Deep Neural Network Modeling of Different Antenna Arrays; Analysis, Evaluation, and Application

Abstract

Array signal processing and systems brings merit for synthesizing the optimal radiation pattern for steering the mainbeam in the direction of interest as well as suppressing interference signals in the direction of not interest while preserving the beamwidth. In this study, we have rigorously represented the deep neural network (DNN) modeling of the different geometries of the antenna arrays in three scenarios. In the first scenario, we briefly review and highlight some of the research methodologies and involved problems of the DNN modeling. The ultimate validation and verification of the DNN modeling have been fulfilled by estimating the DNN outputs and obtaining the radiation patterns of the different array geometries. In the second scenario, we have employed the three concepts of mean squared error (MSE), Shannon error entropy minimization (EEM), and bit error rate (BER) for the performance evaluation and validation of implementing the proposed DNN models for the different array geometries. The concept of Shannon error entropy can be extended to optimize the nonlinear and non-Gaussian problems. In the third scenario, the application of the DNN models in the direction of arrival (DoA) estimation has been analogously represented among the different array geometries.