8:00 Am – 9:30 Am Performance Testing

Measurement and evaluation of biometric device performance is critical to end users and consumers of these devices. Theoretical correlation models can be used to derive variance estimates of biometric performance metrics. In this talk, Dr. Schuckers will review performance metrics for biometric matching systems and discuss statistical methods for assessing these metrics. The focus of the talk will be on the traditional methods for biometric performance including false match and false non-match rates as well as system measures such as failure to enroll and failure to acquire.

What will be covered:

• Statistical methods for biometrics performance Comparison
• Evaluation of biometric systems Failure to Enroll, Failure to Acquire
• False Match Rate, False Non-match Rate, and Receiver Operating Characteristic Curves
• Determining the number of tests to run on a new technology

How you will benefit:

• Gain insight to possible metrics that are used in other areas but are not used in biometrics
• Expand your knowledge of methods for assessing a single metric and progress to methods for comparing metrics from different systems
• Discover how large a sample size should be in determining accuracy

Session Leader:

Dr. Michael Schuckers
Department of Mathematics, Computer Science and Statistics
Saint Lawrence University

9:45 Am – 11:15 Pm Multi-Level Fusion

Robust designs of biometric systems are proposed to address challenges using layered categorization, adaptive and pro-active training strategies, and data fusion. Some of these challenges are related to an expanded biometric processing space that involves less than perfect-sensors, missing information, and/or corrupt data. Scale space and recognition-by-parts support layered categorization, semi-supervised learning and transduction support adaptive and pro-active strategies, and boosting and non-linear mappings support multi-level and multi-layer data fusion. This discussion will introduce predictive quality-based fusion approaches to multi-modal and multi-algorithm data as well as present experimental results to illustrate novel and robust biometrics.

What will be covered:

• Biometrics and Forensics
• Information Theory and Statistical Learning
• Multi-Level Fusion: Feature, Score, and Decision
• Multi-Layer Fusion: Modality, Quality, and Method

How you will benefit:

• Gain insight into promising venues for future R&D
• Examine methods for voting and ensemble
• Enhance your knowledge of Performance Evaluation and Error Analysis

Session Leaders:

11:30 Am – 1:00 Pm Advances In Biometric Fusion (Lunch Will Be Served)

Biometric fusion, or multi-biometrics, refers to the consolidation of evidence presented by different sources of biometric information in order to improve the recognition accuracy of a biometric application. Multi-biometric systems combine the information presented by multiple biometric sensors (e.g. thermal and visible light cameras), algorithms (e.g., minutiae- and ridge-based fingerprint matchers), samples (e.g., frontal- and side-profile images of a person's face), units (e.g., left and right irises) or traits (e.g., face and fingerprint). Besides enhancing matching accuracy, these systems are expected to improve population coverage, deter spoofing and impart fault-tolerance to biometric applications. This workshop will present the fundamentals of multi-biometrics and discuss some of the recent advances in this field.

What will be covered:

• Introduction to the Fundamentals of Biometric Fusion
• Discussion on Multispectral Biometrics and Multi-camera Networks Description of Quality-based Fusion and Database Indexing

How you will benefit:

• Gain insight into the latest advances in multispectral biometrics
• Learn about the most up-to-date use of multi-camera networks applied to biometrics
• Gain an understanding of the requirements for quality-based fusion
• Find out about the most current capabilities of indexing multi-biometric databases.

Session Leader:

Dr. Arun Ross
Associate Professor
West Virginia University

1:15 Pm – 2:45 Pm Biometric System Interoperability (Lunch Will Be Served)

The proliferation of biometric systems in a networked environment has lead to designing of systems that are not homogenous. Instead of stand-alone and monolithic authentication architectures of the past, today’s networked world mixes disparate systems. This raises the issue of interoperability and its effect on performance. For a biometric system, interoperability can affect any of the subsystems of the biometric model: data acquisition, feature extraction, storage, matching and decision making. This workshop will discuss the challenges of interoperability at each stage, previous efforts conducted in the biometrics area with respect to interoperability, and current efforts in the academia and standardization communities to resolve interoperability challenges.

What will be covered:

• Effect of interoperability on biometric adoption
• Evaluation techniques to assess interoperability of different subsystems
• Current efforts in the area of interoperability

What you will learn about:

• Gain insight into how biometric system interoperability issues can effect performance
• Learn about ongoing efforts and advances in field of biometrics to tackle the challenge of interoperability

Session Leaders:

Dr. Shimon K. Modi
Director of Research, Biometrics Standards, Performance & Assurance Laboratory
Purdue University