System of grading hand-written multiple-choice answer sheet based on neural network

Abstract

Abstract: The aim this study is to develop a system of grading hand-written multiple-choice answer sheets based on neural network. Through this system grading can be done by computer with high speed. The study was done by developing a software to read hand-written answers and to classify them into the answer A, B, C, D, or E. The neural network used is the Perceptron model. It is made with Borland Delphi 7 and Matlab. It was found out that the system could identify hand-written marks on the multiple-choice answer sheets with an accuracy level of 68%.
Keywords: grading system, multiple-choice, neural network, Borland Delphi, vector, Optical Mark Recognition (OMR), pattern-mapping, processing unit, binary value