Characterizing ventricular fibrillation signals using direct and seasonal-type autoregressive modeling

Autoregressive modeling was used to more fully characterize the epicardial ventricular electrogram signal during ventricular fibrillation. The authors demonstrate that, for the short time period typically used by automatic implantable defibrillators, bipolar epicardial signals can be characterized as autoregressive (AR) processes of an appropriate order p with white noise excitation. An alternative seasonal-type autoregressive process, where all AR coefficients except the first and last p coefficients are zero, was also examined. Three different criteria, Akaike, Hannan-Quinn, and Rissanen, were then evaluated for determination of the AR model orders.<
>