Intelligent Identification Of Shortwave Communication Protocols

In modern warfare,information confrontation largely determines the direction of war.Battlefield communication mainly relies on short-wave communication,and short-wave communication protocol identification is the technical basis for information countermeasures and rapid civilian emergency networking.With the continuous updating of short-wave communication protocols,more complex forms of short-wave communication have emerged,which put forward higher requirements for protocol identification.Therefore,the introduction of intelligent short-wave communication protocol identification has potential application value.This paper focuses on the identification method of the short-wave communication protocol in the data link layer.In-depth study of the commonly used short-wave communication protocol data frame format,firstly established a short-wave communication protocol identification framework and workflow based on the data link layer.It includes preprocessing module,classification identification module and output module.Under laboratory conditions,it is difficult to obtain real shortwave signals.Therefore,thesis establishes a short-wave communication signal generation module through software.According to the idea of modularization,the physical layer and data link layer of five commonly used shortwave communication protocols are realized,and configurable multipath interference,mutation and noise interference are added in the shortwave channel to simulate the real channel environment.The datasets of training signal and test signal are generated by the signal generation module.In the preprocessing module,direct frame segmentation of the data stream will cause some data frames to be empty.Thesis uses the frame synchronization code as a symbol to divide the data stream into frames.A N-Eclat algorithm suitable for bit stream data is proposed to realize the recognition of frame synchronization code.Test with the data generated by the signal generation module.The results show that the algorithm can accurately identify the frame synchronization code,and can eliminate the empty set phenomenon and multipath interference of the data frame.In the classification recognition module,the parameters and recognition performance of Res Net18,VGG16,Google Net,and Alex Net were analyzed and compared.Res Net18 has a small number of parameters and a high recognition rate.Therefore,a shortwave communication protocol identification scheme based on residual neural network is adopted.Classification and identification of shortwave communication protocols using Res Net18 network.Test with the data generated by the signal generation module.The simulation results show that the recognition rate can reach 92.5% under the condition of 0 d B SNR.The feasibility of using neural network to identify shortwave communication protocols in the data link layer is verified.Considering that in the actual application environment,the data that can intercept short-wave signals per unit time may be relatively small,and at the same time,considering the limited storage resources and computing capabilities of hardware devices.Using the transfer learning method,by adding a pre-trained network to identify shortwave signals,the requirement for the amount of data per unit time is reduced.At the same time,the lightweight neural network Squeeze Net is used for classification and recognition.It reduces the number of parameters of the neural network and the demand for hardware storage and computing power.The experimental results show that under the signal set of small samples generated by the signal simulation module,the recognition rate of Squeeze Net for the five protocols can reach 91.4% when the signal-to-noise ratio is 0 d B.Squeeze Net based on migration learning can reduce the sample size,storage resources and computing power required by the model with a slight decrease in accuracy.