Instead of many algorithms rely on accurate estimation of noise variance, this algorithm estimates noise energy through DCT which extracts local feature.
该算法不依赖于对噪声方差进行估计,直接利用DCT变换对高频各个子带进行局部特征提取,从而估计噪声能量的估计阈值。
The formulas of computing the variance and cross-covariance matrices among local state estimation errors are presented, which are applied to compute the optimal weights.
为了计算最优加权,提出了状态估计误差方差阵和互协方差阵的计算公式。
应用推荐