筆者識別指標の検討を支援するデータベースシステムの開発

Abstract

In criminal investigations, writer-identification to discriminate writers of documents used in crimes has important roles. In the identification processes, writer-unknown documents are compared with writer-known documents. Experts of the process give careful considerations for the comparisons by referring to handwriting-samples, experimental-data, and other vast knowledge collected from past investigations. Characteristics of handwritten strokes written on the documents are highly useful for the experts. Therefore, accurate and efficient managements of the characteristics are important. However, in the daily investigations, the characteristics have been recorded in a huge amount of paper media without indexes, and the amount is growing. In this paper, we propose two prototypes of database system to store and retrieve the characteristics. The database system consists of (1) common digital formats to indicate the characteristics, (2) aggregation functions of the digital indicators for the retrieval, and (3) user interface for the experts. The first prototype is implemented with VBA of Microsoft Excel. We present our experimental results by using 228 testing data sets. The second practical prototype is implemented with XML and Java. We expect that our database system will improve the efficiency and the accuracy of the writer-identification processes, and accelerate information sharing among the experts beyond time and space.