適應性決策迴授廣義旁波帶消除器：效能分析與應用

Abstract

適應性廣義旁波帶消除器（GSC）被廣泛應用在波束形成之抗干擾技術上，由於內在構造的特性，其收歛頗為緩慢，同時，亦對於模型錯配頗為敏感，當發生模型錯配時，會造成訊號抵消（signal cancellation）而令效能嚴重下降，這些問題影響了適應性廣義旁波帶消除器於時變系統之應用，並使其實現更為複雜。在本論文中，我們提出一新技術能有效地對付這些問題。主要的構想是於廣義旁波帶消除器中加入一單權重決策主導等化器（decision-directed equalizer）和一單權重迴授濾波器（feedback filter），而成為決策迴授廣義旁波帶消除器（DFGSC）。當中我們選用最小平方法（LMS）以適應性地實現此方案，並對其收歛特性作全面之分析，這樣證明了其收歛速度會大大地得到改善且訊號抵消現象能避免發生。

於無線通訊中，同頻干擾（CCI）和符元干擾（ISI）為影響系統效能之兩大因素。一般可以使用波束形成器（beamformer）以消除同頻干擾和等化器以消除符元干擾，兩者結合可被稱為時空等化器（STE）。這種裝置通常需要利用訓練序列（training sequence）對其加以訓練後才可以被使用，但在一些特定應用上，接收用戶的空間資訊是可以得知，反而訓練序列不一定存在。在此我們對決策迴授廣義旁波帶消除器進行推廣，提出一適應性決策迴授時空等化器以對付這種情況。我們的方案包含了一適應性決策迴授廣義旁波帶消除器、一盲目決策迴授等化器（blind DFE）和一通道估測器。由於其特殊之構造，此盲目決策迴授等化器可藉由通道估測之輔助，從而有效地對抗決策迴授等化器所造成的錯誤蔓延（error propagation）現象。當中我們也是使用最小平方法適應性地實現此方案，同時對其收歛特性作出分析。實驗模擬證明了我們所提出之適應性決策迴授時空等化器即使處於苛刻的通道環境中亦能有效地對抗同頻干擾和符元干擾。另外，此通道輔助盲目決策迴授等化器之架構也可以應用於適應性之最小均方誤差（MMSE）決策迴授等化器中而成為適應性通道輔助決策迴授等化器（ACA-DFE）。我們亦會證明此通道輔助決策迴授等化器亦適用於多輸入多輸出（MIMO）系統而使其效能優於一般適應性多輸入多輸出之決策迴授等化器。

適應性之平行式干擾消除（PIC）技術近年亦被用於多輸入多輸出系統之訊號檢測。傳統之平行式干擾消除技術使用最小均方誤差之法則來進行優化，但最小均方誤差並不能達致最小位元錯誤率（MBER）之效果。另外，當運作於時變環境中，其效能會深受錯誤蔓延之影響。最後，於本論文中我們提出一基於最小變異（MV）法則之適應性二階段（two-stage）平行式干擾消除檢測法以解決這些問題。其中我們使用決策迴授廣義旁波帶消除器以適應性地實現此演算法。在第一階段之消除時，我們提出使用一雙重的決策迴授廣義旁波帶消除器（dual-DFGSC）架構使其更適用於時變之通道環境。由於第一階段之良好效能，第二階段只需使用一簡單之匹配濾波器便可達接近最優之效能。我們亦選用最小平方法以適應性地實現此方案並分析其收歛特性。實驗模擬證明此適應性二階段平行式干擾消除檢測法之效能大大地優於傳統基於最小均方誤差之平行式干擾消除檢測法。

The adaptive generalized sidelobe canceller (GSC) is a commonly used device for interference cancellation in array beamforming. However, due to its inherent structure, its convergence is slow. Also, it is sensitive to model mismatch. When model mismatch exists, a phenomenon called signal cancellation will occur, and the performance of the adaptive GSC can be seriously affected. These problems limit the use of adaptive GSC in time-variant systems and also complicate real-world implementations. In this dissertation, we propose a new scheme that can effectively solve these problems. The main idea is to introduce a single-tap decision-directed equalizer and a single-tap feedback filter in the GSC structure, resulting in a decision feedback GSC (DFGSC). The least-mean-square (LMS) algorithm is used for adaptation and the convergence behavior of the adaptive DFGSC is fully analyzed. It is shown that the convergence rate can be greatly enhanced and signal cancellation can be completely avoided.

In wireless communications, co-channel interference (CCI) and inter-symbol interference (ISI) are two main factors limiting the system performance. Conventionally, a beamformer is used to reduce CCI while an equalizer is used to compensate ISI. These two devices can be combined into one named space-time equalizer (STE). A training sequence is usually required to train the STE prior to its use. In some applications, however, spatial information corresponding to a desired user is available, but the training sequence is not. Extending the DFGSC approach, we then propose an adaptive decision feedback STE to cope with this problem. Our scheme consists of an adaptive DFGSC, a blind decision feedback equalizer (DFE), and a channel estimator. With a specially designed structure, the proposed blind DFE, aided by the estimated channel, can better resist error propagation effect inherent in a DFE. As previously, adaptation operations are implemented with the LMS algorithm and convergence analysis is given as well. Simulations show that the proposed adaptive decision feedback STE is effective in mitigating both CCI and ISI even in severe channel environments. The channel-aided blind DFE structure can be further applied to the general adaptive minimum mean-squared-error DFE (MMSE-DFE), called the adaptive channel-aided DFE (ACA-DFE). We also demonstrate that the ACA-DFE can be extended to multiple-input multiple-out (MIMO) systems and it can outperform the conventional adaptive MIMO MMSE-DFE.

Adaptive parallel interference cancellation (PIC) has been recently introduced in the signal detection of MIMO systems. Conventional PIC uses the MMSE criterion for parameter adaptation. However, it is known that the MMSE criterion cannot achieve the minimum bit-error-rate (MBER). Also, it suffers from the error propagation problem when operated in time-variant channels. In the last part of the dissertation, an adaptive two-stage PIC detection scheme with the minimum variance (MV) criterion is proposed to solve the problems. Adaptation with the MV criterion is then realized with the DFGSC. In the first-stage cancellation, a dual-DFGSC configuration being effective in time-variant channel environments is developed. Due to the good performance of the first-stage processing, only matched filtering is required in the second stage to achieve near optimum results. The LMS algorithm is employed and its convergence behavior is also examined. Simulation results show that the proposed two-stage DFGSC-PIC detection significantly outperforms the conventional MMSE-PIC detection.