Direction-of-Arrival Estimation With A Vector Sensor Using Deep Neural Networks

A vector sensor is a type of sensor array with six collocated antennas that measures all electromagnetic field components of incident waves. The vector sensor has been shown to be advantageous in estimating both the angle of arrival and polarization of the incident sources. While direction-of-arrival (DOA) estimation with machine learning for linear arrays has been well studied, there has not been a similar solution for the vector sensor. In this paper, we propose a neural network that is capable of determining the number of the sources and estimating the DOA of each source based on the covariance matrix extracted from the received data. Additionally, a solution is proposed to associate the estimated DOAs to their sources, and a customized training loss function to address the field-of-view limit problem. Simulation results show that neural networks can achieve a reasonably accurate DOA estimation measured by the RMSE for up to five sources.