International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882

Volume 5, Issue 1, January 2016

A Hybrid Approach for Biometric Authentication using Fusion of Fingerprint

and Signature
Sujata Kataria1, Ashok K. Goel2
1

Department of Electronics & Communication Engineering, Gaini Zail Singh PTU Campus, Bathinda-151001, Punjab, India
Email: sujatakataria26@gmail.com
2
Professor, Department of Electronics & Communication Engineering, Gaini Zail Singh PTU Campus, Bathinda-151001, Punjab,
India

ABSTRACT
This paper describes a multibiometric system based on
fingerprint and signature recognization. Fingerprint
recognization is the most accepted physiological
characteristic used to recognize a person in biometric
systems. Signature recognition, on the other hand, is the
most popular behavioral characteristic used in biometric
systems. This paper proposes a technique for individual
identification using fusion of fingerprint and signature
data. Signature feature extraction has been done using
SIFT (Scale Invarient Feature Transform) algorithm and
fingerprint feature extraction has been done using
minutia extraction. Fusion has been done on the basis of
Neural Network by using sum rule. System performance
has been evaluated based on False acceptance rate,
False rejection rate and accuracy in MATLAB 7.10
simulation. The results obtained using the proposed
multibiometric approach are better than those obtained
using single biometric system.
Keywords - Biometric, Fingerprint, Minutiae, Neural
Network, Security, SIFT, Signature.

I.

INTRODUCTION

Multimodal biometrics is the combination of two or

more modalities such as signature and speech modalities.
Fingerprint recognition is widely used physiological
characteristic to identify a person in biometric systems,
due to its factors like feasibility, permanence,
distinctiveness, reliability, accuracy, and acceptability.
Signature recognition is the most popular behavioral
characteristic used in biometric systems [1]. Therefore
combination of these two methods will be reliable and
accurate. A weighted fusion scheme is required, which
transforms the scores into a common range, combines
them and giving the final fused score [2].
These modalities are widely accepted and natural to
produce, This combination of multimodal enhances
security and accuracy but, complexity of the system
increases due to increased number of features extracted
out of the multiple samples and it also increases the
system cost [3,4]. These days the major issue is that at
what degree features are to be extracted and how the cost
factor can be minimized. Because the number of features
increases the variability of the intra-personal samples.

Increase in variability of the system will further increase

FAR [5]. To resolve these issues an effective fusion
level and fusion mode is required. The proposed work
tries to present a novel user authentication system based
on a combined acquisition of signature and fingerprint
signals with high accuracy rate.

II.

FINGERPRINT BIOMETRICS

The study of the Fingerprints came into the moon light

in the 18th century. Father of fingerprint matching
technique is Francis Galton. Fingerprint matching is one
of the oldest forms of the biometric technologies that are
being used so widely [6]. Fingerprint technology is used
in fields of criminal investigation, civilian, government
and commercial applications such as passport, license
card, security device etc. [8]. Fingerprints of the human
being are unique and they can never be the identical.
Fingerprint identification is the practice of identification
of personality based on the impression pattern of ridges
and valley on fingers [3]. There is a widespread use of
fingerprint technology yet only modest work has been
done to know the distinctiveness of the fingers. There
are many factors that led to the interruption of
recognition of fingerprints like pressure, small pressing
spot, atmosphere factors, device noise, skin suppleness
etc [9]. A fingerprint image is shown below in Fig. 1.

Figure 1: Fingerprint image

Fingerprints are unalterable and they can diminish only
after death. Fingerprints were used as a signature from
the ancient time [10]. The main features of a biometric
gauge are universality, distinctiveness, durableness,
collectability, satisfactoriness, and recital. Among all the
biometric indicators, fingerprints have one of the highest
levels of reliability and have been extensively used by
forensic experts in criminal investigations.

www.ijsret.org

35

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 5, Issue 1, January 2016

III.

SIGNATURE BIOMETRICS

Handwritten signatures are generally acknowledged as a

method for archive confirmation, approval and
individual check [11]. For legitimacy most archives like
bank checks, travel visas and scholastic declarations
need to have approved handwritten signatures. In
modern world where extortion is common, there is the
requirement for a programmed HSV (Handwritten
signature verification) framework to supplement visual
confirmation [12]. Automated signature verification is as
vital as other programmed distinguishing proof
frameworks. However they vary from different
frameworks that depend on ownership of keys and
learning of particular individual data like passwords [13,
14]. They depend on all around motions and still they
are most socially and lawfully acknowledged type of
individual recognizable proof [4, 5]. A handwritten
signature image is shown in Fig. 2.

V. HYBRID APPROACH FOR BIOMETRIC

AUTHENTICATION AND IMPLEMENTATION
A multimodal biometric system allow the addition of
two or more types of biometric recognition and
verification system in organized manner to assemble
stringent presentation necessities. The multimodal
system might be a mixture of fingerprint confirmation,
face recognition, voice verification and keystroke
dynamics. In this paper, signature and fingerprint
verification system are combined. Fingerprint
recognition is mainly accepted physiological attribute
used to recognize a human being in biometric systems.
Signature recognition is mainly accepted behavioral
characteristic used in biometric systems. Combining the
fingerprint and signature enhances the strength of the
person authentication. The proposed work methodology
is as shown in Fig.3 which explains the work in steps.

Figure 2: Signature Image

IV.

FUSION
IN
MULTIMODAL
BIOMETRIC SYSTEMS

A system that can unite consequences from two

dissimilar biometric traits is called biometric fusion. The
signature and fingerprint verification systems are joint in
the system. This work gives a new user authentication
system which depends upon joint acquisition of
signature and fingerprint signals with elevated accuracy
rate. Diverse level of fusion used in multi biometric
systems are given as below:
1. Sensor level
2. Feature level
3. Match score level
4. Decision level
Fusion at the sensor, match score and decision levels
have been broadly deliberated in the literature. Fusion at
the feature level, is comparatively understudied, it
involves the addition of feature sets matching to
manifold information resources. The feature set have
richer data for the raw biometric data than the match
score or the last decision. Mixing at this level is
predictable to give improved authentication outcome.
The proposed work presents a user authentication system
based on joint acquisition of signatures and fingerprints.
Feature level fusion is used that gives optimal
identification results. But, here many difficulties arise if
the feature sets are created using many biometric
characters.

Figure3: Flowchart of proposed work

1) Fingerprint Recognition
Step 1 : Upload the fingerprint image from the
database. The size of image must be less than
50 kb so that preprocessing can be done in less
time.
Step 2 : After, uploading input image grey scale
conversion is applied on it. And once image is
obtained in grey scale then edge detection is
applied on it. In this canny edge detection is
applied on the image.
Step 3 : Then, binarization is applied on to the image. It
is done to highlight the foreground pattern from
the background. It is the process of conversion
of grayscale image into binary image.

www.ijsret.org

36

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 5, Issue 1, January 2016

Step 4 : After this minutia extraction is applied on to the

binarized image. In this method, minutiae
points of ones fingers are matched with the
others minutiae points. Basically there are 80
minutiae points. Two thumbprint impressions
are matched by compairing minutia matching
points.
2) Signature Recognition
Step 5 : Upload Signature image from the database.
Step 6 : Original image is converted into grey scale
image.
Step 7 : Then edge detection is applied on to the grey
scale image.
Step 8 : After, this SIFT algorithm is applied on to the
previous output image. SIFT is an algorithm in
computer vision to detect and describe local
features in image. In this SIFT key points of
objects are first extracted from a set of
reference images and stored in a database. An
object is recognized in a new image by
individually comparing each feature from the
new image to this database and finding
candidate matching features based on
Euclidean distance of their feature vectors. The
feature vector thus obtained is the signature
template in the knowledgebase.
3) Fusion
Step 9 : The two feature vectors, a feature vector for the
fingerprint generated by the corresponding
subsystem
and the second represents the
feature vector of the signature. These two
feature vectors are fused to present a novel user
authentication system. The sum of feature
vector of fingerprint and signature is performed
by as given below:= fingerprint feature vector
= signature feature vector
= fusion value
=
+
Step 10 : Matching has been done by using hamming
distance. It gives the similar bits containing
between two bit patterns. A decision can be
made using hamming distance that whether
template matches to trained template or not.
Step 11 : Then, apply neural network for classification.
Instead of using the threshold as the judgment
rule, another method to measure the similarity
between the feature vectors of the input
fingerprint and signature with those of the
templates fingerprint and signature stored in the
database is realized by a three-layer BPNN
(Back Propagation Neural Network). Train Set
function has been used for testing. Testing is
done on the original sample and trained

samples are compared with the tested samples

and if they are matched then authentication is
possible otherwise not possible.
5.1. Implementation
In this proposed work, a database consists of 10
fingerprint images and 10 siganture images from ten
different peoples is used. These 10 sample images are
stored in database which are later used for training and
testing the neural network. A fingerprint and signature
image is uploaded from the database. Convert the
original image to grey scale and edge image. Fingerprint
feature extraction is done by using minutia and signature
feature extraction is done by using SIFT algorithm.
Minutia extraction gives total no. of 100 features and
SIFT algorithm gives total no. of 1000 features. The
proposed system select 100 features from both the
feature vectors for fusion. Fused value of fingerprint and
signature features is stored in the database. In this
proposed work, a feed forward back propagation neural
network with 5 hidden layer neurons is used for
classification. This fused value of 100 is used both for
training as well as testing the neural network. For testing
purpose, both modalities extract features & then fuse
them by using this proposed technique and after fusing
the value can be compared with the stored values in the
database and give the results. The proposed system
utilized fused value as a neural network input and as an
output classified data is acquired. If the result and
sampled image match more than 90% only then it is
assumed that individual is identified. Evaluate the
system performance on the basis of parameters such as
FAR, FRR and accuracy.

VI.

RESULTS AND DISCUSSION

This section will present results and analysis of various

parameters used in this work such as FAR, FRR and
accuracy. In this work, a hybrid approach for biometric
authentication using fingerprint and signature is used.
Fingerprint and signature biometrics are combined
together in order to enhance the security and to attain
better results. In this proposed work two different
databases that is Signature database and Fingerprint
database are used. In fingerprint database total 10 images
are taken. In fingerprint database 364*328 pixel size
image has been used. Image is of .tif format and size of
the image is approximately around 117 kb/ image. In
signature database total 10 images are taken. In signature
database 241*80 pixel size image has been used. Image
is of .jpg format and size of the image is approximately
around 6 kb/ image.
In context, the performance of a biometric measure is
usually defined in terms of False accept rate (FAR):
what percentage of times an invalid user is accepted by
the system. Its formula is given as:

www.ijsret.org

37

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 5, Issue 1, January 2016

False rejection rate (FRR): the percentage of times

a valid user is rejected by the system. Its formula is
given as:

Accuracy: Accuracy is a general term used to

describe how accurate a biometric system performs. Its
formula is given as:
In real-world biometric systems the FAR and FRR can
typically be traded off against each other by changing
some parameters. One of the most common measures of
real-world biometric systems is the rate at which both
accept and reject errors are equal: the equal error rate
(EER). The lower the EER, the more accurate the system
is considered to be.
Table 1: Parameters value of Neural Network and
SVM
Parameters
FAR
FRR
Accuracy

Neural Network
0.0953
0.0251
99.9652

SVM
2.1833
2.5234
95.2932

Above table shows various parameters values

classification using Neural Network Technique and
SVM technique and it shows that the proposed system
largely improved the system performance rates. The
proposed system based on classification by Neural
Network gives the FAR, FRR and accuracy respectively
to 0.0953, 0.0251 and 99.9652 as compared to
classification by SVM (support vector Machine) that are
2.1833, 2.5234 and 95.2932.

VII.

CONCLUSION

In this paper, a multi-biometric system has been

presented for fingerprint and signature biometric. In this
work it has been shown that if fingerprint and signature
image of a person is given then the network can
recognize the image. SIFT method has been used for
feature extraction of Signature samples. Minutia
extraction method has been used for feature extraction of
fingerprint samples. Here, feature level fusion is used by
applying sum rule for fusion process. The obtained
results are very interesting. This multimodal fusion
scheme decreased the false acceptance rate, false
rejection rate to 0.0953, 0.0251 and increases the system
accuracy to 99.9652. We have also compared the
results of previous SVM classification technique and
proposed Neural Network classification technique which
shows that this proposed work is better than existing

work. This multibiometric scheme increases the system

performance rate as compared to single biometric
authentication system and also increases the security
level of the system.
Future work could go in the direction to combine other
modalities, other methods of classification, and other
methods of fusion to accomplish a large comparative
study in the multi-biometric domain and to develop a
tool to automate some of the processes which have been
done manually here.

REFERENCES
[1] Davit Kocharyan, et al, A Multimodal Biometric
System Based on Fingerprint and Signature Recognition,
Institute for Informatics and Automation Problems of
NASRA, (2013).
[2] Arun Ross, Anil Jain, Information fusion in biometrics,
Pattern Recognition Letters 24 (2003) 21152125.
[3] Robert Hastings, Ridge Enhancement in Thumbprint
Images Using Oriented Diffusion, IEEE Computer Society
on Digital Image Computing Techniques and Applications, pp.
245-252, (2007).
[4] UnsangParh, et al, Fingerprint Verification using SIFT
Features, SPIE Defense and Security Symposium, (2008).
[5] A.Pokhriyal and S. Lehri," A New Method of Fingerprint
Authentication Using 2D Wavelet", Journal of Theoretical and
Applied Information Technology, Vol. 13 No.2, pp. 131-138,
March, 2010.
[6] Liu Wei, Fingerprint Classification using Singularities
Detection, International Journal of Mathematics and
Computers in Simulation, issue 2, vol. 2, pp. 158-162, (2008).
[7] Hartwing Fronthaler et al, Local Features for
Enhancement and Minutiae Extraction in Fingerprints, IEEE
Transactions on Image Processing, vol. 17, no, 3, pp. 354363, (2008).
[8] M. R. Girgisa, et al, Employing Generic Algorithms for
Precise Fingerprint Matching Based on Line Extraction,
Graphics, Vision and Image Procession Journal, vol. 7, pp. 5159, (2007).
[9] Duresuoquian Miao, et al, Fingerprint Minutiae
Extraction Based on Principal Cures, the Journal of the
Pattern Recognition Letters, vol. 28, pp. 2184-2189, (2007).
[10] Luping Ji and Zhang Yi, Fingerprint Orientation field
Estimation using Ridge Protection, The Journal of the Pattern
Recognition, vol. 41, pp. 1491-1503, (2008).
[11] B. Fang,et al , A Smoothness Index Based Approach for
Off-line Signature Verification (2000).
[12] B. Fang, et al, Off-line Signature Verification by the
Tracking of Feature and Stroke Positions, Pattern
Recognition, vol. 36, pp. 91101, (2003).
[13] Katsuhiko Ueda, Investigation of Off-Line Japanese
Signature Verification Using a Pattern Matching, Proc. Of
the 7th ICDAR, (2003).
[14] Neeraj Shukla, Invariant Features Comparison in Hidden
Markov Model and SIFT for Offline Handwritten Signature
Database, International Journal of Computer Applications
(0975 8887) vol. 2, No.7, (2010).

www.ijsret.org

38