直接序列分碼多重進接無線傳輸系統雜訊干擾抑制之研究

Abstract

直接序列(direct-sequence)分碼多重進接(code division multiple access)傳輸技術已被廣泛地討論並用於實際的無線行動通訊系統。 本論文的主要宗旨是探討在干擾源的情況下， 如何抑制並減緩干擾源的影響， 能成功地解調出直接序列-分碼進接的信號之研究，以提高系統的效能與傳輸容量，我們考慮數種通訊通道中的常見的雜訊與干擾， 針對這些雜訊與干擾， 我們提出許多的抑制方法與技術， 並以電腦模擬驗證理論的結果。 當用戶間的展頻信號波形式互相正交時或只有一個用戶的情形， 使用匹配相關器(matched correlator)解調信號即能達到最佳化的效能。然而， 在非同步的系統中， 用戶之間傳遞訊息的時序是非同步的， 如此一來， 正交性就不存在。更甚的是， 離基地台近端的用戶通常具有較高的功率， 會影響遠端用戶（功率通常較低）的信號接收。 多用戶信號解調的最佳化架構是最有可能性序列檢測器(maximum likelihood sequence detector)， 可是它的複雜度隨用戶個數增加而以指數方式遞增。 線性最小均方誤差(MMSE)接收器是一種被提出來具有次佳抵抗多用戶干擾源能力的接收器， 它的架構類似傳統的匹配相關器， 但是它的效能遠勝於匹配相關器； 它適合用於行動端的接收器， 可以經由適應性演算法則來實現。 我們採用MMSE接收器之架構為本論文所發展的接收器之主體， 發展適應性演算法則以實現此MMSE接收器。 第一部份我們考慮多用戶干擾源(multiuser interference) 與可加性白色高斯雜訊(additive white Gaussian noise, AWGN)。首先， 我們提出一個具有三個平行的有限脈衝響應(finite impulse response)濾波器的MMSE 接收器的架構， 並根據Griffiths演算法則， 發展一個以模式為基礎的Griffiths盲目適應性演算法則(model-based Griffiths' adaptation algorithm, MBGA)，以實現此接收器。 此MMSE接收器可以克服因同步過程產生的時序誤差所造成的通道不匹配的問題， 並提高輸出的信號-干擾比； 除此之外， 該接收器亦可估計此時序誤差的大小。 由第一部份的電腦模擬結果發現， 當接收器與欲被解調的用戶信號之間的時序是完全同步時， Griffiths演算法則將使接收器的濾波器參數收斂至最佳解之附近， 電腦模擬結果也顯示Griffiths演算法則的收斂速度會因用戶的個數增加或干擾源功率增加而變慢， 因此， 在第二部份我們提出一些快速Griffiths適應性演算法則(fast Griffiths' adaptation algorithms)以增進MMSE接收器的收斂速度。如同最小均方演算法（least-mean-square algorithm）一般， Griffiths適應性演算法則會因接收信號之相關矩陣的特徵值幅度( eigenvalue spread )所影響，當多用戶干擾源的功率大時或干擾源的個數多時， 特徵值幅度的問題變得嚴重，所以我們根據轉域信號處理技術 (transform domain signal processing)提出一種快速收斂的Griffiths演算法。 另外， 我們利用可調變步階(variable step-size)的觀念， 提出兩種適應性演算法。 在前面的兩個部份， 吾人假設接收器與欲被解調的用戶信號之間的時序是完全同步的（或接近同步）， 也就是必須等到時序擷取(acquisition)完成後， 才可以開始解調信號， 傳統的做法亦是如此。 本論文的第三部份我們同時考慮展頻碼的擷取與信號解調之問題， 提出一個新型的盲蔽式接收架構與演算法則， 該接收器藉由接收信號的統計特性，將時序量化成幾個假設集，透過我們所提出的演算法， 估計出展頻碼的時序； 一旦展頻碼的時序被估計出來， 此最小均方誤差接收器的濾波器係數向量即可得知。 該展頻碼的時序估計演算法毋需訓練符號序列之助， 與其它盲蔽式的演算法相較之下， 它的複雜度較低， 時序擷取的效能較佳。 在本論文的第四部份我們針對在多重路徑通道下的多重進接干擾(multiple access interference)， 提出一個非同調(non-coherent)的盲蔽式信號解調架構， 該接收器是根據 general side-lobe cancellation (GSC) 之技術而發展的，該接收器利用一平行運算的GSC濾波器組來解調某一個使用者的信號， 每一個GSC濾波器將解調出該使用者的某一個路徑之信號， 最後， 被解調出的信號會依照信號強度的大小， 以最大比例(maximum ratio)的方法結合， 得到較精準的信號檢測。 在論文的第五部份， 我們考慮當接收端前級的背景雜訊不是白色高斯雜訊的情形。 在實際的通道中， 背景雜訊可能是具有脈衝(impulsive)特性的。 首先， 我們分析多用戶干擾源與脈衝雜訊對線性信號相關連形式(correlation-type)的檢測器之效能影響， 我們發現脈衝雜訊比白色高斯雜訊對系統的效能產生更惡化的影響。 又因symmetric -stable 脈衝雜訊的平方均值是無窮大的， 傳統的最小均方誤差演算法並不能用於此雜訊環境下， 因此， 我們提出一個新型的適應性演算法， 使線性檢測器能在symmetric -stable 脈衝雜訊通道中快速地收斂。 該新型演算法的複雜度低， 且收斂速度快。

The direct-sequence code division multiple access (DS/CDMA) transmission technique has been widely investigated and used in practical wireless communication systems. In this thesis, several interference suppression schemes for demodulation of DS/CDMA signals in the presence of interference are investigated. Several kinds of interference in the communication links are considered. Both theoretical derivations and computer simulations are provided. When all the spreading waveforms are orthogonal or only one user exists, the optimal performance of reception can be achieved by the matched correlator. However, the orthogonal property cannot be easily obtained due to random timing offsets between users in an asynchronous system. A nearby interfering user of large power will deteriorate the reception of the highly attenuated signal. The optimal detector for multi-users is the maximum likelihood sequence detector, but its implementation complexity increases vastly. The linear minimum mean-squared error (MMSE) receiver, being a sub-optimal but near-far resistant one, had been presented to mitigate the multiuser interference. It has a similar structure to the matched correlator, but its performance is superior to the latter. The MMSE receiver is suitably used in a mobile terminal, in which the receiver of interest can be implemented by adaptation algorithms. In the thesis, the receivers under development are based on the MMSE receiver structure. In the first part, we consider the DS/CDMA transmission system with multiuser interference together with additive white Gaussian noise (AWGN). First, we a MMSE receiver, in which there are three parallel FIR filters. We propose a novel blind adaptation algorithm (model based Griffiths' adaptation algorithm), modified from the Griffiths' adaptation algorithm, to train this linear MMSE receiver when the channel model is mismatched by the timing error in synchronization. The proposed algorithm can increase the output signal-to-interference ratio. The new algorithm also results in the estimate of the timing error adaptively. As shown in the first part, the numerical results indicate that the Griffiths' algorithm can drive the tap-weight vector of the FIR filter used in the MMSE receiver to converge near the optimum solution when the timing synchronization between the receiver and the desired signal is perfectly achieved. In the meantime, the results reveal the convergence properties depend on the number of active users and the interference level. Therefore, we propose several fast Griffiths' algorithms to accelerate the convergence speed in the second part. The convergence properties of the Griffiths' algorithm, like the least-mean-square algorithm, depend on the eigenvalue spread of the correlation matrix of the received signal vector. When the number of interference or the interference level increases, the problem of eigenvalue spread becomes severe. Therefore, we utilize the transform domain signal processing techniques to accelerate the convergence speed of the Griffiths' algorithm. Besides, two variable step-size Griffiths' algorithms are presented. Demodulation of a DS/CDMA signal must be preceded by timing acquisition, in which the initial coarse alignment of the local symbol clock within one chip of the incoming symbol clock is achieved. That is, after the acquisition is completed, the demodulation is started. In third part, we develop an efficient algorithm for estimating the code timing of a specific desired user without requiring a training sequence. The proposed algorithm quantizes the timing delay uncertainty of interest into a finite set of hypotheses, measures the projection degree of a nominal desired signal vector in the signal subspace of the correlation matrix for each hypothesis, and picks the best hypotheses to estimate the time delay. After the code timing is estimated explicitly, the MMSE demodulator is obtained via the Wiener-Hopf equation. In the fourth part, we propose a blind interference suppression scheme for non-coherent demodulation of differentially encoded DS/CDMA signals over the multipath channels. The proposed receiver employs a generalized side-lobe cancellation (GSC) filter bank, a differential decoder bank and a diversity combiner. Each GSC filter is designed to extract the unique desired signal component from a certain path. The weights of each GSC filter are adapted individually, and a new blind adaptation algorithm is derived for tracking the time-varying environment. Combining rules can be used to form the decision statistic from the differentially-decoded output signals of all GSC filters. The MMSE linear correlation receiver is built upon the assumption that the additive ambient noise is modeled as Gaussian random process. However, the Gaussian noise model is quite inadequate for many practical applications. In some realistic communication links, some natural as well as man-made interference is impulsive with significant probabilities of large interference level. In the fifth part, the performance of the linear correlation receiver operating in the MAI and non-Gaussian noise channel modeled as a class of symmetric stable process is investigated. Because the symmetric - -stable noise has no finite variance, the traditional least mean squares algorithm cannot be used to train the receiver. A new adaptation algorithm will be developed for the soft-limiter correlation receiver.