Autoregressive spectral estimation based on asymmetric two-sided prediction

This work proposes an asymmetric two-sided linear prediction (ATSP)-based autoregressive (AR) spectral estimation method using both past and future time series values asymmetrically to estimate the present value. Simulation results show that the ATSP-based AR spectral estimators may alleviate the problem of undesirable spurious peaks when the one-sided linear prediction (OSP)-based AR spectral estimators are used with a large order to resolve frequencies of very closely spaced sinusoids in the presence of observation noise. The proposed spectral estimators may also address the false peaks and spectral line splitting problems inherent in the symmetric two-sided linear prediction (STSP)-based AR spectral estimators proposed by other authors.