鉴于声反馈冲击响应具有稀疏性，文中提出把IPNLMS算法应用于数字助听器声反馈消除，可以获得比NLMS算法更快的收敛速度。与固定步长NLMS算法一样，其收敛速度和稳态失调是一对矛盾的需求。文中提出一种新的变步长IPNLMS算法，该算法依据滤波器梯度调节IPNLMS算法的全局步长，步长随滤波器系数梯度的减小而减小，有效解决了收敛性能和稳态误差的矛盾。相比其他变步长算法，受噪声影响小，收敛过程稳定，适用于数字助听器声反馈消除。仿真实验说明本算法应用于数字助听器的声反馈消除性能比传统NLMS算法优异许多。%Due to feedback impulse response is sparse,put forward the IPNLMS algorithm used in the digital hearing aid acoustic feedback cancellation,which could get faster convergence speed than NLMS. The requirements of fast convergence and low steady-state misalign-ment are conflict as same as the NLMS. Propose a novel variable step-size algorithm which regulates step size in accordance with filter weight gradient,and a decrease in the gradient of the filter weights causes the step size decrease. The algorithm solves the contradiction between fast convergence and low steady-state misalignment. Compared with other algorithms,it is insensitive to noise and suitable for digital hearing aids acoustic feedback cancellation. The experiments show that the proposed algorithm achieves more excellent perform-ance than traditional NLMS for acoustic feedback cancellation.
展开▼
