Forensic Analysis of Offline Signatures Using Multilayer Perceptron and Random Forest
Abdul Salam Shah, Masood Shah, Muhammad Fayaz, Fazli Wahid, Hira Khan, Asadullah Shah · International Journal of Database Theory and Application · 2017
Forensic applications have great importance in the digital era, for the investigation of different types of crimes.The forensic analysis includes Deoxyribonucleic Acid (DNA) test, crime scene video and images, forged documents analysis, computer-based data recovery, fingerprint identification, handwritten signature verification and facial recognition.The signatures are divided into two types i.e. genuine and forgery.The forgery signature can lead to the huge amount of financial losses and create other legal issues as well.The process of forensic investigation for the verification of genuine signature and detection of forgery signatures in law related departments has been manual and the same can be automated using digital image processing techniques, and automated forensic signature verification applications.The signatures represent any person's authority so the forged signatures may also be used in a crime.Research has been done to automate the forensic investigation process, but due to the internal variations of signatures, the automation of signature verification still remained a challenging problem for researchers.In this paper, we have further extended previous research carried out in [1-2] and proposed a Forensic signature verification model based on two classifiers i.e.Multilayer Perceptron (MLP) and Random Forest for the classification of genuine and forgery signatures.