摘要： 在分析现有基于不同误差准则的最小四次均方算法和最小平均绝对偏差算法的基础上，提出一种可抑制多模噪声的归一化混合范数滤波算法。采用引入混合参数的代价函数调节混合参数得到最佳的滤波算法。建立步长参数随输入信噪比变化的非线性函数关系。仿真结果表明，相对于高斯环境下的滤波，该算法在多模噪声环境下的鲁棒性得到显著提高，能有效地提取信号。

关键词: 多模噪声, 最小均方算法, 噪声抑制, 变步长

Abstract: In order to suppress the of multi-mode noise, this paper proposes a Normalized Mixed-norm Filtering Algorithm(NMNF) based on the analysis of the LMF algorithm and LMAD algorithm based on different error criterion. This algorithm adopts a cost function parameters of the introduction of hybrid, and mixing parameters can be adjusted to get the best filtering algorithm. It establishes of a step change in parameters with the input SNR nonlinear function. Simulation results show that, relative to the Gaussian filtering environment, the algorithm in multi-mode noise environment significantly improves robustness.

Key words: multi-mode noise, Least Mean Square(LMS) algorithm, noise suppression, variable step-size

中图分类号: 

• TP391