改进的NLMS算法及其在自适应预测中的应用

通过分析NLMS1]和VSSLMS7]两种算法控制时变步长的思想及这两种算法的优缺点,文章提出了一种改进的归一化变步长算法,它同时使用输入信号积累和瞬时误差来控制步长更新.该算法的优越性在于收敛速度快,尤其在系统跳变时也能快速收敛,可很好地应用于自适应预测系统中.理论分析与计算机仿真结果都表明该算法收敛速度快、失调量小、稳定性好, 且在低信噪比的环境中比其他同类算法有更好的性能.

Improved Variable Step Size LMS Algorithm and Its Application in Adaptive Prediction

First,the paper introduces the principle of NLMS1] and VSSLMS7] algorithms to control the new variable step size,and their advantages and disadvantages.Based on that,an improved variable step size LMS adaptive filtering algorithm is presented,which makes use of the accumulation of the input signal and temporal error signal to control the new variable step size.The superiority of this algorithm lies in its faster convergence rate,especially in the time varying system;thus,it can be applied perfectly to the adaptive predicted system.Computer simulation results confirm the theoretical analyses and show that this algorithm has faster convergence rate,smaller misadjustment and a better stability,and can achieve better performance in lower SNR situation.