董眾+徐卓異+張善從
摘 要: 一種基于粒子濾波(PF)的正交頻分復(fù)用(OFDM)系統(tǒng)在慢衰落瑞利信道下聯(lián)合信道估計(jì)和載波恢復(fù)的新方法被提出。該算法適用于多徑時(shí)變信道模型以及等效離散時(shí)間信道模型。算法引入了在非線性系統(tǒng)參數(shù)估計(jì)和跟蹤領(lǐng)域上十分有效的PF方法,將Kaman濾波與序貫蒙特卡羅采樣(SMCS)相結(jié)合來估計(jì)信道衰落系數(shù)以及載波頻偏(CFO)的后驗(yàn)概率密度,從而通過計(jì)算得到信道的響應(yīng)函數(shù),并在此基礎(chǔ)上,利用MMSE均衡器消除碼間串?dāng)_(ICI),進(jìn)行碼元估計(jì)。仿真結(jié)果表明了算法的有效性和優(yōu)越性。
關(guān)鍵詞: 正交頻分復(fù)用; 信道估計(jì); 時(shí)延; 載波頻偏; 粒子濾波
中圖分類號(hào): TN951?34 文獻(xiàn)標(biāo)識(shí)碼: A 文章編號(hào): 1004?373X(2014)22?0048?04
Carrier frequency offset and channel joint estimating algorithm for OFDM system
base on PF
DONG Zhong1, XU Zhuoyi2, ZHANG Shancong1
(1. Technology and Engineering Center for Space Utilization, Chinese Academy of Science, Beijing 100094, China;
2. Space Star Technology Co., Ltd, Beijing 100083, China)
Abstract: A new algorithm of joint carrier recovery and channel estimation for orthogonal frequency?division?multiplexing (OFDM) system based on particle fitering (PF) in slow Rayleigh fading channel is proposed in this paper. The algorithm is suitable for both the multipath time?varying channel model and the equivalent discrete?time channel model. PF method effective for parameter estimation and tracking for non?linear model is introduced into the algorithm. In the algorithm, the channel fading coefficients and the posterior probability density of the unknown carrier frequency offset (CFO) are estimated in combination with Kalman filtering and sequential Monte Carlo sampling. The channel response function is got by computation. Based on this, MMSE equalizer is used to estimate the inter carrier interference (ICI), so as to perform code element estimation. The simulation result indicated effectiveness and superiority of the algorithm.
Keywords: OFDM; channel estimation; time delay; carrier frequency offset; particle filtering
0 引 言
OFDM技術(shù)已成為寬帶通信系統(tǒng)(如WiMAX,3GPP,LTE等)的標(biāo)準(zhǔn)技術(shù),而系統(tǒng)載波頻偏估計(jì)是進(jìn)行傳輸信號(hào)解調(diào)和分集合并的前提。至今國內(nèi)外已有大量關(guān)于單天線系統(tǒng)的信道估計(jì)[1?3],且關(guān)于MIMO?OFDM系統(tǒng)的信道估計(jì)也大多是基于導(dǎo)頻的信道估計(jì)方法[4?5]。總體上講,信道估計(jì)方法可分為兩種:直接估計(jì)信道的等效離散時(shí)間抽頭系數(shù);估計(jì)信道信號(hào)傳輸?shù)奈锢韰?shù),如多徑時(shí)延和多徑衰落等[1]。在射頻信號(hào)傳輸過程中,信道時(shí)延在數(shù)個(gè)OFDM數(shù)據(jù)塊中可以認(rèn)為是不變的[6?7],而多徑衰落即使在一個(gè)OFDM數(shù)據(jù)塊內(nèi)也具有明顯的變化?;诖?,國內(nèi)外學(xué)者依據(jù)信道特性,在假設(shè)信道時(shí)延已知的前提下提出了很多信道多徑衰落估計(jì)方法[1,3]。然而這些方法均沒有考慮由于發(fā)射機(jī)和接收機(jī)之間錯(cuò)誤匹配而產(chǎn)生的載波頻偏問題。文獻(xiàn)[4]則提出了一種基于擴(kuò)展Kalman濾波(Extended Kalman Filter,EKF)的載波頻偏和時(shí)延聯(lián)合估計(jì)方法。
本文將Kalman濾波和粒子濾波的重要技術(shù)——序貫重采樣(Sequential Importance Sampling,SIS)技術(shù)相結(jié)合提出了一種按幀處理的方法,來對(duì)信道的多徑衰落和頻偏進(jìn)行估計(jì),該方法無需很長(zhǎng)時(shí)間的累積計(jì)算,具有較好的時(shí)效性和較低的計(jì)算量。進(jìn)而在無需對(duì)信號(hào)進(jìn)行重采樣的情況下,對(duì)載波全頻域帶寬[-12 1 OFDM系統(tǒng)的CFO模型
設(shè)OFDM通信系統(tǒng)子載波數(shù)為[N],循環(huán)前綴長(zhǎng)度為[Ng]。一個(gè)OFDM塊長(zhǎng)度為[T=NbTs],其中[Ts]為采樣間隔,且[Nb=N+Ng]。令系統(tǒng)第[n]個(gè)傳輸符號(hào)為[xn=[xn[-N2],xn[-N2+1],…,xn[N2-1]]T],相應(yīng)的歸一化符號(hào)為[{xn[k]}](其中,[E[xn[k]x*n[k]]=1])。[ΔF]為射頻信號(hào)發(fā)射機(jī)和接收機(jī)間的不同步產(chǎn)生的CFO,對(duì)應(yīng)的歸一化CFO為[v=ΔFNTs]。[yn=[yn[-N2],yn[-N2+1],…,yn[N2-1]]T]為經(jīng)過多徑瑞利信道傳輸后,并對(duì)接收端的信號(hào)去循環(huán)前綴并進(jìn)行離散傅里葉變換后的信號(hào)頻譜[2?3],則可得:
[yn=Hnxn+wn] (1)
式中:[wn=[wn[-N2],wn[-N2+1],…,wn[N2-1]]T]為方差為[σ2IN]的加性高斯白噪聲;[Hn]為信道響應(yīng)矩陣,其每個(gè)元素可由等效抽頭系數(shù)[{hl,n=h(nT)}]表示:
[HNk,m=1Nl=0L′-1hne-j2π(mN-12)?lq=0N-1ej2πm-k+vNq] (2)
也可由信道物理參數(shù)[1]時(shí)延[τl]和衰落[{αl,n=αl(nT)}],得:
[HNk,m=1Nl=0L-1αl,ne-j2π(mN-12)τlq=0N-1ej2πm-k+vNq] (3)
式中:[L′ 在此基礎(chǔ)上,定義如下[L×1]向量[αn=α0,n,...,αL-1,nT],則[αn]間隔[p]的相關(guān)矩陣為[R(p)α=E[αnαHn-p]]為一個(gè)對(duì)角矩陣,且可表示為[R(p)αl,l=σ2αlJ0(2πfdTp)]。據(jù)此,通過簡(jiǎn)單用[L]替換[L′],用時(shí)延[τl]替換[{l,l=0:L′-1}],可推導(dǎo)得到第二種采用物理信道參數(shù)的情況下的結(jié)論。 式(1)提出的觀測(cè)模型對(duì)多徑衰落[αn]是線性的。因此,通過變換處理,可得到下式: [yn=?n(v)?αn+wn] (4) 式中,[?n(v)]為[?n(v)=??diag{Ω(v)}??H?diag{xn}?F],[Ω(v)=ej2π0vN,...,ej2π(N-1)vNT];傅里葉矩陣[F]的元素為[Fk,l=e-j2πkN-12τl],傅里葉矩陣[ω]的元素為[?k,p=1Ne-j2πkpN]。 2 CFO和信道衰落聯(lián)合估計(jì)方法 本文的主要目的即是利用觀測(cè)序列[y1:K={y1,y2,…,yK}],聯(lián)合估計(jì)信道衰落CGs?[α1:K={α1,α2,…,αK}]和載波頻偏CFO?[v],其中[K]為序列幀數(shù)。進(jìn)一步通過MMSE均衡器消除ICI影響,恢復(fù)碼元符號(hào)[1][x1:K={x1,x2,…,xK}]。 由貝葉斯理論可知,后驗(yàn)概率分布(PDF)[p(αn,v|y1:K)]為估計(jì)[n]時(shí)刻參數(shù)[(αn,v)]的主要方法。依據(jù)貝葉斯理論,PDF可表示為: [p(αn,v|y1:K)=p(αn|v,y1:K)×p(v|y1:K)] (5) 式中,PDF[p(αn|v,y1:K)]可通過Kalman濾波直接計(jì)算。然而,由于CFO的存在,觀測(cè)方程為非線性的,導(dǎo)致了PDF[p(v|y1:K)]無法通過Kalman濾波得到。因此,可利用大小為[Nc],系數(shù)為[{v(i)1:n,ω(i)1:n},i=1:Nc]的粒子集,通過粒子濾波來逼近PDF。由于新的粒子是根據(jù)每個(gè)接收符號(hào)由舊粒子的權(quán)重分布得到,因此可直接采用下標(biāo)[1:n]和[v]變量來對(duì)每個(gè)時(shí)間內(nèi)的粒子進(jìn)行區(qū)分。 本文利用組合濾波的方法來對(duì)[n]時(shí)刻的狀態(tài)[(αn,vn)]進(jìn)行估計(jì),通過Kalman濾波來更新信道衰落的估計(jì),并通過經(jīng)典的粒子濾波技術(shù)?SIS來對(duì)CFO估計(jì)值進(jìn)行更新。 2.1 粒子濾波 本文利用SIS來構(gòu)造PDF[p(v|y1:n)]的回歸經(jīng)驗(yàn)近似,通過從重要性函數(shù)中產(chǎn)生粒子,并分配歸一化的重要性權(quán)值。由文獻(xiàn)[8]可得到對(duì)應(yīng)的PDF表達(dá)式為: [p(v(i)1:n|y1:n)≈ω(i)nδ(v1:n-v(i)1:n)] (6) 式中[δ(?)]為迪萊克函數(shù)。利用該P(yáng)DF,可得到CFO的估計(jì)值,則算法可描述為如下步驟。 2.2 Kalman濾波 一階自回歸模型,信道衰落CGs的估計(jì)可建模為高斯白噪聲過程: [αn=a?αn-1+un] (7) 式中:[a=J0(2πfdTp)];[un]為復(fù)高斯向量,其均值為0,協(xié)方差矩陣[U=diag{σ2u0,...,σ2uL-1}],其中[σ2ul=σ2al(1-a2)]。依據(jù)狀態(tài)模型(7)和觀測(cè)模型(4),可通過Kalman濾波實(shí)現(xiàn)對(duì)信道衰落CGs[αn]的自適應(yīng)跟蹤,得到對(duì)應(yīng)[n]時(shí)刻的狀態(tài)后驗(yàn)估計(jì)[αn|n],已經(jīng)相應(yīng)的協(xié)方差[Pn|n]。 2.3 初始化 頻偏[v0]在帶寬的[[-12,12]]中均勻分布。因此,[n=0]時(shí)刻,算法初始化為: [v(i)0~u(-12,12), α(i)0|0=0L,1, P(i)0|0=R(0)α, ω(i)0=1Nc, i=1:Nc]。
2.4 重要性采樣:權(quán)重更新和重采樣
由于重要性函數(shù)[p(vk|v(i)1:n-1,y1:n-1)]沒法準(zhǔn)確得到,根據(jù)文獻(xiàn)[7],可以得到它的近似貝塔分布。由于貝塔分布屬于[[0,1]]范圍,本文引入新的頻偏變量[u=v+0.5],從而[u]也同樣分布在[[0,1]]范圍。因此,可得到頻偏采樣:
[u~β(u,Un,Vn)] (8)
其中,貝塔函數(shù)的參數(shù)[Un,Vn]為:
[Un=unun(1-un)σ2un-1Vn=(1-un)un(1-un)σ2un-1] (9)
其中,[un,σ2un]可通過式(10)得到:
[un=i=1Ncω(i)n-1u(i)n-1 σ2un=i=1Ncω(i)n-1(u(i)n-1-un)2] (10)
得到新的粒子之后,接著更新對(duì)應(yīng)的重要性權(quán)重:
[ω(i)n∝ω(i)n-1?N(yn,m(i)n,R(i)n)] (11)
其中,向量[m(i)n]和矩陣[R(i)n]可表示如下:
[m(i)n=?n(v(i)n)?α(i)n|n-1, R(i)n=D(i)n] (12)
式中權(quán)重[ω(i)n]為非歸一化的權(quán)重。因此,可采用[ω(i)n=ω(i)ni-1Ncω(i)n]進(jìn)行歸一化處理:實(shí)際應(yīng)用中,SIS算法的一個(gè)著名問題為粒子[v(i)n]會(huì)快速退化,在經(jīng)過一定的迭代步驟后,大部分重要權(quán)重均只有很小的值[7][ω(i)n?0]。解決該問題的一個(gè)通常方法為進(jìn)行重采樣[9]。
2.5 CFO和信道衰落CGs估計(jì)
由重要性權(quán)重和CFO的采樣點(diǎn)可通過下式計(jì)算真實(shí)CFO的MMSE估計(jì)[vn=i=1Ncu(i)nω(i)n-0.5]。進(jìn)而,[n]時(shí)刻信道的衰落CGs估計(jì)[αn]可表示為[αn=i=1Ncα(i)n|nω(i)n]。
3 仿真實(shí)驗(yàn)
本節(jié)通過仿真實(shí)驗(yàn)考查利用CFO和CGs進(jìn)行信道估計(jì)的性能,采用估計(jì)的均方差(MSE)和系統(tǒng)符號(hào)判決結(jié)果的誤碼率(BER)作為衡量指標(biāo)。OFDM仿真參數(shù)為:調(diào)制方式4?QAM調(diào)制,子載波數(shù)為[N=64],前綴長(zhǎng)度[Ng=N8],信噪比[SNR=1/σ2],信噪比分貝值為[(SNR)dB=(EbN0) dB+3 dB]。仿真中采用歸一化的慢變多徑瑞利信道,信道的多普勒展寬為[fdT=10-3],多徑數(shù)目為[L=6]。載波CFO在[[-12,12]]均勻分布,本文隨機(jī)選擇了[v=0.3]。本節(jié)分別在粒子數(shù)[Nc=][10, 50, 100, 200]情況下,進(jìn)行CFO和CGs的估計(jì)。相關(guān)仿真結(jié)果如圖1~圖3所示。
圖1 不同粒子數(shù)和信噪比下的CFO估計(jì)情況
圖2 CFO和CGs聯(lián)合估計(jì)結(jié)果
圖1(a)為信噪比10 dB,CFO為0.3時(shí)的仿真結(jié)果,可以看出:CFO的估計(jì)在10~15個(gè)OFDM符號(hào)后就收斂于真實(shí)的CFO;且可以看出,粒子數(shù)為50時(shí),算法收斂于0.292 9;粒子數(shù)為200時(shí),算法收斂于0.299 4;因此隨著粒子數(shù)的增加,算法收斂速度和最終的精度不斷提高。圖1(b)為不同信噪比情況下,粒子數(shù)為100的仿真結(jié)果,可以看出:信噪比為0 dB時(shí),算法在80個(gè)符號(hào)后收斂于0.298 6;信噪比為30 dB時(shí),算法在5個(gè)符號(hào)后就收斂于0.300 7;因此信噪比的提高也可帶來算法收斂速度和精度的提高。
由圖2的仿真結(jié)果可以看出:在不同的條件下,CFO和CGs的估計(jì)值均有變化,且CFO的估計(jì)值會(huì)對(duì)CGs的估計(jì)有較大影響;而CGs的估計(jì)效果則不會(huì)對(duì)CFO有太大影響。
圖3 不同信噪比下系統(tǒng)誤碼率曲線
由圖3的仿真結(jié)果可以看出,與理想的沒有CFO和CGs存在時(shí)的誤碼率相比:沒有CFO的實(shí)際系統(tǒng)中,誤碼率與理想情況幾乎一致;存在CFO時(shí)雖然有所增加,不過也差距很小。仿真結(jié)果表明了本文算法的有效性。
4 結(jié) 語
本文提出了一種基于組合粒子和Kalman濾波的算法,進(jìn)行CFO和CGs聯(lián)合估計(jì),并在此基礎(chǔ)上,利用MMSE均衡器進(jìn)行數(shù)據(jù)符號(hào)檢測(cè)。通過仿真實(shí)驗(yàn)得到的MSE和BER與理想情況進(jìn)行對(duì)比,表明了算法的優(yōu)越性。同時(shí),仿真結(jié)果表明,本文算法在全頻帶CFO情況下,利用大小為50的粒子數(shù)目,在15~20個(gè)OFDM符號(hào)后即可快速準(zhǔn)確地收斂于CFO的實(shí)際值??梢娝惴ㄔ趯?shí)際情況下,有一定的實(shí)用價(jià)值。
參考文獻(xiàn)
[1] HIJAZI H, ROS L. Joint data QR?detection and Kalman estimation for OFDM time?varying Rayleigh channel complex gains [J]. IEEE Transactions on Vehicle Technology, 2010, 58:170?178.
[2] TANG Z, CANNIZZARO R C, LEUS G, et al. Pilot?assisted time varying channel estimation for OFDM systems [J]. IEEE Transactions on Signal Processing, 2007, 5(55): 2226?2238.
[3] HIJAZI H, ROS L. Rayleigh time?varying channel complex gains estimation and ICI cancellation in OFDM systems[J]. European Transactions on Telecommunications, 2009, 6: 111?121.
[4] SIMON E, HIJAZI H, ROS L, et al. Joint estimation of carrier frequency offset and channel complex gains for OFDM systems in fast time?varying vehicular environments [C]// IEEE International Conference on Communications (ICC): Workshop on Vehicular Connectivity. Cape Town, South Africa: IEEE, 2010: 1234?1239.
[5] JAKES W C. Microwave mobile communications [M]. USA: IEEE Press, 1983.
[6] KIM J G, LIM J T. MAP?based channel estimation for MIMOOFDM over fast Rayleigh fading channels [J]. IEEE Transactions on Vehiclre Technology, 2008, 57(3): 234?241.
[7] GHIRMAI T. Sequential Monte Carlo method for fixed symbol timing estimation and data detection [C]// Proceedings of IEEE Information Sciences and Systems. [S.l.]: [s.n.], 2006: 34?39.
[8] GHIRMAI T, BUGALLO N F, MIGUEZ J, et al. A sequential monte carlo method for adaptive blind timing estimation and data detection [J]. IEEE Transactions on Signal Processing, 2005, 53(8): 2855?2865.
[9] DOUCET A, GODSILL S, ANDRIEU C. On sequential Monte Carlo sampling methods for bayesian filtering [J]. Statistics and Computing, 2000, 10:197?208.
[10] SIMON E, ROS L, RAOOF K. Synchronization over rapidly time varying multipath channel for CDMA downlink RAKE receivers in time?division mode[J]. IEEE Transactions on Vehicle Technology, 2007, 56(4): 45?52.
[11] HIJAZI H, ROS L. Time?varying channel complex gains estimation and ICI suppression in OFDM systems [C]// IEEE Global Communications Conference. Washington, USA: IEEE, 2007: 10?15.
[4] SIMON E, HIJAZI H, ROS L, et al. Joint estimation of carrier frequency offset and channel complex gains for OFDM systems in fast time?varying vehicular environments [C]// IEEE International Conference on Communications (ICC): Workshop on Vehicular Connectivity. Cape Town, South Africa: IEEE, 2010: 1234?1239.
[5] JAKES W C. Microwave mobile communications [M]. USA: IEEE Press, 1983.
[6] KIM J G, LIM J T. MAP?based channel estimation for MIMOOFDM over fast Rayleigh fading channels [J]. IEEE Transactions on Vehiclre Technology, 2008, 57(3): 234?241.
[7] GHIRMAI T. Sequential Monte Carlo method for fixed symbol timing estimation and data detection [C]// Proceedings of IEEE Information Sciences and Systems. [S.l.]: [s.n.], 2006: 34?39.
[8] GHIRMAI T, BUGALLO N F, MIGUEZ J, et al. A sequential monte carlo method for adaptive blind timing estimation and data detection [J]. IEEE Transactions on Signal Processing, 2005, 53(8): 2855?2865.
[9] DOUCET A, GODSILL S, ANDRIEU C. On sequential Monte Carlo sampling methods for bayesian filtering [J]. Statistics and Computing, 2000, 10:197?208.
[10] SIMON E, ROS L, RAOOF K. Synchronization over rapidly time varying multipath channel for CDMA downlink RAKE receivers in time?division mode[J]. IEEE Transactions on Vehicle Technology, 2007, 56(4): 45?52.
[11] HIJAZI H, ROS L. Time?varying channel complex gains estimation and ICI suppression in OFDM systems [C]// IEEE Global Communications Conference. Washington, USA: IEEE, 2007: 10?15.
[4] SIMON E, HIJAZI H, ROS L, et al. Joint estimation of carrier frequency offset and channel complex gains for OFDM systems in fast time?varying vehicular environments [C]// IEEE International Conference on Communications (ICC): Workshop on Vehicular Connectivity. Cape Town, South Africa: IEEE, 2010: 1234?1239.
[5] JAKES W C. Microwave mobile communications [M]. USA: IEEE Press, 1983.
[6] KIM J G, LIM J T. MAP?based channel estimation for MIMOOFDM over fast Rayleigh fading channels [J]. IEEE Transactions on Vehiclre Technology, 2008, 57(3): 234?241.
[7] GHIRMAI T. Sequential Monte Carlo method for fixed symbol timing estimation and data detection [C]// Proceedings of IEEE Information Sciences and Systems. [S.l.]: [s.n.], 2006: 34?39.
[8] GHIRMAI T, BUGALLO N F, MIGUEZ J, et al. A sequential monte carlo method for adaptive blind timing estimation and data detection [J]. IEEE Transactions on Signal Processing, 2005, 53(8): 2855?2865.
[9] DOUCET A, GODSILL S, ANDRIEU C. On sequential Monte Carlo sampling methods for bayesian filtering [J]. Statistics and Computing, 2000, 10:197?208.
[10] SIMON E, ROS L, RAOOF K. Synchronization over rapidly time varying multipath channel for CDMA downlink RAKE receivers in time?division mode[J]. IEEE Transactions on Vehicle Technology, 2007, 56(4): 45?52.
[11] HIJAZI H, ROS L. Time?varying channel complex gains estimation and ICI suppression in OFDM systems [C]// IEEE Global Communications Conference. Washington, USA: IEEE, 2007: 10?15.