Keyword Spotting in Cursive Script Using Hidden Markov Models

This paper describes an expansion of the Viterbi algorithm in a cursive script word recognizer based on Hidden Markov Models (HMM) that allows the detection of subwords. The advantage of word recognizers based on HMMs is that there is no need for a segmentation of the word image into single letters. In common systems segmentation still has to take place on a word level which rises problems in many applications. The method described here generalizes the task of the Viterbi algorithm of matching a chain of HMMs to only a subset of feature vectors describing the input image. This ,,wordspotter” has been implemented and tested with street names which showed some promising results. Developing the right strategy of using the benefits of the wordspotter could be the clue to handle recognition tasks with word lists much bigger than those possible today.