Numerical character recognition via matching between coded patterns and feature vectors
L.L. Lee, B.R.M. Sosa · 2002
This paper describes a novel approach for numerical character recognition based on the matching between coded patterns to feature vectors. A one-to-one correspondence among coded patterns, feature vectors and decision regions is established. Each numerical character is composed of eight primitive patterns. The procedure used to identify an unknown numerical character consists of recognizing primitive patterns, which are codewords of a linear code, and being matched to feature vectors.