ASMS_CSC Archives

MS Student


Options: Use Classic View

Use Monospaced Font
Show Text Part by Default
Show All Mail Headers

Topic: [<< First] [< Prev] [Next >] [Last >>]

Print Reply
Tammie T Dudley <[log in to unmask]>
Tue, 1 Mar 2016 20:08:18 +0000
text/plain (15 lines)
Departmental Colloquium at Conference Room 755, 1:30-2:30 pm, March 4th, 2016
Title: "The Latest Trends in Fingerprint Technology".	
Dr. Emanuela Marasco
Postdoc Fellow
Department of Computer Science at the University of North Carolina Charlotte

Abstract: Gender is a valuable demographic characteristic that can aid the recognition performance of primary biometric identifiers. However, predicting gender from fingerprints has been currently a difficult problem, and often the process is not fully automated. Most existing approaches relate gender to a direct measure of ridge density, which may not be robust to factors such as the finger size or image degradation. Features generated through the Discrete Wavelet Transform seem to be the most promising ones; however, they have only been jointly exploited with high-dimensional descriptors such as Principal Component Analysis and Singular Value Decomposition. We discuss a highly accurate algorithm in a low-dimensional space which exploits frequency and matchability information. Experiments are carried out on rolled fingerprints pertaining to 256 males and 238 females collected using three different optical sensors. An accuracy of 97.12% is achieved with a Support Vector Machine classifier. Results show also that the proposed strategy is highly interoperable.

High accurate matching involving fingerprint images acquired using different sensors is currently challenging. Diversity in image resolution, scanning area, arrangements of sensing elements induce variations in the acquired images. Such variations can impact the nature and quality of the features extracted from these images, and subsequently cross-sensor matching performance. This is true even when dealing with fingerprint sensors of the same sensing technology (e.g. inter-optical). We discuss a method that minimizes the impact of low interoperability between optical fingerprint sensors. Experimental results confirm that the proposed approach is able to reduce cross-device match error rates by a significant margin.

Recent research has shown that, commercial fingerprint scanners can be deceived by presenting well duplicated fingerprints. Artificial fingerprints are usually realized using materials which can be scanned (e.g., play-doh, silicone, gelatin, etc.) characterized by a moisture-based texture. In parallel, countermeasures to discriminate between human live fingerprints and spoof artifacts have been developed; hardware- and software-based anti-spoofing approaches have been proposed. In this talk, we review the state-of-the-art techniques involving the employment of successful spoof attacks and the most efficient spoof detection strategies. 

Bio: Emanuela Marasco is currently an Adjunct Professor and a Post-Doctoral Researcher at the Video & Image Analysis Lab of the Department of Computer Science at the University of North Carolina Charlotte (UNC-C).  From February 2011 to January 2015 she was a post-doctoral Associate Researcher at Lane Department of Computer Science and Electrical Engineering, West Virginia University and at the Center for Identification Technology (CiTeR-NSF). She received a five-year degree (Bachelor and M.Sc.) in computer engineering, in March 2006, and a PhD in Computer and Automation Engineering, in December 2010, both at the University of Naples Federico II (Italy) with maximum score. Her research interests focus on Artificial Intelligence, Pattern Recognition, Image Processing, Machine Learning and Biometrics. Specifically, her focus is on fingerprint liveness detection algorithms, fusion schemes to combine multiple biometric modalities, algorithms for estimating age and gender from fingerprints, schemes to enhance interoperability between fingerprint optical sensors, and development of adaptive signal processing strategies for enhancing DNA profiles obtained from degraded DNA samples.