多带结构自适应滤波器的变步长矩阵设计

Variable step-size matrix adaptive filter design based on the multiband structured

为提高归一化子带符号自适应滤波算法的收敛性能,提出了一种基于多带结构的变步长矩阵归一化子带符号自适应滤波设计,该设计方法以后验误差向量的能量为代价函数,求解每次经迭代计算后,使代价函数最小化的各子带最佳步长获得变步长矩阵,并利用滑动平均法给出下次迭代所需的变步长矩阵.无论有无相关信号或是脉冲噪声输入,该算法较传统归一化子带符号自适应滤波算法和已有的一些改进算法相比,具有收敛速度快和稳态失调低,且计算复杂度低的优点.

In order to improve the convergence performance of the normalised sign subband adaptive filter （ NSSAF） , variable step‐size matrix normalised sign subband adaptive filter （ VSSM‐NSSAF） based on the variable multiband structure is proposed . In this method , the energy of the subband posteriori error vector is treated as the cost function and the variable step size matrix as a variable . By solving the optimal step length for each subband to minimize the cost function after each iteration calculation , the best variable step size matrix is obtained . Then we employ the following time‐average scheme to get the variable step‐size matrix before the next iteration . Compared with the traditional NSSAF algorithm and other improved NSSAF algorithms , the proposed algorithm provides a faster convergence speed and lower steady state misadjustment , regardless of whether the input signal is relevant or not , even if it contains impulse noise . Besides , this method has the advantages of a low computational complexity .