CN100477655C - Method for estimating OFDM integral number frequency multiplication bias - Google Patents

Abstract

Present invention relates to a method for estimating integral times frequency offset in communication technology field. It is based on a L isotonic N point OFDM synchronizing training symbol, receiving end reorganization this N point symbol as M number L point OFDM symbol, then according to preceding stage module to obtain decimals times frequency offset estimated value, leading-in decimals times frequency offset corrective term in tradition FFT algorithm structure, thereby designing out FFT algorithm new structure capable of completing decimals times frequency offset compensating, for demodulating part OFDM reorganization symbol which is extracted from M number reorganization symbol through random, equal interval or continuous method, to proceed square merging above-mentioned demodulation result, the merged peak value corresponded frequently point being the unbiased estimate value of integral times frequency offset, in order to further reduce complexity of computation, adopting self-adaptive algorithm base on adjudicate reliability feedback. Said invention has advantage of lower calculation complexity and estimating error ratio, having very high application value in OFDM system.

Description

A kind of method of estimating OFDM integral number frequency multiplication bias

Technical field

The present invention relates to a kind of method of communication technical field, particularly a kind of method of estimating OFDM integral number frequency multiplication bias.

Background technology

At present, the OFDM technology is applied at increasing wired, wireless communication field, and this has many advantages mainly due to the OFDM technology: effectively anti-multipath is disturbed and narrow band interference, and availability of frequency spectrum height, message transmission rate is high.Yet OFDM is for synchronism deviation, and is particularly very responsive to frequency departure.Frequency departure is divided into the decimal overtones band deviation of subcarrier spacing and the integer multiple frequency deviation of subcarrier spacing again, abbreviates fractional part of frequency offset and integer frequency offset below respectively as.Wherein, fractional part of frequency offset can cause and disturb (ICI) between subcarrier; Integer frequency offset can not cause ICI, but can cause the cyclic shift that receives data symbol, makes that the error probability of the information symbol that demodulation is come out is 50%.

The method of common estimating OFDM integral number frequency multiplication bias has three kinds:

(1) based on the synchronous training sequence of specific frequency spectrum pattern, the method need be done fast Fourier transform (FFT) to synchronous training symbol, and to make cyclic shift relevant with the known frequency spectrum pattern then, estimates integer frequency bias by seeking relevant peaks.Referring to document: Schmidl, T.M. etc. " Low-overhead, low-complexity[burst] synchronization for OFDM ", IEEE International Conference onCommunications, Volume 3, and June 1996, Page (s): 1301-1306.(" the OFDM method for synchronous of low accessing cost for data, low complex degree " IEEE international communication technical conference)

(2) based on the virtual subnet carrier wave of ofdm system, ofdm signal only uses a part of subcarrier of whole bandwidth in transmission course, generally reserves some subcarriers as the protection frequency band at band edge, is called virtual subnet carrier wave (virtual carriers).Owing to have orthogonality between the subcarrier of OFDM, so the virtual subnet carrier wave constitutes " zero subspace " of ofdm signal, utilizing inner product between orthogonal sub-carriers is zero character, can extrapolate integer frequency offset.Referring to document: Liu, H. etc., " A high-efficiency carrierestimator for OFDM communications, " IEEE Communications Letters, Volume2, Issue 4, and April 1998, Page (s): 104-106.(" OFDM frequency deviation estimating method efficiently " ieee communication technology press release)

(3) based on the synchronous training symbol structure of the OFDM of L five equilibrium, the method asks phase angle to estimate integer frequency offset by the auto-correlation of the specific delays of calculation training symbol again.Its estimation range becomes along with the increase of L greatly, but estimated accuracy variation thereupon, the also corresponding increase of computation complexity.Referring to document: Heiskala J etc.: OFDM Wireless LANs-A Theoretical and Practical Guide.[M] .Indianapolis USA:Pearson Education Inc, 2002.70-73. (" guidance of OFDM WLAN (wireless local area network)--theory and practice ") is a conventional method to call this method in the following text.

, all there is the too high shortcoming of computation complexity in above-mentioned three kinds of methods.If the sub-carrier number of ofdm system is N, method (1) needs to calculate N point FFT, needs at least

N complex multiplication; Method (2) needs the inner product between the signal calculated, needs complex multiplication at least N time; Method (3) needs the auto-correlation between the computing relay signal, needs complex multiplication at least N-M time.Nowadays, some representative values of N are 512,1024 or 2048 etc., and this makes above-mentioned three kinds of methods run into very big difficulty in actual applications.

Summary of the invention

The objective of the invention is at the deficiencies in the prior art, a kind of method of estimating OFDM integral number frequency multiplication bias is provided, make its by the method for reconstruct symbol and efficiently fft algorithm finish the estimation of integer frequency offset, make full use of ofdm system, guarantee the performance of frequency offset estimating, had lower computation complexity again.

The present invention is achieved through the following technical solutions, and specifically comprises the steps:

Step 1: transmitting terminal produces the synchronous training symbol of N point OFDM of a L five equilibrium, and each five equilibrium of symbol contains M point (M=N/L, and N, L, M are integer);

Step 2: synchronous training symbol is reconstructed, is about to the symbol in the step 1, be reconstructed into M L point ofdm signal symbol;

Step 3: in the reconstruct symbol of step 2 gained, by at random, uniformly-spaced or continuous method, extract M reconstruct symbol (0＜λ≤1 of λ, and λ M is an integer), according to the fractional part of frequency offset estimated value that system's prime obtains, in the traditional FFT algorithm structure, introduce the fractional part of frequency offset correction term, thereby design the fft algorithm new construction that to finish the fractional part of frequency offset compensation, be used for the above-mentioned λ M of a demodulation reconstruct symbol, obtaining a λ M length is the frequency domain sequence that L is ordered;

Step 4: the frequency domain sequence to the step 3 gained, merge according to quadratic sum, the pairing frequency of the peak value after the merging is exactly the estimated value of integer frequency offset.

Can also enter adaptive iteration algorithm behind the completing steps four based on the decision reliability feedback, be the average of calculation procedure three gained frequency domain sequences and the ratio of peak value, reliability index as step 4 gained estimated value, comparative result according to reliability index and a certain threshold value, increase λ adaptively, repeating step three and step 4 obtain integer frequency offset estimated value comparatively reliably at last.

Below the invention will be further described:

(1) the synchronous training symbol of N point OFDM of generation L five equilibrium

The synchronous training symbol of the OFDM of L five equilibrium is a kind of quite common training sequence structure, in being usually used in bit-synchronization algorithm and fractional part of frequency offset algorithm for estimating.Its generation method is as follows:

If the number of sub carrier wave of ofdm system is N, the significant character cycle is T, and the frequency of k subcarrier is f _k=k/T (0≤k≤N-1), a _{I, k}Be i symbol, the frequency domain data that on k subcarrier, loads; b _{I, l}I symbol, the base band time domain data of l sampled point.If establishing i symbol is synchronous training symbol, then insert pilot tone according to formula (1), just can generate the synchronous training symbol of OFDM of L five equilibrium.

$\left\{\begin{array}{ccc}{a}_{i,k}\≠0& k=0& \left(\mathrm{mod}L\right)\\ a_{i,k}=0& k\≠0& \left(\mathrm{mod}L\right)\end{array}\right.---\left(1\right)$

Under the situation of L five equilibrium, counting that the training symbol of each five equilibrium is contained is M (M=N/L), as the formula (2):

b _i，l＝b _i，l+nM(l＝0，1，2，…，M-1，n＝0，1，2，…，L-1) (2)

Usually, a signal of being made up of the M point is called a slot.

(2) synchronous training symbol is reconstructed

In (1) described synchronous training symbol, each slot contains M point, they be labeled as 1,2,3 ..., M.The OFDM symbol reconstruct of indication of the present invention is exactly that the time domain point that label is identical among each slot is extracted out, forms M new OFDM symbol, and each reconstruct OFDM symbol contains L time domain point, and specific practice is as follows:

If r _{I, m}(0≤m≤N-1) is an i symbol of receiving terminal, m time-domain sampling point, and then M reconstruct symbol can be expressed as vector form r _m(1≤m≤M), see formula (3):

$\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="r"\; filenames="C20061002522200161.png,C20061002522200162.png"\; state="\{\{state\}\}">r</figure-callout>}}_1=(r_{i,0},r_{i,M},r_{i,2M},\&CenterDot;\&CenterDot;\&CenterDot;,r_{i,(L-1)M})\\ {\mathrm{<figure-callout\; id="2"\; label="r"\; filenames="C20061002522200161.png,C20061002522200162.png"\; state="\{\{state\}\}">r</figure-callout>}}_2=(r_{i,1},r_{i,M+1},r_{i,2M+1},\&CenterDot;\&CenterDot;\&CenterDot;,r_{i,(L-1)M+1})\\ {\mathrm{<figure-callout\; id="3"\; label="r"\; filenames="C20061002522200161.png,C20061002522200172.png"\; state="\{\{state\}\}">r</figure-callout>}}_3=(r_{i,2},r_{i,M+2},r_{i,2M+2},\&CenterDot;\&CenterDot;\&CenterDot;,r_{i,(L-1)M+2})\\ \begin{array}{cc}.& .\end{array}\\ \begin{array}{cc}.& .\end{array}\\ \begin{array}{cc}.& .\end{array}\\ r_M=(r_{i,M-1},r_{i,M+M-1},r_{i,2M+M-1},\&CenterDot;\&CenterDot;\&CenterDot;,r_{i,(L-1)M+M-1})\end{array}\right.---\left(3\right)$

For discussing for simplicity note:

r _m(n)＝r _i，nM+m (0≤m≤M-1，0≤n≤L-1) (4)

So, r _m=(r _m(0), r _m(1) ..., r _m(n) ..., r _m(L-1)), above restructuring procedure makes sampled data obtain the advantage of time diversity, thereby makes algorithm for estimating robust more.

(3) demodulation reconstruct symbol obtains corresponding frequency domain sequence

Under desirable channel condition, reconstruct OFDM symbol r _mIn the time domain point should be identical.Yet, increase progressively or the trend of successively decreasing owing to exist frequency shift (FS), the phase place of these points to present, as being modulated onto on a certain frequency.So, for r _m, the influence of frequency deviation can equivalence be the baseband modulation process that contains the ofdm system of L subcarrier.Therefore, these reconstruct symbols of demodulation are sought their spectrum amplitude peak value, just can make estimation to integer frequency offset.And the method for demodulating ofdm reconstruct symbol is exactly a L point fft algorithm, and L generally is smaller number, such as 4,8 or the like.Than the fft algorithm of small point just the present invention have the core place of low computation complexity.

To r _mThe utilization fft algorithm:

$R_m\left(y\right)=\mathrm{DFT}\left(r_m\right|y)=\underset{n=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{r}_m\left(n\right)e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="C20061002522200161.png,C20061002522200162.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;ny}/L}---\left(5\right)$

So the estimator of integer frequency offset is expressed as:

Simple L point fft algorithm also is not enough to solve the estimation problem of integer frequency offset completely, especially when fractional part of frequency offset is near 0.5, and R _mIn to have the amplitude of two frequencies all bigger, sharply descend thereby cause adjudicating performance.Therefore, the fft algorithm structure that the present invention makes new advances in conjunction with the design of fractional part of frequency offset estimated value under the prerequisite that does not increase any computing expense, guarantees this algorithm under the situation that correct fractional part of frequency offset is estimated, is not have partially to estimate all the time.

Adopt prior art can realize easily that fractional part of frequency offset estimates, such as based on to the relevant method of asking phase angle of Cyclic Prefix, based on the relevant method of asking phase angle of branch symbols such as 2, based on the method for ML criterion search etc.Said method generally is positioned at the present invention program's prime, and the present invention is according to its fractional part of frequency offset estimated value that provides

Can realize efficiently that integer frequency offset does not have the task of inclined to one side estimation.

In general, the subcarrier of ofdm system is 2 integer power, as: 256,512,1024 etc.In order to guarantee that M is an integer, L also can only be 2 integer power, establishes L=2 ^β(β is a positive integer).2 ^βThe fft algorithm of point can be divided into β level butterfly computation, and p level butterfly computation contains 2 ^P-1Individual different multiple multiplying factor.Original 2 ^βP level in the some fft algorithm, q multiple multiplying factor is:

${\mathrm{<figure-callout\; id="2"\; label="W"\; filenames="C20061002522200161.png,C20061002522200162.png"\; state="\{\{state\}\}">W</figure-callout>}}_{2^{\mathrm{\&beta;}}}^{q,p}=\exp \{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="C20061002522200161.png,C20061002522200162.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}\frac{{\mathrm{<figure-callout\; id="2"\; label="q"\; filenames="C20061002522200161.png,C20061002522200162.png"\; state="\{\{state\}\}">q</figure-callout>}2}^{\mathrm{\&beta;}-\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="C20061002522200161.png,C20061002522200162.png"\; state="\{\{state\}\}">p</figure-callout>}}}{2^{\mathrm{\&beta;}}}\},\left(\begin{array}{c}p=\mathrm{1,2},...,\mathrm{\&beta;}\\ q=\mathrm{0,1,2},...,2^{p-1}-1\end{array}\right)---\left(7\right)$

Corresponding with it p level among the present invention, q multiple multiplying factor is:

As long as according to the butterfly computation structure, the mathematical result former state of each grade is write out, just can prove correctness of the present invention.To r _mThe fft algorithm that utilization is revised can get formula (9):

Formula (9) shows, fft algorithm structure provided by the invention, and equivalence is for to the reconstruct symbol as fractional part of frequency offset being earlier

Phase compensation, remake original FFT computing, reached the purpose that the frequency observation station is moved fractional part of frequency offset, be not have partially to estimate thereby make integer frequency offset of the present invention estimate.As long as the estimation of fractional part of frequency offset is more accurate, so, what new fft algorithm obtained will be not have partially to estimate.And the correction term in the fft algorithm provided by the invention is only carried out add operation on the phase place of plural number, also do not destroy the inherent characteristics of fft algorithm.The function that it is pointed out that fft algorithm of the present invention can realize with following alternative method: estimate according to fractional part of frequency offset

Earlier to r _mAfter doing time domain phase compensation, remake original FFT computing.

Though alternative method has also reached the purpose that the present invention designs new fft algorithm structure, its computation complexity is higher than algorithm provided by the invention.Specifically, for each r _m, alternative method need take advantage of number of times to be again

And the present invention only needs

Inferiorly take advantage of again.

Based on formula (9), the estimator of integer frequency offset is expressed as:

(4) merge frequency domain sequence,, estimate integer frequency offset by seeking peak value

In theory, the present invention only needs 1 r _mUtilization does not have the fft algorithm of estimating partially, just can estimate integer frequency offset, still, owing to influences such as the multipath fading of channel, noises, generally will adopt repeatedly the method for cascading judgement.Concrete operation method is as follows: by at random, uniformly-spaced or the continuous method of selecting, take out M r of λ _m(0＜λ≤1, and λ M is an integer).The random selection method is meant from M OFDM reconstruct symbol, takes out λ M different reconstruct symbol at random.The method of uniformly-spaced selecting is meant from M OFDM reconstruct symbol, determines at random earlier equally spaced to take out λ M different reconstruct symbol again by first reconstruct symbol.The method of selecting continuously is meant from M OFDM reconstruct symbol, determines at random earlier to take out λ M different reconstruct symbol more continuously by first reconstruct symbol.

To a λ M r _mUtilization does not have the fft algorithm of estimating partially, obtains a series of frequency domain data R _m, merging according to quadratic sum again, the pairing frequency of the peak value after the merging is exactly the estimated value of integer frequency offset.

Other merging methods have: the mould value merges, as the formula (12),

And the merging of real part absolute value and imaginary part absolute value sum, as the formula (13).

Because in 1 OFDM symbol, it is constant that signal to noise ratio can be thought, so formula (11) equivalence merges for high specific, this is optimum merging method.And formula (12) and formula (13) are the merging modes of suboptimum, have exchanged the decline of computation complexity for the decline of estimated performance.

(5) adaptive algorithm of feeding back based on decision reliability

When only getting a fixing λ, aforementioned schemes does not have too big advantage than conventional method, so the present invention further proposes the adaptive algorithm based on the decision reliability feedback.

Because error or nonideal channel condition that fractional part of frequency offset is estimated, in the formula (11)

Numerical value advantage with respect to other R (y) can significantly reduce, and causes the estimation of formula (11) unreliable.With the mean value of R (y) with

Ratio as formula (11) reliability index V, according to the comparable situation of V and a certain threshold value η, the method that adopts the multi-stage combination judgement is to improve estimated performance of the present invention.At first, carry out symbol reconstruct, obtain M OFDM reconstruct subsymbol r according to formula (3) _m, be designated as S set, set threshold value η, make n=1 (n represents the judgement of n level).Entering iterations then is the adaptive algorithm based on the decision reliability feedback of X:

From S, take out at random earlier ${\mathrm{\&lambda;}}_{n}M(0<{\mathrm{\&lambda;}}_n\&le;1-{\mathrm{\&Sigma;}}_{j=1}^{n-1}{\mathrm{\&lambda;}}_j$ And λ _nM is an integer) individual r _m, form S set _nS=S-S _nAgain to S _nIn each r _m,, obtain λ with fft algorithm calculating formula (9) _nM R _mThen λ _nM R _mMerge by formula (11), again with the results added of prime, U _n, it contains L element U _n(y):

$U_n\left(y\right)=U_{n-1}\left(y\right)+\underset{s_n}{\mathrm{\&Sigma;}}{\left|{\stackrel{\&OverBar;}{R}}_m\left(y\right)\right|}^2,(y\&Element;[-L/2,L/2),y\&Element;Z)---\left(14\right)$

The peak value frequency points corresponding of searching formula (14) is to f _IMake estimation:

${\hat{f}}_I=\underset{y}{\mathrm{aeg}\max }\left\{U_n\left(y\right)\right\}---\left(15\right)$

With U _n(y) mean value with Ratio as formula (15) reliability index V _n:

$V_n=\frac{\frac{1}{L}\underset{y}{\mathrm{\&Sigma;}}{U}_n\left(y\right)}{U_n\left({\hat{f}}_I\right)}---\left(16\right)$

Work as V _nSmaller or equal to thresholding η, or n=X, algorithm finishes; Otherwise, make n=n+1, repeat said process.

By formula (14) and formula (15) as can be known, the span of η be [1/L, 1), η is more little, the probability that calls back level judgement is just big more, so computation complexity rises, but estimated performance is improved.

It is pointed out that formula (16) is not unique expression of reliability index, other reliability indexs have:

U _n(y) second largest value with

Liken to and be reliability index V _n, as the formula (17),

$V_n=\frac{U_n\left(\underset{y}{\arg \max }\right\{U_n\left(y\right)\left|(y\&NotEqual;{\hat{f}}_I)\right\})}{U_n\left({\hat{f}}_I\right)}---\left(17\right)$

If V _nLess than threshold value α, think that then estimating is that the span of α is (0,1) reliably.

Adjacent two the value mean values with

Ratio as reliability index V _n, as the formula (18),

$V_n=\frac{[U_n({\hat{f}}_I+1)+U_n({\hat{f}}_I-1)]}{2U_n\left({\hat{f}}_I\right)}---\left(18\right)$

If V _nLess than threshold value, think that then estimating is that the span of β is (0,1) reliably.

After having adopted the adaptive algorithm based on the decision reliability feedback, computation complexity of the present invention further reduces, and estimated performance is significantly improved.

The invention has the advantages that: by the reconstruct of OFDM symbol, utilize fully and receive data, have the characteristics of time diversity; With fft algorithm integer frequency offset is done estimation, greatly reduce the computation complexity of system; In conjunction with the new fft algorithm of fractional part of frequency offset design, estimation of the present invention is become does not have estimation partially, has improved the reliability of system.In addition, the present invention adopts the self adaptation merge algorithm based on the decision reliability feedback, makes the present invention have the low and more excellent characteristics of estimated performance of computation complexity concurrently.And emulation shows that the error that the present invention estimates fractional part of frequency offset is insensitive, has good robustness.

Description of drawings

Fig. 1 OFDM baseband modulation and demodulation block diagram

Fig. 2 N=1024, the frequency domain of the synchronous training symbol of OFDM of L=8 and time domain corresponding relation schematic diagram

Fig. 3 N=1024, the synchronous training symbol reconstruct of the OFDM of L=8 schematic diagram

Fig. 4 enforcement block diagram of the present invention

Fig. 5 is used to estimate 8 fft algorithm structures of the inclined to one side estimation of nothing of integer frequency offset

Fig. 6 at random, uniformly-spaced or the systematic function figure of continuation method when selecting OFDM reconstruct symbol

Fig. 7 the present invention is adopting uniformly-spaced method to select OFDM reconstruct symbol, the performance map when having the fractional part of frequency offset evaluated error

Fig. 8 the present invention is adopting uniformly-spaced method to select OFDM reconstruct symbol, and fractional part of frequency offset estimates when entirely true, iterations X=0 based on the adaptive algorithm of decision reliability feedback and the performance comparison diagram of conventional method

Fig. 9 the present invention is adopting uniformly-spaced method to select OFDM reconstruct symbol, and fractional part of frequency offset estimates when entirely true, iterations X=1 based on the adaptive algorithm of decision reliability feedback and the performance comparison diagram of conventional method

Figure 10 the present invention is adopting uniformly-spaced method to select OFDM reconstruct symbol, and fractional part of frequency offset estimates when entirely true, iterations X=2 based on the adaptive algorithm of decision reliability feedback and the performance comparison diagram of conventional method

Embodiment

Provide a concrete OFDM parameter configuration below, set forth performing step of the present invention.Need to prove that the parameter in the following example does not influence generality of the present invention.

The document of 3GPP tissue: TR 25.892 V6.0.0, " Feasibility Study for OrthogonalFrequency Division Multiplexing (OFDM) for UTRAN enhancement (Release6) ", the one group of OFDM parameter that provides is as follows:

System bandwidth B 6.528MHz

Sub-carrier number N 1024

Effective sub-carrier number N _u705

Effective bandwidth 4.495MHz

Subcarrier spacing Δ f 6.375kHz

Cyclic extensions CP 64 (9.803us)

Symbol period T _s156.85+9.81=166.66us

Under the above-mentioned parameter condition, get L=8, then M=128, adopt the quadratic sum merging method of uniformly-spaced selecting method and formula (11), and employing formula (16) is as the computed reliability index, uses 3 level adaptation algorithms based on the decision reliability feedback, and performing step of the present invention is as follows:

(1) according to formula (1), in the subcarrier sequence number is 160,168,176,184,192,200,208,216,224,232,240,248,256,264,272,280,288,296,304,312,320,328,336,344,352,360,368,376,384,392,400,408,416,424,432,440,448,456,464,472,480,488,496,504,520,528,536,544,552,560,568,576,584,592,600,608,616,624,632,640,648,656,664,672,680,688,696,704,712,720,728,736,744,752,760,768,776,784,792,800,808,816,824,832,840,848,856, load pilot data on 864 the frequency, place remainder certificate on other subcarriers, with string and the modular converter 1 of such frequency domain data by Fig. 1, IDFT module 2, parallel serial conversion module 3, just can generate the 8 five equilibriums synchronous training symbol of a shape suc as formula (2), Fig. 2 is the frequency domain of this synchronous training symbol and the schematic diagram of time domain corresponding relation.Insert cyclic prefix module 4, insertion synchronizing information module 5, D/A modular converter 6, send Filtering Processing module 7 through the module of Fig. 1 then, arrive receiving terminal.

(2) receiving terminal is through the processing of accept filter among Fig. 1 processing module 8, A/D modular converter 9, and the bit synchronization information in conjunction with system among Fig. 4 provides will obtain complete r _{I, m}(m=0,1 ..., 1023).According to formula (3), with r _{I, m}Be reconstructed into 128 8 OFDM symbol r _m(m=1,2 ..., 128).Fig. 3 is the schematic diagram of this restructuring procedure.

(3) get λ ₁=0.25, λ ₂=0.25, λ ₃=0.5, X=3.Set threshold value η=0.5, make n=1 (n represents the judgement of n level).

(4) λ _nAfter the M round numbers be

Therefore, by equally spaced method, take out

Individual r _m, the fractional part of frequency offset estimated value that provides in conjunction with prime To above-mentioned

Individual r _mCarry out computing shown in Figure 5, obtain

Individual R _m

(5) to above-mentioned

Individual R _mMerge by formula (14), make the estimated value of integer frequency offset by formula (15).Calculate V by formula (16) _nWork as V _nSmaller or equal to thresholding η, or n=X, algorithm finishes; Otherwise, make n=n+1, get back to (4).

Channel is 8 road rayleigh fading channels, and is as follows:

Postpone (ns) relative power (dB)

Path 10 0

Path 2 153-6.7

Path 3 306-13.3

Path 4 459-19.9

Path 5 612-27.6

Path 6 765-33.3

Path 7 919-39.9

Path 8 1072-46.6

Fig. 6 is that the present invention adopts at random, uniformly-spaced or the systematic function figure of continuation method when selecting OFDM reconstruct symbol, this figure shows, when λ hour (as 0.1 or 0.2), the systematic function that adopts the continuous drawing method is not as other two kinds of methods, when λ is big, adopt at random, uniformly-spaced or the systematic function of continuation method when extracting reach unanimity.All in all, adopt the system of extracted at equal intervals method to have best performance.

To be the present invention adopting uniformly-spaced method to select OFDM reconstruct symbol to Fig. 7, and the performance map when having the fractional part of frequency offset evaluated error, and this figure shows, the error that the present invention estimates for fractional part of frequency offset is also insensitive, and this helps improving the robustness of whole system.

Fig. 8 is the present invention when adopting that uniformly-spaced method is selected OFDM reconstruct symbol, and fractional part of frequency offset estimates when entirely true, iterations X=0 based on the adaptive algorithm of decision reliability feedback and the performance comparison diagram of conventional method.In conventional method, setting time-delay is a slot, is correlated with, and asks phase angle, just can estimate integer frequency offset, and time take advantage of again its needs (N-M)=896.And in the present invention, λ=0.2 o'clock need be taken advantage of about 300 times again, and this is equivalent to 1/3 of conventional method, but obtains the snr gain of 0.5dB than conventional method; When λ=0.6, the present invention need take advantage of about 900 times again, and its computation complexity equates with conventional method, but has obtained the snr gain about 3dB.

Fig. 9 is the present invention when adopting that uniformly-spaced method extracts, and fractional part of frequency offset estimates when entirely true, iterations X=1 based on the adaptive algorithm of decision reliability feedback and the performance comparison diagram of conventional method.In this emulation, establish λ ₁=λ ₂=0.25,3 kinds of situations of investigation η: 0.5,0.75 and 0.875.When η=0.5, the present invention obtains the snr gain of 3dB than conventional method, and the adaptive algorithm emulation of X=1 shows, it needs on average takes advantage of number of times to have only 57% to 64% of conventional method again; When η=0.75, the present invention obtains the snr gain of 2.2dB than conventional method, on average takes advantage of number of times to have only 57% to 58% of conventional method again but it needs; When η=0.875, the present invention obtains the snr gain of 1.4dB than conventional method, on average takes advantage of number of times to have only 57% of conventional method again but it needs;

Figure 10 is the present invention when adopting that uniformly-spaced method extracts, and fractional part of frequency offset estimates when entirely true, iterations X=2 based on the adaptive algorithm of decision reliability feedback and the performance comparison diagram of conventional method.In this emulation, establish λ ₁=λ ₂=0.25, λ ₃=0.5,3 kinds of situations of investigation η: 0.5,0.75 and 0.875.When η=0.5, the present invention obtains the snr gain of 4.7dB than conventional method, and the adaptive algorithm emulation of X=2 shows, it needs on average takes advantage of number of times to have only 57% to 75% of conventional method again; When η=0.75, the present invention obtains the snr gain of 2.6dB than conventional method, on average takes advantage of number of times to have only 57% to 58% of conventional method again but it needs; When η=0.875, the present invention obtains the snr gain of 1.3dB than conventional method, on average takes advantage of number of times to have only 57% of conventional method again but it needs.

Simulation result shows, it is lower that the present invention has a computation complexity, and the lower advantage of estimated error rate, has very high using value in ofdm system.

Claims (7)

1, a kind of method of estimating OFDM integral number frequency multiplication bias is characterized in that, comprises the steps:

Step 1: transmitting terminal produces the synchronous training symbol of N point OFDM of a L five equilibrium, and each five equilibrium of described symbol contains the M point;

Step 2: described synchronous training symbol is reconstructed, be about to the symbol in the step 1, be reconstructed into M L point ofdm signal symbol, the ofdm signal symbol sampler point that is about to each five equilibrium carries out the order number designation, again the identical time domain point of label is extracted out in order, formed new OFDM symbol;

Step 3: λ M the reconstruct symbol that receiving terminal uses the fft algorithm demodulation to pick out from M reconstruct symbol of step 2 gained, obtain the frequency-region signal sequence of a described λ M reconstruct symbol, it is the fractional part of frequency offset estimated value that coupling system prime synchronization module provides, in the multiple multiplying factor of traditional FFT algorithm structure, introduce the fractional part of frequency offset correction term, finish fractional part of frequency offset compensation and demodulating ofdm reconstruct symbol simultaneously; Perhaps, behind the fractional part of frequency offset of employing direct compensation OFDM reconstruct symbol, with the traditional FFT algorithm OFDM reconstruct symbol is carried out demodulation again;

Step 4: the frequency-region signal sequence to FFT computing gained in the step 3 merges, and the pairing frequency of the peak value after the merging is exactly the estimated value of integer frequency offset;

Described N, L, M are the integer greater than 0, M=N/L, and described λ M is an integer, 0＜λ≤1.

2, the method for estimating OFDM integral number frequency multiplication bias according to claim 1 is characterized in that, described step 3 from M L point ofdm signal symbol, is taken out λ M different reconstruct symbol at random; Perhaps from M L point ofdm signal symbol, determine first reconstruct symbol earlier at random, equally spaced take out λ M different reconstruct symbol again; Perhaps from M L point ofdm signal symbol, determine first reconstruct symbol earlier at random, take out λ M different reconstruct symbol more continuously.

3, the method for estimating OFDM integral number frequency multiplication bias according to claim 1, it is characterized in that described step 4, the merging method that merges the frequency-region signal sequence of OFDM reconstruct symbol are that quadratic sum merges, perhaps mould value and merging, the perhaps merging of real part absolute value and imaginary part absolute value sum.

4, the method for estimating OFDM integral number frequency multiplication bias according to claim 1, it is characterized in that, after described step 4, enter adaptive iteration algorithm based on the decision reliability feedback, promptly pass through the reliability of appraisal procedure four gained estimated values, increase the value of λ adaptively, repeating step three and step 4 obtain integer frequency offset estimated value comparatively reliably at last.

5, the method for estimating OFDM integral number frequency multiplication bias according to claim 4, it is characterized in that, the reliability of described appraisal procedure four gained estimated values, its method is the reliability index into step 4 gained estimated value of likening to of the mean value of the frequency-region signal sequence that obtains of calculation procedure four and maximum, if reliability index is less than threshold value η, think that then it is reliable estimating, otherwise, system will increase the value of λ, repeating step three and step 4, obtain integer frequency offset estimated value comparatively reliably at last, the span of described η is [1/L, 1].

6, the method for estimating OFDM integral number frequency multiplication bias according to claim 4, it is characterized in that, the reliability of described appraisal procedure four gained estimated values, the second largest value of the frequency-region signal sequence that obtains in calculation procedure four and maximum liken reliability index to into step 4 gained estimated value, if reliability index is less than threshold value α, think that then it is reliable estimating, otherwise, system will increase the value of λ, repeating step three and step 4, obtain integer frequency offset estimated value comparatively reliably at last, the span of described α is (0,1).

7, the method for estimating OFDM integral number frequency multiplication bias according to claim 4, it is characterized in that, the reliability of described appraisal procedure four gained estimated values, the mean value of adjacent two values of the frequency-region signal sequence that obtains in calculation procedure four and maximum liken reliability index to into step 4 gained estimated value, if reliability index is less than threshold value, think that then it is reliable estimating, otherwise, system will increase the value of λ, repeating step three and step 4, obtain integer frequency offset estimated value comparatively reliably at last, the span of described β is (0,1).

Images