A Power-Efficient Analog Hardware Type-2 Fuzzy Classifier Architecture for Water Potability

This work presents an ultra-low-power analog fuzzy classifier based on Gaussian membership functions, MIN/MAX rule evaluation, and OTA-based defuzzification, operating in the subthreshold region for energy-efficient real-time inference. The architecture is evaluated on a water quality assessment dataset with reduced feature complexity, achieving competitive accuracy of 85.98% compared to the software equivalent of 89.42% and demonstrating suitability for low-power edge applications with a total power consumption of 72nW.