Applying Fourier-Transform Infrared Spectroscopy and Self-Organizing Maps for Forensic Classification of White-Copy Papers

Abstract

White-copy A4 paper is an important kind of substrate for preparation of most formal as well as informal documents. It often encountered as questioned document in cases such as falsification, embezzlement or forgery. By comparing the questioned piece, (e.g. of a contract) against the rest deemed authentic, forgery indicator could be derived from an inconsistent chemical composition.  However, classification and even differentiation of white copy paper have been difficult due to highly similar physical properties and chemical composition. Self-organizing map (SOM) has been proven useful in many published works as a good tool for clustering and classification of samples, especially when involving high-dimensional data. In this preliminary paper, we explore the feasibility of SOM in classifying white copy paper for forensic purposes. A total of 150 infrared spectra were collected from three varieties of white paper using Attenuated Total Reflectance Fourier-transform infrared (ATR-FTIR) spectroscopy. Each IR spectrum composed of over thousands of wavenumbers (i.e. input variables) and resembles chemical fingerprint for the sample. Comparative performance between raw wavenumbers and its reduced form (i.e. principal components, PCs) in SOM modeling also conducted. Results showed that SOM built with PCs is much efficient than built with raw wavenumbers, with the classification accuracy of over 90% is obtained with external validation test. This study shows that SOM coupled with ATR-FTIR spectroscopy could be a potential non-destructive approach for forensic paper analysis.