Unbiased parameter estimation of nonstationary signals in noise

Recent approaches to the modeling of nonstationary signals by means of AR or ARMA models use a representation with time-varying parameters. The time-varying parameters are assumed to be linear combinations of a set of basis time functions so that the model is specified by constant parameters. For stationary signals disturbed by white noise, an approach based upon a modified least-squares method leads to a good unbiased estimator of the parameters. In this correspondence, a similar algorithm deriving the unbiased parameters for nonstationary signals in white noise is given. The experimental results show the good performance of the proposed estimator.