Department of Electronics and Communication Engineering

13th KM Stone, Mysuru-Kanakapura-Bengaluru Road,

Mysuru-570028

Pre-Final Year Mini – Project Synopsis

A Project Work Phase-1 Report on

Enhanced Signature Verification Using OCR and Line Sweep

Techniques Based on Machine Learning

Student Name USN

Hemanthkumar M S 4AD22EC032
Karthik M 4AD22EC039

Koustubh N 4AD22EC040

Nandan G 4AD22EC059

Guide Signature

Group No. 34 Guide Name: Rajeev gowda R

 Table of Contents
Sl. No Contents Page No

1. Introduction ……………………… 1

2. Problem Statement ……………………… 1

3. Objectives ……………………… 2

4. Literature Survey ……………………… 2

5. MethodologyAnd Block ……………………… 7

Diagram

6. Expected Outcomes ……………………… 8

7. References …………………………. 9
Machine Learning Based Signature Verification Process Using OCR And Line Sweep Techniques 2024-25

1. INTRODUCTION

Signature verification and forgery detection systems are used to verify a person's identity by
verifying a signature. These systems can be used both online and offline. Online signature
verification and forgery detection systems are used to verify a person's identity by verifying
signature in an online environment. A reference signature is usually stored in a secure
database and is used to compare a person's signature to ensure its authenticity. The system can
also detect possible fake attempts by comparing the signature with a reference signature.
Offline signature verification and forgery detection systems are used to verify a person's
identity by verifying signature in an offline environment.

2. Problem Statement
Handwritten signatures find extensive use as a method for personal identification and
validation. Numerous documents, such as bank cheques and legal agreements, necessitate the
process of signature authentication. However, given the substantial volume of documents
involved, this becomes a challenging and time-intensive undertaking. As a result, the need for
a reliable automated tool for signature verification becomes imperative, with the goal of
mitigating fraud across various sectors of financial transactions.
It involves developing a signature verification process using machine learning, incorporating.
Role theory and lines weep techniques. The goal is to enhance authentication by leveraging
role-based analysis and lines weep features in signature recognition, addressing issues related
to accuracy, security, and robustness in signature verification systems.

Dept. of ECE, ATMECE, Mysuru Page 1

Machine Learning Based Signature Verification Process Using OCR And Line Sweep Techniques 2024-25

3. Objectives
 OCR and Signature Localization: Develop an OCR system capable of accurately
localizing signatures on documents. The system should effectively extract signature regions
while maintaining the integrity of the signature content.
 Connected Components Analysis: Implement a Connected Components analysis to
segment and isolate individual signature regions from complex documents. This will enable
the system to handle cases where multiple signatures are present.
 Line sweep algorithm & geometric features with Support Vector Machine (SVM):
Utilize the Line sweep algorithm to optimize the detection and extraction of signature
regions. The algorithm should efficiently locate and delineate signatures, even in cases of
overlapping or skewed signatures.
 Enhance Accuracy: Utilize advanced OCR and line-sweep techniques to extract detailed
features from signatures, enabling the system to distinguish between genuine and forged
signatures with high accuracy.
 Automate the Verification Process: Develop an automated system to verify signatures,
reducing the need for manual inspection and thereby saving time and resources.
 Improve Efficiency: Design a scalable verification process capable of handling large
volumes of signature verifications quickly.

4. Literature Survey
[1] Sig Net: Convolutional Siamese Network for Writer Independent Offline Signature
Verification. Sounak Dey, Suman [Link] , Anjan Dutta sep(2017)
In this paper, it proposes a detailed method for detecting forged signatures in offline
signature verification using Siamese neural networks. Siamese grids are designed to capture
two different, similar, or different images and calculate a distance measure between higher.
[2] Julita A., Fauziyah S., Azlina O., Mardiana B., Hazura H., Zachariah [Link]
Signature Verification System Issued April 2009 The study concludes that integrating SVM
for online signature verification offers a reliable method for identity authentication. The
system's design, focusing on real-time data acquisition and analysis of dynamic signature
features, provides a robust framework for distinguishing between genuine and forged
[Link].

Dept. of ECE, ATMECE, Mysuru Page 2

Machine Learning Based Signature Verification Process Using OCR And Line Sweep Techniques 2024-25

[3] Off-line Signature Verification Based on Fusion of Grid and Global Features Using
Neural Networks: Shashikumar KB Raja. (Dec-2010)
This proposes offline signature verification using a neural network (SVFGNN) that
integrates global network and features. It combines generic and network functions to create
a set of signature verification functions. In this case, the test signatures are compared with
the database signatures based on the set of features defined using the neural network [3].
The proposed model is divided into three phases: the preprocessing phase, the feature
extraction phase, and the validation phase. In the preprocessing stage, from the signatures
collected with the scanner, signature features are extracted through a preprocessing step
consisting of denoising, volume normalization, and skeletonization.
[4] Poonam Chaudhary and Vijay Kumar Singh, published in Volume 3, Issue 2, February
2016. of the International Journal for Research in Emerging Science and Technology,
provides a comprehensive overview of online signature verification systemsChandra Sekhar
V, Prerana Mukherjee, Devanur S Guru, Viswanath Pulabaigari. Feb- 2019. Online signature
verification based on writer specific feature selection.
[5] Liyang Xie, Zhongcheng Wu, Xian Zhang, Yong Li Xinkuang Wang Writer-Independent
Online Signature Verification based on 2-D Representation of Time Series Data using Triplet
Supervised Network. The area of Handwritten Signature Verification has been broadly
researched in the last decades, but remains an open research problem. The objective of
signature verification systems is to discriminate if a given signature is genuine (produced by
the claimed individual), or a forgery (produced by an impostor).
[6] S. Priya, [Link], K. Somesh, A Hruday Kumar Signature Verification System
using Different Algorithms Volume-8, Issue6S3, April 2019. In our paper, we proposed a
system for signature verification based on pixels and strokes. Nowadays signature is a basic
and important verification system for every [Link] has a unique signature and
every individual can differ from others. In many areas, this verification system is offline and
not so accurate. Online verification is very efficient than the offline one. Our proposed
system is presented to check signature using pixels, strokes etc. It is efficient enough to
compare signatures easil.
[7] Tushara D, Shridevi Raddy, Shreya KM, Spoorthy Y, Signature Verification System
using Neural Networks, NCCDS- 2021 Conference [Link] proposed system

Dept. of ECE, ATMECE, Mysuru Page 3

Machine Learning Based Signature Verification Process Using OCR And Line Sweep Techniques 2024-25
demonstrates the potential of using dynamic signature features in conjunction with neural
networks for effective online signature verification. While specific performance metrics are
not detailed in the summary, the approach suggests improvements over traditional manual
verification methods.
The proposed system demonstrates the potential of using dynamic signature features in
conjunction with neural networks for effective online signature verification. While specific
performance metrics are not detailed in the summary, the approach suggests improvements
over traditional manual verification methods.
[8] Yuchen Zheng, Brian Kenji Iwana, Muhammad Imran Malik, Sheraz Ahmed, Wataru
Ohyama, Seiichi Uchida Learning the MicroDeformation by Max-Pooling for Offline
Signature Verification October 2021. This research highlights the significance of micro
deformations in distinguishing genuine signatures from forgeries and showcases the
potential of CNNs, particularly through max-pooling operations, to capture these subtle
features, thereby improving the accuracy of offline signature verification systems.
The paper titled "Learning the Micro Deformations by Max-Pooling for Offline Signature
Verification," authored by Yuchen Zheng, Brian Kenji Iwana, Muhammad Imran Malik,
Sheraz Ahmed, Wataru Ohyama, and Seiichi Uchida, was published in Pattern Recognition,
Volume 118, in October 2021.
[9] Chandra Sekhar Vorungunti, Viswanath Pulabaigari, Rama Krishna Sai Subrahmanyam,
Gorthi, Prerana Mukherjee OSVFuseNet: Online Signature Verification by Feature Fusion
and Depth Wise . In this manuscript, owing to large intra-individual variability, a novel
method for OSV based on an interval symbolic representation and a fuzzy similarity
measure grounded on writer specific parameter selection is proposed. The two parameters,
namely, writer specific acceptance threshold and optimal feature set to be used for
authenticating the writer are selected based on minimum equal error rate (EER) attained
during parameter fixation phase using the training signature samples.
[10] Antonio Parziale, Moises Diaz, Miguel A. Ferrer, Angelo Marcelli “SM-DTW This
research underscores the importance of stability regions in the context of signature
verification. By integrating these regions into the DTW framework, the SM-DTW algorithm
offers a refined approach that emphasizes the most consistent parts of a signature, thereby
improving the accuracy of The findings indicate that SM-DTW enhances the performance of

Dept. of ECE, ATMECE, Mysuru Page 4

Machine Learning Based Signature Verification Process Using OCR And Line Sweep Techniques 2024-25
the baseline DTW systemThe study introduces the concept of "stability regions" in
handwritten signatures—segments that remain consistent across multiple instances of a
person's signature. The authors propose that these stable segments are crucial for
distinguishing genuine signatures from forgeries.
[11] Xia Xingua, Song Xiaoyu, Luan Fangun, Zheng Jungang, Chen Zhili, Ma Xiaofu
Discriminative Feature Selection for Online Signature Verification February 2018. This
research highlights the importance of selecting features that are both consistent and
discriminative for enhancing online signature verification systems. By integrating advanced
alignment techniques and a refined DTW matching process, the study presents a robust
approach to accurately verify signatures, thereby contributing to the development of more
reliable biometric authentication system. The proposed methods were evaluated using the MCYT
and SVC2004 Task2 public datasets. The experiments demonstrated that the combination of
consistent and discriminative feature selection, along with the modified DTW approach, led to
improved verification performance compared to existing methods

Dept. of ECE, ATMECE, Mysuru Page 5

 Machine Learning Based Signature Verification Process Using OCR And Line Sweep Techniques 2024-25

5. METHODOLOGY AND BLOCK DIAGRAM

System architecture is the conceptual design that defines the structure and behavior of a
system. An architecture description is a formal description of a system, organized in a way
that supports reasoning about the structural properties of the system. It defines the system
components or building blocks and provides a plan from which products can be procured, and
systems developed, that will work together to implement the overall system.

Cheque Images

OCR Extraction with Signatures

\

Apply Line Sweeping

Images with Signatures Train using CNN Model

CNN

Classify into
Forgery or Original Signature

Measure Accuracy

Performance with Test

Dept. of ECE, ATMECE, Mysuru Page 7

Machine Learning Based Signature Verification Process Using OCR And Line Sweep Techniques 2024-25

6. Expected Outcomes
1. Accurate Verification: High accuracy in verifying the authenticity of signatures, reducing
the risk of false positives and false negatives.

2. Effective Forgery Detection: Reliable detection of forged signatures, preventing

fraudulent activities and protecting sensitive information.

3. Improved Security: Enhanced security measures for various industries, including

banking, healthcare, and government, safeguarding against identity theft and fraud.

4. Increased Efficiency: Automation of the verification process, reducing manual labor and
minimizing the time required for verification.

5. Reduced Costs: Cost savings resulting from reduced manual labor, decreased risk of fraud,
and improved efficiency.

6. Enhanced User Experience: Streamlined verification processes, providing a seamless

experience for users.

7. Compliance with Regulations: Adherence to regulatory requirements, ensuring the

integrity and authenticity of signatures in various sectors.

8. Continuous Improvement: Ongoing refinement of the verification process through

machine learning and data analysis, ensuring the system remains effective and efficient.

9. Protection of Sensitive Information: Safeguarding of sensitive information and

documents, preventing unauthorized access and ensuring the integrity of signature-based
authentication.

Dept. of ECE, ATMECE, Mysuru Page 8

Machine Learning Based Signature Verification Process Using OCR And Line Sweep Techniques 2024-25

7. References

[1] Sounak Deya, Anjan Dutta, J. Ignacio Toledo, Suman [Link], Josep Llados´a,
Umapada Pal SigNet: Convolutional Siamese Network for Writer Independent Offline
Signature Verification” Issue 30, September,2017.
[2] Julita A., Fauziyah S., Azlina O., Mardiana B., Hazura H., Zachariah [Link]
Signature Verification System Issued April 2009.
[3] Shashi Kumar D R, K B Raja, R. K Chhotaray, Sabyasachi Pattanaik Off-line Signature
Verification Based on Fusion of Grid and Global Features Using Neural Networks Vol. 2(12),
2010.
[4] Poonam Chaudhary and Vijay Kumar Singh Online Signature Verification: A Review,
Volume3, Issue-2, Feb 2016.
[5] Liyang Xie, Zhongcheng Wu, Xian Zhang, Yong Li, Xinkuang Wang Writer-Independent
Online Signature Verification based on 2-D Representation of Time Series Data using Triplet
Supervised Network June 2022.
[6] S. Priya, [Link], K. Somesh, A. Hruday Kumar Signature Verification System
using Different Algorithms,Volume-8, Issue6S3, April 2019.
[7] Tushara D, Shridevi Raddy, Shreya KM, Spoorthy Y, Signature Verification System
using Neural Networks, NCCDS.- 2021 Conference Proceedings.
[8] Yuchen Zheng, Brian Kenji Iwana, Muhammad Imran Malik, Sheraz Ahmed, Wataru
Ohyama, Seiichi Uchida Learning the MicroDeformation by Max-Pooling for Offline Signature
Verification October 2021.
[9] Chandra Sekhar Vorungunti, Viswanath Pulabaigari, Rama Krishna Sai
Subrahmanyam, Gorthi, Prerana Mukherjee OSVFuseNet: Online Signature Verification by
Feature Fusion and DepthWise Separable Convolution Based Deep Learning October 2020.
[10] Antonio Parziale, Moises Diaz, Miguel A. Ferrer, Angelo Marcelli SM-DTW: Stability
Modulated Dynamic Time Warping for Signature Verification April 2019.
[11] Xia Xingua, Song Xiaoyu, Luan Fangun, Zheng Jungang, Chen Zhili, Ma Xiaofu
Discriminative Feature Selection for Online Signature Verification February 2018.

Dept. of ECE, ATMECE, Mysuru Page 9