| 摘要: | 在網際網路盛行的今天，網路上所傳送的資料內容已從傳統的檔案與語音資料擴展到多樣化的視訊內容。許多網路設備均已具備強大的運算與資料排程處理能力以應付龐大的視訊流量。然而由於視訊串流具有變動位元率 (variable bit rate) 的特性，這些網路設備亦需及時而充分地掌握視訊串流的流量特性，方能有效提升頻寬資源的管理效率。雖然啟動視訊來源端的位元率控制 (source rate control) 機制可以降低長時間 (long term) 的位元率變化，但是就短時間 (short term) 的區間而言，視訊串流仍然具有明顯的變動位元率現象，進而導致網路端的頻寬資源管理與遞送服務品質管理複雜化。因此，一個能夠有效地降低視訊串流的位元率變動以及能夠及時而準確地將視訊串流的流量特性通知網路端的機制，對於頻寬資源管理而言十分地重要。另一方面，為了提升視訊資料在無線網路中的錯誤回復 (error resilience) 能力，許多視訊編碼標準均提供再同步註記 (Resynchronization Marking) 的方法。然而，在使用再同步註記方法時，若未同時將無線網路目前的傳送品質一併考慮，則無線網路的有效頻寬利用率將會因為再同步註記時間區間的選用不當而下降。此現象在封包檔頭負荷增加或無線頻道錯誤位元率增加時更為明顯。 因此，本論文針對互動／直播／預錄式網路視訊服務，提出一套整合式端點對端點遞送品質控制系統 (IE-VDS) 來有效地解決上述的問題。IE-VDS由四個部份組成：最佳封包長度決定機制 (OPDM)、ρ-域之長跨距訊務預測機制 (ρ-LSP)、智慧型線上訊務平順化機制 (ON-ITS)、確定式離線訊務平順化機制 (OFF-DTS)。OPDM提出一個可以根據無線網路的傳送品質現況而決定最佳封包長度的封閉解 (closed form)。利用此封閉解所得到的結果可以作為視訊編碼端決定再同步註記時間區間及封裝視訊資料時的最佳依據，並且可以使得無線網路有效頻寬利用率最大化。為了有效執行互動與直播視訊訊務的平順化工作，ON-ITS首先利用ρ-LSP進行視訊訊務的預測。ρ-LSP係利用存在於畫面位元率以及非零量化係數與非零移動向量個數之間的線性關係進行準確的訊務預測。在結合ρ-LSP之後，ON-ITS提出一個調適性視窗尺寸決定機制來解決視窗尺寸大小與播放延遲之間的取捨問題，進一步降低視訊串流的位元率變動，並且及時而準確地將視訊串流的流量特性通知網路端。當視訊串流傳送於蜂巢式行動網路時，有限的終端設備記憶體容量往往會限制視訊訊務平順化的效能，尤其是希望利用確定性遞送模式 (deterministic service model) 來達到遞送服務品質的保證時更為明顯。為了解決此問題，OFF-DTS將訊務平順化的觀念納入確定性遞送模式中，並且提出一個智慧型視訊畫面捨棄機制，使得視訊串流在蜂巢式行動網路上的傳送可以同時符合終端設備記憶體的容量限制以及確定性遞送模式的流量特性要求。經由分析與模擬，最佳封包長度封閉解的準確性與有效性獲得充分的驗證。此成果可以同時有效地提升無線網路的頻寬使用率以及增加視訊資料的錯誤回復能力。相同地，透過模擬的結果，ON-ITS在不增加播放延遲的情況下，能夠有效地降低視訊串流的位元率變動，減少與網路端的頻寬協商次數，進而有效地減緩網路設備的處裡負荷。更進一步地，由模擬的數據發現，利用OFF-DTS 可以有效地降低視訊串流在蜂巢式行動網路的峰值頻寬需求以及終端設備的記憶體容量需求。 With improved capabilities of network infrastructures and video compression techniques, rich multimedia applications such as video conferencing, live broadcasting news and streaming contents are dramatically boosted to clients. Many equipment venders provide robust core/access routers and L2/L3 switches for handling the huge multimedia traffics. Yet Variable-Bit-Rate (VBR) compressed video exhibits significant rate variability, even when the rate is computed over time intervals as large as several minutes. This burstiness complicates the design of efficient real-time storage, retrieval, and transport mechanisms capable of achieving high resource utilization. Therefore, a mechanism that can effectively reduce the rate variability of video data and timely send the accurate traffic information of video stream to network facilities is very important for effective resource management. On the other hand, the Resynchronization Marking (RM) scheme proposed in most video coding standards such as MPEG-4 is generally used to improve the error resilience capability of video data in the wireless network. However, if the RM scheme at the application layer is executed independently, the effective throughput of WLAN may be decreased. This phenomenon is particularly true if the multi-layer header overhead increases or the wireless channel condition changes obviously. Therefore, this dissertation proposes an Integrated End-to-End Video Delivery System (IE-VDS) that can provide robust and efficient delivery quality for pre-stored, live broadcasting and real-time videos over networks. IE-VDS includes four major parts, the Optimal Packet size Determination Mechanism (OPDM), the ρ-domain Long-Span Predictor (ρ-LSP), the Intelligent ONline Traffic Smoothing mechanism (ON-ITS), and the Deterministic OFFline Traffic Smoothing mechanism (OFF-DTS). OPDM proposes a simple yet robust closed-form that can determine accurately and timely the optimal video payload length at the video sender based on the current wireless channel condition. Whenever the optimal payload length that can maximize the WLAN throughput is determined, the RM scheme adopts the optimal payload length to packetize the video data. For executing efficient online traffic smoothing, ρ-LSP predicts the bitrate of future frames with high accuracy by utilizing the linear relationship between the cumulated frame bitrates and the total number of nonzero motion vectors and nonzero quantized coefficients of encoded frames. ON-ITS then proposes an intelligent online traffic smoothing mechanism that integrates the proposed adaptive window size method with ρ-LSP to improve the flexibility and performance of online smoothing for live broadcasting and real-time video applications. On the other hand, for delivering pre-stored video contents with deterministic QoS guarantees, OFF-DTS integrates proper traffic smoothing operations with the traditional deterministic traffic modeling scheme to reduce the playback buffer demand and data rate variation. Herein the limitation of playback buffer space, the network delay jitter, the processing load of resource management, and the QoS guarantee are considered in the proposed mechanism. Analytical and simulation results verify the accuracy and effectiveness of the proposed closed-form in OPDM. The contribution effectively improves the WLAN throughput and enhances the error resilience effect of scalable video data simultaneously. From simulation results, ON-ITS effectively reduces the transmission rate variation and the peak rate demand of real-time and live video services while remaining a low playback delay. Moreover, OFF-DTS effectively satisfies the situation of insufficient playback buffer space in mobile clients while still maintaining the advantages of deterministic services for pre-stored videos. |
