Assamese character recognition using zoning feature

Abstract

The aim of this paper is to develop a model to recognize basic Assamese characters using feed-forward neural network. The basic characters included in Assamese language are a set of numeral, a set of vowel, and a set of consonant. An algorithm has been designed to segment the line and individual character of the image and zoning features are extracted from the individual character. The network is trained by gradient descent with momentum and adaptive learning rate backpropagation training function. The network consists of two hidden layers with Sum Square Error (SSE). Finally, the unicode value of the recognized character is written in a text file.