A CNN-Based Direction-of-Arrival Estimation from Bistatic Scattering Pattern

The task of direction-of-arrival (DOA) estimation is to identify the signal source direction. Previous studies demonstrated that the strength of speckle-noise could be related to the size of the resolution cell and other factors, including wind speed and frequencies. In this paper, we consider the effect of resolution cell size in DOA estimation. The advanced integral equation model (AIEM) is applied as the forward model to simulate the radar scattering from sea surface under various observation geometries. To simulate the speckle noise, we evaluate the number of equivalent scatterers per resolution area and generate K-distributed speckle statistics. A convolutional neural network (CNN) is then utilized to establish the relationship of radar measurements with their incident directions. The experimental results show that the CNN structure could be applied to DOA estimation with satisfactory accuracy for high-resolution radar measurement.