Handwriting identification using global and local features for forensic purposes

This paper presents some research work that has been undertaken in the area of software tools for forensic document analysis. The purpose of research into this area is to study the methods used by human experts in the forensic analysis and examination of questioned hand-written documents and to formulate computer tools to assist in this task. The computer tools needed will require image processing and pattern recognition techniques. As part of this research a system for automatic identification of the author of hand-written documents is being investigated. Such systems usually attempt to match a questioned document to a known sample from a large database. It is hoped that by investigating methods that can be used to solve this problem it will be possible to develop methods that will aid the document examiner in his comparison of questioned and known documents. An important set of tools that are being investigated help to extract global and local features from the questioned documents. These techniques are presented in this paper.