A Handwritten Chinese Character Recognition Method Combining Sub-structure Recognition

Abstract

Many Chinese characters are composed of sub-structures. Extracting and recognizing radicals or sub-structures are benefit to character recognition. This paper proposed a new handwritten Chinese character recognition method combining sub-structure recognition. Firstly, a density-based clustering method is adopted to find sub-structure patterns in sub-structure pattern discovering. Secondly, for multiple sub-structure characters, the single Chinese character recognition problem is converted to a sub-structure string recognition problem. By searching the most matched sub-structure string pattern to a character, the additional character recognition candidates are obtained. These candidates and the single character recognition results are combined to yield the final character recognition result. Experiment results on CASIA dataset show that this work is effective on improving handwritten Chinese character recognition as well as sub-structure pattern discovering.