Estimating the parameters of a noisy AR-process by using a bootstrap estimator

This paper deals with the estimation of a process modeled by an autoregressive series of known order. Its output is assumed to be corrupted with noise which is stationary and zeromean but otherwise of unknown statistics. The procedure is based on correlation analysis that assumes a model for the residuals and estimates both the residual and process parameters. It initially estimates the auto-correlation function of the noisy data, the biased process parameters, the residual autocorrelation function and the residual parameters. Then the algorithm iterates to improve the process and residual paramters alternatively which are ´bootstrapped´ together. The algorithm is terminated according to some preselected convergence criterion.