This study evaluated fingerprint quality across two populations, elderly and young, in order to assess age and moisture as potential factors affecting utility image quality. Specifically, the examination of these variables was conducted on a population over the age of 62, and a population between the ages of 18 and 25, using two fingerprint recognition devices (capacitance and optical). Collected individual variables included: age, gender, ethnic background, handedness, moisture content of each index finger, occupation(s), subject's use of hand moisturizer, and prior usage of fingerprint devices. Computed performance measures included failure to enroll, and quality scores. The results indicated there was statistically significant evidence that both age and moisture affected effectiveness image quality of each index finger at a=0.01 on the optical device, and there was statistically significant evidence that age affected effectiveness image quality of each index finger on the capacitance device, but moisture was only significant for the right index finger at a=0.01.

1. AN EVALUATION OF FINGERPRINT IMAGE QUALITY ACROSS AN ELDERLY
POPULATION VIS-A-VIS AN 18-25 YEAR OLD POPULATION
Nathan C. Sickler & Stephen J Elliott, PhD
Purdue University, College of Technology, Department of Industrial Technology
ABSTRACT or behavioral characteristic unique to the user. Therefore,
a biometric system requires something that a person "is",
This study evaluated fingerprint quality across two and not something that the person knows (secret) or has
populations, elderly and young, in order to assess age and (token). The most widely implemented biometric system
moisture as potential factors affecting utility image uses fingerprint recognition technology. The volume of
quality. Specifically, the examination of these variables use of fingerprint recognition technology can be attributed
was conducted on a population over the age of 62, and a to the large number of applications in which it can be
population between the ages of 18 and 25, using two used. Applications include: financial services, health
fingerprint recognition devices (capacitance and optical). care, electronic commerce, telecommunications, and
Collected individual variables included: age, gender, government [2].
ethnic background, handedness, moisture content of each The following are two examples of applications that
index finger, occupation(s), subject's use of hand currently use fingerprint recognition devices. Purdue
moisturizer, and prior usage of fingerprint devices. Employee Federal Credit Union (PEFCU), in West
Computed performance measures included failure to Lafayette, Indiana, integrated ATMs with capacitive-
enroll, and quality scores. The results indicated there was based fingerprint recognition sensors in its One Touch
statistically significant evidence that both age and program (formerly known as TARAtouch). Users can
moisture affected effectiveness image quality of each deposit and withdraw money, and receive account
index finger at a=0.01 on the optical device, and there statements after entering their account number, and
was statistically significant evidence that age affected presenting the fingerprint used to enroll in the One Touch
effectiveness image quality of each index finger on the program. PEFCU's fingerprint ATM enrollees have not
capacitance device, but moisture was only significant for experienced a single case of fraudulent use since the
the right index finger at a= 0.01. deployment of the biometrically enabled ATMs six years
1. INTRODUCTION ago [3]. Furthermore, Arnold states "individuals over the
age of 55 were the most accepting to the idea of gaining
Traditional methods of automatic personal identification access to their money without using passwords" [4]. This
are based on one, or a combination, of the following two is important to understand, since elderly or retired
security measures: a secret or a token. Secret-based individuals generally have more expendable money and
security methods require users to provide information that more time to travel. However, the success of a fingerprint
only they have knowledge of, such as a password or a biometric system deployed in the public, such as point-of-
personal identification number (PIN). Token-based sale or airport identification, would likely fail if the
security methods require users to present an item that is in system discriminates against certain populations that are
their possession, such as a key, security badge or an prone to have poor fingerprint utility image quality
automatic teller machine (ATM) card [1]. Concerns (usefulness of the image, from the system's standpoint),
regarding the security of systems using these methods which includes the elderly population.
arise from the fact that the system cannot determine if the Eight states (Arizona, California, Connecticut, Illinois,
individual providing the secret or the token is, indeed, the Massachusetts, New Jersey, New York, and Texas) have
intended user. Tokens can be lost, stolen, and forged, implemented fingerprint technology in the welfare benefit
while secrets can be compromised and "surprisingly, 25 programs of some counties. The welfare applicants of
percent of people appear to write their PIN on their ATM these counties are required to submit fingerprint samples
cards" [2]. More secure methods of automatic personal in order to receive benefits. The purpose of keeping
identification receiving attention are biometrics. Unlike fingerprint records of applicants is to "eliminate duplicate
secret or token-based systems, a biometric system participation ... deter fraud ... and [restore] the public's
provides the security that the approved user interacted confidence in the integrity of the welfare system [5]."
with the system, by matching or not matching a physical
0-7803-9245-0/05/$20.00 C2005 IEEE
Authorized licensed use limited to: Purdue University. Downloaded on December 14, 2009 at 17:31 from IEEE Xplore. Restrictions apply.

2. Here again, the integrity of a fingerprint system would be to manual labor, injuries, disease, scars or other
reduced if the system failed to enroll, verify, or identify an circumstances [10] such as loose or wrinkled skin. These
individual due to poor fingerprint utility image quality. contact issues may introduce false minutiae points into the
captured image, causing higher FTE, FTA, FMR, and
2. QUALITY AND USAGE ISSUES WITH THE FNMR. Both contact issues can be observed in Figure 1,
ELDERLY in comparison to a normal image (Figure 2).
The general biometric model can be applied to the
fingerprint recognition process; likewise, potential
problem areas of a fingerprint system can be paralleled to
areas of the general biometric model. When placed into
the general biometric model, two areas of the fingerprint
recognition system (data collection and signal processing)
are affected by the failure of interaction between the
system and the user: non-uniform contact and
irreproducible contact. The problem of interaction
between the user and the system affects the sub-category
of presentation within the data collection silo. If the user Figure 1: Dry, worn fingerprint (left) and resulting
cannot present a fingerprint to the device, then enrollment minutiae points (right).
(and subsequent verification or identification attempts) is
not possible through fingerprint recognition.
Irreproducible and non-uniform contacts affect the sub-
categories of feature extraction and quality control within
the signal processing silo. Non-uniform contact tends to
produce images of low quality, resulting in poor feature
extraction of the presented fingerprint, whereas
irreproducible contact may allow for images with high
quality but, depending on the severity of the contact issue,
image quality can be adversely affected.
Moreover, the utility quality of a captured image is one of
the most important aspects for a biometric system, as it is Figure 2: Normal fingerprint (left) and resulting minutiae
this quality parameter that determines whether a captured points (right).
image is acceptable for further use within the biometric
system. The utility quality of a presented fingerprint The resulting utility quality scores of the fingerprint
image is developed and processed by the quality control images in Figures 1 and 2 are displayed in Figures 3 and
function of the biometric system, and a score, based on 4, respectively.
the image's usability, is assigned to that image. It is these
quality scores and captured images that provide the data
used by the biometric system and allow it to make an
accept/reject decision.
Discussion of poor image quality issues of elderly
fingerprints occurs in the biometric literature [2], [6], [7],
[8], [9]. These issues pose problems for fingerprint
recognition systems during the enrollment, verification,
and identification processes. Therefore, the objective of
this study was to determine the impact that particular
variables, namely age and moisture, had on the utility
quality of fingerprint images.
Two of the most common causes of poor utility quality
are attributable to non-uniform and irreproducible contact
(Figure 1), which occur between the fingerprint and the
platen of a fingerprint sensor. Non-uniform contact can
result when the presented fingerprint is too dry or too wet, Figure 3: Image quality calculation using commercially
and irreproducible contact occurs when the fingerprint available software on a dry, worn fingerprint (low utility
ridges are semi-permanently or permanently changed due quality).
Authorized licensed use limited to: Purdue University. Downloaded on December 14, 2009 at 17:31 from IEEE Xplore. Restrictions apply.

3. ridges. The ridge charges and valley non-charges are
converted to pixel values, resulting in a fingerprint image
[12]. The device has an additional feature, which properly
aligns the subject's finger in order to capture the most
distinctive area of the fingerprint. This "ridge lock" is
situated at the base of the device's platen and fits into the
groove of the joint closest to the fingernail of the
presented finger. The capacitance sensor has a 13 mm x
13 mm platen chip grid, and produces an image of 250
dots per inch (dpi) from the converted charges.
The optical sensor acquires an image using a charged
couple device (CCD) camera, light emitting diode (LED)
illumination, and a prism. When a finger is placed on the
platen, which is one side of the prism, the CCD camera
captures an image of the signal reflected by the fingerprint
[12]. The optical sensor has an image acquisition surface
Figure 4: Image quality calculation using commercially of 13 mm x 18 mm and produces an image of 500 dpi
available software on a normal fingerprint image (high from the reflected signal.
utility quality).
3.2. Moisture Checker
Furthermore, the causes of these two contact issues have a The device selected for measuring the moisture content of
higher likeliness of occurring when an elderly user the fingerprint region was Scalar America's MY707S skin
presents the enrolled fingerprint to the fingerprint device moisture checker. This device obtains moisture readings
[10]. As individuals age, their skin becomes drier, sags using electrical conductivity and has a reading accuracy of
from the loss of collagen and elastin fibers, becomes +/- 0.2 percent. Approximately 80 percent of the moisture
thinner and loses fat; all of these conditions decrease the reading is influenced by the top ten microns of the skin,
firmness of the skin, causing wrinkles [11]. Skin and approximately 90 precent by the top 200 microns
exhibiting these symptoms is likely to have incurred semi- [13].
permanent or permanent damage over the life of the
individual. 3.3. Computer Hardware
Another issue involves some elderly individuals being A DellTm Dimension.TM workstation served as the platform
unable to properly present the enrolled fingerprint to the to communicate with the optical sensor and had the
fingerprint device. An elderly individual's ability to use a following specifications: 2 GHz Intel® Pentium® 4
fingerprint device can be severely limited by arthritis or a processor; 512 MB of 2100 double data rate memory; 40-
loss of motor skills, which may affect the quality of GB, 7200-rpm hard drive; and Microsoftg Windowsg
captured images. 2000 operating system with Service Pack 2. The
workstation was loaded with Neurotechnologija's
3. PROCEDURES VeriFinger 4.1 software, the drivers for the optical sensor,
and was used to store the images captured from the
The purpose of this study was to evaluate the fingerprint sensor. A demographic survey and instructions for
quality of an elderly population and compare it to an 18- interacting with each device were also presented to each
to 25-year-old population, in order to assess potential participate on this workstation using Microsoftg
factors affecting utility image quality. This section PowerPoint®'97, a headset with adjustable volume, and a
describes: the software and hardware, the variables 17-inch LCD monitor.
examined, the type of study, and the test procedures A DellTM InspironTm 8200 laptop computer served as the
undertaken to evaluate fingerprint quality. platform to communicate with the capacitance sensor and
had the following specifications: 1.8 GHz Mobile Intelg
3.1. Sensors Pentium® 4 processor, 256 MB of 2100 double data rate
The two fingerprint recognition sensors used in this study memory; 40-GB, 4200-rpm hard drive; and Microsoft®
included a capacitance sensor, and an optical sensor. The Windowsg XP Home Edition operating system. The
capacitance sensor acquires an image using electrical laptop was loaded with image capturing software, and the
charges. When a finger is placed on the capacitive chip drivers for the capacitance sensor; the laptop stored the
grid (platen), electrical charges accumulate at the points images captured from the capacitance sensor.
where the finger ridges contact the chip grid. Absence of
an electrical charge indicates a valley between the finger 3.4. Selected Features to be Recorded
Authorized licensed use limited to: Purdue University. Downloaded on December 14, 2009 at 17:31 from IEEE Xplore. Restrictions apply.

4. Independent variables included age, and moisture content the utility quality was established and incorporated with
of each index finger. The dependent variable was the the previously collected demographic and moisture data.
utility quality of the fingerprint images derived from The data were then analyzed using the GLM function of
Aware, Inc.'s Fingerprint Image Quality API. SAS® 8e.
3.5. Evaluation Classification 4. RESULTS
This study is best described as a scenario evaluation
(Table 1). According to the UK Biometrics Working The study examined two hypotheses. The first hypothesis
Group Best Practice Document 2.01 [14], the goal of a stated that there is no statistically significant difference in
scenario test "is to determine the performance of a the fingerprint quality between the age groups 18-25 and
complete biometric system in a specific application 62+. The second hypothesis stated that there is no
environment with a specific target population." statistically significant difference between the fingerprint
This study satisfies this statement, since two complete moisture content of the age groups 18-25 and 62+. Two
biometric systems are being tested in conditions similar to population age groups were targeted, the elderly (62+)
some e-commerce, ATM banking, and point-of-sale and a younger (18-25 years). No subject was excluded
environments using specific populations (18- to 25-year- from participating based on age, but data from subjects
olds and 62 years and older). not falling into one of these age groups were excluded in
the analysis. The minimum age was set to 18 years old,
Table 1. Scenario Evaluation Criteria since individuals this age and older are considered adults
and do not need a guardian's consent to participate. The
Application Experiment maximum age of the younger population was set to 25
Classification years old, in order to establish the typical age range for
System classification Positive Identification college or university students. The recruitment of the 18-
Cooperative versus Cooperative to 25-year-old population was conducted in the School of
Non-cooperative Technology Department of Industrial Technology, which
Overt versus Covert Overt has a higher percentage of white males than minority
Habituated versus Non- Variable males, white females, or minority females. Consequently,
habituated there was a higher rate of white males among the subjects.
Attended versus Non- Attended Participation for the 62+ age group was open to all
attended individual's partaking in activities at Purdue University's
Ismail Center and residents of Westminster Village. The
Standard Environment Lab environment, room Ismail Center had an approximately equal number of
lighting, temperature males and females, with most members being of a
Public versus Private N/A Caucasian/white ethnic background. Westminster Village
Open versus Closed Closed had approximately three times as many females as males,
with nearly 100 percent of the residents being
3.6. Compliance with Best Practices Caucasian/white. Therefore, a higher rate of Caucasian
This study conformed to recommendations established by females than minority females participated in this study.
the UK Biometrics Working Group Best Practice
Document 2.01. 4.1. Hypothesis 1
For Hypothesis 1, a one-way analysis of variance
3.7. Test Procedures (ANOVA) computation using the GLM function was
After the volunteers consented to participate, they were conducted in order to examine the statement that
shown a Microsoft® PowerPoint® presentation fingerprint image utility quality is not affected by age.
describing the proper interaction with each fingerprint The computation for this one-way ANOVA (Table 2)
device. This presentation was followed by a short survey included data from the capacitance and the optical sensors
used to collect demographic information. There were a for each index finger.
total of four sessions for this study, one enrollment and
three verification; each session was separated by Table 2. Image quality and age ANOVA
approximately one week. The order of the device and Right Index Left Index
index finger used was randomized for each session, with Capacitance F value = 100.16 F value of 116.75
the moisture content being measured before each attempt. p value <.0001* p value <.0001*
Captured images were automatically named using code 39
bar codes encoded with the appropriate identifiers for Optical F value = 180.44 F value = 203.89
each participant and then saved. After image collection, p value <.0001* p value <.0001*
Authorized licensed use limited to: Purdue University. Downloaded on December 14, 2009 at 17:31 from IEEE Xplore. Restrictions apply.

5. * Significant at oc = 0.01 response to age (Figure 6). The graph in Figure 6
illustrates this correlation.
The results of the ANOVA computed for the utility image
quality and age suggested that there was indeed a MDI STURE
101 + +F:
correlation between image quality and age, regardless of
which device or index finger was examined. The Pearson
correlation (r = -.78) was also calculated for the image
utility quality in response to age (Figure 5). The graph in i..
Figure 5 shows that a linear correlation exists.
Ht+ +'t 'X i
~~~~~~~~~~~~~~~+
Z9 X '49 0 P)l n to C.)
ACE
Figure 6: Graphical plot of the relationship between
moisture content and age.
Based upon these findings, it was concluded that
the moisture content data were statistically significant at oc
20 X 40 5 6 Go
4 to s 9a |§ = 0.01 for each index finger using the optical sensor. The
AGE
Figure 5: Graphical plot of the correlation between moisture content data were also statistically significant for
image quality and age. the right index finger using the capacitance sensor, but
were not statistically significant for the left index finger.
Based upon these findings, it was concluded that the Therefore, Hypothesis 2 is rejected at oc = 0.01 for each
image utility quality data were statistically significant at oc index finger using the optical sensor and for the right
= 0.01 for each index finger, as well as for each sensor
index finger using the capacitance sensor.
when tested against age. Therefore, Hypothesis 1 is
5. CONCLUSION
rejected at oc = 0.01.
The purpose of this study was to evaluate the fingerprint
3.1. Hypothesis 2 utility image quality of an elderly population in
For Hypothesis 2, a one-way ANOVA computation (using comparison to an 18- to 25-year-old population baseline.
the SAS GLM function) was conducted in order to During the formulation of this study, two hypotheses were
examine the statement that fingerprint moisture content is generated and examined after the collection and analysis
not affected by age. The computation for this one-way of the data. The first hypothesis states that there is no
ANOVA (Table 3) included data from the capacitance and statistically significant difference in the fingerprint image
the optical sensors for each index finger. utility quality between the age groups 18-25 and 62+. This
Table 3. Moisture content and age ANOVA hypothesis was rejected at oc = 0.01 for each fingerprint
device (capacitance sensor and optical sensor), regardless
Right Index Left Index of the index finger used. The second hypothesis states that
Capacitance F value = 9.10 F value of 2.93 there is no statistically significant difference between the
p value <.0032* p value <.09 fingerprint moisture content of the age groups 18-25 and
Optical F value = 18.22 F value = 10.13 62+. This hypothesis was rejected at oc = 0.01 for both
p value <.0001* p value <.0019* index fingers when used with the optical sensor, and it
* Significant at oc = 0.01 was rejected for the right index finger in conjunction with
the capacitance sensor. However, this hypothesis failed to
The results of the ANOVA computed for the moisture be rejected at oc = 0.01 for the left index finger and the
content and age suggested that there was, in part, a capacitance sensor. Other observations made throughout
correlation between moisture and age, albeit not as strong the study were briefly examined, but only with anecdotal
as image quality vs. age. The Pearson correlation (r= - data.
.38) was also calculated for the moisture content in
Authorized licensed use limited to: Purdue University. Downloaded on December 14, 2009 at 17:31 from IEEE Xplore. Restrictions apply.

6. in Testing and Reporting of Biometric Devices 2.01,"
September 2002, w .
6. REFERENCES /ssf
[1] P. Meenan and R. Adhami, "Fingerprinting For
Security," IEEE Potentials, 2002, 33-38.
[2] A. Jain, L. Hong, and S. Pankanti, "Biometrics:
Promising frontiers for emerging identification market,"
Comm. ACM, 91-98, February 2000.
[3] L. Mearian, "Want Access? Give'em the Finger,"
November 2002,
[4] B. Arnold (private communication), 2002.
[5] D. Burton, R. Salstrom, & J. L. Wayman, "A Proposed
Cost/Benefit Analysis of Finger Imaging Systems in
Social Service Applications," presented at the 12th
Annual CSU-POM Conference, California State
University, Sacramento, 2000.
[6] G. Behrens, "Assessing the Stability Problems of
Biometric Features," presented at the International
Biometrics 2002, Amsterdam, 2002.
[7] D. J. Buettner, "A Large-Scale Biometric
Identification System at the Point of Sale," September
2002,
-2 Final%0dL0Ou,tnr2Bi
f
[8] A. K. Jain and S. Pankanti, Advances in Fingerprint
Technology (2nd ed.). New York: Elsevier, 2001.
[9] X. Jiang and W. Ser, "Online Fingerprint Template
Improvement," IEEE Trans. Pattern Analysis and
Machine Intelligence vol. 24(8), pp. 1121-1126, 2002.
[10] A. Jain, L. Hong, and R. Bolle, "On-Line
Fingerprint Verification," IEEE Trans. Pattern Analysis
and Machine Intelligence, vol. 19(4), pp. 302-314, 1997.
[11] American Academy of Dermatology, "Mature Skin,"
November 2002, http llwww.aa
[12] N. K. Ratha, A. Senior, and R. M. Bolle,
"Automated Biometrics," Proceedings of ICAPR-2001,
2001.
[13] Scalar America, "MY707S Moisture Checker White
Papers," Scalar America Documentation, 2003
[14] A. J. Mansfield, and J. L. Wayman, "Best Practices
Authorized licensed use limited to: Purdue University. Downloaded on December 14, 2009 at 17:31 from IEEE Xplore. Restrictions apply.