Local Directional Strength Pattern for effective Offline Handwritten Signature Verification

Abstract

Offline handwritten signature verification is one the oldest and widespread biometric identification tools. In various daily life applications, the handwritten signature is used for documents approval and identity verification. Systems that automatically achieve this task are mainly composed of feature generation and verification modules. The more features are robust the best the verification score is. Presently, we introduce the use of the Local Directional Strength Patterns (LDSP) for handwritten signature characterization. This descriptor is associated with SVM classifier to perform the signature verification. Experiments conducted on two public datasets reveal the effectiveness of the proposed descriptor which outperforms several state-of-the-art techniques.