Online Character Segmentation Method for Unconstrained Handwriting Strings Using Off-stroke Features
Abstract
In this paper, an online character segmentation method for unconstrained strings is proposed. To recognize unconstrained-expression strings such as those in phrases and informal notations, we designed physical features of segmented patterns, in particular, off-stroke features. Segmented-pattern likelihood was also defined from these features using a probabilistic model. Evaluations using a digital pen system showed that the character segmentation rates were 97.8%, 91.7%, and 75.6% of numerals, Japanese characters, and all characters (numerals, alphabets, symbols, and Japanese characters), respectively.
Loading...