Offline Signature Verification Based on

Geometric Feature Extraction using -Artificial Neural Network

Guided by:Ms. Lima SebastianAssistant ProfessorCSE Dept. AISAT

Submitted by:Cen PaulS7 CSE

13027323

2

Overview

• Introduction

• Types of Signature Forgeries

• Workflow of the System

• Experimental Results

• Conclusion

• References

Introduction• For centuries, handwritten signatures have been an integral part of

validating business transactions , contracts and agreements.

• Among the different forms of biometric recognition systems such as fingerprint, iris, face, voice, palm etc. , Handwritten signature is the most widely used.

• In the era of advanced technology, security is the vital issue to avoid fakes and forgeries.

• The signature verification is classified into online systems and offline systems.

• The signature verification systems help to discriminate between the original and fake signatures.

3

Types of Signature Forgeries

1. Random Forgery

2. Simple Forgery

3. Skilled Forgery

4

Block Diagram

Offline Signature

Verification System 5

5

Workflow of Signature Verification

1. Data Acquisition2. Preprocessing 3. Feature Extraction4. Verification/Comparison

Input Data

Data Preprocessi

ngFeature

Extraction

Comparison/

Verification

Forged or Genuine?

6

1. Data Acquisition

• Signatures from individual person are taken on paper and then scanned with scanner.

• The database contains data from individuals, including genuine signatures and forgeries signatures. • Signatures will be stored as images.

7

2. Preprocessing [1/4]

• Preprocessing is done for noise removal.

• Preprocessing stage includes :i. RGB to gray scale conversion

ii. Binarization

iii. Cropping

8

Preprocessing [2/4]

i. RGB to gray scale conversion RGB image of scanned signature is converted into gray scale intensity signature image to eliminate the hue and saturation information while retaining the luminance.

RGB to Gray-scale Conversion

9

Preprocessing [3/4]

ii. Binarization A gray scale signature image is converted into binary image to count the number of black pixels which make feature extraction simpler

Binarization

10

Preprocessing [4/4]

iii.Cropping Cropping the binary image using the boundary-values returned by bounding box calculation method. This reduces the area of signature to be used for further processing.

Cropping

11

3. Feature Extraction [1/4]

• To extract the feature of signature image using six global features.

• The extracted features of a signature image are based on geometrical features like size and shape.

• Features used in this system :i. Area

ii. Centroid

iii. Standard Deviation

iv. Skewness

v. Kurtosis

vi. Even-Pixels 12

Feature Extraction [2/4]

i. Area

Total number of black pixels present in the binary image.

ii. Centroid

It denotes to the center point of vertical and horizontal of the signature.

13

Feature Extraction [3/4]

iii.Standard Deviation

It measures the amount of variation or dispersion on a set of mean data values. If deviation is closed to the mean data value then the variation is less otherwise spread over a wider range of values.

iv.Skewness

It measure the asymmetricity of the probability distribution of a real valued random variable having positive, negative or may have undefined value.

14

Feature Extraction [4/4]v. Kurtosis

Higher value of kurtosis distribution indicates thicker tails, longer and a sharper peak whereas lower value denotes shorter, thinner tails.

In Image processing kurtosis values are illustrated in combination with resolution and noise measurement. In which high kurtosis values gives low noise and low resolution.

vi. Even Pixels

The positions in the image matrix. Even position refers those matrix positions for which both the coordinates are even .

15

4. Verification

• The geometric feature are extracted and organised as an input array to the back propagation network.

• The selected feature vectors are directed as input to the neural network.

• The trained neural network is used to verify the signature as either genuine or forged.

• If the signature is match then it shows genuine otherwise forgery

16

Experimental Results [1/4]

A. Database• The signature database is collected from MCYT-75 offline signature corpus

database.

• 15 genuine and 15 forgery signature samples are given for each of 75 users in database.

• The forgery signature in the database is the mixture of random, simple and skilled forgeries.

17

Experimental Results [2/4]

B. Performance Measures• The performance measure of the signature verification is measured in terms

of false rejection rate (FRR) and false acceptance rate (FAR).

• False acceptance occurs when forgeries signatures are accepted as genuine while in case of false rejection genuine signature are accepted as forgery.

18

Experimental Results [3/4]

• Accuracy of the system is the mean between percentage of genuine signatures verified as genuine and percentage of forgery signature is verified as forgery.

19

Experimental Results [4/4]

C. Results• Experiments were conducted on 18 different users. Each having 15 genuine

and 15 forgery signatures.

• Total number of 540 signature is taken each having dimension of 850 360 pixels.

20

CONCLUSION

• Explored the application of geometric based feature extraction on offline signature verification.

• The performance of the proposed method is examined using Back propagation learning technique.

• Total accuracy obtained using the proposed method comes out to be above 89.24% .

21

References

• Subhash Chandra , Sushila Maheskar . Offline signature verification based on geometric feature extraction using artificial neural network .3rd Int’l Conf. on Recent Advances in Information Technology RAIT-2016 .

• Mujahed Jarad, Nijad Al-Najdawi, and Sara Tedmori. Offline handwritten signature verification system using a supervised neural network approach. In Computer Science and Information Technology (CSIT), 2014 6th International Conference on, pages 189–195. IEEE, 2014.

• R. Dubey and D. K. Agrawal, “Comparative analysis of off-line signature recognition,” 2012 International Conference on Communication, Information & Computing Technology (ICCICT), pp. 1–6, Oct. 2012.

22

THANKYOU !!

23