An efficient method for estimation of autoregressive signals in noise

The problem of estimating parameters of autoregressive (AR) signals from noisy data is studied in this paper. First, the previous improved least-squares method with direct implementation structure (called ILSD) is revisited with the purpose of establishing its mean convergence. Second, a new and efficient estimation method for noisy AR signals is presented by re-organizing the key equations derived for the ILSD method. The feature of the new scheme is that it is in an non-iterative form, so there is no convergence issue of iteration process involved. Computer simulation results are included to illustrate the performance of the new estimation scheme.