on the neck. Several methods to detect falls were proposed, namely LBP, HOG, video rotation and camera motion derived from optical flow and Random Forests algorithm is used for classification. Each of the proposed methods was tested and evaluated both separately and together with the other methods. To determine the most suited system for a fall detection which can be employed in real-time to assist the living of the elderly, a comparison is made in terms of reliability and efficiency. The best results in terms of reliability were achieved by combining rotation, LBP and motion with 98.3% mean AUC and 92.6% mean accuracy during cross-validation, 98.8% AUC and 91.8% accuracy on the test set for binary classification, 95.1% micro AUC and 94.7% macro AUC during cross-validation and 95.2% micro AUC and 94.9% macro AUC on the test set. However, considering efficiency LBP is slow, therefore our suggest

Item Type:	Thesis (Master's Thesis / Essay)
Supervisor name:	Azzopardi, G. and Talavera Martinez, E.
Degree programme:	Computing Science
Thesis type:	Master's Thesis / Essay
Language:	English
Date Deposited:	02 Nov 2020 08:41
Last Modified:	02 Nov 2020 08:41
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/23552

Actions (login required)

View Item