低信噪比下的变步长最小均方自适应算法及其在时延估计中的应用

为提高变步长最小均方(LMS)自适应算法在噪声干扰下的时变时延跟踪性能,提出改进的变步长LMS自适应算法.该算法对MVSS-LMS算法进行误差均值补偿,改步长因子固定范围约束为动态变化约束；使用HB加权突出自适应滤波器权系数峰值,采用滑动窗遗忘加权减小计算复杂度.自适应时延估计仿真实验和消声水池目标被动定位试验表明:相比于参数固定条件下的MVSS-LMS算法和SVS-LMS算法,改进算法能够获得更好的时变时延跟踪性能.

Variable step-size LMS adaptive algorithm and its application in time delay estimation for low SNR

In order to improve the performance of existing variable step-size LMS-type(least mean square) algorithm for tracking time-varying delay in noise interference was discussed.Based on variable step-size MVSS-LMS algorithm,the time-averaged estimate of error’s autocorrelation was compensated and then the fixed range restrict of step-size was replaced by dynamic change restrict,overcoming fast attenuation of step-size because of less error relativity.Compared to another variable step-size SVS-LMS algorithm,the refrained algorithm was provided with smoother step-size variation and lower steady-state misadjustment.HB weighted method was introduced into the algorithm to give prominence to peaks of adaptive filter’s coefficient and sliding forgetting-weighted window could reduce computational complexity.The results show that,compared to MVSS-LMS and SVS-LMS algorithm being of fixed parameters,the algorithm and its HB weighted method can achieve superior performance for case of Gaussian and impulsive noise interference.