Quasi-straightness based features for off-line verification of signatures

Abstract

Person identification from their signatures or verifying the genuineness of official documents like bank cheques, certificates, contract forms, bonds etc. still remains a challenging task when accuracy and computation time are concerned. In this paper, a novel set of features based on the distribution of the quasi-straight line segments has been presented for off-line signature verification. For the detection of the set of quasi-straight line segments, defining the signature boundary, 8-N chain codes are used. Twelve different classes of quasi-straight line segments are obtained depending upon the orientations of the line segments. Subsequently, the feature set is obtained from those twelve classes. Support Vector Machine (SVM) classifier has been used by us for verification. Results on standard signature databases like CEDAR (Center of Excellence for Document Analysis and Recognition) database and GPDS-100 (Grupo de Procesado Digital de la Senal) are shown to adjudge the fitness of the proposed method.