应用于声反馈抑制的变步长归一化子带自适应滤波算法  

A variable step size normalized subband adaptive filters for acoustic feedback cancellation

作　　者：郑洋 唐加能 ZHENG Yang;TANG Jianeng(College of Engineering, Huaqiao University, Quanzhou 362021, China;College of Mechanical Engineering and Automation, Huaqiao University, Xiamen 361021, China;Centron Communications Technologies Fujian Co., Ltd., Quanzhou 362021, China)

机构地区：[1]华侨大学工学院,泉州362021 [2]华侨大学机电及自动化学院,厦门361021 [3]福建先创电子有限公司,泉州362021

出　　处：《应用声学》2018年第3期356-364,共9页Journal of Applied Acoustics

基　　金：国家自然科学基金项目(61573004);福建省教育厅项目(JA15035);泉州市科技项目(2014Z103;2015Z114)

摘　　要：针对自适应滤波算法中稳态失调量和收敛速度之间的矛盾,提出了一种新的变步长归一化子带自适应滤波算法。该算法在系统噪声抵消原理的基础上,用迭代收缩的方法估计得到无噪先验子带误差的功率,对每个子带步长进行更新。对所提出的算法进行数学分析,可以得出该算法是稳定的和收敛的。在长回声路径和短回声路径两种情况下,将该算法应用于助听器声反馈抑制系统中。相对于其他归一化子带自适应滤波算法,仿真实验表明,所提算法实现了更快的收敛速度,获得了更低的失调量。A new variable step size normalized subband adaptive filtering algorithm is proposed to resolvethe conflict between steady-state misalignment and convergence rate in adaptive filter algorithms. Based onthe principle of cancellation of system noise, the algorithm estimates the power of the noise-free subband priorierror by using the shrinkage denoising method, and updates each sub-band step size. Through mathematicalanalysis of the proposed algorithm, it can be concluded that the algorithm is stable and convergent. In the caseof long echo path and short echo path, the algorithm is applied to the hearing aid acoustic feedback cancellationsystem. Compared with other normalized sub-band adaptive filter algorithms, the simulation results illustratethat the proposed algorithm achieves faster convergence rate and obtains lower misalignment.

关 键 词：变步长 子带自适应滤波 迭代收缩 声反馈消除 

分 类 号：TN912.3[电子电信—通信与信息系统]