Parallel Optical Walsh Expansion in a Pattern Recognition Preprocessor Using Planar Microlens Array

We have developed a parallel optical preprocessor for a pattern recognition system using a planar microlens array and a Walsh orthogonal expansion spatial filter. We experimentally demonstrated the parallel optical Walsh expansion of multiple images made by the planar microlens array with good accuracy, which assures 99% recognition of simple numeral characters in the system. In order to enlarge the tolerance of the recognition rate against the deformation of input patterns, we propose a novel selection method of Walsh expansion coefficients. From the computer simulation, we confirmed that 5% transverse shifts and 20° rotation of input patterns are allowable with the optimization of the spatial filter using this method.