Smart Flexible Turbidity Sensing Based on Embedded Neural Network

This paper presents a flexible turbidity sensing system that implements single and multiple infrared (IR) beam architectures. The hardware includes two pairs of IR LED - IR photodiode whose relative position permits the materialization of different architectures for turbidity optical measurement. The control of the sensing architecture is performed by a microcontroller based I/O board that is wireless connected to a PC for advanced control and data processing based on neural networks. Using intelligent processing procedures, the best turbidity measurement architecture is selected in terms of measurement accuracy for a given turbidity level. Elements of neural network design, optimization and embedding as part of an advanced processing software component are presented.