Research On Interpulse Parameter Sorting Of Radar Signal Based On Machine Learning

In radar electronic warfare,it is very important to sort the intercepted pulse signal quickly and accurately.With the addition of various complex system radars and the development of various anti-detection technologies,it becomes more and more difficult to use the traditional signal sorting algorithm.Therefore,it is very important to sort the received complex radar pulse signals.Interpulse parameter is an indispensable feature parameter in radar signal sorting.In this paper,from the point of view of radar signal pulse parameter characteristics,the machine learning algorithms such as clustering algorithm,principal component analysis and cloud model theory are deeply studied.The main research contents of this paper are as follows:(1)The basic theory of radar signal separation using interpulse parameters is studied.Firstly,various factors affecting radar signal sorting are analyzed,and the general process of sorting using interpulse parameter characteristics is summarized.Then the mathematical modeling of interpulse parameter characteristics is carried out,and some variation rules of each characteristic parameter are abstracted.Finally,some clustering algorithms used in signal sorting are analyzed theoretically,and some non-clustering but classical sorting parameter algorithms are summarized.(2)Based on the research of density peak clustering algorithm and data field,a radar signal pre-sorting algorithm based on potential distance graph is proposed.The algorithm firstly analyzes the relationship between the impact factor in the data field and the truncated distance in the peak density clustering algorithm.The optimal parameter for calculating the potential energy value of a data point is determined.After obtaining the potential energy value of the data points,the interference data points are eliminated based on the potential energy value.Then the distance attribute of data points is defined according to the potential energy,so as to construct the potential distance graph and select the clustering center to complete the clustering.The simulation results show that the radar signal pre-sorting algorithm based on potential distance graph has strong anti-interference ability and can quickly and accurately complete the clustering of each working mode of radar signal.Compared with the current clustering algorithm,it has higher efficiency and accuracy.(3)After the pre-sorting of radar signal is completed based on potential distance graph,a multi-mode radar signal sorting algorithm based on principal component analysis(PCA)and improved cloud model is proposed.In this method,PCA is used to reduce the dimension of interpulse parameters after the pre-sorting,and the new features are extracted.Based on the new features,the improved cloud model theory is used to analyze the membership degree,so as to judge which data clusters are from the same radar and which are from different radars.The algorithm realizes the unification of threshold parameters when using membership degree for similarity discrimination.It also solves the problem that the membership degree of data cluster can not be calculated due to the virtual value of hyper entropy when using cloud model to calculate membership degree.The simulation results show that the algorithm based on PCA and the improved cloud model can complete the multi-mode radar signal sorting better than the existing algorithm.It can solve the problem of "increasing-batch" of multi-mode radar signal sorting,that is sorting different modes of a radar into multiple radars.