Identification of Shortwave Radio Communication Behavior Based on Autocorrelation Spectrogram Features

Table 3

Training process.

Step	Training process
Step1	send the feature matrix obtained by Table 2 to the two inputs of the network
Step2	convolutional calculations, obtain two outputs and , both of them have 256 dimensions
Step3	consider and as two new inputs, and through the add layer, is obtained,
Step4	let pass through the last few layers, the communication behavior identification is completed