Real time Continuous AR parameter estimation from randomly sampled observations

In this paper, a real-time CAR (Continuous AR) parameter estimation method from randomly sampled observations is presented. For this purpose, a new concept is introduced: the Pseudo Correlation Vector. This vector, which is equal to the Correlation Vector in the uniform sampling case, reflects the statistical dependencies between successive values of the CAR signal. As a matter of fact, being the limit of a series recursively related to the observations, its real-time estimation becomes especially easy. An inversion of its dependence on the CAR parameters leads then to an estimate of the latter. Theoretical and simulation results regarding the proposed estimator are given.