An efficient algorithm for parameter estimation of noisy AR processes

The problem of estimating the unknown parameters of an autoregressive (AR) signal observed in white noise, including signal power and noise variance, is studied. A new method is developed for parameter estimation, which is based on a simple technique of estimating the measurement noise variance by increasing the underlying AR model by one dimension. The advantage of the presented method is that consistent estimates can be directly achieved without prefiltering of noisy data and without making any parameter transformation.