- 學位論文
MPEG-DASH在多伺服器架構下之使用者經驗品質最佳化自適性串流演算法
QoE Adaptation of MPEG-DASH over Multiple Servers
摘要
智慧型終端的崛起與網際網路的廣泛運用，使得多媒體串流近期成為一個備受關注的熱門技術。MPEG-DASH標準化客戶端與伺服器端多媒體串流的傳遞服務，客戶端能依照網路狀態和使用者的需求動態調整多媒體串流的內容，進而提供符合使用者需求的多媒體串流體驗。 本論文使用開源式軟體實作一套完整的MPEG-DASH系統，並且提出一套新的在多伺服器架構下的MPEG-DASH自適性串流演算法。此演算法分為兩部分，第一部分依照當下各伺服器的網路狀態，對各伺服器的下載請求進行排程，讓優先順序高的下載請求能優先被完成，第二部分依照當下暫存器狀態選擇最適當的視訊品質，以提供順暢且穩定的視訊串流服務。本論文以視訊品質的平均位元率、平均層數、層數相差的絕對值的標準差和品質層數的變動次數四項作為自適性串流演算法衡量使用者經驗品質(QoE)的客觀指標。模擬結果顯示，在多台伺服器的架構下，本論文可以比過去文獻的方法在平均位元率和平均層數分別有566.28 kbps和0.35層的提升。
並列摘要
Video streaming, owing to the popularity of mobile devices and wide deployment of the Internet, has become the key technology for multimedia applications. The MPEG-DASH defines the transmission protocol for adaptive multimedia streaming services. The client can dynamically select the adequate bit rate for streaming services according to the network conditions or the restrictions imposed by uses. Better quality of experience (QoE) can thus be achieved. In this thesis, we implement a complete open-source MPEG-DASH system. Additionally we propose a novel adaptation algorithm for MPEG-DASH over multiple servers. The first part of the algorithm schedules the download requests of servers according to the priority of the requests and the network conditions. The second part of the algorithm chooses the most appropriate quality level of video according to the buffer status to provide smooth and stable video services. The objective visual quality is measured by the average bit rate, average quality levels, variation of absolute value of difference between quality levels, and transition frequency of quality levels. Simulation results show that the proposed method enhances the average bit rate by 566.28 kbps and the average quality levels by 0.35 quality levels as compared to the other methods in the literature.