A supervised manipuri offline signature verification system with global and local features

Handwritten signature verification is one of the significant research area where writers are verified or identified by their signatures. Handwritten signatures can be found in many official documents in day to day applications where people are fond to use their own scripts for writing the signatures. Usually, human experts look for the pattern of a signature in order to verify an authenticated document. The same expertise or even better can be adopted into an algorithm and run on a computer system where handwritten signatures could be accurately verified with minimum effort and time. As it is a behavioural biometrics trait, therefore writing style would decide the complexity of signature patterns of individual writers. Manipuri or Meithei is one of the official languages of the Indian state Manipur where large number of native people speak Manipuri language. This paper proposes a supervised learning approach for verifying individuals using their handwritten offline signatures. To accomplish this task, a set of local and global features related to the structure of the signature is extracted from offline signature. Further, this set of features is used for matching and classification of signatures using Support Vector Machines. Evaluation is performed on an offline Manipuri signature database containing 630 genuine and 140 forged signatures contributed by 70 individuals. The experimental results are found to be encouraging and effective while a set of local and global features are used for capturing the overall pattern of a Manipuri signature.