基于梯度向量投影的自适应滤波算法的改进及其在多路回波消除中的应用

An Improved Adaptive Filtering Algorithm Based on Projection of Gradient Vectors and Its Application in Multi-channel Acoustic Echo Cancellation

该文首先对Lim(2000)的基于梯度向量正交投影的算法(OGA)进行了分析和改进,在此基础上获得了一种新的自适应滤波算法(MOGA)。新算法使用时变遗忘因子对误差进行指数加权平均来估计均方误差,并使用该因子改变自适应迭代过程中滤波器系数向量的更新方向.然后将改进后的新算法扩展成两路回波消除算法用于多路回波的消除中,获得了良好的效果。仿真结果表明, MOGA不仅对时变或时不变系统具有很好的跟踪能力,克服了Lim(2000)所提算法收敛性不佳甚至有时发散的缺陷,而且应用于多路回波消除时具有计算量小,收敛速度快和精度高等特点,其收敛速度和精度优于J.Benesty(1996)和G.Sankaran(1999)的相应结果。

In this paper, a new adaptive filtering algorithm is proposed based on the analysis of the orthogonal projection of gradient vectors described in Lim(2000). In the new algorithm, a time-variant forgetting factor is introduced to estimate the Mean Square Er-ror(MSE) and change the updating direction of adaptive filter coefficient vector. Furthermore, the new algorithm is extended to two-channel algorithm for multi-channel acoustic echo cancellation. Simulation in MATLAB shows that the new algorithm has good convergence and tracking capability to time-variant and time-invariant system, and can overcome the weakness in convergence and divergence of the Lim's algorithm(2000). Moreover, the extended algorithm is of less computation, faster convergence, and higher accuracy when applied in multi-channel acoustic echo cancellation. Its convergent speed and accuracy are better than that of the algorithms proposed by Benesty(1996) and Sankaran(1999).