Publications:A neural network approach for multifont and size-independent recognition of ethiopic characters

From ISLAB/CAISR
Revision as of 22:40, 30 September 2016 by Slawek (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Do not edit this section

Keep all hand-made modifications below

Title A neural network approach for multifont and size-independent recognition of ethiopic characters
Author Yaregal Assabie and Josef Bigun
Year 2007
PublicationType Conference Paper
Journal
HostPublication Progress in pattern recognition
Conference International Workshop on Advances in Pattern Recognition (IWAPR), Loughborough Univ, Loughborough, England, 2007
DOI http://dx.doi.org/10.1007/978-1-84628-945-3_13
Diva url http://hh.diva-portal.org/smash/record.jsf?searchId=1&pid=diva2:239377
Abstract Artificial neural networks are one of the most commonly used tools for character recognition problems, and usually they take gray values of 2D character images as inputs. In this paper, we propose a novel neural network classifier whose input is ID string patterns generated from the spatial relationships of primitive structures of Ethiopiccharacters. The spatial relationships of primitives are modeled by a special tree structure from which a unique set of string patterns are generated for each character. Training theneural network with string patterns of different font types and styles enables the classifier to handle variations in font types, sizes, and styles. We use a pair of directional filters forextracting primitives and their spatial relationships. The robustness of the proposed recognition system is tested by real life documents and experimental results are reported.