基于独立分量分析的降噪技术

介绍了新兴的独立分量分析技术的基本概念和原理，以及具有代表性的算法，即FastICA算法、EASI算法、非线性PCA算法和基于自然梯度的最大似然估计算法。通过降噪仿真实验，并采用均方误差作为降噪的性能指数，对这些算法与传统的自适应信号处理算法进行比较。所得实验结果表明，独立分量分析算法在降噪上的效果优于自适应信号处理算法。因此在降噪上具有较大的应用价值。

The paper introduces a new technology of signal processing: independent component analysis, including its basic concept, principles, and some representative algorithms, such as FastICA, EASI, Nonlinear PCA, and natural gradient algorithm based maximum likelihood estimation. In a denoising simulation experiment with the mean square error criterion, these algorithms are compared to the classic algorithms of adaptive signal processing, such as LMS and RLS. Results show that in denoising application ICA algorithms are superior to the classic adaptive algorithms. Thus ICA algorithms have large value in denoising application, deservnig further study and promoting.