基于二阶非平稳统计量的船舶噪声信号的盲分离  

The Separation of Ship Noise Signal Based on Second Order Nonstationary Statistic

作　　者：单志超[1] 林春生[2] 向前 

机构地区：[1]海军航空工程学院电子信息工程系,山东烟台264001 [2]海军工程大学兵器工程系,湖北武汉430033 [3]海军武汉军代局701代表室,湖北武汉430000

出　　处：《信号处理》2009年第6期973-976,共4页Journal of Signal Processing

基　　金：国防重点实验室基金项目资助(51444060101JB1108)

摘　　要：针对船舶噪声信号为实的非平稳信号的特点,使用二阶非平稳统计量的方法对传感器阵列所接收到的信号进行了盲分离。首先对原始信号进行了稳健正交化处理,同时根据信号的非平稳性,采用分段加窗的方法求取时滞协方差矩阵,将正交分离矩阵的求解转化为时滞协方差矩阵的联合近似对角化,并最终转化为子空间拟合问题,最后通过Gauss-Newton发进行求解。文中较为详细的给出了针对非平稳实信号的二阶统计量盲分离算法的基本步骤。最后对三艘实测的船舶噪声信号所进行的盲分离仿真试验表明,该算法具有较快的收敛速度和较好的分离效果。Due to the nonstationary of ship noise, the sensor array signals were separated blindly by second-order nonstationary statistic algorithm. Firstly, the original signals were pre-processed by robust orthogonalization algorithem, and then for the nonstationary of signals, the time-lag covariance matrixes of the pre-processed signals are gotten by segmenting and windowing. The solution to the or- thogonalizing separating matrix was transformed as the diagonalization of time-lag covariance matrixes. At last this problem became a subspace fitting question wbieh could be resolved via Gauss-Newton algorithem. The derivation of algorithm to the nonstationary real sig- nal was given. The simulation experiment for separating the three practical ship noises shows the fast convergence rate and good separa- ting effect of the algorithm.

关 键 词：二阶统计量 非平稳信号 盲分离 子空间拟和技术 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]