Least-squares linear estimation of signals from observations with Markovian delays (Q645706)

Assuming no signal equation is available, and that the delay is modeled by a homogeneous discrete-time Markov chain to capture the dependence between delays, the authors study the least-squares linear estimation problem of a signal based on randomly delay measurements. The signal estimation is addressed assuming that the covariance functions of the process are known and the covariance function of the signals expressed in a semi-degenerated kernel form. The proposed recursive filtering and fixed-point smoothing algorithms are obtained using an innovation approach. A numerical simulation example is included.