Application of Artificial Neural Network for Electromagnetic Source Reconstruction

Abstract

In this paper, the artificial neural network is used to reconstruct the electromagnetic source. Firstly, the near-field of the radiation source is obtained, and then, the equivalent magnetic dipoles array is used to predict the radiation from the real source. The information about the near-field’s amplitude and phase is used to find the magnetic moments and locations of the equivalent dipoles, where the artificial neural network is trained for this purpose. In this way, the new near-field pattern generated by equivalent magnetic dipoles is continuously subtracted from the origin near-field pattern until the discrepancy between the origin and new near-fields meets stop criterion. Through experimental results, the accuracy and efficiency of the proposed artificial neural network method are verified.