基于独立分量分析的自适应在线算法

独立分量分析(ICA)是近几年兴起的一种高效的信号处理方法，学习步长的优化问题是自适应ICA重要的一方面，基于变步长思想，定义了一种描述信号分离状态的相似性测度，来衡量输出分量之间的相似性程度，并由此提出一种改进的自适应在线算法。根据相似性程度所反映的信号分离状态自适应调节步长,并建立学习步长和相似性测度变化量的非线性关系，克服了传统算法在信道矩阵变化时对步长自适应调整的不足。性能指标分析和仿真实验证明了算法的收敛性和稳态性能。

Adaptive on-line algorithm based on independent component analysis

ICA is an efficient signal processing method which arose in recent years, an important problem learning in adaptive ICA is opting learning step. According to variable step thinking, this paper defined similarity measure which described the state of signal separation, to measure the level of similarity between output components, and thus developed an improved adaptive line algorithm. Adjusting the learning step on the basis of traditions of degree of signal separation which was reflected by dependent measure, and established the nonlinear relation between learning step and similarity measure variation, and overcame the disadvantages of traditional algorithms in the channel variation circumstances in the process of adaptive step. Performance analysis and simulation results show that separative signal has better performance in convergence and steady.