An Efficient Signature Recognition System Based on Gradient Features and Neural Network Classifier
Ouafae El Melhaoui, Soukaina Benchaou
Abstract
This paper proposes a novel offline signature recognition system (SRS) based on histogram of oriented gradients (HOG) and fuzzy min max classification (FMMC) methods. First of all, the signature image required a preprocessing stage, then the Histogram of Oriented Gradients features are adopted to extract features from the training images. It consists of dividing the image into adjacent cells, for each cell histogram of oriented gradients characteristics are calculated. This technique has been compared with two popular statistical methods such as Loci characteristics and profile projection (PP). The classification is performed using FMMC and it is compared with K nearest neighbors method (KNN). The presented approach achieved a recognition rate of 96% using a diverse signature database.