Skip navigation

Design and analysis of a biometric access control system using an electronic olfactory device to identify human odour characteristics

Design and analysis of a biometric access control system using an electronic olfactory device to identify human odour characteristics

McMillan, Stephen (2000) Design and analysis of a biometric access control system using an electronic olfactory device to identify human odour characteristics. PhD thesis, University of Greenwich.

[img]
Preview
PDF
Stephen_McMillan_2000.pdf - Published Version
Available under License Creative Commons Attribution Non-commercial No Derivatives.

Download (14MB)

Abstract

The use of an electronic olfactory device, termed an electronic 'nose', was investigated for the detection of unique human odour characteristics. The detection of these unique odours was applied to the field of biometrics for access control, where a human's unique characteristics were used to authenticate a user of an access control system. An electronic odour sensing device was designed and constructed using an array of conducting polymer gas sensors in order to facilitate the regular screening of a group of human subjects over a period of six weeks.

A static sampling method was used to measure odour levels from human hands, which were found to contain a reliable source of human odour. Human odour levels were low so dynamic sampling proved to be unsuitable for this application due to the dilution of the odour mixture. Feature analysis results revealed that the features of adsorption and desorption gradient contained discriminatory information in addition to the commonly used maximum divergence. Pattern recognition revealed that neural network architectures produced superior results when compared to statistical methods as a result of their ability to model the non-linearities in the data set. The highest recognition rate was 73% which was produced using a Multi-Layer Perceptron (MLP) neural network compared to 63% obtained using the best statistical method of Parzen windows. The majority of the recognition error was caused by a minority of the humans. Analysis of sensor data revealed that only 30% of the sensor array were contributing discriminatory information so it was deduced that performance would undoubtedly improve if a full array of effective sensors were available.

Exploratory data analysis revealed that human odour changed from day to day and often an increasing divergence with time was observed. A time-adaptive method was devised which increased the recognition to 89%, but was still too low for use as a biometric recognition device. However, use as a verification device demonstrated acceptable levels of performance but resulted in high levels of user frustration caused by a high proportion of users being falsely rejected. This work demonstrated that an olfactory based biometric access control system could be a realistic proposition but requires further work, especially in the areas of sensor development and unique human odour research, before an operational system could be produced.

Item Type: Thesis (PhD)
Additional Information: uk.bl.ethos.638065
Uncontrolled Keywords: machine olfaction, e-nose, biometric access control system,
Subjects: T Technology > T Technology (General)
Pre-2014 Departments: School of Engineering
School of Engineering > Department of Electrical, Electronic and Computer Engineering
Last Modified: 14 Oct 2016 09:31
URI: http://gala.gre.ac.uk/id/eprint/13136

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year

View more statistics