Research On Frequency Hopping Signal Sorting Technology Based On Blind Separation

Frequency hopping communication plays an significant role in electronic warfare because of its low interception capability,anti-interference ability and"variable hop speed".Moreover,it has powerful networking ability and is one of the main technologies of spread spectrum.In the actual complex and changing electromagnetic environment,the application of frequency hopping communication brings immeasurable advantages to electronic warfare and improves the probability of victory,but it also brings some challenges to communication countermeasure technology.In the actual electronic warfare signal detection,frequency hopping signal sorting occupies an important link,because we can provide technical support to the actual combat environment from the actual sorted out frequency hopping signal,which plays a key role in achieving victory in combat,but in the actual electromagnetic environment,due to the existence of various types of interference and synchronous asynchronous network stations,only a messy mixed frequency hopping signal can be intercepted,and it is an urgent and challenging task to carry out research on its sorting technology without any a priori knowledge.To this end,combining the characteristics of frequency hopping signals,presently there are problems of weak precision and ineffective precision in using blind separation algorithm for sorting frequency hopping signals in time and frequency domains are investigated separately for the sorting of multiple mixed frequency hopping signals in time and frequency domains,respectively.（1）Firstly,the relevant tenet of blind source separation and the foundations of frequency hopping communication system theory are researched.It primarily describes the model,categorization and sparse fraction analysis of the blind source separation technique,and the structure of the frequency hopping telecommunication system,the grouping pattern of the frequency hopping signal and the module of it.In which,the short-time Fourier transform（Short-time Fourier transform,STFT）and Wegener variance distribution（Wigner-Ville distribution,WVD）of frequency hopping signals are considered in the timing frequency domain to detect the timing frequency data of frequency hopping signals.The pseudo-Wegener distribution（Pseudo Wigner-Ville Distribution,PWVD）after adding windows is further investigated,including its fundamentals,strengths and weaknesses,and enforcement.（2）Second,the module and treatment of the blind separation domain of the Independent Component Analysis algorithm（Independent Component Analysis,ICA）in the forward or overdetermined site condition are investigated.For the shortcomings of Fast Independent Component Analysis（Fast ICA）,we propose the Fast ICA algorithm（Double Loose Modified Fast ICA,DLM-Fast ICA）which is based on a double-factor refinement and is sensitive to initial values and slow to converge.Firstly,introducing two elements in the Newton iterative approach,and the optimal weight separation matrix is obtained by adaptively adjusting the combination coefficients of the separation matrix,thus improving the initial value sensitivity of the Fast ICA algorithm;Then the fast convergence property of M-Fast ICA（Modified Fast ICA,M-Fast ICA）is used to obtain the extracted signal,Which enhances the segregation efficiency and conversion rate of the algorithm;Finally,the hybrid frequency hopping signal sorting based on DLM-Fast ICA algorithm is accomplished,and the sorting performance of dissimilar algorithms is investigated by comparing and interpreting them experimentally.（3）Again,the system model of the hybrid frequency hopping signal under the underdetermined condition is studied.For the problem of poor performance of hybrid matrix estimation under the underdetermined condition,a blind splitting approach based on an improved time-frequency inspection is submitted.First,the temporal-frequency conversion is applied to the measured signal to modify its temporal-frequency map;Then,in order to obtain the single source time-frequency matrixX_X,1,the adaptive noise threshold determination method is used to judge the noise threshold under the condition of low signal-to-noise ratio.When the signal-to-noise ratio is high,the absolute azimuth difference algorithm is combined to complete the single source detection of the time-frequency matrix.and then the hybrid matrix is evaluated using the method of optimized objective function to improve the estimation accuracy.The simulation experiments show that the matrix estimation error of the proposed method is significantly reduced under the same conditions with low signal-to-noise ratio.（4）Lastly,aiming at the problem of poor sorting performance in source signal recovery in the"two-step method",the improved subspace projection method combined with the source power deviation method is used to complete the sorting.Under the condition of different interference,the number of source signals is 3,and the number of observed signals at the receiving end is 4;When the number of source signals is 3 and the number of observed signals at the receiving end is 5,the hybrid frequency hopping signal sorting experiment is carried out.Through the comparison between the hybrid matrix error and the normalized mean square error,it is verified that the proposed method has better sorting effect than the correlation algorithm under uncertain conditions,which provides a certain guiding value for the feasibility of the actual project.