We consider the problem of Bayesian tracking of radar cross section. The adopted observation model employs the gamma family, which covers all Swerling cases in a unified framework. State dynamics are modeled using a nonstationary autoregressive gamma process. The principal component of the proposed solution is a nontrivial gamma approximation, applied during the time update recursion. The superior performance of the proposed approach...

In the paper the comparison of the popular pitch determination (PD) algorithms for thepurpose of elimination of clicks from archive audio signals using sparse autoregressive (SAR)modeling is presented. The SAR signal representation has been widely used in code-excitedlinear prediction (CELP) systems. The appropriate construction of the SAR model is requiredto guarantee model stability. For this reason the signal representation...

When identifying a nonstationary autoregressive process, e.g. for the purpose of signal prediction or parametric spectrum estimation, two important decisions must be taken. First, one should choose the appropriate order of the autoregressive model, i.e., the number of autoregressive coefficients that will be estimated. Second, if identification is carried out using the local estimation technique, such as the localized version of...

The problem of estimating spectral density of a nonstationary process satisfying local stationarity conditions is considered. The proposed solution is a two step procedure based on local autoregressive (AR) modeling. In the first step Bayesian-like averaging of AR models, differing in order, is performed. The main contribution of the paper is development of a new final-prediction-error-like statistic, which can be used to select...

The problem of identification of nonstationary multivariate autoregressive processes using noncausal local estimation schemes is considered and a new approach to joint selection of the model order and the estimation bandwidth is proposed. The new selection rule, based on evaluation of pseudoprediction errors, is compared with the previously proposed one, based on the modified Akaike’s final prediction error criterion.