Content-based dynamic network bandwidth allocation

In networks that can support only a small number of concurrent video streams (e.g. in wireless and access networks), dynamic resource allocation (DRA) may be used to increase the network utilization. The efficiency of real-time DRA scheme depends on accuracy of stream resource prediction and frequency of renegotiation. We propose a new policy in which visual content of video is used to predict the required network bandwidth. Based on simulation results, it was found that the performance of an off-line content-based DRA scheme is comparable to the off-line RVBR algorithm. However, in real-time, the reduction of 55% to 70% in network resources, as compared to existing approaches (e.g. RVBR and RCBR), could be achieved