Recognition algorithm based on wavelet transform for handprinted Chinese characters

In this paper, we propose a new method for recognition of handprinted Chinese characters. The method absorbs the local variations in handprinting by using the property of the wavelet transform. It is expected that the representation in multiresolution analysis (MRA) with low resolution can absorb the local variation in handprinting than in MRA with high resolution. Furthermore, the representation can detect the horizontal edge and the vertical edge in the original image of a character which are effective in recognition, respectively. Because of these advantageous properties, that is, the capability of absorbing the local variation in handprinting and the ability of detecting the edge lines, we construct the feature vector for recognition by using the wavelet representation in MRA. The performance of the proposed method is superior to that of the directional pattern matching method which is a representative traditional method