分数低阶矩的信号盲分离方法  

Blind signals separation method based on fractional lower order moments

作　　者：张安清[1,2] 邱天爽[1] 章新华[2] 

机构地区：[1]大连理工大学电子与信息工程学院 [2]海军大连舰艇学院信号与信息研究中心

出　　处：《通信学报》2006年第3期32-36,共5页Journal on Communications

基　　金：国家自然科学基金资助项目(60172072;60372081)~~

摘　　要：具有尖峰脉冲特性的α稳定分布信号或噪声不存在二阶及二阶以上的统计量,传统基于二阶统计量(SOS)或高阶统计量(HOS)盲分离和独立成份分析(ICA)算法效果不理想,甚至失效。对脉冲噪声环境下的盲分离技术进行了研究,推导出一种基于分数低阶矩的信号盲分离算法,并进行了仿真检验,结果表明所提出的方法对实现α稳定分布信号或含有α稳定分布噪声的信号盲分离效果很好,并具有良好的韧性。The impulsive signals or noise with a-stable distribution are characterized by the nonexistence of the finite second or higher order statistics. The traditional blind signals separation （BSS） and ICA approaches were poor performance based on second-order moment or higher-order statistics, even more invalidation. Blind signals separation technique in the presence of impulsive noise was investigated, and a new replace BSS algorithm based on fractional lower order moments was proposed. The algorithm was verified by computer simulation. Simulation results indicate that the new method has good performance to separate signals or noise with a-stable distribution, and very robust.

关 键 词：分数低阶矩 Α稳定分布 盲分离 独立成份分析 

分 类 号：TN911.7[电子电信—通信与信息系统]