Offline Signature Verification using Clustering Technique

Signature verification is one of the most basic and heavily used biometric which finds its application in many fields in the day to day life. However, it is also very complex and difficult to correctly classify a signature as genuine or forged because of the discrepancies associated with a signature and due to the skill and precision with which a forgery of the genuine signature is done. The proposed method is based on extraction and analysis of the features of the handwritten signature from the scanned images using Agglomerative Hierarchical Clustering technique. After pre-processing of the images and formation of clusters, based upon the error threshold, which is calculated using the intra-cluster distances, the signatures are classified as either genuine or forged. The proposed method is reliable, quick and does not require large datasets since it is based on an unsupervised approach and has shown promising results while dealing with skilled forgeries.