1D Localization of Mobile Stochastic EM Sources with Variable Radiated Power Using Two-Stage Neural Model

Direction of arrival estimation of mobile stochastic electromagnetic sources with variable radiation powers using hierarchical neural model

“…In all these references, neural models were applied for the DoA estimation of deterministic signals radiated from mobile electromagnetic (EM) sources. Multilayer perceptron (MLP) neural networks were also used for the DoA estimation of deterministic and stochastic EM sources . Regarding the stochastic scenario, in References uncorrelated mobile stochastic sources were considered and it was shown that only the values of elements of the first row of spatial correlation matrix at the receiver were sufficient to develop an accurate neural model.…”

“…All models presented in References were developed with an assumption that stochastic sources are with equal and constant radiation power during their movement. The architecture of the neural model that includes the possibility to perform the DoA estimation of signals emitted from movable stochastic sources with variable radiation powers was given in Reference . It has allowed to model the variation of radiation power of stochastic source through the changeable radiation diagram of antenna elements used to represent the stochastic source.…”

“…Two‐stage concept in the development of the neural model for the DoA estimation, presented in initial form in Reference , is here further enhanced leading to the hierarchical neural model for the DoA estimation of mobile uncorrelated stochastic sources of variable EM radiation. This hierarchical neural model, unlike the model in Reference , does not include in its architecture the impact of autocorrelation elements of the spatial correlation matrix as these elements are more exposed to the noise of the receiving antenna, in comparison with other correlation matrix elements (except in the case when antenna noise is not correlated). Therefore, the proposed hierarchical neural model can be used in real conditions where noise is also present during the signal acquisition at the reception.…”

“…Therefore, the proposed hierarchical neural model can be used in real conditions where noise is also present during the signal acquisition at the reception. In addition, this model considers the more complex variant of the problem analyzed in Reference . It is capable to perform the DoA estimation in cases when sources of stochastic radiation are moving along the arbitrary path with arbitrary direction in the azimuth plane while the model in Reference assumes that the sources are limited to two‐way movement along the straight line.…”

“…1,[10][11][12] Artificial neural networks (ANNs) [13][14][15][16] represent an another direction of research for solving the DoA problems in real-time. [16][17][18][19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34] Beside ANNs, the more general methods in the field of artificial intelligence, such as genetic programming (GP) and support-vector machines (SVM), are used in today research for the same purpose, as demonstrated for the DoA estimation problem solving in Reference 35 and Reference 36, respectively. Although the GP and SVM-based DoA models have more generalization capabilities than the ANN models in modeling highlynonlinear dependences present in DoA problems, DoA ANN models still represent a good alternative.…”