針對平行串接迴旋碼最大概度解碼演算法的有效率估算位元錯誤率技術

Abstract

在本論文中，我們使用Importance Sampling [1, 2, 3]這個模擬技術來做渦輪碼(turbo code)最大概度(maximum-likelihood)演算法解碼 [4]的效能測量。不像Monte Carlo [5]模擬法為了達到可接受的效能需要粗魯地用到相當次數的試驗，Importance Sampling這個模擬技術相對來說是相當有效率的。前題是我們要找出合適的模擬環境及選擇正確的錯誤事件來確保最有效率的模擬結果。我們也觀察到原本因為蒙地卡羅模擬時間太長而放棄的區域的表現。另外從模擬的結果得知，效果最好的模擬環境表現不俗，推究其原因和其模擬通道與數學上計算出來的最佳化通道特性相似。

In this thesis, we applied the Importance Sampling (IS) method [1, 2, 3] to estimate the performance of the MLSDA-turbo decoder proposed in [4]. Unlike the brutal force Monte Carlo (MC) simulation [5], which often requires a very large number of simulation trials to achieve meaningful estimate of system performance, the IS technique can provide an efficient estimate with less trials. The success of the IS technique relies on the proper choice of the biased channel. Hence, the first task of the thesis is to find the right biased channel (or equivalently, the suitable error events) to ensure the efficiency of the IS simulation. By the simulation results, the best IS simulation among the channels we tried in this thesis yielded a moderate improvement in performance estimation. After the selection of the right biased channel, we investigated the time-unreachable performance region (of the MLSDA-turbo decoder) for MC simulation. We conclude that the best biased channel among all simulation channels we used indeed has the channel property similar to the mathematically optimal channel for IS technique.