摘要
为解决在外部环境突变的条件下,最小均方(LMS)算法收敛速度与稳态误差之间难以同时平衡优化这一问题,提出一种基于t分布的新型变步长LMS算法(简称BVTLMS算法)。通过理论推导,对t分布概率密度函数进行变换,得到全新的步长随误差变化的曲线,满足了收敛条件;然后针对不同参数分析对BVTLMS算法的影响,确定最优参数,利用该参数进行仿真。结果表明,在同样的仿真条件下,BVTLMS算法分别与基于分式函数变步长算法和基于正态分布变步长算法相比,具有收敛速度更快、稳态误差更小和更强的跟踪能力优势。采用实际数据进行验证,也证明所提出的BVTLMS算法的优越性。
In order to solve the problem that it is difficult for convergence rate of LMS algorithm and the steady-state error to keep in balance at the same time under the condition of abrupt change, a new variable step size LMS algorithm based on t distribution(BVTLMS algorithm for short)is proposed.By transforming the t distribution function, a new curve of step size varying with error is obtained, and the convergence condition is satisfied.Then according to the influence of different parameter analysis on BVTLMS algorithm, the optimal parameter is determined for simulation.Under the same simulation conditions, compared with the VXLMS algorithm and the variable step size algorithm based on Gaussian distribution, the simulation results show that BVTLMS algorithm has three advantages: faster convergence speed, smaller steady-state error and stronger tracking ability.Finally, the actual data better proved that the actual performance of BVTLMS algorithm was the best.The data obtained by BVTLMS algorithm was better than that of unmodification.
作者
伍彩云
李汶东
WU Caiyun;LI Wendong(Shenyang Ligong University,Shenyang 110159,China)
出处
《沈阳理工大学学报》
CAS
2021年第1期82-87,共6页
Journal of Shenyang Ligong University
基金
国家自然科学基金(61603264)。
关键词
自适应控制
主动降噪
最小均方算法
可变步长
adaptive control
active noise cancel
LMS(Least Mean Squares)algorithm
variable step size
