基于峭度的一种RobustICA算法

Robust Independent Component Algorithm Based on Kurtosis

介绍了由V. Zarzoso and P. Comon提出的一种新的基于峭度的独立成分分析算法RobustICA(Robust Independent Component Analysis),并比较了和FastICA在收敛性和信号质量方面的不同。该算法的主要优点在于可以选取最佳步长,可以选取任何不为零的独立成分,并且解决盲分离信号排序问题,同时提升当信号存在坏点和伪局部极值点时的鲁棒性。仿真实验结果表明了该算法相对于FastICA算法减少了迭代次数和加快了收敛速度,同时在小样本空间下均方误差SMSE也明显优于FastICA算法。

Robust Independent Component Analysis based on Kurtosis is presented by V.Zarzoso and P.Comon is introduced in this paper,The difference in astringents and signal quality are compared with FastICA.The algorithm is the main advantage is that can optimal step-size,can choose any nonzero independent component,and solves the separation signal order,at the same time,improves the robustness when a signal is present saddle point and pseudo local extreme value point.The results show that the algorithm can reduce the iteration times and speed up the convergence speed compared with FastICA,at the same time,The SMSE of RobustICA is significantly better than FastICA in the small sample space.