Students sitting in a biometrics seminar. IEEE Biometrics Council has introduced the Distinguished Lecturers Program (DLP) to support education related activities for the biometrics community. The purpose of DLP is to increase awareness about topics relevant to Biometrics by creating a pool of leading experts who are willing to speak in meetings hosted by IEEE Chapters and Sections. To organize a distinguished lecture, please read the guidelines and submit a completed DLP Speaker Request Form.

DLP Speakers (2022 – 2024)

joseph bigun headshot.

Joseph Bigun

Halmstad University, Sweden

View website →

A summary of main Fourier features along with their importance to human visual intelligence will be given. These features comprise local direction, local (absolute) frequency, and local phase which in turn drive more complex models of shape, also they being dense.

Whereas fingerprint wave structures can be modelled by frequency and direction maps alone, minutiae in the form of ridge ends and bifurcations call for modelling these as first order angular variation of local phase. Likewise, complex global patterns in the form of cores and deltas demand a modelling of dense direction maps using angular variations of order 1 and -1. Similar example applications from other biometrics will be visited, including identity recognition using iris, periocular regions, lip movements, and full faces.

However intricate, the underlying shape models demand a principled and robust estimation of local direction and frequency, independent of each other. With such estimations at hand one can recognize more complex patterns, including those using local phase, as delivered by Gabor filters associated with estimated frequencies and directions. Tools using complex valued filters on complex valued dense maps, easing the practice of more intricate shape recognition will be presented. Examples of shape models built on top of basic dense fields, which in themselves yield dense responses, will be given along their use..

A characteristics of a physical wave is that it is everywhere and cannot be confined to a location. In images, texture is a local property defined by translation invariance of the local appearance. It is hard to define a location inside a texture uniquely by observing the local pattern around the location. Examples in fingerprints are patches with slow orientation and frequency changes at locations far away from deltas and cores. The more it is difficult to define a location by image measurements of an image patch, the more the patch exhibits wave property.

On the other hand, a particle in physics is characterized by its infinitely precise location. In images too, it is sometimes possible to define a location very precisely by measuring image pattern properties around the location, which is the opposite of a texture patch. In fingerprints, delta and core are example patterns which can define a location rather precisely and uniquely at macro scale. Even in micro scale such patterns exist: ridge ends and bifurcations, also known as minutiae. In line with this division of pattern characteristics, there exist image measurements that are more suitable for each of the two cathegories of image patches. Wave properties can be obtained by measuring spectral energy properties, the most known example yielding dense orientation fields. Likewise, patches having particle nature can be described most effectively by their own tools, to the effect that by using them not only one can detect locations precisely but also systemize and explain what one is measuring. Examples include direction estimation in harmonic coordinates yielding precise localization of cores, deltas as well as minutiae, each turning out to be a special type of direction measurement but in non-cartesian coordinates.

Kevin Bowyer headshot.

Kevin W. Bowyer

University of Notre Dame, USA

View website →

Iris recognition is increasingly used in large-scale applications. As a result, presentation attack detection for iris recognition takes on fundamental importance. This survey covers the diverse research literature on this topic. Different categories of presentation attack are described and placed in an application-relevant framework, and the state of the art in detecting each category of attack is summarized. One conclusion from this is that presentation attack detection for iris recognition is not yet a solved problem. Datasets available for research are described, research directions for the near- and medium-term future will be discussed, along with a short list of recommended readings.

Face recognition technology has recently become controversial over concerns about possible bias due to accuracy varying based on race or skin tone. We explore three important aspects of face recognition technology related to this controversy. Using two different deep convolutional neural network face matchers, we show that for a fixed decision threshold, the African-American image cohort has a higher false match rate (FMR), and the Caucasian cohort has a higher false nonmatch rate. We present an analysis of the impostor distribution designed to test the premise that darker skin tone causes a higher FMR, and find no clear evidence to support this premise. Finally, we explore how using face recognition for one-to-many identification can have a very low false-negative identification rate and still present concerns related to the false-positive identification rate. Both the ArcFace and VGGFace2 matchers and the MORPH dataset used in our experiments are available to the research community so that others should be able to reproduce or reanalyze our results.

Mark Nixon headshot.

Mark Nixon

University of Southampton, UK

View website →

Face recognition technology has recently become controversial over concerns about possible bias due to accuracy varying based on race or skin tone. We explore three important aspects of face recognition technology related to this controversy. Using two different deep convolutional neural network face matchers, we show that for a fixed decision threshold, the African-American image cohort has a higher false match rate (FMR), and the Caucasian cohort has a higher false nonmatch rate. We present an analysis of the impostor distribution designed to test the premise that darker skin tone causes a higher FMR, and find no clear evidence to support this premise. Finally, we explore how using face recognition for one-to-many identification can have a very low false-negative identification rate and still present concerns related to the false-positive identification rate. Both the ArcFace and VGGFace2 matchers and the MORPH dataset used in our experiments are available to the research community so that others should be able to reproduce or reanalyze our results.

Many biometrics can be used in forensics, as that is largely where biometrics started as a subject. This keynote concentrates forensics, and use of gait, ear and soft biometrics. We will describe how gait has been presented as evidence, and the limitations and advantages in has provided. We shall also describe how the ear itself can be used as identification evidence, as people can hide their faces but are usually less concerned with hiding their ears. We shall also discuss how soft biometrics are used already as evidence, and how they can be learned from video data. Any consideration for use of biometrics in forensics must consider the provision of evidence. Procedures are described for handcrafted approaches in gait and in ear, though the emergence of deep learning complicates the provision of evidence for newer biometrics and newer implementations. Overall it is likely that biometrics has a rich potential future, and as well as describing the current state of art this talk describes some of the considerations that must be made for generating and using gait, ear and soft biometrics in forensics. As such it is an indicator of a fertile and rewarding area for deployment of biometrics.

With the proliferation of surveillance cameras, society needs means to identify people from the images these cameras provide. Crime solving websites are replete with imagery of criminals who are often disguised and/or at low resolution; terrorist attacks yield more imagery. We noticed this many years ago and were the first to develop systems that aimed to recognise people by their gait, their style of walking. This talk will describe some of the earlier approaches and their motivation, together with the recent works on deep learning. More recently we have moved to recognising people from human descriptions, consistent with eyewitness statements and the limited spatial and temporal resolution of surveillance imagery, and the chance of disguise. We have shown that human descriptions can be used for recognition and retrieval, and formulated ways to make these descriptions more effective. We have so far used descriptions of the face, the body and the clothing, and are our current work shows how the labels can be derived by computer vision and explores the new information available by the interface between semantic description and automated recognition. This talk thus surveys these areas, describing progress in gait and in soft biometrics.

Sharath Pankanti headshot.

Sharath Pankanti

Microsoft, USA

View website →

The world’s population, which is currently estimated to be 7.5 Billion, is very likely to surpass 10 Billion by the turn of the century. While there are several challenges when dealing with a population of this magnitude, the ability to positively establish or refute an individual’s identity is likely to be one of the fundamental expectations of a global society. In this article, we systematically discuss the issues impacting the design, implementation and deployment of a large-scale biometric identification system that can effectively manage and distinguish over 10 Billion identities. In this regard, we identify four technological issues that have to be satisfactorily resolved to design such a system: system scalability, identification accuracy, response time, template security and privacy. We discuss how the lessons learned from ongoing large-scale biometric systems such as UAE’s Border Crossing System and India’s National ID Card Program (Aadhaar) can be leveraged and incorporated into a Global ID system that handles 10B identities. Further, we study existing large-scale pattern recognition and machine learning systems, and determine how the challenges resident in such systems can be effectively addressed for use in the proposed Global ID system. Finally, we assess the gaps that need to be addressed by the research and development community-at-large for designing the Global ID system. We conclude that the outstanding research, engineering and design topics are “Grand Challenges” and, without a serious understanding of the underlying complex issues, simplistic identity infrastructure solutions will be dwarfed by the enormity of the identity problems of the next generation.

Machine learning on encrypted data can address the concerns related to privacy and legality of sharing sensitive data with untrustworthy service providers. Fully Homomorphic Encryption (FHE) is a promising technique to enable machine learning and inferencing while providing strict guarantees against information leakage. Since deep convolutional neural networks (CNNs) have become the machine learning tool of choice in several applications, several attempts have been made to harness CNNs to extract insights from encrypted data. However, existing works focus only on ensuring data security and ignore security of model parameters. They also report high level implementations without providing rigorous analysis of the accuracy, security, and speed trade-offs involved in the FHE implementation of generic primitive operators of a CNN such as convolution, non- linear activation, and pooling. In this work, we consider a Machine Learning as a Service (MLaaS) scenario where both input data and model parameters are secured using FHE. Using the CKKS scheme available in the open-source HElib library, we show that operational parameters of the chosen FHE scheme such as the degree of the cyclotomic polynomial, depth limitations of the underlying leveled HE scheme, and the computational precision parameters have a major impact on the design of the machine learning model (especially, the choice of the activation function and pooling method). Our empirical study shows that choice of aforementioned design parameters results in significant trade-offs between accuracy, security level, and computational time. We strongly believe our work will enable the next generation of biometrics systems affording biometrics recognition from encrypted data and thus overcoming many of the privacy concerns.

Venu Govindaraju headshot.

Venu Govindaraju

University at Buffalo, SUNY, USA

View website →

We present an overview of two decades of innovation in perspective on the evolution of research in this area and the future of the field. We highlight handwriting recognition at the Govindaraju lab at the University at Buffalo and offer our seminal work in handwriting recognition that was at the core of the first handwritten address interpretation system used by the U.S. Postal Service, described as one of the first practical success stories of AI and as a shining example of AI for the Social Good (Eric Horvitz, Microsoft Research). We journey through the HWR landscape, from lexicon-based to lexicon- free approaches, and from heuristics-driven techniques to the principled methodologies that we introduced. We explore a sample of the variety of impactful applications that resulted from our research, from the processing of healthcare forms for the NYS Department of Health for deriving early indicators of outbreaks, to access to historical documents through word spotting, transcript mapping and other indexing schemes for digital libraries, to award-winning pre-processing techniques and multilingual OCR solutions.

Given the pervasive use of e-commerce transactions and personal and protects the privacy and online assets of individuals and organizations. We recommend a data storage in the cloud, society has an urgent need for a robust process that authenticates totally new approach that rethinks the entire ‘science of authentication.’ The biometrics and cyber security communities have approached the challenge from different vantage points. The former focuses on ‘individuality’ and ‘liveness’ of human characteristics whereas the latter has primarily considered encryption and elaborate software protocols. This talk explores methods that go beyond the traditional biometrics of physical and behavioral modalities by integrating tests for humanness and identity in a cognitive framework. We also will show how our holistic process allows for a more practical approach to security within the framework of a continuous authentication scenario.

A ‘smart space’ is one that automatically identifies and tracks its occupants using unobtrusive biometric modalities such as face, gait, and voice in an unconstrained fashion. Information retrieval in a smart space is concerned with the location and movement of people over time. We abstract a smart space by a probabilistic state transition system in which each state records the probabilities of presence of individuals in various zones of the smart space. We carry out track-based reasoning on the states in order to determine more accurately the occupants of the smart space. This leads to a data model based upon an occupancy relation in which time is treated discretely, owing to the discrete nature of events, but probability is treated as a real-valued attribute. Using this data model, we show how to formulate a number of spatio-temporal queries, focusing on the computation of probabilities, an aspect that is novel to this model. We show that a query-dependent metric gives significantly better results for a class of occupancy-related queries compared with query- independent metrics.

Xilin Chen headshot.

Xilin Chen

Institute of Computing Technology, Chinese Academy of Sciences, Beijing, China

View website →

Gesture has emerged as an important component for biometrics authentication and day- to-day interaction in our lives. It is also widely used in traffic control, marine communication, etc. Gesture can even extend to form a kind of language — sign language. Sign language contains a larger vocabulary set and has become a major communication tool among the Deaf. Gesture recognition provides not only human-human communication, but also human-machine interaction for a range of biometrics applications. This talk will firstly introduce challenges in hand gesture recognition, and then discuss the advances in hand gesture and sign language recognition.

(This talk title can be preferably for the undergraduate students or industry) In the past decades, computer vision-based biometrics has become the hottest area in artificial intelligence due to it can offer similar or better results, in some typical tasks such as the recognition of objects or humans, than those performed by humans. However, most current computer vision systems are designed for specific task(s) in the close world, and hard to deal with open world cases. Flat structure for specific task(s) without reasoning and lack of knowledge are the major barriers toward a flexible computer vision system. For this purpose, a key factor is understandable or interpretable. Therefore, objects should be processed in a contextual environment rather than a solo one with a simple identity, and objects should also be associated with relevant concepts. A conceptual mapping of a given image brings enhanced representation, which can support versatile tasks. In this talk, I will briefly review the current state of computer vision, and talk about some open problems. I will then share my points on these relevant problems. Some of our efforts towards understandable computer vision are reported, including hierarchical object detection and categorization, scene graph construction and its application, unseen object inference, etc.

Human Faces are widely used in Biometrics and can reveal range of other features that can not just improve the personal identification but also advance healthcare technologies. Similarly the human actions and gait data can also reveal such intrinsic healthcare parameters for a range of real world applications and assessments. During last several years, our research group has investigated a range such problems and developed applications. This talk will discuss on such challenges and technologies for healthcare advancements.

Yasushi Yagi headshot.

Yasushi Yagi

Osaka University, Japan

View website →

We have been studying human gait analysis for more than 15 years. Because everyone’s walking style is unique, human gait is a prime candidate for person authentication tasks. Our gait analysis technologies are now being used in real criminal investigations.

We have constructed a very large-scale gait database, and proposed several methods of gait analysis. The appearances of gait patterns are influenced by changes in viewpoint, walking direction, speed, clothes, and shoes. To overcome these problems, we have proposed several approaches using a part-based method, an appearance-based view transformation model, a periodic temporal super resolution method, a manifold-based method and score-level fusion. During this talk, I will also present the experimental results to show the efficiency of our approaches on the large-scale gait database.

Human gait can be conveniently acquired from a-distance and also widely accessible from the publicly installed surveillance systems. During this talk, I will present a new focus on the gait analysis to detect emotions from the persons gait patterns. The objective of such study is predict human emotion and mental condition, and for the people surrounding us. During this talk, I will also introduce some studies on its medical applications.