We present a pattern recognition algorithm for hand-printed characters, based on a combination of the classical least squares method and a neural-network-type supervised training algorithm. Characters are mapped, nonlinearly, to feature vectors using selected quadratic polynomilas of the given pixels. We use a method for extracting an equidistributed subsample of all possible quadratic features. This method creates pattern classifiers with accuracy competitive to feed-forward systems trained using back propagation; however back propagation training takes longer by a factor of ten to fifty. (This makes our system particularly attractive for experimentation with other forms of feature representation, other character sets, etc.) The resulting classifier runs much faster in use than the back propagation trained systems, because all arithmetic is done using bit and integer operations.

Date of creation, presentation, or exhibit



This is a pre-print of a paper published by Springer. The final publication is available at via

Copyright 1993 Springer.

Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works in February 2014.

Document Type

Conference Paper

Department, Program, or Center

Chester F. Carlson Center for Imaging Science (COS)


RIT – Main Campus