Artificial neural network-based approach for Mizo character recognition system


This paper examines and compares different types of artificial neural networks for recognition of Mizo script. The Mizo script and English script are very similar in nature but special characters – Â, â, Ê, ê, Î, î, Ô, ô, Û, û, Ṭ, and ṭ – are incorporated in Mizo scripts which are not available in English script. In this experiment, an attempt was made to recognize Mizo script including capital letter, small letter, numerals, and special characters. The approach involves processing input images, conversion of image characters into binary matrix, analyzing and mapping the binary matrix, training and testing with a set of desired Mizo characters using different types of neural networks such as back propagation algorithm, radial basis function, learning vector quantization, and recurrent neural network. The results are compared with each other and suggest the best artificial neural networks algorithm for use in the recognition of Mizo script.

Full Text Download