Research On Interference Identification Technology Of Frequency Hopping Communication

In modern military affairs,the army is becoming more and more electronic,and it is increasingly dependent on the security and stability of communication systems.The confrontation of communication technology also occupies a major battlefield.In electronic warfare,the enemy often carries out active man-made interference to influence or destroy our communication system,which puts us in a disadvantageous position in military command and communication.The accuracy,safety and rapidity of communication are facing great challenges.To avoid and confront active human interference effectively,the premise is to identify the type of interference.Frequency hopping communication has good ability of anti-fading,anti-interference,antieavesdropping,etc.It is one of the main ways to counter electronic interference in electronic warfare and also a research hotspot.Therefore,in order to further improve the anti-interference ability of frequency hopping communication,thesis studies the identification of interference signals.The main contents include.Firstly,the principle of frequency hopping system and active interference are studied.The mathematical models of frequency hopping system,band-stop jamming,single-tone jamming,multi-tone jamming,pulse jamming,linear sweep jamming and follower jamming are established,the main interference parameters are analyzed,the frequency hopping signals under various interference are simulated,moreover the time-domain and frequency-domain characteristics of the interference signals are studied.Secondly,the characteristic parameters of energy,time domain peak-to-average ratio,residual energy,frequency domain kurtosis,spectrum flatness standard deviation and fractional Fourier transform energy concentration are extracted from the interference signal in each transform domain,moreover the performance of each characteristic parameter under different SINR is obtained by simulation.And establish interference signals and their corresponding feature libraries to prepare data sets for classification.Thirdly,interference recognition is carried out by using support vector machine,decision tree and convolutional neural network classifier.The principle of each classifier is studied,the classifier is built,and the classification performance of each classifier under different SINR is tested under generalized data set and targeted data set,and the advantages and disadvantages of different classifiers in research cost and classification are summarized.Fourthly,a multi-classifier interference classification method is proposed.Through experiments,the influence of multi-classifier classification method on recognition rate of support vector machine,decision tree and convolutional neural network is verified.Both support vector machine and convolutional neural network have been improved obviously.Fifthly,an interference classification method based on feature fusion is proposed.By debugging different feature fusion methods and adjusting the number of feature fusion,the performance of feature fusion method is improved compared with that before fusion,and the recognition accuracy can reach more than 95% under 8d B signal to interference ratio.In thesis,typical interference classification methods from feature extraction to classifier are studied,and good classification results are obtained.Furthermore,a new idea of multi-classifier and feature fusion is proposed.Experiments show that the method is effective and has high application potential in other classification and recognition scenarios.