Least squares predictive transform modeling
J.R. Guerci, Erlan H. Feria · 2003
A least-squares error (LSE) linear predictive transform (LPT) method is offered to obtain adaptive time-varying signal models in filtering applications. The method is a direct extension of minimum mean-square error (MSE) LPT modeling to the adaptive case. The LSE LPT method is illustrated with a digital monochrome image filtering example which shows that the resulting time-varying signal model behaves, approximately, as a time-varying whitening filter of the Kalman type.>