Time Frequency Single Source Point Detection Based Method for Radar Signal Sorting

Abstract

Common sorting method have low sorting rates and is sensitive to the Signal-to-Noise Ratio(SNR), detection of time frequency single source point is applied to sort unknown complicated radar signal, high sorting accuracy can be got. The single source points of each radar source signal is detected, then the mixing vector in the corresponding single source point set was estimated by Singular Value Decomposition (SVD), the mixed matrix is estimated simultaneously by cluster validation technique, based on k-means clustering algorithm. Finally, Pulse Description Words (PDW) of each radar signal can be worked out. Experiment results demonstrated that the radar emitter signals extracted by this method showed good performance of noise-resistance and clustering at large-scale SNR. Radar recommaissance is a key part of electromagnetic countermeasure. The mixed signal is sorted by reconnaissance system and the pulse description words of each signal pulse can be calculated to locate radiation sources or assess the threat of each source. In (1), Blind Signal Separation (BSS) was applied in radar signal sorting, a method based on fourth-order cumulant was proposed to accomplish signal sorting and achieved good results, but the number of source signals had been already known in that simulation experiment, the circumstances of unknown signal numbers was not mentioned in that paper. In (2), another blind separation method based on fixed point independent component analysis algorithm was proposed to solve the problem of overdetermined radar signal sorting, but the method can not accomplish underdetermined radar signal sorting. In order to solve the problem of unknown signal numbers in radar signal sorting by BSS, too many methods have been presented. In (3), a robust algorithm under the condition of unknown signal source number was proposed. In that algorithm, the data received by the array was pretreated via the projection transformation to inhibit model errors and reduce data dimension,thus improving the robustness and lightening the computation load. Then, the number of signal could be estimated according to the transformed m-Capon spatial spectrum function. In (4), the author took advantage of the straight line clustering of the sparse source signals, standardized the aliasing signals, then the aliasing signals were formed spherical cluster, thus the linear cluster was turned into density cluster. And the clustering center was searched and obtained by using the ant clustering algorithm, the aliasing matrix can be accurately evaluated. In a word, most methods require clustering to avoid the problem of unknown signal numbers, thus a good clustering algorithm is the key to solve this problem. The problem of underdetermined signal sorting can be settled by two methods: 1. Using the sparsity of source signals; 2. 'Two-step approach'. Both of them work under the condition of sparse signal. In the