Estimation of the Number of Stochastic EM Sources with Partially Correlated Radiation in Far-Field Using Neural Model

Neural model based on multilayer perceptron (MLP) neural networks, capable to determine the number of partially correlated stochastic radiation EM sources present in a monitoring spatial sector, is proposed in the paper. Stochastic radiation sources are moving in the azimuth plane along a line trajectory. Sources are mutually correlated with the same value of correlation level but this value can change during their movement. Also, sources can enter or leave the observed sector so that their number in the sector varies. Signals coming from the sources in the monitoring sector are sampled by a uniform linear antenna array and based on this sampling a spatial correlation matrix is created. The values of correlation matrix elements are used as an input to the proposed neural model that has to estimate the number of stochastic sources present in the observed sector. The neural model presented in this paper can determine a presence of maximum three stochastic sources in the 30° wide sector while the level of mutual correlation of sources can vary between 0.05 and 0.5.