Artificial Neural Networks for Antenna-based Contactless Liquid Classification

Neural networks are enhancing modern day sensors and results for classification and real time monitoring. The complexity of neural networks are enabling the future for image classification, self-driving cars and redefining how information is processed. In this paper, a 4-port antenna configuration was adopted for contactless sensing and characterization of liquids. A standard gain horn antenna with an output power of 17 dBm was placed 50 cm away from a 500 ml plastic container. The response of the system was monitored for two sample liquids: Water and 95% Ethanol. The measured microwave response was then classified across a 3-3.15 GHz frequency range by a trained neural network with a 94.44% accuracy. The results demonstrated the future ability of mixing microwave technology with neural networks for event safety.