“Localization of Stochastic Electromagnetic Sources by using Correlation Matrix Trained MLP Neural Network”

In this paper, MLP neural network-based approach is proposed for an efficient direction of arrival (DOA) estimation of multiple narrow-band electromagnetic sources of stochastic nature in far-field. Neural network is trained to perform the mapping from the space of signals described by correlation matrix, obtained by signal sampling in far-field scan area, to the space of DOA in angular positions. Accuracy and efficiency of the proposed approach is validated on two examples determining the location of one and three stochastic sources in far-field, respectively, placed at fixed angle distance.