A fuzzy neuron and its application to a hand-written character recognition system

A fuzzy neuron in which weighting factors and summation of a conventional neuron model are modified by a membership function and fuzzy logic functions, respectively, is described. An inhibitory synaptic junction is represented by a fuzzy logic complement. This fuzzy neuron facilitates a pattern recognition of a hand-written character by only one layer. It is verified by constructing the pattern recognition hardware system. The distinctive features of this system are: (1) its simplicity of design, (2) its robustness against noises and defects in the pattern, and (3) its high speed (500 ns)