Handwritten character recognition using neural networks

Abstract

Character recognition from scanned images research problems have received lot of attention over a period of time due to its wide range of application. It has been extensively used in computers and hand held devices which have crossed the billion mark. Character recognition due to its importance has gained attention in the field of image processing, pattern recognition and artificial intelligence. In this paper, an optical character recognition technique for handwritten alpha-numeric characters is proposed using neural networks. Neural network has been extensively used in image recognition or pattern recognition due to its high adaptability, learn ability and convergence of solutions. In the proposed method, Radon filter based feature extraction along with feature extraction in the frequency domain has been used for training the networks. Testing image are compared using neural networks along with the feature extracted characters and the percentage of accuracy are presented. Three different network setup such as feed forward, multilayer perceptron and Hopfield are implemented