加权部分更新仿射投影算法

An affine projection algorithm with selective weighted regressors

摘要:

从数据重用因子出发，得到了基于部分更新仿射投影算法(SR-APA)的改进算法。该算法通过加权改变了SR-APA的数据筛选规律，从而降低了等效数据重用因子，并且通过对未加权原始数据的重新利用巧妙地避免了加权带来的条件数增加问题，最终达到了降低稳态均方误差(MSE)的效果。仿真结果表明，该算法不仅MSE比SR-APA低，收敛速度也比SR-APA快。在收敛速度相同时，该算法计算量只有SR-APA计算量的50%左右。

Abstract:

Based on Affine Projection Algorithm with Selective Regressors(SR-APA)，a new algorithm is proposed with a view to data reusing factor. The proposed algorithm introduces a weighting method to change the selective principles. Unweighted original data is reused in the iteration to avoid increasing the condition number of the input data, and finally the steady state Mean Square Error(MSE) is reduced. Simulation results show that the proposed algorithm not only reaches lower steady state MSE than SR-APA, but also accelerates the convergence speed of SR-APA. And with the same performance，the proposed algorithm has only 50% computations of SR-APA.