Hardware Resource Minimization of Triangular Membership Function Based on FPGA

Adaptive Neuro-Fuzzy Inference System (ANFIS) is today's common control and decision-making system. ANFIS is used in several applications related to control, intelligence fields, and biomedical systems. The ANFIS consists of five layers in which the first layer contains the membership function (MF), such as Gaussian, Trapezoidal, bell, sigmoidal, D- sigmoidal, and P-sigmoidal membership functions. The FPGA-based hardware implementations of these membership functions are quite difficult and occupy huge logic elements. This study implements the triangular membership function based on FPGA using five different techniques. Comparing the proposed approaches, the one with the symmetrical property, consuming 52 LUTs and 32 occupied slices, yielded the best results. According to the results obtained, this approach has the lowest hardware resource utilization compared to the other proposed designs and compared to the results demonstrated in the literature.