Return to search

Design and implementation of a portable omnifont reading aid for the blind

The design and implementation of a handheld scanner that can help sight-impaired or even blind users to manually scan and read text is discussed in this dissertation. A thorough investigation of all the elements involved in such a system is presented and optimal solutions are proposed. A unique velocity compensation technique based solely on optical information obtained by the scanning device is discussed and a real time segmentation technique based on topological properties (Quasi-Topological Codes) of connected segments is presented. A skew detection algorithm is discussed that can trace typed and printed text manual1y scanned with skew up to 15 degrees and can guide blind users to properly scan a document. Real time extraction of quasitopological codes for automatic text recognition and the hardware implementation is also discussed in this work. A hierarchical optical character recognition method is proposed which is based on syntactic and metric analysis of the Quasi-Topological Codes and their position in the scanned image. The proposed method can recognize characters stretched to approximately two times their original width or rotated by a few degrees. Finally, an automated iterative learning process is discussed which includes generalization of the recognition logic and dynamic adaptation of the syntactic and metric recognition rules. / Ph. D.

Identiferoai:union.ndltd.org:VTETD/oai:vtechworks.lib.vt.edu:10919/39385
Date16 September 2005
CreatorsAsimopoulos, Nikos
ContributorsElectrical Engineering, Nadler, Morton, Besieris, Ioannis M., Gray, Festus Gail, Claus, Richard O., Smith, Eric P.
PublisherVirginia Tech
Source SetsVirginia Tech Theses and Dissertation
LanguageEnglish
Detected LanguageEnglish
TypeDissertation, Text
Formatv, 129 leaves, BTD, application/pdf, application/pdf
RightsIn Copyright, http://rightsstatements.org/vocab/InC/1.0/
RelationOCLC# 24956815, LD5655.V856_1990.A855.pdf

Page generated in 0.0016 seconds