基于M估计的有效成比例仿射投影算法

Efficient proportionate affine projection algorithm based on M-estimate

在回声消除中,针对回声路径具有稀疏性的优势,有学者提出基于μ准则记忆的改进型成比例仿射投影算法(μ-MIPAPA),但该算法抗脉冲干扰性能较差。为此将稳健估计中的M估计应用于该算法,提出一种基于M估计μ准则记忆的改进型成比例仿射投影算法(μ-MMIPAPA)。为使算法具有更好的稳态性能,采用改良的Huber M估计函数,改进阈值参数的选取方式。有色输入下的稀疏回声信道辨识结果表明了μ-MMIPAPA的良好稳态性能和抗脉冲干扰性能。

For echo cancellation,theμ-law memory improved proportionate affine projection algorithm（μ-MIPAPA）has been proposed to take advantage of the sparseness of the echo path.However,the algorithm has poor performance on combating impulsive interference.The M-estimate of robust estimates was applied to the algorithm,and aμ-law memory improved proportionate affine projection algorithm based on M-estimate（μ-MMIPAPA）was presented.To make the algorithm has better steady-state perfor-mance,a modified Huber M estimation function was used and the way of the threshold parameter selection was improved.The perfor-mance of the proposed algorithm was verified by the simulation of sparse echo channel identification in colored input.