【正文】
osing the most suitable algorithm for different adaptation phases. This method may be applied to all the LMSbased algorithms, although we here consider only several of them.The paper is organized as follows. An overview of the considered LMSbased algorithms is given in Section 3 proposes the criterion for evaluation and bination of adaptive algorithms. Simulation results are presented in Section 4.2. LMS based algorithmsLet us de?ne the input signal vector and vector of weighting coef?cients as .The weighting coef?cients vector should be calculated according to: (1)where 181。} is the estimate of the expected value andis the error at the instant k,and dk is a reference signal. Depending on the estimation of expected value in (1), one de?nes various forms of adaptive algorithms: the LMS,the GLMS,and the SA,[1,2,5,8] .The VS LMS has the same form as the LMS, but in the adaptation the step 181。,GLMS: q ≡ a,SA:q ≡ 181。 μ = . The optimal vectors is generated according to the presented model with ,and with κ = 2. In the first 30 iterations the variance was estimated according to (7), and CA takes the coefficients of SA with μ (SA1). Figure 2(a) shows the AMSD characteristics for each algorithm. In steady state the CA does not ideally follow the SA3 with μ/8, because of the nonstationary problem nature and a relatively small difference between the coefficient variances of the SA2 and SA3. However,this does not affect the overall performance of the proposed algorithm. AMSD for each considered algorithm was: AMSD = (SA1,μ), AMSD = (SA2,μ/2), AMSD = (SA3, μ/8) and AMSD = (Comb). (b) Comparison with VS LMS algorithm [6]: In this simulation we take the improved CA (9) from , and pare its performance with the VS LMS algorithm [6], in the case of abrupt changes of optimal vector. Since the considered VS LMS algorithm[6] updates its step size for each weighting coefficient individually, the parison of these two algorithms is meaningful. All the parameters for the improved CA are the same as in . For the VS LMS algorithm [6], the relevant parameter values are the counter of sign change m0 = 11,and the counter of sign continuity m1 = 7. Figure 2(b)shows the AMSD for the pared algorithms, where one can observe the favorable properties of the CA, especially after the abrupt changes. Note that abrupt changes are generated by multiplying all the system coefficients by ?1 at the 2000th iteration (Fig. 2(b)). The AMSD for the VS LMS was AMSD = , while its value for the CA (CoLMS) was AMSD = .For a plete parison of these algorithms we consider now their calculation plexity, expressed by the respective increase in number of operations with respect to the LMS algorithm. The CA increases the number of requres operations for N additions and N IF the VS LMS algorithm, the respective increase is: 3N multiplications, N additions, and at least 2N IF decisions.These values show the advantage of the CA with respect to the calculation plexity.6. ConclusionCombination of the LMS based algorithms, which results in an adaptive system that takes the favorable properties of these algorithms in tracking parameter variations, is the course of adaptation procedure it chooses better algorithms, all the way to the steady state when it takes the algorithm with the smallest variance of the weighting coefficient deviations from the optimal value.Acknowledgement. This work is supported by the Volkswagen Stiftung, Federal Republic of Germany.基于LMS算法的自適應(yīng)組合濾波器摘要:提出了一種自適應(yīng)組合濾波器。作為正在研究中的濾波器算法比較標(biāo)準(zhǔn),我們采取偏差和加權(quán)系數(shù)之間的方差比。關(guān)鍵詞:自適應(yīng)濾波器;LMS算法;組合算法;偏差和方差權(quán)衡緒論自適應(yīng)濾波器已在信號(hào)處理和控制,以及許多實(shí)際問題[1, 2]的解決當(dāng)中得到了廣泛的應(yīng)用. 自適應(yīng)濾波器的性能主要取決于濾波器所使用的算法的加權(quán)系數(shù)的更新。LMS算法是非常簡便,易于實(shí)施,具有廣泛的用途[13]。每一種基于LMS的算法都至少有一個(gè)參數(shù)在適應(yīng)過程(LMS算法和符號(hào)算法,加強(qiáng)和GLMS平滑系數(shù),各種參數(shù)對(duì)變步長LMS算法的影響)中被預(yù)先定義。這些參數(shù)的選擇主要是基于一種算法質(zhì)量的權(quán)衡中所提到的適應(yīng)性能。也就是說,我們提出了幾個(gè)基于LMS算法的不同參數(shù)的FIR濾波器,并提供不同的適應(yīng)階段選擇最合適的算法標(biāo)準(zhǔn)。本文的結(jié)構(gòu)如下,作者認(rèn)為的LMS的算法概述載于第2節(jié),第3節(jié)提出了自適應(yīng)算法的改進(jìn)和組合標(biāo)準(zhǔn),仿真結(jié)果在第4節(jié)。}是預(yù)期值的估計(jì)。根據(jù)(1)中不同的預(yù)期值估計(jì)在,我們可以得出一種各種形式的自適應(yīng)算法的定義:LMS , ,[1,2,5,8] . 變步長LMS算法和基本LMS算法具有相同的形式,但在適應(yīng)過程中步長μ(k)是變化的[6,7]。我們考慮兩種情況:是一個(gè)常數(shù)(固定的情況下),隨時(shí)間變化(非平穩(wěn)的情況下)。我們假設(shè)變量可以建立模型為,它是隨機(jī)獨(dú)立的零均值,依賴于和自相關(guān)矩陣。定義加權(quán)錯(cuò)位系數(shù),[1–3], 。它可以表示為: (2)根據(jù)(2),是:(3)是加權(quán)系數(shù)的偏差,與方差是零均值的隨機(jī)變量差,它取決于LMS的算法類型,以及外部噪聲方差。在這個(gè)意義上說,在后面的分析中我們將假定只依賴算法類型,及其參數(shù)。對(duì)于自適應(yīng)濾波器,它被賦值,[3]:組合自適應(yīng)濾波器合并后的自適應(yīng)濾波器的基本思想是在兩個(gè)或兩個(gè)以上自適應(yīng)LMS算法并行實(shí)現(xiàn)與每個(gè)迭代之間的最佳選擇,[9]。最好的加權(quán)系數(shù)是1,即在給定的時(shí)刻,向相應(yīng)的維納矢量值最接近。注意,現(xiàn)在我們可以在一個(gè)統(tǒng)一的處理方式(LMS: q ≡ 181。)下?,F(xiàn)在分析最小均方與一些基于相同類型的算法相結(jié)合的自適應(yīng)濾波器,但參數(shù)q是不同的。 (4)(4)中的概率P(κ)依賴κ的值. 例如κ= 2的高斯分布,P(κ)= (兩個(gè)σ規(guī)則)。另一方面,當(dāng)偏置變大,然后中央位置的不同間隔距離很大,而且他們不相交。這個(gè)標(biāo)準(zhǔn)的平衡狀態(tài),從或同一個(gè)數(shù)量級(jí)的,即。第2步:估計(jì)每個(gè)算法的方差。從一個(gè)最大的差異值算法走向與差異較小的值。如果相交,偏差已經(jīng)很小。首先兩個(gè)區(qū)間不相交意味著實(shí)現(xiàn)了取舍標(biāo)準(zhǔn),并選擇最大方差算法。元素的集合Q中最小的數(shù)L= 2。通過增加更多的觀察,這兩個(gè)極端之間,我們可以稍微改進(jìn)算法的瞬態(tài)行為。在仿真中我們估計(jì)[4]式: (7)當(dāng)k = 1,2,... ,L和替代的方法是估計(jì)為 (8)有關(guān)表達(dá)式和在穩(wěn)定狀態(tài)為LMS算法的不同類型,從已知文獻(xiàn)中可以看出。,[3]. 需要注意的是,任何其他估計(jì)對(duì)于濾波器來說是有效的。加權(quán)系數(shù)的計(jì)算并未使并行算法增加計(jì)算時(shí)間,因?yàn)樗怯捎布?shí)現(xiàn)并行執(zhí)行的,從而增加了硬件要求。簡單的分析表明,在CA增加最多的操作步驟,添加了N(L?1)和N(L?1) IF決定增補(bǔ),而且需要添加一些硬件以滿足組成算法。在這里,參數(shù)q是μ,即。我們給個(gè)人平均為方差算法(AMSD),以及它們的結(jié)合,如圖1(a)所示。它引用了未知損壞不相關(guān)零均值高斯噪聲,其中= , SNR = 15 dB, κ = . 在最初的30次迭代的方差估計(jì)根據(jù)式(7)和CA的加權(quán)來計(jì)算LMS的系數(shù)μ。需要注意的是第200和第400迭代,該LMS算法可以采取任何步長根據(jù)不同的認(rèn)識(shí)。組合自適應(yīng)濾波器能夠達(dá)到更好的性能如果該獨(dú)立算法能勝過他們以往所采取的系數(shù)值迭代,即采取由CA所選擇的那些值。為了比較清楚地取得成果,為每次仿真計(jì)算了AMSD,對(duì)于第一個(gè)LMS(μ)是AMSD = ,第二的LMS(μ/10)是AMSD = ,對(duì)CA(CoLMS)是AMSD = ,還有與改進(jìn)的式(9)是AMSD = 。比較聯(lián)合濾波器性能,以組成特定的組合。(a)優(yōu)化加權(quán)時(shí)變向量:提出的想法可能被應(yīng)用到SA算法的非平穩(wěn)情況。 μ = . 根據(jù)最優(yōu)向量生成的模型 ,κ = 2. n的前30次迭代的方差估計(jì)根據(jù)式(7),CA與SA系數(shù)μ(SA1)。在穩(wěn)定狀態(tài)的CA不理想的遵循μ/ 8 SA3,因?yàn)閱栴}的性質(zhì)和非平穩(wěn)之間的SA2和SA3系數(shù)差異相對(duì)較小,但這并不影響該算法的整體性能。(b)比較與VS LMS算法[6]:在仿真中,(9)式,并在最佳載體突然變化的情況下比較其與LMS算法的性能[6]。,對(duì)VS LMS算法[6],有關(guān)的參數(shù)值是變化的且具有符號(hào)的連續(xù)性,m0 = 11,m1 = 7。但要注意的是,突然的變化使系統(tǒng)乘以1到2000次迭代(圖2(b))。與一個(gè)完整的這些算法相比,我們認(rèn)為現(xiàn)在的計(jì)算復(fù)雜度增加了。CA增加了對(duì)N 的補(bǔ)充和N LMS算法,其增加了:3N乘法,N的添加,以及決定至少2N IF 。結(jié)論組合LMS算法,在自適應(yīng)系統(tǒng)中將這些參數(shù)變化的跟蹤與算法的良好性能結(jié)果相結(jié)合,是自適應(yīng)過程中選擇的更好的算法,一直到穩(wěn)定狀態(tài)時(shí)需要從最優(yōu)值與最小方差算法的加權(quán)系數(shù)的偏差。s Shaanxi province pass through a stop on the ancient Silk Road, Gansu39。80s. We sat on