International Journal of Research in Engineering, Science and Management 265

Volume-1, Issue-9, September-2018

www.ijresm.com | ISSN (Online): 2581-5782

Bank Cheque Signature Verification System

Vaibhav Tambade1, Priyanka Varma2, Aditya Sonawale3
1,2,3
Student, Department of Computer Engineering, MGM CET, Navi-Mumbai, India

Abstract—Our Signature Verification System creates a mentioned both in words and figures clearly. The amount
repository of signatures and policy mandates that can be written in words should tally with the amount written in figures.
dynamically accessed for the clearing process. This solution The drawer should sign the cheque properly. The signature
supports defining of accounts, mapping of signatories, scanning of
mandates, and cropping of signature images. Further, it supports given on the cheque should tally with the signature given on the
the definition of complex logics based on different amount bands signature specification car. Image show the signature on
and signatory categories. Using this solution, you can ensure auto- cheque.
validation of these rules based on the amount and signatories
present on the cheque / mandate form. This ensures process
compliance, as well as reduced risk in clearing different payment
instructions. Thus, you can easily define financial rules, as well as
rules for non-financial transactions into the system. Signatures are
imperative biometric attributes of humans that have long been
used for authorization purposes. Most organizations primarily
focus on the visual appearance of the signature for verification
purposes. Many documents, such as forms, contracts, bank
cheques, and credit card transactions require the signing of a
signature. Therefore, it is of upmost importance to be able to Fig. 1. Sample cheque
recognize signatures accurately, effortlessly, and in a timely
manner. In this work, an artificial neural network based on the Handwritten signatures can be verified using online or offline
well-known Back-propagation algorithm is used for recognition schemes. Online signature can be captured using electronic
and verification. To test the performance of the system, the False devices like writing pad or stylus attached to a computer.
Reject Rate, the False Accept Rate, and the Equal Error Rate
Offline signature will have only the digitized signature from
(EER) are calculated. The system was tested with 400 test
signature samples, which include genuine and forged signatures of which required features can be extracted. Since this paper deals
twenty individuals. The aim of this work is to limit the computer with signatures on bank cheques, offline signature verification
singularity in deciding whether the signature is forged or not, and scheme is chosen. A signature which is suspicious is called a
to allow the signature verification personnel to participate in the forged signature.
deciding process through adding a label which indicates the
amount of similarity between the signature which we want to Signature of a banker or a customer can be broadly divided
recognize and the original signature. This approach allows judging into two categories:
the signature accuracy, and achieving more effective results. 1. Initials
This is the short form of signature wherein the signatory put
Index Terms—Feature extraction, Handwriting recognition,
Training, Artificial neural networks, Neurons, Classification son his signature in a very short form. These initials are not
algorithms accepted for wide range of transactions by bankers. These are
used for intra official transactions and on cop-ies of
I. INTRODUCTION instruments. Normally, institutions do not retain these
Signature is the widely used and accepted form of signatures as a specimen.
authentication and it is used even before the usage of
computers. Behavioural biometric identifiers, on the other 2. Full signature
hand, are related to the pattern of behaviour of a person, such These are the signatures widely used in banking transactions
as typing rhythm, gait, and voice. Handwritten signatures are and are retained as a specimen signature by the banker. Bank
used in almost all documents where authentication is required. officials verify the signatures appearing in the instruments
It has low conflict percentage. The number of bank cheque against these specimen signatures. In the cases of corporate or
fraud cases are rising and there are number of modern institutional customers, company seal or stamp is also retained
techniques followed by the fraudulent to counterfeit. Signatures as a part of signature.
are not easy to verify like that of characters. Verifying
signatures in bank cheques is a challenging opportunity in II. APPROACH
image processing. The amount of the cheque should be We approach the problem in two steps. Initially a set of
 International Journal of Research in Engineering, Science and Management 266
Volume-1, Issue-9, September-2018
www.ijresm.com | ISSN (Online): 2581-5782

signatures are obtained from the subject and fed to the system. C. Steps Involved In Signature Verification
These signatures and pre-processed Then the pre-processed 1) Data acquisition
images are used to extract relevant geometric parameters that Data acquisition actually means giving data to the system. In
can distinguish signatures of different persons. These are used this system the data is provided to the system in the form of a
to train the system. The mean value of these features is scanned image. The signature is scanned and stored as a digital
obtained. In the next step the scanned signature image to be image.
verified is fed to the system. It is pre-processed to be suitable
for extracting features. It is fed to the system and various
features are extracted from them. These values are then
compared with the mean features that were used to train the
system. The Euclidian distance is calculated and a suitable
threshold per user is chosen. Depending on whether the input
signature satisfies the threshold condition the system either
accepts or rejects the signature.
Section 3 deals types of signatures and Section 4 explains
signatures verification model and steps to involve that are
extracted followed by the verification procedure by Neural
Network in Section 5. Features use of signature verification
system listed in Section 6. The conclusion and references are
follows in last section.
Fig. 2. Signature model verification
III. TYPES OF VERIFICATION
A. Online/ Dynamic Signature 2) Pre-processing
Data preprocessing describes any type of processing
Online or Dynamic signature publish in early in 1990’s. It
performed on raw data to prepare it for another processing
can be consider only Uses shape, speed and pressure. In online
procedure. There are 4 steps to use to pre-processing steps and
signature system needs special digital surface, pads and pen etc.
steps of pre-processing are followed.
It’s required some numeric data and small amount of storage. It
a) Image Conversion: The first step is to convert the obtained
can use speed, pressure, angle of pen to further exploit
RGB image into gray scale image. This is done to reduce
individuality. In online signature it’s hard to easier to forge.
the complexity and the execution time of the system. It is
Accuracy of signature verification on offline method is around
easy for the system to work with grayscale images rather
99%.
than the RGB images.
B. Off-Line/Static Signature b) Image Cropping: Then a cropping algorithm is applied to
Offline or Static signature publish in early in 1970’s. It can the grayscale image. Cropping is done to separate out
be consider only image signature. In offline signature system precisely the actual region of interest from the complete
no need to use special hardware, ubiquitous etc. It’s required image. This removes the extra pixels from the image thus
large amount of storage. It cannot be trace speed, style, pressure reducing the processing time.
etc. In offline signature it’s easy to easier to forge. Accuracy of
signature verification on offline method is around 95%.

IV. ALGORITHM MODELS AND STEPS

A. Algorithm of Signature Verification
Step 1: Hand written signature is scan using camera or scanner.
Step 2: Signature is pre-processing and converted into binary or
grey scale as per requirement, Removing noise from signatures
and finally normalize the signatures.
Step 3: Special domain feature as high intensity variation points Fig. 3. Basic signature verification system
and cross over points extract from test signatures.
Step 4: Compared features with help of graph matching c) Image Filtering: After obtaining the cropped image the
methods as a classifier. noise that entered while scanning or via any other source
B. Model of Verification has to be removed. For this a Gaussian filter along with the
unsharp filter is used. The Gaussian filters smooth the
The Fig. 2. Shows the signature verification model
image removing the noise and the unsharp filter removes
 International Journal of Research in Engineering, Science and Management 267
Volume-1, Issue-9, September-2018
www.ijresm.com | ISSN (Online): 2581-5782

the blur created by the Gaussian filter. of data, and makes an adjustment. It keeps doing this, repeating
d) Thresholding: This filtered image is then threshold to cycles of data and adjusting the trigger levels, and hopefully
convert it into a binary form. Finally the boundary of the getting closer to a repeatable pattern.
signature is extracted from this binary image using the
canny edge detector. V. FEATURES USED
3) Feature extraction 1. Total time
In pattern recognition and in image processing, feature 2. Signature path length
extraction is a special form of dimensionality reduction. When 3. Path tangent angles
the input data to an algorithm is too large to be processed and it 4. Signature velocity
is suspected to be notoriously redundant then the input data is 5. Signature accelerations
transformed into a reduced representation set of features. 6. Pen-up times & durations
Transforming the input data into the set of features is called 7. [Crane83] proposed 44 while [Parks85] proposed 90
feature extraction. features [Lee96] used 15 static & 34 dynamic
The main function of this step is to generate features which a. None related to shape
can be used as comparison measurements. Since the issue of b. 1% FRR, 20% FAR on timed forgeries
signature verification is a highly sensitive process, more than
one feature has to be generated in order to enhance the accuracy VI. CONCLUSION
of the result.
4) Classification  The new system should be an on-line system
Image classification analyzes the numerical properties of  Shape is an integral part of signature verification, it is
various image features and organizes data into categories. a metric that is most easily imitated by a forger
Classification algorithms typically employ two phases of  Both global & local features should be used
processing: training and testing as show in signature  Different methods have been tried with varying
verification model. In the initial training phase, characteristic results, About 99% at the best
properties of typical image features are isolated and, based on  Great deal of speed improvement to be done
these, a unique description of each classification category, i.e.  Signature segmentation into individual strokes needs
training class, is created. In the subsequent testing phase, these attention
feature-space partitions are used to classify image features.  Multi-expert system to integrate different methods
5) Signature verification using neural network  Analysis on proper setting of thresholds & use of user-
In a neural network, you have a set of inputs, which results specific thresholds
in an output. Between the input and the output are multiple  Sensors have developed to a fair point of saturation
layers of artificial neurons. These neurons take the input, and a  Study on multi-lingual signatures is unfocused
weight inside them works out if it triggers or not. This is then
passed to the next layer which does likewise. The more complex REFERENCES
your problem the more layer and artificial neurons you might [1] S. Chen, and S. Srihari, “Use of Exterior Contour and Shape Features in
need Off-line Signature Verification”, 8th International Conference on
Document Analysis and Recognition (ICDAR ’05), 2005, pp. 1280-1284
[2] Fernando Alonso-Fernandez, Julian Fierrez, Almudena Gilperez, Javier
Galbally, Javier Ortega-Garcia, Robustness of Signature Verification
Systems to Imitators With Increasing skills, IEEE 2009,pp. 728-732.
[3] Vibha Pandey, Sanjivani Shantaiya. “ Signature Verification Using
Morphological Features Based on Artificial Neural Network”, 2012 pp.
288-292
[4] Bradley Schafer, Serestina Viriri, An Off-Line Signature Verification
Fig. 4. Neural network diagram System, IEEE International Conference on Signal and Image Processing
Applications, 2009, pp. 95-100
The whole thing works as a multilayered transformational [5] Daramola Samuel and Prof.Ibiyemi Samuel, Novel feature extraction
technique for off-line signature verification system, IJEST, Vol ED-2(7),
matrix. What happens is you feed through training data, the 2010.
system compares the output it gets with the expected output. It [6] Imran Hussain, Vikash Shrivastava, Vivek Kr. Shrivastava, Review On
then applies feedback which adjusts the weighted trigger levels Offline Signature Verification Methods Based On Artificial Intelligence
Technique, IJOART, Volume 2, Issue 5, May-2013.
of some of the neurons. Then it tries again with another piece