應用於非前置編之正交分頻多天線系統的奇異值分解偵測演算法

Abstract

在這篇論文裡，我們提出一個全新的應用於多天線傳輸系統之奇異值分解偵測演算法。本演算法成功將傳統上限定於前置編碼系統的奇異質分解演算法，成功應用於一般性的非前置編碼多天線傳輸系統。 本演算法所提出的奇異值分解多天線偵測演算法(SVD-based MIMO Detection)，達到PER在0.08下與最大相似演算法誤差在0.5dB以內，並且與傳統之QR分解在同樣套用K-best架構下達到相同的誤差表現。 本演算法是在偵測前，將通道資訊陣列進行埃爾米特矩陣(Hermitain Matrix)轉換，並以此形式進行奇異值分解運算。最後配合以基礎等化器所估出之可能落點進行與星狀圖的偵測及調整，進而完成以寬度優先之搜尋分界的演算法。 實作於IEEE 802.11n的通訊平台上，提供4T4R、64QAM調變，在符合TGN-E所規範的通道模型中進行模擬。並與傳統之QR分解式K-best進行複雜度與擴充性比較，並且搭配矩陣相乘消除法 (Matrix Multiplication Reduction)進行各部分的複雜度降低與效能觀察。並在SVD架構下，能以規律計算之矩陣乘法換取降低矩陣分解之複雜度。

For enhancing spectral efficiency and reducing fading distortions, multiple-input multiple-output orthogonal frequency division multiplexing (MIMO OFDM) is widely adopted for recently wireless communication systems. In this work, a new SVD-based MIMO Detection using a clustering technique, namely SCMD, is proposed, which partitions the transmitted signals into clusters for deciding candidates according to minimum squared Euclidean distance metric. In additions, our SVD decomposition doesn’t require any pre-coding of transmitter to substitute for traditional QR decomposition. Through simulations in an MIMO-OFDM system with frequency-selective fading (100-ns RMS delay spreading; 15 taps), the SNR loss of the proposed SCMD method is within 0.5dB, compared with maximum likelihood (ML) detector. It also indicates the less complexity than K-Best approach (K=12) to achieving the same performance.