A comparative study on signature recognition

Abstract

Contour and shape are the major point for signature recognition. There are a few approaches that have been used for shape detection, particularly to signature recognition. Combination methods of geometric features such as ratio, contour, shape or moment like Zernike moment, Moment Invariants (Hu) usually have been used to identifY the signature. One method never been used for signature recognition, Polar Fourier Transform. In this paper, a comparative study is conducted to compare three methods, Moment Invariants (Hu), Zernike Moment, and Polar Fourier Transform (PFT). Support Vector Machine (SVM) and Multilayer Perceptron (MLP) are used to classifY 20 person data set in each of which consists of 15 genuine signatures and 15 forgery signatures. The result shows that the methods using PFT and SVM achieves accuracy of 86.67%. Whilst the computational time of SVM is faster than MLP. The SVM method had an average value of 0.012 seconds for computational time.