On-line handwritten Chinese character recognition based on inter-radical stochastic context-free grammar

This paper presents a new radical-based recognition method for online handwritten Chinese characters focusing on their hierarchical structure. Inter-radical stochastic context-free grammar (SCFG) is introduced to represent the character generation process where radicals as structure elements. Inter-radical SCFG combines the radical shape likelihood with the relative position likelihood between radicals/meta-radicals. The character pattern is over-segmented by three-layer nested pre-segmentation. Character-radical dictionaries of all character classes are unified into several big tree structures where character-parts (sub-structures) are shared by different character classes. Combining inter-radical SCFG with tree structural character-radical dictionaries, the optimal radical segmentation and recognition result is obtained during hierarchical dynamic programming (DP) search. We have implemented the method to Chinese characters of left-right and up-down structures. Experimental results on a sample set of 5,773 character classes consisting of 1,149 radicals show the proposed method is comparable to our previous method.