复值信号盲分离固定点迭代算法

复值非高斯最大化算法是一种针对任意复值非高斯混合信号的自适应盲源分离算法,算法需要人为设置步长调节参数.在信号源和信道参数未知的条件下,很难选择合适的步长调节参数,步长过小,则收敛过慢,步长过大,则容易发散.针对这一问题,在该算法中引入归一化惩罚函数构造出新的代价函数,直接对复值变量的代价函数求导优化,得到2种新的不需要步长调节参数的固定点复值混合信号盲源分离算法.该算法不但可以同时分离常态和非常态信号,而且计算量小,分离误差小,算法稳定时需要的采样点数少,尤其是不需要设置步长调节参数,使该算法更加具有实用价值.仿真实验验证了算法的有效性.

Fixed-point algorithm for blind separation of complex-value signal sources

Complex maximization of nongaussianity was developed as a self-adaptive blind source separation algorithm for complex-value nongaussianity mixed signals,which have needed set-up of step parameters.Given the conditions,it was difficult to choose step parameters: sources and channels were unknown.Smaller steps decrease convergence speed while larger steps cause divergence.Due to this problem,a new cost function was constructed by adding a penalty function to the algorithm to optimize the cost function that ha...