船舶自动识别系统的单通道信号分离优化研究

摘要

在星载AIS信号的单通道分离研究中,针对传统EMD算法中存在的端点效应问题,提出了基于改进EMD算法和KL散度相结合的单通道分离方法.首先采用改进EMD算法将观测信号分解成一系列频率不同的本征模态函数(IMF);然后根据KL散度(相对熵)的值,提取IMF分量中真实部分,将提取出的IMF分量组成新的多维信号;最后利用FastICA算法实现星载AIS信号分离.方法克服了端点效应,去除了IMF分量中的虚假部分.仿真结果表明,上述算法可以实现多路信号混合下的单通道分离,并且信号相对频偏越大,分离效果越好.%In the study of single-channel separation of AIS signals,a method based on improved EMD algorithm and KL divergence is proposed to improve the end effect problem of the traditional EMD algorithm.The improved EMD algorithm is adopted to decompose the observed signal into a series of different frequency intrinsic mode functions (IMF).Then appropriate IMF components are selected by calculating KL divergence values and a new multi-dimensional signal is formed by these extracted IMF components.Finally,the FastICA algorithm can be applied for multi-channel signal separation.The method overcomes the end effect and removes the false part of the IMF components.The results show that the proposed algorithm can achieve single-channel separation with some mixed signals,and the larger frequency offset is,the better the performance is.