在非理想環境下使用改良式廣義旁瓣消除器之強健式可適性波束成型技術

Abstract

LCMV (Linearly Constrained Minimum Variance)波束成型器是天線信號處理領域相當重要和基礎的技術之一，主要是利用最小化輸出功率並且限制欲接收信號指引向量的增益來做權重運算，因此具有很強的消除干擾信號能力。而廣義旁瓣消除器(Generalized Sidelobe Canceller ,GSC)是由LCMV波束成型器演變而來，兩者擁有相同的權重向量，但是GSC只需要較低的運算量即可達成，更利於現實中使用。 但是當假設的指引向量發生誤差時，會產生信號抵銷現象，使系統效能嚴重衰落。因此我們提出一個由GSC發展而來的MGSC，運用了不同的阻隔矩陣，可以有效的減小信號抵銷現象發生的可能，增加系統的強健性。因為使用了這個阻隔矩陣，使可適性權重向量的維度減小，進而減少所需的運算量。我們也同時也使MGSC擁有估計欲接收信號指引向量的能力，只需要付出較小的運算量即可。接著我們提出將現有的強健性演算法運用至MGSC架構上的做法，不僅使這些強健性演算法計算複雜度降低，還能提升效能。最後我們將MGSC與空間平均法(Spatial Smoothing Technique)結合，使之能同時對抗同調環境以及指引向量誤差等非理想效應。

LCMV (Linearly Constrained Minimum Variance) beamformer is one of the important and basic techniques in the antenna array signal processing. It aims to find the optimal solution for the weight vector by maintaining the desired array response while minimizing the array output power. Thus, the beamformer has a great ability to suppress interference and noise. The GSC (Generalized Sidelobe Canceller) derived from LCMV beamformer provides a simple way for implementing and makes the implementation of the LCMV much more efficient. However, the performance of the aforementioned beamformers is known to be degraded dramatically in the presence of steering vector error because of the signal cancellation phenomenon which leads the desired signal to be suppressed as interference. To improve the robustness, we propose a method called MGSC (modified GSC) evolved from the GSC. The MGSC uses a novel block matrix which can not only avoids the signal cancellation phenomenon but also decreases the computational complexity by diminishing the size of the adaptive weight vector. We also give the MGSC an ability to estimate steering vector and it only needs few computations. Then, we propose a way to make the MGSC combined with the existing algorithms. It can also improve the performance and decrease the computational complexity at the same time. Finally, we combine spatial smoothing technique with the MGSC to against coherent interference and steering vector error simultaneously.