Subspace-based parameter estimation of symmetric noncausal autoregressive signals from noisy measurements (Q2732795)

The problem of parameter estimation of symmetric noncausal autoregressive signals (SNARS) from noisy observations is considered. A subspace fitting approach is used, and it is shown that the subspaces associated with a Hankel matrix formed from the data covariances contain sufficient information to construct consistent estimates of the signal parameters. A multiple signal classification (MUSIC)-like methodology is proposed for SNARS parameter estimation. The approach consists of solving the parameter estimation problem of spectrally equivalent autoregressive moving-average (ARMA) signal models. Efficiency of the method, and in particular a good compromise between computational and statistical properties, is illustrated by means of computer simulations.