CN103107969A - Incremental iterative time-varying channel evaluation and inter carrier interference (ICI) elimination method of fast orthogonal frequency division multiplexing (OFDM) system - Google Patents

Abstract

An incremental iterative time-varying channel evaluation and inter carrier interference (ICI) elimination method of fast orthogonal frequency division multiplexing (OFDM) system is applied in the field of wireless communication channel evolution, and is used for channel evaluation based on pilot frequency in the circumstance that a severe ICI influence of a time-varying channel occurs. The incremental iterative time-varying channel evaluation and ICI elimination method of fast OFDM system is characterized in that the sum of an ICI and a channel noise (SIN) is taken as a denosing object of a Kalman wave filter, and the ICI influence when the Kalman wave filter evaluates is eliminated. In addition, an incremental mode is used by data used for iteration, the number of the sub-carrier waves used for measurement increases slowly along two sides of each pilot frequency point in an iterative process, and thus the influence caused by the ICI is suppressed. The performance of the incremental iterative time-varying channel evaluation and ICI elimination method of fast OFDM system is improved significantly compared with an existing algorithm in the circumstance that signal noise ratio (SNR) is small.

Description

A kind of progressive iteration time varying channel of fast change ofdm system is estimated and the ICI removing method

Technical field

The present invention relates to a kind of progressive OFDM iterative channel estimation method with the ICI elimination.The association area that belongs to channel estimation studies in radio communication.

Technical background

OFDM (Orthogonal Frequency Division Multiplexing, OFDM) be a kind of of multi-carrier modulation, its thought is that channel is divided into several mutually orthogonal subcarriers, convert serial data at a high speed to parallel many groups low rate data streams, transfer to respectively on these subcarriers and transmit, to improve the availability of frequency spectrum.If the signal bandwidth on subcarrier is less than the correlation bandwidth of channel, subcarrier can be regarded the flatness decline as, thereby eliminates intersymbol interference.Due to its high spectrum utilization and good antijamming capability, the OFDM technology has been widely used in audio frequency, video field and the commercial signal communication system of broadcast type, and main application comprises: the digital audio broadcasting (DAB) of asymmetrical Digital Subscriber Loop (ADSL), etsi standard, digital video broadcasting (DVB), high definition TV (HDTV), wireless lan (wlan) etc.

The frequency selective characteristic decline that the OFDM technology can be subject to being caused by the channel multi-path time delay simultaneously, and the impact of the time selective fading that is caused by the doppler spread of channel descend systematic function.Thereby the variation time of advent that frequency selective characteristic causes receiving signal has influence on its amplitude and phase place.The selection of time characteristic causes the orthogonality between the ofdm system subcarrier to be affected, and causes the interference (intersubcarrier interference, ICI) between subcarrier, causes the hydraulic performance decline of system.Particularly in the high-speed mobile situation, in an OFDM symbol, channel also can great changes will take place, ICI to affect meeting more serious.

For this reason, multiple channel estimation method is arranged at present, estimate by the method for inserting pilot tone and difference.But these algorithms may be inapplicable in the ofdm system of high-speed mobile.In the recent period, basis expansion model (Basis Expansion Model, BEM) algorithm is widely used in simulating the time-frequency doubly selective channel, difference according to the base that uses, can be divided into complex exponential base BEM algorithm (CE-BEM), discrete cosine transform BEM(DTC-BEM), polynomial basis BEM(P-BEM), the spherical BEM(DPS-BEM of discrete expansion) and discrete Karhuen-LoeveBEM(KL-BEM).Among these algorithms, the P-BEM algorithm performance is best.In order to reduce ICI, be suggested from technology for eliminating.By information being mapped on one group of subcarrier, producing ICI from elimination, but can cause the reduction of spectrum efficiency.Separately have algorithm that Data Detection is joined in channel estimating, Data Detection is used for iterative algorithm and the data of channel estimating to be recovered, thereby improves the effect of estimating.Iterative algorithm has wherein namely been considered the iterative algorithm of the Kalman filtering of Given information.Data Detection Algorithm is channel matrix to be carried out QR decompose, and revises the error of data to eliminate ICI.But, because ICI can affect the accuracy that frequency domain is estimated, under the environment of fast moving, need to carry out a large amount of iteration, the result of channel estimating is in the situation that signal to noise ratio (snr) is lower also inaccurate.

Summary of the invention

1. the progressive iteration time varying channel that becomes soon ofdm system is estimated and the ICI removing method, it is characterized in that, in the channel estimating of ofdm system, with ICI and the noise sum SIN denoising object as the Kalman filter, from only using the pilot tone dot information, increase progressively the information that is used for iterative computation, realize according to following steps:

Step (1), transmitting terminal produce and send data, pilot data are inserted into according to the Comb Pilot mode send in data:

Transmitting terminal is set as follows: s represents s OFDM symbol, s=1, and 2 ..., s ... S, each OFDM symbol comprises N subcarrier, n=1,2 ..., n ..., N wherein comprises N _pIndividual frequency pilot sign

And N _dIndividual data symbol, N _d+ N _p=N, n _p=1,2 ..., N _p, the location matrix of pilot tone on frequency domain is expressed as:

Wherein

And guarantee N _p〉=L, L are the maximums that channel multi-path is counted l, i.e. l=1, and 2 ...,, l ..., L, N _pIndividual pilot tone is inserted among N carrier wave and remains unchanged in transmitting procedure by average, and the pilot tone point symbol is expressed as in N carrier wave ${x_p}^{\left(s\right)}=x^{\left(s\right)}\left(P_s\right)={[{x_{p_1}}^{\left(s\right)},{x_{p_2}}^{\left(s\right)},\·\·\·,{x_{{p_N}_p}}^{\left(s\right)}]}^T,$

Step (2), data are sent to receiving terminal by ofdm system, after receiving terminal removes Cyclic Prefix, with polynomial basis extended model P-BEM, channel are carried out modeling according to the following steps:

Step (2.1) utilizes polynomial basis extended model P-BEM to describe to have the two time dispersive channels that select characteristic of time-frequency, the channel impulse response h in n subcarrier l footpath of s OFDM symbol ^(s)(n, l) is expressed as:

h ^(s)(n,l)＝QC _l ^(s)+ξ _l ^(s)(n),0≤n≤N-1，

Wherein, ξ _l ^(s)The model error in expression l footpath of each OFDM symbol during modeling, it is worth less than 10 ^-3, ignore when calculating, namely think h ^(s)(n, l)=QC _l ^(s), Q is the orthogonal basis function matrix of a N * B, C _l ^(s)By B coefficient corresponding to basic function

The vector that forms ${C_l}^{\left(s\right)}={[c_{1,l}^{\left(s\right)},{\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">c</figure-callout>}}_{2,l}^{\left(s\right)},\&CenterDot;\&CenterDot;\&CenterDot;,c_{b,l}^{\left(s\right)},\&CenterDot;\&CenterDot;\&CenterDot;,c_{B,l}^{\left(s\right)}]}^T,$

f _maxThe highest frequency of channel, T _sThe sampling time,

Step (2.2) will be shown as following form at the reception signal list of receiving terminal:

y ^(s)＝H ^(s)x ^(s)+W ^(s)，

Wherein, x ^(s)=[x ₁ ^(s), x ₂ ^(s)X _N ^(s)] ^T, y ^(s)=[y ₁ ^(s), y ₂ ^(s)..., y _N ^(s)] ^TRepresent that respectively s on frequency domain is removed the transmitted signal after Cyclic Prefix and receive signal, W ^(s)The white noise on its frequency domain, H ^(s)The channel matrix of N * N:

Wherein, each element of matrix be multipath channel channel impulse response and, account form is as follows:

$H^{\left(s\right)}(m,k)=\underset{l=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{G_l}^{\left(s\right)}(M,K)e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{k-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l},$

M, k represent above-mentioned matrix H ^(s)The value of the capable k of m row, m=1,2 ..., m ..., M, M≤N, k=1,2 ..., k ..., K, K≤N, τ _lThe time delay in l footpath, G _l ^(s)(M, K) impacts corresponding frequency domain presentation matrix for channel, and its each element is calculated as follows:

${G_l}^{\left(s\right)}(m,k)=\frac{1}{N}\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{h}^{\left(s\right)}(n,l)e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(m-k)n/N},$

Step (2.3) will receive signal according to the P-BEM model and carry out modeling again, be expressed as with the expression formula of P-BEM coefficient as follows:

y ^(s)＝Φ ^(s)g ^(s)+W ^(s)，

Wherein,

g ^(s)=[C ₁ ^{(s) T}, C ₂ ^{(s) T}C _L ^{(s) T}] ^T, the coefficient matrix in expression P-BEM algorithm,

After representing modeling again, the coefficient matrix relevant to sending data, its computational methods are as follows:

$Z_l^{\left(s\right)}=\frac{1}{N}[D_1\mathrm{diag}\left(x^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l,\&CenterDot;\&CenterDot;\&CenterDot;,D_b\mathrm{diag}\left(x^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l,\&CenterDot;\&CenterDot;\&CenterDot;,D_B\mathrm{diag}\left(x^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l],$ Wherein,

${\mathrm{\&Gamma;}}_l={\left[\begin{array}{cccc}{e}^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{p_1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}& e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{p_2-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}& \&CenterDot;\&CenterDot;\&CenterDot;& e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{{p_N}_p-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}\end{array}\right]}^T,$ The Fourier transform in l footpath,

Γ=[Γ ₁, Γ ₂..., Γ _L], total Fourier transform matrix in expression L footpath,

Diag (x ^(s)) represent with vector x ^(s)Be the matrix of diagonal element,

Step (3), utilize the AR model to carry out modeling to channel BEM coefficient:

Step (3.1) is calculated as follows C _l ^(s)Correlation matrix:

$R_{C_l}^{\left(j\right)}={\left({\left(Q\right)}^{H}Q\right)}^{-1}{\left(Q\right)}^H{R}_{h(n,l)}^{\left(j\right)}Q{\left({\left(Q\right)}^{H}Q\right)}^{-1},$

Wherein, j represents the exponent number of being correlated with, and namely carries out the mark space of the OFDM symbol of related operation, and the value of j is [1,0,1],

The C that represents respectively current ofdm signal _l ^(s)C with previous symbol _l ^(s-1)Correlation matrix, the C of current ofdm signal _l ^(s)Autocorrelation matrix, the C of current ofdm signal _l ^(s)C with a rear symbol _l ^(s+1) correlation matrix.() ^HExpression Hermitian computing, $R_{h(n,l)}^{\left(j\right)}=E\left[h(n,l)h^{*}(n+j,l)\right]={\mathrm{\&sigma;}}_{h(n,l)}^2{J}_0\left(2\mathrm{\&pi;}{f}_d{T}_{s}j\right),$ E[wherein] the expression average, J ₀The zero Bessel function of () expression first kind, f _d=vf _c/ c is the maximum doppler frequency of the translational speed of terminal when being v, f _cBe carrier frequency, c is the light velocity,

Represent the variance of the channel impulse response in l footpath, and hypothesis

Step (3.2) obtains the state transition equation of channel P-BEM parameter according to the Yule-Walker equation:

g _(s)＝Ag _(s-1)+U _(s)，

Ofdm system is sent the time sequencing g of symbol ^(s)Regard state transitions process g in control system as _(s), i.e. g ^(s)=g _(s), state transition equation coefficient A=diag (a ₁, a ₂..., a _l... a _L),

The matrix of diag (x) expression take vector x as diagonal element, U _(s)Represent the modeling error of the AR model of s OFDM symbol;

Step (4), the Kalman filter is carried out initialization and calculates initial renewal equation:

Step (4.1), according to the following formula the Kalman filter is carried out initialization:

$g_{\left(0\right|0)}=0_{\mathrm{LB},1},$ $P_{\left(0\right|0)}=\mathrm{diag}(R_{C_1}^{\left(0\right)},R_{C_2}^{\left(0\right)}\&CenterDot;\&CenterDot;\&CenterDot;R_{C_L}^{\left(0\right)}),$

Form as

And P _(s|s)The previous s of middle subscript represents that all current state is g _(s), a rear s represents s OFDM symbol,

P _(0|0)For the initial value that calculates,

The g of expression OFDM symbol _(s)Initial value, P _(0|0)Expression

Corresponding error correlation matrix, O _{LB, 1}The null matrix of LB * 1,

Step (4.2) is calculated as follows the initial time renewal equation of Kalman:

i＝1，s＝1，

${\hat{g}}_{\left(s\right)}=A{\hat{g}}_{\left(0\right|0)},$

P _(s)＝AP _(0|0)(A) ^H+V[U _(s)]，

I represents iterations,

State estimation g in expression Kalman equation _(s)Intermediate variable, P _(s)The expression intermediate variable

Corresponding error correlation matrix; Use V[] the expression covariance matrix, V[U _(s)]=diag (u ₁, u ₂U _L),

Step (5), carry out iterative channel estimation computing for the first time, this moment iterations i=1, only use the subcarrier place at pilot tone point place to receive data in this iteration and do channel estimating, data except the subcarrier of pilot tone point place on other subcarriers are considered as ICI, eliminate unknown data to the impact of pilot tone place channel estimating with the SIN method, realize the auxiliary Kalman channel estimating of the pilot tone of disturbing without ICI, concrete steps are as follows:

Step (5.1), the reception signal that only will receive the carrier position that in signal, each pilot tone point is corresponding is used for calculating, and will receive signal and be divided into data on the subcarrier of each pilot tone point place, interference and noise three parts of data on other subcarrier except pilot tone point subcarrier to each pilot tone point place subcarrier are shown below:

${y_p}^{\left(s\right)}=y^{\left(s\right)}\left(P_s\right)={H^{\left(s\right)}}_{[P_s,P_s]}{x_{p_{n_p}}}^{\left(s\right)}+{H^{\left(s\right)}}_{[P_s,d_{n_p{n}^{\&prime;}}]}{x_{d_{n_p{n}^{\&prime;}}}}^{\left(s\right)}+W^{\left(s\right)}\left(P_s\right)$

Wherein, $P_s=[p_1,{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">p</figure-callout>}}_2,\&CenterDot;\&CenterDot;\&CenterDot;,p_{N_p}],$ ${x_{d_{n_p{n}^{\&prime;}}}}^{\left(s\right)}={[x^{\left(s\right)}\left(d_{11}\right),x^{\left(s\right)}\left(d_{12}\right),\&CenterDot;\&CenterDot;\&CenterDot;{,x}^{\left(s\right)}\left(d_{{1N}^{\&prime;}}\right),x^{\left(s\right)}\left(d_{21}\right),\&CenterDot;\&CenterDot;\&CenterDot;,x^{\left(s\right)}\left(d_{N_p{N}^{\&prime;}}\right)]}^T,$ N'=1,2 ..., N' represents the distance between each adjacent pilot tone point,

A N _p* N _pUnit matrix, σ ²White Gaussian noise W ^(s)Variance, in following formula second be data on other subcarrier except pilot tone point subcarrier to the interference ICI of pilot tone point place subcarrier,

Step (5.2) is considered to interchannel noise W with data I CI distracter ^(s)(P _s) a part as the denoising object of filter, the algorithm in step (2) is rewritten order according to the method that SIN estimates

The Kalman observational equation of SIN estimation is expressed as:

$y_{p\left(s\right)}={{\mathrm{\&Phi;}}^{\mathrm{SIN}}}_{\left(s\right)}{g}_{\left(s\right)}+W_{\left(s\right)}^{\mathrm{SIN}},$

Wherein:

${{\mathrm{\&Phi;}}^{\mathrm{SIN}}}_{\left(s\right)}=\frac{1}{N}[{z_1^{\left(s\right)}}_{\mathrm{SIN}},{Z_2^{\left(s\right)}}_{\mathrm{SIN}}\&CenterDot;\&CenterDot;\&CenterDot;{Z_L^{\left(s\right)}}_{\mathrm{SIN}}],$

${Z_l^{\left(s\right)}}_{\mathrm{SIN}}=\frac{1}{N}[D_1^{\mathrm{SIN}}\mathrm{diag}\left({x_p}^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l^{\mathrm{SIN}}\&CenterDot;\&CenterDot;\&CenterDot;D_B^{\mathrm{SIN}}\mathrm{diag}\left({x_p}^{\left(s\right)}{\mathrm{\&Gamma;}}_l^{\mathrm{SIN}}\right)],$

${\mathrm{\&Gamma;}}_l^{\mathrm{SIN}}={\left[\begin{array}{cccc}{e}^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{p_1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_1}& e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{p_2-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}& \&CenterDot;\&CenterDot;\&CenterDot;& e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{{p_N}_p-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}\end{array}\right]}^T,$

${\mathrm{\&Gamma;}}^{\mathrm{SIN}}=\left[\begin{array}{cccc}{{\mathrm{\&Gamma;}}_1}^{\mathrm{SIN}},& {{\mathrm{\&Gamma;}}_2,}^{\mathrm{SIN}}& \&CenterDot;\&CenterDot;\&CenterDot;,& {{\mathrm{\&Gamma;}}_l}^{\mathrm{SIN}}\end{array}\right]$

Step (5.3) is calculated

Covariance matrix

In calculating, hypothesis ICI is white Gaussian noise, order

Because noise and ICI are both separate, so $V\left[W_{\left(s\right)}^{\mathrm{SIN}}\right]=U_{\mathrm{ICI}}+V\left[W^{\left(s\right)}\left(P_s\right)\right];$

U _ICIThe calculating formula of each element in matrix is:

$U_{\mathrm{ICI}}(m,k)=R_{\mathrm{ICI}}^{\left(j\right)}\&ap;4{\mathrm{\&pi;}}^2{T_s}^2{E}_s\left(\underset{l=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_{h(n,l)}^2{\mathrm{\&sigma;}}_{D_l}\right)\mathrm{\&rho;}(\mathrm{\&alpha;},\mathrm{rag},N),$

Wherein, m, the capable k row of the m of k representing matrix, E _sThe power that sends data,

That power is P _vThe time the general function of Doppler's power, f is transmission frequency, it is 0 when calculating for the first time for the marginal date of the iteration distance apart from each pilot tone point of correspondence that α represents, the precision of rag for calculating, rag=[0,1,2,3], and:

ρ(α,rag，N)=ρ(0,rag,N)-ρ ₁(α,rag,N)，

Step (5.4) is calculated kalman gain K by following three formulas respectively _(s), s OFDM symbol transferred to state

The state estimation matrix

With with Corresponding covariance matrix P _(s|s), consist of observation renewal equation group, wherein, Φ=Φ ^SIN, $W_{\left(s\right)}=W_{\left(s\right)}^{\mathrm{SIN}},$

K _(s)＝P _(s)(Φ _(s)) ^H(Φ _(s)P _(s)(Φ _(s)) ^H+V[W _(s)]) ^-1，

${\hat{g}}_{\left(s\right|s)}={\hat{g}}_{\left(s\right)}+K_{\left(s\right)}(y_{\left(s\right)}-{\mathrm{\&Phi;}}_{\left(s\right)}{\hat{g}}_{\left(s\right)}),$

P _(s|s)＝P _(s)-K _(s)Φ _(s)P _(s)，

Step (5.5) calculates channel matrix H according to following formula ^(s)Estimated value:

$H^{\left(s\right)}=\frac{1}{N}\underset{b=1}{\overset{B}{\mathrm{\&Sigma;}}}{D}_b\mathrm{diag}\left(\mathrm{\&Gamma;}{\hat{g}}_{\left(s\right|s)}\right),$

Step (5.6) is utilized following formula to carry out QR to channel matrix and is decomposed, and obtains matrix R ^(s):

H ^(s)＝IR ^(s)

Wherein I is a unit matrix, R ^(s)A upper triangular matrix,

Step (5.7), by following formula, data are carried out the QR Data Detection:

Y ' wherein ^(s)=(I) ^Hy ^(s), With Respectively the detected value of data and the result after the quantification of detected value planisphere, [] _{M, k}Represent the capable k row of m of matrix, [] _mM element of vector, [] _kK element of vector, O () expression demodulation computing, m, k representing matrix H ^(s)The value of the capable k of m row, m=1,2 ..., m ..., M, M≤N, k=1,2 ..., k ..., K, K≤N,

Step (6), iterations i=i+1, iterative computation number of times i〉1 o'clock, iteration is done channel estimating with the data that the subcarrier at place, a pilot tone point place subcarrier place adjacent with its both sides receives for the second time, data on all the other subcarriers are considered as ICI, the data that are used for calculating of the each increase of iteration thereafter, the data that the subcarrier place of the subcarrier both sides at the data place that is that last iteration be used for to calculate receives, data on all the other subcarriers are considered as ICI, and it is as described below that the SIN method is carried out the iterative channel estimation computing:

Step (6.1), to receive signal and be divided into pilot tone point place subcarrier data, the subcarrier data at the data place of be used for calculating is used for the data calculated to pilot tone point and is used for interference and noise four parts of the data places subcarrier of calculating, is shown below:

In following formula, the 3rd is the ICI interference after upgrading,

Step (6.2), the method described in step (5.2) is calculated Φ ^SIN _(s),

Step (6.3) increases with iterations, and the data that are used for iteration increase, and upgrade this moment

Covariance matrix U _ICIBe calculated as follows:

$U_{\mathrm{ICI}}\&ap;4{\mathrm{\&pi;}}^2{T_s}^2{E}_s\left(\underset{l=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_{h(n,l)}^2{\mathrm{\&sigma;}}_{D_l}\right)[\frac{1}{2}\mathrm{\&rho;}(A+m,\mathrm{rag},N)+\frac{1}{2}\mathrm{\&rho;}(A-m,\mathrm{rag},N)]$

Wherein, what A+m represented pilot tone point right side is used for the data calculated to the distance of corresponding pilot tone point, and A-m represents that the data that are used for calculating in pilot tone point left side arrive the distance of the pilot tone point of correspondence,

Step (6.4) is calculated kalman gain K by the described method of step (5.5) _(s), s OFDM symbol transferred to state

The state estimation matrix

With with

Corresponding covariance matrix P _(s|s), consist of observation renewal equation group, wherein, Φ=Φ ^SIN,

K _(s)＝P _(s)(Φ _(s)) ^H(Φ _(s)P _(s)(Φ _(s)) ^H+V[W _(s)]) ^-1，

${\hat{g}}_{\left(s\right|s)}={\hat{g}}_{\left(s\right)}+K_{\left(s\right)}(y_{\left(s\right)}-{\mathrm{\&Phi;}}_{\left(s\right)}{\hat{g}}_{\left(s\right)}),$

P _(s|s)＝P _(s)-K _(s)Φ _(s)P _(s)，

Step (6.5) calculates the estimated value H of channel matrix by the described method of step (5.5) ^(s):

$H^{\left(s\right)}=\frac{1}{N}\underset{b=1}{\overset{B}{\mathrm{\&Sigma;}}}{D}_b\mathrm{diag}\left(\mathrm{\&Gamma;}{\hat{g}}_{\left(s\right|s)}\right),$

Step (6.6) is carried out the QR decomposition by the described method of step (5.6) to channel matrix and is obtained R ^(s):

H ^(s)＝IR ^(s)，

Step (6.7), by the described method of step (5.7), data are carried out the QR Data Detection:

Step (7) judges whether whether all data all have been used for iteration, and if so, algorithm finishes, if not, continue,

Step (8) by the number of times of judgement iteration, determines whether to need to increase the input data of iterative algorithm, and evaluation algorithm is as follows:

Set step delta,

Compare iterations i and Δ μ+2, wherein, $\mathrm{\&mu;}=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,\mathrm{\&Delta;}\&CenterDot;\mathrm{\&mu;}+2<\frac{N}{N_p},$ If i=Δ μ+2, the data that the data at the subcarrier place of the subcarrier both sides at the data place of selecting to be used for to calculate were calculated as next iteration being used for of newly adding, bring the SIN algorithm into together with the data that are used for calculating, be brought in step (6) and again count;

If i ≠ Δ μ+2 do not increase for the data of calculating, return to step (6) and carry out iteration;

Finish.

Description of drawings

Fig. 1 is the ofdm system that the present invention is suitable for;

Fig. 2 is basic principle flow chart of the present invention;

Fig. 3 is the progressive step-length schematic diagram that relates in the present invention.1. reception signal when wherein Fig. 3 (a) expression is calculated for the first time is wherein that the part of black represents frequency place subcarrier

3. the subcarrier that represents the ICI place ${x_{d_{n_p{n}^{\&prime;}}}}^{\left(s\right)}={[x^{\left(s\right)}\left(d_{11}\right),x^{\left(s\right)}\left(d_{12}\right),\&CenterDot;\&CenterDot;\&CenterDot;,x^{\left(s\right)}\left(d_{{1N}^{\&prime;}}\right),x^{\left(s\right)}\left(d_{21}\right),\&CenterDot;\&CenterDot;\&CenterDot;,x^{\left(s\right)}\left(d_{N_p{N}^{\&prime;}}\right)]}^T;$ 1. reception signal when Fig. 3 (b) expression is calculated for the second time wherein represents pilot tone point place subcarrier

2. represent that this iteration newly adds the subcarrier at the data place of calculating

With

3. the subcarrier at expression expression ICI place, 4. represent α, namely is used for the data and the distance between pilot tone point at the edge of iteration, and α=1,5. represent the distance between pilot tone point at this moment

Fig. 3 (c) expression iterations i=Δ μ+2, and the reception signal of i＞2 o'clock, the data places subcarrier that wherein 1. represents to be used to calculate last time

${x^{\left(s\right)}\left(p_{N_p}\right)+[x^{\left(s\right)}\left(d_{11}\right),x^{\left(s\right)}\left(d_{12}\right),\&CenterDot;\&CenterDot;\&CenterDot;,x^{\left(s\right)}\left(d_{1(i-1)}\right),\&CenterDot;\&CenterDot;\&CenterDot;x^{\left(s\right)}\left(d_{n_{p}1}\right),\&CenterDot;\&CenterDot;\&CenterDot;,x^{\left(s\right)}\left(d_{n_{p(i-1)}}\right),\&CenterDot;\&CenterDot;\&CenterDot;,x^{\left(s\right)}\left(d_{N_p(i-1)}\right)]}^T,$ 2. represent that this iteration newly adds the subcarrier at the data place of calculating With

4. represent α, at this moment, α=Δ+1;

Fig. 4 be the present invention with based on Kalman but the performance comparison of the channel estimation method of the ICI that is untreated.Wherein

With

Represent respectively in conventional method 1 time, 3 times and 10 iteration after result,

With

Represent the result after crossing of the present invention 1 time, 3 times and 10 iteration,

The expression data are whole when known,

The theoretical value upper limit of this kind algorithm.

Embodiment

The fast progressive iteration time varying channel that becomes ofdm system is estimated and the ICI removing method, it is characterized in that, in the channel estimating of ofdm system, with ICI and the noise sum SIN denoising object as the Kalman filter, from only using the pilot tone dot information, increase progressively the information that is used for iterative computation, realize according to following steps:

Step (1), transmitting terminal produce and send data, pilot data are inserted into according to the Comb Pilot mode send in data:

Transmitting terminal is set as follows: s represents s OFDM symbol, s=1, and 2 ..., s ... S, each OFDM symbol comprises N subcarrier, n=1,2 ..., n ..., N wherein comprises N _pIndividual frequency pilot sign

And N _dIndividual data symbol, N _d+ N _p=N, n _p=1,2 ..., N _p, the location matrix of pilot tone on frequency domain is expressed as:

Wherein

And guarantee N _p〉=L, L are the maximums that channel multi-path is counted l, i.e. l=1, and 2 ...,, l ..., L, N _pIndividual pilot tone is inserted among N carrier wave and remains unchanged in transmitting procedure by average, and the pilot tone point symbol is expressed as in N carrier wave ${x_p}^{\left(s\right)}=x^{\left(s\right)}\left(P_s\right)={[{x_{p_1}}^{\left(s\right)},{x_{p_2}}^{\left(s\right)},\&CenterDot;\&CenterDot;\&CenterDot;,{x_{{p_N}_p}}^{\left(s\right)}]}^T,$

Step (2), data are sent to receiving terminal by ofdm system, after receiving terminal removes Cyclic Prefix, with polynomial basis extended model P-BEM, channel are carried out modeling according to the following steps:

Step (2.1) utilizes polynomial basis extended model P-BEM to describe to have the two time dispersive channels that select characteristic of time-frequency, the channel impulse response h in n subcarrier l footpath of S OFDM symbol ^(s)(n, l) is expressed as:

h ^(s)(n,l)＝QC _l ^(s)+ξ _l ^(s)(n),0≤n≤N-1，

Wherein, ξ ₁ ^(s)The model error in expression l footpath of each OFDM symbol during modeling, it is worth less than 10 ^-3, ignore when calculating, namely think h ^(s)(n, l)=QC _l ^(s), Q is the orthogonal basis function matrix of a N * B, C _l ^(s)By B coefficient corresponding to basic function

The vector that forms ${{C_l}^{\left(s\right)}=[c_{1,l}^{\left(s\right)},{\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">c</figure-callout>}}_{2,l}^{\left(s\right)},\&CenterDot;\&CenterDot;\&CenterDot;,c_{b,l}^{\left(s\right)},\&CenterDot;\&CenterDot;\&CenterDot;,c_{B,l}^{\left(s\right)}]}^T,$

f _maxThe highest frequency of channel, T _sThe sampling time,

Step (2.2) will be shown as following form at the reception signal list of receiving terminal:

y ^(s)＝H ^(s)x ^(s)+W ^(s)，

Wherein, x ^(s)=[x ₁ ^(s), x ₂ ^(s)X _N ^(s)] ^T, y ^(s)=[y ₁ ^(s), y ₂ ^(s)..., y _N ^(s)] ^TRepresent that respectively s on frequency domain is removed the transmitted signal after Cyclic Prefix and receive signal, W ^(s)The white noise on its frequency domain, H ^(s)The channel matrix of N * N:

Wherein, each element of matrix be multipath channel channel impulse response and, account form is as follows:

$H^{\left(s\right)}(m,k)=\underset{l=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{G_l}^{\left(s\right)}(M,K)e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{k-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l},$

M, k represent above-mentioned matrix H ^(s)The value of the capable k of m row, m=1,2 ..., m ..., M, M≤N, k=1,2 ..., k ..., K, K≤N, τ _lThe time delay in l footpath, G _l ^(s)(M, K) impacts corresponding frequency domain presentation matrix for channel, and its each element is calculated as follows:

$G_l^{\left(s\right)}(m,k)=\frac{1}{N}\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{h}^{\left(s\right)}(n,l)e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(m-k)n/N},$

Step (2.3) will receive signal according to the P-BEM model and carry out modeling again, be expressed as with the expression formula of P-BEM coefficient as follows:

y ^(s)＝Φ ^(s)g ^(s)+W ^(s)，

Wherein,

g ^(s)=[C ₁ ^{(s) T}, C ₂ ^{(s) T}C _L ^{(s) T}] ^T, the coefficient matrix in expression PBEM algorithm,

After representing modeling again, the coefficient matrix relevant to sending data, its computational methods are as follows:

$Z_l^{\left(s\right)}=\frac{1}{N}[D_1\mathrm{diag}\left(x^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l,\&CenterDot;\&CenterDot;\&CenterDot;,D_b\mathrm{diag}\left(x^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l,\&CenterDot;\&CenterDot;\&CenterDot;,D_B\mathrm{diag}\left(x^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l],$ Wherein,

${\mathrm{\&Gamma;}}_l={\left[\begin{array}{cccc}{e}^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{p_1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}& e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{p_2-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}& \&CenterDot;\&CenterDot;\&CenterDot;& e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{{p_N}_p-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}\end{array}\right]}^T,$ The Fourier transform in l footpath,

Γ=[Γ ₁, Γ ₂..., Γ _L], total Fourier transform matrix in expression L footpath,

Diag (x ^(s)) represent with vector x ^(s)Be the matrix of diagonal element,

Step (3), utilize the AR model to carry out modeling to channel BEM coefficient:

Step (3.1) is calculated as follows C _l ^(s)Correlation matrix:

$R_{C_l}^{\left(j\right)}={\left({\left(Q\right)}^{H}Q\right)}^{-1}{\left(Q\right)}^H{R}_{h(n,l)}^{\left(j\right)}Q{\left({\left(Q\right)}^{H}Q\right)}^{-1},$

Wherein, j represents the exponent number of being correlated with, and namely carries out the mark space of the OFDM symbol of related operation, and the value of j is [1,0,1],

The C that represents respectively current ofdm signal _l ^(s)C with previous symbol _l ^(s-1)Correlation matrix, the C of current ofdm signal _l ^(s)Autocorrelation matrix, the C of current ofdm signal _l ^(s)C with a rear symbol _l ^(s+1) correlation matrix.() ^HExpression Hermitian computing, $R_{h(n,l)}^{\left(j\right)}=E\left[h(n,l)h^{*}(n+j,l)\right]={\mathrm{\&sigma;}}_{h(n,l)}^2{J}_0\left(2\mathrm{\&pi;}{f}_d{T}_{s}j\right),$ E[wherein] the expression average, J ₀The zero Bessel function of () expression first kind, f _d=vf _c/ c is the maximum doppler frequency of the translational speed of terminal when being v, f _cBe carrier frequency, c is the light velocity,

Represent the variance of the channel impulse response in l footpath, and hypothesis

Step (3.2) obtains the state transition equation of channel P-BEM parameter according to the YuleWalker equation:

g _(s)＝Ag _(s-1)+U _(s)，

Ofdm system is sent the time sequencing g of symbol ^(s)Regard state transitions process g in control system as _(s), i.e. g ^(s)=g _(s), state transition equation coefficient A=diag (a ₁, a ₂..., a _l... a _L),

The matrix of diag (x) expression take vector x as diagonal element, U _(s)Represent the modeling error of the AR model of s OFDM symbol;

Step (4), the Kalman filter is carried out initialization and calculates initial renewal equation:

Step (4.1), according to the following formula the Kalman filter is carried out initialization:

${\hat{g}}_{\left(0\right|0)}=0_{LB,1},P_{\left(0\right|0)}=\mathrm{diag}(R_{C_1}^{\left(0\right)},R_{C_2}^{\left(0\right)}\&CenterDot;\&CenterDot;\&CenterDot;R_{C_L}^{\left(0\right)}),$

Form as

And P _(s|s)The previous s of middle subscript represents that all current state is g _(s), a rear s represents s OFDM symbol, P _(0|0)For the initial value that calculates,

The g of expression OFDM symbol _(s)Initial value, P _(0|0)Expression

Corresponding error correlation matrix, O _{LB, 1}The null matrix of LB * 1,

Step (4.2) is calculated as follows the initial time renewal equation of Kalman:

i＝1，s＝1，

${\hat{g}}_{\left(s\right)}=A{\hat{g}}_{\left(0\right|0)},$

P _(s)＝AP _(0|0)(A) ^H+V[U _(s)]，

I represents iterations,

State estimation g in expression Kalman equation _(s)Intermediate variable, P _(s)The expression intermediate variable Corresponding error correlation matrix; Use V[] the expression covariance matrix, V[U _(s)]=diag (u ₁, u ₂U _L),

Step (5), carry out iterative channel estimation computing for the first time, this moment iterations i=1, only use the subcarrier place at pilot tone point place to receive data in this iteration and do channel estimating, data except the subcarrier of pilot tone point place on other subcarriers are considered as ICI, eliminate unknown data to the impact of pilot tone place channel estimating with the SIN method, realize the auxiliary Kalman channel estimating of the pilot tone of disturbing without ICI, concrete steps are as follows:

Step (5.1), the reception signal that only will receive the carrier position that in signal, each pilot tone point is corresponding is used for calculating, and will receive signal and be divided into data on the subcarrier of each pilot tone point place, interference and noise three parts of data on other subcarrier except pilot tone point subcarrier to each pilot tone point place subcarrier are shown below:

${y_p}^{\left(s\right)}=y^{\left(s\right)}\left(P_s\right)={H^{\left(s\right)}}_{[P_s,P_s]}{x_{{p_n}_p}}^{\left(s\right)}+{H^{\left(s\right)}}_{[P_s,d_{n_p{n}^{\&prime;}]}}{x_{d_{n_p{n}^{\&prime;}}}}^{\left(s\right)}+W^{\left(s\right)}\left(P_s\right)$

Wherein, $P_s=[p_1,{\mathrm{<figure-callout\; id="2"\; label="p"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">p</figure-callout>}}_2,\&CenterDot;\&CenterDot;\&CenterDot;,{p_N}_p],$ ${x_{d_{n_p{n}^{\&prime;}}}}^{\left(s\right)}={[x^{\left(s\right)}\left(d_{11}\right),x^{\left(s\right)}\left(d_{12}\right),\&CenterDot;\&CenterDot;\&CenterDot;{,x}^{\left(s\right)}\left(d_{{1N}^{\&prime;}}\right),x^{\left(s\right)}\left(d_{21}\right),\&CenterDot;\&CenterDot;\&CenterDot;,x^{\left(s\right)}\left(d_{N_p{N}^{\&prime;}}\right)]}^T,$

N'=1,2 ..., N' represents the distance between each adjacent pilot tone point,

A N _p* N _pUnit matrix, σ ²White Gaussian noise W ^(s)Variance, in following formula second be data on other subcarrier except pilot tone point subcarrier to the interference ICI of pilot tone point place subcarrier,

Step (5.2) is considered to interchannel noise W with data I CI distracter ^(s)(P _s) a part as the denoising object of filter, the algorithm in step (2) is rewritten order according to the method that SIN estimates

The Kalman observational equation of SIN estimation is expressed as:

$y_{p\left(s\right)}={{\mathrm{\&Phi;}}^{\mathrm{SIN}}}_{\left(s\right)}{g}_{\left(s\right)}+W_{\left(s\right)}^{\mathrm{SIN}},$

Wherein:

${{\mathrm{\&Phi;}}^{\mathrm{SIN}}}_{\left(s\right)}=\frac{1}{N}[{z_1^{\left(s\right)}}_{\mathrm{SIN}},{Z_2^{\left(s\right)}}_{\mathrm{SIN}}\&CenterDot;\&CenterDot;\&CenterDot;{Z_L^{\left(s\right)}}_{\mathrm{SIN}}],$

${Z_l^{\left(s\right)}}_{\mathrm{SIN}}=\frac{1}{N}[D_1^{\mathrm{SIN}}\mathrm{diag}\left({x_p}^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l^{\mathrm{SIN}}\&CenterDot;\&CenterDot;\&CenterDot;D_B^{\mathrm{SIN}}\mathrm{diag}\left({x_p}^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l^{\mathrm{SIN}}],$

${\mathrm{\&Gamma;}}_l^{\mathrm{SIN}}={\left[\begin{array}{cccc}{e}^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{p_1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}& e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{p_2-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}& \&CenterDot;\&CenterDot;\&CenterDot;& e^{-\mathrm{<figure-callout\; id="2"\; label="j"\; filenames="DEST\_PATH\_HDA00002930290800031.png"\; state="\{\{state\}\}">j</figure-callout>}2\mathrm{\&pi;}(\frac{{p_N}_p-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}\end{array}\right]}^T,$

Γ ^SIN＝[Γ ₁ ^SIN,Γ ₂, ^SIN…,Γ _l ^SIN]

Step (5.3) is calculated

Covariance matrix

In calculating, hypothesis ICI is white Gaussian noise, order

Because noise and ICI are both separate, so $V\left[W_{\left(s\right)}^{\mathrm{SIN}}\right]=U_{\mathrm{ICI}}+V\left[W^{\left(s\right)}\left(P_S\right)\right];$

U _ICIThe calculating formula of each element in matrix is:

$U_{\mathrm{ICI}}(m,k)=R_{\mathrm{ICI}}^{\left(j\right)}\&ap;4{\mathrm{\&pi;}}^2{T_s}^2{E}_s\left(\underset{l=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_{h(n,l)}^2{\mathrm{\&sigma;}}_{D_l}\right)\mathrm{\&rho;}(\mathrm{\&alpha;},\mathrm{rag},N),$

Wherein, m, the capable k row of the m of k representing matrix, E _sThe power that sends data,

That power is P _vThe time the general function of Doppler's power, f is transmission frequency, it is 0 when calculating for the first time for the marginal date of the iteration distance apart from each pilot tone point of correspondence that α represents, the precision of rag for calculating, rag=[0,1,2,3], and:

ρ(α,rag，N)=ρ(0,rag,N)-ρ ₁(α,rag,N)

Step (5.4) is calculated kalman gain K by following three formulas respectively _(s), s OFDM symbol transferred to state

The state estimation matrix

With with

Corresponding covariance matrix P _(s|s), consist of observation renewal equation group, wherein, Φ=Φ ^SIN $W_{\left(s\right)}=W_{\left(s\right)}^{\mathrm{SIN}},$

K _(s)＝P _(s)(Φ _(s)) ^H(Φ _(s)P _(s)(Φ _(s)) ^H+V[W _(s)]) ^-1，

${\hat{g}}_{\left(s\right|s)}={\hat{g}}_{\left(s\right)}+K_{\left(s\right)}(y_{\left(s\right)}-{\mathrm{\&Phi;}}_{\left(s\right)}{\hat{g}}_{\left(s\right)}),$

P _(s|s)＝P _(s)-K _(s)Φ _(s)P _(s)，

Step (5.5) calculates channel matrix H according to following formula ^(s)Estimated value:

$H^{\left(s\right)}=\frac{1}{N}\underset{b=1}{\overset{B}{\mathrm{\&Sigma;}}}{D}_b\mathrm{diag}\left(\mathrm{\&Gamma;}{\hat{g}}_{\left(s\right|s)}\right),$

Step (5.6) is utilized following formula to carry out QR to channel matrix and is decomposed, and obtains matrix R ^(s):

H ^(s)＝IR ^(s)

Wherein I is a unit matrix, R ^(s)A upper triangular matrix,

Step (5.7), by following formula, data are carried out the QR Data Detection:

Y ' wherein ^(s)=(I) ^Hy ^(s),

With

Respectively the detected value of data and the result after the quantification of detected value planisphere, [] _{M, k}Represent the capable k row of m of matrix, [] _mM element of vector, [] _kK element of vector, O () expression demodulation computing, m, k representing matrix H ^(s)The value of the capable k of m row, m=1,2 ..., m ..., M, M≤N, k=1,2 ..., k ..., K, K≤N,

Step (6), iterations i=i+1, iterative computation number of times i〉1 o'clock, iteration is done channel estimating with the data that the subcarrier at place, a pilot tone point place subcarrier place adjacent with its both sides receives for the second time, data on all the other subcarriers are considered as ICI, the data that are used for calculating of the each increase of iteration thereafter, the data that the subcarrier place of the subcarrier both sides at the data place that is that last iteration be used for to calculate receives, data on all the other subcarriers are considered as ICI, and it is as described below that the SIN method is carried out the iterative channel estimation computing:

Step (6.1), to receive signal and be divided into pilot tone point place subcarrier data, the subcarrier data at the data place of be used for calculating is used for the data calculated to pilot tone point and is used for interference and noise four parts of the data places subcarrier of calculating, is shown below:

In following formula, the 3rd is the ICI interference after upgrading,

Step (6.2), the method described in step (5.2) is calculated Φ ^SIN _(s),

Step (6.3) increases with iterations, and the data that are used for iteration increase, and upgrade this moment

Covariance matrix U _ICIBe calculated as follows:

$U_{\mathrm{ICI}}\&ap;4{\mathrm{\&pi;}}^2{{T_s}^{2}E}_s\left(\underset{l=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_{h(n,l)}^2{\mathrm{\&sigma;}}_{D_l}\right)[\frac{1}{2}\mathrm{\&rho;}(A+m,\mathrm{rag},N)+\frac{1}{2}\mathrm{\&rho;}(A-m,\mathrm{rag},N)]$

Wherein, what A+m represented pilot tone point right side is used for the data calculated to the distance of corresponding pilot tone point, and A-m represents that the data that are used for calculating in pilot tone point left side arrive the distance of the pilot tone point of correspondence,

Step (6.4) is calculated kalman gain K by the described method of step (5.5) _(s), s OFDM symbol transferred to state

The state estimation matrix With with

Corresponding covariance matrix P _(s|s), consist of observation renewal equation group, wherein, Φ=Φ ^SIN,

K _(s)＝P _(s)(Φ _(s)) ^H(Φ _(s)P _(s)(Φ _(s))H+V[W _(s)]) ^-1，

${\hat{g}}_{\left(s\right|s)}={\hat{g}}_{\left(s\right)}+K_{\left(s\right)}(y_{\left(s\right)}-{\mathrm{\&Phi;}}_{\left(s\right)}{\hat{g}}_{\left(s\right)}),$

P _(s|s)＝P _(s)-K _(s)Φ _(s)P _(s)，

Step (6.5) calculates the estimated value H of channel matrix by the described method of step (5.5) ^(s):

$H^{\left(s\right)}=\frac{1}{N}\underset{b=1}{\overset{B}{\mathrm{\&Sigma;}}}{D}_b\mathrm{diag}\left(\mathrm{\&Gamma;}{\hat{g}}_{\left(s\right|s)}\right),$

Step (6.6) is carried out the QR decomposition by the described method of step (5.6) to channel matrix and is obtained R ^(s):

H ^(s)＝IR ^(s)，

Step (6.7), by the described method of step (5.7), data are carried out the QR Data Detection:

Step (7) judges whether whether all data all have been used for iteration, and if so, algorithm finishes, if not, continue,

Step (8) by the number of times of judgement iteration, determines whether to need to increase the input data of iterative algorithm, and evaluation algorithm is as follows:

Set step delta,

Compare iterations i and Δ μ+2, wherein, $\mathrm{\&mu;}=\mathrm{1,2}\&CenterDot;\&CenterDot;\&CenterDot;,\mathrm{\&Delta;}\&CenterDot;\mathrm{\&mu;}+2<\frac{N}{N_p},$ If i=Δ μ+2, the data that the data at the subcarrier place of the subcarrier both sides at the data place of selecting to be used for to calculate were calculated as next iteration being used for of newly adding, bring the SIN algorithm into together with the data that are used for calculating, be brought in step (6) and again count;

If i ≠ Δ μ+2 do not increase for the data of calculating, return to step (6) and carry out iteration;

Finish.

Claims (1)

1. the progressive iteration time varying channel that becomes soon ofdm system is estimated and the ICI removing method, it is characterized in that, in the channel estimating of ofdm system, with ICI and the noise sum SIN denoising object as the Kalman filter, from only using the pilot tone dot information, increase progressively the information that is used for iterative computation, realize according to following steps:

Step (1), transmitting terminal produce and send data, pilot data are inserted into according to the Comb Pilot mode send in data:

Transmitting terminal is set as follows: s represents s OFDM symbol, s=1, and 2 ..., s ... S, each OFDM symbol comprises N subcarrier, n=1,2 ..., n ..., N wherein comprises N _pIndividual frequency pilot sign

And N _dIndividual data symbol, N _d+ N _p=N, n _p=1,2 ..., N _p, the location matrix of pilot tone on frequency domain is expressed as:

Wherein

And guarantee N _p〉=L, L are the maximums that channel multi-path is counted l, i.e. l=1, and 2 ...,, l ..., L, N _pIndividual pilot tone is inserted among N carrier wave and remains unchanged in transmitting procedure by average, and the pilot tone point symbol is expressed as in N carrier wave ${x_p}^{\left(s\right)}=x^{\left(s\right)}\left(P_s\right)={[{x_{p_1}}^{\left(s\right)},{x_{p_2}}^{\left(s\right)},\&CenterDot;\&CenterDot;\&CenterDot;,{x_{{p_N}_p}}^{\left(s\right)}]}^T,$

Step (2), data are sent to receiving terminal by ofdm system, after receiving terminal removes Cyclic Prefix, with polynomial basis extended model P-BEM, channel are carried out modeling according to the following steps:

Step (2.1) utilizes polynomial basis extended model P-BEM to describe to have the two time dispersive channels that select characteristic of time-frequency, the channel impulse response h in n subcarrier l footpath of S OFDM symbol ^(s)(n, l) is expressed as:

h ^(s)(n,l)＝QC _l ^(s)+ξ _l ^(s)(n),0≤n≤N-1，

Wherein, ξ _l ^(s)The model error in expression l footpath of each OFDM symbol during modeling, it is worth less than 10 ^-3, ignore when calculating, namely think h ^(s)(n, l)=QC _l ^(s), Q is the orthogonal basis function matrix of a N * B, C _l ^(s)By B coefficient corresponding to basic function

The vector that forms ${C_l}^{\left(s\right)}={[c_{1,l}^{\left(s\right)},c_{2,l}^{\left(s\right)},\&CenterDot;\&CenterDot;\&CenterDot;,c_{b,l}^{\left(s\right)},\&CenterDot;\&CenterDot;\&CenterDot;,c_{B,l}^{\left(s\right)}]}^T,$

f _maxThe highest frequency of channel, T _sThe sampling time,

Step (2.2) will be shown as following form at the reception signal list of receiving terminal:

y ^(s)＝H ^(s)x ^(s)+W ^(s)，

Wherein, x ^(s)=[x ₁ ^(s), x ₂ ^(s)X _N ^(s)] ^T, y ^(s)=[y ₁ ^(s), y ₂ ^(s)..., y _N ^(s)] ^TRepresent that respectively s on frequency domain is removed the transmitted signal after Cyclic Prefix and receive signal, W ^(s)The white noise on its frequency domain, H ^(s)The channel matrix of N * N:

Wherein, each element of matrix be multipath channel channel impulse response and, account form is as follows:

$H^{\left(s\right)}(m,k)=\underset{l=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{G_l}^{\left(s\right)}(M,K)e^{-j2\mathrm{\&pi;}(\frac{k-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l},$

M, k represent above-mentioned matrix H ^(s)The value of the capable k of m row, m=1,2 ..., m ..., M, M≤N, k=1,2 ..., k ..., K, K≤N, τ _lThe time delay in l footpath, G _l ^(s)(M, K) impacts corresponding frequency domain presentation matrix for channel, and its each element is calculated as follows:

${G_l}^{\left(s\right)}(m,k)=\frac{1}{N}\underset{n=0}{\overset{N-1}{\mathrm{\&Sigma;}}}{h}^{\left(s\right)}(n,l)e^{-j2\mathrm{\&pi;}(m-k)n/N},$

Step (2.3) will receive signal according to the P-BEM model and carry out modeling again, be expressed as with the expression formula of P-BEM coefficient as follows:

y ^(s)＝Φ ^(s)g ^(s)+W ^(s)，

Wherein,

g ^(s)=[C ₁ ^{(s) T}, C ₂ ^{(s) T}C _L ^{(s) T}] ^T, the coefficient matrix in expression P-BEM algorithm,

After representing modeling again, the coefficient matrix relevant to sending data, its computational methods are as follows:

$Z_l^{\left(s\right)}=\frac{1}{N}[D_1\mathrm{diag}\left(x^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l,\&CenterDot;\&CenterDot;\&CenterDot;,D_b\mathrm{diag}\left(x^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l,\&CenterDot;\&CenterDot;\&CenterDot;,D_B\mathrm{diag}\left(x^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l],$ Wherein,

${\mathrm{\&Gamma;}}_l={\left[\begin{array}{cccc}{e}^{-j2\mathrm{\&pi;}(\frac{p_1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}& e^{-j2\mathrm{\&pi;}(\frac{p_2-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}& \&CenterDot;\&CenterDot;\&CenterDot;& e^{-j2\mathrm{\&pi;}(\frac{{p_N}_p-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}\end{array}\right]}^T,$ The Fourier transform in l footpath,

Γ=[Γ ₁, Γ ₂..., Γ _L], total Fourier transform matrix in expression L footpath,

Diag (x ^(s)) represent with vector x ^(s)Be the matrix of diagonal element,

Step (3), utilize the AR model to carry out modeling to channel BEM coefficient:

Step (3.1) is calculated as follows C _l ^(s)Correlation matrix:

$R_{C_l}^{\left(j\right)}={\left({\left(Q\right)}^{H}Q\right)}^{-1}{\left(Q\right)}^H{R}_{h(n,l)}^{\left(j\right)}Q{\left({\left(Q\right)}^{H}Q\right)}^{-1},$

Wherein, j represents the exponent number of being correlated with, and namely carries out the mark space of the OFDM symbol of related operation, and the value of j is [1,0,1],

The C that represents respectively current ofdm signal _l ^(s)C with previous symbol _l ^(s-1)Correlation matrix, the C of current ofdm signal _l ^(s)Autocorrelation matrix, the C of current ofdm signal _l ^(s)C with a rear symbol _l ^(s+1)Correlation matrix.() ^HExpression Hermitian computing, $R_{h(n,l)}^{\left(j\right)}=E\left[h(n,l)h^{*}(n+j,l)\right]={\mathrm{\&sigma;}}_{h(n,l)}^2{J}_0\left(2\mathrm{\&pi;}{f}_d{T}_{s}j\right),$ E[wherein] the expression average, J ₀The zero Bessel function of () expression first kind, f _d=vf _c/ c is the maximum doppler frequency of the translational speed of terminal when being v, f _cBe carrier frequency, c is the light velocity,

Represent the variance of the channel impulse response in l footpath, and hypothesis

Step (3.2) obtains the state transition equation of channel P-BEM parameter according to the Yule-Walker equation:

g _(s)＝Ag _(s-1)+U _(s)，

Ofdm system is sent the time sequencing g of symbol ^(s)Regard state transitions process g in control system as _(s), i.e. g ^(s)=g _(s), state transition equation coefficient A=diag (a ₁, a ₂..., a _l... a _L),

The matrix of diag (x) expression take vector x as diagonal element, U _(s)Represent the modeling error of the AR model of s OFDM symbol;

Step (4), the Kalman filter is carried out initialization and calculates initial renewal equation:

Step (4.1), according to the following formula the Kalman filter is carried out initialization:

${\hat{g}}_{\left(0\right|0)}=O_{\mathrm{LB},1}{P}_{\left(0\right|0)}=\mathrm{diag}(R_{C_1}^{\left(0\right)},R_{C_2}^{\left(0\right)}\&CenterDot;\&CenterDot;\&CenterDot;R_{C_L}^{\left(0\right)}),$

Form as And P _(s|s)The previous s of middle subscript represents that all current state is g _(s), a rear s represents s OFDM symbol,

P _(0|0)For the initial value that calculates,

The g of expression OFDM symbol _(s)Initial value, P _(0|0)Expression

Corresponding error correlation matrix, O _{LB, 1}The null matrix of LB * 1,

Step (4.2) is calculated as follows the initial time renewal equation of Kalman:

i＝1，s＝1，

${\hat{g}}_{\left(s\right)}=A{\hat{g}}_{\left(0\right|0)},$

P _(s)＝AP _(0|0)(A) ^H+V[U _(s)]，

I represents iterations,

State estimation g in expression Kalman equation _(s)Intermediate variable, P _(s)The expression intermediate variable

Corresponding error correlation matrix; Use V[] the expression covariance matrix, V[U _(s)]=diag (u ₁, u ₂U _L)

Step (5), carry out iterative channel estimation computing for the first time, this moment iterations i=1, only use the subcarrier place at pilot tone point place to receive data in this iteration and do channel estimating, data except the subcarrier of pilot tone point place on other subcarriers are considered as ICI, eliminate unknown data to the impact of pilot tone place channel estimating with the SIN method, realize the auxiliary Kalman channel estimating of the pilot tone of disturbing without ICI, concrete steps are as follows:

Step (5.1), the reception signal that only will receive the carrier position that in signal, each pilot tone point is corresponding is used for calculating, and will receive signal and be divided into data on the subcarrier of each pilot tone point place, interference and noise three parts of data on other subcarrier except pilot tone point subcarrier to each pilot tone point place subcarrier are shown below:

${y_p}^{\left(s\right)}=y^{\left(s\right)}\left(P_s\right)={H^{\left(s\right)}}_{[P_s,P_s]}{{{x_p}_n}_p}^{\left(s\right)}+{H^{\left(s\right)}}_{[P_s,d_{n_p{n}^{\&prime;}}]}{x_{d_{n_p{n}^{\&prime;}}}}^{\left(s\right)}+W^{\left(s\right)}\left(P_s\right)$

Wherein, $P_s=[p_1,p_2,\&CenterDot;\&CenterDot;\&CenterDot;,p_{N_p}],$ ${x_{d_{n_p{n}^{\&prime;}}}}^{\left(s\right)}={[x^{\left(s\right)}\left(d_{11}\right),x^{\left(s\right)}\left(d_{12}\right),\&CenterDot;\&CenterDot;\&CenterDot;{,x}^{\left(s\right)}\left(d_{{1N}^{\&prime;}}\right),x^{\left(s\right)}\left(d_{21}\right),\&CenterDot;\&CenterDot;\&CenterDot;,x^{\left(s\right)}\left(d_{N_p{N}^{\&prime;}}\right)]}^T,$

N'=1,2 ..., N' represents the distance between each adjacent pilot tone point,

A N _p* N _pUnit matrix, σ ²White Gaussian noise W ^(s)Variance, in following formula second be data on other subcarrier except pilot tone point subcarrier to the interference ICI of pilot tone point place subcarrier,

Step (5.2) is considered to interchannel noise W with data I CI distracter ^(s)(P _s) a part as the denoising object of filter, the algorithm in step (2) is rewritten order according to the method that SIN estimates

The Kalman observational equation of SIN estimation is expressed as:

$y_{p\left(s\right)}={{\mathrm{\&Phi;}}^{\mathrm{SIN}}}_{\left(s\right)}{g}_{\left(s\right)}+W_{\left(s\right)}^{\mathrm{SIN}},$

Wherein:

${{\mathrm{\&Phi;}}^{\mathrm{SIN}}}_{\left(s\right)}=\frac{1}{N}[{z_1^{\left(s\right)}}_{\mathrm{SIN}},{Z_2^{\left(s\right)}}_{\mathrm{SIN}}\&CenterDot;\&CenterDot;\&CenterDot;{Z_L^{\left(s\right)}}_{\mathrm{SIN}}],$

${Z_l^{\left(s\right)}}_{\mathrm{SIN}}=\frac{1}{N}[D_1^{\mathrm{SIN}}\mathrm{diag}\left({x_p}^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l^{\mathrm{SIN}}\&CenterDot;\&CenterDot;\&CenterDot;D_B^{\mathrm{SIN}}\mathrm{diag}\left({x_p}^{\left(s\right)}\right){\mathrm{\&Gamma;}}_l^{\mathrm{SIN}}],$

${\mathrm{\&Gamma;}}_l^{\mathrm{SIN}}={\left[\begin{array}{cccc}{e}^{-j2\mathrm{\&pi;}(\frac{p_1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}& e^{-j2\mathrm{\&pi;}(\frac{p_2-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}& \&CenterDot;\&CenterDot;\&CenterDot;& e^{-j2\mathrm{\&pi;}(\frac{{p_N}_p-1}{N}-\frac{1}{2}){\mathrm{\&tau;}}_l}\end{array}\right]}^T,$

Γ ^SIN＝[Γ ₁ ^SIN,Γ ₂, ^SIN…,Γ _l ^SIN]

Step (5.3) is calculated

Covariance matrix

In calculating, hypothesis ICI is white Gaussian noise, order

Because noise and ICI are both separate, so $V\left[W_{\left(s\right)}^{\mathrm{SIN}}\right]=U_{\mathrm{ICI}}+V\left[W^{\left(s\right)}\left(P_s\right)\right];$

U _ICIThe calculating formula of each element in matrix is:

$U_{\mathrm{ICI}}(m,k)=R_{\mathrm{ICI}}^{\left(j\right)}\&ap;4{\mathrm{\&pi;}}^2{T}_s^2{E}_s\left(\underset{l=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_{h(n,l)}^2{\mathrm{\&sigma;}}_{D_l}\right)\mathrm{\&rho;}(\mathrm{\&alpha;},\mathrm{rag},N),$

Wherein, m, the capable k row of the m of k representing matrix, E _sThe power that sends data,

That power is P _vThe time the general function of Doppler's power, f is transmission frequency, it is 0 when calculating for the first time for the marginal date of the iteration distance apart from each pilot tone point of correspondence that α represents, the precision of rag for calculating, rag=[0,1,2,3], and:

ρ(α，rag，N)=ρ(0,rag,N)-ρ ₁(α，rag,N)

Step (5.4) is calculated kalman gain K by following three formulas respectively _(s), s OFDM symbol transferred to state

The state estimation matrix

With with

Corresponding covariance matrix P _(s|s), consist of observation renewal equation group, wherein, Φ=Φ ^SIN $W_{\left(s\right)}=W_{\left(s\right)}^{\mathrm{SIN}},$

K _(s)＝P _(s)(Φ _(s)) ^H(Φ _(s)P _(s)(Φ _(s)) ^H+V[W _(s)]) ^-1，

${\hat{g}}_{\left(s\right|s)}={\hat{g}}_{\left(s\right)}+K_{\left(s\right)}(y_{\left(s\right)}-{\mathrm{\&Phi;}}_{\left(s\right)}{\hat{g}}_{\left(s\right)}),$

P _(s|s)＝P _(s)-K _(s)Φ _(s)P _(s)，

Step (5.5) calculates channel matrix H according to following formula ^(s)Estimated value:

$H^{\left(s\right)}=\frac{1}{N}\underset{b=1}{\overset{B}{\mathrm{\&Sigma;}}}{D}_b\mathrm{diag}\left(\mathrm{\&Gamma;}{\hat{g}}_{\left(s\right|s)}\right),$

Step (5.6) is utilized following formula to carry out QR to channel matrix and is decomposed, and obtains matrix R ^(s):

H ^(s)＝IR ^(s)

Wherein I is a unit matrix, R ^(s)A upper triangular matrix,

Step (5.7), by following formula, data are carried out the QR Data Detection:

Y ' wherein ^(s)=(I) ^Hy ^(s),

With

Respectively the detected value of data and the result after the quantification of detected value planisphere, [] _{M, k}Represent the capable k row of m of matrix, [] _mM element of vector, [] _kK element of vector, O () expression demodulation computing, m, k representing matrix H ^(s)The value of the capable k of m row, m=1,2 ..., m ..., M, M≤N, k=1,2 ..., k ..., K, K≤N,

Step (6), iterations i=i+1, iterative computation number of times i〉1 o'clock, iteration is done channel estimating with the data that the subcarrier at place, a pilot tone point place subcarrier place adjacent with its both sides receives for the second time, data on all the other subcarriers are considered as ICI, the data that are used for calculating of the each increase of iteration thereafter, the data that the subcarrier place of the subcarrier both sides at the data place that is that last iteration be used for to calculate receives, data on all the other subcarriers are considered as ICI, and it is as described below that the SIN method is carried out the iterative channel estimation computing:

Step (6.1) will receive signal and be divided into pilot tone point place subcarrier data, and the subcarrier data at the data place that is used for calculating is used for the data calculated to pilot tone point and is used for interference and noise four parts of the data places subcarrier of calculating,

Be shown below:

In following formula, the 3rd is the ICI interference after upgrading,

Step (6.2), the method described in step (5.2) is calculated Φ ^SIN _(s),

Step (6.3) increases with iterations, and the data that are used for iteration increase, and upgrade this moment

Covariance matrix U _ICIBe calculated as follows:

$U_{\mathrm{ICI}}\&ap;4{\mathrm{\&pi;}}^2{{T_s}^{2}E}_s\left(\underset{l=0}{\overset{L-1}{\mathrm{\&Sigma;}}}{\mathrm{\&sigma;}}_{h(n,l)}^2{\mathrm{\&sigma;}}_{D_l}\right)[\frac{1}{2}\mathrm{\&rho;}(A+m,\mathrm{rag},N)+\frac{1}{2}\mathrm{\&rho;}(A-m,\mathrm{rag},N)]$

Wherein, what A+m represented pilot tone point right side is used for the data calculated to the distance of corresponding pilot tone point, and A-m represents that the data that are used for calculating in pilot tone point left side arrive the distance of the pilot tone point of correspondence,

Step (6.4) is calculated kalman gain K by the described method of step (5.5) _(s), s OFDM symbol transferred to state

The state estimation matrix

With with

Corresponding covariance matrix P _(s|s), consist of observation renewal equation group, wherein, Φ=Φ ^SIN, $W_{\left(s\right)}=W_{\left(s\right)}^{\mathrm{SIN}},$

K _(s)＝P _(s)(Φ _(s)) ^H(Φ _(s)P _(s)(Φ _(s)) ^H+V[W _(s)]) ^-1，

${\hat{g}}_{\left(s\right|s)}={\hat{g}}_{\left(s\right)}+K_{\left(s\right)}(y_{\left(s\right)}-{\mathrm{\&Phi;}}_{\left(s\right)}{\hat{g}}_{\left(s\right)}),$

P _(s|s)＝P _(s)-K _(s)Φ _(s)P _(s)，

Step (6.5) calculates the estimated value H of channel matrix by the described method of step (5.5) ^(s):

$H^{\left(s\right)}=\frac{1}{N}\underset{b=1}{\overset{B}{\mathrm{\&Sigma;}}}{D}_b\mathrm{diag}\left(\mathrm{\&Gamma;}{\hat{g}}_{\left(s\right|s)}\right),$

Step (6.6) is carried out the QR decomposition by the described method of step (5.6) to channel matrix and is obtained R ^(s):

H ^(s)＝IR ^(s)，

Step (6.7), by the described method of step (5.7), data are carried out the QR Data Detection:

Step (7) judges whether whether all data all have been used for iteration, and if so, algorithm finishes, if not, continue,

Step (8) by the number of times of judgement iteration, determines whether to need to increase the input data of iterative algorithm, evaluation algorithm

As follows:

Set step delta, Compare iterations i and Δ μ+2, wherein, $\mathrm{\&mu;}=\mathrm{1,2},\&CenterDot;\&CenterDot;\&CenterDot;,\mathrm{\&Delta;}\&CenterDot;\mathrm{\&mu;}+2<\frac{N}{N_p},$ If i=Δ μ+2, the data that the data at the subcarrier place of the subcarrier both sides at the data place of selecting to be used for to calculate were calculated as next iteration being used for of newly adding, bring the SIN algorithm into together with the data that are used for calculating, be brought in step (6) and again count;

If i ≠ Δ μ+2 do not increase for the data of calculating, return to step (6) and carry out iteration;

Finish.

Images