CITeR 2004

Projects:

Developing the PRESS ‐ Methodologies for Estimating Error Rates of Biometric Devices
Michael Schuckers (St. Lawrence University)

Scaling Analysis of Iris Codes using Large Deviations Approach
Natalia Schmid and Bojan Cukic (WVU)

Multibiometric Fusion at the Feature Extraction Level and Face Indexing
Arun Ross and Natalia Schmid (WVU)

Strategic Business Directions in Biometrics: Research with Vendors, Government and Corporate Buyers
Virginia Kleist and Richard Riley (WVU)

Acquisition and Understanding of Nonideal Iris Imagery
Lawrence Hornak, Xin Li, Gamal Fahmy, Natalia Schmid (WVU), Stephanie Schuckers (Clarkson) and A. Realini (WVU Eye Center)

Utilizing Soft Biometric Traits for User Recognition
Anil K. Jain and Sarat C. Dass (MSU)

Face Image Quality Assessment System
Gamal Fahmy, Hany Ammar (WVU) and Abdel‐Mottaleb (U of M)

On the Independence of Biometric Modalities
Arun Ross (WVU) and Anil K. Jain (MSU)

Geometric Coding and Processing of Biometric Images (Phase II)
Xin Li (WVU)



Summaries:



Developing the PRESS ‐ Methodologies for Estimating Error Rates of Biometric Devices

Michael Schuckers (St. Lawrence University)

The next phase of research on the Program for Rate Estimation and Statistical Summaries (PRESS) will add capabilities to plot ROC curves (both on the original scale and on the log scale), to determine the appropriateness of the Beta‐binomial distribution for a given data set, and to estimate the EER. In addition, we plan to do an algorithmic analysis of the software in order to make PRESS more efficient.



Scaling Analysis of Iris Codes using Large Deviations Approach

Natalia Schmid and Bojan Cukic (WVU)

This project explores the scaling and prediction of the capacity of iris biometric systems. We plan to apply the large deviations approach to asymptotically predict the performance of these biometric systems and use the derived results to evaluate performance limits of large‐scale identification systems based on iris.



Multibiometric Fusion at the Feature Extraction Level and Face Indexing

Arun Ross and Natalia Schmid (WVU)

This project will develop techniques to perform multibiometric fusion at the feature extraction level. If necessary, feature selection methods will be used to reduce the dimension of the fused feature set. We will also develop an efficient face‐indexing (classification) scheme that will rely on the geometric attributes of the human face to narrow the search to a limited number of faces (classes) in the database.



Strategic Business Directions in Biometrics: Research with Vendors, Government and Corporate Buyers

Virginia Kleist and Richard Riley (WVU)

This proposed research will design, plot, and collect a large scale data set related to strategic business issues in the biometrics industry. We will develop an in‐depth research tool aimed at surveying perceptions, best practices, and directions from biometric vendors, government, and private buyers. The results should provide a detailed understanding of potential risks and solutions for vendors, government buyers and other users.


Acquisition and Understanding of Nonideal Iris Imagery

Lawrence Hornak, Xin Li, Gamal Fahmy, Natalia Schmid (WVU), Stephanie Schuckers (Clarkson) and A. Realini (WVU Eye Center)

This work will revisit the fundamental information content in the iris and its variability in order to explore means of iris classification and matching through nonideal iris imagery. Automated location of nonideally oriented irises in facial images will be investigated and achievement of imagery of the prescribed quality from lower quality video frame sequences will be explored using deblurring and super resolution techniques.



Utilizing Soft Biometric Traits for User Recognition

Anil K. Jain and Sarat C. Dass (MSU)

The performance of a biometric system can be improved by utilizing ancillary information about the users such as their height, weight, age, gender, ethnicity, and eye color, referred to as soft biometric traits. We will (i) design a prototype recognition system that automatically extracts these soft biometric traits along with the primary biometric (e.g., fingerprint), (ii) develop a mathematical framework for integrating the soft biometric information with the primary biometric system for improving the recognition accuracy, and (iii) derive variable weights for the soft biometric traits based on their distinctiveness and permanence. Initial experiments show that the use of additional user information like gender, ethnicity, and height improves the recognition performance of a fingerprint system by ≈ 5%.



Face Image Quality Assessment System

Gamal Fahmy, Hany Ammar (WVU) and Abdel‐Mottaleb (U of M)

Poor imaging from biometric systems contribute to the difficulty in detecting features from the image or due to the poor quality of the detected features. While fingerprint identification systems overcome this problem by passing the finger print through a quality assessment stage, there are no such techniques for assessing the quality of images for face recognition to determine whether the image is suitable for recognition or not. In this project we develop a prototype that automatically measures the quality of face images for most well known face recognition algorithms.



On the Independence of Biometric Modalities

Arun Ross (WVU) and Anil K. Jain (MSU)

We investigate the degree of correlation between multiple traits in a multibiometric system. It is generally agreed that the performance gain in a multiple classifier system (MCS) is directly related to the extent of independence between constituent classifiers. However, in the context of multibiometrics, no systematic study has been conducted to ascertain the common assumption of independence between multiple modalities. The goal of this project is to develop a framework to conduct such a study and to derive the degree of dependence between various biometric modalities.


Geometric Coding and Processing of Biometric Images (Phase II)

Xin Li (WVU)

The objective of this project is to continue our investigation of geometric coding and processing of biometric images and demonstrate the potential of signal processing techniques for improving the performance and ergonomics of biometric systems. The research plan for phase II shifts from a single copy case to a multi‐copy case, i.e., how to exploit the concept of diversity to improve the coding efficiency and subjective quality of biometric images. Specifically, our study consists of two parts: image registration, which addresses the problem of resolving the geometric relationship among multiple copies and image manipulation, which intelligently processes the image information based on the discovered geometric relationship.

CITeR: Center for Identification Technology Research

Research Overview

2014 Projects

2013 Projects

2012 Projects

2011 Projects

2010 Projects

2009 Projects

2008 Projects

2007 Projects

2006 Projects

2005 Projects

2004 Projects

2003 Projects

2002 Projects



CITeR Contact Information:
Clarkson Phone: 315-268-6536
WVU Phone: 304-293-1455
Arizona Phone: 520-621-5818
UB Phone: 716-645-1558
More Information
National Science Foundation