Handwritten Numerical Characters Recognition Using Biorthogonal Spline Wavelets

Suzete E. N. Correia, João M. de Carvalho

DOI: 10.14209/sbrt.2000.5250251

Evento: XVIII Simpósio Brasileiro de Telecomunicações (SBrT2000)

Keywords:

Abstract

"In this paper, a novel approach for recognition of handwritten numerical characters using the biorthogonal spline wavelets Cohen-Daubechies-Feauveau (CDF) 3/7 is proposed. The main purpose is to optmize character recognition methods and develop a system which is able to efficiently absorb the variations presented in this type of data. The proposed scheme consists of three stages: scale normalization, feature extraction and classification. The classifier used is a multilayer cluster neural network. In order to verify the performance of the method, experiments were realized with the numerical database of Concordia University of Canada, obtaining an recognition rate of 94.7 %."

Download