Feasibility of Passive Wireless Sensors Based on Reflected Electro-Material Signatures

摘要

In this paper, a neural network is applied to reconstruct the permittivity profile of a three-section passive sensor for RFID applications. Input reflection coefficients of the wave backscattered from a RF tag, over the frequency range 1-5 GHz, are used to estimate the material parameters. A neural network incorporating the Levenberg Marquardt algorithm is evaluated in terms of average absolute error, regression analysis and computational efficiency. Suitability of the algorithm is verified using both simulated and measured data, and accurate results are obtained while avoiding computational complexity. The methodology developed in this paper can be successfully used for passive sensing applications involving RFID technology to investigate and reconstruct a material profile altered by environmental variables.