System for Optical Character Recognition of Macedonian Cyrillic Texts Aimed for Helping Visually Impaired Persons
Abstract: System for optical character recognition (OCR) of Macedonian Cyrillic texts is presented in this thesis. It's aimed for helping visually impaired persons as a part of a larger system, which includes text-to-speech conversion for Macedonian language, and Braille printing. The recognition is based on character segmentation and isolated character recognition. Character classification is performed using weighted voting of adaptive logic networks. Contextual postprocessing on word level using a lexicon for Macedonian language is also performed.
Keywords: optical character recognition, OCR, adaptive logical network, ALN, classifier combining, weighted voting, contextual postprocessing, computer lexicon
download 8MB gzipped postscript (in Macedonian)