A PROJECT REPORT ON

SIGNATURE VERIFICATION BASED ON ONE REAL

SIGNATURE

SUBMITTED TO THE SAVITRIBAI PHULE PUNE UNIVERSITY, PUNE

IN THE PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE AWARD OF THE DEGREE
OF

BACHELOR OF ENGINEERING (COMPUTER

ENGINEERING)

SUBMITTED BY

MITESH PARMAR Exam No:

NUPUR PURANIK Exam No:
DHRUVA JOSHI Exam No:
SONAL MALPANI Exam No:

DEPARTMENT OF COMPUTER ENGINEERING

STES’S SINHGAD ACADEMY OF
ENGINEERING KONDHWA(BK.), PUNE 411048

SAVITRIBAI PHULE PUNE UNIVERSITY

 SAVITRIBAI PHULE PUNE
UNIVERSITY 2019-20

CERTIFICATE

This is to certify that the project report entitled

SIGNATURE VERIFICATION BASED ON ONE REAL SIGNATURE

Submitted by
MITESH PARMAR Exam No:
NUPUR PURANIK Exam No:
DHRUVA JOSHI Exam No:
SONAL MALPANI Exam No:

is a bonafide student of this institute and the work has been carried out by Students
under the supervision of Prof B.S. Thakare and it is approved for the partial fulfilment
of the requirement of Savitribai Phule Pune University, for the award of the degree of
Bachelor of Engineering (Computer Engineering).

(Prof. B.S. Thakare)

Internal External Examiner
Dept. of Computer Engineering

(Dr. B. B. Gite) (Dr. K. P. Patil)

Head, Principal
Dept. of Computer Engineering Sinhgad Academy of Engineering

Place: Pune
Date:
 ABSTRACT

Currently a lot of time is needed for the verification of signature manually. The need of
developing an automated checking system is felt because of signature forgery in various
transactions. The dynamic signature is a biometric trait which is used in identification.
Automatic signature verification is an application of Image Processing. The aim of the
model is identifying correct signature for reducing fraudulent transactions. It is difficult
to have multiple signatures of the same person so the idea is to duplicate a given
signature number of times by inducing variations and using verifiers to give a
similarity. For duplication of signature the approach of using the Cognitive Inspired
Model is used. The approach for creating human like signatures can be done by
introducing Intra-Component and Inter-Component variability. For verification,
Structural Similarity Index (SSIM) and Image Hashing techniques are used.
Keywords: SRSS, Cognitive Inspired Model, Signature Duplication, SSIM, Image
Hashing
 ACKNOWLEDGEMENTS

It gives us great pleasure in presenting the preliminary project report on ‘SIGNATURE

VERIFICATION BASED ON ONE REAL SIGNATURE’.

I would like to take this opportunity to thank my internal guide Prof. B. S. Thakare for
giving me all the help and guidance I needed. I am really grateful to him for his kind
support. His valuable suggestions were very helpful.

I would also like to thank my project co-ordinator Prof. Santosh Shelke for giving me
the necessary guidance and support. I would also like to thank him for his useful inputs
and tips which really helped this project.

I am also grateful to Prof. B. B. Gite, Head of Computer Engineering Department,

SINHGAD ACADEMY OF ENGINEERING for his indispensable support,
suggestions.

Mitesh Parmar
Nupur Puranik
Dhruva Joshi
Sonal Malpani
(B.E. Computer Engg.)
 INDEX

1 INTRODUCTION 1

OVERVIEW 2

MOTIVATION 3

PROBLEM DEFINITION AND OBJECTIVE 3

PROBLEM SCOPE AND LIMITATIONS 4

METHODOLOGIES OF PROBLEM SOLVING 5

Dataset 5

Image Processing 7

OpenCV 7

p-Hash and d-Hash 8

Structural Similarity Index 8

2 LITERATURE SURVEY 9

3 SOFTWARE REQUIREMENTS SPECIFICATIONS 13

ASSUMPTIONS AND DEPENDENCIES 14

FUNCTIONAL REQUIREMENTS 14

System Features 14

EXTERNAL INTERFACE REQUIREMENTS 15

User Interfaces 15

Hardware Interfaces 15

Software Interfaces 15

NON-FUNCTIONAL REQUIREMENTS 16

Performance Requirements 16
Safety Requirements 16

Security Requirements 16
 Software Quality Attributes 16

SYSTEM REQUIREMENTS 17

Database Requirements 17

Software Requirements 17

Hardware Requirements 17

ANALYSIS MODELS: SDLC MODEL TO BE APPLIED 18

4 SYSTEM DESIGN 20

SYSTEM ARCHITECTURE 21

MATHEMATICAL MODEL 22

DATA FLOW DIAGRAM 23

ENTITY RELATIONSHIP DIAGRAMS 24

UML DIAGRAMS 25

Use-Case Diagram 25

Activity Diagram 28

Sequence Diagram 29

Class Diagram 30

State Transition Diagram 31

Deployment Diagram 32

5 PROJECT PLAN 33

PROJECT ESTIMATES 34

Reconciled Estimates 34

Project Resources 34

RISK MANAGEMENT 36

Risk Identification and Analysis 36

Overview of Risk Mitigation, Monitoring, Management 37

PROJECT SCHEDULE 38
 Project Task Set 38

Task Network 38

Timeline Chart 39

TEAM ORGANISATION 39

Team Structure 39

Management Reporting and Communication 40

6 PROJECT IMPLEMENTATION 41

OVERVIEW OF PROJECT MODULES 42

TOOLS AND TECHNOLOGIES USED 42

Software Tools 42

Programming Languages 43

ALGORITHM DETAILS 44

Algorithm for Sign-Up 44

Algorithm for Sign-In 44

Image Recognition 44

Signature Verification 44

7 SOFTWARE TESTING 45

TYPES OF TESTING 46

User Acceptance Testing 46

Unit Testing 46

Integration Testing 46

Black Box Testing 46

Manual Testing: 47

TEST CASES AND TEST RESULTS 47

8 RESULTS 49

OUTCOMES 50
 SCREENSHOTS 50

Food Freak Application 50

MySQL Database 54

9 CONCLUSIONS 54

CONCLUSIONS 55

FUTURE WORK 55

APPLICATIONS 56

APPENDIX A LABORATORY ASSIGNMENTS ON PROJECT ANALYSIS OF

ALGORITHM DESIGN 57

APPENDIX B REVIEWERS COMMENTS OF PAPER SUBMITTED 60

APPENDIX C PLAGIARISM REPORT 63

APPENDIX D LABORATORY ASSIGNMENTS ON PROJECT QUALITY AND

RELIABILITY TESTING OF PROJECT DESIGN 65

LABORATORY ASSIGNMENT NO. 1 67

Corrective Actions suggested 67

Review and changes based on corrective actions suggested 67

LABORATORY ASSIGNMENT NO. 2 68

Platform 68

Tools Used 68

Libraries Used 68

Deployment 68

PyCharm Setup 69

XAMPP Setup 70
Database 71

LABORATORY ASSIGNMENT NO. 3 72

Functions 72

Resultant GUI 72

LABORATORY ASSIGNEMENT NO. 4 73

Test Tool Selection 73

User Acceptance Testing 73

REFERENCES 77
 LIST OF FIGURES

1.1 Dataset Images 5

3.1 Agile Model 19

System Architecture 22

Data Flow Diagram 24

Entity Relationship Diagram 25

Use Case Diagram 27

Activity Diagram 28

Sequence Diagram 29

Class Diagram 30

State Transition Diagram 31

Deployment Diagram 32

Task Network 38

Sign-up Screen 50

Log-in Screen 50

User Profile 51

Output 52
 Anaconda Setup 69

XAMPP Setup 70

D.3.2 Resultant GUI 71

 LIST OF TABLES

1.1 Dataset 5

2.1 Literature Survey 9

3.1 Hardware Requirement 18

4.1 Use-Case 26

Estimate of Cost 34

Project Implementation Plan 34

Risk Table 36

Risk Probability Definition 36

Risk Impact Definition 39

Team structure and Responsiblities 39

Suggestions and Remarks 40

7.1 Unit Test Cases and Result 47

D.4.2 User Acceptance Testing 73

CHAPTER 1

INTRODUCTION
OVERVIEW

Since the time when authentication came into existence Signatures became an integral
part of the process. Be it any legal documents, agreements, bank documents like
cheques, forms, wills the need of Signature is of utmost importance. The 21st century
has bought great advancements in the development of various processes of
authentication like iris recognition, fingerprint detection but yet Signatures seem to be
inevitable. This gave the rise to improve systems for detecting signatures. On a positive
this has been a remarkable achievement showing the capability of us humans and shows
how much science and technology has advanced. The need of Detection was because of
forged signatures. Any person trying to duplicate signatures is known as forgery. This
needed to be checked.

In this project the implantation of Image Processing has been utilized in Python. It is
not convenient for a person to give Signatures with slight differences is not possible.
Image Processing is one of the upcoming and efficient ways of processing images
which uses various modification representation techniques and algorithms to identify
and change images with a few changes as needed making the task of the user easier and
efficient. Image Processing has been implemented in python with is an object orient
language and a structured programming language. Python uses dynamic typing,
reference counting and a cycle-detecting garbage collector for memory management. It
also features late binding, which binds method and variable names during program
execution. Python was designed to be highly extensible.
MOTIVATION

● Signature is one of the most important needs in the process of authentication.

● Since Signatures are a physical form, it can be forged with skill making it
difficult to check the authenticity
● Signatures need to be checked using a verifier for prediction of genuine or
forged.
● This verification system can be helpful for detecting forgeries.
● The main motivation of our project is to capture a picture of a signature,
duplicate it by adding variations and finally check if a new signature has
similarity with the duplicates to predict the result.
● So the variations are added by using Image Processing techniques, thus only
One Signature per Signer is required which is the most convenient way as taking
multiple signatures is not feasible in any system.

PROBLEM DEFINITION AND OBJECTIVE

The title for our project is ‘Signature Verification Using One Real Signature’. The main
purpose of the project is to verify a signature which requires the user to click the picture
of his/her signature and upload it to the application, here the application creates
duplicates which are used in the verification process and when a new signature from the
same signer is given as an input, similarity is found to give the result as forged or
genuine.

The objectives of our project are:

● To understand and implement the working of Image Processing for duplication
of signature.
● To create duplicates based on One Real Signature.
● To verify signatures using Similarity between the duplicated signatures and any
new signature from the same signer.
PROBLEM SCOPE AND LIMITATIONS

The scope of our project is to help in verification process of the signature. Here the
single signature system is unique and needed as it is not possible to collect various
signatures. Signature plays an important role in authentication process. Due to
advancement in technology the forgery rate has also gone up. This is a counter-measure
to authenticate correct information. While we are taking only one signature, the goal is
to create maximum duplicates as the signature of the same person also varies under
different circumstances. It would be helpful in banks where cheques have to be
processed, also in case of wills and legal documents. The information provided by the
user in the form of signature would be processed and a final output would be given.
This output is given to the banks or any organization using the system to differentiate
between real and duplicate signatures. This is done so as to prevent forgeries and use
signatures in a safer way for authentication.

Limitations-
- It creates less differences if the signature is complex and hence accuracy id
affected.
- If external reference of any object is present then it might result into reduced
prediction accuracy.
METHODOLOGIES OF PROBLEM SOLVING

Dataset

The dataset we used is a self-generated data set. Every time a new user registers a
signature is provided by the user. This signature is taken for processing. We have used
OpenCV for creating duplicates of the signature which comes as an input. The
duplicates are created by adding variations to each part of the signature. The most
variations are observed in the curves of any signature and hence maximum duplicates
have differences in curves. The number of duplicates per signature depends widely on
the number of components. However on an average 50 duplicates per signature is
formed. This is indeed a complex dataset, as when the verification process needs to be
done, the new signature is compared with only its duplicates that were previously
created. Our proposed system gives a correct result 89.6% of the times when checked
with 20 different signers. To calculate this result we have taken into account the real as
well as forged signature of the signer.
Figure 1.1 Dataset Images
 Image Processing:

Image processing deals with image-to-image transformation. The input and output of image
processing are both images. Image Processing is a method to perform some operations on the
image, in order to get an enhanced image or extract useful information from this image. It is a
type of signal processing in which the input is an image and the output can be an image or its
important characteristics. The process involves 3 main steps: Importing the image. Analysing
and manipulating the image. Output with the necessary changes. For the proposed system, the
input data are the signatures which need to be read and given further for addition of
duplications. The duplications formed here is a process of Image Modification based on certain
logic. For this we have used OpenCV. The manipulations are done using features of OpenCV
and the duplicates are created.

OpenCV:

OpenCV (Open Source Computer Vision) is a library of programming functions mainly aimed
at real-time computer vision. It is library used for Image Processing. It is mainly used to do all
the operation related to Images. OpenCV was originally developed in C++. In addition to it,
Python and Java bindings were provided. OpenCV runs on various Operating Systems such as
windows, Linux, OSx, FreeBSD, Net BSD, Open BSD, etc.OpenCV is used for all sorts of
image and video analysis, like facial recognition and detection, license plate reading, photo
editing, advanced robotic vision, optical character recognition,verification.

p-Hash

pHash is an open source software library released under the GPLv3 license that implements
several perceptual hashing algorithms, and provides a C-like API to use those functions in your
own programs. Perceptual hashing can be applied to a wide variety of situations. Similar to
comparing images for copyright infringement, a group of researchers found that it could be used
to compare and match images in a database. A perceptual hash, is a generated string (hash) that
is produced by a special algorithm. This perceptual hash is a based on some input picture, that
can be used to compare images by calculating the Hamming distance (which basically counts the
number of different individual bits). If you use another hashing technique for comparing images,
making the slightest change to the picture, will generate a totally different hash. Perceptual hash
uses similar approach but instead of averaging relies on discrete cosine transformation.
 d-Hash

Difference hash uses the same approach as a-hash, but instead of using information about
average values, it uses gradients (difference between adjacent pixels). The difference is that with
image hashing, if two pictures look practically identical but are in a different format, or
resolution (or there is minor corruption, perhaps due to compression) they should hash to the
same number. Despite the actual bits of their data being totally different, if they look parctically
identical to a human, they hash to the the same thing.

SSIM

SSIM stand for Structural Similarity index. The Structural Similarity Index (SSIM) is a
perceptual metric that quantifies image quality degradation* caused by processing such as data
compression or by losses in data transmission. It is a full reference metric that
requires two images from the same image capture— a reference image and a processed image.
The processed image is typically compressed. It may, for example, be obtained by saving a
reference image as a JPEG (at any quality level) then reading it back in. SSIM is best known in
the video industry, but has strong applications for still photography. Any image may be used,
including those of Imatest test patterns such as Spilled Coins or Log F-Contrast. SSIM actually
measures the perceptual difference between two similar images. It cannot judge which of the
two is better: that must be inferred from knowing which is the “original” and which has been
subjected to additional processing such as data compression
 CHAPTER 2

LITERATURE SURVEY
Authors Main Features Database Approach Results/Conclusions

Moises Diaz Neuromuscular SUSIG-Visual Sub- Dynamic Time False Rejection

corpus, SUSIG-Blind Rate(FRR):2.15
et al.[1] Model, Sigma- Sub-corpus, SVC- Warping(DTW),
Lognormal Model Task1 Sub-corpus, Manhattan Distance False Acceptance
SVC-Task2 Sub- Rate(FAR):2.10
corpus, MCYT100 Based Verifier, Hidden
Sub-corpus, SG- Markov Model(HMM)
NOTE Database
Kamlesh Discrete Cosine Region Based Neural Networks, HMM Dynamic features
Kumari, et al. Transform, Linear Signature Database cannot be extracted.
[10] discriminate
analysis, SURF

Moises Diaz Cognitive Inspired GPDS-300, Support Vector FRR: 1.9%

et al.[9] Model, Intra MYCT-75 Machine(SVM),HMM
component and inter
component
variability

Moises Diaz Cognitive inspired MCYT HMM Improves the

et al.[8] duplicate model, accuracy
Bresenham Line
drawing algorithm,
Local Binary pattern,
Local Derivative
pattern.
Anastasia Discrete Wavelet Local database HMM K-Nearest Neighbour
Beresneva , et Transform, Discrete (KNN) performs
al[.7] Fourier Transform better.

Miguel A. Cognitive Model, Synthesizer Data HMM,SVM,DTW HMM performs

Ferrer , et al. Neuromotor Model Set, better.
[3] MCYT330,NISDCC,
SVC2004, SUSIG
BLIND, SUSIG
VISUAL
Mohsen Broyden–Fletcher– SVC2004, SVS2004 SVM Equal Error
Fayyaz, et al. Goldfarb–Shanno Rate(EER) is
[14] algorithm, Auto reduced.
encoder

Syed Khaleel Dynamic Feature Local Database NN FAR: 5.17%

Ahmed, et al. Extraction, Self FRR:31.33%
[11] Organization
Map(SOM)

Napa Sae Bae, Histogram Feature MCYT-100, SUSIG HMM FAR: 4.66%
et al.[15] Extraction, User
Template
Generation,
Manhattan Distance

B.S Thakare, SIFT,LBP for GPDS-300 Markov Random Model FRR: 16.62%
H.R feature extraction, FAR:14.33
Deshmukh[15 Computer Vision
]

Bhushan S. Simulated annealing CPDS & CEDAR SVM FRR: 17%

Thakare,Dr.
H.R.
Deshmukh[12
]

Zhihua Xia, et Two step verification SGNOTE, K-Nearest Neighbour 2 step verification
al. [2] MCYT-100 (KNN) improves the
accuracy

Luiz G. LBP gradient, MCYT,CEDAR, SVM FAR:4.5 %

Hafemann et Simulated annealing. GPDS-160 FRR: 4.36%
al. [16] FGM - Fast gradient
method,
CNN Signet for
feature extraction
Taraggy. M. HOG, Random Local Database SVM FRR: 6%
Ghanim et al. Classifier, Bagging
[17] Tree Classifier

Yasmine Conventional CEDAR, SVM classifier CEDAR:

Serdouk et al. Artificial Immune GPDS-300 FAR:19.60%
[18] Recognition System FRR:8%
for classification,
Gradient local binary GPDS-300:
pattern for signature FAR:30.69%
characterization FRR:9.16%

Victor K.S.L. Leaky rectified IRONOFF-300 Full Convolutionary

Melo1 et al. linear units as Neural Network(CNN)
[19] activation function

Songxuan Lai 1. DTW MYCT-100 Recurrent Neural

2. Gated Auto Network (RNN)
et al.[20]
Regressive Units
(GARU)
BiswajitKar et 1. Skeleton tree MCYT -100 for SVM 1.0% EER
al.[6] matching. skilled forgery
2. Dynamic
time
warping.
3. Gaussian
mixture
model.
 CHAPTER 3

SOFTWARE REQUIREMENTS SPECIFICATIONS

 ASSUMPTIONS AND DEPENDENCIES

Assumptions:
1. The end user device should be a laptop.
2. Additionally, the end user has used the application before for his duplicates to be present in
the database.
3. The system has a dataset inbuilt in the backend which stores all duplicates for
verification and prediction purpose.

Dependencies:

1. The system is dependent on the end user device.

2. The image clicked by the user is dependent on our system for duplication.
3. The prediction and analysis purpose are dependent on the types of algorithms
used.
4. The image recognition is dependent on the shape, size and colour of the
document image.

FUNCTIONAL REQUIREMENTS

System Features

● Provides a medium to duplicate a given signature and verify it.

● Uses a software application for the interface between end user and
image recognition techniques.
● Simple, easy to use desktop application for the end users.
● Uses a relational database system (SQLLite3) to store the data.
 EXTERNAL INTERFACE REQUIREMENTS

User Interfaces

The user needs to give his original signature while signing up. User will click a picture
of signature and upload it. The picture will be sent to the program for processing and
duplicating. The duplicates are stores as dataset for later use. When the signer gives the
signature for verification, this new signature is checked with the duplicates and the
result is given as an output.

Hardware Interfaces

The Personal Computer (PC) is used to access the desktop application. It acts as a
medium through which the user can interact with the database. User inputs his/her data
as required by the application and is received by PC and later the system stores this data
into the database.

Software Interfaces

The desktop application interacts with the Flask, MYSQLLite server where the database
is accessed for login/signup purpose. Here the records are stored as given by the user.
These records are retrieved by the desktop application for further analysis and
prediction.
 NON-FUNCTIONAL REQUIREMENTS

Performance Requirements

Accuracy: The system can verify a signature to be forged or real with a accuracy of
89.6% using p-hash, d-hash and SSIM verifiers.

Privacy: The authentication is provided in form of user id, password which ensures that
no unauthorized person can access other user’s records.

Usability: The system requires the user to click the picture of signature. Main objective
is to duplicate using single reference system and predict the result.

Safety Requirements

● The image which is being detected needs to be of good quality i.e the image of
document with signature needs to be visible.
● When clicking the image the signature, any part of the document can be visible
in the image but it may extract the exact image.
● When cropping the image, only the part of the image which has to be detected
should be cropped by the user.

Security Requirements

Third-party access to the dataset should be restricted and care should be taken to check
if unauthorized access to the application is also prohibited.

Software Quality Attributes

● Availability: The system does not crash and performs its function in every
possible scenario.
 ● Robustness: The system reduces the impact of operational mistakes or invalid
data.

● Extensibility: The system can further be extended to detect any number of signatures.
● Portability: The system can operate and adapt in new environments efficiently.

SYSTEM REQUIREMENTS

Database Requirements

The database is required to be created and maintained in SQLLite Server. SQLLite is a

software library that provides a relational database management system. The lite in
SQLLite means light weight in terms of setup, database administration, and required
resource. SQLLite was designed originally on August 2000. Dwayne Richard
Hipp (born April 9, 1961) is a software developer and the primary author of SQLLite.
SQLLite is an open source relational database management system (RDBMS). SQLLite
is written in C.

Software Requirements

Platform:
1. Operating System: Windows 7/8/10 (32 or 64bit)
2. IDE: Anaconda(Jupyter Notebook, Spyder)
3. Programming Language: Python, JavaScript, HTML, MySQLs, jQuery

Hardware Requirements

Sr. Parameter Minimum

No. Requirement
1. PC(Desktop) Should support internet
2. Smart Phone 512 MB – 1GB
RAM
Table 3.1 Hardware requirements
 ANALYSIS MODELS: SDLC MODEL TO BE APPLIED

Agile SDLC model is a combination of iterative and incremental process models with focus on
process adaptability and customer satisfaction by rapid delivery of working software
product. Agile Methods break the product into small incremental builds. These builds are
provided in iterations. These builds are provided in iterations. Each iteration typically lasts from
about one to three weeks. Every iteration involves cross functional teams working
simultaneously on various areas like −

● Planning

● Requirements Analysis

● Design

● Coding

● Unit Testing and

● Acceptance Testing.

Figure 3.1 Agile Model

Following are the Agile Manifesto principles −
1. Individuals and interactions − In Agile development, self-organization and
motivation are important, as are interactions like co-location and pair
programming.

2. Working software − Demo working software is considered the best means of

communication with the customers to understand their requirements, instead of
just depending on documentation.

3. Customer collaboration − As the requirements cannot be gathered completely

in the beginning of the project due to various factors, continuous customer
interaction is very important to get proper product requirements.

4. Responding to change − Agile Development is focused on quick responses to

change and continuous development.

All these phases are intervened to each other in which progress is seen as flowing
steadily like a spiral through the phases. The next phase is started and we come back to
that phase if any changes are required, so the name Agile Model. In this model phases
do overlap.
 CHAPTER 4

SYSTEM DESIGN
SYSTEM ARCHITECTURE

Figure 4.1: System architecture

 MATHEMATICAL MODEL

Input x, Output y
Therefore, y=f(x)

I: Set of Inputs = {I1,I2,I3}

where,

I1: Login
I2: User details
I3: Image of Signature

O: Set of Outputs = {O1}

where,

O1 : Prediction if Forged or
Authentic

F : Set of Functions ={F1,F2,F3,F4,F5,F6,F7}

F1 : The user signs up into the application.

F2 : The application asks the user to upload signature for first
time
F3 : The image is processed and its duplicates are formed and
saved.
F4 : The user when logs in is asked to upload signature to be checked
F5 : The image is checked with its duplicates previously formed and similarity is calculated.
F6 : Based on similarity, output is given
F7 : Signature authenticity is displayed.

Let S be the solution perspective of the class

S={s, e, i, o, f, success, failure}
s=Start of program
e = The end of program
i= Data of the user and image to be tested
o=Signature authenticity is displayed.
success= Accurate prediction within deadline
failure= Incorrect prediction of signature.
 DATA FLOW DIAGRAM

A data flow diagram(DFD) is a graphical representation of the flow of data through an

information system. DFDs can also be used for the visualisation of data
processing(structural design). It is therefore quite different from a flowchart, which
shows the flow of control through an algorithm.
A DFD provides no information about a timing of processes, or about whether
processes will operate in sequence or in parallel. It allows a reader to determine what
operations will be performed, in what order and under what circumstances, but not what
kinds of data will be input to and output from the system, nor where the data will come
from and go to, nor where the data will be stored.

Figure 4.2 Data Flow Diagram

 ENTITY RELATIONSHIP DIAGRAMS

An entity relationship model (ER model for short) describes interrelated things of
interest in a specific domain of knowledge. A basic ER model is composed of entity
types (which classify the things of interest) and specifies relationships that can exist
between instances of those entity types. In software engineering, an ER model is
commonly formed to represent things that a business needs to remember in order to
perform business processes. Consequently, the ER model becomes an abstract data
model, that defines a data or information structure which can be implemented in a
database, typically a relational database.

Figure 4.3 Entity Relationship Diagram

 UML DIAGRAMS

Use-Case Diagram

A use case diagram in the Unified Modelling Language (UML) is a type of a behaviour
diagram defined by and created from a use case analysis. Its purpose is to present a
graphical overview of the functionality provided by a system in terms of actor, their
goals (represented as use cases), and any dependencies between those use cases. The
main purpose of a use case diagram is to show what system functions are performed for
which actor. Roles of the actors in the system can be depicted.

Sr Use-Case Description Actor Assumptions

. s
N
o.
1. Image Capture New user sign-up to the User User is a
system and clicks the client and
signature image using a therefore
camera of which the system need to
duplicates need to be made have one
and signature needs to be original
verified. signature.

2. Select The user is made to bound User User

Bounding the signature image using a
has
Box bounding rectangle which uploaded the
document
crops the image if not
with signature
extracted exactly. to the system.
3. Image The system then recognises Syste User crops the
Signature from the document m Image within
Extractio
and extracts for the process of box.
n duplication.
4. Image After recognising the curves Syste System
m
Duplication variations are added to observes
induce intra and inter similarity in
personal variability. The Images and
variations make duplicates of adds
 signature. modifications.
.
5. Signature System predicts if the new Syste Similarity is
Prediction m found by using
signature is forged or real by
SSIM, p-hash
finding similarity between and
d-hash
duplicates and image.
techniques.
6. Confirmation Question is popped to the user User Signature
of prediction for the purpose of
is
confirmation
predicted.
of predicted
7. Input For Verification user needs User User confirms
to login, thus the system log-in details.
of
checks if the new image
User Details
given at this time is similar
with previous duplicates of
the signer.
8. View The system keeps record of User Using the
Statistics
correctly displayed signature. input from
user, statistics
are displayed.

Table 4.1 Use-Cases

Figure 4.4 Use Case Diagram
 Activity Diagram

Activity diagrams are graphical representation of workflow of stepwise activities and

action with support of choice, iteration and concurrency. In the UML, activity diagram
are intended to model both computational and organizational processes. Arrows run
from the start towards end and represent the order in which activity happens. Hence
they can be regarded as form of flowchart technique lack construct for expressing
concurrency. However the join and the splits symbol in activity diagrams only resolve
this from a simple cases, the meaning of a model is not clear when they are arbitrarily
combined with decision or hope.

Figure 4.5 Activity Diagram

 Sequence Diagram

A sequence diagram in UML is a kind of interaction diagram that shows how processes
operate with one another and in what order. It is a construct of message sequence chart.
A sequence diagram shows object interaction arranged in time sequence. It depicts the
objects and classes involved in the scenario and sequence of messages exchanges
between the objects needed to carry out the functionality of scenario. Sequence
diagrams are typically associated with use case realizations in the logical view of the
system under development.

Figure 4.6 Sequence Diagram

 Class Diagram

A class diagram in the UML is a type of static structure diagram that describes the
structure of a system by showing the system classes, their attributes and the
relationships between the classes. This diagram shows various classes or main entities
involved in the system and also their relationship with each other. It depicts the
attributes and operations each class can carry out, individually and with the help of
other classes in the system designed

Figure 4.7 Class Diagram

 State Transition Diagram

UML state machine diagrams depict the various states that an object may be in and the
transitions between those states. Infact, in other modelling languages it is common for
this type of a diagram to be called a state transition diagram or even a simple state
diagram. A state represents a stage in the behaviour pattern of an object. A transition is
a progression from one state to another and will be triggered by an event that is either
internal or external to the object.

Figure 4.8 State Transition Diagram

Deployment Diagram

Figure 4.9 Deployment Diagram

CHAPTER 5

PROJECT PLAN
 PROJECT ESTIMATES

Reconciled Estimates

Cost Estimates

Compone Cost in
nt INR
Paper Publication fees per person ₹2000
Project Competition -
Table 5.1: Estimate of Cost

Time Estimates

Activity Duration
Project Team Identification 10 days
Topic Selection 2 weeks
Research for proposed system 3 weeks
Design of the proposed system 1 month
Implementation of the signature 2 months
verification application.

Testing 3 weeks
Documentation 1 week
Table 5.2: Project Implementation Plan

Project Resources

Everyone and everything involved in the project:

People
● People
- Mitesh Parmar
- Nupur Puranik
- Dhruva Joshi
- Sonal Malpani
 ● Project Guide
- Prof. Bhushan S. Thakare

Software Tools
● Windows 7/8/10 (32/64 bit)
● Anaconda

Programming Languages
● Python
● JavaScript
● HTML
● MySQL
● jQuery
 RISK MANAGEMENT

Risk Identification and Analysis

The risks for the Project can be analysed within the constraints of time and quality

I Risk Description Probabilit Impa

D y ct
Schedul Quality Overall
e
1 Signature Mediu Very Very Very
. Prediction m High High High
Failure
2 Low Very Very Very
Signature
. High High High
Duplication
Failure

3 Software Mediu Low Low Low

. Database m
interfacing
failure
Table 5.3: Risk Table

Probabili Probability of
ty occurrence
High > 75%
Medium 26 – 75 %
Low < 25%
Table 5.4 Risk Probability Definitions

Impact Value Descripti

on
Very high > 10% Schedule impact or Unacceptable quality
High 5-10% Schedule impact or some parts of the project have low
Quality
Medium < 5% Schedule impact or Barely noticeable degradation in
quality Low impact on schedule or quality can be
incorporated

Table 5.5 Risk Impact Definitions

Overview of Risk Mitigation, Monitoring, Management

Following are the details for each risk.

Risk ID 1
Risk Description Signature Prediction Failure
Category Technology
Source Software requirement specification document
Probability Medium
Impact Very High
Response Monitor
Strategy Training the system with more and complex images will help
improve the results.
Risk Status Identified

Risk ID 2
Risk Description Signature Duplication Failure
Category Technology
Source Software requirement specification document
Probability High
Impact High
Response Mitigate
Strategy Using a better dataset with approximately correct
corresponding calorie value for training.
Risk Status Identified
 Risk ID 3
Risk Description Software Database interfacing failure
Category Technology
Source Identified during early development and testing
Probability Medium
Impact Low
Response Accept and Improve means to store local data
Strategy Identify database servers and problems, technology used
and mitigate issue
Risk Status Identified

+PROJECT SCHEDULE

Project Task Set

● Task 1: Design and deploy an application for interaction with end user.

● Task 2: Design and maintain the schema and database connection with front end
for signature and credential record management.
● Task 3: Testing the connection and interfacing, checking for proper functionality
and working before deployment.
Figure 5.1 Task Network
 TEAM ORGANISATION

Team Structure

Sr. Modu Responsible Student

No. le
1 Log-in module Dhruva Joshi
2 Sign-up module Nupur Puranik & Sonal
Malpani
3 Image Extraction module Mitesh Parmar & Dhruva
Joshi
4 Mitesh Parmar
Image Duplication Module

5 Presentation module Nupur Puranik

6 Similarity Module Nupur Puranik & Mitesh
Parmar
7 Prediction Module Mitesh Parmar
8 Database schema and design Dhruva Joshi

Table 5.6 Team structure and responsibilities

Management Reporting and Communication

Sr Activity Scheduled Date Remarks and

. Suggestions
N
o
1) Review I

2) Review II

3) Mock Project Stage-I -

4) Project Stage-I exam -

5) Review III

6) Paper conference

7) Paper conference

Table 5.7 Suggestions and Remarks

 CHAPTER 6

PROJECT IMPLEMENTATION
6.1 OVERVIEW OF PROJECT MODULES

The project is comprised of following modules:

1. User sign-up module.
2. User login module.
3. Image recognition module.
4. Signature verification module.

6.2 TOOLS AND TECHNOLOGIESUSED

The tools used for the development of this project are:

6.2.1 Software Tools

● Windows 7/8/10(32/64bit)
The project was implemented on a Laptop running Windows 10 64-bit, on a 6 th
generation Intel core i5 processor 5Ghz with 8GB RAM.

● Spyder and Jupyter Notebook

Spyder is an open source cross-platform integrated development environment
for scientific programming in the Python language. The Jupyter Notebook is an
open-source web application that allows you to create and share documents that
contain live code, equations, visualizations and narrative text.

● Database Server
A website hosting is deployed. It gives a panel for managing web pages, PHP
scripts and also provides SQLLite database server for querying and record
storage.
6.2.2 Programming Languages

● Python
Python is an interpreted, high-level, general-purpose programming language.
Python uses dynamic typing, reference counting and a cycle-detecting garbage
collector for memory management. It also features late binding, which binds method
and variable names during program execution. Python was designed to be highly
extensible.

● JavaScript
JavaScript is the programming language of HTML and the Web. It is a prototype-
based, multi-paradigm, dynamic language, supporting object-oriented, imperative,
and declarative (e.g. functional programming) styles.

● HTML
The Hypertext Markup Language (HTML) is the most widely used language on
Web to develop web pages. It is a standard markup language for creating web pages
and web applications. With Cascading Style Sheets (CSS) and JavaScript, it forms a
triad of cornerstone technologies for the World Wide Web.

● MySQL
MySQL is an Oracle backed open-source relational database management system
(RDBMS) based on Structured Query Language (SQL).MySQL runs on virtually
all platforms, including Linux, UNIX and Windows. MySQL is an important
component of an open source enterprise stack called LAMP.

● jQuery
jQuery is a JavaScript library designed to simplify HTML DOM tree traversal and
manipulation, as well as event handling, animation, and Ajax. jQuery is a light
weight, ”write less, do more”, JavaScript library. The purpose of jQuery is to make
it much easier to use JavaScript on your website. jQuery also simplifies a lot of the
complicated things from JavaScript, like AJAX calls and DOM manipulation.
6.3 ALGORITHMDETAILS

6.3.1 Algorithm for Sign-Up

● Enter details name, user name and password
● User name and password used for Sign-in purpose.

6.3.2 Algorithm for Sign-In

● The Email-ID is the username.
● The password is which is given by the user.

6.3.3 Image Recognition

● Pre-processing Data(Normalization)
● Feature Extraction
● Image Recognition
● Cropping of image
● Features extracted are stored in a vector form
● Images are recognized according to its category.

6.4.4 Signature Verification

● The image is entered by the user which is then cropped.
● The cropped image is detected and then it is compared with the
database.

● After comparing with the database the result message is displayed

whether the signature is Genuine or forged.
 CHAPTER 7

SOFTWARE TESTING
 TYPES OF TESTING

7.1 TYPES OFTESTING

7.1.1 User Acceptance Testing

User acceptance testing is concerned with what is to be testes from the users view point
of what the system does. This is not a technical description of the software, but the
USERS view of the functions.

7.1.2 Unit Testing

Unit testing focuses verification effort on the smallest unit of the software design the
software component or module. The important control paths are tested to uncover errors
within the boundary of the module. The unit test is white-box oriented. In our software
product the following components would undergo unit testing:

7.1.3 Integration Testing

Integration testing is a systematic technique for constructing the program structure

while at the same time conducting tests to uncover errors associated with interfacing.
The objective is to take unit tested components and build a program structure that has
been dictated by design. Therefore, after the unit testing of the individual modules, the
integration of these modules would be carried out. The integration strategy used will be
Bottom-Up Integration. GUI testing also carried out in this phase.

7.1.4 Black Box Testing

Black-box testing is a method of software testing that tests the functionality of an

application as opposed to its internal structures or working. Specific knowledge of the
internal working of the applications code is not required. The tester is only supposed to
be aware of what the software is supposed to do, but not how. This testing strategy is
used to check whether the system provides the expected output with the predefined
inputs.
7.1.5 Manual Testing:

Manual testing is the process of manually testing software for defects. It requires a
tester to play the role of an end user and use most of all features of the application to
ensure correct behavior. The manual testing is very basic type of testing which helps to
find the bugs in the application under test. It is preliminary testing, must be carried out
prior to start automating the test cases and also needs to check the feasibility of
automation testing.

7.2 TEST CASES AND TESTRESULTS

T Objectiv Pre- Descripti Te Expect Actu

e
es Conditi on st ed al
t on Da Result Res
C ta ult
a
s
e
ID
.:
1 User User is To store Userna Encrypt Password
Sign- Up using the user me passwor creation
applicati login and d and and user
on for the credentia passw successf created
first time ls or d ul user successfull
creation y
2 User User To Userna User The
Logi already authentic me authentic authentic
n has ate user and ati-on ati on
signed up passw based on process
or d data sent successfu
to l
server
3 Image User To Signat Correct Signature
recognit has check if ure predicti is
io n upload the image on of predicted
ed the Signatu Signatu correctly.
image re is re.
predicte
d
correctly
4 Signat The To The The The
ure cropped compar user’s Duplicate Duplicate
Dupli image is e the signat s of s are
cation used for signatur ure signature generated
duplicati e based on and stored
ons. variation successful
s. ly
5 Signatur The User To The The Signature
e has verify signatu signature is verified
uploaded
verificat the image the re is is successful
ion. signatur compa verified ly
e red correctly.
whether within
It is databa
genuine se
or
forged
6 Databas Internet To - Databas Databas
e connecti connect e server e server
connecti on to connecti connecti
vity databas on on
e successful
Server
7 All tests - All the - All Cleared all
above tests the test
mentione must successfull
d test be y
cases cleare
should be d.
passed.

Table 7.1 Unit Test Cases and Results

CHAPTER 8

RESULTS
OUTCOMES

● A simple e-signature application recognizes the signature and predicts if any

new signature by the same user is genuine or forged
● An accuracy graph is given to show correctly VS non-corrected predicted
signatures
● Duplicates are formed based on One Real Signature.

SCREENSHOTS

Signature Verification Application

Figure 8.1 Sign-up Screen

Figure 8.1 Log-In Screen

Figure 8.3 User Profile Screen

Figure 8.4 Output Screen
MySQL Database

Figure 8.13 User’s Database

CHAPTER 9

CONCLUSIONS
CONCLUSIONS

Signature Verification System is an advancing field and has varied usage such as
Banking sector, legal documents etc. A user while signing any document does not
have the same signature always and has a few variations. It is necessary to develop a
system that can incorporate these variations and can accurately predict if the
signature is forged or genuine. One Real Signature is given emphasis in our
experiment as it is not feasible to collect multiple signatures. Our system modifies
the One Real Signature by adding slight variations. Further when the same signer
gives an input for the second time, this new image is checked with its previously
made duplicates. Finally we used SSIM and Image Hash to give similarity and
predict the output. Thus using Cognitive Inspired Model for image duplication gives
a better approach rather that previously used techniques which uses mathematical
formulae as this model gives a cognitive insight

FUTURE WORK

● The proposed system would be enhanced when we use more complex

algorithms for signature duplication so that more accurate duplications are done.

● The work can be further extended to multiple languages. Currently, the

duplications is majorly done by the process of component labeling and then
identifying the curves in those components. It can be extended to multi-lingual
signatures.

● The process of signature verification can be further improved by enhancing the

verification techniques. If techniques can identify dynamic signatures, the pen-
up and pen-down movement all the other factors that affect signature change can
also be considered.

● As fast response is one of the most important factors nowadays, the

computational part of the application (i.e., feature extraction, duplication, image
verification etc) which takes longer time. With the use of cloud, an application
can be expected which takes less time to produce the result by performing large
computations.
APPLICATIONS

● Banks: For the account holders of the banks who can collect their signature once
and use it later for verification purpose.

● Legal Documents, Wills: To verify authentic signatures and avoid forgery.

● Attendance System: Can be used by schools and colleges to check if student’s

signature is authentic.
 APPENDIX A

LABORATORY ASSIGNMENTS ON PROJECT ANALYSIS OF

ALGORITHM DESIGN
Aim:
Project problem statement feasibility assessment using NP-Hard, NP-Complete or
satisfy ability issues using modern algebra and/or relevant mathematical models.

1. P type Problem:
The Collection of problems that can be solved in polynomial time is called P.
This polynomial is small degree. The problems belonging to this class are easy
to solve and can be solved using tractable input.

2. NP type Problem:
The algorithm in which every operation may not have unique result, rather there
can be specified set of possibilities for every operation. Such an algorithm is
called non-deterministic algorithm. Non-deterministic algorithm is a two stage
algorithm.

A: Non-deterministic (Guessing) Stage: Generate an arbitrary string that can be

thought of as a candidate solution.

B: Deterministic (Verification) Stage: In this stage, it takes as input the candidate

solution and the instance to the problem and returns yes if the candidate solution
represents actual solution.
There are two types of NP-type problem:

A. NP-complete: They are harder to compute rather than to verify; they could not
be solved in polynomial time but they can be verified in polynomial time.

B. NP-hard: They need not have solutions verifiable in polynomial time.

Mathematical Model:

Input x, Output y
Therefore, y=f(x)

I: Set of Inputs = {I1,I2,I3}

where,

I1: Login
I2: User details
I3: Image of Signature

O: Set of Outputs = {O1}

where,

O1 : Prediction if Forged or
Authentic

F : Set of Functions ={F1,F2,F3,F4,F5,F6,F7}

F1 : The user signs up into the application.

F2 : The application asks the user to upload signature for first
time
F3 : The image is processed and its duplicates are formed and
saved.
F4 : The user when logs in is asked to upload signature to be checked
F5 : The image is checked with its duplicates previously formed and similarity is calculated.
F6 : Based on similarity, output is given
F7 : Signature authenticity is displayed.

Let S be the solution perspective of the class

S={s, e, i, o, f, success, failure}
s=Start of program
e = The end of program
i= Data of the user and image to be tested
o=Signature authenticity is displayed.
success= Accurate prediction within deadline
failure= Incorrect prediction of signature.
 APPENDIX B

REVIEWERS COMMENTS OF PAPER SUBMITTED

PAPER 1

1. Paper Title: State of the Art Literature Survey 2019 on Signature Verification

2. Name of the Conference/ Journal where paper submitted: AJCT

3. Paper Accepted/rejected: Accepted

4. Review comments by reviewer:

 Suggested about the various data sets that can be used.
PAPER 2

1. Paper Title: Image Processing Based Signature Duplication and its Verification.

2. Name of the Conference/ Journal where paper submitted:

3. Paper Accepted/rejected: In-Process

4. Review comments by reviewer:

 APPENDIX C

PLAGIARISM REPORT
 APPENDIX D

LABORATORY ASSIGNMENTS ON PROJECT QUALITY

AND RELIABILITY TESTING OF PROJECT DESIGN
 LABORATORY ASSIGNMENT NO. 1

AIM- Review of design and necessary corrective actions taking into consideration the
feedback report of Term 1 assessment, and other competitions/conferences participated
like IIT, Central Universities, University Conferences or equivalent centres of
excellence etc.

Corrective Actions suggested

● Suggested about the use of various optimizer and trial with each one to predict
the output.
● Trial of other Neural Networks architecture was suggested to converge to a
positive outcome.

Review and changes based on corrective actions suggested

● Adam’s optimizer is used to get better accuracy in results.

● A fully connected 6-layer architecture of Convolutional Neural Network is
created to give promising results.
 LABORATORY ASSIGNMENT NO. 2

AIM: Project workstation selection, installations along with setup and installation
report preparations.

Platform

D.2.1.1 Windows 7/8/10 (32/64 bit)

Standard platform used. Linux or Mac OS X can also be used. Windows was primarily
chosen because of its popularity and the convenience of IDEs that can be installed and
run on it.

Tools Used

Anaconda
SQLLITE Database
Flask

Libraries Used
1. Python 3.6
2. Pandas
3. Numpy
4. Keras
5. Matplotlib
6. TensorFlow
7. OpenCV
8. Flask
Anaconda Setup
XAMPP Setup
 Database

D.2.3.1 SQLLITE

A website with SQLLITE database connectivity was used for the specific purpose of
data management. Structured data, primarily relational based is used for record storage.
The records contains data about the login credentials, images, correctly predicted and
accordingly constructs the accuracy analysis data.
 LABORATORY ASSIGNMENT NO. 3

AIM: Programming of the project functions, interfaces and GUI (if any) as per 1 st Term
term-work, submission using corrective actions recommended in Term-1 assessment of
Term-Work

Functions
1. Login & SignUp: Function that check the credentials and logs into the account
for Login and Inserts details of a new user for SignUp option.
2. Detection: Function that takes a cropped image as input and performs
duplications and gives the prediction after image recognition.
3. Duplication: Function that duplicates the signature given by user and stores it in
system.
4. Similarity Prediction: This gives the similarity between 2 images and predicts if
the signature uploaded is authenticated or not.

Resultant GUI:
 LABORATORY ASSIGNEMENT NO. 4

AIM: Test Tool selection and testing of various test cases for the project performed and
generate various testing result charts, graphs etc, including reliability testing,

Test Tool Selection

Selenium – It is a free (open source) automated testing suite for web applications across
different browsers and platforms. Selenium is a suite of software tools to automate Web
Browsers.

User Acceptance Testing

User acceptance testing is concerned with what is to be tested from the users view point of
what the system does. This is not a technical description of the software, but a USERS view
of the functions.
Sr. Feature to be Level of testing
No. tested required
1 Login Page High
2 Sign-up Page High
3 Button Low
4 User Input Low
5 User interface Medium
6 Link High
7 Test Suite (All High
tests)

1. Login Page – PASSED

2. Signup – PASSED

3. Button – PASSED
4. User Input-PASSED

5. User Interface-PASSED
6. Link-PASSED

7. Selenium Katalon Recorder

REFERENCES
[1] Moises Diaz, Member, IEEE, Andreas Fischer, Member, IEEE,
Miguel A. Ferrer, and Réjean Plamondon, Fellow, IEEE, “Dynamic Signature
Verification Based on One Real Signature”, 2168-2267 _c 2016 IEEE.

[2] Moises Diaz*, Andreas Fischertt §, Rejean Ferrer*,“Towards an Automatic On-

Line Signature Verifier Using Only One Reference Per Signer”, 978-1-4799-
1805-8/15/$31.00 ©2015 IEEE
[3] Andreas Fischer, Member, IEEE, and R´ejean Plamondon, Fellow, IEEE,
“Signature Verification Based on the Kinematic Theory of Rapid Human
Movements”, 2168-2291 © 2016 IEEE
[4] Kamlesh Kumari, V.K. Shrivastava, ”Factors Affecting the accuracy of
Automatic Signature Verification”, 978-9-3805-4421-2/16/$31.00 c_2016 IEEE
[5] Moises Diaz-Cabrera, Miguel A. Ferrer, Aythami Morales, “ Cognitive Inspired
Model to Generate Duplicated Static Signatures”, 2167-6445/14 $31.00 © 2014
IEEE
[6] Moises Diaz, Miguel A. Ferrer, George S. Eskander and Robert Sabourin
Member, IEEE, “Generation of Duplicated Off-Line Signature Images for
Verification System”, 0162-8828 (c) 2016 IEEE
[7] Javier Galbally, Julian Fierrez, Marcos Martinez-Diaz, and Javier Ortega-Garcia,
“Improving the Enrollment in Dynamic Signature Verification with Synthetic
Samples”, 978-0-7695-3725-2/09 $25.00 © 2009 IEEE
[8] Anastasia Beresneva, Anna Epishkina, DarinaShingalova “Handwritten
Signature Attributes for its Verification” 978-1-5386-4340-2/18/$31.00©2018
IEEE .
[9] Moises Diaz, Miguel A. Ferrer, Eskander , Robert Saburin “Generation of
Duplicated Offline Signature Images for Verification Systems” 0162-8828©
2016 IEEE.
[10] Bhushan S. Thakare, Dr. Hemant R. Deshmukh “A Novel End-To-End
Approach For Offline Signature Verification System” 978-1-5386-4273-
3/18/$31.00 ©2018 IEEE
[11] Zhihua Xia, TuanjiaShi ,Neal N. Xiong, Xingmuing Sun, Byeungwoo Jeun,
“A Privacy-preserving handwritten signature verification method using
combinational features and secure KNN”
[12] Miguel A. Ferrer, Moises Diaz, Cristina Carmona-Duarte and Aythami
Morales “A Behavioral Handwriting Model for Static and Dynamic Signature
Synthesis” 0162-8828 (c) 2016 IEEE.
[13] Biswajit Kar, Anirban Mukherjee , and Pranab K. Dutta “Stroke Point
Wrapping-Based Reference Selection and Verification of Online Signature”
0018-9456 © 2017 IEEE.
[14] Mohsen Fayyaz, Mohammad Hajizadeh Saffar, Mohammad Sabokrou, M.
Hoseini and M. Fathy “Online Signature Verification Based on Feature
Representation”.978-4799-8818-1/15/$31.00 ©2015 IEEE.
[15] Napa Sae-Bae and NasirMemon, Fellow, IEEE “Online Signature Verification
on Mobile Devices” 1556-6013 © 2014 IEEE.
[16] Bhushan S. Thakare, Dr. Hemant R. Deshmukh “A combined feature
extraction model using SIFT and LBP for offline signature verification system”
978-1-5386-4273-3/18/$31.00 ©2018 IEEE.
[18] Yashmine Serdouk, Hassiba Nemmourand Youcef Chibani “An Improved
Artificial Immune Recognation System for Off-
line Handwritten Signature

[17] Songxuan Lai, LianwenJin “Recurrent Adaption Networks for Online

Signature Verification” 1556-6013 (c) 2018 IEEE

[18] Victor K.S.L Melo, Byron Leite Dantas Bezerra,Donato Impedovo,Giuseppa

Pirlo ,Antonio Lundgren “Deep Learning Approach to Generate Offline
Handwritten Signatures based on Online samples” IET Biom., 2019, Vol. 8 Iss. 3,
pp. 215-220