CN100493053C - Method for channel estimation in multi-antenna system - Google Patents

Abstract

A channel calculation method in a multi-antenna system is related to a pilot design and channel calculation method of a multi-antenna system under a double-selected decline channel environment. At the sending end, the discrete Fourier transformation matrix is used to set up the optimum circulation orthogonal pilot sequence of the least square to be inserted into the sending data interleave to constitute double circulate time slot structure, at the receiving end, the circulation orthogonal property is utilized to carry out the least square channel calculation of the least mean square error, the resolving of the received pilot matrix is used to realize the channel calculation quickly and discrete cosine transformation is used to carry out more accurate channel calculation and noise variance calculation of the pilot band to obtain the calculation of the data band channel parameters with the interpolates of its domain.

Description

The method of channel estimating in the multiaerial system

Technical field

The present invention relates to a kind of by using a plurality of send/receive antennas to come the wide-band mobile communication system of transmitting high speed data, relate in particular to a kind of multiaerial system under double selectivity fading channel environment pilot design and the method for channel estimating.

Background technology

The needs that be to adapt to future development, super 3 g mobile communication system must: support all-IP high-speed packet data transmission, data rate is tens of million bps (bit per second) even hundreds of million bps; Support high terminal mobility, translational speed is up to hundreds of kilometer per hour; Support high transmission quality, the error rate of data service is lower than 10 ^-6Provide the high availability of frequency spectrum, more than every hertz of number bit; High power efficiency is provided, and transmitting power reduces more than the 10dB; Be supported in the variation of aspect great dynamic ranges such as user data rate, user capacity, service quality and translational speed effectively.

In order to improve the availability of frequency spectrum of system, the air interface mechanism that adopts many antenna transmission and many antennas to receive is a kind of effective solution.Yet even under many antenna environment, for reliable and effective support high speed data transfer, super 3 g mobile communication system still needs very high bandwidth.Wideband transmit has increased the weight of the frequency selective fading phenomenon of channel, thereby causes serious multipath to disturb; Then increased the weight of the time selective fading of channel by the Doppler frequency shift phenomenon that high-speed mobile caused of terminal.Therefore, in super 3 g mobile communication system, the decline of channel is a double selectivity.

Receiver in the communication system is divided into two kinds of coherent receiver and noncoherent receivers.Coherent receiver need be at the impulse response coefficient of receiving terminal known channel, thereby need carry out channel estimating at receiving terminal; Noncoherent receiver then need be not the quadrature modulation mode but require to send signal, and have the loss of 3-4dB on performance at the impulse response coefficient of receiving terminal known channel.The present invention mainly considers prevailing coherent reception mode in super 3 g mobile communication system.

In order to realize coherent reception, need carry out channel estimating at receiving terminal.In order to estimate channel parameter timely and accurately, the normal channel estimation methods that adopts based on pilot frequency sequence of actual communication systems.Its basic thought is: intermittently insert pilot tone in the transmitting terminal appropriate location, receiving terminal utilizes pilot tone to recover the channel information of pilot frequency locations, utilizes certain processing means (as interpolation, filtering, conversion etc.) to obtain the channel information of all periods then.Here relate generally to three problems: the selection and the insertion of (1) transmitting terminal pilot tone; (2) mode obtained of receiving terminal pilot frequency locations channel information; (3) channel information that obtains by pilot frequency locations recovers all information of channel constantly.The present invention has mainly provided a kind of optimal performance and low technical scheme of implementation complexity of approaching with regard to these three problems.

Summary of the invention

Technical problem: the method that the purpose of this invention is to provide channel estimating in a kind of multiaerial system, this method can improve precision of channel estimation effectively, and the performance of improving receiver is the performance of receiver under the conventional channel method of estimation high speed that is difficult to guarantee and the speed change situation of movement particularly.

Technical scheme: the method for channel estimating in the multiaerial system of the present invention, at transmitting terminal, utilize discrete Fourier transform (DFT) (DFT) matrix construction to go out cyclic orthogonal experiment pilot frequency sequence optimum on least square (LS) meaning, and it is inserted the transmission data off and on to form bicirculating structure of time slot; At receiving terminal, utilize the characteristic of cyclic orthogonal experiment sequence, carry out least square channel estimating optimum on least mean-square error (MMSE) meaning with low implementation complexity, utilize the decomposition that receives pilot matrix, carry out the quick realization of channel estimating, utilize discrete cosine transform (DCT) to carry out pilot more precise channels estimation and Noise Variance Estimation, adopt discrete cosine transform (DCT) interpolation to obtain the estimation of data segment channel parameter again

This method comprises following step:

Step 1), at transmitting terminal, according to the number NT of multiaerial system transmitting antenna and the multipath number P of channel, construct length and be

Optimum cyclic orthogonal experiment pilot frequency sequence s on the least square meaning, and generate the pilot frequency sequence of each transmitting antenna according to the following rules:

$s_n\left(l\right)=s_{{\left((l-\mathrm{nP})\right)}_{L_P}},(n=\mathrm{0,1},...,N_T-1,l=\mathrm{0,1},...,L_P-1);$

Step 2), at receiving terminal, try to achieve the channel impulse response parameter Estimation of each pilot in the time slot by following formula

${\hat{H}}^{\left(k\right)}=\frac{1}{L_P}{Y}^{\left(k\right)}{X}^H,(k=\mathrm{0,1},...,K),$

$X=P_{\mathrm{\μ}}^H\left[\begin{array}{cc}{I}_{N_{T}P}& 0_{N_{T}P\×(L_P-N_{T}P)}\end{array}\right]P_{\mathrm{\α}}^H(I_Q\⊗W)P_{\mathrm{\β}}^H\mathrm{\Λ}(I_Q\⊗W^H)P_{\mathrm{\α}};$

Step 3), at receiving terminal, utilize step 2) channel impulse response that estimates estimates an interchannel noise variance by following formula to each pilot, and obtains the Noise Variance Estimation of current time slots:

${\widehat{\mathrm{\σ}}}_k^2=\frac{1}{N_R(L_P-N_{T}P)}{\left|\right|Y^{\left(k\right)}-{\hat{H}}^{\left(k\right)}X\left|\right|}_F^2,(k=\mathrm{0,1},...,K)$

${\widehat{\mathrm{\σ}}}^2=\frac{1}{K+1}\underset{k=0}{\overset{K}{\mathrm{\Σ}}}{\widehat{\mathrm{\σ}}}_k^2;$

Step 4), at receiving terminal, utilize step 2) channel impulse response of all pilots of estimating, it is carried out the denoising and the Noise Variance Estimation of pointwise in the DCT territory, and by obtain the channel impulse response of data segment in discrete cosine transform DCT territory interpolation

Cyclic orthogonal experiment pilot frequency sequence s is constructed by fourier transform matrix in the described method, and satisfies the cyclic orthogonal experiment characteristic; The pilot frequency sequence of each transmitting antenna is obtained by the s cyclic shift; Described least square is meant the quadratic sum minimum of evaluated error.The estimating step 2 of described channel impulse response parameter is carried out sub-slots, utilize to receive the following decomposition of pilot matrix, the channel impulse response parameter estimate at quick implementation algorithm;

Cyclic orthogonal experiment pilot frequency sequence s is constructed by fourier transform matrix in the described method, and satisfies the cyclic orthogonal experiment characteristic; The pilot frequency sequence of each transmitting antenna is obtained by the s cyclic shift; Described least square is meant the quadratic sum minimum of evaluated error.The estimation of the channel impulse response parameter described in the step 2 is carried out sub-slots; Utilize to receive the following decomposition of pilot matrix, the channel impulse response parameter estimate at quick implementation algorithm;

$X=P_{\mathrm{\μ}}^H\left[\begin{array}{cc}{I}_{N_{T}P}& 0_{N_{T}P\×(L_P-N_{T}P)}\end{array}\right]P_{\mathrm{\α}}^H(I_Q\⊗W)P_{\mathrm{\β}}^H\mathrm{\Λ}(I_Q\⊗W^H)P_{\mathrm{\α}}.$

The estimating step 4 of described channel impulse response parameter is carried out in a time slot; The denoising of channel impulse response parameter and interpolation are all carried out in discrete cosine transform DCT territory.

Beneficial effect: the invention provides a kind of building method and channel estimation methods that can be used for the pilot frequency sequence of multi-aerial transmission system channel estimating, according to the pilot frequency sequence that the inventive method generates, can realize channel estimating optimum on the least square meaning with lower computational complexity; Utilize the time domain correlation properties of channel simultaneously,, obtain the approximate optimal solution on the MMSE meaning of low complex degree, further improved channel estimated accuracy by the processing of transform domain.Channel estimation methods provided by the invention compared with prior art can improve precision of channel estimation effectively, and the performance of improving receiver is the performance of receiver under the conventional channel method of estimation high speed that is difficult to guarantee and the speed change situation of movement particularly.This channel estimation methods need not long pilot frequency sequence, and operand and memory space are all very little, are convenient to hardware and realize.

Description of drawings

Fig. 1 is the pilot time slot structure at intermittence that adopts among the present invention.Recycling-guard section G, pilot P, data segment D are wherein arranged.Sub-slots number K adjusts according to the translational speed self adaptation of terminal.

Fig. 2 is the application process schematic diagram of being constructed each antenna pilot sequences among the present invention by basic sequence.

Fig. 3 is the schematic diagram of a kind of quick implement device of least square channel estimating among the present invention.The accent preface module that wherein has pair signal sequence to adjust; The phase place rotary module that signal phase is rotated; FFT group and IFFT group.

Fig. 4 is that a kind of channel estimating of the present invention is specifically installed block diagram.Least square channel estimation module at each pilot received signal is wherein arranged; All time domain estimated values are carried out the DCT module of discrete cosine transform; Single-point denoising of DCT territory and Noise Variance Estimation module; Signal is carried out the IDCT module of inverse discrete cosine transformation.

Embodiment

For making the purpose, technical solutions and advantages of the present invention clearer, be described in further detail below in conjunction with the enforcement of accompanying drawing to technical scheme:

At transmitting terminal, utilize discrete Fourier transform (DFT) (DFT) matrix construction to go out cyclic orthogonal experiment pilot frequency sequence optimum on least square (LS) meaning, and it is inserted the transmission data off and on to form bicirculating structure of time slot; At receiving terminal, utilize the characteristic of cyclic orthogonal experiment sequence, carry out least square channel estimating optimum on least mean-square error (MMSE) meaning with low implementation complexity, utilize the decomposition that receives pilot matrix, carry out the quick realization of channel estimating, utilize discrete cosine transform (DCT) to carry out pilot more precise channels estimation and Noise Variance Estimation, adopt discrete cosine transform (DCT) interpolation to obtain the estimation of data segment channel parameter again.

This method comprises following step:

Step 1), at transmitting terminal, according to the number NT of multiaerial system transmitting antenna and the multipath number P of channel, construct length and be

Optimum cyclic orthogonal experiment pilot frequency sequence s on the least square meaning, and generate the pilot frequency sequence of each transmitting antenna according to the following rules:

$s_n\left(l\right)=s_{{\left((l-\mathrm{nP})\right)}_{L_P}},(n=\mathrm{0,1},...,N_T-1,l=\mathrm{0,1},...,L_P-1);$

Step 2), at receiving terminal, try to achieve the channel impulse response parameter Estimation of each pilot in the time slot by following formula

${\hat{H}}^{\left(k\right)}=\frac{1}{L_P}{Y}^{\left(k\right)}{X}^H,(k=\mathrm{0,1},...,K),$

$X=P_{\mathrm{\μ}}^H\left[\begin{array}{cc}{I}_{N_{T}P}& 0_{N_{T}P\×(L_P-N_{T}P)}\end{array}\right]P_{\mathrm{\α}}^H(I_Q\⊗W)P_{\mathrm{\β}}^H\mathrm{\Λ}(I_Q\⊗W^H)P_{\mathrm{\α}};$

Step 3), at receiving terminal, utilize step 2) channel impulse response that estimates estimates an interchannel noise variance by following formula to each pilot, and obtains the Noise Variance Estimation of current time slots:

${\widehat{\mathrm{\σ}}}_k^2=\frac{1}{N_R(L_P-N_{T}P)}{\left|\right|Y^{\left(k\right)}-{\hat{H}}^{\left(k\right)}X\left|\right|}_F^2,(k=\mathrm{0,1},...,K)$

—4—

${\widehat{\mathrm{\σ}}}^2=\frac{1}{K+1}\underset{k=0}{\overset{K}{\mathrm{\Σ}}}{\widehat{\mathrm{\σ}}}_k^2;$

Step 4), at receiving terminal, utilize step 2) channel impulse response of all pilots of estimating, it is carried out the denoising and the Noise Variance Estimation of pointwise in the DCT territory, and by obtain the channel impulse response of data segment in DCT territory interpolation

Cyclic orthogonal experiment pilot frequency sequence s is constructed by fourier transform matrix in the described method, and satisfies the cyclic orthogonal experiment characteristic; The pilot frequency sequence of each transmitting antenna is obtained by the s cyclic shift; Described least square is meant the quadratic sum minimum of evaluated error.The estimating step 2 of described channel impulse response parameter is carried out sub-slots, utilize to receive the following decomposition of pilot matrix, the channel impulse response parameter estimate at quick implementation algorithm; The estimating step 4 of described channel impulse response parameter is carried out in a time slot; The denoising of channel impulse response parameter and interpolation are all carried out in the DCT territory.

1. system model

Fig. 1 has provided transmission signal break pilot time slot structure.If K sub-slots arranged in each time slot, then the number of its pilot is K+1.Sub-slots is counted K and can be set according to the translational speed of terminal.

In mimo system, the number of establishing transmitting antenna is N _T, the number of reception antenna is N _R, the length of channel impulse response sequence is P, then each receive path treats that the estimated channel number of parameters is N _T* P, correspondingly, pilot sequence length (wherein

Expression is not less than the smallest positive integral of x).

Use s _n=[s _n(0), s _n(1) ..., s _n(L _P-1)] ^T, (n=0,1 ..., N _T-1) pilot frequency sequence of expression n root transmitting antenna, then $S={[{\mathrm{<figure-callout\; id="0"\; label="s"\; filenames="C200510039284E00134.png"\; state="\{\{state\}\}">s</figure-callout>}}_0,s_1,...,s_{N_T-1}]}^T$ The pilot signal transmitted of representing all antennas.The received signal of k pilot after receiving terminal removes recycling-guard can be expressed as:

${\stackrel{\&OverBar;}{Y}}^{\left(k\right)}=\underset{p=0}{\overset{P-1}{\mathrm{\&Sigma;}}}{H}_p^{\left(k\right)}{S}_p+Z^{\left(k\right)},(k=\mathrm{0,1},...,K)$ [formula 1]

Y wherein ^(k)And Z ^(k)Be N _R* L _PMatrix, the pilot signal and the variance that receive of expression is σ respectively ²Additive white Gaussian noise; $H_p^{\left(k\right)}=\left[h_{m,n}^{\left(k\right)}\left(p\right)\right]$ Be N _R* N _TMatrix,

Be illustrated in the channel tap coefficient in p footpath between k pilot, m root reception antenna and the n root transmitting antenna; $S_p=S\left[\begin{array}{cc}0& I_{L_p-\mathrm{<figure-callout\; id="0"\; label="p\; I\; p"\; filenames="C200510039284E00134.png"\; state="\{\{state\}\}">p</figure-callout>}}& \\ I_p{}_{}& 0\end{array}\right]$ Be that row ring shift right p position by S obtains.

Order $H^{\left(k\right)}=[H_0^{\left(k\right)},H_1^{\left(k\right)},...,H_{P-1}^{\left(k\right)}]$ Be illustrated in all channel parameters of k pilot, $X={[S_0^T,S_1^T,...,S_{P-1}^T]}^T,$ Then formula 1 can be rewritten as:

Y ^(k)=H ^(k)X+Z ^(k)[formula 2]

The least square (LS) that is obtained channel parameter by formula 2 described linear models is estimated as:

${\hat{H}}^{\left(k\right)}=Y^{\left(k\right)}{X}^H{\left({\mathrm{XX}}^H\right)}^{-1}$ [formula 3]

Work as L _PN _TDuring P, the unbiased estimator that can obtain noise variance is:

${\widehat{\mathrm{\&sigma;}}}_k^2=\frac{1}{N_R(L_P-N_{T}P)}{\left|\right|Y^{\left(k\right)}-{\hat{H}}^{\left(k\right)}X\left|\right|}_F^2$ [formula 4]

‖ ‖ wherein _FThe F norm of representing matrix.Theory analysis shows, when ${\mathrm{XX}}^H=L_P{I}_{N_{T}P}$ The time, top LS estimates to have optimum performance, and has avoided matrix inversion operation.

2. optimum pilot frequency sequence structure

If $s={[{\mathrm{<figure-callout\; id="0"\; label="s"\; filenames="C200510039284E00134.png"\; state="\{\{state\}\}">s</figure-callout>}}_0,s_1,...,s_{L_P-1}]}^T.$ Be the long L that is _PThe cyclic orthogonal experiment sequence, we as basic sequence, construct the pilot frequency sequence of each antenna with it by following criterion

$s_n\left(l\right)=s_{{\left((l-\mathrm{nP})\right)}_{L_P},(n=\mathrm{0,1},...,N_T-1,l=\mathrm{0,1},...,L_P-1)}$ [formula 5]

Wherein ((n)) _NExpression n is to the modular arithmetic of asking of N.Can verify that when s was the cyclic orthogonal experiment sequence, the pilot frequency sequence of constructing according to formula 5 satisfied the orthogonality condition of front.

The structure of relevant cyclic orthogonal experiment sequence, existing in the literature report, we find L _P=2 ²ⁿAnd L _P=2 ^2n-1The cyclic orthogonal experiment sequence can directly obtain from N=2 ＂ point DFT matrix element.If W is a N point DFT matrix, its element W _M.n=e ^{-J2 π mn/N}, and establish W=[W ₀W ₁], W wherein ₀And W ₁Be 2 ⁿ* 2 ^N-1Submatrix.If L _P=2 ²ⁿ, order $\tilde{W}=W,$ Then $s=\mathrm{vec}\left\{{\tilde{W}}^T\right\}$ Be that length is 2 ²ⁿThe cyclic orthogonal experiment sequence of vectors; If L _P=2 ^2n-1, order $\tilde{W}=({\mathrm{<figure-callout\; id="0"\; label="W"\; filenames="C200510039284E00134.png"\; state="\{\{state\}\}">W</figure-callout>}}_0+j{W}_1)/\sqrt{2},$ Then $s=\mathrm{vec}\left\{{\tilde{W}}^T\right\}$ Be that length is 2 ^2n-1The cyclic orthogonal experiment sequence of vectors.Here, vec{} is the stretching operator.

3. the implementation method of least square channel estimating

X satisfies orthogonality condition when matrix ${\mathrm{XX}}^H=L_P{I}_{N_{T}P}$ The time, formula 3 can be reduced to:

${\hat{H}}^{\left(k\right)}=\frac{1}{L_P}{Y}^{\left(k\right)}{X}^H$ [formula 6]

Receiving terminal is being received Y ^(k)Can directly utilize formula 6 to make channel estimating afterwards,, avoid complicated matrix inversion operation than the calculating of formula 3.Further study the structure of matrix X, we find that it has following two kinds of decomposed forms:

$X=P_{\mathrm{\&mu;}}^H\left[\begin{array}{cc}{I}_{N_{T}P}& 0_{N_{T}P\&times;(L_P-N_{T}P)}\end{array}\right]{\mathrm{F\&Gamma;F}}^H$ [formula 7]

$X=P_{\mathrm{\&mu;}}^H\left[\begin{array}{cc}{I}_{N_{T}P}& 0_{N_{T}P\&times;(L_P-N_{T}P)}\end{array}\right]P_{\mathrm{\&alpha;}}^H(I_Q\&CircleTimes;W)P_{\mathrm{\&beta;}}^H\mathrm{\&Lambda;}(I_Q\&CircleTimes;W^H)P_{\mathrm{\&alpha;}}$ [formula 8]

In the formula 7, F is L _PThe DFT matrix of point, $\mathrm{\&Gamma;}=\frac{1}{L_P}\mathrm{diag}\left\{\mathrm{Fa}\right\},$ P _μBe the N that generates by displacement μ _TP rank permutation matrix; In the formula 8, Λ is L _PThe rank diagonal matrix, P _α, P _βBe respectively by displacement α, the L that β generates _PThe rank permutation matrix, Q=L _P/ N.Wherein diag{d} represents that main diagonal element is the diagonal matrix of d,

The Kronecker product of representing matrix.The create-rule of above-mentioned each displacement is as follows:

$\mathrm{\&mu;}\left(k\right)=\left\{\begin{array}{cc}{\left(\left({\mathrm{kN}}_T\right)\right)}_{(N_{T}P-1)}& 0\&le;k<N_{T}P-1\\ N_{T}P-1& k=N_{T}P-1\end{array}\right.$

$\mathrm{\&alpha;}\left(k\right)=\left\{\begin{array}{cc}{\left(\left(\mathrm{kQ}\right)\right)}_{(L_P-1)}& 0\&le;k<L_P-1\\ L_P-1& k=L_P-1\end{array}\right.$

$\mathrm{\&beta;}\left(k\right)={\left((k+N\&CenterDot;{\left(\left(k\right)\right)}_N)\right)}_{L_P},0\&le;k\&le;L_P-1$

Diagonal matrix Λ can obtain its diagonal element by formula 8 counter pushing away.By formula 6,7,8, we obtain three kinds of algorithms of initial LS channel estimating:

(1) directly calculate (formula 6), its complex multiplication operation number of times is N _RL _PN _TP;

(2) fast algorithm one (formula 7), its complex multiplication operation number of times is N _RL _P(1+1og ₂(QN));

(3) fast algorithm two (formula 8), its complex multiplication operation number of times is N _RL _P(1+log ₂N).

Fig. 2 has provided a kind of installation drawing of being realized the LS channel estimating by formula 8.

4. more precise channels is estimated

Least square channel estimating and Noise Variance Estimation method have more than been provided based on single pilot.In two circulation adaptive slot structures, a plurality of pilots are arranged, utilize the temporal correlation that estimates channel parameter, can obtain more precise channels estimation.In addition, work as L _P=N _TDuring P, can't utilize formula 4 to carry out the estimation of noise variance, can utilize the temporal correlation of channel parameter to carry out the estimation of noise variance this moment.

Note ${\hat{h}}_{m,n}\left(p\right)={[{\hat{h}}_{m,n}^{\left(0\right)}\left(p\right),{\hat{h}}_{m,n}^{\left(1\right)}\left(p\right),...,{\hat{h}}_{m,n}^{\left(k\right)}\left(p\right)]}^T,$ Expression the (n, m) K+1 channel parameter of acquisition gone up in p of transmission channel footpath, then the (m, all channel parameters that n) obtain on the transmission channel can be written as:

${\hat{h}}_{m,n}={[{\hat{h}}_{m,n}^T\left(0\right),{\hat{h}}_{m,n}^T\left(1\right),...,{\hat{h}}_{m,n}^T(P-1)]}^T$

Then have:

${\hat{h}}_{m,n}=h_{m,n}+{\mathrm{\&eta;}}_{m,n}$ [formula 9]

Wherein, h _{M, n}For with

Corresponding ideal communication channel vector, η _{M, n}Be zero-mean white Gauss noise vector, the variance of its each element is σ ²/ L _P

By formula 9 as can be known, h _{M, n}Least mean-square error (MMSE) be estimated as:

${\tilde{h}}_{m,n}=R{(R+\frac{{\mathrm{\&sigma;}}^2}{L_P}{I}_{P(K+1)})}^{-1}{\hat{h}}_{m,n}$ [formula 10]

Wherein $R=E\left\{h_{m,n}{h}_{m,n}^H\right\}.$ Here, we suppose that each transmission channel has identical power time-delay spectrum (PDP), further, utilize the channel statistical characteristic in the separable character of time-domain and frequency-domain, and top relevant battle array R can be decomposed into: $R=R_{\mathrm{ISI}}\text{\&CircleTimes;}{R}_{\mathrm{DPR}},$ Wherein $R_{\mathrm{ISI}}=\mathrm{diag}\{{\mathrm{\&rho;}}_0^2,{\mathrm{\&rho;}}_1^2,...,{\mathrm{\&rho;}}_{P-1}^2\},$

Be channel p footpath power; R _DPRIt is the channel time domain statistical property of determining by Doppler frequency shift.Utilize the decomposition of R, formula 10 can dimensionality reduction realizes, that is:

${\tilde{h}}_{m,n}\left(p\right)=R_p{(R_p+\frac{{\mathrm{\&sigma;}}^2}{L_P}{I}_{K+1})}^{-1}{\hat{h}}_{m,n}\left(p\right),(p=\mathrm{0,1},...,P-1)$ [formula 11]

Wherein $R_p=E\left\{h_{m,n}\left(p\right)h_{m,n}^H\left(p\right)\right\}={\mathrm{\&rho;}}_p^2{R}_{\mathrm{DPR}}.$

Because R _pBe the Hermite battle array, so but feature decomposition be: R _p=U ^HΛ _pU, wherein U is an orthogonal matrix, Λ _p=diag{ λ _{P, 0}, λ _{P, 1}..., λ _{P, K}.Utilize this feature decomposition, formula 11 can be rewritten as:

${\tilde{h}}_{m,n}\left(p\right)=U^H{\mathrm{\&Gamma;}}_{p}U{\hat{h}}_{m,n}\left(p\right)$ [formula 12]

Γ wherein _p=diag{ γ _{P, 0}, γ _{P, 1}..., γ _{P, K}, ${\widetilde{\mathrm{\&gamma;}}}_{p,k}={\mathrm{\&lambda;}}_{p,k}/({\mathrm{\&lambda;}}_{p,k}+{\mathrm{\&sigma;}}^2/L_P).$

In order to realize h _{M, n}(p) MMSE estimates, needs the relevant battle array of actual measurement R _p, and it is carried out feature decomposition.Consider that different transmission channels has identical R _pSo, can directly utilize sample on the space to R _PEstimate, that is:

${\hat{R}}_p=\frac{1}{N_T{N}_R}\underset{m=0}{\overset{N_R-1}{\mathrm{\&Sigma;}}}\underset{n=0}{\overset{N_T-1}{\mathrm{\&Sigma;}}}{\hat{h}}_{m,n}\left(p\right){\hat{h}}_{m,n}^H\left(p\right)$

Matrix character decomposition operation for fear of complexity, on the basis of the relevant battle array of research time domain characteristic, we approach the feature decomposition of relevant battle array with discrete cosine transform (DCT), orthogonal matrix U above also the DCT matrix of promptly ordering with K+1 replaces, and, obtain h by anti-dct transform then in denoising and Noise Variance Estimation that dct transform domain carries out pointwise _{M, n}(p) approximate solution of MMSE estimation.Theory analysis and simulation result confirm that all the method is effective.

5. the channel estimating of data segment

After the channel parameter that has obtained pilot, need follow the tracks of or predict the channel parameter of data segment.Method commonly used has: linear interpolation, Gauss's linear interpolation, weighting multi-slot average (WMSA), these methods all are simple linear process, their common drawback is, when mobile station speed is too fast, the conversion of channel fading is very fast, or nonlinear change appears, makes and utilizes pilot channel can not reflect channel variance situation truly as the data channel that linear process obtains.

Here we carry out interpolation in the DCT territory to channel parameter, and detailed process is as follows: at first the DCT of ordering with K+1 will

Transform to the DCT territory, obtain the vector of K+1 dimension, fill (K+1) (L-1) individual neutral element at its end, obtain the vector of (K+1) L dimension, wherein L is the interpolation factor at data segment, that is insert out L-1 value in each sample value back, the anti-dct transform of using (K+1) L to order at last again returns its conversion to time domain.Because the edge effect of DCT interpolation, (L-1) individual data of afterbody are not very accurate, and considering does not need these data in follow-up detection, so (L-1) individual data of deletion afterbody.The channel parameter length that obtains after the interpolation is KL+1.In conjunction with formula 12, above-mentioned processing procedure can be described as with formula:

[formula 13]

Wherein

Be the expanded DCT matrix that (K+1) L is ordered,

It is the delivery channel parameter after the interpolation.Work as K+1=8, during L=4, the real multiplications number of times of formula 13 is 224.

Claims (5)

1, the method for channel estimating in a kind of multiaerial system, it is characterized in that: at transmitting terminal, utilize discrete Fourier transform (DFT) " DFT " matrix construction to go out cyclic orthogonal experiment pilot frequency sequence s optimum on least square " LS " meaning, and it is inserted the transmission data off and on to form bicirculating structure of time slot; At receiving terminal, utilize the characteristic of cyclic orthogonal experiment sequence, carry out least square channel estimating optimum on least mean-square error " MMSE " meaning with low implementation complexity, utilize the decomposition that receives pilot matrix, carry out the quick realization of channel estimating, utilize discrete cosine transform " DCT " to carry out pilot more precise channels estimation and Noise Variance Estimation, adopt discrete cosine transform " DCT " interpolation to obtain the estimation of data segment channel parameter again, this method comprises following step:

Step 1), at transmitting terminal, according to the number NT of multiaerial system transmitting antenna and the multipath number P of channel, construct length and be

Optimum cyclic orthogonal experiment pilot frequency sequence s on the least square meaning, and generate the pilot frequency sequence of each transmitting antenna according to the following rules:

$s_n\left(l\right)=s_{{\left((l-\mathrm{nP})\right)}_{L_P}},(n=\mathrm{0,1},...,N_T-1,l=\mathrm{0,1},...,L_P-1);$

Step 2), at receiving terminal, try to achieve the channel impulse response parameter Estimation of each pilot in the time slot by following formula

${\hat{H}}^{\left(k\right)}=\frac{1}{L_P}{Y}^{\left(k\right)}{X}^H,(k=\mathrm{0,1},...,K)$

Wherein: the matrix that X is made up of pilot signal, Y ^(k)Be k the pilot signal that sub-slots receives;

Step 3), at receiving terminal, utilize step 2) channel impulse response that estimates estimates an interchannel noise variance by following formula to each pilot, and obtains the Noise Variance Estimation of current time slots:

${\widehat{\mathrm{\&sigma;}}}_k^2=\frac{1}{N_R(L_P-N_{T}P)}{\left|\right|Y^{\left(k\right)}-{\hat{H}}^{\left(k\right)}X\left|\right|}_F^2,(k=\mathrm{0,1},...,K),$

${\widehat{\mathrm{\&sigma;}}}^2=\frac{1}{K+1}\underset{k=0}{\overset{K}{\mathrm{\&Sigma;}}}{\widehat{\mathrm{\&sigma;}}}_k^2,$

Wherein: N _RExpression reception antenna number;

Step 4), at receiving terminal, utilize step 2) channel impulse response of all pilots of estimating, it is carried out the denoising and the Noise Variance Estimation of pointwise in the DCT territory, and by obtain the channel impulse response of data segment in DCT territory interpolation

Wherein: Represent between n root transmitting antenna and m root reception antenna p footpath channel impulse response, U and Represent DCT matrix and expansion DCT matrix respectively, Γ _pIt is filtering matrix.

2, the method for channel estimating in the multiaerial system according to claim 1 is characterized in that cyclic orthogonal experiment pilot frequency sequence s is constructed by fourier transform matrix in the described method, and satisfies the cyclic orthogonal experiment characteristic; The pilot frequency sequence of each transmitting antenna is obtained by the s cyclic shift.

3, the method for channel estimating in the multiaerial system according to claim 1 is characterized in that described least square is meant the quadratic sum minimum of evaluated error.

4, the method for channel estimating in the multiaerial system according to claim 1 is characterized in that step 2) described channel impulse response parameter Estimation carries out sub-slots; Utilize to receive the following decomposition of pilot matrix, the channel impulse response parameter estimate at quick implementation algorithm:

$X=P_{\mathrm{\&mu;}}^H\left[\begin{array}{cc}{I}_{N_{T}P}& 0_{N_{T}P\&times;(L_P-N_{T}P)}\end{array}\right]P_{\mathrm{\&alpha;}}^H(I_Q\&CircleTimes;W)P_{\mathrm{\&beta;}}^H\mathrm{\&Lambda;}(I_Q\&CircleTimes;W^H)P_{\mathrm{\&alpha;}},$

Wherein: P _α, P _β, P _μBe permutation matrix, W is the DFT transformation matrix, and Λ is a diagonal matrix, and Q is a constant.

5, the method for channel estimating in the multiaerial system according to claim 1 is characterized in that step 2) described channel impulse response parameter Estimation carries out in a time slot; The denoising of channel impulse response parameter and interpolation are all carried out in discrete Fourier transform (DFT) " DFT " territory.

Images