A Neural Network Approach to Signature Verification with Mathematical Moments

The present work develops a method for identity authentication and verification of static signatures stored in the database employing artificial neural network. The present method uses mathematical moments for feature extraction such as mean, variance, skewness and kurtosis. First of all, the method suggests to scan the signature images, then after a sequence of preprocessing steps the resulting images are subjected to feature extraction, however, at present already existing standard databases have been used. Subsequently, the system is trained from the genuine signature of individuals, and then an ANN is used to classify the signature images. The suggested method's effectiveness has been demonstrated by comparison with the four current methodologies and experimental findings.