隨著網路上多媒體應用日益風行，即時性視訊的應用也逐漸成為矚目的焦點，然而視訊訊務對延遲以及封包遺失皆非常敏感，而且其資料流具有變動位元率的特性，又具有高變動的資料叢集現象，更因其為即時性的應用，無法事先得知訊務的分布特性，造成欲在網路上傳送即時視訊且要有良好的品質是件深具挑戰性的事。許多文獻提出各種即時性視訊訊務的預測技術，但很少討論到利用僅以預測的結果及該頻寬協商的機制，如何保證其畫面延遲並且符合使用者緩衝區容量大小的接受範圍之內。 在本篇論文中，我們針對即時性視訊服務提出一種新的即時性視訊訊務預測技術以及頻寬協商機制。即時性視訊訊務預測技術是依據畫面複雜度及偵測場景變換為基礎來做視訊訊務的預測，而頻寬協商機制則加入使用者緩衝區的使用限制，整合所提之視訊訊務預測技術並提出一套頻寬協商演算法則，使預測結果能同時滿足畫面延遲的限制以及使用者緩衝區的容量限制。實驗結果顯示，我們所提出的視訊訊務預測機制在畫面群位元率預測下，比傳統利用調適性最小錯誤均方線性預測器，可以減少10％至40％的預測錯誤量，更能利用簡單的參數設定，來調整重新協商次數與頻寬利用率的關係；而在使用者緩衝區的使用限制下，所提出的頻寬協商演算法確可在良好的頻寬利用率與合理的重新協商次數之下，滿足視訊訊務的延遲要求。 Variable bit-rate (VBR) compressed video traffic is difficult to manage because it has strict delay and loss requirements. In particular, we cannot attain the bandwidth requirements for future frames in real-time video applications. Therefore, it is necessary to use a traffic prediction algorithm to estimate how much bandwidth should be reserved. Up to now, many literatures have proposed many video traffic prediction methods for bandwidth reservation. However, client buffer constraint and delay requirements are not considered in these methods. In this thesis, we propose a new video traffic prediction scheme and a bandwidth negotiation scheme for real-time video applications. The video traffic prediction scheme is based on picture complexity analysis. The bandwidth negotiation scheme is based on the client buffer size constraint and the predictions from the proposed video traffic prediction scheme. It must decide when to renegotiate its service rate and what the new service rate should be. The performance of the strategy is studied using renegotiated constant bit-rate (RCBR) network service model. Simulation results show that using the proposed prediction scheme for predicting GOP rates reduces the prediction errors from 10% to 40% as compared to the conventional methods. The proposed bandwidth negotiation scheme also achieves high bandwidth utilization with reasonable negotiation times.
