A low power consumption and compact mixed-signal Gaussian membership function circuit for neural/fuzzy hardware

Abstract

In this paper a new mixed-signal cmos integrated circuit implementation of a Gaussian membership function for neural/fuzzy hardware is presented. This GMF circuit is based on upgraded version of a linear adjustable transconductor and implements any four types of membership function: s-function, z-function, trapezoidal and triangular function with all various parameters (slope, width, height and position) independently and continuously tunable. Programmability of purposed GMF circuit can be acceded by varying the reference voltages and the size of transistors in the differential pairs. The computer simulations that verify the performance of this circuit are given, showing high speed with low power consumption. The purposed circuit was developed in CMOS 0.35µm technologies.