ABSTRACT
This chapter shows that the random noise affecting the measurements is Gaussian or normally distributed with zero mean and finite variance. It discusses Kalman filters (KF) algorithms for the correlated noise processes. Wiener filtering is based on designing the filter optimal gains using the frequency domain method to solve the important problem of estimating a signal process on the basis of measurements additively corrupted by noise. The least squares method which is regarded as a deterministic method is based on the assumption that the dynamic system parameters are constant and/or slowly varying with time which implies that during the measurement period the system is quasi-stationary. The Fisher information matrix is introduced as a measure of the relative accuracy of the estimated parameters and is equal to the second gradient of the cost function.
