基于ICA的相位编码信号欠定盲分离算法

摘要

A method for the mixing matrix estimation and source recoveration in the underde-termined source separation is proposed in which the shaped phase coded signals are not sparse in time domain or frequency domain. By means of particularity among the shaped phase coded signals, a method is presented, which uses tapped delay lines to turn an underdetermined con-volutive blind source separation problem into an overdetermined instantaneous blind source separation problem which can be solved by independent component analysis (ICA) algorithm. Furthermore, the sources can be estimated by using kurtosis choosing and least mean square (LMS) algorithm precisely. The method can overcome the difficulty that the sources are not sparse enough to estimate the matrix and recover signal in time or frequency domains.%针对二相编码信号时域或频城上不充分稀疏的情况,提出了欠定盲源分离中估计混合矩阵和恢复源信号的新方法.首先,利用二相编码信号成型模型的特异性,将欠定盲分离问题转化成卷积盲分离问题,然后通过抽头延时将其转化为线性瞬时混叠问题,通过独立分量分析(ICA)方法对延时后的观测信号进一步处理.为了准确地分离出源信号,利用峭度准则对聚类后的ICA分量两两分组,最后利用最小均方误差(LMS)算法将分组后的ICA分量和观测信号对消得到源信号.该算法克服了欠定条件下时域和频域中二相编码信号无法利用稀疏性进行盲源分离的困难.仿真结果验证了算法的有效性.