摘要
针对稀疏未知系统的辨识问题,提出了一种基于lp(0<p<1)范数的稀疏约束变步长最小均方自适应滤波算法,并对其收敛性进行了理论分析。该算法将系统迭代过程中产生的预测误差的平方根引入到步长控制中,设置了平衡系数以平衡系统的收敛速度和稳态误差,使步长在迭代过程中能够得到实时的调整。同时,将lp范数作为惩罚约束项作用在代价函数中,使得自适应过程具有吸引零滤波器系数的能力。由于lp范数约束比l1范数更加接近l0范数,系统辨识结果较为精确。仿真结果表明,该算法在系统辨识中收敛速度和稳定性均优于现有的稀疏系统辨识方法。
In order to improve the performance of sparse adaptive filtering(SAF)algorithms for system identification when the system is sparse,a new square root variable step-size lp(0<p<1)norm constraint least mean square(LMS)algorithm is proposed.The square root of iteration error is adopted into the step-size length,and a parameter Vth is used to balance the convergence speed and the mean-square error of identification.The lp norm is used as penalty in the cost function of the LMS algorithm to exploit the sparsity of the system.On the basis of maintaining the excellent convergence speed and steady-state performance of the traditional LMS algorithm,this algorithm further improves the convergence speed and accuracy of system identification.The convergence analysis of the proposed algorithm is presented,and the stability condition is derived.Simulation results show that the proposed algorithm outperforms different LMS-based identification.
作者
周其玉
张爱华
曹文周
张瑞哲
ZHOU Qiyu;ZHANG Aihua;CAO Wenzhou;ZHANG Ruizhe(School of Electronic and Information Engineering,Zhongyuan University of Technology,Zhengzhou 450007,China;School of Information and Communication Engineering,University of Electronic Science and Technology of China,Chengdu 611731,China;School of Information Engineering,Zhengzhou University,Zhengzhou 450001,China)
出处
《电讯技术》
北大核心
2020年第2期137-141,共5页
Telecommunication Engineering
基金
国家自然科学基金资助项目(61501530)
河南省高等学校重点科研项目(16A510012)
河南省高校科技创新团队支持计划(18IRTSTHN013)
关键词
稀疏系统辨识
变步长
最小均方算法
lp范数惩罚
sparse system identification
variable step-size
least mean square
lp norm penalty
