CAR identification from nonuniformly sampled values using LMS
Elisabeth Lahalle, Daniel Poulton, J. Oksman · 2005
In this paper a new CAR LMS identification algorithm for irregularly sampled signals is proposed. The proposed method uses implicit numerical integration formulas to build an adaptive predictor from the stochastic differential equation of the CAR model. Formulas that may adapt to the irregular sampling case have been considered. The performances of the proposed method have been evaluated for both Poisson and jitter sampling schemes.