Handwritten word recognition by image segmentation and hidden Markov models

In this paper, we propose a method for the recognition of handwritten literal amount on various bank checks. We present the pre-processing of the original 256 gray-levels image, containing inhomogeneous background, and the locating of the handwritten information. For recognition of the amount, we choose a Markovian approach, which is first applied to the sequences of words. The first results allow to consider an extension of the method to the sequences of graphems in words in order to improve the recognition rate.<>