HANDWRITTEN GE’EZ CHARACTER RECOGNITION USING ARTIFICIAL NEURAL NETWORK

Achamie Aynalem

Abstract: In this paper, I’ve used artificial Neural Networks to recognize handwritten Ge’ez characters. In the paper, I’ve used one or two hidden layer feed forward Neural Network architecture with different learning algorithms to study the best performance parameters. I’ve made MATLAB user interface with adjustable network parameters to study the best performance by varying network parameters.I’ve used MATLAB image processing techniques for basic image preprocessing operations like noise removal, normalization, binarization, adaptive background removal and image border enhancement. Characters are segmented from the processed image and encoded to binary vector and fed to the multilayer ANN for training. After training the network with handwritten Ge’ez characters, a sample test data is fed to the network and tested for the performance of the network.

Keywords: handwritten character recognition, neural network, multilayer neural network, Ge’ez character recognition, image processing.

DOI: https://doi.org/10.15623/ijret.2018.0707014