A study of video source modelling for 3G mobile communication systems

Video is expected to be a significant part of information carried by broadband integrated networks and 3G mobile communication systems. It is therefore important to provide accurate and efficient models for variable-bit-rate (VBR) video traffic in order to assess new algorithms in 3G mobile systems. Modelling VBR traffic as a stochastic process is a challenging task since such traffic exhibits a complex statistical property with co-existence of long-term and short-term dependencies. Three video traffic-modelling techniques (the FARIMA-based model, GEAR model and wavelet-based model) are presented and analysed in terms of first- and second-order statistics, with considerations of computing resource requirements. Comparative evaluation of these models is presented based on statistical analysis and observations in a simulation scenario.