低密度奇偶校驗碼之自適應分群解碼演算法

Abstract

本篇論文針對低密度奇偶校驗碼提出了一種新的分群信度傳遞解碼演算法。我們提出之演算法在每次疊代解碼時皆會更新分群狀況。為達成此目的我們提出的演算法中包含了3 個步驟。首先先找出相對可靠的變數點(variable node)，再利用這些可靠的變數點找出可靠的校驗點(check node)，最後再選出能連接到最多可靠校驗點的變數點將其分成同一群，同時進行資訊更新。所有的變數點在每一次解碼疊代必會被選中且更新。藉由模擬可知，比起現有之演算法，本論文所提出的演算法能有較佳的效能。

We propose a new group shuffled belief propagation (BP) decoding algorithm for low-density parity check (LDPC) code. In our method, the variable nodes (VNs) will be re-grouped in each iteration. First, we decide if each message sent from a VN to a check node (CN) is reliable. Then, we further identify the CN which can provide reliable information. Based on the filtering of the reliable messages, we select the VNs which are able to obtain most reliable messages in to a group. After VN updating, the above procedure will be re-performed until all VNs are updated in one iteration. Numerical results show that the proposed shuffled BP decoding outperforms the existing shuffled BP decoders for LDPC codes and performance improvements become more distinctive at higher rate.