基於猛禽碼架構之通道編碼結合與參數最佳化

Abstract

可調式視訊編碼能夠將視訊內容編碼為許多層，每一層具有不同的重要性，為了在不可靠的通套上傳輸這樣的資料結構，我們需要使用擁有不均等錯誤保護特性的通道編碼。在這篇論文中，我們基於Raptor Codes的架構，結合Convolutional LDPC Codes和LMRC創建了新的編碼，並為此編碼設計一個參數最佳化流程。為了完善這個最佳化流程，我們需要一系列可靈活調適碼率的Convolutional LDPC Codes，但現有的規律方法只能提供我們少數的選擇，因此我們以非規律的方式改善它，使其能夠滿足我們的需求。有了改良後的非規律方法，就能擁有足夠多的碼率來進行最佳化，我們從小型的實驗中歸納出一些規則來幫助我們降低參數排列組合的數量，並基於這些規則來設計最佳化流程，使我們的編碼在有限的資源下可以提供最好的不均等錯誤保護。

The transmission of video content over network is an important issue nowadays, especially for Scalable Video Coding (SVC) content. The SVC encode the video into different layers which can be seen as different importances. To trasmit such data structure through unreliable channel, we use the technology of Unequal Error Protection (UEP) channel coding to encode the data with different importances. To more efficiently protect those data, we construct a new code based on structure of Raptor Codes with two individual codes, convolutional Low Density Parity Check (LDPC) codes and Layer-aligned Multipriority Rateless Codes (LMRC). We design a procedure to optimize proposed UEP Raptor Codes. In this optimization, however, we need sufficient amount of rates of convolutional LDPC codes to be the parameter choices. The exist regular method can generate a series codes with different code rates from a base matrix. However, it just provided a few number of choices. So we developed an irregular version based on regular method. With help of the proposed irregular method, we have a lot of rate choices in optimization. We designed some experiments to help us conclude some rules which can reduce the number of parameter permutations we need to search. We build an optimization procedure utilized those rules to find the best parameter setting of UEP Raptor Codes, and provide a best UEP performance under a given resource limitation.