A blind separation algorithm for complex-value signal sources
Yanfei Jia · Harbin Gongcheng Daxue Xuebao/Journal of Harbin Engineering University · 2008
Current blind source separation algorithms have limited ranges of application and their convergence rate is too slow.To overcome this,the authors developed a kind of fast blind separation algorithm that can be used to separate any non-Gaussian statistically independent complex-value signal.The object function of this algorithm maximizes kurtosis of complex signals,optimizing the object function using a complex quasi-Newtonian iterative algorithm.Circularly symmetrical signals,non-circularly symmetrical complex signals,and a mixture of these signals were used separately as sources in a simulation.The results illustrate that this algorithm,when compared to currently available algorithms,can not only separate non-circularly symmetrical complex signals and circularly symmetrical complex signals better,but also their mixed signals,and it is not necessary to set any step parameters.Its convergence rate is fast and it produces only small errors.