Abstract: It is well known that the convergence rate and steady-error are crucial performance indexes for sequential Blind Source Separation（BSS） algorithms. In order to speed up the convergence rate and improve tracking ability, it proposes a novel adaptive step-size BSS algorithm based on Nonlinear Principal Component Analysis（NPCA）. The proposed algorithm utilizes an adaptive step-size whose value is adjusted in sympathy with the time-varying dynamics of the input signals and the separating matrix. Simulation results show that the proposed algorithm has faster convergence rate and better tracking ability compared with existed NPCA algorithm.

Key words: blind source separation, adaptive step-size, nonlinear principal component analysis

摘要： 收敛速度和稳定误差是在线盲源分离算法的两个重要的性能指标。为了加快算法的收敛速度，提高算法的跟踪性能，提出一种基于NPCA的自适应变步长盲源分离算法。该算法的迭代步长随着输入信号和混合矩阵的变化而变化，因而具有更好的跟踪性能。仿真结果表明，该算法提高了NPCA算法的收敛速度和跟踪性能。

关键词: 盲源分离, 自适应变步长, 非线性主成分分析