On the use of a priori knowledge to character recognition

A machine-printed character recognition system that relies on prior knowledge of character shapes is presented. Because the classification performance is strongly dependent on the input feature set, the focus is on the creation of features based on curvature estimation. A set of `clean' characters from multiple fonts was used to emphasize the authors' belief that a clean set of characters can be used to build an inference engine to recognize noisy characters. Clustering of a clean character set from multiple fonts led to 263 unique character contours