تمييز الحروف العربية باستخدام شبكة الانتشار العكسي المعتمدة على العزوم السبعة

	تمييز الحروف العربية باستخدام شبكة الانتشار العكسي المعتمدة على العزوم السبعة
Tikrit Journal of Pure Science
Article 1, Volume 14, Issue 1, 2009, Pages 170-176
Abstract
In this research, it has been using seven moments to extract attributes of Arabic character images to calculates values of these moments, and then enter these values to the backpropagation neural network to recognize Arabic characters by intelligence methods. Also it has been building a simulating program for backpropagation neural network for recognition of Arabic characters moments experiments. It has been implementing the programs under windows98 environment to exploitation computer memory. The programs was written using mathlab compiler version 6. The backpropagation neural network was trained in different shapes of group of character that include ( ا، ب ، ح ، ع ، س ، ظ ، ر ، ذ ، ف، ك) using several types of training functions that was used with backpropagation neural network like Gradient descent backpropagation function, Gradient descent w/adaptive lr backpropagation function, Gradient descent w/momentum backpropagation function and BFGS quasi-Newton backpropagation function. The last was working slower with the data of the network but achieved better results during testing of character that was not trained previously by the neural network and it was reached goal in training by fewer iterations from other functions.

Statistics Article View: 188 PDF Download: 150