蜂巢正交分頻多工系統中之同頻干擾消除

Abstract

本論文考慮蜂巢正交分頻多工系統中之同頻干擾消除。由於嚴重的干擾，傳統的通道估計方法效果欠佳。首先我們提出兩種通道估計演算法，其分別採用了符元級干擾消除和編碼級干擾消除。配合干擾消除的機制，我們採用了資料壓縮取樣的技術來估計時域通道，如子空間追蹤 (subspace pursuit SP) 和逐步最佳子空間追蹤 (step-wise optimal subspace pursuit SOSP)。此外在接收端我們採用了渦流等化器來增強效能。當它採用基於匹配的排序連續干擾消除濾波器 (MF-OSIC) 來實施軟性解映射時，其效能可以逼近基於最小平方誤差排序連續干擾消除法 (MMSE-OSIC) 且只需要較低的計算複雜度。然而在干擾數目增加且數據速率提高的環境下，它的效果將會嚴重的下降，故我們提出了基於QR分解的連續干擾消除法和基於Cholesky分解的連續干擾消除法來解決這個問題。此外我們提出了一個基於最陡下降法 (steepest descent) 之最小平方誤差排序連續干擾消除法 (SD-MMSE-OSIC) 來降低複雜度。模擬結果可以看到其效能與最小平方誤差排序逐步干擾消除法相近，但計算複雜度降低許多。

This thesis considers co-channel interference (CCI) cancellation in reuse-1 cellular OFDM systems. Due to strong interference, conventional channel estimation methods result in poor performance. We first propose two channel estimation algorithms with symbol-level interference-cancellation (SLIC) and code-level interference-cancellation (CLIC). Synergy with interference cancellation, the compressive sensing technique, such as subspace pursuit (SP) and step-wise optimal subspace pursuit (SOSP), are used to conduct the time-domain channel estimation. To enhance performance, turbo equalization is conducted in the receiver. It has been shown that match-filter-based ordered-successive-interference-cancellation (MF-OSIC) can be used to conduct soft demapping. Its performance approaches that of minimum-means-square-error-based OSIC (MMSE-OSIC), and the required computational complexity is lower. However, its performance is seriously degraded for the scenario of high QAM modulation. We then propose QR-decomposition-based SIC (QRD-SIC) and Cholesky-decomposition based SIC (CLD-SIC) to solve the problem. In addition, we propose a steepest-descent-based MMSE-OSIC (SD-MMSE-OSIC) to reduce the complexity. Simulations show that while the computational complexity of SD-MMSE-OSIC is much lower than that of MMSE-OSIC, the performance is similar.