Zoran Stankovic

Loaded cylindrical metallic cavities modeling using neural networks

The feasibility of using neural networks for resonant frequencies determination in loaded cylindrical metallic cavities is presented. The load in the form of a homogeneous dielectric slab with losses, which is located on the bottom of the cavity, is considered. The use of the classical multilayer perceptron (MLP) network is illustrated through the example of TM/sub 113/ modes determination in a microwave cavity with circular cross-section. An original approach for decreasing the neural network training set is used.

Modelling of the cylindrical metallic cavity loaded by lossy dielectric slab using neural networks

In this paper, the loaded cylindrical metallic cavity loaded by lossy dielectric slab is modelled using multilayer perceptron networks. A proper neural model is defined, which includes the dielectric slab losses and is used for modelling of the cavity with circular cross-section when lossy dielectric slab is placed at the bottom of the cavity. An original approach for significantly decreasing neural network training set is suggested. The results obtained show that the suggested approach provides microwave cavity modelling with high accuracy.

Efficient DOA estimation of impinging stochastic EM signal using neural networks

In this paper a method for the accurate and fast determination of direction of arrival (DOA) of impinging electromagnetic signal radiated from stochastic sources in the far-field is proposed. The method is based on neural models using MLP (Multi-Layer Perceptron) artificial neural network. To illustrate the applicability of the proposed method, two MLP models for one-dimensional (1D) DOA estimation (in azimuth plane) are presented: MLP model for the estimation of angle position of one stochastic source and MLP model for the estimation of two stochastic sources position at fixed angle distance. Presented models perform very fast 1D DOA estimation and therefore they are very suitable for the real time applications. The architecture of developed models, their training results and simulation results are described. in details.

Modeling of patch antennas using neural networks

In this paper patch antennas are modeled using neural model based on multi-layer perceptrons (MLP) network. Neural model is trained by data, which are obtained by electromagnetic simulation of antennas using HFSS 9.0 software. This model has four input parameters: patch antenna length L, patch antenna width W, deep of patch antenna slot l and width of patch antenna slot s and it enables quick and correct calculation of resonant frequency f/sub r/ and minimum value of S/sub 11/ parameter (S/sub 11min/).

The hybrid-neural empirical model for the electromagnetic field level prediction in urban environments

The application of multilayer perceptron networks to calculating the electromagnetic wave path loss in an urban environment for propagation through an area with low or high buildings is presented. A hybrid neural-empirical model, created in two phases, is proposed. The first phase implies the realization of an approximate (coarse) propagation model based on measured values. This model determines the propagation loss from the beginning of the area, based on the distance from the area beginning, the average building density, the partial loss of a single building, the distance from the transmitter and the exponential loss index of the area. In the second phase, a neural network and the approximate model are integrated in the hybrid (fine) model of the propagation area. The input parameters for the neural network are the distance from the area beginning and the average height of buildings in that area, while the output parameter is the partial loss of a single building. This value is used in the approximate model, in order to obtain the propagation area model with higher accuracy.

Modeling of the cylindrical metallic cavity with circular cross-section using neural networks

In this paper, a loaded cylindrical metallic cavity with circular cross-section is modeled using classical multi-layer perception (MLP) network. The load in the form of a homogeneous dielectric slab with losses located on the bottom of the cavity is considered. Several training approaches of the neural model are applied, where an original approach for decreasing the neural network training set is used. The obtained results are discussed and compared. Also, the modeling results are experimentally verified.

Neural networks-based DOA estimation of multiple stochastic narrow-band EM sources

Localization of multiple stochastic narrow-band electromagnetic sources in the far-field is considered in the paper. Artificial neural networks-based approach is proposed to allow for an efficient direction of arrival (DOA) determination of electromagnetic signals radiated from stochastic sources as one of the key steps in the source localization procedure. It uses correlation matrix, obtained by signal sampling via antenna array in far-field scan area, to train an appropriate model based on MLP (Multi-Layer Perceptron) neural network. Proposed approach is validated on the example of a neural model performing accurate and fast one-dimensional (1D) DOA estimation of the position of three stochastic sources placed at fixed angle distance in azimuth plane.

Near-Earth Propagation Loss Prediction in Open Rural Environment using Hybrid Empirical Neural Model

New hybrid empirical neural (HEN) model for prediction EM field level of wireless communication transmitter in frequency range 150-1500 MHz is presented in this paper. Previous empirical Okumura-Hata model, which has been very often used, does not consider all influences of environment that signal goes throughout. Therefore, the results obtained by this model can disagree with the measured results. Considering Okumura-Hata model as approximate empirical knowledge holder and connecting it with the artificial neural network, in this paper we will realize the HEN model whose estimated results are more closed to measured results than Okumura-Hata results.

Neural Network Approach for Efficient DOA Determination of Multiple Stochastic EM Sources in Far-field

An efficient approach for determination of incoming direction of electromagnetic (EM) signals radiated from multiple stochastic sources in far-field is presented in this paper. The approach is based on using a neural model realized by the Multi-Layer Perceptron (MLP) artificial neural network. MLP neural model, successfully trained by using correlation matrix of signals sampled by receiving antenna array, can be used to accurately determine a direction of arrival (DOA) of radiated EM signals and afterward a location of each of multiple stochastic sources in azimuth plane. Presented model is suitable for real-time applications as it performs fast the DOA estimation. The model architecture, results of its training and testing as well as simulation results are described in details in the paper.

Neural network model for efficient localization of a number of mutually arbitrary positioned stochastic EM sources in far-field

An efficient direction of arrival (DOA) estimation of multiple electromagnetic sources by using artificial neural network (ANN) approach is presented in the paper. Electromagnetic sources considered here are of stochastic radiation nature, mutually uncorrelated and at arbitrary angular distance. The approach is based on training of the ANN in which the calculation of correlation matrix in the far-field scan area is done by using the Green function and the correlation of antenna elements feed currents used to describe stochastic sources radiation and then mapping this matrix to the space of DOA in angular coordinate. Once successfully trained, the neural network model is capable to perform an accurate DOA estimation within the training boundaries. Presented example verifies the accuracy of the proposed neural network model.