Publication

Abstract : Character recognition is an active research area for recognizing handwritten or optical characters using computers. The number of languages in the world is about 7000, which demands different technologies to deal with the characters. Different tools help to get the scanned documents in the form of images and then technologies are used to recognize the characters. In this paper, a convolution neural network (CNN) model is proposed for recognizing 52 Telugu characters. The performance of the proposed model has been evaluated by optimizers such as Adam, Adagrad, Adadelta, and Stochastic Gradient Descent. The preprocessing step is also supported in improving the accuracy. The CNN model is compared with VGG-16 and the accuracy obtained is less. A highest accuracy of 90.8% is achieved by Adam for the proposed CNN model.