An analog fuzzy processor using neuron-MOS technology

A simple CMOS fuzzy processor using a neuron-MOS center-of-mass detector as the defuzzifier has been proposed and fabricated test circuits have experimentally verified the concept. The processor receives the membership values in analog voltages and fuzzy inference is conducted by an array of MIN/MAX circuits. The neuron-MOS defuzzifier yields a crisp output as a digital code, thus providing a smooth interfacing between the analog sensory inputs and digital systems. The rules are fully programmable on the hardware and parallel processing is carried out by analog voltage-mode circuitries, enabling compact, fast and flexible implementation of a fuzzy accelerator. Test circuits were fabricated by a 0.8-μm double polysilicon CMOS process and the circuit operation has been demonstrated.