Traffic modeling for layered video

In this paper we propose a traffic model for delivering scalable video encoded with multiple layers on heterogeneous networks. The model is based on Markovian arrival process with marked transitions. The state of the underlying Markov chain of the video arrival process is derived from the correlation feature found in the video data. The base layer and enhancement layer video frame size pairs are decided by a cluster detection algorithm; each cluster corresponds to one state of the Markov chain. The joint base and enhancement layer video frame size distribution for each state of the Markov chain is approximated by multivariate normal distribution. Simulation study on the traffic model data and the video trace data is carried out and compared with the model. The results show that the proposed traffic model can predict the network performance with good accuracy.