一种基于牛顿迭代的自适应复盲源分离算法

An Algorithm for Adaptive Complex Blind Source Separation Based on Newton Update

为了解决源数动态变化情况下的复数盲源分离的问题,首先采用一种基于交叉验证技术的源数动态估计方法;利用牛顿迭代法推导了一种自适应的复数盲源分离算法,由于在分析过程不需要对复值源信号做任何限制或约束性的假设前提,因此该算法适合于分离服从正则或非正则分布的超高斯和亚高斯信号;提出的算法通过了源数动态变化仿真实验的验证.另外,在此基础上提出了一种基于复盲源分离的波达方位动态估计法,该方法适用于在源数未知且动态变化的情况下对目标源的波达方位进行动态估计.

To solve the problems of complex blind source separation in the condition of time-varying numbers of source ,a dynamic number estimation method for complex-valued source based on cross-validation technique was proposed .In this method ,an adaptive complex blind source separation algorithm based on kurtosis of complex source is derived by using a Newton update method .Due to the innovative complex blind separation algorithm without any restrictions or constraints on source signals ,it can be used to separate sub-and sup-Gaussian sources with canonical or noncanonical distribution .The effectiveness of the proposed meth-ods was validated by simulation of time-varying numbers of complex source .Based on the experimental result ,an estimation method of dynamic direction of arrival （DOA ） was proposed to online estimate DOA orientation under the case that the number of sources is unknown in advance and time-varying .