抗冲击噪声的核对数最小绝对差算法  

Kernel least logarithmic absolute difference algorithm aganist impulsive noise

作　　者：林云[1] 雷洋[1] 曾俊俊[1] 

机构地区：[1]重庆邮电大学移动通信技术重庆市重点实验室,重庆400065

出　　处：《电子技术应用》2016年第2期78-80,84,共4页Application of Electronic Technique

摘　　要：提出了一种鲁棒核自适应滤波算法,其结合了核空间和最小对数绝对差(LLAD)算法,使用对数代价函数来解决冲击噪声对算法收敛的影响,从而提高算法的抗干扰性能。核对数最小绝对差(KLLAD)算法实现了类似核最小均方误差(KLMS)算法的收敛性能,而且KLLAD算法具有很强的抗干扰能力,在非线性系统辨识中的鲁棒性和收敛方面具有很好的表现。This paper presents a robust kernel adaptive filter algorithm, which combines the kernel space and the least logarithm absolute difference（ LLAD） algorithm, it uses logarithm as its cost function to solve the impact on the convergence of algorithm which caused by the impulsive noise, then improve the anti- interference performance of the algorithm. Kernel least logarithm abso-lute difference（ KLLAD） algorithm achieves the comparable convergence performance with the kernel least mean square（ KLMS） algo-rithm while the KLLAD algorithm is robust against impulsive interference. The good performances in terms of robust and conver-gence performance are demonstrated in nonlinear system identification.

关 键 词：自适应滤波 系统辨识 鲁棒性 

分 类 号：TN911.72[电子电信—通信与信息系统]