Experimental verification of an ann‐based model for 2D DOA estimation of closely spaced coherent sources

Abstract

This article presents development and validation of an artificial neural network (ANN)‐based model for accurate two‐dimensional direction of arrival estimation of two coherent sources. The ANN model is developed using simulated signals and its performance is evaluated by measurements inside an anechoic chamber. Simulations and experiments show good agreement in estimating both azimuth and elevation angles of coherent sources. The proposed model demonstrates ability to separate two closely spaced sources and outperforms MUSIC with spatial smoothing preprocessing in the case of small number of signal snapshots. © 2014 Wiley Periodicals, Inc. Microwave Opt Technol Lett 56:2558–2562, 2014

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

“…Better accuracy of neural models in comparison with the root MUSIC model under noisy conditions can be explained not only with the fact that neural models were trained by using the first row of the correlation matrix without the autocorrelation element but also with the fact that when the signal sources are mutually very close, then the super‐resolution algorithm make significantly higher errors than neural models as shown in Reference .…”

“…In References and , neural models based only on the radial basis function (RBF) neural network were suggested, in References RBF neural model from References and was enhanced with the introduction of modular structure and division of the monitoring space into a number of smaller sectors, while in Reference , a hybrid RBF‐AML model for the DoA estimation was presented aiming to combine good properties of ML method and pure RBF approach. In all these references, neural models were applied for the DoA estimation of deterministic signals radiated from mobile electromagnetic (EM) sources.…”

“…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.…”

“…Therefore, some researches have been focused to adapt some of the previously mentioned super‐resolution algorithms, based on time‐consuming inverse matrix calculations and Eigen‐value decompositions, to the strict real‐time demands . Artificial neural networks (ANNs) represent an another direction of research for solving the DoA problems in real‐time . 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 and Reference , respectively.…”

Efficient sparse representation algorithm for accurate DOA estimation of multiple targets with single measurement vector

“…Direction‐of‐arrival (DOA) estimation has become one of the important topics in the array signal processing field for radar, sonar, and so forth . There have been some effective techniques for DOA estimation of multiple targets as subspace‐based algorithms including multiple signal classification (MUSIC) and estimation of signal parameter via rotational invariance (ESPRIT) techniques.…”