基于峭度的FastICA改进算法

An Improved Kurtosis FastICA Algorithm

独立分量分析(ICA)是信号处理技术的新发展,而FastICA是独立分量的一种快速算法,因其收敛速度快而备受关注,但存在步长μ选取不当可能导致算法收敛速度减慢甚至不收敛的问题,本文提出了一种改进的优化学习算法,在牛顿迭代方向上增加精确线性搜索,从而使得算法的收敛性不依赖于μ的人为选择.将改进的FastICA算法应用到语音信号处理中,结果表明该方法迭代次数大大少于FastICA算法,具有收敛速度快的特点.

Independent component analysis（ICA） is a new development of signal processing technology.FastICA is a fast algorithm of ICA.As its fast convergence,it has attracted broad attraction,but if step-size μ was chose incorrectly,the algorithm may be having slow convergence even no convergence.To overcome the drawbacks and improve the learning algorithm,exact line search was imposed on the direction of Newton iterative.The improved algorithm can ensure the convergence of the results and is robust to μ.When the improved algorithm is used to separate audio signal,the experimental results show it has fast convergence and is robust to outline.