Research On Identification Of Typical Communication Jamming Signals

With the advent of the information age,the people are increasingly relying on wireless communications.For military wireless communications,the enemy often releases some hostile jamming signals to interfere with our wireless communications.If the type of the jamming signal can be accurately identified,the resulting destruction of the jamming signal may be effectively suppressed as much as possible.Therefore,the idenficaiton of jamming signal is one of the key technologies of anti-jamming communicaitons.This dissertation focused on the identification and paramters estimation of typical hostile jamming singal.In Chapter 2,the mathmatical model of typical jamming signals is first described.In order to improve the identification performance,the pre-processing,such as signal power normalization,is then introduced.Finally,the feature of the jamming signals and the corresponding extraction methods are given.In the third chapter,the handcrafted feature extraction-based jamming signals identification is addressed.The support vector machine(SVM)based decision tree and backpropagation(BP)neural network based jamming signal identification methods are discussed.The principle and mathematical models of these two methods are described.The idenfication performance of these two methods are verified with respect to the number of different training samples,different feature preprocessing methods,and the different JNR distribution of training samples.As some feature important information may be lost in handcrafted feature extraction,deep matchine learning based jamming signals recognition methods are proposed in Chapter 4.We design a convolutional neural network(CNN)structure that can be used to identify the jamming signals.We invesitgate the influence of different convolution input and JNR distribution of different training samples on the recognition performance.Moreover,the proposed CNN based methtod is compared with the handcrafed feature extraction based recognition method.Simulation results show that the proposed CNN based method significantly outperform the handcrared freature extraction method.After the jamming signal is recognized,the parameters estimation of the corresponding jamming signal is further investigated in Chapter 5.For singletone and mutlitone jamming signals,we propose a two-threshold based iterative method to estimate the number of tones,and the discrete frequency domain based method to esitmate the frequency and JNR.For linear FM jamming signal,the amplitude,linear sweep frequency and JNR are estimated by performing secondary demodulation of the jamming signal and discrete frequency domain processing.For partial-band and noise frequency modulated jamming signals,the signal power spectrum is first reconstcted and smoothed using a smoothing window,and then the occupied bandwidth is estimated based a relative threshold based method.Simulation results validate the proposed jamming signals parameters estimation methods.In Chapter 6,this dissertation is summarized and some potential researches are provide for future work.