一种基于高阶互累计量的遗传盲反卷积算法

A Blind Deconvolution Method Based On High Order Cross Cumulants and Genetic Algorithm

提出了一种基于高阶互累计量的遗传盲反卷积算法,旨在解决现有许多基于独立分量分析盲反卷积算法中存在的两个共同缺陷.一是算法中引入的非线性函数依赖于源信号的峭度性质,当观测信号为超高斯信号与亚高斯信号的卷积混合时,算法性能急剧下降.二是算法中大多采用梯度法对分离矩阵序列进行寻优,初始值和步长的设定对搜索性能影响较大,使得寻优过程易陷入局部极优值,从而降低算法的分离效果.在阐述了算法的相关原理和设计思路之后,通过对比试验验证了算法的正确性和有效性.

The anticle introduced a kind of blind deconvolution method based on high order cross cumulants and Genetic Algorithm, which is dedicated to resolve the two common problems in most of existing BDMs based on independent component analysis. One is that the choosing of none-linear function in them depends on the kurtosis of original signals, which degrades the performance of separation seriously when observed signals are the mixture of Super-Gaussian and Sub-Gaussians signals. The other is gradient, as the optimization approach for searching separation matrixs, it has shortcomings that configuration of initial value and the length of pace will make the he separation algorithms strap into local optimum values. After describing the relevant theories and design of this new method, it shows its correctness and validity by simulated comparable experiment results.