CN101588338B - OFDM carrier frequency offset estimation method suitable for packet transmission - Google Patents

Abstract

The present invention provides an OFDM carrier frequency offset estimation method for packet transmission, which comprises the following steps: 1) performing IFFT conversion on frequency domain symbols to obtain time domain training symbols of two same repetition modules; 2) adding cyclic prefixes with the length of L on the front of the time domain training symbols to form integral OFDM symbols, and obtaining received OFDM signals carrying frequency offset epsilon through a multipath channel; 3) rotating phases with pi epsilon by N/2 spaced received sample values in the received OFDM signals to obtain fine frequency offset estimation; 4) performing fine frequency offset compensation on the received OFDM signals to obtain OFDM signals after the compensation; 5) performing FFT conversion on training symbol parts of the compensated OFDM symbols after the cyclic prefixes are removed so as to obtain receiving end modulation frequency domain symbols; 6) detecting peaks of cyclic shift of the receiving end frequency domain symbols and sending end frequency domain symbols to obtain coarse frequency offset estimation; and 7) performing coarse frequency offset compensation on the receiving end modulation frequency domain symbols in a receiving end frequency domain, and summing up the coarse frequency offset estimation and the fine frequency offset estimation to obtain the carrier frequency offset estimation.

Description

A kind of OFDM carrier frequency bias estimation that is applicable to transmitted in packets

Technical field

The present invention relates to be applicable in a kind of orthogonal frequency division multiplex OFDM system the OFDM carrier frequency bias estimation of transmitted in packets, belong to the broadband wireless communication technique field, can be applicable in the next generation mobile communication system.

Background technology

OFDM (OFDM) technology is owing to have good anti-multipath interference performance and high spectrum utilization ratio, be subjected to research [document [1] Van Nee R widely, Prasad R.OFDM for Wireless Multi-media Communications[M] .Boston, London:Artech House, 2000:73-93.].But ofdm system is to very responsive [document [2] Moose P H.A technique for orthogonal frequency division multiplexing frequency offset correction[J] the .IEEETrans.on Communications of frequency deviation, 1994,42 (10): 2908-2914.].Because when having frequency departure, orthogonality between subcarrier is destroyed, cause inter-carrier interference, thereby have a strong impact on performance [document [2] Moose P H.Atechnique for orthogonal frequency division multiplexing frequency offset correction[J] the .IEEE Trans.on Communications of OFDM receiving terminal, 1994,42 (10): 2908-2914., document [3] Modelski, J, Oziewicz, M.Distortions ofthe OFDM Sub-carriers in SFN Baseband Channel[C] .International Conference on Computer as aTool, Warsaw, Sep.9-12,2007,919-925.].Document [document [4] Van de Beek J J, Sandell M, Brjesson P O.ML estimation of timing and frequency offset in OFDM systems[J] .IEEE Trans.on Signal Processing, 1997,45 (7): 1800-1805.] the analysis showed that, reach negligible degree if the signal to noise ratio that requires to be caused by carrier wave frequency deviation descends, so carrier frequency shift should be limited in subcarrier spacing 2% within.

There are a lot of documents that the Nonlinear Transformation in Frequency Offset Estimation that is fit to the OFDM transmitted in packets is inquired into.These methods are mostly by designing a time domain training symbol that comprises some replicated blocks, obtain Nonlinear Transformation in Frequency Offset Estimation at receiving terminal according to the phase place rotation of replicated blocks then.Modal training symbol is exactly Schmidl[document [5] Schmidl T M, Cox D C.Robust frequency and timingsynchronization for OFDM[J] .IEEE Trans.on Communications, 1997,45 (12): 1613-1621.] She Ji the training symbol that comprises two identical replicated blocks, its frequency offset estimation range are ± 1 subcarrier spacing.But in real system, the absolute value of normalization carrier wave frequency deviation may be greater than 1, so Schmidl designed second training symbol and enlarge frequency offset estimation range, but the number that increases training symbol can reduce the efficiency of transmission of block transmission system.

For improving the efficient of transmission, document [[6] Morelli M, Mengali U.An improved frequency offset estimatorfor OFDM applications[J] .IEEE Communications Letters, 1999,3 (3): 75-77., document [7] Song H K, You Y H, Paik J H et al..Frequency-offset synchronization and channel estimation for OFDM-basedtransmission[J] .IEEE Communications Letters, 2000,4 (3): 95-97.] designed frequency offset estimating scheme based on a training symbol, this training symbol comprises several (more than or equal to 2) modules.Wherein, Morelli and Mengali (M﹠amp; M) [document [6] Morelli M, Mengali U.An improved frequency offset estimator for OFDM applications[J] .IEEE Communications Letters, 1999,3 (3): 75-77.] do not have inclined to one side criterion according to optimum linearity and designed a frequency offset estimator, its estimated accuracy has had certain raising, estimation range also enlarges, but computation complexity also improves thereupon, and estimation range is limited.Song[document [7] Song H K, You Y H, Paik J H et al..Frequency-offset synchronization andchannel estimation for OFDM-based transmission[J] .IEEE Communications Letters, 2000,4 (3): 95-97.] use the thought of multistage frequency offset estimating to enlarge estimation range and improve estimated accuracy, and M﹠amp; M is the same, and also there is the high and little deficiency of estimation range of computation complexity in it.

In recent years, many new methods are proposed.Document [[8] Yi Q M, Shi M.A New Algorithm of frequency offsetEstimation[C] .International Conference on Machine Learning and Cybernetics, Hong Kong, Aug.19-22,2007,2,813-817.] by the cost function of design with the frequency deviation variation, make frequency offset estimating be easy to hardware and realize, but do not enlarge estimation range.Document [[9] Yu Z H, Chen K, Huang Y M, Zhang Yang.OFDM Timing andFrequency Offset Estimation Based on Repeated Training Sequence[C] .International Conferenceon Wireless Communications, Networking and Mobile Computing, Shanghai, Sep.21-25,2007,264-266] designed the frequency offset estimating algorithm of suitable multipath channel, make estimated accuracy improve, but estimation range is still limited.Document [[10] Guo Y, Liu G, Ge J H.A novel time and frequency synchronization scheme for OFDMsystems[J] .IEEE Trans.on Consumer Electronics, 2008,54 (2): 321-325.] frequency deviation estimating method of Ti Chuing, enlarge estimation range, but but do not improved estimated accuracy.

Summary of the invention

The objective of the invention is at the deficiencies in the prior art, propose the OFDM carrier frequency bias estimation that is applicable to transmitted in packets that a kind of estimation range is big and estimated accuracy is high.

A kind of OFDM carrier frequency bias estimation that is applicable to transmitted in packets, it carries out following steps in ofdm system:

1) by frequency domain symbol S _l(l=0,1 ..., N-1) through the IFFT conversion, obtain comprising the time domain training symbol (s of two identical replicated blocks _n, n=0,1 ..., N-1);

2) time domain training symbol (s _n, n=0,1 ..., N-1), adding length is the complete OFDM symbol of Cyclic Prefix formation of L, i.e. { s _-L, s _-L+1... s _-1, s ₀, s ₁..., s _N-1; Through multipath channel, because the difference of transmit end receive end crystal oscillation frequency obtains carrying the reception OFDM symbol (r of frequency deviation information (being ε after the normalization) _n, n=-L ,-L+1 ..., N-1);

3) according to receiving OFDM symbol (r _n, n=-L ,-L+1 ..., the sample value of the N/2 of being separated by in N-1) obtains thin frequency offset estimating to there being the phase place rotation of π ε

4) obtain thin frequency offset estimating

Afterwards, to receiving OFDM symbol (r _n, n=-L ,-L+1 ..., N-1) carry out after thin compensate of frequency deviation is compensated the OFDM symbol (r ' _n, n=-L ,-L+1 ..., N-1);

5) to the compensation after the OFDM symbol (r ' _n, n=-L ,-L+1 ..., remove in N-1) training symbol part behind the Cyclic Prefix (r ' _n, n=0,1 ..., N-1) carry out the FFT conversion, obtain the frequency domain symbol (R of receiving terminal demodulation _k, k=0,1 ..., N-1);

6) utilize the good autocorrelation performance of pseudo random sequence, by receiving terminal frequency domain symbol (R _k, k=0,1 ..., N-1) with transmitting terminal frequency domain symbol (S _l, l=0,1 ..., the N-1) correlation peak detection of cyclic shift can obtain thick frequency offset estimating

7) according to thick frequency offset estimating Directly at the receiving terminal frequency domain to receiving terminal demodulating data (R _k, k=0,1 ..., N-1) do the related offset compensation, Nonlinear Transformation in Frequency Offset Estimation is thick frequency offset estimating

With thin frequency offset estimating

Sum.

The time domain training symbol that the present invention utilizes the inventor to design is divided into the thin frequency offset estimating of receiving terminal time domain to the frequency offset estimating of OFDM and two steps of thick frequency offset estimating of frequency domain realize.The present invention is by the relevant information of corresponding module in the relevant information of replicated blocks and Cyclic Prefix and the training symbol in the comprehensive utilization training symbol, carry out thin frequency offset estimating, utilize thick frequency offset estimating to expand frequency offset estimation range to whole signal bandwidth by ± 1 subcarrier spacing then.

The time domain training symbol that uses among the present invention, with Schmidl[document [5] Schmidl T M, Cox D C.Robust frequencyand timing synchronization for OFDM[J] .IEEE Trans.on Communications, 1997,45 (12): 1613-1621.] the time domain training symbol similar in the method, the module that also comprises two identical repetitions, but production method difference.Fig. 1 has provided the generation frame principle figure of these two kinds of method time domain training symbols.Two kinds of methods the odd indexed of frequency domain 1,3 ..., all put 0 on the N-1} subcarrier, and the even number sequence number 0,2 ..., then send different sequences on the N-2} subcarrier.What send in the Schmidl method is pseudorandom QPSK symbol $\stackrel{\&RightArrow;}{Q}=({\mathrm{<figure-callout\; id="0"\; label="Q"\; filenames="H2009100206380E00011.png,H2009100206380E00022.png"\; state="\{\{state\}\}">Q</figure-callout>}}_0,{\mathrm{<figure-callout\; id="1"\; label="Q"\; filenames="H2009100206380E00011.png"\; state="\{\{state\}\}">Q</figure-callout>}}_1,...,Q_{N/2-1}),$ And what send in the method that we propose is pseudo random sequence $\stackrel{\&RightArrow;}{P}=({\mathrm{<figure-callout\; id="0"\; label="P"\; filenames="H2009100206380E00011.png,H2009100206380E00022.png"\; state="\{\{state\}\}">P</figure-callout>}}_0,{\mathrm{<figure-callout\; id="1"\; label="P"\; filenames="H2009100206380E00011.png"\; state="\{\{state\}\}">P</figure-callout>}}_1,...,P_{N/2-1}).$

In the carrier frequency bias estimation that we propose, the pseudo random sequence of use has good autocorrelation performance, and corresponding frequency domain symbol is provided by following formula

$S_l=\left\{\begin{array}{c}\sqrt{2}{P}_m,l=2m,m=\mathrm{0,1,2},...,N/2-1\\ 0,l=\mathrm{1,3},...,N-1\end{array}\right.---\left(1\right)$

Wherein N is the subcarrier number, P _m(m=0,1,2 ..., N/2-1) be m value in the pseudo random sequence, and P _m∈ 1,1}, S _l(l=0,1 ..., N-1) be value on l subcarrier of frequency domain symbol.

Next modulation produces time domain training symbol (module that comprises two identical repetitions) through IFFT

$s_n=\frac{1}{N}\underset{l=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{S}_l\exp \left(\frac{j2\mathrm{\&pi;nl}}{N}\right),n=\mathrm{0,1},...,N-1---\left(2\right)$

Wherein N is the subcarrier number, s _n(n=0,1 ..., N-1) be n sample value of time domain training symbol, S _l(l=0,1 ..., N-1) be value on l subcarrier of frequency domain symbol, the j representative

For eliminating intersymbol interference, before each training symbol, insert the Cyclic Prefix of length L, L is greater than the corresponding sample value number of multipath channel maximum delay expansion, that is and, the OFDM symbol of actual transmissions is: { s _-L, s _-L+1... s _-1, s ₀, s ₁..., s _N-1, a preceding L sample value is duplicating of L the sample value in back.

At receiving terminal, if regularly accurately, the reception OFDM symbol when to have normalization frequency deviation ε be the ratio of actual frequency deviation and subcarrier spacing is so

$r_n=s_n\exp \left(\frac{\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n\&epsiv;}}{N}\right)+w_n,n=-L,-L+1,...,\mathrm{0,1},...,N-1---\left(3\right)$

Wherein N is the subcarrier number, and L is a circulating prefix-length, the j representative

w _n(n=-L ,-L+1 ..., 0,1 ..., be that average is 0 N-1), variance is σ _w ²N sample value of additive white Gaussian noise, s _n(n=-L ,-L+1 ..., 0,1 ..., N-1) be n sample value of time domain OFDM symbol, ε is that the normalization frequency deviation is the ratio of actual frequency deviation and subcarrier spacing, r _n(n=-L ,-L+1 ..., 0,1 ..., N-1) be n the sample value that receives the OFDM symbol.

Because the existence of carrier wave frequency deviation, make in OFDM symbol period of receiving terminal, be separated by the sample value of N/2 to there being the phase place rotation of π ε, identical replicated blocks is relevant in the comprehensive utilization training symbol, and the corresponding sequence of first relevant in Cyclic Prefix and the training symbol, can obtain thin frequency offset estimating, promptly

${\mathrm{\&gamma;}}_1=\underset{n=0}{\overset{N/2-1}{\mathrm{\&Sigma;}}}{r}_n^{*}{r}_{n+N/2}---\left(4\right)$

${\mathrm{\&gamma;}}_2=\underset{n=-L}{\overset{-1}{\mathrm{\&Sigma;}}}{r}_n^{*}{r}_{n+N/2}---\left(5\right)$

${\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}=\frac{1}{\mathrm{\&pi;}}\frac{N/2}{N/2+L}\mathrm{angle}\left({\mathrm{\&gamma;}}_1\right)+\frac{1}{\mathrm{\&pi;}}\frac{L}{N/2+L}\mathrm{angle}\left({\mathrm{\&gamma;}}_2\right)---\left(6\right)$

Subscript symbol wherein ^*Complex conjugate is got in expression, and the argument of () is got in angle () expression, and N is the subcarrier number, and L is a circulating prefix-length, r _n ^*Be the complex conjugate that receives n sample value of OFDM symbol, r _N+N/2Be n+N/2 the sample value that receives the OFDM symbol, γ ₁Be the correlation that receives the identical replicated blocks of training symbol in the OFDM symbol, γ ₂Be Cyclic Prefix and the right correlation of the corresponding sequence of training symbol first,

Estimated value for thin frequency deviation.

Because the periodicity of phase function is so actual frequency deviation has following properties

$\mathrm{\&epsiv;}={\mathrm{\&epsiv;}}_{\mathrm{coarse}}+{\mathrm{\&epsiv;}}_{\mathrm{fine}}=2g+{\mathrm{\&epsiv;}}_{\mathrm{fine}}\&ap;2g+{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}---\left(7\right)$

Wherein g is an integer, ε _FineAnd ε _CoarseBe respectively thin frequency deviation and thick frequency deviation, ε is the ratio of actual frequency deviation and subcarrier spacing for actual normalization frequency deviation,

Be thin frequency offset estimating value.

If after thin frequency offset estimating, the absolute value of the normalization carrier wave frequency deviation that obtains can guarantee that then frequency offset estimating only just can be obtained by thin frequency offset estimating less than 1.Otherwise, after thin frequency offset estimating, also to carry out following thick frequency offset estimating.

After thin frequency offset estimating, need to carry out thin compensate of frequency deviation to receiving the OFDM symbol earlier, promptly

$r_n^{\&prime;}=r_n\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)$

$=s_n\exp \left(\frac{\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}(\mathrm{\&epsiv;}-{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}})}{N}\right)+w_n\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)$

$=s_n\exp \left(\frac{\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}\left(2g\right)}{N}\right)+w_n\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)---\left(8\right)$

Wherein N is the subcarrier number, r _nBe n the sample value that receives the OFDM symbol, r ' _nBe n sample value of the reception OFDM symbol behind meticulous compensate of frequency deviation,

Be thin frequency offset estimating value, s _nBe n sample value of time domain OFDM symbol, w _nBe n sample value of additive white Gaussian noise, ε is actual normalization frequency deviation, and g is the number of ambiguity period, the j representative

Then, to the compensation after the OFDM symbol (r ' _n, n=-L ,-L+1 ..., remove in N-1) training symbol part behind the Cyclic Prefix (r ' _n, n=0,1 ..., N-1) do the FFT conversion; (r ' _n, n=0,1 ..., N-1) transform to frequency domain after because the existence of even number frequency deviation, cyclic shift will appear in the frequency domain symbol that recovers to obtain, promptly

$R_k=\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{r}_n^{\&prime;}\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;kn}}{N}\right)$

$=S_{{(k-2g)}_N}+W_k^{\&prime;},k=\mathrm{0,1},..,N-1---\left(9\right)$

Wherein N is the subcarrier number, and g is the number of ambiguity period, the j representative

R ' _n(n=0,1 ..., N-1) be n sample value of the reception training symbol behind the thin compensate of frequency deviation, R _k(k=0,1 ..., N-1) be value on k the subcarrier of the frequency domain symbol that obtains of receiving terminal demodulation,

Be the value S on k subcarrier of transmitting terminal frequency domain symbol _kWith N move to right value behind the 2g of loop cycle, W ' _kBe the additive white Gaussian noise of k subchannel, promptly

$W_k^{\&prime;}=\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\left(w_n\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)\right)\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;kn}}{N}\right)---\left(10\right)$

Because what send on the even subcarriers of transmitting terminal frequency domain is pseudo random sequence, the even-multiple frequency deviation only can cause the cyclic shift of receiving symbol, so utilize the good autocorrelation performance of pseudo random sequence, thick frequency offset estimating can detect by following peak value and obtain

${\widehat{\mathrm{\&epsiv;}}}_{\mathrm{coarse}}=2\left\{\underset{g}{\arg \max }\right[\frac{{\left|\underset{k=0}{\overset{N/2-1}{\mathrm{\&Sigma;}}}{R}_{2k}^{*}{S}_{{(2k-2g)}_N}\right|}^2}{{\left(\underset{k=0}{\overset{N/2-1}{\mathrm{\&Sigma;}}}{\left|R_{2k}\right|}^2\right)}^2}\left]\right\},g=-N/4,-N/4+1,...,N/4---\left(11\right)$

Wherein N is the subcarrier number, and g is the number of ambiguity period, the subscript symbol ^*Complex conjugate, R are got in expression _2k ^*Be the complex conjugate of 2k value of the frequency domain symbol that obtains of receiving terminal demodulation,

Be the value S on 2k subcarrier of transmitting terminal frequency domain symbol _2kWith N move to right result behind the 2g of loop cycle,

Be the estimated value of thick frequency deviation,

The g value of correspondence when { } maximum is got in expression;

By formula (11), by thick frequency offset estimating, frequency offset estimation range can expand whole signal bandwidth to by ± 1 subcarrier spacing, and promptly (N/2, N/2).

At last, by (6) and (11) formula, the Nonlinear Transformation in Frequency Offset Estimation that can obtain ofdm system is

$\widehat{\mathrm{\&epsiv;}}={\widehat{\mathrm{\&epsiv;}}}_{\mathrm{coarse}}+{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}---\left(12\right)$

Wherein

Be the estimated value of the thick frequency deviation that obtains by (11) formula,

Be the estimated value of the thin frequency deviation that obtains by (6) formula,

It is the estimated value of actual normalization frequency deviation.

At the receiving terminal frequency domain to thick frequency deviation

Compensate, just directly to frequency domain symbol (R after the receiving terminal demodulation _k, k=0,1 ..., N-1) carry out corresponding cyclic shift, promptly

$R_k^{\&prime;}=R_{{(k+{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{coarse}})}_N},k=\mathrm{0,1},...,N-1---\left(13\right)$

Wherein N is the subcarrier number,

Be the estimated value of the thick frequency deviation that obtains by (11) formula, R ' _k(k=0,1 ..., N-1) be value on k the subcarrier of the receiving terminal demodulation frequency domain symbol behind the thick compensate of frequency deviation,

Be the value R on k subcarrier of receiving terminal demodulation frequency domain symbol _kWith N is that loop cycle moves to left

After value.

Outstanding advantage of the present invention is: only adopt a training symbol, improved the validity of transmission; Frequency offset estimation range expands whole signal bandwidth to; The frequency offset estimation accuracy height.Under multidiameter fading channel, when signal to noise ratio reaches 12dB, can make inherent spurious frequency deviation be limited in subcarrier spacing 2% in.

Description of drawings

The generation functional-block diagram of Fig. 1 time domain training symbol;

The Nonlinear Transformation in Frequency Offset Estimation model of this algorithm of Fig. 2;

The thin frequency offset estimating block diagram of this algorithm of Fig. 3;

Fig. 4 is under the A channel, when Doppler frequency shift is 0, and when signal to noise ratio is 20dB, estimation range of this algorithm and Schmidl, M﹠amp; The comparison of M and Song algorithm estimation range;

Fig. 5 is under the A channel, when Doppler frequency shift is 0, and when signal to noise ratio is 20dB, this algorithm and Schmidl, M﹠amp; The contrast that the frequency offset estimating mean square error of M and Song algorithm changes with the normalization frequency deviation;

Fig. 6 is under the A channel, when Doppler frequency shift is respectively 0 and during 172Hz, this algorithm and Schmidl, M﹠amp; The contrast that the frequency offset estimating mean square error of M and Song algorithm changes with signal to noise ratio;

Fig. 7 is for being under 0 condition at Doppler frequency shift, when multipath channel is respectively A and B channel, and this algorithm and Schmidl, M﹠amp; The contrast that the frequency offset estimating mean square error of M and Song algorithm changes with signal to noise ratio;

Fig. 8 a is under the A channel, when Doppler frequency shift is 0, and when signal to noise ratio is respectively 6dB, the tracking performance of the frequency deviation of this algorithm;

When Fig. 8 b is 9dB, the tracking performance of the frequency deviation of this algorithm;

When Fig. 8 c is 12dB, the tracking performance of the frequency deviation of this algorithm.

Embodiment

The invention will be further described below in conjunction with accompanying drawing and embodiment.

The present invention is directed to the estimated accuracy that exists in the Nonlinear Transformation in Frequency Offset Estimation and the contradiction of estimation range, proposed a kind of effective Nonlinear Transformation in Frequency Offset Estimation scheme.

The present invention has additionally used the redundant information of Cyclic Prefix in thin frequency offset estimating, i.e. the corresponding sequence of first relevant in Cyclic Prefix and the training symbol obviously improved the estimated accuracy of carrier wave frequency deviation.And uniting of thin frequency offset estimating of time domain and the thick frequency offset estimating of frequency deviation makes the estimation range of carrier wave frequency deviation expand whole signal bandwidth to.Thereby the contradiction of the estimated accuracy of efficiently solving and estimation range.

A kind of OFDM frequency deviation estimating method that is applicable to transmitted in packets, in ofdm system, carry out following steps:

1) according to Fig. 1 by frequency domain symbol (S _l, l=0,1 ..., N-1) comprise the time domain training symbol (s of two equal modules through the IFFT shift design _n, n=0,1 ..., N-1);

2) time domain training symbol (s _n, n=0,1 ..., N-1), adding length is the complete OFDM symbol of Cyclic Prefix formation of L, i.e. { s _-L, s _-L+1... s _-1, s ₀, s ₁..., s _N-1.Through multipath channel, because the difference of transmit end receive end crystal oscillation frequency obtains carrying the reception OFDM symbol (r of frequency deviation information (being ε after the normalization) _n, n=-L ,-L+1 ..., N-1);

3) receive OFDM symbol (r _n, n=-L ,-L+1 ..., the sample value of the N/2 of being separated by in N-1) is to { r _n, r _N+N/2(n=-L ,-L+1 ..., N/2-1) exist the phase place of π ε to rotate, obtain thin frequency offset estimating according to Fig. 3

4) obtain thin frequency offset estimating

Afterwards, according to Fig. 2, to receiving OFDM symbol (r _n, n=-L ,-L+1 ..., N-1) carry out after thin compensate of frequency deviation is compensated the OFDM symbol (r ' _n, n=-L ,-L+1 ..., N-1);

5) to the compensation after the OFDM symbol (r ' _n, n=-L ,-L+1 ..., remove in N-1) training symbol part behind the Cyclic Prefix (r ' _n, n=0,1 ..., N-1) carry out the FFT conversion, obtain the frequency domain symbol (R of receiving terminal demodulation _k, k=0,1 ..., N-1);

6) utilize the good autocorrelation performance of pseudo random sequence, by receiving terminal frequency domain symbol (R _k, k=0,1 ..., N-1) with transmitting terminal frequency domain symbol (S _l, l=0,1 ..., the N-1) correlation peak detection of cyclic shift can obtain thick frequency offset estimating

7) according to thick frequency offset estimating

Directly at the receiving terminal frequency domain to receiving terminal demodulation symbol (R _k, k=0,1 ..., N-1) do the related offset compensation, Nonlinear Transformation in Frequency Offset Estimation is thick frequency offset estimating

With thin frequency offset estimating

Sum.

In the described step 1), in the carrier frequency bias estimation, the pseudo random sequence of use has good autocorrelation performance, and corresponding frequency domain symbol is provided by following formula

$S_l=\left\{\begin{array}{c}\sqrt{2}{P}_m,l=2m,m=\mathrm{0,1,2},...,N/2-1\\ 0,l=\mathrm{1,3},...,N-1\end{array}\right.---\left(1\right)$

Wherein N is the subcarrier number, P _m(m=0,1,2 ..., N/2-1) be m value in the pseudo random sequence, and P _m∈ 1,1}, S _l(l=0,1 ..., N-1) be value on l subcarrier of frequency domain symbol;

Next modulation produces the time domain training symbol through IFFT, and it comprises the module of two identical repetitions

$s_n=\frac{1}{N}\underset{l=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{S}_l\exp \left(\frac{\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;nl}}{N}\right),n=\mathrm{0,1},...,N-1---\left(2\right)$

Wherein N is the subcarrier number, s _n(n=0,1 ..., N-1) be n sample value of time domain training symbol, S _l(l=0,1 ..., N-1) be value on l subcarrier of frequency domain symbol, the j representative

Step 2) in, for eliminating intersymbol interference, insert the Cyclic Prefix of length L before each training symbol, L is greater than the corresponding sample value number of multipath channel maximum delay expansion, that is, the OFDM symbol of actual transmissions is: { s _-L, s _-L+1... s _-1, s ₀, s ₁..., s _N-1, a preceding L sample value is duplicating of L the sample value in back;

At receiving terminal, if regularly accurately, the reception OFDM symbol when to have normalization frequency deviation ε be the ratio of actual frequency deviation and subcarrier spacing is so

$r_n=s_n\exp \left(\frac{\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n\&epsiv;}}{N}\right)+w_n,n=-L,-L+1,...,\mathrm{0,1},...,N-1---\left(3\right)$

Wherein N is the subcarrier number, and L is a circulating prefix-length, the j representative

w _n(n=-L ,-L+1 ..., 0,1 ..., be that average is 0 N-1), variance is σ _w ²N sample value of additive white Gaussian noise, s _n(n=-L ,-L+1 ..., 0,1 ..., N-1) be n sample value of time domain OFDM symbol, ε is that the normalization frequency deviation is the ratio of actual frequency deviation and subcarrier spacing, r _n(n=-L ,-L+1 ..., 0,1 ..., N-1) be n the sample value that receives the OFDM symbol.

In the step 3), because the existence of carrier wave frequency deviation, make in OFDM symbol period of receiving terminal, be separated by the sample value of N/2 to there being the phase place rotation of π ε, identical replicated blocks is relevant in the comprehensive utilization training symbol, and right relevant of the corresponding sample value of first in Cyclic Prefix and the training symbol, can obtain thin frequency offset estimating, promptly

${\mathrm{\&gamma;}}_1=\underset{n=0}{\overset{N/2-1}{\mathrm{\&Sigma;}}}{r}_n^{*}{r}_{n+N/2}---\left(4\right)$

${\mathrm{\&gamma;}}_2=\underset{n=-L}{\overset{-1}{\mathrm{\&Sigma;}}}{r}_n^{*}{r}_{n+N/2}---\left(5\right)$

${\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}=\frac{1}{\mathrm{\&pi;}}\frac{N/2}{N/2+L}\mathrm{angle}\left({\mathrm{\&gamma;}}_1\right)+\frac{1}{\mathrm{\&pi;}}\frac{L}{N/2+L}\mathrm{angle}\left({\mathrm{\&gamma;}}_2\right)---\left(6\right)$

Subscript symbol wherein ^*Complex conjugate is got in expression, and the argument of () is got in angle () expression, and N is the subcarrier number, and L is a circulating prefix-length, r _n ^*Be the complex conjugate that receives n sample value of OFDM symbol, r _N+N/2Be n+N/2 the sample value that receives the OFDM symbol, γ ₁Be the correlation that receives the identical replicated blocks of training symbol in the OFDM symbol, γ ₂Be Cyclic Prefix and the right correlation of the corresponding sample value of training symbol first,

Estimated value for thin frequency deviation;

Because the periodicity of phase function is so actual frequency deviation has following properties

$\mathrm{\&epsiv;}={\mathrm{\&epsiv;}}_{\mathrm{coarse}}+{\mathrm{\&epsiv;}}_{\mathrm{fine}}=2g+{\mathrm{\&epsiv;}}_{\mathrm{fine}}\&ap;2g+{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}---\left(7\right)$

Wherein g is an integer, ε _FineAnd ε _CoarseBe respectively thin frequency deviation and thick frequency deviation, ε is the ratio of actual frequency deviation and subcarrier spacing for actual normalization frequency deviation,

Be thin frequency offset estimating value.

In the step 4), after thin frequency offset estimating, need to carry out thin compensate of frequency deviation to receiving the OFDM symbol earlier, promptly

$r_n^{\&prime;}=r_n\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)$

$=s_n\exp \left(\frac{\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}(\mathrm{\&epsiv;}-{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}})}{N}\right)+w_n\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)$

$=s_n\exp \left(\frac{\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}\left(2g\right)}{N}\right)+w_n\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)---\left(8\right)$

Wherein N is the subcarrier number, r _nBe n the sample value that receives the OFDM symbol, r ' _nBe n sample value of the reception OFDM symbol behind meticulous compensate of frequency deviation,

Be thin frequency offset estimating value, s _nBe n sample value of time domain OFDM symbol, w _nBe n sample value of additive white Gaussian noise, ε is actual normalization frequency deviation, and g is the number of ambiguity period, the j representative

In the step 5), to the compensation after the OFDM symbol (r ' _n, n=-L ,-L+1 ..., remove in N-1) training symbol part behind the Cyclic Prefix (r ' _n, n=0,1 ..., N-1) do the FFT conversion; (r ' _n, n=0,1 ..., N-1) transform to frequency domain after because the existence of even number frequency deviation, cyclic shift will appear in the frequency domain symbol that recovers to obtain, promptly

$R_k=\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{r}_n^{\&prime;}\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;kn}}{N}\right)$

$=S_{{(k-2g)}_N}+W_k^{\&prime;},k=\mathrm{0,1},..,N-1---\left(9\right)$

Wherein N is the subcarrier number, and g is the number of ambiguity period, the j representative

R ' _n(n=0,1 ..., N-1) be n sample value of the reception training symbol behind the thin compensate of frequency deviation, R _k(k=0,1 ..., N-1) be value on k the subcarrier of the frequency domain symbol that obtains of receiving terminal demodulation, Be the value S on k subcarrier of transmitting terminal frequency domain symbol _kWith N move to right result behind the 2g of loop cycle, W ' _kBe the additive white Gaussian noise of k subchannel, promptly

$W_k^{\&prime;}=\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\left(w_n\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)\right)\exp \left(\frac{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="H2009100206380E00011.png,H2009100206380E00021.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;kn}}{N}\right)---\left(10\right)$

In the step 6), because what send on the even subcarriers of transmitting terminal frequency domain is pseudo random sequence, the even-multiple frequency deviation only can cause the cyclic shift of receiving symbol, so utilize the good autocorrelation performance of pseudo random sequence, thick frequency offset estimating can detect by following peak value and obtain

${\widehat{\mathrm{\&epsiv;}}}_{\mathrm{coarse}}=2\left\{\underset{g}{\arg \max }\right[\frac{{\left|\underset{k=0}{\overset{N/2-1}{\mathrm{\&Sigma;}}}{R}_{2k}^{*}{S}_{{(2k-2g)}_N}\right|}^2}{{\left(\underset{k=0}{\overset{N/2-1}{\mathrm{\&Sigma;}}}{\left|R_{2k}\right|}^2\right)}^2}\left]\right\},g=-N/4,-N/4+1,...,N/4---\left(11\right)$

Wherein N is the subcarrier number, and g is the number of ambiguity period, the subscript symbol ^*Complex conjugate, R are got in expression _2k ^*Be the complex conjugate of 2k value of the frequency domain symbol that obtains of receiving terminal demodulation,

Be the value S on 2k subcarrier of transmitting terminal frequency domain symbol _2kWith N move to right value behind the 2g of loop cycle,

Be the estimated value of thick frequency deviation,

The g value of correspondence when { } maximum is got in expression;

By formula (11), by thick frequency offset estimating, frequency offset estimation range can expand whole signal bandwidth to by ± 1 subcarrier spacing, and promptly (N/2, N/2).

In the step 7), the Nonlinear Transformation in Frequency Offset Estimation of ofdm system is

$\widehat{\mathrm{\&epsiv;}}={\widehat{\mathrm{\&epsiv;}}}_{\mathrm{coarse}}+{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}---\left(12\right)$

Wherein

Be the estimated value of the thick frequency deviation that obtains by (11) formula,

Be the estimated value of the thin frequency deviation that obtains by (6) formula,

It is the estimated value of actual normalization frequency deviation;

At the receiving terminal frequency domain to thick frequency deviation

Compensate, just directly to receiving terminal demodulation symbol (R _k, k=0,1 ..., N-1) carry out corresponding cyclic shift, promptly

$R_k^{\&prime;}=R_{{(k+{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{coarse}})}_N},k=\mathrm{0,1},...,N-1---\left(13\right)$

Wherein N is the subcarrier number,

Be the estimated value of the thick frequency deviation that obtains by (11) formula, R ' _k(k=0,1 ..., N-1) be value on k the subcarrier of the receiving terminal demodulation frequency domain symbol behind the thick compensate of frequency deviation,

Be the value R on k subcarrier of receiving terminal demodulation frequency domain symbol _kWith N is that loop cycle moves to left

After value.

According to present embodiment the present invention is carried out performance simulation, Fig. 4 and Fig. 5 are the present invention program and Schmidl, M﹠amp; The contrast of M, Song frequency offset estimation range under the same conditions.As can be seen from the results, the estimation range of Schmidl is ± 1 subcarrier spacing, M﹠amp; M is identical with the estimation range of Song method, all be ± 4 subcarrier spacings, and frequency offset estimating of the present invention and desirable curve almost matches, and promptly the actual frequency deviations estimation range is ± 128 subcarrier spacings, can reach whole signal bandwidth.Exceed the estimation range of each frequency offset estimator when actual frequency deviation after, adopt Schmidl, M﹠amp; When M and Song method, it is very big that mean square error will become, almost unacceptable, and the frequency offset estimator of the inventive method does not have this restriction, because its estimation range can reach whole signal bandwidth.For example, when the normalization frequency deviation is 6, Schmidl, M﹠amp; M, the mean square error of Song and the inventive method is approximately 38,65,65,4 * 10 respectively ^-6

Fig. 6 is under same channel, during different Doppler frequency shift, and the present invention program and Schmidl, M﹠amp; M, the contrast that the frequency offset estimating mean square error of Song changes with signal to noise ratio.Normalization frequency deviation ε=0.3.As seen from the figure; when Doppler frequency shift is 0; channel is constant channel when being; the mean square error of frequency offset estimating increases almost linear decline with signal to noise ratio; and when Doppler frequency shift is 172Hz, influenced by time varying channel, the frequency offset estimating mean square error descends slack-off; even when signal to noise ratio during greater than 25dB, it is very slow to descend.But can find out that under two kinds of Doppler frequency shifts the present invention all can make frequency offset estimation accuracy improve a lot, especially when signal to noise ratio is lower than 25dB.

Fig. 7 is under identical Doppler frequency shift, during different channels, and the present invention program and Schmidl, M﹠amp; M, the frequency offset estimating mean square error of Song is with the contrast of signal to noise ratio situation of change.Normalization frequency deviation ε=0.3.As seen from the figure, under channel B condition, owing to be subjected to the influence of intersymbol interference, the frequency offset estimating mean square error of four kinds of methods all is higher than their situations under channel A, especially when signal to noise ratio is higher.This is because under high s/n ratio, the performance of frequency offset estimating mean square error depends primarily on intersymbol interference.It can also be seen that from Fig. 7 under channel B condition, when signal to noise ratio during greater than 20dB, frequency offset estimating mean square error of the present invention descends very slow, performance is not as Schmidl, M﹠amp; M and Song, but under the low signal-to-noise ratio of channel B, and under any signal to noise ratio of channel A, frequency offset estimating mean square error of the present invention all is lower than Schmidl, M﹠amp; M and Song.Therefore, the present invention's (it happens frequently in the reality) under the low signal-to-noise ratio condition has more advantage.

It is under the A condition that Fig. 8 a, 8b, 8c have provided at channel, and when Doppler frequency shift was 0, signal to noise ratio was respectively 6dB, 9dB, and during 12dB, the tracking performance of frequency offset estimating of the present invention.Initial normalization frequency deviation ε=0.3.As seen from the figure, compare with other three kinds of algorithms, the present invention can make under different signal to noise ratios the frequency deviation shake littler, and when signal to noise ratio is 12dB, inherent spurious frequency deviation (actual frequency deviation and the difference of estimating frequency deviation) is controlled within 2%.

Fig. 4-Fig. 8 a, 8b, 8c are the simulation result of MATLAB software to ofdm system.Simulation parameter is: the subcarrier number N=256 of ofdm system, signal bandwidth 5.76MHz, circulating prefix-length L=64, the sample time is spaced apart 173.6ns[document [11] IEEE Standard for Local and metropolitan area networks Part 16:Air Interface for FixedBroadband Wireless Access Systems[S] .2004.].Multipath channel is got M.1225 Vehicle Channel A and B channel [document [12] Recommendation ITU-R M.1225, Guideline for evaluation of radio transmissiontechnologies for IMT-2000[S] .1997. of ITU-R].

Claims (5)

1. OFDM carrier frequency bias estimation that is applicable to transmitted in packets is characterized in that it carries out following steps in ofdm system:

1) utilizes pseudo random sequence to have good autocorrelation performance, the time domain training symbol that comprises two identical replicated blocks (s is provided _n, n=0,1 ..., N-1), time domain training symbol (s _n, n=0,1 ..., N-1) by frequency domain symbol (S _l, l=0,1 ..., N-1) conversion obtains through IFFT; In the formula, N is the subcarrier number, s _n(n=0,1 ..., N-1) be n sample value of time domain training symbol, S _l(l=0,1 ..., N-1) be value on l subcarrier of frequency domain symbol;

Corresponding frequency domain symbol is provided by following formula

$S_l=\left\{\begin{array}{c}\sqrt{2}{P}_m,l=2m,m=\mathrm{0,1,2},...,N/2-1\\ 0,l=\mathrm{1,3},...,N-1\end{array}\right.---\left(1\right)$

Wherein N is the subcarrier number, P _m(m=0,1.2 ..., N/2-1) be m value in the pseudo random sequence, and P _m∈ 1,1}, S _l(l=0,1 ..., N-1) be value on l subcarrier of frequency domain symbol;

Next modulation produces the time domain training symbol through IFFT, and it comprises the module of two identical repetitions

$s_n=\frac{1}{N}\underset{l=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{S}_l\exp \left(\frac{j2\mathrm{\&pi;nl}}{N}\right),n=\mathrm{0,1},...,N-1---\left(2\right)$

Wherein N is the subcarrier number, s _n(n=0,1 ..., N-1) be n sample value of time domain training symbol, S _l(l=0,1 ..., N-1) be value on l subcarrier of frequency domain symbol, the j representative

2) at time domain training symbol (s _n, n=0,1 ..., N-1) preceding adding length is L Cyclic Prefix constitutes complete OFDM symbol, i.e. { s _-L, s _-L+1... s _-1, s ₀, s ₁..., s _N-1; Through multipath channel, because there is frequency departure ε in the difference of transmit end receive end crystal oscillation frequency, ε is the normalization frequency deviation, and the ratio of promptly actual frequency deviation and subcarrier spacing obtains carrying the reception OFDM symbol (r of frequency deviation information _n, n=-L ,-L+1 ..., 0,1 ..., N-1), in the formula, r _n(n=-L ,-L+1 ..., 0,1 ..., N-1) be n the sample value that receives the OFDM symbol;

3) according to receiving OFDM symbol (r _n, n=-L ,-L+1 ..., the sample value of the N/2 of being separated by in N-1) is to { r _n, r _N+N/2(n=-L ,-L+1 ..., N/2-1) exist the phase place of π ε to rotate, obtain thin frequency offset estimating

R in the formula _n, r _N+N/2(n=-L ,-L+1 ..., N/2-1) be respectively n and n+N/2 the sample value that receives the OFDM symbol; Because the existence of carrier wave frequency deviation makes in OFDM symbol period of receiving terminal, be separated by the sample value of N/2 to there being the phase place rotation of π ε, identical replicated blocks is relevant in the comprehensive utilization training symbol, and right relevant of the corresponding sample value of first in Cyclic Prefix and the training symbol, can obtain thin frequency offset estimating, promptly

${\mathrm{\&gamma;}}_1=\underset{n=0}{\overset{N/2-1}{\mathrm{\&Sigma;}}}{r}_n^{*}{r}_{n+N/2}---\left(4\right)$

${\mathrm{\&gamma;}}_2=\underset{n=-L}{\overset{-1}{\mathrm{\&Sigma;}}}{r}_n^{*}{r}_{n+N/2}---\left(5\right)$

${\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}=\frac{1}{\mathrm{\&pi;}}\frac{N/2}{N/2+L}\mathrm{angle}\left({\mathrm{\&gamma;}}_1\right)+\frac{1}{\mathrm{\&pi;}}\frac{L}{N/2+L}\mathrm{angle}\left({\mathrm{\&gamma;}}_2\right)---\left(6\right)$

Subscript symbol wherein ^*Complex conjugate is got in expression, and the argument of () is got in angle () expression, and N is the subcarrier number, and L is a circulating prefix-length,

Be the complex conjugate that receives n sample value of OFDM symbol, r _N+N/2Be n+N/2 the sample value that receives the OFDM symbol, γ ₁Be the correlation that receives the identical replicated blocks of training symbol in the OFDM symbol, γ ₂Be Cyclic Prefix and the right correlation of the corresponding sample value of training symbol first,

Estimated value for thin frequency deviation;

Because the periodicity of phase function is so actual frequency deviation has following properties

$\mathrm{\&epsiv;}={\mathrm{\&epsiv;}}_{\mathrm{coarse}}+{\mathrm{\&epsiv;}}_{\mathrm{fine}}=2g+{\mathrm{\&epsiv;}}_{\mathrm{fine}}\&ap;2g+{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}---\left(7\right)$

Wherein g is an integer, ε _FineAnd ε _CoarseBe respectively thin frequency deviation and thick frequency deviation, ε is the ratio of actual frequency deviation and subcarrier spacing for actual normalization frequency deviation,

Be thin frequency offset estimating value;

4) obtain thin frequency offset estimating

Afterwards, to receiving OFDM symbol (r _n, n=-L ,-L+1 ..., N-1) carry out after thin compensate of frequency deviation is compensated the OFDM symbol (r ' _n, n=-L ,-L+1 ..., N-1);

5) to the compensation after the OFDM symbol (r ' _n, n=-L ,-L+1 ..., remove in N-1) training symbol part behind the Cyclic Prefix (r ' _n, n=0,1 ..., N-1) carry out the FFT conversion, obtain the frequency domain symbol (R of receiving terminal demodulation _k, k=0,1 ..., N-1);

6) utilize the good autocorrelation performance of pseudo random sequence, by receiving terminal frequency domain symbol (R _k, k=0,1 ..., N-1) with transmitting terminal frequency domain symbol (S _l, l=0,1 ..., the N-1) correlation peak detection of cyclic shift can obtain thick frequency offset estimating Because what send on the even subcarriers of transmitting terminal frequency domain is pseudo random sequence, the even-multiple frequency deviation only can cause the cyclic shift of receiving symbol, so utilize the good autocorrelation performance of pseudo random sequence, thick frequency offset estimating can detect by following peak value and obtain

${\widehat{\mathrm{\&epsiv;}}}_{\mathrm{coarse}}=2\left\{\underset{g}{\arg \max }\right[\frac{{\left|\underset{k=0}{\overset{N/2-1}{\mathrm{\&Sigma;}}}{R}_{2k}^{*}{S}_{{(2k-2g)}_N}\right|}^2}{{\left(\underset{k=0}{\overset{N/2-1}{\mathrm{\&Sigma;}}}{\left|R_{2k}\right|}^2\right)}^2}\left]\right\},g=-N/4,-N/4+1,...,N/4---\left(11\right)$

Wherein N is the subcarrier number, and g is the number of ambiguity period, the subscript symbol ^*Complex conjugate is got in expression,

Be the complex conjugate of 2k value of the frequency domain symbol that obtains of receiving terminal demodulation,

Be the value S on 2k subcarrier of transmitting terminal frequency domain symbol _2kWith N move to right value behind the 2g of loop cycle, Be the estimated value of thick frequency deviation,

The g value of correspondence when { } maximum is got in expression;

By formula (11), by thick frequency offset estimating, frequency offset estimation range can expand whole signal bandwidth to by ± 1 subcarrier spacing, and promptly (N/2, N/2);

7) according to thick frequency offset estimating

Directly at the receiving terminal frequency domain to receiving terminal frequency domain symbol (R _k, k=0,1 ..., N-1) make corresponding thick compensate of frequency deviation, Nonlinear Transformation in Frequency Offset Estimation is thick frequency offset estimating With thin frequency offset estimating

Sum.

2. the OFDM carrier frequency bias estimation that is applicable to transmitted in packets as claimed in claim 1, it is characterized in that, described step 2) in, for eliminating intersymbol interference, before each training symbol, insert the Cyclic Prefix of length L, L is greater than the corresponding sample value number of multipath channel maximum delay expansion, that is, the OFDM symbol of actual transmissions is: { s _-L, s _-L+1... s _-1, s ₀, s ₁..., s _N-1, a preceding L sample value is duplicating of L the sample value in back;

At receiving terminal, if regularly accurately, the reception OFDM symbol when to have normalization frequency deviation ε be the ratio of actual frequency deviation and subcarrier spacing is so

$r_n=s_n\exp \left(\frac{j2\mathrm{\&pi;n\&epsiv;}}{N}\right)+w_n,n=-L,-L+1,...,\mathrm{0,1},...,N-1---\left(3\right)$

Wherein N is the subcarrier number, and L is a circulating prefix-length, the j representative

w _n(n=-L ,-L+1 ..., 0,1 ..., be that average is 0 N-1), variance is N sample value of additive white Gaussian noise, s _n(n=-L ,-L+1 ..., 0,1 ..., N-1) be n sample value of time domain OFDM symbol, ε is that the normalization frequency deviation is the ratio of actual frequency deviation and subcarrier spacing, r _n(n=-L ,-L+1 ..., 0,1 ..., N-1) be n the sample value that receives the OFDM symbol.

3. the OFDM carrier frequency bias estimation that is applicable to transmitted in packets as claimed in claim 1 is characterized in that, in the described step 4), after thin frequency offset estimating, needs to carry out thin compensate of frequency deviation to receiving the OFDM symbol earlier, promptly

$r_n^{\&prime;}=r_n\exp \left(\frac{-j2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)$

$=s_n\exp \left(\frac{j2\mathrm{\&pi;n}(\mathrm{\&epsiv;}-{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}})}{N}\right)+w_n\exp \left(\frac{-j2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)$

$=s_n\exp \left(\frac{-j2\mathrm{\&pi;n}\left(2g\right)}{N}\right)+w_n\exp \left(\frac{-j2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)---\left(8\right)$

Wherein N is the subcarrier number, r _nBe n the sample value that receives the OFDM symbol, r ' _nBe n sample value of the reception OFDM symbol behind meticulous compensate of frequency deviation, Be thin frequency offset estimating value, s _nBe n sample value of time domain OFDM symbol, w _nBe n sample value of additive white Gaussian noise, ε is actual normalization frequency deviation, and g is the number of ambiguity period, the j representative

4. the OFDM carrier frequency bias estimation that is applicable to transmitted in packets as claimed in claim 1 is characterized in that, in the described step 5), to the compensation after the OFDM symbol (r ' _n, n=-L ,-L+1 ..., remove in N-1) training symbol part behind the Cyclic Prefix (r ' _n, n=0,1 ..., N-1) do the FFT conversion; (r ' _n, n=0,1 ..., N-1) transform to frequency domain after because the existence of even number frequency deviation, cyclic shift will appear in the frequency domain symbol that recovers to obtain, promptly

$R_k=\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{r}_n^{\&prime;}\exp \left(\frac{-j2\mathrm{\&pi;kn}}{N}\right)$

$=S_{{(k-2g)}_N}+W_k^{\&prime;},k=\mathrm{0,1},...,N-1---\left(9\right)$

Wherein N is the subcarrier number, and g is the number of ambiguity period, the j representative

R ' _n(n=0,1 ..., N-1) be n sample value of the reception training symbol behind the thin compensate of frequency deviation, R _k(k=0,1 ..., N-1) be value on k the subcarrier of the frequency domain symbol that obtains of receiving terminal demodulation,

Be the value S on k subcarrier of transmitting terminal frequency domain symbol _kWith N move to right result behind the 2g of loop cycle, W ' _kBe the additive white Gaussian noise of k subchannel, promptly

$W_k^{\&prime;}=\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}\left(w_n\exp \left(\frac{-j2\mathrm{\&pi;n}{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}}{N}\right)\right)\exp \left(\frac{-j2\mathrm{\&pi;kn}}{N}\right)---\left(10\right)$

5. the OFDM carrier frequency bias estimation that is applicable to transmitted in packets as claimed in claim 1 is characterized in that in the described step 7), the Nonlinear Transformation in Frequency Offset Estimation of ofdm system is

$\widehat{\mathrm{\&epsiv;}}={\widehat{\mathrm{\&epsiv;}}}_{\mathrm{coarse}}+{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{fine}}---\left(12\right)$

Wherein

Be the estimated value of the thick frequency deviation that obtains by (11) formula,

Be the estimated value of the thin frequency deviation that obtains by (6) formula,

It is the estimated value of actual normalization frequency deviation;

At the receiving terminal frequency domain to thick frequency deviation

Compensate, just directly to receiving terminal demodulation symbol (R _k, k=0,1 ..., N-1) carry out corresponding cyclic shift, promptly

$R_k^{\&prime;}=R_{{(k+{\widehat{\mathrm{\&epsiv;}}}_{\mathrm{coarse}})}_N},k=\mathrm{0,1},...,N-1---\left(13\right)$

Wherein N is the subcarrier number,

Be the estimated value of the thick frequency deviation that obtains by (11) formula, R ' _k(k=0,1 ..., N-1) be value on k the subcarrier of the receiving terminal demodulation frequency domain symbol behind the thick compensate of frequency deviation,

Be the value R on k subcarrier of receiving terminal demodulation frequency domain symbol _kWith N is that loop cycle moves to left After value.

Images