在無線網路中為大的I型式畫面所設計之具有型式判別能力的錯誤控制

Abstract

針對互動式影像服務我們提出了要傳送不同大小影像畫面的錯誤控制. 錯誤控制的行為例如重傳和增加冗餘碼都被考慮. 為了得到低延遲和小延遲差距, 我們選用的SFQ排程器是一個適合處理具有變動位元流量的影像應付的排程器. 當嚴密的檢查問題, 我們發現大的I型畫面, RS碼以及影像畫面之間短暫的相互關連性這些要素深深地影響畫面失去率,有效率地處理它們是非常的具挑戰性的. 在這篇論文裡, 我們延續了之前針對固定大小畫面, 基於馬可夫的錯誤控制的成果來做. 在我們的模擬裡, 在寬頻考量下, 將通道特徵化成兩個狀態的模型, 且狀態的停留時間是根據二階Erlang分佈. 我們的數值模擬展示了我們提出的方法在所有模擬設定降低了失去畫面的比率. 除此之外, 我們的數值模擬有效的降低了重傳次數和使得碼率多元化.

We propose an error control for delivering video frames of different sizes for interactive video service. The error control actions such as retransmission and adding redundancy codes are considered. For the purpose of low delay and small jitter, we use start-time fair queuing which is a suitable scheduling algorithm for video applications that have variable bit rate. When examining closely, we find that factors of the large I-frame, redundancies of RS-code, the temporal correlation between video frames significantly impacts the frame loss rate but dealing with them effectively is very challenging. In this thesis, we extend the early work of MDP-based error control for fixed size frames to do so. In our simulation, characterizing the channel by a two-state model where state durations follow the Erlang distribution of second order under wideband is considered. Our simulation shows that our proposed method reduces frame loss rate in all the simulation settings. Besides, it effectively reduces retransmission times and makes code-rate diversity.