A novel two-step MPEG traffic modeling algorithm based on a GBAR process

In this paper we propose a two-step MPEG video sources modeling algorithm based on three autoand cross-correlated gamma-beta autoregressive processes of order one GBAR(1). This algorithm is an extension of models originally proposed by Heyman [1] and Lombardo et al. [2]. We extend the model proposed by Heyman to the case of periodic frame sequence at the output of the codec. In order to do that we use information about the correlation between frame sizes within the group of pictures. Based on this information we introduce three cross-correlated GBAR(1) processes each of which models a particular type of the frame. The proposed algorithm captures the basic statistical characteristics of the empirical data well: the distribution of frame sizes and the autocorrelation function. Our algorithm is fairly simple computationally efficient and therefore, can be used in simulation studies.