For the contradition of existing variable step-size adaptive algorithms between convergence speed and steady-state error,a new variable step-size least mean square(LMS) adaptive filtering algorithm is proposed. Based on a function of the class S,this new algorithm introduces a reference of adjustment factor P to perform real-time adjustments of the shape of step function,and adjusts the step according to error’s autocorrelation mean value so that the convergence is faster innitially and the step changes more smoothly at steady state. The introduction of maximum likelihood weighted algorithm further suppresses the spurious peaks of adaptive filter weights. The new algorithm with maximum likelihood weighted is applied in adaptive time delay estimation experiment,and the result shows under the condition of the existing fixed parameters, the newly-proposed algorithm has faster convergence and smaller steady-state error. Meanwhile,accurate delay estimation when SNR is above -3 dB can be achieved effectively.%针对已有的变步长自适应算法收敛速度和稳态误差矛盾的问题，提出了一种新的变步长最小均方自适应滤波算法。新的算法在类S函数的基础上，引入调节因子P对步长函数的形状进行实时调整，并以误差的自相关时间均值估计调节步长，使得算法在初始时具有较快的收敛速度，稳态时有更平滑的步长变化。在新算法中引用最大似然加权算法进一步抑制自适应滤波器权系数伪峰。将新算法和最大似然加权应用在自适应时延估计的实验中，结果表明：在已有参数固定的条件下，新提出的算法具有更快的收敛速度和更小的稳态误差。同时，时延估计实验中能有效地实现信噪比-3 dB以上的准确时延估计。
展开▼
