Research On Handwritten Chinese Character Synthesis

The recognition and generation of handwritten character have always been the difficult problems in the field of pattern recognition. The research of handwritten recognition is helpful to solve some basic problems of pattern recognition. The research of handwritten generation is helpful to handwritten recognition, it can provide training samples for machine learning and pattern recognition, and also we can use the synthetic characters to generate personal-style documents.Most of the existing algorithms for character synthesis deal with English, and they are not suitable for Chinese character synthesis. In this paper, we proposed two different ways to synthesize Chinese character; the first is about shape statistical method which has been used to synthesize English characters. It contains three aspects: representation of shape variation of Chinese character, representation of stroke's shape and statistic of spatial layout of stroke.The second is a method for modeling and synthesizing Chinese characters based on hierarchical representation of character. Using a three-level hierarchical representation, each character is decomposed into basic components, which forms the stroke database and radical database. In the synthesis process, we use a wavelet-based approach to select proper strokes and radicals, and some aesthetic constraints are defined based on the relationships between components, then genetic algorithm is employed to search for the optimal results which best match the aesthetic constraints. The details lie on four aspects:(1) Three levels in representation for Chinese character Many thousands of Chinese characters can be composed from a relatively small number of basic strokes and radicals. Our approach decomposes Chinese character into the three levels:"characterâ†'radicalâ†'stroke".(2) Modeling of relationships between components Since the relationships between components are import to personal style Chinese character, we define some aesthetic constraints. Using these constraints, we search the optimal result.(3) Features extraction based on wavelet decomposition The low frequencies of Wavelet decomposition denote the profile information. Extract these more important profile information as features to select the most likely radicals and strokes for candidates.(4) Optimal Combination With a Genetic Algorithm Search the optimal result of synthetic character is very time-consuming, we employ Genetic Algorithm to search the approximate optimal result.