It has been reported that variable bit rate (VBR) video traffic exhibits long-range dependence(LRD). Various processes have been proposed for modeling traffic with LRD and analyzing its effects on network performance. However, in the previous models it was not possible to identify the effects of short range and long-range correlation. But recently, some reserchers presented a video traffic model based on the shift-level process. They observed that the auto-correlation function(ACF) of an empirical video trace is accurately captured by the shift-level process with combined function :an exponential function in short range and a hyperbolic function in long range. Especially, they present an accurate parameter matching algorithm for JPEG- coded video traffic.
In this paper, we present a verification of shift-level process modeling for MPEG VBR video traffics. By the shift-level process modeling, the probability distribution of scene lengths in video traffic can be obtained explicitly from ACF of video traffic. For the ACF, the matching parameter are obtained for the approximation of ACF by the combined model of exponential and hyperbolic functions. The approximated ACF of video traffic and the obtained probability of scene lengths are compared with measured ACF and probability distribution for several video traces. And also we present scene change detection algorithm and the methods of getting p.d.fs of data size and of scene lengths.
By the simulation results, it is shown that the ACF of the shift-level process modeling is very much dependent on the characteristics of video trace for MPEG-coded video traffic.