Implementation of the fuzzy ART neural network for fast clustering of radar pulses
M.-A. Cantin, Yves Blaquière, Yvon Savaria, Éric Granger, Pierre Lavoie · 2002
A real time radar signal clustering problem is resolved by a dedicated hardware implementation of the fuzzy ART neural network. This novel architecture implements a reformulated algorithm for high speed clustering. The proposed dedicated digital VLSI system is composed of cascadable integrated circuits, each one containing several neural processors, comparators, a divider and blocks of RAM. This efficient solution was designed and will be implemented in the near future. The basic component requires 74 K gates and occupies an area of 52.5 mm/sup 2/ in a 0.8 /spl mu/m BiCMOS technology. Each chip process an input pattern for 32 neurons every 2 /spl mu/s.