基于系数相关性的多尺度Kalman滤波器组的GPS共视观测数据算法

Multiscale Kalman filters algorithm for GPS common-view observation data based on correlation structure of discrete wavelet coefficients

运用基于小波系数相关性的多尺度Kalman滤波器组算法处理GPS共视观测数据。在假设GPS共视钟差数据具有1/f分形特性的条件下,用基于小波变换的算法估计GPS钟差数据的自相似参数H。当0<H<1时,GPS共视钟差数据是一个具有1/f分形特性的高斯、零均值、非静态随机过程。在此条件下,在多尺度Kalman滤波器参数估计过程中讨论小波系数列的相关性。并在考虑相关性的基础上进行钟差数据的估计。分别对单通道和多通道共视数据进行处理,并与C ircular T数据进行了比对。结果表明本文方法是可行的、有效的。

The GPS common-view observation data were processed using the multiscale Kalman filters algorithm based on a correlative structure of the discrete wavelet coefficients. Supposing the GPS common-view observation data has the 1/f fractal characteristic, the algorithm of wavelet transform is used to estimate the self-similar Hurst parameter H of GPS clock difference data. When 0 〈 H 〈 1, the 1/f fractal characteristic of the GPS clock difference data is a Gaussian Zero-mean and non-stationary stochastic process. Thus, the discrete wavelet coefficients can be discussed in the process of estimating the multi-scale Kalman coefficients, and further the discrete clock difference can be estimated. The single-channel and multi-channel common-view observation data were processed respectively. Comparisons were made between the results obtained and the Circular T data. Simulation results show that the algorithm discussed in this paper is both feasible and effective.