OFF-LINE CURSIVE WORD RECOGNITION

After surveying the state of the art in handwriting recognition, especially in cursive word recognition, some basic notions are reviewed in the field of picture recognition, particularly of line image recognition, stating the usefulness of "regular" versus "singular" classes of features. These notions are applied to obtain a graph, G, representing a line image, by a new technique, and also to find an "axis" as the regular part of G. The complements to G of the axis are the "tarsi," singular parts of G. A segmentation of the graph is obtained, giving a symbolic description chain (SDC) A "tarsus" (singularity) is an informative feature of a cursive word, often a letter of the word. Using one or more as robust anchors, possible words are selected in a list of words. Candidate words are examined to see if the other letters fit to the remains of the SDC Good results are obtained for clean images of words written by several persons. But the problem of recognition of a word on a busy or noisy background is still an open one.