Signature verification using wavelet, gabor, convolution neural network (CNN)

Abstract

Signature is one of the most popular personal attributes for authentication. It is basic, shabby and adequate to individuals, official associations and courts. However, Signature Verification Systems experience the ill effects of factors that influence the execution, for example, composing instrument, paper, and physical state of the author. On the other head, for this system to be accurate, a considerably large number of samples is required. This report focuses on offline signature verification to determine whether the signature is genuine or forgery. In our research we use Convolution Neural Network (CNN) to classify handwritten digits in two types of datasets: the MNIST database, and UTSIG database. To obtain better accuracy, we propose to preprocess the data in the wavelet domain and in the Gabor filter comparing outputs of both CNN and classify one results which is 99% accuracy.