Predictive bandwidth control for MPEG video: a wavelet approach for self-similar parameters estimation

The measurements of various types of network traffic are found to exhibit self-similar characteristics. A key parameter characterizing self-similar processes is the Hurst parameter H, which is designed to capture the degree of self-similarity. In order to determine if a given time series exhibits self-similarity, a method is needed to estimate H for a given time series. We present an estimation tool by using the wavelet transform. An important feature of the wavelet-based tool is the conceptual and practical simplicity, consisting essentially in measuring the slope. Moreover, the Hurst parameter can be accurately estimated from the power-law behavior of the wavelet coefficients. We use an FIR multilayer network to predict the next incoming data, then we apply the wavelet-based tool to measure the Hurst parameters of the predicted data. Based on these Hurst parameters, we apply the Norros (see IEEE JSAC, vol.15, no.2, p.200-8, 1997) formula to estimate the bandwidth requirement for each predicted data. Finally, we utilize the random early detection (RED) algorithm to traffic congestion control based on the predictive bandwidth