Handwritten Numerals Recognition, New Approach

There are many important applications for recognizing the handwritten numerals, one of the most important and significant application is the automation of postal services especially in countries that use different languages at the post offices, such as India. The purpose of this study is to develop a recognition method for recognizing the handwritten numerals using neural networks and wavelets transform. This study is going to present a method that uses a directional continuous two dimensional wavelet transform besides the neural networks. The transform will be used as an extraction technique of the proper features that are required for the classification before using them as an input to the feed-forward neural network. The results are promising and significant improvements could be achieved to enhance the recognition performance. A recognition rate of 90.3% could be achieved. Key wrods: Neural Networks, Character Recognition, Wavelet Transform.