Recognition of Arabic Handwritten Characters using Discrete Wavelet Transforms and Neural Network

New approach of recognizing Arabic handwritten characters have been proposed, includes two main steps. In the first step, the input image which includes Arabic handwritten characters is acquisitioned using digital scanning, and then the scanned image is transformed using discrete wavelet transform (DWT) to extract a specific vector of features from the transformed image which reflects the texture of the scanned characters. While the second step performs the classification job using neural network based on the extracted features from first step. The classification process using ANN is trained by 208 images and other 70 images for testing. The simulation results of classification prove the capability of the proposed approach to perform significant percent of accuracy especially when Coiflet3 filter is considered for DWT and number of layers in ANN is increased to 6 or layers.