應用於多天線偵測之叢集演算法

Abstract

在這篇論文我們學習多層級叢集為基礎的多輸入多輪出偵測演算法，並且應用在多輸入多輸出系統和多輸入多輸出正交分頻多工系統。在偵測的過程中，以非線性的方法來降低搜尋空間。透過多層級的N-QAM 星狀圖架構將接收的符號分割成叢集，並且以階層式的策略來搜尋叢集，藉以降低搜尋的空間。 另一方面，我們也在多層級叢集為基礎的方法上探討降低複雜度的技術。藉由設定適當的限制，在偵測的過程中可以限制搜尋的叢集候選人數目。再進一步透過分支限界的技術來減輕歐幾里德距離的計算。這兩個方法不但可以降低搜尋空間也可以減輕運算複雜度。 最後，我們整合以上技術，提出一個具有可接受複雜度與僅次於最佳化效能的多輸入多輸出偵測演算法。

The multilevel cluster-based MIMO detection technique for the MIMO and MIMO-OFDM system was discussed in this thesis. The non-linear detection method was proposed to reduce search space in the detection process. According to the multilevel N-QAM structures, the multi-dimensional symbols were partitioned into clusters in each dimension; by this way, the cluster candidate can be searched in hierarchical strategy to reduce search space. On the other hand, the low complexity technique for cluster-based detection was addressed. With a suitable constraint, the cluster candidate number was limited in the detection process. Moreover, branch and bound technique was also applied to decrease distance metric calculation. These two techniques not only reduce the search space, but also lower the computation complexity. Finally, the above techniques were integrated to design the MIMO detection algorithm with acceptable complexity and sub-optimal performance.