Video sources will become one of the major traffic in future ATM networks. The video sources should be transmitted using variable bit rate (VBR) since it maintains a constant video quality and has a high statistical multiplexing gain. Video characterization is an essential step before estimating whether the video traffic satisfies the required quality of services (QOS). A good video model is a good matching of the histogram and autocorrelation function that leads to an accurate estimation of QOS. In this thesis, models are built for both motion-JPEG sequence and MPEG-1 sequence. It is found that the motion-JPEG sequence is highly correlated and MPEG-1 sequence is less correlated....[ Read more ]

Video sources will become one of the major traffic in future ATM networks. The video sources should be transmitted using variable bit rate (VBR) since it maintains a constant video quality and has a high statistical multiplexing gain. Video characterization is an essential step before estimating whether the video traffic satisfies the required quality of services (QOS). A good video model is a good matching of the histogram and autocorrelation function that leads to an accurate estimation of QOS. In this thesis, models are built for both motion-JPEG sequence and MPEG-1 sequence. It is found that the motion-JPEG sequence is highly correlated and MPEG-1 sequence is less correlated.

The aggregate traffic of homogeneous motion-JPEG video is modeled by the correlated Bernoulli process and the aggregate traffic of non-homogeneous motion-JPEG can be modeled by the correlated histogram approach. The random sources model is proposed for the modeling of the aggregate of homogeneous MPEG-1 video. A single MPEG-1 video sequence can also be generated by cross- correlation coefficient. Simulations are performed using BONeS Designer to find the cell loss probability. Closed-form approximations are also obtained when the buffer is not large.

In order to provide a good QOS, a new approach is proposed to reduce the cell loss probability in an ATM network using interleaved error-correcting codes. In an ATM network with bursty traffic, the occurrence of cell loss is also bursty due to an instantaneous loss of cells during buffer overflow. Interleaving can reduce the burstiness and thus the cell loss probability.