此篇論文提出一個創新的最佳接收機，此接收機能夠在空間相關性的Nakagami-m 衰落通道下以最大比率結合 (Maximal-Ratio Combining , MRC)接收時，其時間平均的訊雜比(Signal-to-Interference-plus-Noise Ratio , SINR)會有所增加，透過帶有對共變異數矩陣進行特徵值分解的主成分分析(Principal Component Analysis , PCA)，可以得到一組完整的標準正交基底，之後對其進行去相關過程並分析，然後將接收到的訊號投影到基底函數所跨越的空間中，根據不相關理論(Theorem of Irrelevance)，我們提出一個名為主成分選擇組合的方法(Principal Component Selection Combining , PCSC)，此方法可以在少數維度中移除訊雜比較低的分量，並在隨後的最大比率結合接收提高時間平均的訊雜比，此方法還可以在雜訊與不同分支具有相關性時，可以避免雜訊增強。為了印證此篇論文中所提出的最佳接收器其性能，我們對其訊雜比的分佈，振幅穿越速率（Level-Crossing Rate ,LCR）和平均衰落持續時間（Average Fade Duration ,AFD）進行模擬，而我們所模擬的模型是基於創新的散射模型，根據分析評估出來的協方差矩陣，開發出以空間相關的Nakagami-m衰落通道所組成的模擬環境，並透過電腦模擬可以看出此篇論文所提出的最佳接收器器不僅減少了不相關子空間的干擾和雜訊，還實現在時間平均下更高的SINR和更低的AFD，而且還大大降低了後續訊號處理所需的複雜度。;This paper proposes a novel optimum combiner that raises the time-averaged signal-to-interference-plus-noise ratio (SINR) for maximal-ratio combining (MRC) reception over spatially correlated Nakagami-m fading channels. By exploiting the principal component analysis (PCA) with eigenvalue decomposition (EVD) of the covariance matrix analytically evaluated offline, a complete set of orthonormal basis functions can be obtained. A decorrelation process analyzes and then projects the received signal into the space spanned by the basis functions. In accordance with the theorem of irrelevance, a principal component selection combining (PCSC) method is proposed to remove components in a few dimensions in which SINRs are considered low to raise the resulting time-averaged SINR on the subsequent MRC reception. The proposed technique also avoids noise enhancement occurring with MRC reception in the scenario where noises on different branches are correlated. The SINR distribution, level-crossing rate (LCR) and average fade duration (AFD) are derived. Based on a novel scattering model interpretation, a simulator consisting of spatially correlated Nakagami-m fading channels is developed according to the analytically evaluated covariance matrix. Computer simulations show that the proposed optimum combiner not only reduces interference and noise from the irrelevant subspace to achieve higher time-averaged SINR and lower AFD, but also significantly reduces the complexity required for subsequent signal processing.
