Research On Intelligent Cognitive Method Of Communication Jamming Under Non-Gaussian Noise

Wireless communication suffers from various jammings from the outside world.In order to improve the reliability and security of the communication system,it is necessary to conduct in-depth research on the jamming in communication.Through the deep perception and cognition of communication jamming,it provides a priori information for the anti-jamming of the communication system.In addition,communication systems are often affected by non-Gaussian noise,so this paper studies the cognitive method of communication jamming under non-Gaussian noise,focusing on the detection of communication jamming and parameter estimation of communication jamming in non-Gaussian noise environment.The main work is:To solve the problem that the existing communication jamming detection methods are affected by noise and the detection performance is unstable,a communication jamming signal detection method under non-Gaussian noise environment is proposed.The method first uses a fast myriad filtering algorithm based on fixed-point search to suppress non-Gaussian noise.Then,based on the memory of the long-short-term memory network and the regularity of the time-domain amplitude of the communication jamming signal,the long-short-term memory network is trained,and the test statistic is designed through the error between the network output predicted value and the actual value of the signal amplitude.Finally,the trained network is used to input the noise multiple times to obtain the detection threshold,thereby realizing the detection of communication jamming signals under non-Gaussian noise.The simulation results show that the method significantly reduces the influence of noise on the detection performance.To solve the problem that the existing bandwidth estimation methods have poor estimation effect under the condition of low signal-to-noise ratio,a bandwidth estimation method of communication jamming signal in non-Gaussian noise environment based on pyramid pooling network is proposed.The method first uses weighted myriad filtering to suppress non-Gaussian noise.Then the spectrum is scanned to obtain the rough estimated bandwidth of the communication jamming signal,and the frequency band is divided to obtain the bandwidth sub-intervals,so as to obtain the corresponding relationship between the communication jamming signal spectrum and the bandwidth sub-intervals.The bandwidth estimation problem is transformed into a classification problem,and the classification and estimation of bandwidth is realized through the pyramid pooling network.When the generalized jamming-to-noise ratio is 0d B,the estimation accuracy of the bandwidth sub-interval of communication jamming reaches 100%.To solve the problem that the existing center frequency estimation methods have poor estimation effect under the condition of low signal-to-noise ratio,a center frequency estimation method of communication jamming signal under non-Gaussian noise based on convolutional neural network is proposed.The method first uses weighted myriad filtering to suppress non-Gaussian noise.Then,the Fourier transform value and the point number position corresponding to each transform value are combined to obtain the feature matrix of the input deep convolution network network,and the output of the network is the point number position corresponding to the estimated center frequency.Finally,the estimated center frequency is calculated from this point position.The simulation results show that when the generalized jamming-to-noise ratio is greater than 0d B,the estimation error of the center frequency of the communication jamming signal is lower than0.5×10^-4.To solve the problem that the existing power estimation methods have poor estimation effect under the condition of low signal-to-noise ratio,a power estimation method of communication jamming signal under non-Gaussian noise based on deep neural network is proposed.This method calculates the covariance matrix of the signal after the weighted myriad filter suppresses the non-Gaussian noise,and extracts the part of the covariance matrix that only contains the signal power through singular value decomposition.The reconstructed matrix is obtained as the input data of the deep residual network,and the output of the network is the estimated power.Simulation results show that the proposed method performs better than existing estimation methods at low signal-to-noise ratios.