Research Of Passive Detection Signal Sorting Based On Clustering Algorithm

The essence of radar signal sorting is that it needs to isolate the pulse sequence from the large pulses which are intercepted by the passive detection system, and the pulse sequence belongs to every radar. Radar signal sorting is the part of all the passive detection system. We can only estimate the parameters of the radar signal when the sorting is right, then the parameters can be places in the library of radar so it can be used for positioning, tracking and analysis.However, with the development of technology and the need of the modern military, every parameter of the radar changes randomly, and the number of the radar which is multifunctional and multipurpose becomes lager and lager, all these things lead to that the pulse becomes more and more, which is a high requirement for the signal sorting.This paper introduces the traditional radar signal sorting algorithm, and illustrates briefly the advantages and disadvantages of each algorithm. The traditional radar signal sorting algorithms include: dynamic correlation method, histogram method(CDIF, SDIF), PRI transform method and plane transform method. Next, aiming at the shortcomings of traditional PRI sorting algorithms, the clustering algorithm is applied to the field of radar signal sorting,clustering algorithm based on grid is used for radar signal sorting of pre-processing, two algorithms are introduced fit for unknown radiation source signal sorting: the radar signal sorting algorithm based on directional similarity clustering and the radar signal sorting algorithm based on hierarchical clustering, these two algorithms and the pre-processing based on gird are simulated.The directional similarity clustering algorithm defines a cost function through the similarity measure between the clustering center and the data, then transforms the optimal clustering center for the sake of extremum problems, computes the recursive formula of the optimal clustering center through the lagrange multiplier method, start with the assumption that each point is the initial cluster center, and then iterate through the calculated iteratively repeated iteration time and time again. Hierarchical clustering algorithm classifies through defining two data points similar to each other, then changes the number of clusters by increasing the threshold, and calculates evaluation index in the process, finds the optimal clustering and gets the optimal number of clusters according to the evaluation index. The simulation results show that we can get the automatic clustering while these two kinds of clustering algorithms is applied to the radar signal sorting. There is no need to set the number of clusters and the fixed radar parameters,which will not affect the sorting accuracy of the algorithm. Whether parameters are close or notand the proximity have influence on the sorting accuracy of the algorithm. The sorting accuracy of the algorithm is very high when the parameters are not close, so the two algorithms can be applied to the radar signal sorting when the parameters are not close to others and the radar is complex.