基于Q函数的变步长LMS算法

Variable Step-size LMS Algorithm Based on Q Function

为提高变步长最小均方算法(least mean square,LMS)的抗干扰能力,并降低其稳态误差,提出了基于Q函数的变步长LMS算法,通过带有补偿项的相对误差互相关函数控制步长更新,有效克服噪声干扰;利用归一化输入信号代替原始信号降低输入信号动态范围对稳态误差的影响;给出了保证算法收敛的步长界限.仿真结果表明,与既有变步长LMS算法相比,所提出算法的稳态误差小,且收敛速度较快.

In order to improve the anti-jamming ability of LMS algorithm and reduce its steadystate error,a variable step-size LMS algorithm based on Qfunction is proposed.The relative error cross-correlation function with compensation is used to control the step update,which can overcome the interference of noise effectively.In order to reduce the impact of the large dynamic range of input signal on algorithm,the original signal is replaced by the normalized input signal.Meanwhile,the step upper bound is presented in order to ensure that the algorithm can converge.Compared with other variable step LMS algorithms,the steady-state error of the proposed algorithm is smaller and the rate of convergence is faster.