e-ISSN: 2582-5208

International Research Journal of Modernization in Engineering Technology and Science

( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:03/Issue:07/July-2021 Impact Factor- 5.354 www.irjmets.com

A PROJECT BASED APPROACH TOWARDS OFF-LINE SIGNATURE

AUTHENTICATION
Hemant Singh Rawat*1, Ayush Sharma*2, Manan Gupta*3,
Tarun Kumar*4, Aditya Tandon*5
*1,2,3,4Student, Computer Science Engineering, Krishna Engineering College, Ghaziabad, U.P, India.
*5Professor, Department Of Computer Science Engineering, Krishna Engineering College,
Ghaziabad, U.P, India.
ABSTRACT
In this paper, we present a system to verify whether the given signature belong to a valid person or not (off-line
signature authentication system) using machine learning logistic regression algorithm by taking out some
global feature from the signature and put them on the model.
Keywords: Logistic Regression, Feature Extraction, Handwritten Signature, Pre-Processing.
I. INTRODUCTION
A signature (also called sign) of any person is a handwritten representation of someone's nickname, name, or
other signs that a person writes on paper as a identity proof. The person whose signature is on any paper is
called a signer. A signature work on any document indicates the work as readily recognise its creator. There is a
minor confusion between a sign and an autograph, which is an artistic sign. This confusion among the people
has both an autograph and signature and some people keep their signatures private because of security reasons
while fully used their autograph.
A. Function of Signature
A signature can has many different functions but one of the most important functions is to permanently attach
to a document of a person uniquely identifiable, undeniable identity as physical witness of that person witness
and proof of the content of all the things in the document. For ex, the part of a signature in any contracts isn’t
just to provide the proof of identity to both the parties, but also to provide a proof of consideration and
agreement. On legal documents, an illiterate person can make a ‘mark’ or people can put a thumb expression on
documents as an alternative of a signature. In most of the countries, signatures are noticed and observed in the
appearance of a notary officer to carry additional legal formality.

Fig 1: Photos of Signatures

B. Online Signature
In online signature, the content on the document is mixed up with the private key of the user, this process is
also called hashing, then the other user check with the help of public key of the sign user whether the document
is sign by the user or not. Users can set one or more lines of custom text known as a signature block to be
automatically appended to their messages. The content usually includes hashed data public key and every
change is done in hexadecimal coding. Here, both the parties don’t have to be together but all the security stuff

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[1361]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:03/Issue:07/July-2021 Impact Factor- 5.354 www.irjmets.com
is handled by the third party, through which the signature has been done. But the third party need cash to do it,
they don’t provide free service. This type of signature is also called digital signature. And it is very hard, almost
impossible, to alter the data in the signed document.
We introduced an Off-line handwritten signature system and it has two parts:
(i) Training the signatures
(ii) Verification of given signature.
The block diagram of the off-line handwritten signature system is given in below figure Fig 2.

Fig 2: Block diagram of the system

II. PRE-PROCESSING
The pre-processing process is included in both testing and training of the signature. The scanned signatures are
converted into grey scale mode. This is because to make signatures easy and ready to apply the feature
extraction method. The pre-processing process has four steps an they are given below.
A. Background Elimination
The area of the signature must be cropped, it is necessary to extract the features. A threshold needs to be set to
capture signature density wise from the background. After setting the threshold, the pixels of the signature
would be categorize either “1” or “0” based on the density of the pixel.
B. Noise Reduction
To eliminate the noise, we need to check all the adjacent neighbour of a pixel i.e., 8-neighbours(up, down, left
and right). We check it on the grey scale image. If a pixel has more black pixel than white then the pixel is black
otherwise it consider a white.
C. Width Normalisation
The size or dimensions of different image may be different. So to overcome this problem, we readjust the image
width and height in order to have proper calculation. But during changing the height and width, we must have
same height to width ratio. At the end of width normalization width dimension is adjusted to 100.
D. Thinning
In this process, we want to decrease the thickness of the ink on the signature. The thickness can be change very
often while using pen. We need a constant thickness of every line in the signature for proper calculation.
III. FEATURE EXTRACTION
In this phase, extracted features are the inputs of training phase.
A. Signature Area
The area of the signature is the total number of pixel in the image of signature. This feature provides
information about the signature density.
B. Signature height to width ratio
It is the ratio of the height of the signature to the width of the signature. A person signature can have different
height and width in different signature but the ratio of all the signature are approximately equal.
C. Maximum Horizontal histogram and Vertical histogram
www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science
[1362]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:03/Issue:07/July-2021 Impact Factor- 5.354 www.irjmets.com
For each row, which has highest ‘1’ values than the rest of the row is taken to be maximum horizontal
histogram. And for each column, which has highest ‘1’ values than the rest of us is taken to be maximum vertical
histogram.
D. Vertical and Horizontal centre of the signature
The centre of the row of an image is called horizontal centre of the signature and the centre of the column of an
image is called vertical centre of the signature
E. Local maxima numbers of the signature
The number of local maxima of the vertical and horizontal histogram is calculated.
F. Edge point numbers of the signature
Edge point is the pixel which has only one neighbour, which belongs to the signature, in 8-neighbor.
IV. SIGNATURE DATABASE
For training and testing of the Signature recognition and authentication system 100 signature are used (50 are
real and 50 are forged).The signature are taken from the 10 persons.
For training the system 4 out of five real and forged signatures of a person is used and the last one is used for
testing.
So there are total 50 real signature of 10 persons and 50 forges signature of 10 persons.
V. RESULT
A. Training phase
In the Training phase, the model extract the global features of the first 4 image of a person from real and 4
images from the forged dataset. And use the features of these images as a training dataset for the model.
B. Testing phase
In testing phase, the last signature from the real and forge dataset is taken and classify using the coefficient, we
find in training phase, whether it is forged or real signature.
Here the model used logistic regression machine learning algorithm and train it coefficient by using gradient
descent.
So out of 100 dataset of 10 persons, 80 signatures is taken for training the model and 20 is for testing. Here is a
confusion matrix of the testing data:
N=20 Predicted NO Predicted YES
Actual NO 6 4
Actual YES 2 8
Fig.3: Confusion matrix of testing data
In this confusion matrix, total correct predictions are 14 and total wrong predictions are 6. This show the
efficiency of this system i.e., 70 %.
VI. CONCLUSION
We present an offline handwritten signature authentication technique which is based on logistic regression
machine learning algorithm. The efficiency of this model is 70% as we can see in Fig.3 . Carefully chosen
discriminating features of signatures made this system more powerful. Also we need to be careful while taking
image of the signature otherwise system efficiency may vary (decrease if not taken carefully). Also taken the
signature as clean paper as possible.
VII. REFERENCES
[1] Prashanth CR,KB Raja,KR Venugopal, LM Patnaik,”Standard Scores Correlation based Offline signature
verification system”, International Conference on advances in computing, control and
telecommunication Technologies 2009 .
[2] R. Plamondon and S.N. Srihari, "Online and Offline Handwriting Recognition: A Comprehensive Survey",
IEEE Tran. on Pattern Analysis and Machine Intelligence, vol.22 no.1, pp.63-84, Jan.2000.
www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science
[1363]
 e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science
( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:03/Issue:07/July-2021 Impact Factor- 5.354 www.irjmets.com
[3] J Edson, R. Justino, F. Bortolozzi and R. Sabourin, "An off-line signature verification using HMM for
Random,Simple and Skilled Forgeries", Sixth International Conference on Document Analysis and
Recognition, pp.1031-1034, Sept.2001. 211-222, Dec.2000.
[4] S.Srihari. K. M. Kalera. and A. XU, “Offline Signature Verification and Identification Using Distance
Statistics,” International Journal of Pattern Recognition And Artificial Intelligence ,vol. 18, no. 7, pp.
1339–1360, 2004.
[5] Ramachandra A. C ,Jyoti shrinivas Rao”Robust Offline signature verification based on global features”
IEEE International Advance Computing Conference ,2009.
[6] “An Introduction to Artificial Neural Systems” by Jacek M. Zurada, West Publishing Company 1992.
[7] OZ, C. Ercal, F. and Demir, Z. Signature Recognition and Verification with ANN.
[8] Golda, A. 2005. Principles of Training multi-layer neural network using back propagation.
[9] Ozgunduz, E., Karsligil, E., and Senturk, T. 2005.Off-line Signature Verification and Recognition by
Support Vector Machine. Paper presented at the European Signal processing Conference.
[10] Pacut, A. and Czaja, A. 2001. Recognition of Human Signatures. Neural Network, in proceedings of the
International Conference on Neural Network, IJCNN’01, vol.2, pp 1560-1564.
[11] D.Z. Lejtman. On-line Handwritten Signature Verification Using Wavelets and Back-propagation Neural
Networks. Proceedings of the Sixth International Conference on Document Analysis and Recognition,
pp 992, 2001.

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[1364]