A Cross-band Method for Intelligent Realtime Frequence Prediction of HF Communication

High-frequency (HF) communication is affected by many complex channel factors, and the communication frequency needs to be selected in real time. However, the existing frequency prediction method using neural network needs to obtain the communication data of the whole frequency band during communicating, which can't make cross-band frequency prediction based on the known frequency band data. In this paper, we propose an intelligent cross-band frequency prediction method to solve this problem. First the correlation between the data in known frequency band and full frequency band is analyzed, and a multi-laver network prediction model is built based on it. Then, the network loss function is adjusted according to the prediction demand. The neural network training process is optimized by selecting the appropriate optimizer and adding regularization methods. Finally, the validity of the proposed model is verified by analyzing the test results of real sample data.