一种变正则化矩阵的改进多带结构子带自适应滤波算法

A Variable Regularization Matrix for the Improved Multiband-Structured Subband Adaptive Filter

定正则化因子的改进多带结构子带自适应滤波(IMSAF)算法在取得收敛速度快和稳态失调误差小之间存在冲突.根据系统噪声抵消原理,设定子带后验误差功率等于子带噪声功率,本文提出了变正则化矩阵的IMSAF算法来解决这一问题.仿真结果证明,所提算法可以同时达到收敛速度快、稳态失调误差小以及追踪速度快等优势.

The conflict between fast convergence rate and small steady-state misadjustment is a problem of the improved multiband-structured subband adaptive filter（ IMSAF） algorithm with a fixed regularization factor. Based on the principle of minimizing the system noise,the power of the subband posteriori error is assumed to equal the power of the subband noise. Under this assumption a variable regularization matrix for the IMSAF algorithm is proposed to solve this problem. Simulation results showthat the proposed method is superior to the IMSAF algorithm in terms of convergence rate,steady-state misadjustment and tracking capability.