VBR broadcast video traffic modeling-a wavelet decomposition approach

We propose a wavelet decomposition approach for modeling VBR broadcast video traffic. The proposed method decomposes video traffic into two parts via a wavelet transformation, and models each part separately. The first part, which is modeled by an AR(1) process, serves to capture the long-term trend of the traffic; the second part, classified via vector quantization, addresses the short-term behavior of the traffic. Compared with other VBR video models, our model has three advantages. First, it allows the separate modeling of long- and short-term behavior of the video traffic; second, it preserves the periodic coding structure in traffic data; and third, it provides an unified approach for the frame- and slice-level traffic modeling. We demonstrate the validity of our model by statistical measurements and network performance simulation