MPEG Video Traffic Models: Sequentially Modulated Self-Similar Processes

1 New traffic models are called for to facilitate the design of effective admission and flow control algorithms, and network performance evaluation in accommodating video traffic. We propose a new approach, sequentially modulated self-similar processes (SMSSPs), to model MPEG coded video traffic. SMSSPs are shown to be able to capture both short range dependency (SRD) and long range dependency (LRD) of the video traffic accurately. Traffic data are decomposed according to the MPEG data structure, into several parts, each modeled as a self-similar process. These processes are then modulated sequentially in a manner similar to how the frames are grouped into the GOP (Group of Pictures) pattern.