非高斯噪声环境中基于梯度范数的自适应滤波算法  

An Adaptive Filtering Algorithm Based on Gradient-norm in Non-Gaussian Environment

作　　者：冯子昂 胡国平[1] 匡旭斌 周豪[1] 

机构地区：[1]空军工程大学防空反导学院 [2]93861部队

出　　处：《弹箭与制导学报》2017年第3期93-96,100,共5页Journal of Projectiles,Rockets,Missiles and Guidance

基　　金：国家自然科学基金(61372166);陕西省自然科学基础研究计划(2014JM8308)资助

摘　　要：针对传统最小均方类算法在非高斯噪声中自适应滤波性能下降的问题,提出了一种基于梯度范数的变步长归一化最小平均p范数算法。算法将α稳定分布作为非高斯噪声分布模型,依靠梯度范数与均方权值偏差(MSD)的关系自适应调整步长,加快收敛速度的同时减小稳态误差,理论推导证明了算法的有效性。仿真结果表明,在非高斯噪声条件下,该算法具有更好的收敛性能和抗突变能力以及更小的稳定误差。Conventional least mean square（LMS） algorithms meet declines of performance in non-Gaussian environment. A new variable step-size normalized least mean p-norm algorithm based on gradient-norm is proposed. The new algorithm assumes that the non-Gaussian noise satisfies alpha stable distribution, and the step size is adaptively adjusted by the relationship between mean square departure （MSD） and the gradient-norm. Through the relationship, the convergence rate is accelerated and the steady state error is decreased at the same time. The performance of the proposed algorithm is confirmed by theoretical derivation. Simulation results show that the proposed method has faster convergence rate, smaller steady state error and better performance of anti-saltation in non-Ganssian environment.

关 键 词：a稳定分布 自适应滤波 梯度范数 归一化最小平均p范数算法 

分 类 号：TN911.72[电子电信—通信与信息系统]