An adaptive filtering based on moving window covariance estimation is introduced after the shortcomings of covariance matrices formed by windowing residual vectors, innovation vectors and correction vectors of the dynamic states are analyzed.

Aiming at the shortcoming of the estimating covariance which is obtained by given measurement variance,a new algorithm of sensor management based on time-varying measurement variance is presented.

For the environment in which the clutter power changes slowly with distance,a novel weighting maximum likelihood estimation(WMLE) algorithm is proposed that uses the Bayes criterion to improve the approximative covariance matrix estimation for Space-Time Adaptive Processing(STAP).

For the clutter power slowly changed environment—range-dependent nonhomogeneity,the weighting covariance matrix estimation algorithm is proposed which is based on the relative distance criterion.

We give a sufficient and necessary condition for the covariance improve- ment estimator of the unknown regression coefficient and their estimable functions under the reduced model remains covariance improvement estimator in the partitioned seemingly unrelated regression model.

For a chass of seemingly unrelated regression system(1),when the design matrix X_1 is ill-conditioned,Wang song-gui has proposed improvement estimators _1(k)、_1(T,k) of covariance improvement estimators _1、_1(T) for covariance matrix V is known or unknown in[1].

In this paper we try to obtain the covariance-improved estimator sequence and the convergence of two-stage covariance-improved estimator sequence.

The exact finite sample result of MSEM of the two-stage covariance-improved estimator introduced by Liu and Wang (1999) is obtained in SUR model with the multivariate t-distribution.