Oo-line Cursive Handwriting Recognition Using Recurrent Neural Networks 2 Handwriting Recognition 10 2.1 a Taxonomy of Handwriting Recognition Problems

Computer handwriting recognition ooers a new way of improving the human-computer interface and of enabling computers to read and process the many handwritten documents that must currently be processed manually. This thesis describes the design of a system that can transcribe handwritten documents. First, a review of the aims and applications of computer handwriting recognition is presented, followed by a description of relevant psychological research. Previous researchers' approaches to the problems of oo-line handwriting recognition are then described. A complete system for automatic, oo-line recognition of handwriting is then detailed, which takes word images scanned from a handwritten page and produces word-level output. Methods for the normalization and representation of handwritten words are described, including a novel technique for detecting stroke-like features. Three probability estimation techniques are described, and their application to handwriting recognition investigated. The method of combining the probability estimates to choose the most likely word is described, and performance improvements are made by modelling the lengths of letters and the frequency of words in the corpus. The system is tested on a database of transcripts from a corpus of modern English and recognition results are shown. Recognition is described both with the search constrained to a xed vocabulary and with an unlimited vocabulary. The nal chapter summarizes the system and highlights the advances made before assessing where future work is most likely to bring about improvements. Declaration This thesis describes research carried out at Cambridge University Engineering Department between October 1991 and September 1994. It is the result of my own work and contains no work done in collaboration. The length of this thesis, including references and gure captions, is thirty-seven thousand words. Acknowledgements First of all, I would like to express my gratitude to the late Professor Frank Fallside, for supervising me during the rst half of this thesis and for providing the original inspiration for this work. I am also indebted to Dr Tony Robinson who has supervised me admirably for the latter half of this thesis with enthusiastic guidance and support, particularly in the last few weeks. I would like to thank everyone else in the Speech, Vision and Robotics group which Frank Fallside created. The group has been an ideal environment , both socially and technically, in which to conduct research. Those in the group who have helped in the creation of this thesis are too numerous to mention individually. Special thanks must …