Research On Fpga Implementation Of Malicious Jamming Signal Detection System

In recent years,driven by the application demand of the Internet of everything,wireless communication technology has developed rapidly.The accompanying electromagnetic environment is also increasingly complicated,which brings unprecedented challenges to the reliability and effectiveness of information transmission.Among many jamming threats,artificial malicious jamming signal will seriously damage the system transmission performance,in addition to the "hard" anti-jamming,the antijamming means of communication system are all based on the detection of jamming signals.Therefore,this thesis studies the typical malicious jamming signal recognition and parameter estimation technique and its FPGA implementation scheme,on the basis of the malicious jamming signal detection system is constructed.The main research content of this thesis includes the following four parts:Firstly,aiming at the problem of reliable recognition of jamming signals,two jamming signal recognition algorithms are studied based on BP Neural Network and Convolutional Neural Network(CNN).First of all,three signal preprocessing methods,including DC removal,power normalization and windowing,are studied to improve the accuracy of jamming detection.Then a BP neural network based on artificial extraction of seven jamming features is designed.Aiming at the problem of fuzzy feature selection criteria,a CNN was designed to automatically extract jamming features,and two modular structures were integrated to enhance the classification ability of the network.The simulation results show that the jamming signal recognition algorithm based on CNN automatic feature extraction requires less sampling points than BP neural network,and the recognition performance is better.Secondly,aiming at the problem of efficient estimation of jamming signal parameter,the parameter estimation algorithm of jamming signal is studied based on the different characteristics of three kinds of signals.A parameter estimation algorithm based on Forward Consecutive Mean Excision(FCME)is studied for time-frequency domain continuous jamming signals,and the FCME algorithm is optimized to facilitate FPGA implementation.For tonal jamming signals,the algorithm of tonal frequency estimation based on DFT transform and the algorithm flow of multi-tone jamming frequency point estimation are presented.Aiming at Linear Frequency Modulation Jamming(LFMJ),a method of parameter estimation of multi-segment continuous sampling sequence is proposed to solve the problem that the sampling time is less than one sweep period and the sampled signal crosses the period.The simulation results show that the proposed parameters estimation algorithms have good performance.Thirdly,Aiming at the FPGA implementation of the jamming signal recognition and parameter estimation algorithm,the implementation scheme of the key algorithm is studied.For BP neural network,an approximate sigmoid activation function of 15 broken lines is proposed.In Register Transfer Level(RTL)design,neurons can be activated only by shifting and splicing.For CNN,the basic unit of CNN is realized by using the advantage of FPGA parallel computing,and the serial reuse and pipeline processing are carried out to realize complicated CNN with low resource consumption.The FCME algorithm is improved,no longer need to iterate the decision threshold repeatedly.Fourthly,based on Xilinx XC7K325 T FPGA chip hardware platform,the circuit test of jamming signal recognition and parameter estimation algorithm is completed,and the hardware resource consumption,measured performance,processing delay and other key performance indicators are given.The results show that compared with the simulation performance,the measured performance is less than 2d B back,meeting the application requirements.Compared with the simulation performance,the Jamming-to-Noise Ratio(JNR)of the measured performance of jamming recognition is no more than 2d B,and the measured performance of jamming parameter estimation is close to the simulation performance,which meets the application requirements.