International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395 -0056
Volume: 04 Issue: 04 | Apr -2017
p-ISSN: 2395-0072
www.irjet.net
Comparative Study of Various Signature Verification Algorithms Namita Gupta1, K. N. Saravanan 2 1Student,
Department of Computer Science, Christ University, Bengaluru, Karnataka, India Department of Computer Science, Christ University, Bengaluru, Karnataka, India
2Professor,
---------------------------------------------------------------------***--------------------------------------------------------------------2. STRUCTURE OF AUTOMATIC SIGNATURE VERIFICATION or name which can distinguish his/her identity from others
Abstract - Signature is a way of writing one’s own initials
and can be used for authentication purposes. It is the best way of authenticating a person since each individual possesses a different style of writing. Two signatures can differ from each other in terms of the pressure exerted while signing, the shape of loops, the speed of writing, and various other features. Several algorithms have been written to verify these signatures based on different sets of features extracted as well as different classifiers used for classification. This paper compares some of such signature verification algorithms which focused on different sets of features and used different classification algorithms. Key Words: Automatic Signature Verification (ASV), Minimum Distance Classification, K-Nearest Neighbor (KNN), Support Vector Machine (SVM)
Scanned Image
Preprocessing
Training, Classification & Recognition
Feature Extraction
Verified Signature
1. INTRODUCTION
Fig -1: Structure of ASV
Signature is one of the behavioural biometrics which is concerned with the identification of a person based on the pattern of behaviour of his/her characteristics. According to the history of biometrics, before the emergence of signatures the most familiar way for human identification was face recognition. After an increase in population, identifying a person became a challenging task, so they introduced the concept of fingerprints, palm, footprints and signature recognition. Signature verification is categorized into two classes according to how the data is acquired, namely – offline and online. Online method is also known as Dynamic method since it captures the signature at the same moment of writing along with some extra details such as movements of pen and the pressure exerted on the paper. This method needs a special setup to record the signature. The Offline approach is also known as Static approach. In this method the signature is captured on a sheet of paper, and is scanned using scanner to translate it into digital format. As per the history of biometrics, the first automated signature recognition system was first developed by North America Aviation in 1965. Since then many researchers have experimented and tested different ways to enhance the efficiency and applicability of signature verification system.
I. Signature Acquisition: The signature is converted into a digital format using an optical scanner. It can be viewed as an image of M×N pixels.
© 2017, IRJET
|
Impact Factor value: 5.181
|
II. Preprocessing: This step is done to make the acquired signature ready for the next step which is feature extraction. Various researchers use variety of methods within preprocessing step that is suitable for their further feature extraction module. But the steps that are commonly followed by majority of researchers are: i. Noise Reduction – to remove the noise that comes while scanning. ii. Resizing – to adjust the size according to required template. iii. Binarization – to bring all images (colored/grayscale) under one category i.e. black and white. iv. Thinning – to take out the thickness differences of pen. III. Feature Extraction: This step plays a major role in determining the accuracy and efficiency of any ASV system. Features can broadly be divided into three categories – Local features, Global features and Geometric features. Some researchers have experimented with only one kind of features while others have tried and tested several combinations of different kinds of features.
ISO 9001:2008 Certified Journal
|
Page 2146