摘要
当LMS算法用于时变系统时,其跟踪性能不佳,常常不能满足应用要求。其主要原因是步长的选取影响收敛速度与稳态失调。而固定步长往往使二者不可兼得。本文提出一种具有快速跟踪能力和允许大输入动态范围的归一化变步长自适应滤波的新算法(NVS).计算机仿真计算表明,新算法的性能明显优于LMS算法,而计算量增加甚少。
It is shown that the tracking ability of LMS adaptive algorithm is not so better when it is used
in time-varing systems and unsatisty the requirements in use, the main reason is that the convergent speed and
steady-state misadjustment are effected by the stepsize and can not be improved by constant stepsize of same
time. Thus a new algorithm having fast tracking ability and permiting large dynamic range input is presented in
this paper, which is named as normalized variable stepsize algorithm (NVS) . The tracking ability of NVS
algorithm is more better than LMS that is verified by computer simulation, but the computing amount is only
inereased slightly.
出处
《信号处理》
CSCD
北大核心
1992年第2期112-118,共7页
Journal of Signal Processing
