Generalized Harmonic Analysis for Pattern Recognition: A Biologically Derived Model.
AIR FORCE INST OF TECH WRIGHT-PATTERSON AFB OH SCHOOL OF ENGINEERING
Pagination or Media Count:
An image classification model based on cross-correlation and discrete Walsh-Fourier techniques is presented, and shown to have biological relevance. A review of some known properties of Walsh functions and Hadamard matrices, a new result for discrete Walsh-Fourier transformations, and a discussion of sequency-filtering are included. A computer simulation of the model, using hand-written capital letters as images, and the results obtained from the simulation are described. It is concluded that discrete Walsh-Fourier transforms can be filtered to facilitate reliable automatic classification of alphabetic characters. Author
- Electrical and Electronic Equipment