Research On Separation And Recognition Algorithm Of Aliasing LPI Radar Signal

Low Probability of Intercept Radar(LPI)uses power management,pulse compression technology and other technologies.It has many advantages such as anti-reconnaissance,anti-radiation strikes and anti-jamming,so it is widely used in modern battlefield.However,with the large number of LPI radar,the battlefield electromagnetic environment has become increasingly intensive.It is prone to overlap when the radar emitter signals reach the receiver.In the absence of priori information,the separation and recognition of overlapping signals has become a hot issue in the field of electronic reconnaissance signal processing.First of all,this paper summarizes the present situation of research and development of LPI radar.Based on the key technology of LPI radar,the paper analyzes the different styles of signal reach the receiver when they are overlapped.According to the diversity of LPI signal mixed mode,the paper focuses on the blind source separation of transient and convolution mixed signal models.Finally,based on the wavelet ridge,the paper analyzes the modulation recognition of single component signal after the separation.In regard of the separation of transient mixed signals,the traditional serial radar sorting algorithm has some weakness in overlapping signal separation.The idea of independent component analysis(ICA)is introduced for blind source separation processing.According to the information maximization criterion of negative entropy,FastICA is applied to aliasing signal separation.Aimed at the defect of poor performance of the algorithm convergence,the paper designs an improved algorithm for joint penalty function and one-dimensional search,which reduces the sensitivity of the initial value selection and improves the convergence of the original algorithm.In regard of the separation of convolution mixed signals,the paper follows the idea of blind source separation when dealing with convolution signals in the time domain.The mixed signal is separated based on the natural gradient algorithm,and the crosstalk error index is used to evaluate the separation effect.When the convolutional mixed signal is processed in the frequency domain,the time domain convolution signal is transformed into transient mixing of a series of frequency points in the frequency domain by STFT.For the problem of uncertainty about frequency points rearrangement,this paper transforms the long sequence of frequency points into several strong correlation signal segmentations by sparse decomposition.An improved algorithm based on joint DOA and correlation coefficient method is used to solve the problem of accuracy and reliability of frequency points rearrangement.Based on the wavelet ridge,the paper analyzes the modulation recognition of single component signal after the separation.The instantaneous frequency is an important parameter that describes the characteristics of the signal modulation.The wavelet ridge is extracted based on the instantaneous frequency by wavelet transform.By identifying the slope of the wavelet ridge,the number of jump points and the distribution of information,the wavelet ridge completes the modulation of the signal recognition.For different modulation methods of radar signals,the least squares method is used to complete the modulation of the signal recognition by fitting the straight line and calculating the characteristic parameters.The method achieves good recognition effect under different signal-to-noise ratio.