![]() | Only 14 pages are availabe for public view |
Abstract Due to its strong cursive nature recognition of the arabic writing is more difficult than that of latin.This problem is being investigated in this research using the established method of walsh transformation of the vertical projection of the characters images. The application of walsh transform has helped through the selection of the appropriate coefficients to recognize an arabic text in different sizes and allow a little variation in fonts. The results show that this approach is promissing as a degree of success is achieved.Important techniques are used including the neural network back propagation algorithm to build a classifier trained using the best walsh coefficients as input features. |