4th International Conference on Electrical Engineering (ICEE 2015)

IGEE, Boumerdes, December 13th -15th, 2015

Remote Multimodal Biometric Identification

Based on the Fusion of the Iris and the
Fingerprint
KamelAiziMohamed OuslimAhmed Sabri
LARESI Lab. Electronics Dept. LMSE Lab. Electronics Dept. LMSE Lab. Electronics Dept.
USTO,Oran, Algeria USTO,Oran, AlgeriaUSTO,Oran, Algeria
aizi.kamel@gmail.comouslim@yahoo.comahmedustopg@hotmail.com
AbstractWith the development of various services through
the Web and especially with the emergence of electronic
commerce, all suppliers of products and services are providing
considerable efforts to secure against all possible fraudulent
intrusions. It appears that biometrics is the only method that can
satisfy the requirements of remote identity in terms of relevance
and reliability.In this paper, we propose a client-server network
architecture for a remote multimodal biometric identification. As
a matter of fact, we use two modalities, namely, the human iris
and his fingerprint in order to strengthen the security, since the
unimodal biometric systems cannot always be used reliably to
perform recognition. However, the association of the information
presented by the various modalities may allow a precise
recognition of the identity. Concerning the fusion of these two
modalities, we used a new approach at thescores level based on a
classification method by the decision tree and a combination
method by the sum. The results obtained confirm that the
proposed method helpedsignificantly to optimize the
performance of the identification.
Index TermsClient-Server Architecture; Fingerprint; Iris;
Fusion; Remote Identification.

I. INTRODUCTION
For security reasons the manhas used several types of
security such as the handwritten signature, password,
magnetic cards. But they have limits and involve the risks
such as loss, theft, oblivion. Biometrics is required
undeniably as the future technology in the field of security
[1]. Concrete cases of application of biometrics on the
Internet exist, they are still few, but the need for this
technology keeps increasing.
There are primarily two computer solutions that are possible
forthe application of biometrics on the Internet; one to identify
a user and the other to authenticate access of a user [2]. The
identification uses a remote database that aggregates users
biometric data stored on a server. This operation can be done
in two distinct ways. One would be to send biometric data
entered on the centralized server to proceed to the
verification.The other allows the verification locally by
retrieving the biometric template from the centralized server
which allows to identify the user through the recorded data.
Concerning the authentication of a user, the procedure is

simple in this case the user could provide his biometric data
and have a smart card [3]. This captured data is compared with
the biometric data of the person stored on the card. This one
to one verification is used to authenticate the user [4].
Biometric systems can automatically identify people using
physical or behavioral characteristics [5], but we cannot
currently guarantee a perfect recognition rate with biometric
systems of a single biometric signature. Moreover, these
systems are often affected by problems like poor one-time
quality of data, not universality of the chosen characteristics,
the similarities between classes, the poor results or the
sensitivity to attacks [6][7]. To overcome these drawbacks, a
solution is to use multiple biometric modalities within a single
system, we speak then of multimodal biometric system [8].
The objective of our work is to propose a client-server
architecture for a remotely multimodal biometric
identification, while offering an approach to combine the iris
and the fingerprint in order to optimize the recognition rate.
In the following sections, we will develop a literature
review related to this work and we will describe the complete
chain of treatment for both modalities the iris and the
fingerprint and the fusion of their scores, and we will describe
the remote identification technique, then we will present the
experimental results, and finally we will finish by a conclusion
and perspectives for the future extensions to this work.
II. PREVIOUS RELATED WORK
Several methods of fusion have been proposed at scores
level in regards to both approaches [9] [10] [11]. In [12], they
used a classifier based on a support vector machine to
combine scores of the face and the voice. In [13], the proposed
system simultaneously acquires the finger-vein and low
resolution fingerprint images and it combines these two
evidences using a novel score level combination strategy. In
[14], a minutiae and texture based fingerprint fusion study was
carried out using a Quality Weighted Sum (QWS) rule for the
score level fusion. In [15], they proposed a biometric health
care security system using the iris and face fusion which was
performed at the matching score level using weighted
scores.In [16], a score-level fusion method for fingerprint
matching has been proposed. The trend and discrimination of
scores were used as prior knowledge with a sigmoid function
to search the optimal fusion parameters.

2015 IEEE

Similarly,several methods havebeen proposed for the

remote biometric identification/authentication [17] [18]
[19].In [20], they discuss the remote biometric identification
and its security aspects, and they propose a solution which
enforces security and promotes a good practice for
designers.In [21], they present biometrics based client-server
architecture for continuous user authentication in e-learning
environments. In [22], they propose a biometric-based multifactor scheme to be used for key agreement and remote mutual
authentication between two parties over an open network. The
scheme properly combines biometrics with a PIN and a token
to achieve high level of security and robustness. In [23], they
propose an improvement of Dass biometric-based remote user
authentication scheme using smart cards in order to withstand
his design flaws.
III. THE MULTIBIOMETRIC SYSTEM
The proposed system is made of two parts. The first part
deals with the fingerprint identification while the second is
based on the iris identification. Each of which follows a
determined process to generate a vector of scores. The fusion
is applied at scores level, and depending on the result of
thisfusion, the identification decision will be made.
A. Iris Identification
The iris is the annular colored area between the pupil and
the white part of the eye.The identification of persons by iris is
performed by the method of Daugman [24] [25]. This method
is made of four main steps namely the segmentation, the
normalization, the feature extraction and finally the
comparison step.
The segmentation is used to extract the region of the iris
from the rest of the image. The iris is located between two
circles which are detected by a Canny filter after smoothing
the image, then the circular Hough transform is applied to
determine the rays and the centers of these two circles.
The normalization is used to transform the iris into a
rectangular 2-D image of a fixed-size which is independent of
the position of the iris in the eye image. This operation of
normalization of the iris image ( , ) is achieved by
transforming polar coordinates ( , ) to Cartesian coordinates
( , ) of the points of the radial r and the angular position in
the normalized model [39], as shown in equation (1).
( , )=(
( , )=(

) ( )+
) ( )+

( )
( )

(1)

Where ( , ) and ( , )are defined as linear combinations of

the set of points of the border of the pupil ( ( ), ( ))and
by the border of the iris ( ( ), ( )), detected in the
segmentation step of the iris.
The feature extraction consists of the coding of the iris
characteristics by applying a convolution of thenormalized iris
model with a Gabor filter. The Gabor filter analyzes the
texture of the iris following different resolutions and different
angles, leading to complex values. The Gabor Wavelet is
)by the equation (2).
defined in polar coordinates (

) /

) /

(2)

Where and
are parameters of the dimensions of the
Gabor analysis window,
the Gabor Wavelet frequency.
Finally and
represent the coordinates of the points of the
Gabor Wavelet analysis.
We proceed by quantification of these coefficients in order to
obtain a binary image representing the corresponding iris code
with a resolution of 480 x 20 pixels.
The comparison or the calculation of the score is based on
the calculation of the Hamming distance (HD) between the
prototype iris code and the iris code under test according to
equation (3).
=

||(

||

||

||

(3)

Where codeA and codeB are two codes calculated from two
iris images and maskA and maskB represent their respective
masks which take into consideration the occlusion
problem[25].
B. FingerprintIdentification
The fingerprintidentification is made of three main steps,
namely the preprocessing, the feature extraction and the
comparison step [26] [27].
The preprocessing is divided into two main steps which are
the normalization of the fingerprint image, and the location
and the framingof the central point of the fingerprint
image.The normalization is used to eliminate the effects of
noise and distortion when capturing the image from the
fingerprint sensor. The original image is normalized by its
mean M and its variance VAR, the matrix G (I) given
byequation (4)indicates the normalized grayscale image and G
(i, j) is its value at pixel (i, j).Where
and
are the
desired mean and variancevalues,respectively.
(, )
=

( (, )

((, )

(, )

(4)

The following step is used to locate and frame the central

point of the fingerprint.The central point detection algorithm is
summarized as follows:
- Estimate the orientation field
- Calculate the field strength of the loop at each point
in the orientation field, using the expanded field of
the hidden orientation.
- Normalize the resistance loop field row in a range
from 0 to 1.
- Perform a thresholding on thefield loop to locate both
the kernel and the center of the region.
In order to extract the relevant features of the fingerprint,
the Gabor filter was applied on the framed part of the
fingerprint following 8 different directions that is (0, 22.5 ,

45 , 67.5 , 90 , 112.5 ,135 , 157.5 ). The results are

complex values which were encoded in order to obtain a
binary vector of size 1024, representing the main features of
the fingerprint image.
Similarly, the comparison of the fingerprints was performed
by the Hamming distance in order to obtain a score that
represents the degree of dissimilarity between the prototype
fingerprint code and the code of the fingerprint under test.

V.

N-Input scores

Classification by decision
tree

No

Identical
Class

MULTIBIOMETRIC F USION APPROACH

It is at a scores level of classification for each unimodal

biometric system that we have carried out the fusion. This
operation was performed using the classification approach and
the combination approach. We have to note that the scores
generated by the two unimodal identification systems, do not
require a prior step of normalization [29] before the
combination by the sum, because these scores are
homogeneous (dissimilarity distance), and their values are
within the same interval [0-1].
The classification method by the decision tree is configured
on the two used databases (the iris and the fingerprint) using a
specific setting of a decision threshold. We cut space of scores
in three zones by making tests on the learning databases; zone
1 is the certainty zone (Identical class), zone 2 is the
uncertainty zone (undefined class), zone 3 is the certainty zone
(Different class). The cutting of the space of the scores is
based on a double thresholding principle as it is represented in
Fig.1.

Decision: unidentified
person (Different)

Yes
No

More than one class

Identical

Decision: person is
identified (Identicali)

Yes

Combination of scores by the sum

(iris + fingerprint) for classes
'Identical'

Search the small combined score (iris

+ fingerprint)

Decision: person is identified

(Identicali)

End
Fig. 2.Flowchart of the fusion method

Iris score

Zone3 : certainty zone

(Different class)

This fusion method consists of two steps, a fusion step and a

decision step[28]. We used a decision tree, developed after
preliminary experiments carried out on the database used, as it
is shown in Fig.3. The main role of this tree is to help to
classify the 2-D N-scores vector, in one of the two classes,
either Identical (if the identity of the person is verified) or
Different (if his identity is not verified).

0.34
Zone2 :uncertainty zone
(Undefined class)
0.31
Zone1 :certainty
zone(Identical
class)
0

Input scores vector

0.30

0.41

Fingerprint score
ISi

Fig. 1.Principle of the double thresholding

0.31
No

Yes

The main advantage of this type of thresholding is to allow

the decision tree to define the undefined class of zone 2
(uncertainty zone) based on two scores, the iris scoreand the
fingerprint score, since the decision in this zonefor unimodal
biometric systems, is not sufficiently reliable.
Fig. 2 shows the general flowchart of this fusion method.
The N-input scores are made of N iris scores and N fingerprint
scores, where N is the total number of persons registered in
the database used.

FSi
Yes
Identical
Class

FSi

0.30
No

Yes

FSi

0.41

Yes

No

Identical
Class

0.30

ISi

0.34

Yes

No

No
Different
Class

Different
Class
Identical
Class

Different
Class

Fig. 3. The obtained decision tree: ISi-iris score, FSi fingerprint score

Then, the decision is made according to the result of the

classification by the decision tree. In the case where there is
no identical class, then the person is not identified (Different).
In the case where there is a single identical class, then the
person is identified (Identical person i)and in the case where
there is more than one identical class, then the combination
method of the scores by the sum is used only to combine both
scores of the iris and the fingerprint corresponding to the
found Identical classes. The purpose of this combination by
the sum, is to seek the smallest combined score (iris score +
fingerprint score) among the N combined scores of the
identical classes. This obtained score represents the
comparison score for the identified person (Identical person i).

it consists of two tables: Table1 is dedicated to codes of the

iris and the fingerprint, while Table2 is dedicated to the
information about the identified person (user). The flowchart
concerning the website and the graphical interface is given in
figure 5.
Website

Graphical Interface

Start

Start

Connecting to
the MySQL
database

Connecting to
the MySQL
database
No

IV. REMOTE IDENTIFICATION TECHNIQUE

In this technique we have used a central database which
contains the biometric data of the users stored on a central
computer unit, a server. To identify the user one must capture
his biometric data and compare it with the set of biometric
data existing on the centralized database. It is therefore a
Client-Server architecture which is proposed and it is
illustrated in figure 4. In this architecture, the communication
is done between a dynamic website and a graphical interface
through a database. To achieve this task, we have used the
WampServer which is a Web development platform in
windows for dynamic Web applications using the Apache
Server, PHP scripts language and a MySQL database and
phpMyAdmin for Web administration of MySQL databases
[3].
Our
website
is
primarily
created
in
PHP
language(Hypertext Preprocessor) which is used to produce a
dynamic web site, as it is associated with the MySQL
database. When the PHP page is executed by the server, then it
returns to the client a web page that contains the HTML
(Hypertext Markup Language) code, this language works on
the assembly and the combination of tags.
The user interface was carried out under the development
environment C ++ Builder which comes with different ways to
interact with the databases such as MySQL.

No

Table1
receives the
data

The codes are

loaded and the
Send button is
pressed

Yes
Reading the data from
table1

Yes
Transmission of the iris
and the
fingerprintcodesto the

Ta
bl
e1

Emptying the table1

Parallel Processing
Comparison of the Comparison of the
fingerprint code
iris code

No
Table2
receives the
data

Centralized Biometric
Database

Yes
Reading the information
concerning the person
from table2
Emptying the table2
Display of the identity
of the person

Vector of scores
Ta
bl
e2

Scores fusion
Transmission of
information concerning
the person to the Table2

Fig. 5.Flowchart of thegraphical interface and the website.

Client Computer

User

Server Computer

Request
Internet
Navigator
(Client)

WampServer
- Apache
- PHP
- MySQL

Response
Graphical Interface

Biometric
Database

Fig. 4.Client-Server Architecture.

The MySQL database that we have created acts as an

intermediary between the website and the graphical interface,

First of all, in this remote identification technique, the user

one must present his code of the iris and the fingerprint in a
client computer, after requesting the web page. So, he can load
his biometric data in the website which is connected
automatically to the MySQL database. Then, he should send
his biometric data to Table1 of the MySQL database by
pressing a send button. At that time, the graphical interface is
also connected to the MySQL database, it can directly retrieve
these biometric data and make the comparison with all
existing biometric data on the centralized database. At the end
of the comparison the graphical interface starts the fusion of
the scores by the proposed method. Then, it sends user
informations (name, first name, sex, age, code and his picture)
to Table2. Finally, the website can retrieve and display
directly this information.

1
EFAR
FRR
FAR

0.9
0.8
0.7
0.6
Error rate

Several tests were undertaken using the two modalities

separately and then applying the proposed fusion method.
During the experiments we used two standard databases:
(1)The iris database CASIA-IrisV4 (figure 6), which
contains 6 subsets of different types. We used 220
samples of individuals (CASIA-Iris-Interval)with 5
samples for each individual, which gives 1100 iris
images.
(2) Thefingerprint database CASIA-FingerprintV5 (figure
7) which contains 20,000 fingerprint images (left and
right thumb / index / middle / ring finger) from 500
individuals (each individual contributed 5 samples
per finger).Similarly,we used 220 samples of
individuals (left index finger), which gives 1100
fingerprint images.
In our case, we created a multimodal database, which
contains 220 virtual individuals, represented by the iris and the
fingerprint signatures, by associating with each individual 5
images of each modality. We divided this database into two
parts, the first one contains 180 individuals who were
registered in the database, and the second part contains 40
individuals who were not registered in the database, i.e. 40 x 5
= 200 attacks by non-enrollees. Then, for each modality of
each registered individual, we enrolled three extracted vector
features (from 3 samples) in the database, and we left 2
samples for the test (180 x 2 = 360 access by enrollees).
The objective is to fuse the two unimodal systems on this
multimodal database and to compare the performance of the
multimodal system with those of unimodal systems.

according to the decision threshold for the iris and the

fingerprint unimodal systems respectively in Figures 8 and 9.

0.5
0.4
0.3
0.2
0.1
0

0.1

0.2

0.3

0.4
0.5
0.6
Decision threshold

0.7

0.8

0.9

Fig. 8.Curve of changes in error rates according to the decision threshold for
unimodal system of iris
1
EFAR
FRR
FAR

0.9
0.8
0.7
0.6
Error rate

VI. RESULTS AND COMMENTS

0.5
0.4
0.3
0.2
0.1
0

0.1

0.2

0.3

0.4
0.5
0.6
Decision threshold

0.7

0.8

0.9

Fig. 9.Curve of changes in error rates according to the decision threshold for
unimodal system of fingerprint

Fig.6 Example iris images in CASIA-IrisV4-Interval

Fig.7. Example fingerprint images in CASIA-FingerprintV5

To evaluate the performance of the proposed system, we

used three metrics. The FAR (False Accept Rate) is the
probability that an impostor is accepted as a genuine
individual, the FRR (False Reject Rate) is the probability that
a genuine individual is rejected as an impostor, and the EFAR
(Enrollee False Accept Rate) isthe probability that an enrollee
is accepted as another enrollee [30]. Furthermore, we
calculated the Recognition Rate TR.
In order to adequately select the decision thresholds, we
carried out preliminary tests that allow us to empirically
represent the three selected metrics in terms of the decision
thresholds. As a result, the three error rate types are plotted

In Fig. 8 and Fig. 9, we can clearly notice that whenthe

decision threshold is low, the EFAR and FAR are both low
and the FRR is approximately high in the case of the
twounimodal systems. However, when the decision threshold
is high greater than 0.5, the EFAR and FAR are both
approximately high and the FRR is relatively low.
Once the threshold values were selected for both unimodal
systems and the decision tree with the principle of the double
thresholding, corresponding respectively to the intervals [0.31
- 0.34] for the iris and [0.30 - 0.41] for the fingerprint, we
computed the three performance metrics, as well as the
recognition rate by exercising the database used, for both
unimodal systems and the multimodal system based on the
proposed method of fusion. Table I summarizes the obtained
results.
TABLE I
REPRESENTATION OF THE PERFORMANCE METRICS
Rate
EFAR
FRR
FAR
TR
Model
Unimodal System
Iris
Unimodal System
Fingerprint
Multimodal
System

4.44 %

10.83 %

7.50 %

84.73 %

7.78 %

16.67 %

13.5 %

75.55 %

1.39 %

4.44 %

1.50 %

94. 17 %

As it is illustrated in Table I, we note that the final fusion

system improved significantly the three error rates, by giving
values less than 1.5% for the EFAR and the FAR and a value
less than 4.5% for the FRR and it enhanced also the
recognition rate to 94.17% compared to the best result given
by the iris unimodal corresponding to 84.73 %. These results
allowed us to confirm that the proposed method of fusion
provides a multibiometric system that outperforms both
unimodal systems.
V. CONCLUSION
In our work, we have proposed and implemented a ClientServer architecture for a remotely multimodal biometric
identification. We used the integration of two biometric
modalities the fingerprint and the iris, based on the statistical
method of classification and the combination method, while
showing the positive contribution of this multibiomeric
approach relative to the unimodality by increasing the
performance of the multibiometric systems.
The application of the biometrics on the internet still raises
several problems. Among these problems there are the security
issues. One must secure the biometric data by finding the ways
to restrict the access to these data. This problem will be the
subject of our next work, which consists to combine the
biometrics with the cryptographic methods. The cryptography
transforms the data using a mathematical process difficult to
reverse[20] [31]. It is therefore possible to combine
cryptography with biometric in order to secure the access
procedure.
REFERENCES
[1]

A. Ross, A. Jain, Information fusion in biometrics,Pattern

Recognition Letters, vol. 24, n. 13, pp. 2115-2125, 2003.
[2] A. Ross, A. K. Jain, Multimodal biometrics : An overview,
Proceedings of 12th European Signal Processing Conference
(EUSIPCO), pp. 1221-1224, Vienna, Austria, 2004.
[3] N. Naim, A. Yassin, W. Zamri, MySQL Database for Storage of
Fingerprint Data, 13thInternational Conference on Modelling and
Simulation UKSim, 2011.
[4] M. Mohammed, M. Elsadig, A Multi-layer of Multi Factors
Authentication Model for Online Banking Services,International
Conference on Computing, Electrical and Electronic Engineering
(ICCEEE), 2013.
[5] A. K. Jain, A. Ross, S. Prabhakar, An introduction to biometric
recognition,IEEE Transactions on Circuits and Systems for Video
Technology, vol. 14, n. 1, pp. 4-20, 2004.
[6] A. K. Jain, A. Ross, Multibiometric systems, Communications of the
ACM, special issue on multimodal interfaces, vol. 47, n. 1, pp. 34-40,
2004.
[7] L. I. Kuncheva, C. J. Whitaker, C. A. Shipp, R. P. W. Duin, Is
independence good for combining classifiers?, Proceedings of
International Conference on Pattern Recognition (ICPR), vol. 2, pp. 168171, Barcelona, Spain, 2000.
[8] M.J. Sudhamani, M.K. Venkatesha, K.R. Radhika, Revisiting Feature
level and Score level Fusion Techniques in Multimodal Biometrics
System,Proceedings of International Conference on Multimedia
Computing and Systems (ICMCS), pp. 881-885, 2012.
[9] Y. Wang, T. Tan, A. Jain, Combining face and iris biometrics for
identity verification, Proceedings of Fourth International Conference
on Audio- and Video-Based Authentication (AVBPA), pp. 805-813,
Guildford, U.K., 2003.
[10] Y. Tong, F.W. Wheeler, X. Liu, Improving Biometric Identification
Through Quality based Face and Fingerprint Biometric Fusion,
Proceedings of IEEE Computer Society Conference on Computer Vision
and Pattern Recognition Workshops, pp. 53-60, 2010.

[11] Z. Yaghoubi, K. Faez, M. Eliasi, A. Eliasi, Multimodal Biometric

Recognition Inspired by Visual Cortex and Support Vector Machine
Classifier, Proceedings of International Conference on Multimedia
Computing and Information Technology, pp. 93-96, 2010.
[12] C. Sanderson, K. Paliwal, Information fusion and person verification
using speech and face information, Tech. Rep. IDIAP-RR 02-33,
IDAIP, 2002.
[13] A. Kumar, Y. Zhou, Human Identification Using Finger Images,IEEE
Transactions on Image Processing, vol. 21, n. 4, pp. 2228-2244. 2012.
[14] K. Nandakumar, A. Ross, A. K. Jain, Incorporating ancillary
information in multi biometric systems,Handbook of Biometrics, (New
York: Springer- Verlag, 2007, 335-355)
[15] S. JadAllah, Al-Hijaili, M. AbdulAziz, Biometric in health care
security system, Iris-Face fusion system,international journal of
academic research, vo1. 3, n. 1, pp.1-11, 2011.
[16] Y. Zang, X. Yang, K. Cao, X. Jia, N. Zhang, J. Tian, A Score-Level
Fusion
Method
with
Prior
Knowledge
for
Fingerprint
Matching,Proceedings of International Conferenceon Pattern
Recognition (ICPR), pp. 2379 2382, Tsukuba, Japan, 2012.
[17] M. Sarvabhatla, M.Giri, C. S. Vorugunti, A Secure Biometrics-Based
Remote User Authentication Scheme for Secure Data Exchange, IEEE
International Conference on Embedded Systems (ICES), pp. 110-115,
2014.
[18] H.B. Tang, X.S. Liu, Y. Li, Cryptanalysis Of A Biometric-Based
Remote User Authentication Scheme,International Conference
onInformation Science and Control Engineering (ICISCE), pp. 1-5,
2012.
[19] S. Barkathunisha, R. Meenakumari, Secure Transmission of Medical
Information U sing IRIS Recognition and Steganography, IEEE
International Conference on Computation of Power, Energy, Information
and Communication (ICCPEIC), pp. 89 92, 2013.
[20] P. Bodnar, R. Joint, E. Center, A Solution To Remote Biometric
Identification,Proceedings of 1st International Conference on
Information Technology (IT), pp. 1-4, Gdansk, Poland, 2008.
[21] A. Moini, A. M. Madni, Leveraging Biometrics for User
Authentication in Online Learning: A Systems Perspective,IEEE
Systems Journal, vol. 3, n. 4, pp. 469 476, 2009.
[22] H. Al-Assam, S. Jassim,Robust Biometric Based Key Agreement and
Remote Mutual Authentication, IEEE 11th International Conference on
Trust, Security and Privacy in Computing and Communications, pp. 5965,Liverpool, 2012.
[23] L. Jiping, D. Yaoming, L. Shouyin, An Noval Biometric-based Remote
User AuthenticationScheme for C/S System, IEEE 2nd International
Symposium on Instrumentation and Measurement, Sensor Network and
Automation (IMSNA), pp. 163 167, Toronto, 2013.
[24] J. Daugman, High confidence recognition of persons by rapid video
analysis of iris texture, European Convention on Security and
Detection, pp. 244-251, 1995.
[25] J. Daugman, How Iris Recognition Works,IEEE Transactions on
Circuits and Systems for Video Technology, vol. 14, n. 1, pp. 21-30,
2004.
[26] L. Chaorong, F. Bo, L. Jianping, Y. Xingchun,Texture-Based
Fingerprint Recognition Combining Directional Filter Banks And
Wavelet,International Journal of Pattern Recognition and Artificial
Intelligence, vol. 26, n. 4, pp. 1-20, 2012.
[27] A. K. Jain, S. Prabhakar, L. Hong, S. Pankanti, Filterbank-based
fingerprint matching,IEEE Transactions on Image Processing, vol. 9,
n. 5, pp. 846-859, 2000.
[28] K. AIZI, M. Ouslim, A New Multibiometric Identification Method
Based on a Decision Tree and a Parallel Processing
Strategy,International Review on Computers and Software,vol. 8, n.
10, pp. 2507-2514,October 2013.
[29] A. Jain, K. Nandakumar, A. Ross, Score normalization in multimodal
biometric systems,Pattern Recognition, vol. 38, n. 12, pp. 2270-2285,
2005.
[30] T. Murakami, K. Takahashi, Fast and Accurate Biometric
Identification Using Score Level Indexing and Fusion,International
Joint Conference on Digital Object Identifier, pp. 1-8, 2011.
[31] W. Scheirer, B. Bishop, T. Boult, Beyond PKI: The Biocryptographic
Key Infrastructure,Proceedings of the IEEE International Workshop on
Information Forensics and Security (WIFS), pp. 16, 2010.