基於索引層級的疊代訊源通道解碼機制

Abstract

在實際的數位通訊系統，訊源編碼器的設計受限於複雜度與延遲需求，其編碼輸出參數仍有殘餘冗息存在。為了加強系統的強健性，我們在合併訊源通道解碼平台下，引入渦輪原則並據以設計一索引層級的疊代演算法。核心元件包括軟性輸出通道解碼器、軟性位元訊源解碼器以及索引交錯器三個工作單元。其關鍵在於利用訊源殘餘冗息和通道編碼提供的保護位元，在不同解碼器之間逐次交換外源訊息以累積更多可靠的判定內容。通道解碼器根據分段柵狀碼圖而發展索引層級的BCJR演算法，讓外源訊息的交換得以提升至更具效率的索引層級。模擬結果顯示，索引層級疊代解碼演算法的強健效能明顯優於位元層級的疊代解碼演算法。

This thesis presents a new method of iterative source-channel decoding (ISCD) technique. The ISCD consists of a channel decoder followed by a source decoder, which are connected with a feedback loop. The main attraction of ISCD is that reliability gains resulting from both artificial and source residual redundancy can be exchanged iteratively in a turbo-like process. The extrinsic information produced by each decoding algorithm is used to strengthen the a priori information of another decoder. However, conventional ISCD system computes the extrinsic information in bit-level and therefore the correlation of source parameters cannot be fully utilized. A new index-based ISCD is presented which consists of a softbit source decoder (SBSD) and a channel decoder based on the modified BCJR algorithm using sectionalized code-trellis. The index-based ISCD does not produce bit-level information in its decoding process and therefore improves the way of utilizing the source correlation. Simulation results show that significant improvements of error robustness are achieved by ISCD compared to conventional concatenated decoding schemes. Furthermore, the index-based ISCD outperforms the bit-based ISCD under all channel-SNR conditions.