自适应实时型复值混合信号盲分离算法研究

利用非圆信号的协方差和伪协方差都为非零对角线矩阵的特点,构造代价函数,通过复值梯度推导得到一种针对复值混合信号的自适应实时算法.该算法能分离任何包含非圆信号源的复值混合信号;与强不相关变换算法相比,该算法结构简单,具有实时性,分离误差小,同时不需要特征值和奇异值分解,即使信源谱系数相同也能成功分离出复值混合信号.实验仿真证明了算法的有效性.

Research on adaptive real-time blind separation algorithm of complex mixed signals

Because the covariance and pseudo-covariance of non-circular signals are nonzero diagonal matrixes,the contrast function is constructed via the property.A new adaptive real-time algorithm for complex mixed signals is proposed through complex grads deduction.This algorithm can separate any mixed complex-value signals including non-circular signals.Compared with the strong-uncorrelating transform algorithm,it has the real-time feature,simple structure and less error.Moreover,eigenvalues and singularvalue decomposition are unnecessary,especially when the spectral coefficients are the same,the complex mixed signal can be still separated.Simulation results prove the efficiency of the proposed algorithm.