Authors: Stanojević, Milan
Milenković, Ivan
Starčević, Dušan
Stanojević, Bogdana 
Title: Optimization of thresholds in serial multimodal biometric systems
Journal: 2016 6th International Conference on Computers Communications and Control, ICCCC 2016
First page: 140
Last page: 146
Conference: 6th International Conference on Computers Communications and Control, ICCCC 2016; Hotel PresidentBaile Felix-Oradea; Romania; 10 May 2016 through 14 May 2016
Issue Date: 20-Jun-2016
Rank: M33
ISBN: 978-1-509-01735-5
DOI: 10.1109/ICCCC.2016.7496752
Abstract: 
Multimodal biometric verification systems use information from several biometric modalities to verify an identity of a person. The false acceptance rate (FAR) and false rejection rate (FRR) are metrics generally used to measure the performance of such systems. In this paper we propose a novel approach to determine the upper and lower acceptance thresholds in sequential multimodal biometric matching, in such a way that the expected values of FAR and FRR for the entire system are minimized. We linearize locally the score distributions of both genuine users and impostors using the least squares method, and derive formulas for the approximated FAR and FRR for each matcher. Further, we aim to minimize both probabilities for entire processing chain. In order to find the best compromise between them, we analyze the efficient solutions to the associated bi-objective programming problem. The results of our experiments are also reported in the paper. They showed a good performance of the sequential multiple biometric matching system based on optimized thresholds comparing with the widely adopted parallel fusion multimodal biometric systems.
Keywords: multi-criteria optimization | multi-modal biometrics | sequential fusion
Publisher: IEEE
Project: Multimodal Biometry in Identity Management 
Optimization of Distributive and Reverse Flows in Logistic Systems 

Show full item record

SCOPUSTM   
Citations

2
checked on Jan 30, 2023

Page view(s)

30
checked on Jan 31, 2023

Google ScholarTM

Check

Altmetric

Altmetric


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.