應用擴散訊息機制於低密度奇偶檢查碼排程解碼之研究

Abstract

由於低密度奇偶檢查碼具備強大的錯誤更正能力，故在眾多通訊標準中目前皆有採用此碼做為保護傳輸資料之用。而在解碼端最常被使用的是運用和積演算法來幫助解碼。在一些研究中發現傳統動態排程可以得到比和積演算法更快的收斂速度或得到更好的錯誤率。 在傳統動態排程中根據差值的形式可分為兩類，一類是以訊息的穩定度為考量的排程；另一類是以訊息可靠度為考量的排程。以此兩類為基礎的動態排程中每次子疊代都會從全部的訊息中來挑選訊息進而更新，但在本篇論文中我們發現在疊代的過程中一旦找到不穩定的位元則應該優先對此位元做更新的動作，因此提出擴散的概念。由實驗結果顯示將擴散的想法加入傳統動態排程中對於收斂速度或錯誤率有改善。

Since low-density parity-check (LDPC)codes have the powerful capability of error correction, they have been widely adopted for data protection in many communication standards. The sum-product algorithm is often used for decoding LDPC codes. In some researches, it has been found that dynamic scheduling can result in faster convergence speed or better error rate performance than the sum-product algorithm. There are two kinds of dynamic scheduling based on different forms of residual; one is about the stability of message , and the other one is about the reliability of message. In these two kinds of dynamic scheduling, the message which has the largest residual will be picked up for updating in each sub-iteration. The purpose of this study is to propose a greedy way of belief propagation in the process of sub-iteration. By experimental results, our research found that adding the idea of greedy way belief propagation in dynamic scheduling can improve the convergence speed or error rate performance compared with the conventional designs.