The Application of Filtered Transforms to the General Classification Problem
Joseph W. Carl, Charles Francis Hall · IEEE Transactions on Computers · 1972
An image classification model based on nearest prototypes in filtered Fourier and Walsh transform domains is presented. A computer simulation of the model applied to handwritten English letters, Russian letters, numerals, and electromagnetic signals is also presented. Experiments to date fail to refute the working hypothesis that generalized harmonic analysis can be used to reliably classify alphabet characters, time-varying signals, and other images.