In this paper, we present a handwritten character string recognition system on medical prescription reading. The medicine's words are recognized as an entire because there is no extra space between characters. The lexicon contains 520 medicine's words, which are stored in a trie structure. The recognition, the text line image is matched with the lexicon entries to obtain reliable segmentation and retrieve valid medicine's words. The pre-segmentation, the text line image is separated into aboriginal segments by connected component analysis. Lexicon matching, coherent segments are dynamically combined into candidate character patterns. The character classifier is embedded in lexicon matching to select characters matched with candidate pattern from dynamic series category. In the experiment result on 5200 handwritten character images achieved correct rate of 74.13%.
展开▼
