Deep learning-based direction finding in the presence of direction-dependent mutual coupling

In this paper, the problem of direction finding is considered for millimeter-wave communication with direction-dependent mutual coupling. For direction-of-arrival estimation, a low-complexity deep neural network (DNN)-based regression model is proposed, taking an array output as the DNN input with a straightforward preprocessing step. In addition to designing DNN building blocks, a quantitative analysis is performed to enhance the generalization capability of the DNN trained with finite-size training data. A theoretical performance bound is derived in terms of the Cramér-Rao lower bound (CRLB). The effectiveness of the proposed approach is verified by comparison with existing methods and the CRLB.