A Multiple-Choice Test Recognition System Based on Android and RBFNN

Abstract

Optical Mark Recognition (OMR) is often used by teachers to help mark students’ test in the form of multiple-choice test. This article describes how to build an offline OMR system which run in Android based smartphone. The input is the digital image of student’s answer sheet which was taken from smartphone’s camera. The system implements image processing techniques to detect and determine not only student’s handwritten for his/her ID, name, and form’s ID, but also for the answer of each test number. Gradient feature is used as features for the handwritten parts, while pixel density and area are used for the marked parts. Radial Basis Function Neural Network (RBFNN) is implemented to recognize of the students’ handwriting. The experiment result shows that the system has successfully identified the answers with 99.69% success rates, but only has 82.28% and 72.25% success rates for digits and uppercase letters recognition respectively.