摘要
本文提出用前向和后向预测均值滤波法估计噪声中多个实指数衰减正弦波参数。算法是将频率和衰减因子的后向预测估计值和前向预测估计值取均值。蒙特卡洛模拟计算表明:均值滤波在两个方面优于改进的K-T法。 1.可在很宽的归一化频率范围內对较大衰减因子得到无偏估计。除个别点外,其方差特性近似同改进的K-T法。 2.进一步减小了频率和较小衰减因子的估计偏差。且所有参数的估计偏差受归一化频率变化影响很小。 伹这种方法的计算量比改进的K-T法增加了一倍。
Abstract The mean filter(MF) of forward predictor(FP) and backward predictor (BP) is suggested. In this algorithm the mean values of the FP est imate and the BP estimate to frequencies and damping factors of multiple rean exponen-tially damped sinusoids are selected in stead of the BP estimate, The advan-tage of the mean filter, which is shown by Monte-Car10 simulatiol when SNR=10db, is:1. The estimate of larger damping factor is unbiased in wide range of normalized frequency, The standard derivation of the estimate for larger damping factor by MF algorithm is slightly larger than that by improved K-T algorithm except for a certain frequency.2. The bias of frequencies and smaller damping factor is reduced still further and the variance of the estimate for frequencies and smaller damping factors is similar to that by KT algorithm. The bias of the estimate for all parameters change much less with the normalized frequency than by improved KT algorithm.But the amount of computation is increased y two times in comparison with improved KT algorithm.
出处
《信号处理》
CSCD
北大核心
1990年第1期17-24,共8页
Journal of Signal Processing
基金
国家自然科学基金的资助
