The computation of complete and reduced sets of orthogonal spectral coefficients for logic design and pattern recognition purposes

The classic methods of computing the spectral coefficient values for any n-variable binary function, involving a 2ⁿ × 2ⁿ transform matrix, are reviewed. From this established starting point methods of generating the full set of spectral coefficients from decompositions of the function are considered, followed by possible methods of generating a reduced set of coefficients which may be sufficient for certain practical applications.