Biografo: An integrated tool for forensic writer identification

Abstract

The design and performance of a practical integrated tool for writer identification in forensic scenarios is presented. The tool has been designed to help forensic examiners along the complete identification process: from the data acquisition to the recognition itself, as well as with the management of large writer-related databases. The application has been implemented using JavaScript running over a relational database which provides the whole system with some very desirable and unique characteristics such as the possibility to perform all type of queries (e.g., find individuals with some very discriminative character, find a specific document, display all the samples corresponding to one writer, etc.), or a complete control over the set of parameters we want to use in a specific recognition task (e.g., users in the database to be used as control set, set of characters to be used in the identification, size of the ranked list we want as final result, etc.). The identification performance of the tool is evaluated on a real-case forensic database showing some very promising results.