一种新的多信号卷积混合信号盲分离算法

对于含噪声情况下多个源信号卷积混合盲分离，由于混合矩阵比较复杂，分离算法会出现迭代次数增加、收敛速度变慢等问题。在对多信号卷积混合进行合理简化的基础上，提出一种以四阶累积量为独立准则的多信号卷积混合的新的时域盲源分离算法。由于采用高阶累积量为独立准则，该算法对高斯噪声具有良好的抑制作用，改善了信噪比。其次，算法也建立了步长因子的选取与二次残差之间的非线性函数关系，使得算法既获得了较快的收敛速度，也得到较高的分离精度。仿真数据表明提出的算法对于多个源信号卷积混合具有良好的分离效果。

A new blind source separation algorithm for convolved multiple source signals

For the convolution mixture of multiple sources with noise signal, the mixed matrix is complex. The iteration number increases and convergence speed is more slow in the separation process.In this paper,a blind source separation algorithm for convolution mixture of multiple sources in time domain is proposed. This algorithm takes fourth-order cumulant as judgment criterion, so the algorithm can inhibit Gaussian white noise as the criterion of fourth-order cumulant. The nonlinear function between the step-size of the algorithm and REQ is established, so convergence rate of the algorithm is faster and separation accuracy is higher. Simulation results illustrate the good performance of this algorithm for the convolution mixture of multiple source.