Handwritten numeral recognition using multi-wavelets and neural networks
Kambiz Rahbar, Muhammad Rahbar, Farhad Muhammad Kazemi · 2006
Abstract: In this paper, we develop a handwritten numeral recognition using multiwavelets and neural networks. We first transform the character image into polar coordinate ( r, θ) using the center of mass of the character as origin, then by performing the Fourier transform, characters spectrum achieved; afterward the multi-wavelet transformation is applied to get the proper features; to classify them into predefined classes they are fed to a feed-forward neural network. We apply this method on the MNIST dataset. The results show that there are some significant improvements on recognition performance. We could achieve 93.2 % recognition rate. Keywords:- OCR, Multi-wavelets, Neural networks, Pattern recognition, Invariant descriptor