使用於多媒體串流之重疊無限制比率前向糾錯碼分析模型

Abstract

多媒體串流的資料是有可能源源不絕的，而現有的TCP重新傳送機制來傳送封包，會使得時間會有延遲。使用前向糾錯碼來取代TCP傳送多媒體串流資料是一個很好的選擇。在傳送資料前，我們無法得知通道的封包遺失率，所以無限制比率前向糾錯碼很適合應用於此環境，再加上此前向糾錯碼的編碼與解碼速度相較快於Reed-Solomon碼，使用此前向糾錯碼來傳送多媒體串流的資料是有利的。對於多媒體串流資料，我們先提出一種分析模型於一小段時間內的資料上，來分析有權重的選擇編碼方式，可否能達到更低的解碼失敗率。在有此分析模型之後，放到整個多媒體串流來預估怎樣的權重序列，才會使得整體的解碼失敗率降低。最後用實驗來驗證有權重方式的多媒體串流會優於無權重方式。

The amount of the multimedia streaming data could be endless. If we use TCP scheme to transmit, it may cause long delay due to possible retransmission. Forward Error Correction (FEC) codes are very suitable candidates. Before transmitting the encoded packets over the error prone channel, we do not know the packet loss rate. So the rateless codes could apply in this environment with its rateless feature, and the complexity of the rateless codes’ encoding / decoding is much less than the well-known Reed-Solomon codes. For the multimedia streaming, we propose an analytic model of the unequal overlapped rateless codes on a short time data. From the simulation results, the weighted selection could achieve lower decoding failure probability. With this analytic model, better encoding strategy for the multimedia streaming could be formed to ensure low decoding failure probability across multiple sections. The experiment results show that the weighted selection is better than the uniform selection in the multimedia streaming.