Search

Video and videoconferencing are expected to represent a substantial part of traffic carried on future IP networks, therefore requiring accurate and reliable traffic models for network performance evaluation. Previous attempts to model video have been based on pre-recorded video clips, with modelling performed at the frame level. A unified method for video traffic modelling over IP networks, through acquisition of MPEG-4 encoded videoconference Ethernet packets collected experimentally, is proposed.

An online video traffic model based on the wavelet transform is proposed and evaluated. The key to the online model support is the removal of least significant wavelet components, within a coarse time resolution, as necessitated by a priori model parameter estimation. This, in turn, is achieved through knowledge gained from the autocorrelation function of the base video traffic trace.

A cellular call capacity optimisation technique for dynamic channel allocation (DCA) schemes that are either blocking limited or dropping limited is described. A Hopfield neural network based DCA is used for illustrating the concept.