Novel parameter estimation of autoregressive signals in the presence of noise

This paper proposes a new method for estimating the parameters of an autoregressive (AR) signal from observations corrupted by white noise. The feature of the new method is that the observation noise variance estimate is converted into the only solution of a nonlinear equation to yield unbiased estimate of the AR parameters. Moreover, a convergent Newton iterative algorithm with a deterministic initial point is presented for efficient implementation of the proposed new estimation method. As a result, the proposed new method can minimize the error of estimating the variance of the observation noise. Since more accurate estimates of this observation noise variance can be attained at earlier stages, the proposed method can achieve a good performance in estimating the AR signal parameters. Numerical results demonstrate that the proposed new algorithm is more effective in terms of accuracy and robustness against noise than conventional algorithms.