CMOS cells with continuously adjustable parameters for implementation of fuzzy and neurofuzzy systems

The use of fuzzy logic, neural networks, and more recently, neurofuzzy systems, has increased the necessity of counting with efficient architectures to solve the problems present in many research areas, not only at software level, but at hardware level, which presents a higher performance in most applications. In this manuscript we present a novel set of basic analog CMOS cells to generate triangular membership functions, which can adjust continuously their parameters in current-mode. This characteristic results crucial when implementing neurofuzzy systems such as NEFCLASS, NEFPROX and NEFCON architectures, which are aimed at in this work, nevertheless our cells will work well in any fuzzy/neuro fuzzy system using triangular, left and right shoulder membership functions. The proposed cells are extensions of those originally presented in early research, in such a way that the current gain can be electrically modified and set to the optimal value required by the neurofuzzy system.