Approximating properties of artificial neural network in time domain for the analysis of electromagnetic fields reflected from model of human body surface

Abstract

The approximating properties of artificial neural network for the problem of determination of thickness a layer of layered medium by means of the analysis of reflected impulse electromagnetic fields are considered. The impulse fields are reflected from the model of human body surface. The normal incidence of plane wave with Gaussian time form on the layered medium with losses is considered. The reflected electromagnetic field is obtained by FDTD method. Initial data for neural network analysis are the values of amplitude of electrical component of reflected field in different moments of time. So, the analysis is performed in time domain directly. As an example, the network is trained to determine the thickness of one of the layers of the medium. The stability of the determination in presence of interferences, experimental errors and instabilities of medium parameters is investigated. The approximating properties of the artificial neural network are considered for the smooth change of the second layer thickness.