基于带双窗N-1/N重叠频域FIR滤波器的自适应除噪

ANC Based on N-1/N Overlapping FIR Filter with Dual Windows in Frequency Domain

基于一种带双窗重叠自适应滤波器 ,将重叠滤波思想引入LMS算法 ,给出了窗加权重叠LMS (WO LMS)算法。与传统的LMS算法相比 ,WO LMS算法在同样收敛速度的前提下可以得到较低的稳态均方误差。理论分析了算法的收敛性 ,实验中通过和LMS算法的比较 ,验证了WO LMS算法的优越性。

A novel overlapping filter is proposed and the equivalent structure is deduced out. Based on such overlapping adaptive filter with double windows, windowed overlapping LMS(WO-LMS) algorithm is advanced which combines the windowed overlapping idea and LMS algorithm. It is theoretically proved that WO-LMS achieves higher convergence speed under the same condition as LMS, however with smaller step range. ANC with WO-LMS is introduced and experiments on different correlation between noise in main and sub channel are made to verify WO-LMS algorithm that can obtain lower steady mean square error at the same convergence speed. Result that de-noise effect is relevant to correlation between noise in main and sub channel is brought out.