Fractional Brownian motion usage for smoothed video traffic bandwidth allocation

We investigate the fractional Brownian motion (fBm) accuracy to model individual video compressed source for link bandwidth allocation considering small network buffers. These video streams are smoothed by the source to decrease the rate variability. The fBm correlation degree represented by the Hurst (1951) parameter was achieved but we loose the first-order statistics that is fundamental for the bandwidth allocation scheme. The solution to this problem was the introduction of some video coding parameters providing a fBm version that well-fitted the video cell distribution while conserving the correlation degree information. Our results show that smoothed video sources have negative Hurst values. We also demonstrate that we increase the statistical multiplexing gain for negatively correlated video sources even for small buffer sizes.