自适应IIR滤波器的韧性梯度估计算法

Robust gradient estimation algorithm for adaptive IIR filters

对于自适应ＩＩＲ滤波器的输出误差模型结构，ＨＡＲＦ算法和ＲＰＥ算法是两种基本算法．ＨＡＲＦ算法受严格正实条件限制．ＲＰＥ算法则有稳定性问题；经分析，其原因在于其梯度的不坚韧性．为此，提出了一种能够稳定收敛的韧性梯度估计算法；其计算量小，是一种实时算法．计算机仿真表明，它比已有算法更为优越，扩大了自适应ＩＲ滤波器的应用范围．

HARF and RPE are the two basic algorithms for the output error model structure of adaptive IIR filters. HARF algorithm is restricted by the strict positive real condition, while RPE algorithm has the problem of stability, the reason of which lies in the non robustness of its gradients. Therefore, a Robust Gradient Estimation (RGE) algorithm is put forward, which is of stable convergence, modest computational cost and is a real time algorithm. It is indicated in the computer simulation that RGE algorithm is superior to the existing algorithms and will enlarge the application scope of adaptive IIR filters.