Elimination of skilled forgeries in off-line systems: a breakthrough

Presents a novel approach to automatic verification of signatures in off-line systems. This new approach is able to eliminate skilled forgeries at a very low false alarm rate. A zero false alarm rate is also possible at a very high correct rejection, and the average error rate has approached 2% for skilled forgeries. The new approach is based on matching the suspect signature with reference knowledge (RN) represented by a reference image (RI) constructed from all the available training samples. A set of newly developed global and local features, extracted using a new matching technique called Ammar matching technique (AMT), are used