Altitude Measurement of Low-Angle Target Using Beamspace Based on Deep Neural Network

In the low elevation angle height measurement problem of a very-high-frequency (VHF) radar, beamspace (BS) processing can reduce the computational complexity of classical angle estimation algorithms and lower the complexity of the radar system through dimensionality reduction. However, when the target is at low altitude, the angle measurement performance of classical BS direction of arrival estimation (DOA) algorithms will decrease due to the influence of strong coherent multipath signals. Therefore, we proposes a beamspace angle estimation algorithm based on deep neural network (DNN). The algorithm uses deep neural network to learn the potential features of beam domain signals and reconstructs the BS covariance matrix by enhancing the direct wave signal features. The simulation results and measured data show that compared with general classical physics-driven algorithms, this method not only reduces the complexity of radar systems, but also has higher DOA estimation accuracy and good adaptability to array defects.