Off-line signature verification based on principal component analysis and multi-layer perceptrons

Abstract

The off-line signature verification approach based on Principal Component Analysis (PCA)  and Muti-Layer Perceptrons (MLP)  is proposed in this work. The proposed approach involves three major phases. In the first phase, the signature image is subjected to preprocessing, such as binarization, noise elimination, skew correction followed by normalising the image by a series of thinning and dilating with a structuring element of size 3X3. The principal component analysis is employed on the preprocessed signature samples and the features extracted are termed as eigen-sign feature vectors, in second phase. The multilayer perceptrons, a neural network based approach is trained with eigen-sign feature vector and subsequently  used for verification of signature samples. Extensive experimentation has been conducted on the publicly available signature datasets namely, CEDAR, GPDS-100 and MUKOS, a regional language dataset. The state-of-art off-line signature verification  methods are considered for comparative study and objective analysis through experimental results is provided to justify the accuracy of the proposed approach.