CN105897640A - Frequency offset estimation method and device - Google Patents

Abstract

The present invention discloses a frequency offset estimation method. The method comprises a step of carrying out sliding window storage of maximum likelihood frequency offset estimations and obtaining a maximum likelihood frequency offset estimation set, and a step of using sliding filtering on the maximum likelihood frequency offset estimation set to combine, and obtaining a final maximum likelihood frequency offset estimation. The invention also discloses a frequency offset estimation device.

Description

A kind of frequency deviation estimating method and device

Technical field

The present invention relates to wireless communication field, particularly relate to a kind of frequency deviation estimating method and device.

Background technology

For any digital communication system, synchronization is premise and the important guarantee of reliable data transmission, synchronicity The quality of energy will directly affect the performance of whole communication system.But in mobile communication system, due to transmitting terminal Frequency difference between equipment and receiving device, and the Doppler frequency-shift etc. that sending and receiving end equipment relative movement is brought Impact so that there is frequency deviation between the frequency of carrier frequency and local crystal oscillator.In order to ensure the reliable of data Transmission, it is necessary to signal frequency deviation is accurately estimated and is compensated for.

The Long Term Evolution of third generation cooperative programme (3rd Generation Partnership Project, 3GPP) (Long Term Evolution, LTE) is descending have employed the OFDM that the availability of frequency spectrum is higher (Orthogonal Frequency Division Multiplexing, OFDM) modulation technique, OFDM symbol It is to be formed by multiple sub-carrier signal superpositions, utilizes the orthogonality between subcarrier to be demodulated at receiving terminal, Thus the orthogonality between ofdm system sub-carrier proposes strict requirements.In actual transmissions, due to Do not mate, between Doppler frequency shift and transceiver local oscillator, the frequency departure brought, OFDM can be destroyed Orthogonality between system subcarrier, causes inter-sub-carrier interference (ICI)；Timing slip then can cause intersymbol Interference (ISI), reduces the effectiveness of Cyclic Prefix (CP), therefore, synchronizes for ofdm system very Important.

Frequency synchronization method in ofdm system substantially can be divided into two classes: blind synchronized algorithm and based on training The synchronized algorithm of sequence；Wherein, blind synchronized algorithm mainly utilizes Cyclic Prefix specific to ofdm system Matter completes to synchronize to estimate, synchronized algorithm based on training sequence utilizes the known pilot symbols inserted to realize Synchronize.In view of including that the practical communication system such as TD-LTE standard are widely used all kinds of training sequence, because of This, the frequency synchronization algorithm around pilot aided the most more has the using value of reality.? Under the precondition of receiving terminal known training sequence, maximum likelihood frequency deviation is estimated to tend to obtain more preferable frequency deviation Estimate performance, and become the emphasis that all kinds of institute is paid close attention to.Around the maximum likelihood frequency under the conditions of OFDM Bias estimation is a lot, the essentially identical likelihood function of these method choice, different frequency excursion algorithm Difference is mainly reflected in how to be obtained concrete frequency excursion algorithm by the likelihood function derivation given.

Comprehensive analysis is the most both at home and abroad around the frequency Synchronization achievement of ofdm system, at given frequency Partially and on the basis of channel joint likelihood function, the most existing a large amount of practicable frequency deviations and channel estimate calculation Method is available for using for reference, but prior art focuses simply on design and the selection of frequency excursion algorithm, and does not pay close attention to Associate feature between the estimation of adjacent frequency deviation, frequency deviation is estimated by the associate feature between estimating due to adjacent frequency deviation The impact of performance is relatively big, and therefore, the frequency deviation of prior art estimates that performance is relatively low.As can be seen here, how to utilize Associate feature between adjacent frequency deviation is estimated improves the Developing Tendency that frequency deviation estimation performance is Frequency Synchronization research Gesture.

Summary of the invention

In view of this, embodiment of the present invention expectation provides a kind of frequency deviation estimating method and device, it is possible to increase frequently The precision partially estimated and performance.

For reaching above-mentioned purpose, the technical scheme is that and be achieved in that:

Embodiments providing a kind of frequency deviation estimating method, the method includes:

Estimate to carry out sliding window storage to maximum likelihood frequency deviation, it is thus achieved that maximum likelihood frequency deviation estimates collection；

Use smothing filtering that maximum likelihood frequency deviation being estimated, collection merges, obtain final maximum likelihood frequency deviation Estimate.

In such scheme, described to maximum likelihood frequency deviation estimate carry out sliding window storage, it is thus achieved that maximum likelihood frequency deviation Estimate collection, including:

Maximum likelihood frequency deviation is estimated ε⁽ⁿ⁾It is stored in the FIFO FIFO sliding window memorizer of a length of K, it is thus achieved that Maximum likelihood frequency deviation estimates collection { ε^(n-k), k=0,1 ..., K-1}；

Wherein, n, K are positive integer.

In such scheme, maximum likelihood frequency deviation is estimated that collection merges by described employing smothing filtering, obtains Whole maximum likelihood frequency deviation is estimated, including:

According toAsk for maximum likelihood frequency deviation to estimate to concentrate all maximum likelihood frequency deviations to estimate Meansigma methods, and described meansigma methods is estimated as final maximum likelihood frequency deviation；Wherein, n, K are positive integer.

In such scheme, described method also includes:

The likelihood function estimated according to frequency deviation obtains maximum likelihood frequency deviation and estimates.

In such scheme, the described likelihood function estimated according to frequency deviation obtains maximum likelihood frequency deviation and estimates, including:

To pilot symbol transmitted matrix D⁽ⁿ⁾, the front L row of discrete Fourier transform (DFT) matrix F and described matrix F The matrix constitutedCarry out pretreatment, it is thus achieved that matrix P⁽ⁿ⁾；

Frequency pilot sign matrix R will be received⁽ⁿ⁾And described matrix P⁽ⁿ⁾It is respectively stored into the FIFO sliding window of a length of M In memorizer；

According to the reception frequency pilot sign matrix R obtained from described sliding window memorizer⁽ⁿ⁾, matrix P⁽ⁿ⁾And build The likelihood function that M frequency deviation matrix calculus frequency deviation is estimated, and the likelihood function or cumulative estimated according to frequency deviation The likelihood function that frequency deviation is estimated obtains maximum likelihood frequency deviation and estimates；Wherein, M, n are positive integer.

According to said method, the embodiment of the present invention additionally provides a kind of frequency deviation estimation device, and this device includes: Sliding window memory element, smothing filtering unit；Wherein,

Described sliding window memory element, for estimating to carry out sliding window storage to maximum likelihood frequency deviation, it is thus achieved that maximum is seemingly So frequency deviation estimates collection；

Described smothing filtering unit, is used for using smothing filtering that maximum likelihood frequency deviation being estimated, collection merges, Obtain final maximum likelihood frequency deviation to estimate.

In such scheme, described sliding window memory element, specifically for estimating ε by maximum likelihood frequency deviation⁽ⁿ⁾It is stored in In the FIFO FIFO sliding window memorizer of a length of K, it is thus achieved that maximum likelihood frequency deviation estimates collection {ε^(n-k), k=0,1 ..., K-1}；

Wherein, n, K are positive integer.

In such scheme, described smothing filtering unit, specifically for basisAsk for maximum Likelihood frequency deviation estimates the meansigma methods concentrating all maximum likelihood frequency deviations to estimate, and using described meansigma methods as finally Maximum likelihood frequency deviation estimate；Wherein, n, K are positive integer.

In such scheme, described device also includes:

Acquiring unit, the likelihood function for estimating according to frequency deviation obtains maximum likelihood frequency deviation and estimates.

In such scheme, described acquiring unit, specifically for pilot symbol transmitted matrix D⁽ⁿ⁾, discrete Fourier The matrix that the front L row of leaf transformation matrix F and described matrix F are constitutedCarry out pretreatment, it is thus achieved that matrix P⁽ⁿ⁾；

Frequency pilot sign matrix R will be received⁽ⁿ⁾And described matrix P⁽ⁿ⁾It is respectively stored into the FIFO sliding window of a length of M In memorizer；

According to the reception frequency pilot sign matrix R obtained from described sliding window memorizer⁽ⁿ⁾, matrix P⁽ⁿ⁾And build The likelihood function that M frequency deviation matrix calculus frequency deviation is estimated, and the likelihood function or cumulative estimated according to frequency deviation The likelihood function that frequency deviation is estimated obtains maximum likelihood frequency deviation and estimates；Wherein, M, n are positive integer.

The frequency deviation estimating method of embodiment of the present invention offer and device, estimate to carry out sliding window to maximum likelihood frequency deviation Storage, it is thus achieved that maximum likelihood frequency deviation estimates collection；Use smothing filtering that maximum likelihood frequency deviation being estimated, collection closes And, obtain final maximum likelihood frequency deviation and estimate.So, the embodiment of the present invention utilize joint pilot characteristic and Associate feature between the estimation of adjacent maximum likelihood frequency deviation, is primarily based on maximum likelihood frequency deviation and estimates to obtain maximum Likelihood frequency deviation estimates collection；Then use smothing filtering to ask for maximum likelihood frequency deviation to estimate to concentrate all maximum likelihoods The meansigma methods that frequency deviation is estimated, estimates as final maximum likelihood frequency deviation, it is possible to increase the precision that frequency deviation is estimated And performance.

Accompanying drawing explanation

Fig. 1 be embodiment of the present invention frequency deviation estimating method realize schematic flow sheet；

Fig. 2 is the schematic diagram of embodiment of the present invention sliding window storage；

Fig. 3 is the schematic diagram that the embodiment of the present invention uses that Indirect-approach Method search maximum likelihood frequency deviation is estimated；

Frequency deviation when Fig. 4 is that under embodiment of the present invention TD-LTE standard, pilot resources takes 2 estimates mean square deviation Can schematic diagram；

Frequency deviation when Fig. 5 is that under embodiment of the present invention TD-LTE standard, pilot resources takes 20 estimates mean square deviation Can schematic diagram；

Fig. 6 is the signal that embodiment of the present invention pilot resources takes frequency deviation estimated probability distribution function corresponding when 2 Figure；

Fig. 7 is the signal that embodiment of the present invention pilot resources takes frequency deviation estimated probability distribution function corresponding when 20 Figure；

Fig. 8 is the composition structural representation of embodiment of the present invention frequency deviation estimation device.

Detailed description of the invention

In the embodiment of the present invention, estimate to carry out sliding window storage to maximum likelihood frequency deviation, it is thus achieved that maximum likelihood frequency deviation Estimate collection；Use smothing filtering that maximum likelihood frequency deviation being estimated, collection merges, obtain final maximum likelihood Frequency deviation is estimated.

Below in conjunction with the accompanying drawings the detailed description of the invention of the present invention is illustrated.

The embodiment of the present invention proposes a kind of frequency deviation estimating method, includes as it is shown in figure 1, implement step:

Step S100: the likelihood function estimated according to frequency deviation obtains maximum likelihood frequency deviation and estimates.

The embodiment of the present invention be applied to time-division Long Term Evolution (Time Division-Long Term Evolution, TD-LTE) substandard descending multiple-input and multiple-output (Multiple-Input Multiple-Output, MIMO) -OFDM (Orthogonal Frequency Division Multiplexing, OFDM) system and/ Or Uplink MIMO single-carrier frequency division multiple access (Single Carrier-Frequency-Division Multiple Access, SC-FDMA) reception in system synchronizes, under conditions of assuming receiving terminal sign synchronization, right For Uplink MIMO SC-FDMA system with descending MIMO-OFDM system, although LTE's is upper Descending pilot frequency number of symbols, the regularity of distribution in each subframe, the antenna number of up-downgoing transmitting terminal phase not to the utmost With, but the time-domain received signal characteristic after reception antenna removal Cyclic Prefix is identical.

The embodiment of the present invention is as a example by MIMO-OFDM system, to the likelihood letter how to estimate according to frequency deviation Number obtains maximum likelihood frequency deviation and estimates, is described in detail:

Preset N_T、N_RBeing respectively transmitting antenna number and the reception antenna number of MIMO-OFDM system, N is The sub-carrier number of OFDM.

Step one: to pilot symbol transmitted matrix D⁽ⁿ⁾, discrete Fourier transform (DFT) matrix F and described matrix The matrix that the front L row of F are constitutedCarry out pretreatment, it is thus achieved that matrix P⁽ⁿ⁾。

Concrete, according to equation below (1) to pilot symbol transmitted matrix D⁽ⁿ⁾, discrete Fourier transform (DFT) DFT The matrix that the front L row of matrix F and described matrix F are constituted carries out pretreatmentObtain matrix P⁽ⁿ⁾:

Wherein, $D^{\left(n\right)}=\frac{1}{\sqrt{N_T}}{[D_1^{\left(n\right)},...,D_{N_T}^{\left(n\right)}]}_{N\×{\mathrm{NN}}_T},D_p^{\left(n\right)}=\mathrm{diag}\left(d_p^{\left(n\right)}\right)=\mathrm{diag}\left({[d_{p,0}^{\left(n\right)},...,d_{p,N-1}^{\left(n\right)}]}^T\right)$ For pth root Launch the n-th pilot tone sign matrix that antenna sends,The n-th of antenna transmission is launched for pth root Pilot tone symbol；Matrix F ∈ C^N×NFor DFT matrix, (l, m) individual element is the of matrix F $F_{l,m}=\frac{1}{\sqrt{N}}{e}^{-j(2\mathrm{\π}(l-1)(m-1)/N)};$ Matrix $\tilde{W}={(W\⊗I_{N_T})}_{{\mathrm{NN}}_T\×{\mathrm{LN}}_T},$ W represents the front L row of DFT matrix F, i.e. F=[W | V], W ∈ C^N×L, V ∈ C^N×(N-L), and W^HV=0, WW^H+VV^H=I, H represent that conjugation turns Put.

Step 2: frequency pilot sign matrix R will be received⁽ⁿ⁾And described matrix P⁽ⁿ⁾It is respectively stored into the elder generation of a length of M Enter first to go out in (First In First Out, FIFO) sliding window memorizer；

Concrete, frequency pilot sign matrix R will be received⁽ⁿ⁾And matrix P⁽ⁿ⁾It is stored in a length of M's of correspondence respectively In FIFO sliding window memorizer, as shown in Figure 2, it is thus achieved that following two groups of data:

{P^(n-k), k=0,1 ..., M-1} (2)

{R(^n-k), k=0,1 ..., M-1}

Wherein, $R^{\left(n\right)}={[r_1^{\left(n\right)},...,r_{N_R}^{\left(n\right)}]}_{N\×N_R};$

Under conditions of system receiving terminal sign synchronization, the n-th reception that q root reception antenna receives is led Frequency symbol is:

$r_q^{\left(n\right)}=\underset{p=1}{\overset{N_T}{\mathrm{\Σ}}}{E}^{\left(n\right)}{F}^H{D}_p^{\left(n\right)}{\mathrm{Wg}}_{q,p}^{\left(n\right)}+n_q---\left(3\right)$

Wherein, E⁽ⁿ⁾It is the frequency deviation matrix that the n-th frequency pilot sign is corresponding, E=diag ([1, e^j2πε/N,…,e^{j2π(N-1)ε/N}]^T), ε is normalization frequency deviation value, L_nIt it is the n-th frequency pilot sign first Sequence number corresponding to individual sampling time；Represent the n-th frequency pilot sign (q, p) The L footpath time-domain channel gain experienced during transmission between antenna pair, T represents transposition, n_qIt is that q root receives sky N × 1 dimension zero-mean and every one-dimensional variance that line receives areMultiple Gaussian noise, j²=-1.

Step 3: according to the reception frequency pilot sign matrix R obtained from described sliding window memorizer⁽ⁿ⁾, matrix P⁽ⁿ⁾And The likelihood function that M the frequency deviation matrix calculus frequency deviation built is estimated, and according to the likelihood function of frequency deviation estimation or The likelihood function that cumulative frequency deviation is estimated obtains maximum likelihood frequency deviation and estimates；Wherein, M, n are positive integer.

Here, first make, $G^{\left(n\right)}={\left[\begin{array}{c}{g}_1^{\left(n\right)}\\ ..\\ ..\\ ..\\ g_{N_T}^{\left(n\right)}\end{array}\right]}_{{\mathrm{LN}}_T\×N_R}---\left(4\right)$

$g_q^{\left(n\right)}={[g_{1,q}^{\left(n\right)},...,g_{N_R,q}^{\left(n\right)}]}_{L\×N_R}---\left(5\right)$

$g_{1,q}^{\left(n\right)}={{\left[g_{1,q}^{\left(n\right)}\right[0],...,g_{1,q}^{\left(n\right)}[L-1\left]\right]}^T}_{L\×1}---\left(6\right)$

$\begin{array}{cc}N=[n_1,...,n_{N_R}]& \tilde{W}={(W\⊗I_{N_T})}_{{\mathrm{NN}}_T\×{\mathrm{LN}}_T}\end{array}----\left(7\right)$

$D^{\left(n\right)}=\frac{1}{\sqrt{N_T}}{[D_1^{\left(n\right)},...,D_{N_T}^{\left(n\right)}]}_{N\×{\mathrm{NN}}_T}---\left(8\right)$

$R^{\left(n\right)}={[r_1^{\left(n\right)},...,r_{N_R}^{\left(n\right)}]}_{N\×N_R}---\left(9\right)$

Wherein, G⁽ⁿ⁾Represent time-domain channel gain；

Here, formula (4) obtain to (9):

$R^{\left(n\right)}=E^{\left(n\right)}{F}^H{D}^{\left(n\right)}{\tilde{W}G}^{\left(n\right)}+N---\left(10\right)$

Wherein, M frequency deviation matrix of structure is:

$E^{(n-k)}=e^{j2\mathrm{\π\ϵ}{L}_{n-k}/N}\·\mathrm{diag}\left({[1,e^{j2\mathrm{\π\ϵ}/N},...,e^{j2\mathrm{\π}(N-1)\mathrm{\ϵ}/N}]}^T\right),k=\mathrm{0,1},...,M-1---\left(11\right)$

Wherein, ε is normalization frequency deviation value, L_n-kRepresent first sampling time institute of (n-k) individual frequency pilot sign right The sequence number answered, T represents transposition.

Concrete, how according to the reception frequency pilot sign matrix R obtained from described sliding window memorizer⁽ⁿ⁾, matrix P⁽ⁿ⁾And the likelihood function that M the frequency deviation matrix calculus frequency deviation built is estimated, and estimate according to cumulative frequency deviation Likelihood function obtain maximum likelihood frequency deviation and estimate, the concrete the following two kinds mode that uses:

Mode one, first theoretical according to maximal possibility estimation, may certify that ML channel estimator and frequency deviation Estimate to meet following relation:

Wherein,Represent pseudoinverse, according to the reception frequency pilot sign matrix R in formula (10)⁽ⁿ⁾, can obtain Signal is measured to revising:

${\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}={\left(G^{\left(n\right)}\right)}^H{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}+{\tilde{N}}^{\left(n\right)}---\left(13\right)$

Measure signal according to revising, N can be obtained_RRoot reception antenna receives the conditional probability density letter of signal Number:

$\mathrm{\Λ}\left({\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}\right|G^{\left(n\right)})=\frac{1}{{\left({\mathrm{\π\σ}}_n^2\right)}^{N\×N_R}}\exp \{-\frac{1}{{\mathrm{\σ}}_n^2}\mathrm{tr}\left[\begin{array}{c}{({\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}-{\left(G^{\left(n\right)}\right)}^H{\tilde{W}}^H{D}^{\left(n\right)H}{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)})}^H\·\\ ({\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}-{\left(G^{\left(n\right)}\right)}^H{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)})\end{array}\right]\}---\left(14\right)$

Theoretical according to maximal possibility estimation, can construct and obtain following channel estimation likelihood function:

$\begin{array}{c}L\left({\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}\right|G^{\left(n\right)})=-\mathrm{tr}\left[{({\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}-G^{\left(n\right)H}{\tilde{W}}^H{D}^{\left(n\right)H}{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)})}^H({\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}-G^{\left(n\right)H}{\tilde{W}}^H{D}^{\left(n\right)H}{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)})\right]\\ =-\mathrm{tr}\left[\begin{array}{c}{\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}{\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}-{\left(G^{\left(n\right)}\right)}^H{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}{\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}\\ -{\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}{\left(G^{\left(n\right)}\right)}^H{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}+{\left(G^{\left(n\right)}\right)}^H{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}{\left(G^{\left(n\right)}\right)}^H{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}\end{array}\right]\end{array}$

Because, $\frac{\∂\mathrm{tr}\left({\left(G^{\left(n\right)}\right)}^H{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}{\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}\right)}{\∂G^{\left(n\right)*}}={\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}{\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}---\left(15\right)$

$\frac{\∂\mathrm{tr}\left({\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}{\left(G^{\left(n\right)}\right)}^H{{\tilde{W}}^H\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}\right)}{\∂G^{\left(n\right)*}}={\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}{\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}---\left(16\right)$

$\frac{\∂\mathrm{tr}\left({\left(G^{\left(n\right)}\right)}^H{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}{\left(G^{\left(n\right)}\right)}^H{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}\right)}{{\∂G}^{\left(n\right)*}}=2{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}{\left(G^{\left(n\right)}\right)}^H{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}---\left(17\right)$

ByAnd formula (15) is to (17), ML channel estimator can be obtained full Be enough to lower restriction relation:

$\begin{array}{c}{\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{\hat{G}}^{\left(n\right)}\left(\mathrm{\ϵ}\right){\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}={\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{\hat{G}}^{\left(n\right)}\left(\mathrm{\ϵ}\right){\left({\hat{G}}^{\left(n\right)}\right)}^H\left(\mathrm{\ϵ}\right){\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{\hat{G}}^{\left(n\right)}\left(\mathrm{\ϵ}\right)\\ \⇒{\left(R^{\left(n\right)}\right)}^H{R}^{\left(n\right)}={\left({\hat{G}}^{\left(n\right)}\right)}^H\left(\mathrm{\ϵ}\right){\tilde{W}}^H{\left(D^{\left(n\right)}\right)}^H{D}^{\left(n\right)}\tilde{W}{\hat{G}}^{\left(n\right)}\left(\mathrm{\ϵ}\right)\end{array}---\left(18\right)$

In order to the restriction relation that met by the ML channel estimator in formula (18) being estimated obtain maximum Likelihood frequency deviation is estimated, can be by the ML channel estimator in formula (12)Bring formula (18) into Obtain following relation:

The likelihood function then deriving frequency deviation estimation is:

λ⁽ⁿ⁾(ε)=| | (R⁽ⁿ⁾)^HE⁽ⁿ⁾(P⁽ⁿ⁾-I)(E⁽ⁿ⁾)^HR⁽ⁿ⁾||_F (19)

Wherein, | | | |_FRepresenting the F norm taking matrix, H is conjugate transpose, E⁽ⁿ⁾It it is the n-th frequency pilot sign pair The frequency deviation matrix answered,E=diag ([1, e^j2πε/N,…,e^{j2π(N-1)ε/N}]^T)；

Estimate performance for improving frequency deviation further, the many groups frequency pilot sign in adjacent sub-frame can be combined, will frequency The likelihood function λ partially estimated⁽ⁿ⁾(ε) store in the FIFO sliding window memorizer of a length of M, obtain frequency deviation and estimate The likelihood function collection λ of meter^(n-k)(ε):

{λ^(n-k)(ε), k=0,1 ..., M-1} (20)

Then, the likelihood function estimated according to cumulative frequency deviationAnd use Indirect-approach Method to obtain Obtain maximum likelihood frequency deviation and estimate ε⁽ⁿ⁾:

${\mathrm{\ϵ}}^{\left(n\right)}=\underset{\mathrm{\ϵ}}{\arg \min }\left\{\underset{k=0}{\overset{M-1}{\mathrm{\Σ}}}{\mathrm{\λ}}^{(n-k)}\left(\mathrm{\ϵ}\right)\right\}---\left(21\right)$

Concrete, use Indirect-approach Method to obtain maximum likelihood frequency deviation and estimate ε⁽ⁿ⁾, as it is shown on figure 3, specifically Implementation is as follows:

Presetting initial frequency deviation estimation range is (-ε_max,ε_max)；

All frequency deviation hunting zone is divided into when search every time P interval, calculates successively and compare likelihood functionValue size at each interval endpoint；

Change from big to small and by little change with the selection value of different normalization frequency deviation estimated values according to likelihood function Big process, determines the frequency deviation region (ε that next round frequency deviation is searched for₁,ε₂), wherein, ε₁,ε₂Choose and likelihood Function value variation relation meets $\underset{n=1}{\overset{M-1}{\mathrm{\Σ}}}{\mathrm{\λ}}^{\left(n\right)}\left({\mathrm{\ϵ}}_1\right)>\underset{n=1}{\overset{M-1}{\mathrm{\Σ}}}{\mathrm{\λ}}^{\left(n\right)}\left({\mathrm{\ϵ}}_0\right),\underset{n=1}{\overset{M-1}{\mathrm{\Σ}}}{\mathrm{\λ}}^{\left(n\right)}\left({\mathrm{\ϵ}}_2\right)>\underset{n=1}{\overset{M-1}{\mathrm{\Σ}}}{\mathrm{\λ}}^{\left(n\right)}\left({\mathrm{\ϵ}}_0\right),$ Wherein, ε₀=(ε₁+ε₂)/2；

Determining new frequency deviation region (ε₁,ε₂On the basis of), repeat above-mentioned search step, until current frequency Step-size in search partially has met the frequency offset estimation accuracy requirement of system, exports current frequency offset hunting zone (ε₁,ε₂) Intermediate value ε₀The final valuation ε searched for for this⁽ⁿ⁾。

Mode two, first according to the signal model in formula (10), N can be obtained_RRoot reception antenna receives letter Number conditional probability density function:

$\mathrm{\Λ}\left(R^{\left(n\right)}\right|G^{\left(n\right)},\mathrm{\ϵ})=\frac{1}{{\left({\mathrm{\π\σ}}_n^2\right)}^{N\×N_R}}\exp \{-\frac{1}{{\mathrm{\σ}}_n^2}\mathrm{tr}[{(R^{\left(n\right)}-E^{\left(n\right)}{F}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)})}^H(R^{\left(n\right)}-E^{\left(n\right)}{F}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)})\left]\right\}---\left(22\right)$

Theoretical according to maximal possibility estimation, the channel that can be constructed as follows and the joint likelihood function of frequency deviation:

$\begin{array}{c}L\left(R^{\left(n\right)}\right|\mathrm{\ϵ},G^{\left(n\right)})=-\mathrm{tr}\left[{(R^{\left(n\right)}-E^{\left(n\right)}{F}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)})}^H(R^{\left(n\right)}-E^{\left(n\right)}{F}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)})\right]\\ =-\mathrm{tr}[R^{\left(n\right)H}{R}^{\left(n\right)}-R^{\left(n\right)H}{E}^{\left(n\right)}{F}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}-G^{\left(n\right)H}{\tilde{W}}^H{D}^{\left(n\right)H}{\mathrm{FE}}^{\left(n\right)H}{R}^{\left(n\right)}+G^{\left(n\right)H}{\tilde{W}}^H{D}^{\left(n\right)H}{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}\end{array}---\left(23\right)$

By $\frac{L\left(R^{\left(n\right)}\right|\mathrm{\ϵ},G^{\left(n\right)})}{{\∂G}^{\left(n\right)*}}=0,$ And $\frac{\∂\mathrm{tr}\left[R^{\left(n\right)H}{E}^{\left(n\right)}{F}^H{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}\right]}{{\∂G}^{\left(n\right)*}}=0,\frac{\∂\mathrm{tr}\left[G^{\left(n\right)H}{\tilde{W}}^H{D}^{\left(n\right)H}{\mathrm{FE}}^{\left(n\right)H}{R}^{\left(n\right)}\right]}{{\∂G}^{\left(n\right)*}}={\tilde{W}}^H{D}^{\left(n\right)H}{\mathrm{FE}}^{\left(n\right)H}{R}^{\left(n\right)},$ $\frac{\∂\mathrm{tr}\left[G^{\left(n\right)H}{\tilde{W}}^H{D}^{\left(n\right)H}{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)}\right]}{{\∂G}^{\left(n\right)*}}={\tilde{W}}^H{D}^{\left(n\right)H}{D}^{\left(n\right)}\tilde{W}{G}^{\left(n\right)},$ Can obtain ML channel estimator as follows:

Wherein,Represent generalized inverse；

WillThe likelihood function that the available frequency deviation of substitution formula (23) is estimated:

Wherein,

Formula (24) is simplified further and obtains:

L (ε)=-tr (R^(n)HR⁽ⁿ⁾-R^(n)HE⁽ⁿ⁾B^(n)HB⁽ⁿ⁾E^(n)HR⁽ⁿ⁾) (25)

The likelihood function L (ε) then estimated according to frequency deviation obtains maximum likelihood frequency deviation and estimates ε⁽ⁿ⁾:

${\mathrm{\ϵ}}^{\left(n\right)}=\arg \underset{\mathrm{\ϵ}}{\max }\left[\mathrm{tr}\left(R^{\left(n\right)H}{E}^{\left(n\right)}{B}^{\left(n\right)H}{B}^{\left(n\right)}{E}^{\left(n\right)H}{R}^{\left(n\right)}\right)\right]---\left(26\right)$

Step S101: estimate to carry out sliding window storage to maximum likelihood frequency deviation, it is thus achieved that maximum likelihood frequency deviation estimates collection.

Here, between estimating due to adjacent maximum likelihood frequency deviation, there is Statistical Distribution Characteristics, therefore, it can Associate feature between utilizing joint pilot feature and adjacent maximum likelihood frequency deviation to estimate, improves frequency deviation and estimates The precision of meter and frequency deviation estimate performance.

Concrete, maximum likelihood frequency deviation is estimated ε⁽ⁿ⁾It is stored in the FIFO sliding window memorizer of a length of K, it is thus achieved that Maximum likelihood frequency deviation estimates collection { ε^(n-k), k=0,1 ..., K-1}；Wherein, n, K are positive integer.

Here, maximum likelihood frequency deviation estimates that collection includes that maximum likelihood frequency deviation estimates ε⁽ⁿ⁾And estimate with maximum likelihood frequency deviation Meter ε⁽ⁿ⁾Adjacent maximum likelihood frequency deviation is estimated.

Step S102: use smothing filtering that maximum likelihood frequency deviation being estimated, collection merges, obtain final Maximum-likelihood frequency deviation is estimated.

This step is particularly as follows: use smothing filtering according to formulaAsk for maximum likelihood frequency deviation Estimate the meansigma methods concentrating all maximum likelihood frequency deviations to estimate, and using the meansigma methods that obtains as final maximum Likelihood frequency deviation is estimated

Wherein, n, K are positive integer.

Association between the embodiment of the present invention utilizes joint pilot characteristic and adjacent maximum likelihood frequency deviation to estimate is special Property, first estimate to carry out sliding window storage to maximum likelihood frequency deviation, it is thus achieved that maximum likelihood frequency deviation estimates collection；Use again Smothing filtering asks for the meansigma methods that maximum likelihood frequency deviation is estimated to concentrate all maximum likelihood frequency deviations to estimate, as Whole maximum likelihood frequency deviation is estimated；So, it is possible to improve precision and the performance that frequency deviation is estimated.

Fig. 4 and Fig. 5 is that the substandard frequency deviation of TD-LTE estimates mean square deviation performance schematic diagram, the embodiment of the present invention pair The frequency deviation that the maximum likelihood frequency deviation of formula (26) obtains after estimating to carry out sliding window storage and smothing filtering is estimated mean square Error performance curve, as shown in the real point line in Fig. 4 and Fig. 5；Prior art is according to the maximum of formula (26) seemingly So frequency deviation estimates that the frequency deviation obtained estimates mean square error performance curve, as shown in the imaginary point line in Fig. 4 and Fig. 5； Fig. 4 and Fig. 5 is all to be standard testing channel ETU300HZ at channel, and the normalization frequency deviation value of default is ε=0.05, MIMO dual-mode antenna is configured to 1 × 2, and OFDM sub-carrier number is 512, pilot sub-carrier number N_p Under conditions of being 240, pilot resources corresponding for Fig. 4 with Fig. 5 is respectively RB=2 and RB=20, by Fig. 4 and Fig. 5 Visible, the embodiment of the present invention is relative to prior art, it is possible to significantly improves frequency deviation and estimates performance.

Fig. 6 and Fig. 7 is the schematic diagram of corresponding frequency deviation estimated probability distribution function, and the embodiment of the present invention is corresponding Frequency deviation estimated probability distribution function curve, shown in solid as in Fig. 6 and Fig. 7；The frequency deviation that prior art is corresponding Estimated probability distribution function curve, as shown in the dotted line in Fig. 6 and Fig. 7；From Fig. 6 and Fig. 7, the present invention Embodiment can not only significantly improve frequency deviation and estimate performance, moreover it is possible to reduce the dynamic span that frequency deviation is estimated.

For realizing said method, embodiments provide a kind of frequency deviation estimation device, owing to device solves The principle of problem is similar to method, and therefore, the enforcement of device may refer to the enforcement of preceding method, repeats it Place repeats no more.

The structural representation of the Fig. 8 frequency deviation estimation device for providing in the embodiment of the present invention, as shown in Figure 8, This device includes: sliding window memory element 800, smothing filtering unit 801；Wherein,

Described sliding window memory element 800, for estimating to carry out sliding window storage to maximum likelihood frequency deviation, it is thus achieved that Maximum-likelihood frequency deviation estimates collection；

Described smothing filtering unit 801, is used for using smothing filtering that maximum likelihood frequency deviation being estimated, collection closes And, obtain final maximum likelihood frequency deviation and estimate.

For convenience of description, each several part of the above frequency deviation estimation device with function be divided into various module or Unit is respectively described.A kind of preferred implementation that the above dividing mode only embodiment of the present invention is given, merit The dividing mode of energy module is not construed as limiting the invention.

In being embodied as, described sliding window memory element 800, specifically for estimating ε by maximum likelihood frequency deviation⁽ⁿ⁾Deposit Enter in the FIFO FIFO sliding window memorizer of a length of K, it is thus achieved that maximum likelihood frequency deviation estimates collection {ε^(n-k), k=0,1 ..., K-1}；Wherein, n, K are positive integer.

In being embodied as, described smothing filtering unit 801, specifically for according to formulaAsk Take the meansigma methods that maximum likelihood frequency deviation is estimated to concentrate all maximum likelihood frequency deviations to estimate, and described meansigma methods is made Estimate for final maximum likelihood frequency deviationWherein, n, K are positive integer.

In being embodied as, described device also includes:

Acquiring unit 802, the likelihood function for estimating according to frequency deviation obtains maximum likelihood frequency deviation and estimates.

In being embodied as, described acquiring unit 802, specifically for pilot symbol transmitted matrix D⁽ⁿ⁾, discrete The matrix that the front L row of fourier transform matrix F and described matrix F are constitutedCarry out pretreatment, it is thus achieved that square Battle array P⁽ⁿ⁾；

Frequency pilot sign matrix R will be received⁽ⁿ⁾And described matrix P⁽ⁿ⁾It is respectively stored into the FIFO sliding window of a length of M In memorizer；

According to the reception frequency pilot sign matrix R obtained from described sliding window memorizer⁽ⁿ⁾, matrix P⁽ⁿ⁾And build The likelihood function that M frequency deviation matrix calculus frequency deviation is estimated, and the likelihood function or cumulative estimated according to frequency deviation The likelihood function that frequency deviation is estimated obtains maximum likelihood frequency deviation and estimates；Wherein, M, n are positive integer.

In actual applications, described sliding window memory element 800, smothing filtering unit 801, acquiring unit 802 Can be by being positioned at the centre of receiving device in MIMO SC-FDMA system or MIMO-OFDM system Reason device (CPU), microprocessor (MPU), digital signal processor (DSP) or field-programmable Gate array (FPGA) realizes.

Although preferred embodiments of the present invention have been described, but those skilled in the art once know base This creativeness concept, then can make other change and amendment to these embodiments.So, appended right is wanted Ask and be intended to be construed to include preferred embodiment and fall into all changes and the amendment of the scope of the invention.

Obviously, those skilled in the art can carry out various change and modification without deviating from this to the present invention Bright spirit and scope.So, if the present invention these amendment and modification belong to the claims in the present invention and Within the scope of its equivalent technologies, then the present invention is also intended to comprise these change and modification.

Claims (10)

1. a frequency deviation estimating method, it is characterised in that described method includes:

Estimate to carry out sliding window storage to maximum likelihood frequency deviation, it is thus achieved that maximum likelihood frequency deviation estimates collection；

Use smothing filtering that maximum likelihood frequency deviation being estimated, collection merges, obtain final maximum likelihood frequency deviation Estimate.

Method the most according to claim 1, it is characterised in that described maximum likelihood frequency deviation is estimated into Row sliding window stores, it is thus achieved that maximum likelihood frequency deviation estimates collection, including:

Maximum likelihood frequency deviation is estimated ε⁽ⁿ⁾It is stored in the FIFO FIFO sliding window memorizer of a length of K, it is thus achieved that Maximum likelihood frequency deviation estimates collection { ε^(n-k), k=0,1 ..., K-1}；

Wherein, n, K are positive integer.

Method the most according to claim 2, it is characterised in that described employing smothing filtering is to maximum seemingly So frequency deviation estimates that collection merges, and obtains final maximum likelihood frequency deviation and estimates, including:

According toAsk for maximum likelihood frequency deviation to estimate to concentrate all maximum likelihood frequency deviations to estimate Meansigma methods, and described meansigma methods is estimated as final maximum likelihood frequency deviation；Wherein, n, K are positive integer.

Method the most according to claim 1, it is characterised in that described method also includes:

The likelihood function estimated according to frequency deviation obtains maximum likelihood frequency deviation and estimates.

Method the most according to claim 4, it is characterised in that the described likelihood estimated according to frequency deviation Function obtains maximum likelihood frequency deviation and estimates, including:

To pilot symbol transmitted matrix D⁽ⁿ⁾, the front L row of discrete Fourier transform (DFT) matrix F and described matrix F The matrix constitutedCarry out pretreatment, it is thus achieved that matrix P⁽ⁿ⁾；

Frequency pilot sign matrix R will be received⁽ⁿ⁾And described matrix P⁽ⁿ⁾It is respectively stored into the FIFO sliding window of a length of M In memorizer；

According to the reception frequency pilot sign matrix R obtained from described sliding window memorizer⁽ⁿ⁾, matrix P⁽ⁿ⁾And build The likelihood function that M frequency deviation matrix calculus frequency deviation is estimated, and the likelihood function or cumulative estimated according to frequency deviation The likelihood function that frequency deviation is estimated obtains maximum likelihood frequency deviation and estimates；Wherein, M, n are positive integer.

6. a frequency deviation estimation device, it is characterised in that described device includes: sliding window memory element, smooth Filter unit；Wherein,

Described sliding window memory element, for estimating to carry out sliding window storage to maximum likelihood frequency deviation, it is thus achieved that maximum is seemingly So frequency deviation estimates collection；

Described smothing filtering unit, is used for using smothing filtering that maximum likelihood frequency deviation being estimated, collection merges, Obtain final maximum likelihood frequency deviation to estimate.

Device the most according to claim 6, it is characterised in that described sliding window memory element, specifically uses In maximum likelihood frequency deviation is estimated ε⁽ⁿ⁾It is stored in the FIFO FIFO sliding window memorizer of a length of K, it is thus achieved that Maximum-likelihood frequency deviation estimates collection { ε^(n-k), k=0,1 ..., K-1}；

Wherein, n, K are positive integer.

Device the most according to claim 7, it is characterised in that described smothing filtering unit, specifically uses In basisAsk for maximum likelihood frequency deviation to estimate to concentrate the flat of all maximum likelihood frequency deviations estimation Average, and described meansigma methods is estimated as final maximum likelihood frequency deviation；Wherein, n, K are positive integer.

Device the most according to claim 6, it is characterised in that described device also includes:

Acquiring unit, the likelihood function for estimating according to frequency deviation obtains maximum likelihood frequency deviation and estimates.

Device the most according to claim 9, it is characterised in that described acquiring unit, specifically for To pilot symbol transmitted matrix D⁽ⁿ⁾, the front L row of discrete Fourier transform (DFT) matrix F and described matrix F constitute MatrixCarry out pretreatment, it is thus achieved that matrix P⁽ⁿ⁾；

Frequency pilot sign matrix R will be received⁽ⁿ⁾And described matrix P⁽ⁿ⁾It is respectively stored into the FIFO sliding window of a length of M In memorizer；

According to the reception frequency pilot sign matrix R obtained from described sliding window memorizer⁽ⁿ⁾, matrix P⁽ⁿ⁾And build The likelihood function that M frequency deviation matrix calculus frequency deviation is estimated, and the likelihood function or cumulative estimated according to frequency deviation The likelihood function that frequency deviation is estimated obtains maximum likelihood frequency deviation and estimates；Wherein, M, n are positive integer.