基于稀疏活跃度分段的卷积频域盲分离排序

Blind Source Separation Permutation in Frequency Domain Based on Sparse Active Degree Segmentation

针对复杂电磁环境下高密度非平稳雷达信号的卷积混合问题,首先建立其频域盲分离模型,其次利用稀疏分解的思想,将混合信号分解成时频稀疏单元,再利用分离信号包络描述的单元活跃性将信号分段,综合基于到达方位角估计和相关系数法各自的优劣性,使算法在每段的准确度和鲁棒性得到最优化,以解决频域卷积盲分离频点重排问题。仿真实验从信号分离前后的相关系数与误差指标的角度对算法质量进行了评价,并提出了算法复杂度方面的不足。结果表明,该算法能够极大改善信号的分离效果,且可靠性高,适用于分离非平稳体制的复杂雷达信号。

Aiming at the high density non stationary radar signal convolution mixed problem under the complex electromagnetic environment. First of all, establish the frequency domain blind separation model, then by the use of sparse decomposition, mixed signal is decomposed into time-frequency sparce unit, again using the separated signal envelope to describe the unit activity, so the signal was segmented, and because it was based on the respective advantages and disadvantages of azimuth estimation and correlation coefficient method, the algorithm make the best optimization in the accuracy and robustness of each section for solving the frequency rearrangement problem of frequency domain eonvolutive blind source separation. Simulation evaluate the quality of the algorithm from points of view of the correlation coefficient and error indicators of separated signal and put forward the algorithm complex degree of problems. The results show that, the algorithm in this paper can greatly improve the signal separating effect and high reliability, which is suitable for separating non stationary system of complex radar signal.