動態自適性串流架構下具 QoE 感知之最佳化排程設計

Abstract

動態自適性串流已成為現今最熱門的影音串流標準,此方法允許使 用者端動態調整影片解析度;然而,現今大多數文獻提出的品質自適 性策略,主要最佳化使用者品質經驗,尚未解決如何花費最小資源達 到影片播放的流暢與高畫質等特性。本論文探討網路頻寬變動較小的 居家環境,如協定網路電視的網路中,解決最小資源達到最佳化視覺 體驗問題,預設網路頻寬與緩衝時間計算傳送排程以最佳化使用者經 驗。透過上界演算法,找到最少所需花費資源,包含最小緩衝時間、 最小緩衝區與最小使用者可播放影片的網路頻寬;此外,利用最佳品 質演算法、L2H 與 L2HB,根據使用者滿意度標準最佳化傳送排程。 最後,客觀與主觀使用者滿意度實驗,顯示本論文提出的演算法能夠 讓使用者有較佳的視覺體驗。

DASH (Dynamic Adaptive Streaming over HTTP) is now the most pop- ular standard in video streaming. This method allows the client adjusting the quality of the requested stream dynamically. To the best of our knowledge, most heuristic algorithms proposed for DASH consider how to maintain a good user experience when playing videos; however, minimum resources for smooth and high quality video playback have not been resolved in the previ- ous studies. In this paper, we are interested in transmitting DASH videos over residential networks with small bandwidth (such as DSL based network for IPTV). The goal is to solve the QoE maximization with minimum resource problem and design a transmission schedule for a given transmission rate and initial delay such that QoE metrics are optimized. We show that upper-bound algorithm can find minimum resources (such as the minimum buffer size, ini- tial delay, and the network bandwidth) for video playback. Besides, maxi- mum quality algorithm, L2H with and without buffer size constraint optimize the given QoE metrics. The results of objective and subjective experiments show that the given algorithms achieve better QoE than previous research.