A Multiscale Model for MPEG-4 Varied Bit Rate Video Traffic

Abstract

This paper studies modeling approach of MPEG-4 VBR video traffic based on multifractal multiplicative model. Multiscale analysis reveals that the multiplier distribution is different in style on different time scales. Based on statistical characteristics of the multipliers, a composite modeling approach is proposed: Gaussian distribution is used to fit multiplier distribution at large time scales, a new statistical distribution-Symmetric Pareto distribution to fit multiplier distribution at small time scales and a linear model to model frame traffic. Simulations are performed to validate the good effect of this approach.