Liquid Characterization Using A Dielectric Slot Waveguide And Machine Learning For Sensing In The Terahertz Regime

Abstract:
In this paper, we characterize liquids by measuring their transmittance in the Terahertz (THz) regime using machinelearning models. A Vector Network Analyzer (VNA) based setup, which emits THz radiation, is used in combination with a Dielectric Slot Waveguide (DWSG) as a sensor. The high confinement within the slot is utilized to maximize the interaction between the electric field and a Liquid Under Test (LUT). Here, we present a comparative study of different machine learning algorithms used on the extracted transmission coefficients within the frequency range of 260 GHz to 400 GHz. We compare the performance of three algorithms—k-Nearest Neighbours (k-NN), Support Vector Machine (SVM), and eXtreme Gradient Boosting (XGBoost)—as candidate classifiers and report their performances in classifying different liquids. The algorithms work on the S21 (scattering parameter for transmission) measurements and attain accuracies over 90%.