biometricauthentication.doc

8/17/2019 biometricAuthentication.doc

1/33


2/33

Abstract

Biometric systems have been researched intensively by many organization and

institution. It overcomes the conventional security systems by identify “who you are”.

This paper discusses the current image based biometric systems. It first gives some

information about why biometric is needed and what should people look for in biometric

systems. everal popular image based biometric systems have been e!amined in this

paper. The techni"ue used in each system for data ac"uisitions# feature e!traction and

classifiers are briefly discussed. The biometric systems included are face# fingerprint#

hand geometry# hand vein# iris# retina and signature. The paper concludes by e!amining

the benefits of multi$modal biometric systems# it is found that there is no one good

biometric systems each have its advantages and disadvantages and the performance of

each biometric system is summarized.

%


3/33

Introduction

&s technology advances and information and intellectual properties are wanted by many

unauthorized personnel. &s a result# many organizations have being searching ways for

more secure authentication methods for user access. 'urthermore# security has always

been an important concern to many people. 'rom Immigration and (aturalization ervice

)I(* to banks# industrial# military systems# and personal are typical fields where security

is highly valued. It is soon realized by many# that traditional security and identification

are not sufficient enough# people need to find a new authentic system in the face of new

technological reality +%,.

-onventional security and identification systems are either knowledge based like a

social security number or a password# or token based such as keys# I/ cards. The

conventional systems can be easily breached by others# I/ cards and passwords can be

lost# stolen or can be duplicated. In other words# it is not uni"ue and not necessary

represent the rightful user. Therefore# biometric systems are under intensive research for

this particular reason.

What is Biometric

umans recognize each other according to their various characteristics for ages. 1eople

recognize others by their face when they meet and by their voice during conversation.

These are part of biometric identification used naturally by people in their daily life.

Biometrics relies on “something you are or you do”# on one of any number of uni"ue

characteristics that you can2t lose or forget. It is an identity verification of living# human

individuals based on physiological and behavioral characteristics. In general# biometric

system is not easily duplicated and uni"ue to each individual. It is a step forwards from

identify something you have and something you know# to something you are + 3,.

General biometric system

Biometrics uses physical characteristics# defined as the things we are and personal traits#

it can consists of following +4rror5 6eference source not found,#

3


4/33

Table %. Biometric characteristics.

1hysical characteristics 1ersonal traits

• chemical composition of body

odor

• facial features and thermal

emissions

• features of the eye $ retina and

iris

• fingerprints

• hand geometry

• skin pores

• wrist7hand veins

• handwritten

signature

• keystrokes or

typing

• voiceprint

ame as many recognition systems# a general biometric system can consists of following

sections# data collection# transmission# signal processing storage and decision +8,# see

'igure %. It can considered that each section function independently# and errors can be

introduced at each point in an additive way.

'igure %. 9eneralized biometric system.

/ata collection consists of sensors to obtain the raw biometric of the sub:ect# and can

output one or multidimensional signal. ;sually# data are obtained in a normalized

fashion# fingerprints are moderately pressed and rotation is minimized# faces are obtained

in frontal or profiled view# etc. /ata storage is usually separated from point of access#

therefore the data have to be transmitted or distributed via a communication channel. /ue

to bandwidth# data compression may be re"uired. The signal processing module takes the

8


5/33

original biometric data and converts it into feature vectors. /epend on the applications#

raw data might be stored as well as the obtained feature vectors. The decision subsystem

compares the measured feature vectors with the storage data using the implemented

system decision policy. If measures indicate a close relationship between the feature

vector and compared template# a match is declared +4rror5 6eference source not found,.

'alse matching and false non$matching error can occur# although for different systems

error e"uation varied# a general e"uation can be developed +4rror5 6eference source not

found,+ be the number of independ biometric measures the probability of false

match FMRSR against any single record can be given by#

∏==

M

j

j jSR FMR FMR%

*)τ

?here *) j j FMR τ e"ual single comparison false match rate for the jth biometric and

threshold τ . The probability for not making any false match in comparison in multiple

records can be e!pressed as#

PN

SYS FMR FMR *%)% −=−

?here FMRSYS is the system false match rate# and N and P is number of records and

percentage of the database to be searched respectively. 'or the single record false non$

match rate#

∏=

−=− M

j

j jSR FMR FNMR%

*,)%+% τ

>ore commonly used biometric system reliability inde!es are '66 )'alse 6e:ect 6ate*

which is the statistical probability that the system fails to recognize an enrolled person

and '&6 )'alse &ccept 6ate* which is the statistical probability that an imposter is

recognized as an enrolled person. '66 and '&6 are inversely dependent on each other

and as within modern biometric systems identification5

@,A@#000%.0+

@,%.0@#000%.0+

∈

∈

FAR

FAR

They are very reliable# promising# universal and tampering resistant.


6/33

Image based biometric techniques

There are many biometric systems based on different characteristics and different part of

the human body. owever# people should look for the following in their biometrics

systems +A,#

;niversality $ which means that each person should have the characteristic

;ni"ueness $ which indicates that no two persons should be the same in

terms of the characteristic

1ermanence $ which means that the characteristic should not be changeable

-ollectability $ which indicates that the characteristic can be measured

"uantitatively

'rom the above# various image based biometric techni"ues has been intensively studied.

This paper will discuss the following techni"ues# face# fingerprints# hand geometry# hand

veins# iris# retina and signature.

Face

'ace recognition technology )'6T* has applications in many wide ranges of fields#

including commercial and law enforcement applications. This can be separate into two

ma:or categories. 'irst is a static matching# e!ample such as passport# credit cards# photo

I/2s# driver2s licenses# etc. econd is real$time matching# such as surveillance video#

airport control# etc. In the psychophysical and neuroscientific aspect# they have concerned

on other research field which enlighten engineers to design algorithms and systems for

machine recognition of human faces. & general face recognition problem can be# given

still or video images of a scene to identify one or more persons in the scene using a stored

database of faces +,. ?ith additional information such as race# age and gender can help to

reduce the search. 6ecognition of the face is a relatively cheap and straightforward

method. Identification based on ac"uiring a digital image on the target person and

analyzing and e!tracting the facial characteristics for the comparison with the database.

Carhunen$=oeve )C=* e!pansion for the representation and recognition of faces is said to

generate a lot of interest. The local descriptors are derived from regions that contain the

eyes# mouth# nose# etc.# using approaches such as deformable templates or eigen$

A


7/33

e!pansion. ingular value decomposition )D/* is described as deterministic counterpart

of C= transform. &fter feature e!traction# recognition is done# early approach such as

feature point distance or nearest neighbor rule is used# and later eigenpictures using

eigenfaces and 4uclidean distance approach is e!amined. Ether methods using yperB'

network a neural network approach# dynamic link architecture# and 9abor wavelet

decomposition methods are also discussed in +4rror5 6eference source not found,.

& back$propagation neural network can be trained to recognize face images. owever# a

simple network can be very comple! and difficult to train. & typical image recognition

network re"uires N F m ! n input neurons# one for each of the pi!els in an m ! n image.

These are mapped to a number of hidden$layer neurons# p +G,. These in turn map to n

output neurons# at least one of which is e!pected to fire on matching a particular face in

the database. The hidden layer is considered to be a feature vector.

4igenfaces is an application of principal component analysis )1-&* of an n$dimensional

matri!. tart with a preprocessed image I ) x# y*# which can be considered as vector of

dimension N 3. &n ensemble of images then maps to a collection of points in this huge

space. The idea is to find a small set of faces )eigenfaces* that can appro!imately

represent any point in the face space as a linear combination. 4ach of the eigenfaces is of

dimension N ! N # also can be interpreted as an image +4rror5 6eference source not

found,. &n image can be reduced to an eigenvector ib B = which is the set of best$fit

coefficients of an eigenface e!pansion. 4igenvector is then used to compare each of those

in a database through distance matching# such as -artesian distance.

9abor wavelet is another widely used face recognition approach# it can be described by

the e"uation#

*e!p)3

HHe!p*)

3

33

# xk i

xk x

k

−=

σ ψ

σ

where k is the oscillating fre"uency of the wavelet# and the direction of the oscillation. σ

is the rate at which the wavelet collapses to zero as one moves from its center outward.

The main idea is to describe an arbitrary two$dimensional image function I ) x# y* as a


8/33

linear combination of a set of wavelets. The x# y plane is first subdivided into a grid of

non$overlapping regions. &t each grid point# the local image is decomposed into a set of

wavelets chosen to represent a range of fre"uencies# directions and e!tents that “best”

characterize that region +4rror5 6eference source not found,. By limiting k to a few

values# the resulting coefficients become almost invariant to translation# scale and angle.

The finite wavelet set at a particular point forms a feature vector called a jet # which

characterize the image. &lso with elastically distorted grid# best match between two

images can be obtained +4rror5 6eference source not found,.

Fingerprints

Ene of the oldest biometric techni"ues is the fingerprint identification. 'ingerprints were

used as a means of positively identifying a person as an author of the document and areused in law enforcement. 'ingerprint recognition has a lot of advantages# a fingerprint is

compact# uni"ue for every person# and stable over the lifetime. & predominate approach

to fingerprint techni"ue is the uses of minutiae +,# see 'igure 3.

'igure 3. >inutiae# ridge endings and ridge bifurcations.

The traditional fingerprints are obtained by placing inked fingertip on paper# now

compact solid state sensors are used. The solid state sensors can obtain patterns at 800 !

G


9/33


10/33

ABG


11/33

Ether enhancement algorithm such as preprocessing# mathematic algorithm and etc# have

been discussed by +%3,# +%8, and +%edium cost# only needs a platform and a low7medium resolution --/ camera.

3. It uses low$computational cost algorithms# which lead to fast results.

8. =ow template size5 from J to 3A bytes# this reduces the storage needs.


12/33


13/33


14/33


15/33

(ormalize the background Brightness surface is subtracted# leaving only the vein

structure and background

Threshold out the vein pattern >orphological gradient is applied to obtain a threshold

value that separates the vein and background

6emove artifacts# fill holes Binary alternating se"uential filter# is used to remove

threshold artifacts and fill holes in vein structure

Thin the patter down to its medial a!is >odified hang and uen algorithm is used1rune the medial a!is &utomatic pruning algorithm is used

It is noted that significant horizontal positional noise during docking for different

registration process +4rror5 6eference source not found,. The proposed matching

approach compares medial a!is and coding algorithm# $!n%trained %e&uential $!rrelati!n.

It is a variation on the traditional correlation methods used for template matching. The

reference or library signature is first dilated by a he!agonal structuring element. The test

signature is then superimposed on this reference buffer and the percentage of pi!els

contained within the buffer determined. /ue to horizontal translation error# test signature

is se"uentially translated horizontally and compared against the reference buffer + 4rror5

6eference source not found,. The horizontal translation is limit to S 80 pi!els. The

highest match percentage is said to be the forward similarity# where reverse similarity is

obtained by dilate the test signature and reference signature is se"uentially correlated

until the ma!imum measure is obtained. By setting the forward and reverse minimum

percentage to GA@ and 0@ respectively# the resultant '66 is G.A@ and '&6 is 0@. If

forward percentage is lowered to G0@# '66 improved down to A@ and '&6 remains the

same +4rror5 6eference source not found,.

'igure . =eft is original captured image# right is vein structure after prune the medial a!is

-onventional method uses low pass filter follow by high pass filter# after that threshold is

applied with bilinear interpolation and modified median filter to obtain hand vein in

%


16/33

region of interest )6EI* +4rror5 6eference source not found,. The 9aussian low pass filter

is a 8 ! 8 spatial filter with e"uation#

∑=

=J

%

*)*)*A)i

i ' i( '

JG

BA<

83%

JG

BA<

83%

( ( (

( ( (

( ( (

' ' '

' ' '

' ' '

ima)e Filtered ×=

/ue to heavy computation load +4rror5 6eference source not found, introduce a way of

enhancing the algorithm. Both the coefficients for the 9aussian low pass filter and the

low pass filter are designed to have G$tap -/ )canonical signed digit* codes at the

ma!imum. &lso# for the normalization# the decimation method is used. It is said that#

-/ code is an effective code for designing a 'I6 filter without a multiplier. The proposed preprocessing algorithm follows the same steps as the conventional method#

e!cept that the coefficients for each filter are made of -/ codes. The general -/ code

is e"ual to#

∑=

−= j M

i

i

i j S (

%

3

where j F %# 3# U# %3%# S i∈O$%# 0# %N and M is an integer. Instead a 8 ! 8 9aussian low

pass filter# a %% ! %% spatial filter is applied instead. Is it found that 9aussian filter is

J


17/33

'igure J. )a* Iris image )b* iris localization )c* unwrapped te!ture image

)d* te!ture image after enhancement

Iris is highly randomized and its suitability as an e!ceptionally accurate biometric derives

from its +30,#

• e!tremely data$rich physical structure

• genetic independence# no two eyes are the same

• stability over time

• physical protection by a transparent window )the cornea* that does not inhibit

e!ternal view ability

There are wide range of e!traction and encoding methods# such as# /augman >ethod#

multi$channel 9abor filtering# /yadic wavelet transfmor + 3%,# etc. &lso# iris code is

calculated using circular bands that have been ad:usted to conform to the iris and pupil

boundaries. /augman is the first method to describe the e!traction and encoding process

+33,. The system contains eight circular bands and generates A%3$byte iris code# see 'igure

%0 +4rror5 6eference source not found,.

%


18/33

a* b*

'igure %0. a* /ougman system# top circular band# bottom iris code b* demodulation code

&fter boundaries have been located# any occluding eyelids detected# and reflections or

eyelashes e!cluded# the isolated iris is mapped to size$invariant coordinates and

demodulated to e!tract its phase information using "uadrature 3/ 9abor wavelets +4rror5

6eference source not found,. & given area of the iris is pro:ected onto comple!$valued

3/ 9abor wavelet using#

∫ ∫ −−−−−−= ρ φ

β φ θ α ρ φ θ φ ρ ρ φ ρ d d eee I * r i+ 33

033

00 7*)7*)*)

ImNO6e#ImNO6e# *#)sgn

where *O6e#ImN can be regarded as a comple!$valued bit whose real and imaginary parts are

either % or 0 )sgn* depending on the sign of the 3/ integral. I ) ρ #φ * is the raw iris image in

a dimensionless polar coordinate system that is size$ and translation$invariant# and which

also corrects for pupil dilation. α and β are the multi$scale 3/ wavelet size parameters#

spanning a $fold range from 0.%Amm to %.3mm on the iris# and w is wavelet fre"uency

spanning 8 octaves in inverse proportion to β . )r 0#θ 0* represent the polar coordinates of

each region of iris for which the phasor coordinates *O6e#ImN are computed +4rror5

6eference source not found,. 3#0


19/33

'or two identical iris codes# the / is zeroR for two perfectly unmatched iris codes# the

/ is %. 'or different irises# the average / is about 0.A +4rror5 6eference source not

found,. The observed mean / was p F 0.


20/33

e!tracted as te!ture features. Enly five low resolution levels# e!cluding the coarsest level#

are used. This makes the 3 e!tracted features robust in a noisy environment +4rror5

6eference source not found,. ?eighted 4uclidean /istance is used as classifier#

∑=

−=

N

ik

i

k

ii . . k (/-%

3*)

3*)

*)*)*)

δ #

where . i denotes the ith feature of the unknown iris# . i

)k * and δi0k1 denots the ith feature and its

standard deviation of iris k # N is the total number of features e!tracted from a single iris.

It is found that# a classification rate of J8.@ was obtained when either all the


21/33


22/33

'igure %3. Vuantities for testing curvilinear bands.

where p and n represent a vessel candidate and one of the eight neighbors from its

neighborhood N p# respectively# e p and en are their corresponding nearest background

pi!el. The two measures are defined by#

n p N n

n p

n p

N nn p

N n

eed

e pe pe pe pe pe pan)le

p

p p

∈

∈∈

=

⋅⋅⋅==

ma!

arccos%I0ma!*#)ma!π

φ

The overall improvement result can be seen in 'igure %8 below#

'igure %8. 1roposed approach versus global thresholding.

3%


23/33

Signature

ignature differ from above mentioned biometric system# it is a trait that characterize

single individual. ignature verification analyzes the way a user signs his or her name.

This biometric system can be put into two categories# on$line and off$line methods. En$

line methods take consideration of signing features such as speed# velocity# rhythm and

pressure are as important as the finished signature2s static shape +3,. ?here as# off$line

classification methods are having signature signed on a sheet and scanned. 1eople are

used to signatures as a means of transaction$related identity verification# and most would

see nothing unusual in e!tending this to encompass biometrics. ignature verification

devices are reasonably accurate in operation and obviously lend themselves to

applications where a signature is an accepted identifier +4rror5 6eference source not

found,. Darious kinds of devices are used to capture the signature dynamics# such as thetraditional tablets or special purpose devices. pecial pens are able to capture movements

in all 8 dimensions. Tablets are used to capture 3/ coordinates and the pressure# but it has

two significant disadvantages. ;sually the resulting digitalized signature looks different

from the usual user signature# and sometimes while signing the user does not see what

has been written so far. This is a considerable drawback for many )une!perienced* users.

& proposed off$line classification method to compensate the less information is raised by

+4rror5 6eference source not found,. The proposed method utilizes idden >arkov

>odels )>>* as the classifiers. Before# >> is applied# scanned signature image

have to go through the following#

%. (oise filtering# to remove the noise including noise added by scan process.

3. -orrecting the inclination of the sheet in the scanner.

8. Binarization of the graphic.


24/33

the straight direction line is fished. The direction vector usually have 800 elements

+4rror5 6eference source not found,.

'igure %


25/33

useful as a method for the selection of features for signature verification +4rror5

6eference source not found,.

&lternative approach could be wavelet base method# where the signature to be tested is

collected from an electronic pad as two functions in time ) x)t *# y)t **. It is numerically

processed to generate numbers that represent the distance between it and a reference

signature )standard*# computed in a previous enrolment stage. The numerical treatment

includes resampling to a uniform mesh# correction of elementary distortions between

curves )such as spurious displacements and rotations*# applying wavelet transforms to

produce features and finally nonlinear comparison in time )/ynamic Time ?arping*.

The decomposition of the functions x)t * and y)t * with wavelet transform generates

appro!imations and details like those showed in 'igure %A to an original e!ample of x)t *

+4rror5 6eference source not found,.

'igure %A. 'unction x)t * after wavelet transform.

4ach zero$crossing of the detail curve at the < th level of resolution )this level was chosen

empirically# by trial and error*# three parameters are e!tracted5 its abscissa# the integral

between consecutive zero$crossings#

∫ −

=k

k

'4

'4

k dt t (-i

%

*)<

and the corresponding amplitude to the same abscissa in the appro!imation function at 8 rd

level#

*)8 k k 5$(Aa =

&s it has been demonstrated that this information suffices to a complete reconstruction of

the nontransformed curve +4rror5 6eference source not found,. Before measuring

3


26/33


27/33

assumptions by using different appro!imations +80,. In +4rror5 6eference source not

found, it is discussed that different threshold or weights can be given to different user# to

reduce the importance of less reliable biometric traits. It is found by doing this# '66 can

be improved. &s well it can reduce the failure to enroll problem by assigning smaller

weights to those noisy biometrics. &lso# in +4rror5 6eference source not found,# the

proposed integration of face and fingerprints overcomes the limitations of both face$

recognition systems and fingerprint$verification systems. The decision$fusion scheme

formulated in the system enables performance improvement by integrating multiple cues

with different confidence measures# with '66 of J.@ and '&6 of 0.00%@. Ether fusion

techni"ues have been mentioned in +4rror5 6eference source not found,# these are Bayes

theory# clustering algorithms such as fuzzy C$means# fuzzy vector "uantization and

median radial basis function. &lso vector machines using polynomial kernels and

Bayesian classifiers )also used by +8%, for multisensor fusion* are said to outperform

'isher2s linear discriminant +4rror5 6eference source not found,. (ot only fusion between

biometric# fusions within a same biometric systems using different e!pert can also

improve the overall performance# such as the fusion of multiple e!perts in face

recognition +83, and +88,.

Conclusion

/epend on application different biometric systems will be more suited than others. It is

known that there is no one best biometric technology# where different applications re"uire

different biometrics +4rror5 6eference source not found,. ome will be more reliable in

e!change for cost and vise versa# see 'igure % +8


28/33

'igure %. -ost vs accuracy

1roper design and implementation of the biometric system can indeed increase the overall

security. 'urthermore# multiple biometric fusions can be done to obtain a relative cheaper

reliable solution. The imaged base biometric utilize many similar functions such as 9abor

filters and wavelet transforms. Image based can be combined with other biometrics to

give more realible results such as liveliness )4-9 biometric* or thermal imaging or 9ait

based biometric systems. & summary of comparison of biometrics is shown in table

below +4rror5 6eference source not found,#

Table edium igh igh

%ignature igh -hanging

signatures

igh Dery high >edium >edium

Face >edium =ighting# age# hair#glasses

igh >edium >edium >edium

3G


29/33


30/33


31/33

4!n.eren$e# 3# A08 $A0.

%3 -onnell# M..# 6atha# (.C.# Bolle# 6.>. 3003. 'ingerprint image enhancement using weak models.

Ima)e Pr!$e%%in); =>>=; Pr!$eedin)%; =>>= Internati!nal 4!n.eren$e# %# I$>; Pr!$eedin)%; 8?t* Internati!nal 4!n.eren$e# arcos# &. 3000. Biometric identification through

hand geometry measurements. Pattern Analy%i% and Ma$*ine Intelli)en$e6 I/// #ran%a$ti!n%# 33 )%0*#

Ect# %% $%%G%.

% anchez$6eillo# 6. 3000. and geometry pattern recognition through 9aussian mi!ture modeling.

Pattern Re$!)niti!n6 =>>>; Pr!$eedin)%; 8?t* Internati!nal 4!n.eren$e# 3# J8G $J.# mith# -.=. %JJA. Thermographic imaging of the subcutaneous vascular network of the

back of the hand for biometric identification. Se$urity #e$*n!l!)y6 899?; Pr!$eedin)%; In%titute !.

/le$tri$al and /le$tr!ni$% /n)ineer% =9t* Annual 899? Internati!nal 4arna*an 4!n.eren$e# %$30

Ect# 30 $8A.

% ang Cyun Im# yung >an 1ark# et . al . 300%. &n biometric identification system by e!tracting hand

vein patterns. !urnal !. t*e K!rean P*y%i$al S!$iety. >arch# 8 )8*# 3$G3.

%J Zong hu# Tieniu Tan# Zunhong ?ang. 3000. Biometric personal identification based on iris

patterns. Pr!$eedin)% 8?t* Internati!nal 4!n.eren$e !n Pattern Re$!)niti!n. 3# 0%$.# -hmielewski# T.&.# Mr.# et . al . 3000. &n iris biometric system for public and personal use.

4!mputer # 88 )3*# 'eb# G0 $GA.


32/33


33/33

Internati!nal 4!n.eren$e# esser# C. 3003. 'usion of multiple e!perts in multimodal biometric personal identity

verification systems. Neural Net+!rk% .!r Si)nal Pr!$e%%in)6 =>>=; Pr!$eedin)% !. t*e =>>= 8=t* I///

(!rk%*!p# 8 $%3.

88 -zyz# M.# Cittler# M.# Dandendorpe# =. 3003. -ombining face verification e!perts. Pattern

Re$!)niti!n6 =>>=; Pr!$eedin)%; 8Ct* Internati!nal 4!n.eren$e# 3# 3 $8%.

8< 1ankanti# .# Bolle# 6.>.# Main &. 3000. Biometrics5 The 'uture of Identification. 4!mputer # 88 )3*#

'eb#

biometricauthentication.doc

Documents