Persian/Arabic Handwritten Digit Recognition Using Neural Network

Optical character recognition (OCR1) is one of the most commonly used topics in field of image processing, which has undergone many researches in this today. The OCR includes three main steps, preprocessing, feature extraction and learning. In preprocessing, normalization of images and changing them in to monochrome image version are done. In feature extraction, for each sample a feature vector is created which represents the main attributes of that sample image. In the learning step, multilayer perceptron neural network that is trained with samples (a set of handwritten Persian/Arabic digits) match the input test images to the digits. In this paper presents a method for persian/arabic handwritten digit recognition using neural network. The proposed system consists of three phases, preprocessing, feature extraction and learning features. Learning step is done by multilayer perceptron neural network (MLP2) and a training set of digit image samples. That uses to detect Persian/Arabic digits. The proposed method are tested by MAHDBase database and the experimental results shows that the proposed algorithm has a good accuracy

کلیدواژه ها:

OCR ، Persian/Arabic Digit Recognition ، MLP