Matching Medium Design for In-Body Communications Using Artificial Neural Networks

Matching media are located between wearable antennas and the human body to enhance implant communications. The selection of the matching medium is a complicated problem, and there is yet no well-established approach. The simulations are computationally expensive, and the theoretical work is limited. Hence, this work proposes a novel approach by using artificial neural networks for determining the effect of various matching media. For this aim, a wearable repeater antenna, human tissue blocks with varying relative permittivities, and matching medium layer with varying relative permittivities and thicknesses are utilized. Employing more than 200 simulated designs, optimum matching medium designs are proposed, and the matching medium concept has been shown to increase the average transmitted power by 17.6%. Moreover, it is shown that the trained neural network model can predict the test cases with 4.5% mean error and the computational cost has been decreased by 91% compared to the empirical method.