Abstract de la publi numéro 3833

The problem addressed in this paper is the reconstruction of a continuous-time stationary process from noisy interlaced sampled observations. This work extends the one of Yen who gave an exact reconstruction of a process or a function with spectral support within (-Kpi,Kpi) subjected to interlaced sampling i.e. when sampling times are such that t_{kn}=n+a_{k}, n?Z, k=1,...,K. In this paper, we derive the linear mean square estimator (LMSE) of a continuous-time process based on the observations of noisy interlaced samples. Two different cases are studied. The first one deals with noise samples coming from a unique stationary process. In the second one, K uncorrelated noises with different spectra are considered. In both cases, the specific sampling leads to technics linked to cyclostationarity properties.