Totally unconstrained handwritten numeral recognition via fuzzy graphs
Ibrahim S. I. Abuhaiba, P. Ahmed · 2002
The performance of a novel recognition system which uses fuzzy constrained character graph models (FCCGMs) is investigated for totally unconstrained and handwritten numeral recognition. The system was tested on 1812 unnormalized samples. The reliability, recognition, substitution error, and rejection rates of the system were 97.1%, 90.7%, 2.9%, and 6.4%, respectively.>