訊務平順化 (Traffic smoothing)是視訊傳輸前，充分考量網路服務能力與使用者資源，有效減少視訊串流叢集性與頻寬浪費的重要技術之一。在視訊壓縮標準規範中，如MPEG視訊標準的Video Buffering Verifier (VBV)、或H.263視訊標準的Hypothetical reference decoder (HRD)，對於所提供的串流順暢傳輸機制，是基於固定的傳輸率與已知的使用者緩衝區大小的假設下予以建議，如此受侷限的假設環境，難以適用於現今廣泛、複雜的網路環境與眾多視訊傳輸應用的需求。因此，一個能提供有效平整訊務、適應網路環境變化、兼顧使用者播放能力，甚至適用於即時性視訊傳輸的應用，如線上球賽轉播、線上監視系統…等之視訊傳輸服務技術，已成為迫切研究發展的議題。 在本篇論文中，針對線上視訊服務提出一套有效訊務平順化之整合方案，其中包含視訊訊務預測以及因應可變延遲網路之智慧型線上平順調整機制。有別於其他相關文獻，本論文突破線上訊務平順化技術的瓶頸，具有更低的傳輸率變動與保有低峰值頻寬需求的優點，更重要地，具有對抗網路延遲之能力。實驗結果顯示，傳輸率變動造成的協商次數大幅降低最多可達80%，改善之傳輸率規劃可有效地因應網路的延遲，使得畫面丟棄率大幅降低約20%以上，進而有效提升使用者播放的視訊品質。 The technique of on-line video traffic smoothing is a significant method to improve the network BW utilization and reduce the burstiness characteristic of live video data. An efficient traffic smoothing scheme should not only consider the characteristics of video coding, but also consider the QoS capacity of IP network and the limitation of server and client buffer space. Regarding the methods of delivering video data used by current video coding standards, a constant transmission rate and a fixed network delay are assumed whether the Hypothetical Reference Decoder (HRD) in H.263 or the video buffering verifier in MPEG is considered. These assumptions are quite rigid and unrealistic due to the variety of current network environments, especially for live on-line video applications. Therefore, this paper proposes an Integrated On-line traffic Smoothing (IOS) system that consists of the Long-Span Prediction (LSP) scheme, the Smart Hopping (SH) approach, the Middle Tracking (MT) method, and the Dynamic Bound Adjustment with network Delay Tracking (DBA-DT) strategy. Compared with traditional schemes, the proposed IOS can provide more efficient smoothing performances with low variance of transmission rates and low peak rate requirement. More importantly, IOS provide the robust capability against the dynamic network delay. Simulation results reveal that IOS can effectively reduce the number of negotiations with network up to 80% and low the frame loss rate up to 20%.
