Hand Written Signature Verification based on Geometric and Grid Features

Abstract

The fact that the signature is widely used as a means of personal verification
emphasizes the need for an automatic verification system. Verification can be
performed either Offline or Online based on the application. Offline systems work on
the scanned image of a signature. In this paper an Offline Verification of handwritten
signatures which use set of simple shape based geometric features. The features used
are Mean, Occupancy Ratio, Normalized Area, Center of Gravity, Pixel density,
Standard Deviation and the Density Ratio. Before extracting the features,
preprocessing of a scanned image is necessary to isolate the signature part and to
remove any spurious noise present. Features Extracted for whole signature first, then
extracted for every part after dividing the signature into four sections. For verification,
statistical verification techniques are used (Euclidean Distance, Hellinger Distance
and Square Chord Distance). The system is trained on three datasets of signatures.
The first and the second datasets have English signatures while the third one is
collected from people; it contains Arabic and English signatures. The system has been
tested on every dataset. The experimental results show that the Euclidean Distance
has the average accuracy of 94.42, the Hellinger Distance has the average accuracy of
95.27 and the Square Chord Distance has the average accuracy of 93.14. That result
for whole the image and the following average accuracy for image using grid the
Euclidean Distance has the average accuracy of 93.54, the Hellinger Distance has the
average accuracy of 95.87, and the Square Chord Distance has the average accuracy
of 95.93.