基于恒模特性的单通道时频重叠信号盲分离算法  

A single-channel blind separation algorithm based on constant modulus for two time-frequency overlapped signals

作　　者：吴芳[1] 芮国胜[1] 张嵩[1] 凌青[2] 

机构地区：[1]海军航空工程学院电子信息工程系,烟台264001 [2]海军航空工程学院研究生管理大队,烟台264001

出　　处：《中国科学：信息科学》2013年第3期418-428,共11页Scientia Sinica(Informationis)

基　　金：国家自然科学基金(批准号:60572161);航空基金(批准号:20055184005)资助项目

摘　　要：针对单通道同调制方式、同调制参数时频重叠信号的盲分离问题,基于MSK信号的恒模特性,及对MSK混合信号分量幅度的准确估计,构建了单通道信号盲分离欠定方程组.为解决该方程组的解模糊问题,根据MSK信号相位连续特性,提出了相位方差最小模糊解消除准则.但相位对噪声较为敏感,该准则可能会产生严重的误判现象,又提出了一种基于最小斜率的补充准则.基于以上两种模糊解消除准则,可对单通道信号盲分离欠定方程组的模糊解进行合理选择,有效地实现了MSK时频重叠信号的单通道盲分离.仿真结果表明:在信噪比大于5 dB的条件下,该算法对混合信号具有较好的分离效果.并且该算法具有复杂度低、计算量小的优点.Blind signal separation is a challenging topic in the communication field, especially for wireless radio signal reconnaissance and monitoring. Considering a blind signal separation problem concerning co-modulated time-frequency overlapped mixed signals on a single channel, a signal blind separation algorithm is proposed based on the constant modulus and estimated amplitude of the minimum shift keying modulated signals. The blind separating algorithm is based on dissolving the underdetermined equations under a constant modulus constrained condition. Two criteria of minimum phase variance and minimum slope are presented to eliminate solution illegibility, which can reasonably lead to the right solution of the blind signal separation equations. Simulation results show good separation performance when SNR is greater than 5 dB. The algorithm minimizes computing complexity.

关 键 词：单通道 恒模特性 盲分离 模糊解 欠定方程组 

分 类 号：TN911.7[电子电信—通信与信息系统]