应用于声反馈抑制的变步长归一化子带自适应滤波算法*

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 variable step size normalized sub-band adaptive filters for acoustic feedback cancellation

Abstract: A new variable step size normalized sub-band adaptive filtering algorithm is proposed for the confliction between steady-state misalignment and convergence rate in adaptive filtering algorithm. Based on the principle of system noise cancellation, the method of shrinkage denoising was used to obtain the power of the noise-free sub-band priori error, and the steps of each sub-band are updated. Through the mathematical analysis of the proposed algorithm, it can be concluded that the algorithm is stable and convergent. In the case of long echo path and short echo path, the algorithm is applied in acoustic feedback cancellation in hearing aids. Compared with other normalized sub-band adaptive filtering algorithm, the simulation results show that the proposed algorithm achieves faster convergence rate and obtains lower misalignment.