A New Mind of Wavelet Transform for Handwritten Chinese Character Recognition

Abstract

Featureextraction and classification recognition are the most important parts in the process of off-line handwritten Chinese character recognition. Wavelet transform is put forward by predecessors and is a kind of feature extractional gorithm. It is said that we can use separately one dimensional wavelet transform along the rows and columns two directions and get the result. This paper presents a new direction, Wavelet transform along diagonal direction. It can solve the problem that left falling and right falling can not be directly separated. In addition this paper also proposes a new mind, that is radial Wavelet transform. Used to extract the outer border of the Chinese characters.