多媒體通訊系統中之即時平順演算法

Abstract

隨著頻寬的增加與視訊壓縮技術的進步和電腦網路技術的提升,透過網路觀賞高品質即視訊服務的應用大大增多,網路必須分送不同的影像資料以滿足不同的應用.例如,包含聲音的即時影像要求、多媒體短片與電影。這類資料會使用變動位元速率的視訊壓縮方法,這種壓縮方式雖然可以保有每張圖片品質的一致,但是在傳送時會有速率突然增加的情況(burst), 這會加重網路的負擔,同時降低了網路的使用效率,也使得網路的管理更加困難。平順演順法可以將burst變成一個平緩的資料流,對整個網路的負擔會減少很多.本論文主要是延伸已有的stored video平順演算法,原本只能平順傳送已經存成檔案的影音資訊外,現在也能夠處理即時產生的資料.我們一開始先回顧基本的平順演算法,說明其運作方式及特性,接著再探討即時資料和已存好的資料在平順處理上有何差異,針對這些不同之處改進原有的演算法.並且實際將演算法實作在一個傳輸系統上,原始的平順演算法必須因應實作的需求和限制而有修改,才能相容於以封包為單位的傳輸系統.最後我們對系統所使用的無線網路做了關於延遲及jitter等的量測,並從MAC的層面解釋網路對平順演算法的結果有哪些影響.

With rapid increases in network bandwidth and major advances in video compression technology, the number of the applications that provide consumers with high quality audiovisual contents has gone up dramatically. Hence, the network now has to support the transmissions of different types of data, such as real-time conferences, multimedia clips, and Hollywood movies. These different types of data, however, do have a common characteristic, “VBR”, which results from modern compression techniques to ensure that the quality of the contents maintains the same throughout the entire presentation. Since the bitrate of VBR contents is not constant, the traffic generated when transmitting them will obviously be bursty. This burstiness will add burdens to the network, decrease the utilization efficiency of the network resources, and complicate the task to manage the network. To solve this problem, the smoothing technique that can smooth bursty traffic in effect and hence remove some of the burdens caused by VBR contents has been proposed and developed. In this work, the stored video smoothing technique that originally can only generate a smoother plan for stored audiovisual contents is extended. The proposed online smoothing technique can deal with real-time captured data with only minor buffering. In the beginning of the paper, the general smoothing techniques are reviewed, and their mechanisms and characteristics are illustrated. Subsequently, the major differences that stored and real-time captured contents can make to the smoothing algorithms are discussed. With the discoveries of these differences, how the original smoothing technique is changed so it can adapt to the online case is presented. Moreover, this extended algorithm is further modified so it can be implemented in a real transmission system. Specifically, since in the real world, the data are transmitted in packets but not in sequential bytes, the smoothing algorithm has to be changed to packet-based. Finally, after the technique is implemented in the real system, measurements of delays and jitters in the wireless environment are taken. Their effects to the smoothing technique are explained in terms of the mechanisms of MAC layer.