適應性盲蔽等化和干擾消除:方法與性能分析

Abstract

首先，在代價函數與搜尋方式方面，本文提出以高階統計為基底的方法。 第一種是以快速濾波器為架構的最小均值平方-牛頓法，第二種則加入GAL 濾波器於LMS演算法之中。這些演算法的運算量只與濾波器的長度成正比 ，且其穩態均值平方誤差均小。我們同時也討論證明參數的收斂情形與預 測MSE的軌跡變化。在步進值方面，為了達到收斂速度快與穩態誤差小的 目標，文中提出以MSE與ISI為指標的可變步進值的盲蔽演算法。在濾波器 的結構方面，由於IIR濾波器比FIR濾波器具有較好效能與節省運算量的優 點，我們提出三種以IIR濾波器為基礎的盲蔽等化器結構。除了雜訊與符 際干擾之外，無線通訊所傳送的寬頻訊號也可能被其它同通道的窄頻使用 者或任何其它無線電波所干擾。雖然展頻通訊本身對干擾具有抑制作用， 但是如果在解展頻之前即對干擾加以清除，對系統效能的提昇將有莫大助 益。本文的最後一部份對直接系列系統提出一具體可行的方法。我們假設 通道可視為一FIR濾波器，而干擾形態則為CW或由白高斯雜訊所驅動的AR 過程。我們所提的窄頻干擾去除技術是以PN解展頻輸出為基礎所導出，所 以易於實現且可對付符際干擾。我們並就新方法與現存之LMS方法分析其 穩態誤差並預測其錯誤率。 Our first approach involves the second order only. More Specially, we design a filter which has the effect of self- orthogonalizing and whose structure is similar to that of the fast transversal filter. The second approach, considering both the first and the higher-order statistics. We also propose the use of three forms of infinite impulse response filters with HOC-based cost function and study their convergence properties. Our third approach emphasizes the step size. It is well known that a large step size in the transient period and a small one in the convergence period will yield small steady state error while providing fast convergence rate. Finally, we deal with a more complicated situation where, in addition to thermal noise and ISI, the transmitted direct-sequence spread spectrum signal suffers from distrubance. We provide a workable solution. It is simple to implement and is capable of removing ISI. Moever, both simulation and analysis point out that our algorithm does render better MSE and bit error probability performance.