Binary filters for pattern classification

Generalized Fourier correlators imposing finite system space-bandwidth products are described, and a class of binary filters is proposed. In pattern classification and signal registration applications, it is shown that for a particular class of signals, the binary filters yield the same asymptotic performance as the matched filter. It is hence adduced that a dynamic range of a single bit in the filter suffices for classification purposes. The effects of statistical sidelobe fluctuations and a finite system space-bandwidth product are included in the analysis. It is demonstrated that performance improves in a natural fashion with increase in the space-bandwidth product for both the binary filter and the matched filter.<>