PORLA METODU İLE TAHMİN EDİLEN ARMA MODEL PARAMETRELERİ ÜZERİNDE PENCERE FONKSİYONLARININ ETKİSİ

The effect of window functions on the ARMA (Autoregressive Moving Average) model parameters estimated by PORLA (Pure Order Recursive Ladder Algorithm) method is presented. The PORLA method has an algorithm structure, in which the tracking and the modelling problems are treated as independent sub algorithms. In this method, first, the tracking of the nonstationary data is performed by the time-recursive calculation of the input/output data covariance block matrix. Second, the modelling problem is solved by the two-channel PORLA method in which the ARMA modelling problem is embedded. Error propagation in time can not occur in the PORLA method. Arbitrary windowing techniques can be easily incorporated to control the fast start-up and tracking capability. To illustrate the effect of window functions on the ARMA model parameters estimated by PORLA method, the simulation results are given for the different window functions such as the rectangular, triangular, Bartlett, Hanning, Hamming, exponential, modified Barnwell, Blackman and Kaiser windows.