Adaptive Beamforming Algorithm Based on Residual Neural Networks
摘要
In the field of array signal processing, adaptive beamforming technology is commonly used to suppress sidelobe interference. However, when facing challenges such as an increased number of echoes, rapid changes in the angle of arrival, and significantly greater interference signal strength than the desired signal, traditional beamforming methods may encounter increased computational burdens and performance degradation. To overcome these limitations, deep neural networks have been applied to the field of adaptive beamforming. In radar array signal processing, the deep neural network adaptive beamforming algorithm has improved computational speed to some extent. However, as the complexity of the model increases, issues such as gradient explosion or overfitting may arise, and the lack of training samples under low snapshot conditions may affect the performance of the algorithm. To address these issues, this paper proposes a Residual Neural Network Adaptive Beamforming (ResNetABF) algorithm, which effectively mitigates gradient vanishing and overfitting through the use of residual blocks and skip connections, while also compensating for the limitations caused by insufficient training samples in low snapshot scenarios. Simulation experiments show that the ResNetABF algorithm can generate nulls of over