PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Method of verifying declared identity in optical answer sheets
Joseph Levi, Yosef Solewicz, Yael Dvir and Yisca Steinberg
International Journal of Soft Computing Volume Special Issue on Soft Computing for Digital Information Forensics, 2009.


In the use of Optical answer sheet examinations, occasionally the identity of the examinee is in question. A novel method for characterizing personal quality of mark shapes in optically-read answer sheets is described. The application of the method is for identifying imposters in multiple choice examinations. All the marks are segmented and measured in multiple parameters including area, dimensions, perimeter, optical density. Imposter decisions are made on the collected data by comparing an identified test form against the form in question in comparison with a population using SVM (Support Vector Machine) modeling. In testing the method 300 test forms from 100 examinees from past tests were used. An EER of 15.5% was found. While performance of the presented method is currently insufficient for practible purposes, future research options are mentioned.

EPrint Type:Article
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Machine Vision
Learning/Statistics & Optimisation
ID Code:4856
Deposited By:Yosef Solewicz
Deposited On:24 March 2009