Direction of arrival estimation in multipath environments using deep learning

Abstract

This article aims to present a novel direction of arrival (DOA) estimation strategy in multipath environments using deep learning. Eigen decompositionbased algorithms, such as multiple signal classification (MUSIC), have highresolution DOA estimation performance, but they fail to work in the case of coherent signals. These algorithms require extensive computation and are difficult to implement in real time. Neural networks (multilayer perceptron [MLP] and radial basis function neural network [RBFNN]) are also applied to DOA estimation problem, and they are found to be faster than the conventional techniques, but they fail to ensure the desired accuracy in multipath environments when the training data set contains a huge number of samples and the number of incident signals was unknown. To enhance the DOA estimation performance, an efficient convolutional neural network (CNN)-based smart antenna is proposed. This smart antenna is composed of a uniform linear array (ULA), an intelligent DOA estimator, and an efficient adaptive beamformer, and the space is decomposed into five space sectors. The intelligent DOA estimator contains six CNN networks. One network is used as a classifier to select one or more space sectors, and five networks are used to calculate the DOAs of unknown number of coherent or noncoherent incident signals that received in the selected space sectors. The simulation results demonstrate that the proposed DOA estimator enables reliable DOA estimation despite very challenging multipath environment, and the CNN substantially reduces the CPU time for the DOA estimation computations especially when the number of incident signals is large.

Improving the Accuracy of Direction of Arrival Estimation with Multiple Signal Inputs Using Deep Learning

“…DL models have a superior predictive performance, and once trained, the network requires only simple inference to complete estimation tasks [12,13]. Current DL-based DOA estimation studies typically use the received signals and their processed forms as network inputs, with the outputs being the corrected received signals or predicted angles [14,15]. Furthermore, DL has also been used to enhance traditional DOA estimation methods.…”

Deep Learning for Inverse Design of Broadband Quasi-Yagi Antenna

“…In addition to improving image-processing applications, the CNN can also be applied in the antenna design. Harkouss applied CNN in smart antenna design to enhance the direction of arrival (DOA) estimation performance [19]. Sahedian et al used CNN to collect spatial information from the images for plasmonic structures design [20].…”