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

Abstract

Hierarchical neural model for an efficient one‐dimensional direction of arrival (DoA) estimation of stochastic electromagnetic (EM) sources with a variable radiation power is proposed. Model is trained to provide an azimuth position of such sources based on a spatial correlation matrix obtained by a signal sampling at a reception point and then used as an input to a neural model. It consists of two hierarchical levels realized by the multilayer perceptron (MLP)‐based neural networks. The first level is responsible to reduce the dimensionality of the considered DoA problem which allows for its easier solution at the second level. Accuracy and run‐time of the proposed model is verified on an example of determining the azimuth position of two stochastic EM sources in noisy conditions through comparison with a standalone MLP model and a model based on the root MUSIC algorithm.

Advanced ANN model for DoA estimation in smart textile wearable antenna array subsystem

“…As already shown in, 26,28 only cross-correlation elements in the first row of a normalized matrix C 0 without an autocorrelation element are sufficient to determine the positions of EM sources in the azimuthal plane. The real and imaginary parts of these elements are the inputs to the MLP network rather than their complex values, due to much easier training of the neural model.…”

“…On the other hand, in the version of the MLP DoA module developed in, 30 only the change in the distance between two antenna elements was considered and the change in the gains of the antenna elements was neglected. MLP networks were chosen to implement the DoA modules since they are suitable for modeling multidimensional nonlinear regression problems, such as DoA estimation, which was confirmed in our previous research 25,26 …”

“…For its determination it is not necessary to know the gains of all antenna elements, but only the gain ratio of the i th ( i = 2, 3, …, M ) and the first antenna element, G i / G 1. The normalization of the matrix C does not change the results of the MUSIC algorithm for DoA estimation 26,29 . By introducing the variable g i —root gain ratio of the i th antenna element, as , and by specifying the initial distance between the antenna elements, d , in wavelengths, d λ , the matrix C ′ is obtained as follows: …”

“…In our previous research, we proposed to solve the DoA estimation problem for the TWAA in real‐time using artificial neural networks 22–24 . There are two reasons for this: (1) artificial neural networks (ANNs) can be trained to estimate DoA with only modest hardware platforms, and (2) once trained, ANNs are able to provide the same accuracy as super resolution algorithms but much faster 25–34 . Two basic versions of the DoA module, which is an integral part of the smart TWAA DoA subsystem with two antenna elements that perform fast DoA estimation of the RG signals and determination of the RG location in the azimuthal plane, were presented in 28–30 .…”

Source Enumeration Approaches Using Eigenvalue Gaps and Machine Learning Based Threshold for Direction-of-Arrival Estimation

“…These machine learning techniques have been applied to DOA estimation and source enumeration. The authors of [27][28][29] applied neural networks to DOA estimations, and the results showed that their neural networks based schemes can improve the performance of DOA estimations. Yang et al [30] proposed eigenvalue based deep neural networks for source enumeration, and the results showed that the proposed networks can achieve significantly better performance than the state-of-the-art methods in the low SNR regime.…”