BTAS 2015 will feature the following Special Sessions. The feasibility of a Special Session will be based on the number of accepted papers related to the topic of the Special Session.

Important: Authors who would like their papers to be considered in one or more Special Sessions, must select the appropriate Primary and/or Secondary Subject Areas when submitting their paper.

  1. 4D’s of Machine Learning for Biometrics

  2. Biometrics & Forensics: A Path Forward

  3. Biometric Data Quality Assessment

  4. Mobile Biometrics

  5. Reproducible Research in Biometrics

4D’s of Machine Learning for Biometrics

Deep Learning, Dictionary Learning, Domain Adaptation, Distance Metric Learning

[Organizers: Mayank Vatsa, IIIT-Delhi, India; Kevin Bowyer, University of Notre Dame, USA]

With the availability of inexpensive biometric sensors, computing power, and memory, it is becoming increasingly clear that biometrics technology will have broader usage, and therefore also broader scope of future research in addressing newer challenges and pushing the boundaries. If we perceive biometrics as a fundamental problem in science and engineering with broad economic and scientific impact, then designing efficient algorithms and systems will require a multidisciplinary effort in signal processing, pattern recognition, machine learning, sensor design, embedded systems, and information fusion. Recent advances in machine learning have seen widespread development of algorithms in four specific areas: deep learning, dictionary learning, domain adaptation, and distance metric learning. As a consumer of these 4-D paradigms, the likelihood of exploring new avenues of research is immense. This special session focuses on bringing together researchers and practitioners in biometrics and machine learning to showcase the progress, algorithms, and  applications of deep learning, dictionary learning, domain adaptation, and distance metric learning in biometrics. Topics appropriate for this special session include (but not necessarily limited to):

  1. 1.Novel feature representation using deep learning, dictionary learning for face, fingerprint, ocular, and/or other biometric modalities

  2. 2.Novel algorithms for heterogeneous biometric recognition such as (a) matching visible images to near-infrared images, (b) matching cross-resolution images, and (c) matching sketches with digital face images

  3. 3.Novel algorithms for transferring knowledge from one biometric domain to another, including transfer learning and other semi-supervised learning algorithms

  4. 4.Novel distance metric learning algorithms for biometrics modalities

  5. 5.Applications of these paradigms in biometric systems

Biometrics & Forensics: A Path Forward

[Organizers: Anil K. Jain, Michigan State University, USA; Christophe Champod, University of Lausanne, Switzerland]

Forensic science – and especially topics within it that are related to human traits such as fingerprint or palmprint identification, handwriting and signature attribution, facial and gait recognition, speaker verification, etc. – is currently under scrutiny. It has been strongly suggested that forensic science has not benefited from structured and systematic research allowing characterization and validation of evidence with the appropriate quantitative measures. On the other hand, progress in biometric research has advanced the state of the art to the point that biometric systems can now handle automatic recognition tasks in less and less controlled environments. Forensic scientists deal often with low quality and small quantity of information which are currently not very amenable to automate. There is hence a fertile ground for interdisciplinary research and mutual benefit. Bringing together forensic experts with their holistic expertise and knowledge of the issues and biometric experts versed into signal processing and machine learning gives a new ground for cross-fertilization and improvement. For this special session, we invite original contributions that will present, consolidate or expand the boundaries of research between forensic science and biometrics.

Biometric Data Quality Assessment

[Organizers: Christophe Charrier, University of Caen Basse Normandie, France; Christophe Rosenberger, ENSICAEN, France]

Performance of biometric systems is dependent on the quality of the acquired input samples. If quality can be improved, either by sensor design, by user interface design, or by standards compliance, better performance can be realized. For those aspects of quality that cannot be designed-in, an ability to analyze the quality of a live sample is needed. This is useful primarily in initiating the reacquisition from a user, but also for the real-time selection of the best sample, and the selective invocation of different processing methods. It is the key component in quality assurance management, and because quality algorithms often embed the same image (or signal) analyses needed to assess conformance to underlying data interchange standards, they can be used in automated image screening applications.

The goal of this special session is to bring together researchers and practitioners working in the area of biometric quality. We are soliciting original contributions, which address a wide range of theoretical and practical issues related to systems based on fingerprint, iris, face, voice, hand, handwriting, gait and other modalities including soft biometrics, biometric fusion and mobile biometrics, but not limited to

• Biometric performance measurement

• Biometric sample quality

• Validation approaches to quality assessment of biometric data

• Template selection and update

• Multispectral biometric data quality

• Combination of quality effects

• Quality measures and their relationship to recognition performance

• Minutiae Interoperability exchange

• Quality of 3D face acquisition

• Quality of biometric databases

• Quality fusion of multi-modalities systems

• Study on quality assessment of biometric sensors

• Spoof detection and quality

  1. Impact of capture context on biometric quality

  2. Size and quality inter-relation

Mobile Biometrics

[Organizers: Thirimachos Bourlai, West Virginia University, USA; Dimitris Metaxas, Rutgers University, USA; Mayank Vatsa, IIIT-Delhi, India]

Smartphones and mobile devices are now an integral component of day to day life. With the advent of mobile technology, the quality of mobile devices including sensors, processing power, and user interaction has improved, which has instigated research in biometrics to build mobile-based authentication-related applications. Apple’s TouchID, Samsung’s fingerprint technology and DeltaID’s ActiveIRIS are some examples of successful amalgamation of mobile technology with biometrics. Recent market surveys predict that the global mobile biometrics market is set to expand with an annual growth rate of over 150%. While there are some successful applications, there are several important aspects that require further research such as mobile device centric algorithms, circumvention analysis, permanence analysis, facial movement and expressions and user acceptability. This special session focuses on both theoretical contributions and applications of mobile-based biometrics. Topics of interest include, but are not limited to:

Biometric authentication on mobile devices using face, fingerprint, iris and other modalities such as facial movement and expressions

Novel approaches for mobile biometric authentication such as swipe pattern recognition

Algorithms for mobile data quality analysis

Security analysis and algorithms of mobile biometric systems

Spoofing and anti-spoofing for mobile biometrics

Low cost mobile biometrics hardware design

Human computer interaction (HCI) aspects of designing mobile biometric systems (e.g. user interaction, human performance evaluation, user centric studies)

Real world large scale applications

Reproducible Research in Biometrics

[Organizers: Sebastien Marcel, Idiap Research Institute, Switzerland; Andre Anjos, Idiap Research Institute, Switzerland; Mauro Barni, University of Siena, Italy]

Biometrics research is an interdisciplinary field that combines expertise from several research areas. Starting with signal and image processing to perform preprocessing and feature extraction, passing by the field of machine learning with means of subspace projections or data modeling, to the field of information theory, including pattern recognition and distance computations, biometrics research might require a lot of different expertises. Additionally, to make results comparable, a proper implementation of the required evaluation protocols of biometric databases and merit figures need to be provided. This makes biometrics a particularly difficult research topic, especially when comparable results should be provided. More often than not, biometric algorithms are tested only on a few of the available databases. Research results can not be reproduced since because researchers themselves do not publish all of the meta-parameters of their algorithms.

One attempt to tackle issues surrounding research irreproducibility is the concept of Reproducible Research (, RR). An RR paper is comprised of several aspects [1], which makes it possible and easy to exactly reproduce experiments:

(a) a research publication that describes the work in all relevant details,

(b) the source code to reproduce all results,

(c) the data required to reproduce the results,

  1. (d)instructions how to apply the code on the data to replicate the results on the paper.

A strong reason for providing RR, besides easing research itself and providing for longer term re-use, is the visibility of the resulting scientific publications. As [1] showed, the average number of citations for papers that provide source-code in IEEE Transactions on Image Processing (IEEE-TIP) is seven times higher that of papers that do not.

There have been attempts to foment reproducibility of research results in the biometric community with the release of public software and datasets. Various biometric communities organize open challenges, for which web-based solutions for data access and result posting are particularly attractive. Some dataset providers also publish an aggregation of the results of different algorithms on their web pages. However, cases where those components are used in a concerted effort to produce a reproducible publication remain rare.

It is our strong belief that reproducible research in Biometrics should be promoted. A special session exclusively composed of reproducible research papers following the above-mentioned principles would be a unique and impactful event.

[1] Patrick Vandewalle, Jelena Kovacevic, and Martin Vetterli. Reproducible research in signal processing - what, why, and how. IEEE Signal Processing Magazine, 26, 2009.


Special Sessions

7th IEEE International Conference on
Biometrics: Theory, Applications and Systems 
(BTAS 2015)
September 8 - 11, 2015