適用於多輸入多輸出系統之低複雜度K-Best 球體解碼演算法

Abstract

這篇論文中，我們在維持和傳統K-Best球體解碼演算法及最大概似偵側(ML detection)相近的效能的前提下提出了兩個化簡K-Best 球體解碼演算法的方法。其中可變動式K-Best 球體解碼演算法提供利用接收訊號來決定K值大小的方式。 而分群式K-Best球體解碼演算法利用接收訊號的統計特性僅僅需要粗略排序的比較器就可以替換運算複雜的排序電路。藉由 4x4 64-QAM的系統模擬，位元錯誤率(BER)訂在5x10-4 的條件下與傳統的64-Best 球體解碼演算法做比較，使用可變動式K-Best 球體解碼演算法可以化簡23.65% 到 52.22% 的計算複雜度，並且僅造成 0.13dB到1.18dB的效能衰減。使用分群式K-Best球體解碼演算法可以化簡計算複雜度超過99%，並且僅造成0.09dB的效能衰減。

In the thesis, two low-complexity techniques of K-best SD algorithm are proposed while remain similar performance to conventional K-best SD algorithm and ML detection. Adaptive K-Best SD algorithm provides a means to determine the value K according to the received signals. Clustered K-Best SD algorithm uses the statistics knowledge of the received signal, and the clustering technique replaces the high complexity of the sorter with a few comparators. As compared with conventional 64-Best SD algorithm for 4x4 64 -QAM system, the adaptive K-Best SD algorithm can reduce complexity ranges from 23.65% to 52.22% within 0.13dB and 1.18dB performance degradation, whereas the clustered K-Best SD algorithm can reduce over 99% complexity within 0.09dB performance degradation.