IRJET - An Enhanced Signature Verification System using KNN

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 03 | Mar 2020 www.irjet.net p-ISSN: 2395-0072
© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 2168
AN ENHANCED SIGNATURE VERIFICATION SYSTEM USING KNN
S.Kuppusamy1, M.Bala Krishnan2, S.Mukthair Basha3, M.Indhumathy4
1,2,3Dept. of Information Technology, Rajiv Gandhi College of Engineering and Technology, Puducherry, India.
4Assistant Professor, Dept. of Information Technology, Rajiv Gandhi College of Engineering and Technology,
Puducherry, India
---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - Handwritten signatures have proved to be
important in authenticating a person's identity, who issigning
the document. Here proposeasystemforsignatureverification
using KNN. Nowadays signature is a basic and important
verification system for every individual. Everyonehasaunique
signature and every individual can differ from others. An
automated verification process would enable banks andother
financial institutions to significantly reduce check and money
order forgeries, which account for a large monetary loss each
year. Simulation are carried out using Matlab.
Key Words: Handwritten, signature verification, KNN …
1.INTRODUCTION
Handwriting is a skill that is highly personal to individuals
and consists of graphical marks on the surface in relation to
a particular language. Many researchers have been done on
this topic. Signatures of the same person can vary with time
and state of mind. A method proposed a signature
verification system which extracts certain dynamic features
derived from velocity and acceleration of the pen together
with other global parameters like total time taken, number
of pen-ups. The features are modeled by fitting probability
density functions i.e., by estimating the mean and variance,
which could probably take care of the variations of the
features of the signatures of the same person with respect to
time and state of mind. Handwritten signature is a form of
identification for a person A method is introduced where a
signature image is first segmented (vertical and horizontal)
and then data is extracted from individual blocks.Herethese
data is then compared with the test signature.Signaturesare
composed of special characters and flourishesandtherefore
most of the time they can be unreadable. Also intrapersonal
variations and the differences make it necessary to analyze
them as complete images and not as letters and words put
together.
The handwritten signature is a particularly
important type of biometric trait, mainly due to its
ubiquitous use to verify a person’s identity in legal, financial
and administrative areas. One of the reasons for its
widespread use is that the process to collect handwritten
signatures is non-invasive, and people are familiar with the
use of signatures in their daily life. Signature verification
systems aim to automatically discriminate if the biometric
sample is indeed of a claimed individual. In other words,
they are used to classify query signatures as genuine or
forgeries. Forgeries are commonly classified in three types:
random, simple and skilled (or simulated) forgeries. In the
case of random forgeries, the forger has no information
about the user or his signature and uses his own signature
instead. In this case, theforgerycontainsa differentsemantic
meaning than the genuine signatures from the user,
presenting a very different overall shape. In the case of
simple forgeries, the forger has knowledge of the user’s
name, but not about the user’s signature. In this case, the
forgery may present more similarities to the genuine
signature, in particular for users that sign with their full
name, or part of it. In skilled forgeries, the forger has access
for both the user’s name and signature, and often practices
imitating the user’s signature. This result in forgeries that
have higher resemblance to the genuine signature, and
therefore are harder to detect. Depending on the acquisition
method, signature verification systems are divided in two
categories: online (dynamic) and offline (static). In the
online case, an acquisition device, such as a digitizing table,
is used to acquire the user’s signature. The data is collected
as a sequence over time, containing the position of the pen,
and in some cases including additional information such as
the pen inclination, pressure, etc. In offline signature
verification, the signature is acquired after the writing
process is completed. In this case, the signature is
represented as a digital image.
FIG.1 SIGNATURE
For any legal transactions the authorization is done
by the signature. So the need of the signature verification
increases. The handwritten signatures are unique for
individuals and which is impossible to duplicate. The
technology is easy to explain and trust. The primary
advantage that signature verification systems have over
other type’s technologies is that signatures are already
accepted as the common method of identity verification

2. RELATED WORK
This work proposes a novel system for off-line handwritten
signature verification. A new descriptor founded on a quad-
tree structure of the Histogram Of Templates (HOT) is
introduced. For the verification step, we propose a robust
implementation of the Artificial ImmuneRecognitionSystem
(AIRS). This classifier is inspired from the natural immune
system, which generates antibodies to protect the human
body against antigens. The AIRS training develops new
memory cells that are subsequently used to recognize data
through a k NearestNeighbor(kNN)classification.Presently,
to get a robust verification, the kNN classification is
substituted by a Support Vector (SV) decision, yielding the
AIRSV classifier [1]. Recognition of signature is a method of
identification, whereas verification takes the decision about
its genuineness. Though recognition and verification both
play important role in forensic sciences, however,
recognition is of special importance to the banking sectors.
In this paper, we present a methodology to analyse 3D
signatures captured using Leap motion sensor with the help
of a new feature-set extracted using convex hull vertices
enclosing the signature. We have used k-NN and HMM
classifiers to classify signatures. Experiments carried out
using our dataset as well aspubliclyavailabledatasetsreveal
that the proposed feature-set can reduce the computational
burden significantly as compared to existing features. It has
been observed that a 10-fold computational gain can be
achieved with non-noticeable loss in performance using the
proposed feature-set as compared with the existing high-
level features due to significant reduction in the feature
vector size [2]. In this work a new online signature
verification system based on Mellin transform in
combination with an MFCC is presented. In the first step we
extract signals x(t) and y(t) fromeachsignatureandthenthe
novel pre-processing algorithm by Mellin transform is
performed. The key property of Mellin transformisthescale
invariance which makes the features insensitive to different
signature scale. The feature is extracted by Mel Frequency
Cepstral Coefficient (MFCC). Subsequently, feature
extraction is used to extract coefficient for each signature to
construct a feature vector. These vectors are then fed into
two classifiers: Neural network with multi-layer perception
architecture and linear classifier used in conjunction with
PCA and then results are compared. In order to evaluate the
effectiveness of the system several experiments are carried
out. Online signature database from signature verification
competition (SVC) 2004 is used during all of the tests [3]. A
system which does computing and is combines with basic,
and highly coincidental processing elements which use the
data to get a highly relevant and faster response from the
inputs taken. Artificial neural network models are a subpart
of the machine learning models which are motivated by the
functioning of the brain. Neural networks generally work
like the neurons of the brain and the connected neurons will
work in a network process to collect and processthedata for
providing the necessary output. There will be an input layer
to the system which consists of all the patterns in which the
system should process and also the necessary inputs and it
communicates with the hidden layer as shown in the below
figure and the hidden layers use the patterns and inputs by
the input layer and are used to find out a relevant function
for the task to be performed and then they communicate
with the output layers to display the final output.
2.1 FEEDFORWARD MECHANISM:
This mechanism does not form circles like many artificial
neural networks. This mechanism goes in a single way from
the input to the hidden layers to the output and do not form
any loops or circles in the process.
FIG.2 FLOWCHART
The important factor in preprocessing stage is to
build standard signature which is prepared for extraction of
features. Image processing application, pre-processing is
required to remove discrepancies, from the input image.
2.2 NORMALIZATION OF SIGNATURE
It is possible that signature can be fractured due to
imperfections in image scanning and capturing. It is also
possible that the dimensions of signature can vary from
person to person and even the same person can sometimes
have different sizes based on the mood and environmental
factors. So a process is required to overcome the size
variation problem and achieve a standard signature size for
all signatures.

2.3 THINNING
It is possible that the signature is written on different pen
and the thickness thus varies from one pen to another. The
purpose of thinning is to eliminate thickness differences in
signature by making all of them one pixel thick. Thinning is
used to enhance the object’s global properties and to
transform the input image into a compact form.
2.4 FEATURE EXTRACTION
This method is used for extracting the necessary and
essential features from the input image. A feature vector is
created from the features extracted. Each signature has a
unique feature vector. These features are extracted as
follows
The feature extraction module uses moment
invariants to extract texture features of the image using
central moment and derived invariant moment.
2.5 NEURAL NETWORK TRAINING
The features extracted are fed to natural network as inputs.
Before that the networks are trained with data sets. Each
neural network has a corresponding user to it. So a user has
two neural networks one with feedforward mechanism and
the other with feedback mechanism. The user’s features are
given as input to both the neural networks and the output is
recorded.
In this method it construct a neural network by optimizing
some existing neural networks and it will havea usethedata
structure tree along with nodes similar to human eye which
has neurons and it used for recognition of patterns.
3. PROPOSED SYSTEM
The signature verification system using k-nearest neighbor
is proposed. In pattern recognition, the k-nearest neighbor
algorithm is a method for classifyingobjectsbasedonclosest
training examples in the feature space. The intuition
underlying nearest neighbour classification is quite
straightforward, signatures are classified based on the class
of their nearest neighbours.
3.1 KNN
K Nearest Neighbor(KNN) is a very simple, easy to
understand, versatile and one of the topmost machine
learning algorithms. In KNN, K is the number of nearest
neighbors. The number of neighbors is the core deciding
factor. K is generally an odd number if the number of classes
is 2. When K=1, then the algorithm is known as the nearest
neighbor algorithm
FIG.3 BLOCK DIAGRAM
3.2 IMAGE ACQUISITION
Here For every person we have collected n signature
samples for database. It is better if we can collect more
signature samples for database. Then for verification collect
test signatures against the sample signatures. These test
signatures we have to verify if it is genuine or forgery. Each
of the signatures (Samples and corresponding test) has to
take within a same sized area on paper by pen and collect
the image of that particular area.
3.3 PREPROCESSING
The preprocessing of the signature images is related to the
removal of noises, and thinning. The goal of thinning is to
eliminate the thickness differences of pen by making the
image one pixel thick. To remove noises and enhance, the
images are preprocessed by filtering techniques. The
principle objective of the image enhancement is to process
an image for a specific task so that the processed image is
better viewed than the original image. Preprocessing is
process which helps us to reduce the background noise.
Intensity of the image should be normalized. By Enhancing
input image or image captured by digital camera, is to
remove the backgroundnoise,imagecangetenhancedvisual
appearance of input images. By this enhancement process
artifact image can be highlighted. Image preprocessing is
used to create an enhanced and please full version of the
captured image. The image preprocessing steps used in the
system are:
1) Conversion of RGB image to gray image
2) Resizing of the image
3) Filtering of the image.
In RGB color model, each color appears in its primary
spectral components of red, green, and blue. The color of a
pixel is made up of three components; red, green, and blue
(RGB). The disadvantages of RGB models are, it requires

large space to store and it will take more time to process. So
there is a need for converting the RGB model to Gray model.
Resizing is an important step in image preprocessing. The
acquired image is resized according to the requirement of
the system. Resizing is nothing but, changingthedimensions
of an image. The captured image is resized using some
resizing methods according to the requirement of the
system.
3.4 FEATURE EXTRACTION
Extracted features in this stage are used for clustering the
signature images for verification stage. Features will haveto
be extract from both sample images and Test image.
3.5 SIGNATURE HEIGHT WIDTH RATIO:
The ratio is obtained by dividing signature height to
signature width. The height is the maximum length of the
columns obtained from the cropped image. Similarly the
width is also calculated considering the row of maximum
length. Signature heightand widthcanchange.Butheight-to-
width ratios of an individual’s signatures are approximately
constant.
3.6 SIGNATURE OCCUPANCY RATIO:
It is the ratio of number of pixels which belong to the
signature to the total pixels in the signature image. This
feature provides information about the signature density.
3.7DISTANCERATIOCALCULATIONATBOUNDARY:
After cropping, the pixels in closest proximity to the
boundaries (left, right, upper & bottom) are determinedand
their distance from the left & bottom boundaries are
evaluated, i.e. for the upper leftmost pixel its distance from
bottom boundary(L1) & for the bottom left most pixel the
distance from right boundary is calculated(L2).Thesevalues
are used later in verification process.
3.8 CLASSIFICATION
The classifier we have used is KNN which stands for k-
nearest neighbours. It is basically a classification algorithm
that means it assigns a class to a test image based on its
feature values. A person may have many sample signature
images. We create separate clusters for set of sample
signatures for each person. Here we use K-Nearest
Neighbors’ (KNN) clustering Technique for verifying a test
signature belongs which cluster. The k-nearest neighbours’
algorithm uses Euclidian distance method to find the
distance between two training points. Thus using Euclidian
distance we find k nearest neighbouring training points of
our test point based on its features and the class with
maximum number of occurrences is taken as the decision
class for that test image and is assigned to that image.
It is often useful to take more than one neighbour into
account so the technique is more commonly referredto ask-
nearest neighbour (k-NN) classification where k nearest
neighbours are used in determining the class. Since the
training signatures are needed at run-time, i.e. they need to
be in memory at run-time; it is sometimes also called
memory-basedclassification.Becauseinductionisdelayedto
run time, it is considered a lazy learning technique. Because
classification is based directly on the training signaturesitis
also called example-based classification or case-based
classification. So, k−NN classificationhastwostages;thefirst
is the determination of the nearest neighbours and the
second is the determination of the class using those
neighbours. This approach to classification is of particular
importance today because issues of poor run-time
performance is not such a problem these days with the
computational powerthatisavailable.Duringtheenrollment
phase, a set of reference signatures are used to determine
user dependent parameters characterizing the variance
within the reference signatures. The reference set of
signatures, togetherwiththeseparameters,arestoredwitha
unique user identifier in the system’s database. In the
training phase we choose a number of genuine and forged
signatures for training the K-NN classifier. Intheverification
phase when a test signature is input to the system, it is
compared to each of the reference signatures of the claimed
person. The person is authenticated if the resulting
dissimilarity measure is below or equalsa thresholdvalueof
the classifier, otherwise denied.
4. IMPLEMENTATION
In the project work, the experiments are carried using
Matlab coding. We have prepared a GUI layout with a list of
menus. Clicking on each menu will perform an independent
function.
FIG.4 OUTPUT GUI

FIG.5 MOST LIKELY GENUINE
FIG.6 DEFINITELY FORGED
5. CONCLUSION
In this project, we used Image processing which is one ofthe
most trending and most used domain nowadays for
functions like image detection, fingerprint verification etc.
This project helps in controlling human errors in signature
verification and also makes the signature verification
accurate, easy and faster. It also makes the work easier for
understanding and executing it by anyone without any
knowledge of image processing. If any bank or any company
uses this system the customers will feel much more secure
and trustworthy. Thus, here propose that this system brings
a change in the working of several banks, companies etc. It
is better if future works extracts more features that may
provide a combination to achieve higher accuracy. Future
works should include the use of different features and
classifiers such as deep learning neural network. By
increasing the numbers of hidden layers, theperformanceof
the neural network can be expected to be better but time for
training and testing may increase.
REFERENCES
1. S. Yin, A. Jin, Y. Han, and B. Yan, “Image-based handwritten
signature verification using hybrid methods of discrete
Radon transform , principal component analysis and
probabilistic neural network,” Appl. Soft Comput. J., vol. 40,
pp. 274–282, 2016.
2. K. Wrobel, R. Doroz, P. Porwik, J. Naruniec, and M.
Kowalski, “Engineering Applications ofArtificial Intelligence
Using a Probabilistic Neural Network forlip-basedbiometric
verification,” Eng. Appl. Artif. Intell., vol. 64, no. January, pp.
112–127, 2017.
3. D. Suryani, E. Irwansyah, R. Chindra, D. Suryani, E.
Irwansyah, and R. Chindra, “ScienceDirect O fflfflineine
Signature Signature Recognition Recognition and and
Verification Verification System System using usingfficient
Fuzzy Kohonen Clustering Network ( EFKCN ) Algorithm E
fficient Fuzzy Kohonen Clustering Network ( EFKCN )
Algorithm,” Procedia Comput. Sci., vol. 116, pp. 621–628,
2017.
4. Y. Serdouk, H. Nemmour, and Y. Chibani, “New off-line
Handwritten Signature Verification method based on
Artificial Immune Recognition System,” Expert Syst. Appl.,
vol. 51, pp. 186–194, 2016.
5. Y. Serdouk, H. Nemmour, and Y. Chibani, “Handwritten
signature verification using the quad-tree histogram of
templates and a Support Vector-based artificial immune
classification ଝ,” Image Vis. Comput., vol. 66, pp. 26–35,
2017.
6. P. Porwik, R. Doroz, and T. Orczyk, “Signatures veri fi
cation based on PNN classi fi eroptimised byPSOalgorithm,”
vol. 60, pp. 998–1014, 2016.
7. S. Kumar, D. Prosad, and P. Pratim, “Fast recognition and
verification of 3D air signatures using convex hulls,” Expert
Syst. Appl., vol. 100, pp. 106–119, 2018.
8. N. Khera and S. A. Khan, “Microelectronics Reliability
Prognostics of aluminum electrolytic capacitors using arti fi
cial neural network approach,” Microelectron. Reliab., vol.
81, no. October 2017, pp. 328–336, 2018.
9. A. Fallah, M. Jamaati, and A. Soleamani, “A new online
signature verification system based on combining
Mellintransform , MFCC and neural network,” Digit. Signal
Process., vol. 21, no. 2, pp. 404–416, 2011.
10. D. Dabrowski, “Condition monitoring of planetary
gearbox by hardware implementation of artificial neural
networks,” Measurement, vol. 91, pp. 295–308, 2016.
[11] Vance Faber , “Clustering And the Continuous K-means
Algorithm”, Los Alamos Science, Number 22 , 1994
[12] P.S.Bradley, Usama M.Fayyad,“RefiningInitial Pointsfor
K-Means Clustering”, International Conference on Machine
Learning (ICML98), J. Shavlik (ed.), pp. 91-99, 1998

[13] Dorin Comaniciu, Peter Meer, “ Mean Shift: A Robust
Approach Toward Feature Space Analysis ”, IEEE
transactions on Pattern Analysis and Machine Intelligence,
Vol. 24, No. 5, May, 2002
[14] Pierre-Alain Moëllic, Jean -Emmanuel Haugeard,
Guillaume Pittel, “Image Clustering Based on a Shared
Nearest NeighborsApproach for Tagged
Collections”,Proceedings of the 2008 international
conference on Content-based image and video retrieval,
pp:269-278 ISBN:978-1-60558-070-8 , 2008
[15] Ujjwal Maulik, Sanghamitra Bandyopadhyay,
“Performance Evaluation of Some ClusteringAlgorithmsand
Validity Indices”, IEEE TRANSACTIONS ON PATTERN
ANALYSIS AND MACHINE INTELLIGENCE, VOL. 24, NO. 12,
DECEMBER 2002
[16] H. Liu1, J. Li and M. A. Chapman, “Automated Road
Extraction from Satellite Imagery Using Hybrid Genetic
Algorithms and Cluster Analysis”, Journal of Environmental
Informatics 1 (2) 40-47, 2003
[17] M.Liwicki and H.Bunke, “Combining Online and off-line
Systems for Handwriting Recognition”, 9th International
conference on Document Analysis and Recognition (ICDAR
2007) , Curitiba, Brazil, Vol-I, pp:372-376
[18] Vu Nguyen, Michael Blumenstein, Vallipuram
Muthukkumarasamy,Graham Leedham, “Offline Signature
Verification Using Enhanced Modified Direction Features in
Conjunction with Neural Classifiers and Support Vector
Machines”, 9th International conference on Document
Analysis and Recognition (ICDAR 2007) , Curitiba, Brazil,
Vol-II, pp:734-738.

IRJET - An Enhanced Signature Verification System using KNN

More Related Content

What's hot (20)

Similar to IRJET - An Enhanced Signature Verification System using KNN (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET - An Enhanced Signature Verification System using KNN