Feature extraction is one of the principal components of any pattern recognition system. This thesis focuses on extraction and use of features coming from a wavelet transform applied on character recognition. The followed objective consists of proposing and studying new variants of processing methods of this feature type.; In these methods, we have used Fourier transform followed by wavelet transform. We have extracted the most discriminant features and have validated them by the help of two classifier types: 1NN and SVM with validation. Interesting results have been obtained, indicating that considered variants of these feature families are promising. |