A Method of JTIDS Feature Extraction Based on Nonlinear Transformation of High Order Cumulant

In the view of the target dense region, the feartures of different target's time difference are not obvious, which leads to the problem of JTIDS multi-user sorting useless based on time difference. A nonlinear transformation sorting algorithm of high-order cumulant based on doppler frequency shift is proposed: First of all,make the nonlinear transformation based on the second order cumulant,the sixth order cumulant, the normalized skewness and the normalized kurtosis for single station signals respectively, achiece a sharp rise occurs in the degree of the frequency shift; Then,build the eigenvector according to the transformation;Finally,use spectral clustering algorithm to cluster the datasets. Simulation experiments were carried out,based on four targets which could not be sorted by the time difference. The result showed that compared with the traditional method,the success rate of our method was significantly improved, and the effectiveness of the algorithm was verified.