(2009) APPLICATION OF BIOMETRIC TECHNOLOGIES: INTEGRATING APPLIED RESEARCH INTO A GRADUATE LEVEL COURSE

APPLICATION OF BIOMETRIC TECHNOLOGIES: INTEGRATING
APPLIED RESEARCH INTO A GRADUATE LEVEL COURSE

Shimon K. Modi, Ph.D. 1, Stephen J. Elliott, Ph.D. 2

Abstract  This paper presents a case study for integrating Science and Engineering department. West Virginia offers a
applied research into a graduate level course in biometrics. Bachelor of Science in Biometric Systems through its
The biometrics course at Purdue University has a diverse Computer Science and Electrical Engineering department.
range of students, with majors varying from Computer University of Notre Dame has also incorporated biometrics
Science, Computer and Information Technology, Industrial into its curriculum through the Computer Science and
Technology, and Information Security. Therefore, the Engineering department. The Electrical Engineering
knowledge that students bring to the class, with respect to Department at U.S. Naval Academy offers a course in
statistical knowledge, etc varies tremendously. Many times biometrics at the undergraduate level. These courses have
the students are senior undergraduates or first year focused more on the basic research related to algorithm and
graduate students, and they have not been exposed much to sensor development for biometric systems. However, there is
research activities. The challenge for the instructors is to an increasing demand for biometric technologies in security
incorporate applied research principles, wrapped around infrastructures which necessitates development of courses
the concepts of biometric technologies and modalities that related to application of biometric technologies as well. This
students can then use in their respective disciplines, and at suits our students well – their backgrounds are in majors
the same time have a greater understanding of biometrics such as Computer Information Technology, Industrial
and their use within their majors. Biometrics is defined as Technology, Computer Science, and Information Security.
automated recognition of humans using physiological or The students typically will graduate to work in information
behavioral characteristics. The field of biometrics has technology, banking, manufacturing and healthcare, all of
received increasing attention in the last decade which has which have seen an increase in the deployment of biometrics.
led to engineering courses integrating biometrics into their
respective curricula. Biometric technologies are still in
developmental stages, and courses teaching biometric MOTIVATION
technologies have to be cognizant of its dynamic nature. The
This course has been offered for over 6 years, and typically
goal of this course is to provide these students with an
changes with the addition of new biometric equipment that
avenue to become involved in the research activities of the
arrives in the lab each semester. As the lab’s research
lab.
footprint and capabilities have matured, so has the course. In
fact, the course is now part of the research footprint of the
Index Terms –biometrics, case study, applied research, lab. That research emphasis is also influenced by the
graduate course. development of research goals for the College of
Technology, which is typically understood to be applied
INTRODUCTION
research. The research framework was inspired by ‘Pasteur’s
Biometrics is defined as the automated recognition of Quadrant’. D.E. Stokes conceptualized scientific research as
humans using physiological or behavioral characteristics. falling into one of 4 categories that could be represented as a
Examples of biometric technologies include fingerprint 4 quadrant chart [2] as shown in Figure 1 below. One
recognition, face recognition, and iris recognition. The use of quadrant contains scientists who conduct pure basic research
biometric technologies extends from applications for logical and have little interest in the potential uses of the research
access like signing in to a computer to physical access like findings for the real world. Physicist Niels Bohr, a scholar-
entering a secured facility. The field of biometrics has scientist who developed a model of the atom, is used
received increasing attention in the last decade which has led example of a researcher who fits within this quadrant.
to engineering and science programs integrating biometrics Thomas Edison, who conducted pure applied research in
into their respective curricula. Michigan State University has order to be able to market electric lighting and who had little
offered courses related to biometrics through its Computer interest in the scientific aspects of his work, is used as an

1
Shimon K. Modi, Purdue University, Industrial Technology, 401 N Grant St, W Lafayette, IN, 47906, USA, shimon@purdue.edu
2
Stephen J. Elliott, Ph.D., Purdue University, Industrial Technology, 401 N Grant St, W Lafayette, IN, 47906, USA, elliott@purdue.edu

example of pure applied research quadrant. A third quadrant 2) Comprehension: defined as the ability to grasp the
contains work that is neither overtly theoretical nor applied. meaning of the material being taught and articulate it.
It might contain taxonomic or classificatory work, which is 3) Application: defined as the ability to apply the
worthwhile but not driven by the desire either to advance comprehension of knowledge in specific situations.
knowledge or to develop practical solutions. This quadrant, 4) Analysis: defined as the ability to break down material
named for the "use-inspired basic science" of Louis Pasteur, into its component parts in order to understand its
is also labeled as “Pasteur’s Quadrant”. One of the organizational structure.
objectives of the course described in this paper was to 5) Synthesis: defined as the ability to put previously learned
expose students to the concepts of “use-inspired basic parts in order to create new knowledge. Students are creating
science” and pure applied research through the activities in new patterns and drawing inferences, such as formulating
this course. hypothesis from a set of data.
6) Evaluation: defined as ability to make critical judgments
and offer constructive perspective.
This six-level pattern governed the structure for class, using
the initial level to teach the students basic knowledge and
definition. The research component of the course
concentrated on the higher levels such as 5 and 6.

COURSE STRUCTURE
The course was offered over a 16-week period with a weekly
3 hour mandatory meeting time for all participants of the
class. The course comprised of the following modules:
• Basic knowledge of biometrics modalities like
fingerprint, face, iris, hand recognition, etc .
• Performance evaluation of biometric systems.
• Design of experiments & data analysis.
FIGURE. 1. CATEGORIZATION OF RESEARCH • Semester Project.

Biometric Modalities

COURSE FRAMEWORK The first module of the course focused on teaching the
students the basics of different biometric modalities, and
processes involved in using a biometric system. Biometric
This dynamic nature of biometric technologies poses several modalities include several different types, but this module
challenges from a pedagogical perspective. Unlike courses concentrated on fingerprint recognition, face recognition, iris
that focus on basic concepts of biometric technologies which recognition, hand recognition, voice recognition, signature
can be taught using traditional classroom lectures, the course verification, and vascular recognition. These modalities were
was structured to teach students about application of chosen since they are deployed in operational environments
biometric technologies and conducting applied research for and there is industry interest in those technologies.
biometric systems. The course had the following goals: Furthermore, there are examples of each of these modalities
1. Provide students with basic knowledge of biometric in the biometrics lab for students to interact with. The scope
systems. for each of these topics included learning about the history of
2. Provide students with concepts of design of the modality, understanding the basics of the modality and
experiment and data analysis. the advantages/disadvantages of deploying each technology
3. Provide students with the ability to perform applied in an operational scenario. Algorithm and sensor
research of biometric systems and critically analyze development was beyond the scope of this course. Students
the results. were taught about architecture of biometric systems and how
4. Provide students with the ability to select biometric to decompose any biometric system into its basic sub-
systems based on the requirements of the systems. Knowledge of the basic building blocks of any
application. biometric system is essential in order to understand the
The course objectives were based on Bloom’s taxonomy. different problems that can affect a biometric system.
The six levels of the structure were defined as follows [1]: Students were also taught about the process of enrollment,
1) Knowledge: defined as the lowest level of learning and identification and verification. Biometric technologies have
primarily relies on recall. several myths and misconceptions associated with it, and one

of the aims of this module was to dispel them. Some problem, selection of response variable, choice of factors,
examples are: there is a best biometric system for all levels and range, choice of experimental design, performing
applications, biometrics mean 100% security etc. There are the experiment, statistical analysis of data and forming
several different biometric modalities that are commercially conclusions. The step of performing the experiment requires
available and it is extremely important to understand the a data collection activity. Due to the time constraints and
advantages and disadvantages of these modalities depending policies regarding collection of human data, students used
on its intended application. For example, face recognition is data that collected previously in other experiments. The two
affected by different levels of ambient illumination and not main components of interest were developing a hypothesis
considering these factors has a detrimental effect. A based on the type of data they were presented with, and
combination of books and reports were used as source performing data analysis. Data analysis is extremely
materials for this class [4-6]. important in synthesizing the results of biometric system
output and drawing conclusions regarding performance of
Performance Evaluation of Biometric Systems the system. Most biometric system output can be distilled
Biometric technologies are continually improving with new into error rates, but the statistical significance of the
systems claiming to have higher accuracy [4]. The definition differences needs to be understood as well. The BSPA Lab
of accuracy is subjective and dependent on the context it is collects data from human subjects for various experiments
applied to. This module taught the students the definition of throughout the semester. Students in this course were taught
performance related terms like false accept rates, false reject the concepts of data analysis by analyzing real biometric
rates, failure to acquire errors etc. Biometric systems, at its data. The students had to process the raw biometric data,
core, can be described as a pattern recognition system. perform sample quality analysis and perform matching
Biometric system performance evaluation involves operations for the processed biometric data using different
understanding different error rates like false accept rates, commercially available systems. The processed data was
false reject rates, failure to acquire rates, etc. Students were used to demonstrate the concept of normality. The
taught about error rates related to system evaluation, the researchers of BSPA Lab also collect meta-data like age,
sources of these errors, and the interplay between these error gender, occupation etc. from individuals who participate in
rates. Biometric system error rates are essentially a tradeoff the data collection activities. Graphical analysis concepts
between security of the system and convenience of using the like scatter-plots and box-plots were taught using the sample
system. Comprehension of performance evaluation concepts quality score datasets as well. These aided the students in
and metrics are extremely important in making a decision understanding outlier analysis, and relate the cause for
which biometric system to use. The BSPA Laboratory has abnormal data back to the source of the biometric data.
been at the forefront of evaluating and researching An example of activity related to strategy of experimentation
performance of biometric systems when they are exposed to involved fingerprint data collected by researchers from the
a diverse demographic of users, environmental conditions, BSPA Lab of individuals between 18-25 years old, and 62
and different habituation and training levels etc. Performance years or older. The students were given an overall objective
results of any biometric system are also dependent on the of understanding the impact of age on fingerprint sample
type of test, application, and population that participated in quality. The students had to recognize the statement of
the test. This module concluded by teaching students how to problem, formulate the hypothesis, and decide which
write a performance evaluation report for a biometric system. statistical tests to perform. The students had to generate
Some important components of a performance evaluation sample quality scores for fingerprints of these two datasets,
report include details of the system tested, type of test and then statistically test for equality of quality scores
evaluation, demographics of the participants, system between the two datasets. The students performed normality
thresholds used for decision making, etc. [7] was used as test on the quality scores of the two datasets, and then
reference material for reporting performance evaluation performed a 2 sample t-test for testing equality. This activity
reports. also laid the foundation for teaching students about
parametric and non-parametric statistical test methods.
Design of Experiments & Data Analysis
Final Project
Students were taught the strategy of experimentation, which
is the general approach planning and conducting of The final project was designed keeping in mind the different
experiments. Students learned how to perform statistical phases of knowledge, comprehension, application, analysis,
design of experiments so that appropriate data could be synthesis, and evaluation from Bloom’s taxonomy, as well as
analyzed by statistical methods and draw objective concentrating on the application based research quadrant
conclusions [8]. Designing experiments requires knowledge described in [2]. Four research projects were created for the
of the general guidelines of experimentation. The following course, and the students were assigned to the projects based
guidelines were taught: recognition of the statement of on their interests. The projects commenced in the second half

of the semester. All four projects involved performance components of this course and the required readings for this
evaluation of a dataset collected in a live setting. At this course were a combination of book chapters and research
point in the semester the students had covered the basics of papers. Even given these challenges, we have had success.
biometric modalities, performance evaluation of biometric For undergraduates, we are going to develop a research
systems and design of experiments and data analysis. The opportunities program which will provide these students with
first phase of the project required the students to formulate more experience in understanding research methodologies.
the research question and hypothesis for their specific The first cohort of students will enter in Spring 2009.
projects. The second phase of the project involved
processing the raw biometric data using the appropriate CONCLUSIONS
biometric system. Due to time constraints students could not
perform the protocol development or data collection The biometric modalities module formed the knowledge
activities. Instead students were given the protocol and all level of Bloom’ s Taxonomy. The biometric modalities and
related information for the dataset from the actual data performance evaluation modules formed the comprehension
collection activity. This included demographic information and application levels of Bloom’ s Taxonomy. The design of
of the data, environmental conditions during data collection, experiment and data analysis module formed the analysis
and any deviations from the protocol. After understanding level of Bloom’ s Taxonomy. The term project formed the
the relations between all the collected data, the appropriate synthesis and evaluation levels of Bloom’ s Taxonomy. The
statistical data analysis was performed to test the hypothesis. overriding objective of the class was to give students
Using the non statistical knowledge about biometric systems, experience with applied research and the course and
the groups had to analyze the results and draw conclusions activities reflected this objective. This course shows an
from the results. Each group had to submit a final report that application of Bloom’ s taxonomy to an applied research
was structured according to the performance evaluation course. This course is designed to accommodate students
report. This project was designed to allow students to from multiple disciplines, and the applied research nature of
demonstrate the knowledge they had acquired as part of the this course provided an opportunity for students from diverse
class. For example, in one of the projects students were education backgrounds to contribute significantly. This
given 4 datasets of fingerprints collected from the same set course gave students experience with commercial systems,
of individuals from 4 different sensors. The students were which students indicated as an advantage for this type of a
given the overall objective of evaluating the problem of course. The active learning component of this course was
matching fingerprints from different sensors. With this well received by students as it gave them a chance to apply
information, the students formulated the hypothesis, classroom knowledge to real world systems. It is expected
processed the fingerprint data, calculated the error rates that this course will be taken by students from different
associated with fingerprint matching, and identified the disciplines that will help reflect the interdisciplinary nature
appropriate statistical tests for evaluating performance for of biometrics.
this experiment. The students had to report the process of
their experiment, along with their results and conclusions REFERENCES
using the performance evaluation report template.
1. McDaniel, T.R., Designing Essay Questions for
CHALLENGES
Different Levels of Learning. Improving College
This course was faced with several challenges. Biometrics is and University Teaching, 1978. 27(3): p. 120-123.
an interdisciplinary subject, and students registered in this 2. Stokes, D.E., Pasteurs Quadrant: Basic Science
class reflected that concept. Ensuring that all students were and Technological Innovation. 1997, Washington,
able to understand the technical and non-technical concepts D.C.: Brookings Institution Press.
of biometric technologies, and learn about applied research 3. IBG, Biometrics Market and Industry Report. 2007,
was a challenge. The educational challenge was to make IBG: NY. p. 224.
course content relevant to students from different disciplines. 4. Bolle, R.M., et al., Guide to Biometrics. 2004, New
Students typically enrolled in this course do not have the pre- York, N.Y.: Springer.
requisite knowledge of research methodologies and 5. Jain, A., P. Flynn, and A. Ross, eds. Handbook of
statistical methodologies. There are entire courses devoted to Biometrics. 2007, Springer. 556.
these topics, and we had the challenge of distilling the 6. Dunstone, T. and N. Yager, Design, Evaluation,
important concepts from those topics so that students would And Data Mining. 2008, New York, New York:
only concepts germane to biometrics. Another educational Springer-Verlag 288.
challenge for this course was the lack of a single textbook 7. Mansfield, A. and J. Wayman, Best Practices.
which would suffice the reading requirements. There are 2002, National Physics Laboratory: Middlesex. p.
excellent books and papers which address different 32.

8. Montgomery, D.C., Design and Analysis of
Experiments. 4th ed. 1997, New York: John Wiley
& Sons. 704.

(2009) APPLICATION OF BIOMETRIC TECHNOLOGIES: INTEGRATING APPLIED RESEARCH INTO A GRADUATE LEVEL COURSE

More Related Content

What's hot (17)

Similar to (2009) APPLICATION OF BIOMETRIC TECHNOLOGIES: INTEGRATING APPLIED RESEARCH INTO A GRADUATE LEVEL COURSE (20)

More from International Center for Biometric Research (20)

Recently uploaded (20)

(2009) APPLICATION OF BIOMETRIC TECHNOLOGIES: INTEGRATING APPLIED RESEARCH INTO A GRADUATE LEVEL COURSE