基于峭度的独立分量逐次提取梯度算法

Gradient Algorithm for Sequential Extraction of Independent Component Based on Kurtosis

独立分量分析是一种有效的盲源分离和特征提取技术,在许多领域已获得成功应用。结合快速固定点算法和极大似然自然梯度算法的特点,提出了一种基于峭度的独立分量逐次提取梯度算法,编制了相应的计算程序,并设计仿真试验,试验结果表明,在信源满足独立分量分析的前提条件时,该算法具有较好的收敛性能,且分离效果较好。

Independent component analysis is an effective technology for blind source separation and feature extraction, and has been successfully applied to a wide range of researches. Firstly, the principle and realization of independent component analysis are briefly introduced in the paper. Then we put forward a gradient algorithm for sequential extraction of independent component based on kurtosis. The algorithm integrates the merits of the fast fixed- point algorithm and the natural gradient algorithm based on maximum likelihood estimation. We compile the corresponding calculation program and conduct a simulation experiment. The results have shown that this algorithm has good convergence and separation effect when the source signals could meet with the prerequisite for independent component analysis.