Abstract
當我們在設計一個無線展頻分碼多工( CDMA ) 系統時,消除窄頻干擾的技術是非常重要; 在過去,這類技術不是因為效果不彰,就是難以實現。在本論文中,我們提出一些計算/硬體複雜度( computational/hardware complexity ) 都差不多的方法來抑制無線展頻分碼多工系統的窄頻干擾,這些方法包含一個基本架構和兩個加強版。 基本架構是一個非線性預測器,它是由一個量化器和一個適應性線性濾波器所組成; 第一個加強架構引用了一個非線性內插濾波器, 而第二個加強版則是在基本架構的輸出加上一個偏移量。電腦模擬結果顯示這兩種加強版在高訊號雜訊比時( SNR )的環境下,不管使用者人數多寡,二者的性能都插不多; 一般而言,在低訊號雜訊比情況下,非線性內插濾波器的表現較好,不過在訊號雜訊比極低且使用者人數眾多的情況之下,偏移式非線性預測器則有較佳的性能。 我們所提出的這兩種方法的性能比傳統線性濾波器優越甚多,而且計算/複雜度增加不多;當與近似條件平均( ACM ) 非線性濾波器比較時, 其在高訊號雜訊比的情況下亦具有相當之性能, 不過當使用者人數增加很多或訊號雜訊比極低時,性能有些許的衰減,然而,我們所提出的架構在計算/複雜度方面卻有很大的優勢。 若我們考慮這些干擾消除器的實現難易度時可以發現,傳統線性預測器最容易; 而近似條件平均非線性濾波器最難以實現;我們所提出之架構僅次線性預測器,也是很容易實現。Narrowband interference cancellation or suppression isimportant in design of a wireless spread-spectrum code-division multiple-access (CDMA) system. Most relatedtechniques described previously are either withunsatisfactory performance or difficult for hardwareimplementation. In this thesis, we propose some methods withcomparable computational/hardware complexity to suppressnarrowband interference in wireless direct-sequence CDMAsystems, including a basic scheme and two performance-enhanced versions. The basic scheme is a nonlinear predictorthat consists of an (N+1)-level quantizer and an adaptivelinear filter, where N is the number of users in the CDMAsystem. The first enhanced scheme is an adaptive nonlinearinterpolator, and the second one is an adaptive nonlinearpredictor with offset outputs. Computer simulation resultsshow that these two approaches have approximately thesame performance for high signal-to-noise-ratio (SNR) cases,regardless of the number of users. However, the proposedinterpolator outperforms the proposed offset predictor formost low SNR cases, except for the case where the SNR is verylow and the number of users is large. Both approaches havemuch better performance than the conventional linearfiltering approach. As compared to the approximateconditional mean (ACM) nonlinear filter, they achievecomparable performance when the SNR is large, but involvemuch less computational/hardware complexity.