Offline handwritten signature verification using sift features

Abstract

In this research we evaluate the use of SIFT features in offline handwritten signature verification. For each known writer we take a sample of three genuine signatures and extract their SIFT descriptors. We calculate the intra-class Euclidean distances (measure of variability within the same author) among SIFT descriptors of this known signatures. The keypoints Euclidean distances, the image distances and the intra class thresholds are stored as templates. We evaluate use of various intra-class distance thresholds like the maximum, average, minimum and range. For each signature claimed to be of the known writers, we extract its SIFT descriptors and calculate the inter-class distances, that is the Euclidean distances between each of its SIFT descriptors and those of the known template and image distances between the test signature and members of the the genuine sample. The intra-class threshold is compared to the inter-class threshold for the claimed signature to be considered a forgery. A database of 90 signatures consisting of a training set and a test set is used. The training set is made up of 54 genuine signatures from 18 known writers each contributing a sample of 3 signatures. The testset consists of 36 signatures, 18 genuine signature and 18 forged signature. The specificity and sensitivity of the verifier is measured and compared with the results from the analysis of human expert.