盲源分离算法及其在雷达信号分选中应用.pdfVIP

摘要 摘要 盲源分离是指在未知源信号和传输信道先验信息的情况下，根据输入源信号 的统计特性，仅由观测到的混合信号来恢复或估计出源信号。盲源分离涉及到统 计信号处理、信息理论以及人工神经网络等理论，现已成为众多领域的研究热点。 本文重点研究盲源分离算法以及在雷达信号分选中的应用，主要工作概括如下： 首先，研究了瞬时混合盲源分离算法，包括独立分量分析法和联合对角化方 法。通过仿真实验，验证了各个算法的有效性及其适用条件，并比较了算法的分 离性能和实时性。 其次，研究了卷积盲源分离算法，包括时域和频率方法。重点探讨了频域卷 积混合中的排序不确定性，针对逐次排序算法的不足，提出了一种基于区域划分 的排序方法，并根据相邻频率点之间的相关性，估计错误分离频率点的源信号。 最后，针对常规雷达信号分选方法的不足，研究了基于盲源分离的雷达信号 分选算法。对各种线性调频、相位编码、非线性频率调制等 LPI 雷达信号混合模 型，进行信号分离仿真实验，验证了算法的有效性。 关键字：盲源分离 瞬时混合 卷积混合 排序模糊 雷达信号分选 Abstract Abstract Blind source separation (BSS) is an approach to restore or estimate the original source signals only using the information of the mixed signals observed in each input channel with the original source signals and input channels unknown. Involving theories like Statistical Signal Processing, information theory and Artificial Neural Network etc, BSS has become one of the key topics in many fields. This thesis describes in detail the algorithms of BSS and studies their application in radar signal sorting. Contents can be summarized as follows: This thesis will study the algorithm of instantaneous mixing BSS, including Independent Component Analysis (ICA) and the method of Joint Diagonalization. In this thesis, the validity and conditions for application of each algorithm will be verified through simulation experiment; a comparison of separation and real-time performance of algorithms will be conducted at the same time. This thesis will make a research on convolutive BSS