Li S.Z., Jain A.K. (eds.) Encyclopedia of Biometrics

Подождите немного. Документ загружается.

Standardization

In order to allow for better global interoperability,

several efforts to standardize the biometric match-on-

card technology are on-going. There is effort at the

international standards body ISO/IEC JTC1 SC17

WG11 to introduce the match-on-card standards. In

addition, the effort on biometrics is also on-going, such

as at the ISO/IEC JTC1 SC37 level to develop the com-

pact ﬁngerprint template format suitable for smartcards.

In 2005, several standards for biometric data inter-

change format have been published including ﬁnger

minutiae data, ﬁnger pattern spectral data, ﬁnger image

data, face image data, iris image data, ﬁnger patter n

skeletal data etc. ISO/IEC 19794-1 [10 ] is intended to

describe the framework for deﬁning biometric data

interchange formats. In ISO/IEC 19794-2 ﬁnger min-

utiae data [11], compact card format is included in the

speciﬁcation to support ﬁngerprint authentication

with smartcard. The document ISO/IEC 7816-11

[12], published in 2004, speciﬁes basic operations for

performing personal veriﬁcation through biometric

methods using smartcard. However, the above stan-

dards are not sufﬁcient for biometric match-on-card.

A standard with more in-depth speciﬁcation is

needed for deployment of match-on-card with better

interoperability.

In 2006, a new work group (WG) 11 in Sub-

Committee 17 under Joint Technical Committee 1

(JTC1) of ISO/IEC was formed. The role of WG11 is to

deﬁne the functional blocks and components for the use

of integrated circuit (IC) cards in applications where

the matching of biometric identiﬁers is to be performed

on-card. The document entitled ‘‘24787 Information

technology - Identiﬁcation cards: On-Card matching’’

is still under committee draft stage as of February 2008.

Technical Challenges

MOC technology is challenging to develop due to the

limited resources – computational power, memory,

and power supply, in the smartcards. For example,

today’s PC has powerful speciﬁcation while for smart-

card has relatively much lower processing capability.

For example, one of the high-end conﬁgurations is

only 16-bit, 25 MHz processor with 8Kb RAM and

1Mb ﬂash memory. There are few 32-bit smartcards

but the price are quite expensive. The most widely used

smartcard is the 8-bit card due to its low cost. More-

over, applications in the smartcard have to share

resources especially the limi ted stat ic memory for

runtime execution. For a contactless smartcard, avail-

ability of RF power is crucial. If the peak power

demanded by the intensive computation is not met or

that the computational duration is longer than what

the power can be sustained from the reader through

induction, then the match ing process will fail. As a

consequence, the user will experience a ‘‘false rejec-

tion’’ even though the rejection is not due to the

outcome of the biometric matching. Moreover, soft-

ware optimization is very crucial for MOC implemen-

tation to achieve good matching performance. The

optimization in term of speed, resource allocation

and code size are necessary during the system design

phase. Nevertheless, ﬁngerprint match-on-card can al-

ready be realized today on an off-the-shelf Java card

having 8-bit, 5 MHz CPU core, 5k bytes RAM and

32Kb EEPROM or better w ith Java OS. The following

optimization methods are commonly employed in

development of match-on-card technology:

1. Reduce the size of the template: Reduce the amount

of information to be matched during on- card

matching can reduce the overall matching time.

For example, ﬁngerprint match-on-card can restrict

the maximum number of minutiae to less than

60 minutiae to be matched per template. However,

information reduction may degrade the accuracy of

the matcher. Developer should be aware of how

much information should be reduced to achieve

acceptable accuracy.

2. Work-sharing biometric match-on-card: Some

low-end smartcards are not able to handle the

whole biometric matching algorithm within

acceptable timing. In this case, work-sharing archi-

tecture which has been introduced previously can

be used to speed up the matching process. The idea

of work-sharing architecture is to assist the smart-

card to compute those computation intensive

functions of the matching algorithm, such as

template alignment, using the biometric terminal.

The ﬁnal biometric comparison, such as the calcu-

lation of matching score, shall be computed inside

the smartcard. The smartcard can send intermedi-

ate data or information other than enrolment

template to the terminal using secure channel.

The developer should be aware of the security

1020

On-Card Matching

requirement to design matching algorithm using

work-sharing architecture for smartcard.

3. Biometric codeword or hashing: Some researchers

investigate algorithms to generate codeword such

as ﬁnger code [13] or biohashing [14 ]. In this case,

the complexity of the matching algorithm in the

smartcard can be reduced. However, the stability of

the biometric codeword is still not as robust as

ﬁngerprint matching due to alignment and defor-

mation of the biometric presentation. Hence,

developer should be aware of whether the accuracy

is sufﬁcient for particular application.

Summary

Biometric match-on-card technolo gy holds great

promise in offeri ng good security and privacy protec-

tion. The technology has come a long way to become

feasible today at an attractive cost and more can still

be done to make it better and cheaper. It provides a

good platform for the launch of a nation wide strong

identity authentication capability which will open up

many other new applications and business possibilities

that will provide better convenience, security and pro-

tection to the users as compared to what is being used

today. The re is also a foothold of this technology in the

global push for machine readable travel documents

which hopefully will lead to a global opportunity in

biometric system level application.

Related Entries

▶ Authentication

▶ Enrolment

▶ Identiﬁcation

▶ Match-on-Card

▶ Veriﬁcation

References

1. Hines, M.: LexisNexis break-in spurs more calls for reform.

ZDNet.com, http://news.zdnet.com/2100-1009-5606911.html

(2005)

2. Vijayan, J.: Scope of credit card security breach expands.

Computerworld, http://www.computerworld.com/securitytopics/

security/ stor y/0,10801,101101,00.html (2005)

3. Scheuermann, D.: Identiﬁcation cards - Integrated circuit cards

with contacts - Part 11: Personal veriﬁcation through biometric

methods, ISO/IEC 7816-11:2004. International Organization for

Standardization/International Electrotechnical Commission

(2006)

4. Allah, M.M.A.: A fast and memory efﬁcient approach for ﬁnger-

print authentication system. IEEE Conf. Adv. Video Signal Based

Surveill. pp. 259–263 (2005)

5. Krivec, V., Birchhauer, J., Marius, W., Bischof, H.: A hybrid

ﬁngerprint matcher in memory constrained environments

2(18-20), 617–620 (2003)

6. Rikin, A.S., Li, D., Isshiki, T., Kunieda, H.: A Fingerprint Match-

ing Using Minutia Ridge Shape for Low Cost Match-on-Card

Systems. IEICE Trans. E88-A(5), 1305–1312 (2005)

7. Mimura, M., Ishida, S., Seto, Y.: Fingerprint veriﬁcation system

on smartcard. In: International Conference on Consumer Elec-

tronics pp. 182–183 (2002)

8. Bistarelli, S., Santini, F., Vaccarelli, A.: An Asymmetric Finger-

print Matching Algorithm for Java Card

TM.

Pattern Anal.

Appl. J. 9(4), 359–376 (2006)

9. Grother, P.W.S., Watson, C., Indovina, M., Flanagan, P. (eds.):

Performance of Fingerprint Match-on-Card Algorithms

Phase II Report. National Institute of Standards and

Technology (2008)

10. Stardards: Information technology - Biometric data interchange

formats - Part 1: Framework, ISO/IEC 19794-1:2006. Interna-

tional Organization for Standardization/International Electro-

technical Commission (2006)

11. Stardards: Information technology - Biometric data interchange

formats - Part 2: Finger minutiae data, ISO/IEC 19794-2:2006.

International Organization for Standardization/International

Electrotechnical Commission (2006)

12. Stardards: Usability of Biometrics in Relation to Electronic Sig-

natures, ISO/IEC JTC1/SC17 N1793. International Organization

for Standardization/International Electrotechnical Commission

(2000)

13. Jain, A.K., Prabhakar, S., Hong, L., Pankanti, S.: Filterbank-

based Fingerprint Matching, IEEE Transactions on Image Pro-

cessing. IEEE Trans.Image Process. 9(5), 846–859 (2000)

14. Teoh, B.J.A., Ngo, C.L.D., Goh, A.: BioHashing: two factor

authentication featuring ﬁngerprint data and tokenised random

number. Pattern Recognit. 37, 2245–2255 (2004)

One-to-Many Identiﬁcation

One-to-many matching is a process of searching a

dataset to match the target image with those of more

than one persons to identify an individual. See also

▶ Identiﬁcation.

If a biometric sample is compared with n templates,

where n is a positive integer, the matching process is

One-to-Many Identiﬁcation

1021

referred to as a 1:n matching. In the case of n =1,itis

equivalent to a veriﬁcation process. If n > 1, it is

regarded as an identiﬁcation process and sometimes

expressly described as 1:many matching.

▶ Biometrics, Overv iew

One-to-One Verification

One-to-many matching is a process of matching the

target image against those of the claimed person to

verify an individual. See also

▶ Veriﬁcation .

▶ Biometrics, Overv iew

Online Learning

▶ Incremental Learning

Open-Set Identification

It is unknown whether the subject presented to the

biometric system for recognition has enrolled in the

system or not. Therefore, the system needs to decide

whether to reject or recognize him as one of the

enrolled subject. It is the opposite of ‘‘Closed-Set

Identiﬁcation.’’

▶ Performance Evaluation, Overview

Operational Tests

Operational tests are those in which a biometric system

collects and processes data from actual system users in

a real ﬁeld application. Operational tests differ funda-

mentally from technology and scenario tests in that the

experimenter has limited control over data collection

and processing. Because operational tests should not

interfere with or alter the operational usage being eval-

uated, it may be difﬁcult to establish ground truth at the

subject or sample level. As a result, operational tests may

or may not be able to evaluate match or enrollment

rates, FRR, or FTE; instead they may be able to directly

measure acceptance rates (without distinction between

genuine and impostor) and operational throughput.

▶ Performance Testing Methodology Standardization

Operational Times

STEPHEN J. ELLIOTT

,ERIC P. K UKULA

ICHARD T. LAZARICK

Department of Industrial Technology, Purdue

University, West Lafayette, IN, USA

Chief Scientist-GSS Identity Labs Computer Science

Corporation Camp Hill, PA

Synonyms

Biometric decision time, and the external operation

time; Biometric subsystem transaction time; Biometric

transaction time; External operation time; Subject in-

teraction time; Total transaction time

Definition

There are a number of associated deﬁnitions for ‘‘time’’

as they relate to an operational biometric system(s).

These ‘‘time’’ metrics include the total transaction

time, the overt biometric transaction time, the subject

interaction time, the biometric subsystem processing/

transaction time, the biometric decision time, and the

external operation time.

Introduction

One deﬁnition of time is given as ‘‘the length of time

taken to complete an activity’’ [1]. For a biometric

system, ‘‘time’’ can be segmented in alignment with a

particular activity or function. Assuming that the bio-

metric model in Fig. 1 is for an access control system,

1022

One-to-One Verification

and that the steps of that model are shown, there are

ﬁve different types of event time. These include the

subject interaction time, the biometric subsystem pro-

cessing time, the biometric decision time, the biomet-

ric transaction time, and the total transaction time.

This document outlines the various timing metrics

that are used in biometric veriﬁcation applications.

They include the overall transaction time, the subject

interaction time, the biometric subsystem processing

time, and the biometric decision time. Although the

document includes examples from a technical contribu-

tion presented to Working Group 5 in the international

biometric standards committee (ISO/IEC JTC 1 SC37)

on the developments of timing metrics in scenario test-

ing for biometric access control systems, the metrics can

be applied to generic biometric veriﬁcation systems.

Total Transaction Time

The total transaction time is a sum of all the subcom-

ponent periods of time associated with the biometric

application system. For a biometric veriﬁcation sys-

tem, the overall transaction time is initiated when the

user makes a claim or presents an identity (i.e., swipes

a card or enters a PIN). The overall transaction time is

completed when the last measurable component has

been satisﬁed, in the physical access control system

when the door strike is activated.

Biometric Transaction Time

This begins with the biometric sample presentation

and ends with the biometric decision. Therefore, this

includes the presentation of the biometric trait portion

of the subject interaction time, biometric subsystem

processing time, which includes sample acquisition

and sample processing time, and the biometric deci-

sion time.

Subject Interaction Time

Using the same access control system as described

earlier, the subject interaction time commences when

a claim of identity is made (or presented), that is,

swiping a card or entering a PIN by the user. The

time ends when the in dividual has presented hi s/her

biometric characteristic(s) and the sensor begins to

acquire the sample.

Biometric Subsystem Processing Time

The biometric subsystem processing time is the time

taken for the system to acquire the biometric sample, to

evaluate the quality of the sample, and if the quality is

satisﬁed, to process that sample for comparison. For the

samples of bad quality, the biometric system requests

the subject to submit the biometric trait. The biometric

subsystem processing time ends when either a compari-

son score or a request for re-submission is generated.

Biometric Decision Time

The biometric decision time is the time required by the

biometric subsystem to generate an accept or reject

response based on the comparison score and the deci-

sion logic. The decision logic could be a simple thresh-

old or a more complex methodology such as fusion

logic. In biometric identiﬁcation where the biometric

subsystem generates a list of matched candidates, the

biometric decision time is the time required to search

Operational Times. Figure 1 Types of transaction times.

Operational Times

1023

the subject from the database of enrolled subjects. In

this case, it depends on both t he size of the database

and the strategy in search, for example, the classiﬁca-

tion or binning of biometric samples.

External Operation Time

Pro vided that the biometric decision is an accept,

the external operation time is the time required to

complete the application transaction. In the physical

access control system, it is the time required by the phy-

sical electro-mechanical components to act according

to the decision of the biometric subsystem and other

access control privilege criteria to complete the access

transaction. In biometric identiﬁcation, the external op-

eration time is not required because the external opera-

tion is typically a manual function.

Timing Illustrations

According to a Technical Contribution from the ISO/

IEC JTC 1 SC 37 WG5 Special Group on 19795–5 [2],

timing can be related to the type of testing to be

undertaken. There are three major types of testing:

technolog y, scenario, and operational.

For the purpose of evaluating the performance

of a biometric subsystem in a scenario evaluation, it

is appropriate to deﬁne a metric that only measures the

biometric subsystem performance and is not inﬂu-

enced by the presence or performance of other access

control devices, policies, and functions. Thus, the dis-

tinction is made between biometric transaction time

and total transaction time. This distinction is made in

timing diagrams illustrating the components of a bio-

metric transaction and a total access control system

transaction shown in Figs. 2–4.

Operational Times. Figure 2 Timeline showing total and biometric subsystem transaction times for a one-attempt

transaction.

Operational Times. Figure 3 Timeline showing total and biometric subsystem transaction times for a one-attempt

transaction with longer access control time.

Operational Times. Figure 4 Timeline showing total and biometric subsystem transaction times for a two-attempt

transaction.

1024

Operational Times

Related Entries

▶ Attempt

▶ Failure to Acquire (FTA)

▶ Failure to Enroll (FTE)

▶ General Biometric Model

▶ Match rates

▶ Performance Testing and Evaluation – Technology,

Scenario, and Operational Testing

▶ Presentation

▶ Threshold

▶ Transaction

References

1. Soanes, C., Hawker S.: Compact Oxford English Dictionary of

Current English (3rd ed.). Oxford University Press, New York

(2005)

2. International Organization for Standardization.: ISO/IEC JTC1

SC37 Comments from the Special Group Established to Review

19795–5 (N-1770) (No. SC37N1853). Geneva: ISO/IEC (2006)

Optical Flow

Optical ﬂow is a vector ﬁeld that represents the motion

of pixels between two images. A variety of algorithms

have been developed for computing optical ﬂow. It is

commonly computed for video, in which the motion

from one frame to the next is relatively small.

▶ Face Variation

▶ Gait Recognition, Silhouette-Based

Optical Target

Optical target refers to an object the optical character-

istics of which are well known. It is used to perform an

optical calibration of lenses or illuminators.

▶ Biometric Sensor and Device, Overview

Optimal Hyperplane

▶ Support Vector Machine

Optimization

Given a cost function, different strategies could be used

to obtain the estimate. This is called the optimization

strategy and the solution often depends upon the exact

strategy that is used.

▶ Face Tracking

Ordinal Measure

In ordinal measure, variables are required to be mono-

tonically related so that they can be rank-ordered. In

palmprint identiﬁcation, ordinal measure qualitatively

compares neighborhood image pixels or regions and

preserves their ordinal relationship. This yields a sym-

bolic representation of the relations.

▶ Palmprint Features

Orthographic Scanning

The procedure of recording two-dimensional images

of an object with the goal of capturing the object’s

three-dimensional structure. In a hand-geometry de-

vice, for example, it is performed with the help of a

mirror that projects the lateral surface of the hand into

the visual ﬁeld of the camera and allows it to record

both the side and top views of the hand in a single

image.

▶ Hand-Geometry Device

Orthographic Scanning

1025

Osmology

Osmology is the study of odors and the sense of smell;

phrase coined by the Polish Forensic Police for the ﬁeld

of science where trained dogs compare scent traces that

the perpetrator of a crime leaves at the crime scene

to the odor of a person suspected of that crime in

a line-up procedure.

▶ Odor Biometrics

Output Noise Variance (ONV)

The output noise variance is the variance of the noise

of a correlation plane. If the input noise is stationary

and w ith Gaussian distribution, the output noise

on correlation plane is also going to be stationary

and Gaussian distributed. The output noise variance

describes the output noise variance at all pixels in the

output. If we would like to see a noise-tolerant peak on

the correlation plane for the authentic comparison, it

is desirable to make the noise variance on the correla-

tion plane as small as possible on average. That is why

would like to incorporate the criteria for minimizing

ONV in the design of the correlation ﬁlters.

▶ Iris Recognition Using Correlation Filters

Outsole Pattern Matching

▶ Footwear Recognition

Overfitting

The phenomenon that the learning result performs

very good on training data but poorly on unseen new

data, which is caused by that the learning approach has

ﬁt the training data too much, such that some malign

particularities that prevent good generalization has

also been captured by the learning result.

▶ Ensemble Learning

▶ Mainfold Learning

1026

Osmology

Paleoanthropology

Paleoanthropology is the study of ancient human

beings, with the evidences such as bones, and

footprints.

▶ Skull, Forensic Evidence of

Palm Dorsal Vein

▶ Hand Veins

Palm Segment

It refers to one of the three palm portions: lower palm,

upper palm, and writer palm.

▶ Fingerprint, Palmprint, Handprint and Soleprint

Sensor

Palm Vein

MASAKI WATANABE

Fujitsu Laboratories Ltd, Kawasaki, Japan

Synonyms

Palm vein authentication; Palm vein recognition

Definition

Palm vein authentication is one of a modality of

biometric authentications, and is classiﬁed as a phy-

siological biome tric authentication. It uses palm vein

patterns, a vascular image of a person’s palm which can

be seen as a kind of pattern, as personal information.

The palm vein patterns are normally captured

under near-infrared illumination using the reﬂection

method, in which

▶ near infrared rays are emitted

from a person’s palm and the reﬂected light is cap-

tured. The places on the palm where veins occur are

captured as dark parts because veins absorb more near-

infrared illumination while emitting only little. With

this reﬂection method, a

▶ contactless type of pattern-

capturing and user identiﬁcation can be realized.

Because veins are inside the human body, they are

secure and hard to be stolen or duplicated. Moreover,

because palm vein patterns are varied and complex,

they have sufﬁcient information to identify one indi-

vidual among many people and palm vein authentica-

tion is highly accurate.

A contactless type of identiﬁcation is suitable for

applications that require a high level of hygiene and

for public-use applications. Several banks in Japan

use palm vein authentication to identify customers

since July 2004. In addition, the method has been

used in a variety of applications including door secu-

rity systems of ofﬁces and condominiums, login man-

agement systems for PCs, and to identify patients

in hospitals.

Introduction

Palm vein authentication is one of the vascular pattern

authentication technologies. It uses palm vein patterns,

which are difﬁcult to be seen by human on a person’s

palm, as personal information. Because palm vein pat-

terns are information that are found within someone’s

body, it is hard for that information to be stolen.

2009 Springer Science+Business Media, LLC

This means that forgery is very difﬁcult under usual

conditions.

Palm vein patterns are unique to each individual;

even identical twins have different palm vein patterns.

Furthermore, palm vein patterns do not change within

lifetime of a human except in certain cases of inju ry or

disease. Althoug h these facts and any other inﬂuences

including physiological growth have not been medi-

cally proven, as with the ﬁngerprints, irises, and other

internal identiﬁcation methods, experimental results

based on extensive data and large-scale operational

results prove practically that palm vein authentication

has the advantages of consistency and accuracy as a

method of personal identiﬁcation.

A patent for hand vein authentication was ﬁled in

1985 by Joseph Rice in the United States [1]. The ﬁrst

device for palm vein authentication was presented by

Advanced Biometrics, Inc. in the United States in 1997,

and in 2003, a remarkable contactless device was re-

leased by Fujitsu in Japan. In 2004, Japanese ﬁnancial

institutions adopted Fujitsu’s technology for conﬁrm-

ing the identiﬁcation of its customers. This was the

ﬁrst major app lication in Japan in which a private

enterprise adopted biometric authentication in a ser-

vice for the general public. Palm vein authentication

and ﬁnger vein authentication have received a great

deal of attention in Japan compared to other biometric

authentication methods such as ﬁngerprint, iris, and

face recognition methods.

Fujitsu’s implementation of a contactless sensor

and its concept was awarded the ‘‘Wall Street Journal’s

2005 Technology Innovation Award for Security in

Networks’’ [2].

Palm Vein

A person’s palm has a w ide and complex vein pattern,

and it contains sufﬁ cient information to identify an

individual among many people. If other parts of the

hand are used for authentication, additional informa-

tion, such as the relative position of that part of the

hand relative to a vein sensor will be needed because

the vein pattern of other parts of the hand does not

contain sufﬁcient information for identiﬁcation.

Compared with the back of the hand or the back of

a ﬁnger, the palm is a good area to use because it does

not normally have any hair on it to obscure the vein

pattern.

Sensing

Vein patterns in the subcutaneous tissue of a person’s

palm are captured using near-infrared rays. This tech-

nology is called near-infrared spectroscopy (NIRS) and

imaging. This has been investigated as a technology of

in vivo measurements in the last ten or so years [3].

Hemoglobin is grouped into two types; (1) oxyge-

nated hemoglobin that is present in arteries, which

contains oxygen; and (2) deoxygenated hemoglobin

present in veins that does not contain oxygen, and in

particular it absorbs light with a wavelength of about

760 nm (Fig . 1)[4, 5]. When capturing an image of a

palm using near-infrared rays which include the wave-

length of light, the veins appear like a shadow on the

palm, seen to be darker than the surrounding area

(Fig. 2). In other experiments using near-infrared

rays with a wavelength of 880 nm , a vein with a

1 mm diameter could be seen even if it is 3 mm

below the surface of the skin [6].

Palm vein patterns are captured using a reﬂection

method, which illuminates a person’s palm from the

front and also captures an image from the front of the

palm. A palm image can be captured using light trans-

mission method, which illuminates the palm from the

back of the hand and captures an image from the front

[6]. While in the transmission method the illumina-

tion device and the capture device are separated facing

each other across a palm, in the reﬂection method,

the illumination device and the capture device can

be gathered together more compactly because the

Palm Vein. Figure 1 Absorption spectra of hemoglobin.

(Adapted from Wray et al. (1988) by K. Shimizu,

Hokkaido University.)

1028

Palm Vein

direction of the illumination is the same as the direc-

tion of image capturing.

In Fujitsu’s implementation [7, 8], a palm vein

authentication sensor is made in the shape of a small

box 35 mm deep  35 mm wide  27 mm high.

Capturing is executed in a contactless manner. Users

do not need to touch the sensor; they only have to hold

their palms above it. To obtain a clear image of the

palm vein pattern of a hand ﬂoating in the air, the

capturing is controlled according to the movement

and height of the hand above the sensor, and the

illumination is controlled according to the light

around the sensor.

In contactless manner, worries of the user regarding

their sensitiveness about hygiene or any emotional

concerns can be eliminated. It enables the identiﬁca-

tion method to be used in environments where a high

standard of hygiene is required, such as in hospitals or

food factories. In addition, sufﬁcient consideration is

given to individuals who are reluctant to come into

direct contact with publicly used devices.

The intensity of the near-infrared ray emitted from

the sensor is also safe. It is lessser than the intensity

speciﬁed in the ‘‘Light and Near-Infrared Radiation’’

guidelines of the American Conference of Governmen-

tal Industrial Hygienists (ACGIH).

Matching

When ﬁrst matching a palm to a database of images, a

palm vein pattern is extracted from a near-infrared

image. The pattern is picked up among the dark

lines, which are obtained by morphologically tracing

valleys of brightness in the palm area in an image.

The similarity of the captured vein pattern with one

in a database is given a score to identify whether the

vein pattern s that have been registered are the same as

that has been captured from a person being authenti-

cated. This rating is based on the sum of the Euclidean

distance between pixels that compose the two palm

vein patterns.

For a veriﬁcation (one-to-one matching), by way

of example, the identity of the user is proved if the

similarity score is greater than or equal to a predeter-

mined threshold. Otherwise, the user is regarded as an

imposter.

For an identiﬁcation (one-to-many matching),

similarity scores are calculated between the palm vein

pattern captured from the person being identiﬁed

and all or some palm vein patterns that have been

registered in a database, and the identity of the palm

vein pattern which has the maximum similarity score

is regarded as being the one of the person being

identiﬁed.

Performance

Using the data of 150,000 palms from 75,000 people

[7], it is suggested that a typical palm vein authentica-

tion system can achieve false acceptance rate of less

than 0.00008% and a false rejection rate of 0.01%,

provided that the palm is held over the sensor two

times during registration and one retry is allowed for

comparison during authentication [7]. The se results

are certainly promising and conﬁrm the individuality

of ﬁnger vein features in a large user population.

In addition, ability of the sensor to perform pe r-

sonal authentication w as veriﬁed using the following

data: (1) data from individuals ranging from 5 to

85 years old, including people in various occupations,

in accordance with the demographics released by the

Statistics Center of the Statistics Bureau in Japan;

(2) data from foreigners living in Japan in accordance

with world demographics released by the Unit ed

Nations; (3) data that trace daily ch anges in the

palm vein pattern over several years; and (4) data

taken in various situations in daily l ife, f or example

after drinking alcohol, taking a bath, going outside, or

waking up.

Palm Vein. Figure 2 Infrared ray image of a palm.

Palm Vein

1029