A Deep Learning-Based Adaptive Beamforming Method for Measured Shortwave Signals

Shortwave communication has been widely applied in various fields due to its long-distance transmission capability without the need for relays, as well as its advantages of low cost and high reliability. However, in long-distance transmission scenarios, factors such as background noise, electromagnetic interference, and multipath effects significantly degrade the quality of the received signal, leading to reduced signal-to-noise ratio (SNR) and limited communication performance. To address these challenges, this paper proposes a Measured Shortwave Signals Adaptive Beamforming (MSS-ABF) method based on deep learning, aiming to enhance spatial signal quality through adaptive processing of array signals. The proposed approach tackles the limitations of traditional beamforming techniques in weight vector design and nonlinear modeling capability.Specifically, the method adopts an end-to-end architecture integrating time-domain convolutional neural networks (CNNs) to improve nonlinear representation ability, incorporates pooling layers and attention mechanisms to enhance feature extraction and accelerate model convergence, and introduces a custom loss function inversely proportional to the output SNR, which transforms the original SNR maximization objective into a minimization task suitable for gradient-based optimization. Simulation and real-world shortwave signal experiments demonstrate that the proposed method significantly improves the SNR of array outputs and effectively suppresses directional interference, enabling adaptive enhancement of practical shortwave array signals.