Underwater acoustic channel prediction through cross-frequency-domain grouping and deep learning

Underwater acoustic (UWA) channel prediction has played an important role in UWA adaptive communication network and underwater environment target perception. The existing underwater acoustic channel prediction is usually carried out in the time domain, and the prediction performance will decrease when the channel presents a non-sparse structure. This paper proposes a channel prediction method through cross-frequency-domain grouping and deep learning (CFDG-DL). The proposed CFDG-DL method uses the cross-frequency coherence matrix to divide channel’s all frequency points into several groups, and each group uses a deep learning model containing several fully connected layers and long short-term memory layers for channel prediction. The proposed method can avoid the influence of channel non-sparseness and improve the prediction performance by using the correlation between frequency points. The performance of the proposed method is verified by using the KAU2 and BCH1 public at-sea experiment datasets. Experimental results show that the proposed method has lower prediction error and computational complexity than the back propagation neural network and long short-term memory network prediction methods.