Traffic modeling and bandwidth allocation for MPEG video sources in ATM networks

We extend the autoregressive (AR) model to the projected AR (PAR) model for the purpose of fitting the histogram of VBR Motion Picture Expert Group (MPEG) and at the same time preserving the autocorrelation property. We also propose a bandwidth allocation scheme based on a prediction method where the frame sequence and the autocorrelation between consecutive frames are considered. Our experimental results show that the PAR model is capable of capturing the statistics of both the histogram and autocorrelation. The PAR model is then used as the traffic generator of a video source for an exhaustive simulation on the effectiveness of the proposed bandwidth allocation scheme. The proposed bandwidth allocation method, which predicts the bandwidth by the frame correlation (FC), demonstrates a quite good performance when comparing to a previous scheme, especially for the video scenes with the combination of intraframes and interframes. Bandwidth allocation of a multiplexer connected with several video sources is also studied and the results show that the proposed method is better than that of the fixed bandwidth allocation and is suitable for, the application of MPEG video services