Efficient Handwritten Digit Recognition based on Histogram of Oriented Gradients and SVM

Abstract

Automatic Handwritten Digits Recognition (HDR) is the\rprocess of interpreting handwritten digits by machines. There\rare several approaches for handwritten digits recognition. In\rthis paper we have proposed an appearance feature-based\rapproach which process data using Histogram of Oriented\rGradients (HOG). HOG is a very efficient feature descriptor\rfor handwritten digits which is stable on illumination variation\rbecause it is a gradient-based descriptor. Moreover, linear\rSVM has been employed as classifier which has better\rresponses than polynomial, RBF and sigmoid kernels. We\rhave analyzed our model on MNIST dataset and 97.25%\raccuracy rate has been achieved which is comparable with the\rstate of the art.\r