Recognition of Handprinted Numerals by Two-Stage Feature Extraction

An optical character recognition system for handprinted numerals of noisy and low-resolution measurement is proposed. The system consists of the two-stage feature extraction process. In the first stage a set of primary features insensitive to the quality and format of a black-white bit pattern are extracted. In the second stage, a set of properties capable of discriminating the character classes is derived from primary features. The system is simple and reliable in that only three kinds of primary features are needed to be detected. The recognition is based on the decision tree which tests the logic statements of secondary features.