Lexicon-driven handwritten word recognition using Choquet fuzzy integral

The core process in a lexicon-driven handwritten word recognition system is a matching algorithm that finds the best segmentation of a word image matching to each word string in the lexicon under the criterion defined by a match function. The match function aggregates the character matching scores to obtain the word matching score. In this paper, we use the Choquet fuzzy integral as a match function for the matching process and develop a fast approximate matching algorithm using a dynamic programming technique. We then demonstrate the recognition rate improvement through some experiments using the standard SUNY CDROM data set