中文摘要：

常见的野值剔除方法即卡尔曼滤波算法条件苛刻,不易实现。在分析了系统方程之后,利用小波变换的基本原理,将系统信号分解到各个尺度上,在各尺度上分别进行卡尔曼滤波,之后进行小波重构从而得到各尺度下的滤波估值。此算法将小波去噪与卡尔曼滤波算法结合起来,能够更有效地剔除野值。针对卡尔曼滤波算法的复杂性,还提出了一种有效的滤波算法,利用该算法进行小波多尺度分解和滤波和重构,信号野值也可以得到很好的剔除。经仿真实验验证,这两种多尺度小波变换算法都能够很好地剔除野值,效果明显。

英文摘要：

Kalman filtering algorithm which is a common method of outliers elimination needs harsh conditions and is hard to achieve. After analyzing the system equation, using the basic principles of wavelet transform,the system signal is discomposed to each scale where the disassembled signal is estimated by Kalman filter, and then wavelet reconstruction is applied to get the best estimation in various scale. This algorithm combining wavelet de-noising with Kalman filtering algorithm could eliminate outliers more effectively. In light of the complexity of the Kalmari filter algorithm,an efficient filtering algorithm is proposed, which is able to eliminate outliers well after wavelet multiseale decomposition and reconstruction. Simulation showed both two multiscale wavelet transform algorithms have good results in elimination of outliers.