Abstract: Echo cancellation system use self-adapting filter to simulate the echo channel, the order of the adaptive filter have great impacts on the echo cancellation result, also the proper order can improve it. According to the weights of active region, it can simulate the basic echo channel. Firstly, the modified NLMS algorithm can frame the order of initial weight value and calculate the mean value and joggling states; then, it can adjust the active region by comparing threshold result and abandon inactive region; eventually, it can adjust the order of filter by the total number of frames. the modified NLMS algorithm not only retain the advantage of NLMS algorithm including simple structure and fast convergence, but also adjust the filter order at real time, it enhances the echo cancellation result effectively. The echo cancellation simulation results show that the filter misalignment coefficient of the improved NLMS algorithm is lower than the NLMS and PNLMS algorithm.

Key words: echo cancellation system, filter order, weights of active parts

摘要： 回声消除系统使用自适应滤波器模拟回声信道，自适应滤波器的阶数对回声消除效果具有重要的影响，合适的阶数可以提高回声消除效果。根据滤波器活跃区域权值可以模拟出基本的回声信道的原理，改进的NLMS算法首先将滤波器的初始权值阶数分帧，并求取每帧内权值的均值与抖动状态，然后通过比较阈值结果调整滤波器权值的活跃区域，摒弃非活跃权值区域，最终根据活跃权值的总帧数调整滤波器的阶数。改进的NLMS算法不仅保留了NLMS算法的结构简单、快速收敛的优点，而且实时调整滤波器阶数，降低稳态误差。通过回声消除仿真对比实验显示，算法的滤波器失准系数低于NLMS、PNLMS算法。

关键词: 回声消除系统, 滤波器阶数, 权值活跃部分