1257735 玖、發明說明： 發明所屬之枝術頜域 本發明是有關於一種線性陣列天線的波束成形(Beamforming) 架構， 且特別是有關於一種階層式線性陣列天線 波束成形架構。 先前技術 本發明相關先即之波束成形技術(Beam-forming)可分 爲以下三大類，如下表1: 表1.波束成形技術彙整表[1] [2；| MMSE Max SNR LCMV 基準 將陣列的輸出與期望 響應(desired response) 間的誤差縮減到最 小。 將目標訊號功銮辯雜訊 藉由一個線性限制條 件，將陣列的輸出差異 最小化。 功率的比率最大化。 主値函數 (Cost function) J(w) = E[\y(t)-d(t)\2] 此處MO是陣列的輸 出，而作)是期望響 應（desired response) 〇 ^⑼= 此處'是 w Rsw 表示雜訊的變異炬陣 (covariance matrix )，而 A是目標訊號的變異矩 W (covariance matrix) 〇 = 接受一個線 性限制條件心)=容. ，當g = 1，這演算法被稱 爲 MVDR (Minimum Variance Distortionless Response)波束產生 器。 最佳化解 w = R~lp itM R-E[u(t)uH(t)] 而 p =你(W(〇] ’ 吨)表示輸入資料向 量。 此處是最大的A特 徵値(eigen value)〇 w = R-lc[cHrlcTlg 此處c = (彳)是限制方向 的操控向量。 優點 不需要DOA的資訊。 具有最大的SNR 〇 一般化的限制技術。 缺點 需凄產生參考訊號。 必須知道目標訊號的 DOA資訊，與雜訊的統 計資料。 必須知道目標訊號的 DOA資訊。 上表中的MMSE(最小均平方誤差，Minimum Mean Square Error)演算法是利用縮小輸出訊號與期望響應間的 誤差來達成波束成形的技術，其優點是不需要訊號源的 11939TWF.DOC 5 1257735 來源指向（DOA: Direction of Arrival)資訊’缺點是需要模 擬產生一些參考訊號。MaxSNR是利用將訊雜比（SNR)BACKGROUND OF THE INVENTION 1. Field of the Invention This invention relates to a beamforming architecture for a linear array antenna, and more particularly to a hierarchical linear array antenna beamforming architecture. Prior Art The beam-forming technique of the present invention can be divided into the following three categories, as shown in Table 1: Table 1. Beamforming technology summary table [1] [2;| MMSE Max SNR LCMV benchmark will be arrayed The error between the output and the desired response is reduced to a minimum. The target signal is ridiculed to minimize the output difference of the array by a linear constraint. The ratio of power is maximized. Cost function J(w) = E[\y(t)-d(t)\2] where MO is the output of the array, and is the desired response 〇^(9)= Where 'w rsw denotes the covariance matrix of the noise, and A is the coefficient of variation of the target signal W (covariance matrix) 〇 = accept a linear constraint condition) = 容., when g = 1, this calculus The method is called a MVDR (Minimum Variance Distortion Less Response) beam generator. The optimal solution w = R~lp itM RE[u(t)uH(t)] and p = you (W(〇] ' t) represents the input data vector. Here is the largest A feature eigen value〇 w = R-lc[cHrlcTlg where c = (彳) is the steering vector that limits the direction. Advantages do not require DOA information. Maximum SNR 〇 generalized limiting technique. Disadvantages need to generate reference signals. Must know the target signal DOA information, and statistics of noise. Must know the DOA information of the target signal. The MMSE (Minimum Mean Square Error) algorithm in the above table is achieved by reducing the error between the output signal and the expected response. The beamforming technology has the advantage of not requiring the signal source 11939TWF.DOC 5 1257735 Source pointing (DOA: Direction of Arrival) information 'The disadvantage is that the analog signal needs to be generated. MaxSNR is to use the signal to noise ratio (SNR)
最大化來達成波束成形的技術，優點是具有訊號源最大 的訊雜比，缺點是需要訊號DOA資訊。LCMV (Lineady Constrained Minimum Variance)演算法是只需利用一個簡 單的線性條件將輸出訊號的變異最小化即可達成波束成 形的技術，其優點是使用一般簡單的線性技術即可達成’ 而不需太過複雜的數學演算法技術，缺點也是需要訊號 DOA資訊。以上三種技術各有其優、缺點與相關的限制 條件。在本發明中將使用階層式的設計架構來改善LCMV 中的 MVDR (Minimum Variance Distortionless Response ) 波束成形演算法。 LCMV中的MVDR波束成形演算法能夠將輸出的訊 號雜訊比（SNR)最大化和讓偵測方向傳送之訊號無失真。 此種演算法的主要特徵是在偵測方向上以一個限制條件 將平均的輸出功率減至最小，以取得陣列的權重向量。 因此我們可以了解到MVDR的效能評估是根據最小化演 算法的收斂速率與誤差調整率決定的。一般我們常利用 最小均平方（Least Mean square，LMS)或遞迴最小平方 (Recursive Least Squares，RLS)等演算法來最小化平均輸 出功率。但一般的LMS與RLS演算法在先天上會有些 限制與缺點，而影響MVDR波束成形演算法的整體效 能。這些限制與缺點在下一段落中我們將會做進一步說 明。 LMS是一個十分簡單的演算法，由於它使用 11939TWF.DOC 6 1257735 steepest-descent的方法更新權重向量，而此種方法只使 用前一次的均平方誤差(mean square error)來調整權重向 量，因此它有較低的計算複雜度。然而，LMS因爲具有 大的eigenvalue spread，所以它的缺點是收斂速度較於緩 慢，且當天線元素增加時，它的收斂效果也會逐漸變差。 因此，它通常被用於變化較慢的通道上(channels)。 RLS演算法是Kalman filter的一個特例，它不同於LMS 演算法，因爲LMS演算法是用steepest-descent的方法來 更新它的權重向量，而RLS是使用least square的方式來 調整權重向量。RLS演算法的重要特徵是它使用包含在輸 入訊號中的資訊，因此使得它的收斂速率比單純的LMS 要來的快速。但這相對的也使得RLS演算法的計算複雜度 增加。RLS演算法因爲具有快速的收斂率和免除eigenvalue spread的特性，是一個十分受到歡迎與重要的演算法。因 此，它通常被用於變化較快的通道上。 發明內容The technology to maximize beamforming has the advantage of having the largest signal-to-noise ratio of the signal source. The disadvantage is that the signal DOA information is required. The LCMV (Lineady Constrained Minimum Variance) algorithm is a technique for beamforming by minimizing the variation of the output signal with a simple linear condition. The advantage is that it can be achieved using a generally simple linear technique. Too complicated mathematical algorithm technology, the shortcoming is also the need for signal DOA information. Each of the above three technologies has its own advantages, disadvantages and related restrictions. In the present invention, a hierarchical design framework will be used to improve the MVDR (Minimum Variance Distortion Less Response) beamforming algorithm in LCMV. The MVDR beamforming algorithm in LCMV maximizes the output signal-to-noise ratio (SNR) and the distortion-free signal transmitted in the direction of detection. The main feature of this algorithm is to minimize the average output power in a detection direction with a constraint to obtain the weight vector of the array. Therefore, we can understand that the performance evaluation of MVDR is determined by the convergence rate and error adjustment rate of the minimum algorithm. In general, we often use algorithms such as Least Mean Square (LMS) or Recursive Least Squares (RLS) to minimize the average output power. However, the general LMS and RLS algorithms have some limitations and disadvantages in nature, which affects the overall efficiency of the MVDR beamforming algorithm. These limitations and shortcomings are further explained in the next paragraph. LMS is a very simple algorithm, because it uses the 11939TWF.DOC 6 1257735 steepest-descent method to update the weight vector, and this method only uses the previous mean square error to adjust the weight vector, so it Has a lower computational complexity. However, since the LMS has a large eigenvalue spread, its disadvantage is that the convergence speed is slower, and as the antenna element increases, its convergence effect gradually deteriorates. Therefore, it is usually used for channels that change slowly. The RLS algorithm is a special case of the Kalman filter, which is different from the LMS algorithm because the LMS algorithm updates its weight vector with the steepest-descent method, and the RLS uses the least square method to adjust the weight vector. An important feature of the RLS algorithm is that it uses the information contained in the input signal, thus making its convergence rate faster than that of a pure LMS. But this also makes the computational complexity of the RLS algorithm increase. The RLS algorithm is a very popular and important algorithm because of its fast convergence rate and the elimination of eigenvalue spread. Therefore, it is usually used on channels that change faster. Summary of the invention
有鑑於使用LMS與RLS有以上之限制與缺點，因此 本發明的目的在提出一種階層式線性陣列天線波束成形架 構，以改良LMS與RLS先天上的缺點，進而提升MVDR 波束成形演算的執行效能。 波束成形技術在智慧型天線中是一個十分重要的空間 濾波技術，這是因爲它能區隔不同來源的訊號（這些訊號 中包含了目標源訊號、干擾源訊號和雜訊），並能有效地 將不必要的干擾訊號消除，使天線系統能正確偵測到目標 源訊號。由於，此種濾波技術能區隔出不同訊號間的空間 11939TWF.DOC 7 1257735 特性並能有效地將干擾降低，使得頻率再甩率增加，因此 能大幅提升系統的容量。 在波束成形技術中，常使用到的LMS與RLS演算法 皆有其先天上之限制與缺點。因此，在本發明所提出之階 層式線性陣列天線波束成形架構改善上述演算法的能力與 降低計算複雜度，以提升波束成形的執行效能。根據本發 明實際的模擬與評估，證實導入階層式設計的演算法優於 標準的演算法。 本發明提出一種階層式線性陣列天線波束成形架構， 此架構係採「階層式」的設計架構來發展新的演算法模型。 此種多階層式MVDR線性陣列天線波束成形設計架構原 理是將Μ個天線元素在第一階層中分成兩個部份處理， 第1個天線元素被用來滿足入射角度的限制條件（louk-aiigle constraint)，目的是用來萃取真正的目標訊號。其餘 的M-1個元素被分割成群組處理，此部份的處理，目的是 用來將干擾與雜訊消除。階層式的MVDR與傳統的MVDR 間主要的差異是階層式的MVDR將剩餘的M-1個元素切 割成群組的方式處理，如此可以使得權重的收斂速率較 快，並且能夠有效地改善波束成形的能力。在每一個獨立 地群組中干擾與雜訊的輸出功率被減到最小。以上所述之 階層式架構可根據不同需求調整爲多階層處理；同時任一 層之任一群組亦不限定一定數目之天線個數。 本發明因採用「階層式」的設計架構，因此Μ個天 線元素，在第一階層中覆分成兩個部份處理，第〇個天線 元素被用來滿足訊號源入射角度的限制條件，目的是用來 11939TWF.DOC 8 1257735 萃取真正的目標訊號；其餘的Μ“個元素被分割成多個群 組處理（每個群組之天線元素個數任意），使權重之收斂速 率較快，且能夠有效地改善波束成形能力，而每一獨立地 群組中干擾與雜訊的輸出功率亦減至最小。此Μ個天線 元素及Ν階層式波束成形設計架構，最後一階層匯集第〇 個天線元素及前Ν-1層之天線群組輸入信號之演算法演算 輸出結果，再如前項所述處理及獲致一定之性能效益。在 階層式MVDR線性陣列天線波束成形架構的槪念架構裡， 由於在每一個天線群組內的天線元素個數減少，使得特徵 値(eigenvalue)的展開減小，mis-adjustment rate /zM；lev/2的 比例也被減小和七分別是代表步距step size與相關矩陣 的平均特徵値）。因此，使得權重向量的收斂速度加快， 空間響應的能力也獲得明顯改善。 爲讓本創作之上述和其他目的、特徵、和優點能更明 顯易懂，下文特舉二較佳實施例，並配合所附圖式，作詳 細說明。在圖中，當元件被指爲”連接”或”耦接”至另一元 件時，其可爲直接連接或耦接至另一元件，或可能存在介 於其間之元件。相對地，當元件被指爲”直接連接”或”直 接耦接”至另一元件時，則不存在有介於其間之元件。 實施方式 第1J»是依據本發明階層式線性陣列天線波束成形架 構所舉出之一較佳實施例所繪示之一種圖。本發明採「階 層式」的設計架構來發展新的演算法模型，如第1圖所示。 階層式設計的重要關鍵，是將第二部分長度爲M-1的天線 元素110分成許多群組處理，並將這些群組建立成邏輯上 11939TWF.DOC 9 1257735 的階層樹150,而每一個群組分別執行選定的適應性演算 法(LMS、RLS或其它演算法）。並以第1層群組120的輸 出（必須注意的是，此處的輸出是第1層輸入訊號乘上更 新過後的權重），作爲第2層群組130的輸入，而第2層 群組130的輸出則是濾波器真正的輸出。由於此種MVDR 架構原理是將天線元素分割成群組處理，所以權重向量的 收斂速度會較快，並且能使得空間響應（spatial response) 獲得顯著的改善。 第2圖是依據本發明階層式線性陣列天線波束成形架 構所舉出之一較佳實施例所繪示之一種二階層式LMS MVDR線性陣列天線波束成形架構圖，我們可在第1層220 將M-1個天線元素210分成λ/^Τ個群組，而第二層230 是由第一層220的每一個群組的輸出功率形成另一個群 組。從圖中我們可看出，％所表示的意義是指在第/層第/ 個群組的第y個天線210訊號。在每一個群組內所使用的 權重w丨’同樣也是使用以上的方式定義。在空間響應部份 則是 201og1()|w⑻咖)|定義之，而化⑷：= 风⑷，化⑻，&⑻，…，九^⑻），圮⑻它所代表的意義是天線元 素210的真正權重。而在這個階層式的方法中，每一個天 線元素210真正權重的計算方法是由第一層220天線元素 的權重與其相對應第二層230權重的乘積。例如： ^i(^) = K(n)K(n)^2(n) = ♦)=物♦)，"·，（丨㈨=^泥>^。 第1個元素屯⑻是爲了符合入射角度的限制條件， 11939TWF.DOC 10 1257735 圮⑻=w〇 = 1 -笮1命» •(彡）。 /=1 在階層式的槪念架構裡，對於階層式LMS(HierarchiCal LMS，HLMS)而言由於在每一個天線群組內的天線元素個 數減少，使得特徵値的展開減小，錯誤調節率/^aflV/2的比 例也被減小。其中//和;L分別是代表步距與相關矩陣的平 均特徵値。因此，使得權重向量的收斂速度加快，空間響 應的能力也獲得明顯改善。 對於階層式RLS(Hierarchical RLS，HRLS)而言，由於 第1層220每一個群組其收斂速率比標準RLS使用M-1 個陣列元素的收斂速率快。因此輸入到第2層230雜訊也 比非階層式低（因爲第1層220已過濾一次雜訊），所以第 2層230的收斂速度可以很快，且有較低之均平方誤差。 因此可以明顯改善標準RLS演算法的計算複雜度與收斂速 度，提升空間響應的能力。 本發明階層式線性陣列天線波束成形架構之階層式架 構可以明顯地提升MVDR演算法的波束成形能力。依照 本發明所舉之一較佳實施例，以下開始描述本實施例模擬 的模型，並展示模擬的結果。首先，我們先描述一些基本 的控制條件： 1·目標源與干擾源的入射角分別是sinloe和0。 2·操控向-4 是以 Z⑼= [l，e”'e-'···，€，，表 示，而分=π sin(0)。 3·陣列天線的元素訊號是以基頻的形式表示，如： w⑻=砗 exp(/〇 + 為 exp〇 么 + γ) + ν〇)，π = 1，2,3,4,5。此處的 4 與4分別代表的是目標訊號和干擾訊號的振幅，是相 11939TWF.DOC 11 1257735 關聯的相位變化，⑷是一個高斯雜訊値。 4·目標雜訊比（TNR)被固定控制在10dB。干擾雜訊比是可 變的，假設它們的數値分別是20, 30, 40dB。 第3圖與第4圖分別顯示的MVDR與本實施例之階 層式的MVDR在M=17和n=200條件下的空間響應，結 果我們可看出階層式的MVDR在干擾源的角度有較低的 響應，也就是說階層式的方法有較好的干擾消除能力。如 果我們將天線元素Μ增加至26，本實施例之階層式的 MVDR仍然能夠很精確地指出干擾源與訊號源的方向（如 第6圖所示），但習知的MVDR則無法做到同樣效果(見第 5圖）。第7圖、第8圖中所顯示的是當天線元素M= 17， 且iteration次數很少時(在n=25,30)，本實施例之階層式 架構比習知之架構有較好的干擾消除能力。第10圖中所 顯示的是當天線元素Μ增加時，本發明之階層式架構仍然 能夠在較少的iteration次數條件下，將干擾消除。在本實 施例中，例如:我們將Μ增加至26時，且在非常少的iteration 數目時就將干擾消除(《 = 20,25,30)，但習知的MVDR則無法 做到(如第9圖所示）。 雖然本發明已以一較佳實施例揭露如上，然其並非用 以限定本發明，任何熟習此技藝者，在不脫離本發明之精 神和範圍內，當可作些許之更動與潤飾，因此本發明之保 護範圍當視後附之申請專利範圍所界定者爲準。 圖式簡單說明 第1圖是依據本發明階層式線性陣列天線波束成形架 11939TWF.DOC 12 1257735 構所舉出之一較佳實施例所繪示之一種多階層式MVDR 線性陣列天線波束成形架構圖。 第2圖是依據本發明階層式線性陣列天線波束成形架 構所舉出之一較佳實施例所繪示之一種二階層式LMS MVDR線性陣列天線波束成形架構圖。 第3圖是習知之適應性MVDR波束成形器 (beamformer)的空間響應，其干擾雜訊比（11^价代1^64〇-noise ratio)=20，30, 40dB.(” = 200，M = 17) 〇 第4圖是本發明階層式線性陣列天線波束成形架構之 較佳實施例中，階層式的適應性MVDR波束成形器的空 P奇響應，其干擾雜訊比=20, 30, 40 dB· (” = 200，M = 17)。 第5圖是習知之適應性MVDR波束成形器的空間響 應，其干擾雜訊比=20, 30, 40 dB.(n = 200，M = 26) 第6圖是本發明階層式線性陣列天線波束成形架構之 較佳實施例中，階層式的適應性MVDR波束成形器的空 間響應，其干擾雜訊比=20,30,40 (18.(/2 = 200，从=26) 第7圖是習知之適應性MVDR波束成形器的空間響 應，其干擾雑訊比=20 dB.(w = 20,25,30，M = 17) 第8圖是本發明階層式線性陣列天線波束成形架構之 較佳實施例中，階層式的適應性MVDR波束成形器的空 間響應，某于擾雜訊比=20 dB· 〇 = 20,25,30，M = 17) 第9圖是習知之適應性MVDR波束成形器的空間響 應，其干擾雜訊比=20dB和= 20,25,30，M = 26) 第10圖是本發明階層式線性陣列天線波束成形架構 之較佳實施例中，階層式的適應性3^¥£)11波束成形器的 11939TWF.DOC 13 1257735 空間響應，其干擾雜訊比=20 dB· (n = 20,25,30, 圖式標記說明= 110、210 :天線元素 120、220 ··第一階層群組 130、230 ··第二階層群組 140 :最後階層群組 150 :階層式線性陣列天線波束成形架構 11939TWF.DOC 14In view of the above limitations and disadvantages of using LMS and RLS, the object of the present invention is to propose a hierarchical linear array antenna beamforming architecture to improve the inherent disadvantages of LMS and RLS, thereby improving the performance of MVDR beamforming calculation. Beamforming technology is a very important spatial filtering technique in smart antennas because it can separate signals from different sources (these signals contain target source signals, interference source signals and noise), and can effectively Eliminate unnecessary interference signals, so that the antenna system can correctly detect the target source signal. Because this filtering technology can distinguish the space between different signals 11939TWF.DOC 7 1257735 and can effectively reduce the interference and increase the frequency re-frequency, it can greatly increase the capacity of the system. In beamforming technology, the commonly used LMS and RLS algorithms have their inherent limitations and shortcomings. Therefore, the layered linear array antenna beamforming architecture proposed by the present invention improves the capability of the above algorithm and reduces the computational complexity to improve the performance of beamforming. According to the actual simulation and evaluation of the present invention, it is confirmed that the algorithm for introducing the hierarchical design is superior to the standard algorithm. The invention proposes a hierarchical linear array antenna beamforming architecture, which adopts a "hierarchical" design architecture to develop a new algorithm model. The multi-layer MVDR linear array antenna beamforming design architecture principle is to divide two antenna elements into two parts in the first level, and the first antenna element is used to meet the constraint of the incident angle (louk-aiigle) Constraint), the purpose is to extract the real target signal. The remaining M-1 elements are divided into group processing, which is used to remove interference and noise. The main difference between hierarchical MVDR and traditional MVDR is that hierarchical MVDR processes the remaining M-1 elements into groups, which can make the weight convergence rate faster and effectively improve beamforming. Ability. The output power of interference and noise is minimized in each of the independent groups. The hierarchical architecture described above can be adjusted to multi-level processing according to different requirements; at the same time, any group of any layer does not limit the number of antennas. The invention adopts a "hierarchical" design architecture, so that one antenna element is divided into two parts in the first layer, and the second antenna element is used to satisfy the limitation condition of the incident angle of the signal source. Used to extract the true target signal from 11939TWF.DOC 8 1257735; the remaining Μ "elements are divided into multiple groups (the number of antenna elements in each group is arbitrary), so that the convergence rate of the weight is faster and can Effectively improve beamforming capability, and the output power of interference and noise in each independent group is also minimized. This antenna element and Ν hierarchical beamforming design architecture, the last level gathers the second antenna element And the algorithmic output of the antenna group input signal of the front Ν-1 layer, and the processing and the performance benefit obtained as described in the previous item. In the mourning structure of the hierarchical MVDR linear array antenna beamforming architecture, The number of antenna elements in each antenna group is reduced, so that the expansion of the feature eigenvalue is reduced, and the ratio of mis-adjustment rate /zM;lev/2 is also reduced. And seven are the average features of the step size and the correlation matrix respectively. Therefore, the convergence speed of the weight vector is accelerated, and the ability of the spatial response is also significantly improved. To make the above and other purposes, features, and The advantages of the invention will be more apparent from the following description of the preferred embodiments. It may be directly connected or coupled to another element, or there may be intervening elements. In contrast, when the element is referred to as being "directly connected" or "directly coupled" to another element, there is no An intervening component. Embodiment 1J» is a diagram illustrating a preferred embodiment of a hierarchical linear array antenna beamforming architecture according to the present invention. The present invention adopts a "hierarchical" design architecture. Develop a new algorithm model, as shown in Figure 1. An important key to hierarchical design is to divide the second part of the antenna element 110 of length M-1 into a number of group processes, and to build these groups into a hierarchical tree 150 of logically 11939TWF.DOC 9 1257735, and each group The group performs the selected adaptive algorithm (LMS, RLS or other algorithms) separately. And the output of the first layer group 120 (it must be noted that the output here is the layer 1 input signal multiplied by the updated weight) as the input of the layer 2 group 130, and the layer 2 group The output of 130 is the true output of the filter. Since the MVDR architecture principle divides the antenna elements into group processing, the weight vector converges faster and can achieve a significant improvement in the spatial response. FIG. 2 is a schematic diagram of a two-layer LMS MVDR linear array antenna beamforming architecture according to a preferred embodiment of the hierarchical linear array antenna beamforming architecture of the present invention. The M-1 antenna elements 210 are divided into λ/^ groups, and the second layer 230 is formed by the output power of each group of the first layer 220 to form another group. As can be seen from the figure, the meaning represented by % refers to the signal of the yth antenna 210 in the /th group/th group. The weight w丨' used in each group is also defined in the above manner. In the spatial response part is 201og1()|w(8) coffee)|definition, and (4):= wind (4), (8), & (8), ..., nine ^ (8)), 圮 (8) which means the antenna element The true weight of 210. In this hierarchical approach, the true weight of each antenna element 210 is calculated by the product of the weight of the first layer 220 antenna element and its corresponding second layer 230 weight. For example: ^i(^) = K(n)K(n)^2(n) = ♦)=object ♦),"·,(丨(九)=^泥>^. The first element 屯(8) is In order to meet the constraints of the angle of incidence, 11939TWF.DOC 10 1257735 圮(8)=w〇= 1 -笮1命» •(彡). /=1 In the hierarchical structure, for hierarchical LMS (HierarchiCal LMS, In HLMS), since the number of antenna elements in each antenna group is reduced, the expansion of the feature 减小 is reduced, and the ratio of the error adjustment rate /^aflV/2 is also reduced. Among them, // and ;L are respectively It represents the average feature of the step and correlation matrix. Therefore, the convergence speed of the weight vector is accelerated, and the spatial response capability is also significantly improved. For Hierarchical RLS (HRLS), since the first layer 220 each The convergence rate of the group is faster than that of the standard RLS using M-1 array elements. Therefore, the input to the second layer 230 is also lower than the non-hierarchical (because the first layer 220 has filtered the noise), so the first The convergence speed of the 2nd layer 230 can be fast and has a low mean square error. Therefore, the standard RLS algorithm can be significantly improved. Complexity and convergence speed, ability to improve spatial response. The hierarchical architecture of the hierarchical linear array antenna beamforming architecture of the present invention can significantly improve the beamforming capability of the MVDR algorithm. According to a preferred embodiment of the present invention, The model simulated in this embodiment will be described below, and the results of the simulation will be shown. First, we will first describe some basic control conditions: 1. The incident angles of the target source and the interference source are sinloe and 0, respectively. It is represented by Z(9)=[l,e"'e-'···, €,, and =π sin(0). 3. The element signal of the array antenna is expressed in the form of a fundamental frequency, such as: w(8)=砗Exp(/〇+ is exp〇+ γ) + ν〇), π = 1,2,3,4,5. Here 4 and 4 represent the amplitude of the target signal and the interference signal, respectively, which is phase 11939TWF .DOC 11 1257735 Associated phase change, (4) is a Gaussian noise 値 4. The target noise ratio (TNR) is fixedly controlled at 10 dB. The interference noise ratio is variable, assuming their numbers are 20, respectively. 30, 40dB. The MVDR shown in Fig. 3 and Fig. 4 respectively shows the order of this embodiment. The spatial response of the MVDR under the condition of M=17 and n=200, we can see that the hierarchical MVDR has a lower response at the angle of the interference source, that is to say, the hierarchical method has better interference cancellation. Capability. If we increase the antenna element 26 to 26, the hierarchical MVDR of this embodiment can still accurately indicate the direction of the interference source and the signal source (as shown in Figure 6), but the conventional MVDR cannot do it. The same effect (see Figure 5). It is shown in Fig. 7 and Fig. 8 that when the antenna element M = 17, and the number of iterations is small (at n = 25, 30), the hierarchical architecture of this embodiment has better interference than the conventional architecture. Eliminate ability. It is shown in Fig. 10 that the hierarchical architecture of the present invention is still capable of canceling interference with fewer iterations as the antenna element Μ increases. In this embodiment, for example, we increase the Μ to 26 o'clock, and the interference is eliminated when there are very few iterations ("20, 25, 30), but the conventional MVDR cannot do it (such as Figure 9 shows). Although the present invention has been described above in terms of a preferred embodiment, it is not intended to limit the invention, and it is obvious to those skilled in the art that the present invention may be modified and retouched without departing from the spirit and scope of the invention. The scope of the invention is defined by the scope of the appended claims. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a multi-layer MVDR linear array antenna beamforming architecture diagram according to a preferred embodiment of a hierarchical linear array antenna beamforming frame 11939TWF.DOC 12 1257735 according to the present invention. . Figure 2 is a diagram showing a two-layer LMS MVDR linear array antenna beamforming architecture according to a preferred embodiment of the hierarchical linear array antenna beamforming architecture of the present invention. Figure 3 is the spatial response of a conventional adaptive MVDR beamformer with an interference noise ratio of 11^64〇-noise ratio = 20, 30, 40dB. (" = 200, M = 17) FIG. 4 is an empty P-single response of a hierarchical adaptive MVDR beamformer in a preferred embodiment of the hierarchical linear array antenna beamforming architecture of the present invention, with an interference noise ratio of 20, 30, 40 dB· (" = 200, M = 17). Figure 5 is the spatial response of a conventional adaptive MVDR beamformer with an interference noise ratio = 20, 30, 40 dB. (n = 200, M = 26). Figure 6 is a hierarchical linear array antenna beam of the present invention. In a preferred embodiment of the forming architecture, the spatial response of the hierarchical adaptive MVDR beamformer has an interference noise ratio = 20, 30, 40 (18. (/2 = 200, from = 26). Figure 7 is The spatial response of a conventional adaptive MVDR beamformer with an interference ratio = 20 dB. (w = 20, 25, 30, M = 17) Figure 8 is a comparison of the beamforming architecture of the hierarchical linear array antenna of the present invention. In a preferred embodiment, the spatial response of a hierarchical adaptive MVDR beamformer, a disturbing noise ratio = 20 dB · 〇 = 20, 25, 30, M = 17) Figure 9 is a conventional adaptive MVDR beam Spatial response of the former, its interference noise ratio = 20 dB and = 20, 25, 30, M = 26) Figure 10 is a hierarchical adaptation of the preferred embodiment of the hierarchical linear array antenna beamforming architecture of the present invention. Sexual 3^¥£) 11 beamformer 11939TWF.DOC 13 1257735 spatial response with interference noise ratio = 20 dB· (n = 20, 25, 30, typographical description = 110, 210: day Line elements 120, 220 · · First level group 130, 230 · Second level group 140: Last level group 150 : Hierarchical linear array antenna beamforming architecture 11939TWF.DOC 14