CN1200521C - Method for inserting pilot frequency and estimating parameters of channel for equilibrium in frequency doman - Google Patents

Abstract

The present invention relates to a method for pilot frequency insertion and channel parameter estimation for frequency domain equilibrium, which belongs to the technical field of wireless communication. The method for pilot frequency insertion and channel parameter estimation for frequency domain equilibrium is characterized in that a data signal after modulated and mapped by a code is reconnected with a pilot frequency signal at the sending terminal of a communication system. Pilot frequency information is only present on some frequency points after orthogonal transformation and interleaving, while data do not have frequency components. Then, each data block is added with a cyclic prefix at the sending terminal. The cyclic prefix is then removed at a receiving terminal so as to eliminate interference among the data blocks and interference among frequency domain signals, which are caused by multipath time delay expansion. The extracted pilot frequency information of the frequency domain is subsequently utilized for the channel parameter estimation on a corresponding frequency point. In this way, the channel parameter estimation is then performed on all the frequency points by an interpolation method. The method for pilot frequency insertion and channel parameter estimation for frequency domain equilibrium has the advantages of simple calculation method and realization and high accuracy of channel estimation.

Description

Be used for the pilot tone insertion of frequency domain equalization and the method for channel parameter estimation

Technical field

Be used for the pilot tone insertion of frequency domain equalization and the method for channel parameter estimation and belong to wireless communication technology field.

Background technology

The baseband transmission system of a reality can not satisfy desirable waveform transmission distortionless condition fully, thereby to crosstalk almost be inevitable.Cause when having a strong impact on when crosstalking, must proofread and correct the transfer function of whole system, promptly balanced, make it near undistorted transmission conditions.Equilibrium is exactly that the equalizer of receiving terminal produces the characteristic opposite with the characteristic of channel, is used for offsetting the interference that the time change multipath transmisstion characteristic of channel causes, can be divided into two kinds of time domain equalization and frequency domain equalizations.Before equilibrium, need to estimate channel parameter, utilize pilot signal to carry out channel estimating usually.

The signal that sends in the communication system mainly can be divided into pilot signal and data, wherein data are information signals of actual transmission, pilot signal is an additional known information signal, thereby is used for obtaining balanced weight coefficient in the estimation that receiving unit carries out channel parameter.

Usually adopt pilot tone to take the insertion method of a quadrature code channel in traditional method such as the 3G system, adopt many code channel transmission data, when channel multi-path is serious, serious crosstalking arranged between code channel.Before equilibrium, crosstalking between pilot tone code channel and data code channel makes that the channel parameter estimate according to pilot signal is inaccurate, and according to the time domain equalization that the channel parameter of gained carries out, convergence rate tap number slow and transversal filter is too many, it is too complicated to realize.

Need to improve the method for pilot tone insertion and channel estimating for this reason, eliminate crosstalking between pilot signal and data-signal, improve accuracy of channel estimation, reduce the complexity of algorithm and realization simultaneously.

Summary of the invention

The objective of the invention is:, proposed a kind of realize simply being used for the pilot tone insertion of frequency domain equalization and the method for channel parameter estimation in order to obtain higher precision of channel estimation.

The pilot tone that is used for frequency domain equalization that the present invention proposes is inserted the method with channel parameter estimation, it is characterized in that: in wireless communication system, at transmitting terminal, each data signal sequence and pilot signal sequence multiple connection after the coded modulation mapping form a data block, after orthogonal transform and interweaving, again each data block is added Cyclic Prefix; At receiving terminal, remove Cyclic Prefix again, to eliminate interference between the data block that the multidiameter delay expansion causes and the interference between frequency-region signal, so that utilize the frequency-domain pilot information that extracts to make channel estimating; Particularly, it contains successively and has the following steps,

Transmitting terminal:

(1) data signal sequence and pilot signal sequence multiple connection after the coded modulation mapping:

Is known random signal train X that the pilot signal sequence of time domain is inserted among the data sequence D after the coded modulation in order uniformly, form a data block s, establishing X length is M, and D length is (MN-M), wherein M and N are positive integer, and then data block s length is N _B=MN, note X=(x ₀, x ₁, Λ x _M-1), D=(d ₀, d ₁, Λ d _MN-M-1), s=(s ₀, s ₁, s ₂, s ₃, Λ s _MN-1), between per two pilot signals N-1 data are arranged, establish the 0th pilot signal x ₀Be inserted in L position, 0≤L≤N-1 then has

(2) signal after the above-mentioned multiple connection is carried out the conversion of N orthogonal dimension;

(3) signal that above-mentioned orthogonal transform is obtained is made interlace operation to change the order of burst, promptly writes in the matrix by row, reads by row again;

(4) data block after above-mentioned interweaving is added Cyclic Prefix;

Receiving terminal:

(3) remove the Cyclic Prefix that adds at transmitting terminal;

(4) channel parameter estimation:

Earlier all frequencies of pilot frequency information are made channel parameter estimation, and then the frequency of data message is made channel parameter estimation, to obtain the channel frequency domain characteristic of all frequencies by interpolation;

In said method, the conversion of described N orthogonal dimension can be expressed as: B _{N * 1}=WA _{N * 1}, wherein A is the N dimensional vector, is made up of N-1 data-signal and pilot signal, B is the signal vector after the orthogonal transform, W is the matrix of N * N of answering of N orthogonal dimension transfer pair, if pilot signal is positioned at L the position of vectorial A, then N element of matrix W L row is satisfied:

$w_{n,L}=\frac{1}{\sqrt{N}}\·e^{j\frac{2\mathrm{\πkn}}{N}},n=\mathrm{0,1,2}\mathrm{\ΛN}-1$

Form, w wherein _{N, L}Be the element of the capable L row of n of matrix W, k is an integer arbitrarily;

In said method, described interlace operation changes the order of burst, and elder generation writes the data order in the matrix by row, and N data of every row are write the M row altogether, read by row again, and the data block of establishing before and after interweaving is q=(q ₀, q ₁, q ₂, q ₃, Λ q _MN-1) ^TAnd g=(g ₀, g ₁, g ₂, g ₃, Λ g _MN-1) ^T, wherein T represents vectorial transposition, then:

g _n·M+m＝q _m·N+n，(0≤m≤M-1，0≤n≤N-1)；

In said method, described to add Cyclic Prefix operation be operation to the data block g after interweaving, and establishing circulating prefix-length is N _g, with the last N of g _gIndividual signal replication is to the front end of data block g, and the length of the data block g after interweaving equals N _B, then generating a length that includes Cyclic Prefix is N _s=N _B+ N _gData block f, promptly

f＝T _CP·g

Wherein

$T_{\mathrm{CP}}=\left[\begin{array}{c}{I}_{\mathrm{CP}}\\ I_{N_B}\end{array}\right]$

Use I _NBExpression N _B* N _BUnit matrix, use I _CPExpression I _NBLast N _gOK, T _CPBe by I _CPAnd I _NBThe matrix of combining and forming.

Experimental results show that this method that adopts, can obtain higher channel estimating and balanced precision with lower complexity, thereby improve the signal to noise ratio of baseband transmission, help system and adopt high system modulation to reach the availability of frequency spectrum efficiently.

Description of drawings

Fig. 1 represents transmitting terminal baseband digital signal handling procedure flow chart of the present invention.

Fig. 2 represents receiving terminal baseband digital signal handling procedure flow chart of the present invention.

Fig. 3 represent to interweave order of algorithm data read-write.

Embodiment

Utilize a plurality of mutual isolation and relatively independent subcarrier in the communication system, each carrier structure is basic identical, the baseband digital signal of each subcarrier is handled the method for the present invention that adopts, be each data signal sequence and the data block of pilot signal sequence multiple connection formation after the coded modulation mapping, by orthogonal transform with interweave, make pilot frequency information only appear on some frequency, and data do not have these frequency components, again each data block is added Cyclic Prefix making a start, receiving end is removed it, thereby interference between the data block that the multidiameter delay expansion causes and interference between frequency-region signal have been eliminated, guaranteed that the pilot frequency information that receives is not subjected to the interference of data-signal at frequency domain, and then utilizing the frequency-domain pilot information that extracts to make channel estimating and frequency domain equalization, the data-signal that orthogonal transform obtains sending is conciliate in deinterleaving again.

Pilot tone provided by the present invention is inserted with channel estimation methods as follows in the baseband digital signal processing procedure of transmitting terminal:

1, data and pilot tone multiple connection

Be known random signal train X that the pilot signal sequence of time domain is inserted among the data sequence D after the coded modulation in order uniformly at first, form a data block s, establishing X length is M, and D length is (MN-M), wherein M and N are positive integer, and then data block s length is N _B=MN.Note X=(x ₀, x ₁, Λ x _M-1), D=(d ₀, d ₁, Λ d _MN-M-1), s=(s ₀, s ₁, s ₂, s ₃, Λ s _MN-1), between per two pilot signals N-1 data are arranged, establish the 0th pilot signal x ₀Be inserted in L position, 0≤L≤N-1, then

s＝(s ₀，s ₁，s ₂，s ₃，Λs _M·N-1)＝(d ₀，Λd _L-1，x ₀，d _L，Λ，d _N+L-2，x ₁，d _N+L-1，Λd _M·N-M-1)

Be formulated pilot tone and insert operation:

2, N orthogonal dimension conversion

$q=\left[\begin{array}{c}{q}_0\\ q_1\\ M\\ q_{M\·N-1}\end{array}\right]=\stackrel{\mathrm{\ρ}}{W}\·s=\left[\begin{array}{cccc}W& 0& \mathrm{\Λ}& 0\\ 0& W& \mathrm{\Λ}& 0\\ M& M& O& M\\ 0& 0& \mathrm{\Λ}& W\end{array}\right]\·\left[\begin{array}{c}{s}_0\\ s_1\\ M\\ s_{M\·N-1}\end{array}\right].$

Wherein W is a N orthogonal dimension transformation matrix, Diagonal on M W matrix arranged, N element of the L of W row is satisfied:

$w_{n,L}=\frac{1}{\sqrt{N}}\·e^{j\frac{2\mathrm{\πkn}}{N}},n=\mathrm{0,1,2}\mathrm{\ΛN}-1$

Form, w wherein _{N, L}Be the element of the capable L row of n of matrix W, k is an integer arbitrarily.Pay special attention to, when making the matrix multiplication of orthogonal transform, the element of W L row and the pilot signal among the vectorial s multiply each other, and other N-1 column element and data-signal multiply each other.

3, interweave

Interlace operation changes the order of burst, and elder generation writes the data order in the matrix by row, and N data of every row are write the M row altogether, read by row again, and the data block of establishing after interweaving is $g={(g_0,g_1,g_2,g_3,\mathrm{\Λ}{g}_{N_B-1})}^T,$ Represent interlace operation with J, i.e. g=Jq is formulated as:

g _n·M+m＝q _m·N+n，(0≤m≤M-1，0≤n≤N-1)。

4, add Cyclic Prefix

Last N with data block _gIndividual signal replication is to its front end, if former data block length is N _B, then generating a length that includes Cyclic Prefix is N _s=N _B+ N _gData block f, promptly

$f=T_{\mathrm{CP}}\·g=T_{\mathrm{CP}}\·J\·\stackrel{\mathrm{\ρ}}{W}\·s---\left(1\right)$

Wherein

$T_{\mathrm{CP}}=\left[\begin{array}{c}{I}_{\mathrm{CP}}\\ I_{N_B}\end{array}\right]$

T _CPBe by N _B* N _BUnit matrix I _NBAnd I _NBLast N _gRow (is used I _CPExpression) combines and the matrix that forms.

With the transmission of the data block of matrix notation signal under multipath channel, used herein is the discrete equivalent model of system, supposes at receiving terminal below, comprise carrier wave, code element and sampling clock various all be accurately synchronously, then work as t _mCan get sampled signal during=mT:

$v^{\′}\left(t_m\right)=\underset{k=-\∞}{\overset{+\∞}{\mathrm{\Σ}}}f\left(k\right)h(\mathrm{mT}-\mathrm{kT},\mathrm{mT})+n\left(\mathrm{mT}\right)---\left(2\right)$

Wherein $h(\mathrm{\τ},t)=\underset{l=1}{\overset{L}{\mathrm{\Σ}}}{h}_l\left(t\right)\·\mathrm{\δ}(\mathrm{\τ}-{\mathrm{\τ}}_l)$ Be the time domain impulse response of channel, n (t) is an interchannel noise.

Suppose that channel impulse response changes slowly, promptly in a data block, suppose that channel is constant, but from a data block to next data block, channel changes.In addition, the sample sequence of channel impulse response always can be approximated in the scope that precision allows and have finite length.The length of supposing this sequence is K+1, then can use vectorial h _i=[h _i(0), Λ h _i(K)] ^TBe illustrated in the whole channel impulse response sequence in i the data block, wherein h _i(m)=and h (mT, iTNs), TN wherein _sBe the length that comprises the whole data block of Cyclic Prefix.In order to eliminate the interference between the data block, in system design, the length of Cyclic Prefix always will obtain the maximum additional delay more than or equal to channel, promptly will satisfy N _g〉=K.

With I the data block that expression sends,

I the data block that expression receives by (2) formula, and utilized the characteristic of channel impulse response for finite length, and the relation that can obtain between transmitting block and the reception piece is as follows:

Because the interference between the data block that multipath channel is introduced makes To depend on simultaneously With In the formula

Be corresponding noise vector, C ₀And C ₁Be respectively N with following form _s* N _sMatrix, the element on its diagonal all equates:

$C_0=\left[\begin{array}{ccccc}{h}_i\left(0\right)& 0& 0& \mathrm{\Λ}& 0\\ M& h_i\left(0\right)& 0& \mathrm{\Λ}& 0\\ h_i\left(k\right)& \mathrm{\Λ}& O& \mathrm{\Λ}& M\\ M& O& M& O& 0\\ 0& \mathrm{\Λ}& h_i\left(K\right)& \mathrm{\Λ}& h_i\left(0\right)\end{array}\right]$

$C_1=\left[\begin{array}{ccccc}0& \mathrm{\Λ}& h_i\left(K\right)& \mathrm{\Λ}& h_i\left(1\right)\\ M& O& 0& O& M\\ 0& \mathrm{\Λ}& O& \mathrm{\Λ}& h_i\left(K\right)\\ M& O& M& O& 0\\ 0& \mathrm{\Λ}& 0& \mathrm{\Λ}& 0\end{array}\right]$

The baseband digital signal processing procedure of receiving terminal is as follows:

1, removes Cyclic Prefix

Wherein

The Cyclic Prefix operation is removed in expression, Be N _B* N _gNull matrix, I _NBBe N _B* N _BUnit matrix.

As seen, work as N _gDuring 〉=K, remove the Cyclic Prefix operation and also will eliminate inter-block-interference,

To only depend on

At this moment, we just can carry out independent demodulation to each data block, and the mark i of expression data block sequence number will be omitted.

(1) formula substitution (3) Shi Kede:

Wherein Be noise vector, Being the compound channel impulse response, is a N _B* N _BCircular matrix, have following special shape

$\stackrel{\mathrm{\ρ}}{C}=\left[\begin{array}{ccccccc}{h}_i\left(0\right)& 0& \mathrm{\Λ}& 0& h_i\left(K\right)& \mathrm{\Λ}& h_i\left(1\right)\\ M& h_i\left(0\right)& 0& \mathrm{\Λ}& O& O& M\\ M& M& O& O& \mathrm{\Λ}& O& h_i\left(K\right)\\ h_i\left(K\right)& M& M& O& O& \mathrm{\Λ}& 0\\ 0& h_i\left(K\right)& M& M& O& O& M\\ M& O& O& M& M& O& 0\\ 0& \mathrm{\Λ}& 0& h_i\left(K\right)& \mathrm{\Λ}& \mathrm{\Λ}& h_i\left(0\right)\end{array}\right]$

Just

(k, l) individual element is h _i((k-l) mod N _B), mod represents modulo operation here.

Utilize the character of circular matrix, can resolve into following form:

Wherein F and F ^-1Be respectively N _B* N _BFourier transform and inverse fourier transform matrix, Λ is the transfer function matrix of compound channel, is a diagonal matrix.If h=[h (0), Λ h (K), 0, Λ, 0] ^TN _BPoint Fourier is transformed to H=[H (0), Λ H (N _B-1)] ^T, then

$\mathrm{\Λ}={\left[\begin{array}{cccc}H\left(0\right)& 0& \mathrm{\Λ}& 0\\ 0& H\left(1\right)& \mathrm{\Λ}& 0\\ M& M& O& M\\ 0& 0& \mathrm{\Λ}& H(N_B-1)\end{array}\right]}_{N_B\×N_B}$

The received signal that (5) formula substitution (4) Shi Kede is removed behind the Cyclic Prefix is:

2, channel estimating

Obtain the received signal of frequency domain by Fourier transform:

By (6) formula as can be known

$V_{\mathrm{mN}+k}=\sqrt{N}\·H(\mathrm{mN}+k)\·p_m+n_{\mathrm{mN}+k},0\≤m\≤M-1$

Wherein

Be the noise vector of frequency domain,

$\left[\begin{array}{c}{p}_0\\ p_1\\ M\\ p_{M-1}\end{array}\right]=F\·\left[\begin{array}{c}{x}_0\\ x_1\·e^{-j2\mathrm{\πk}/\mathrm{MN}}\\ M\\ x_{M-1}\·e^{-j2\mathrm{\πk}(M-1)/\mathrm{MN}}\end{array}\right]$ It is the frequency-domain pilot information that sends.

F is a fourier transform matrix, (x ₀, x ₁, Λ x _M-1) be the time-domain pilot signal identical with transmitting terminal, x _ne ^{-j2 π kn/MN}In k equal parameter k among the orthogonal transform matrix W.

The pilot frequency information that receives only appears on the frequency of mN+k (m is an integer), and there are not these frequency components in data, extract the signal on these frequencies, $\left\{{\hat{p}}_m\right|{\hat{p}}_m=V_{\mathrm{mN}+k},m=\mathrm{0,1,2},\mathrm{KM}-1\},$ Infer the frequency domain characteristic of the corresponding frequency of channel according to the variation of pilot frequency information on frequency domain of pilot frequency information that receives and transmission, just with the pilot tones domain information of the pilot tones domain information that receives divided by transmission, can obtain the least-squares estimation of the channel frequency response H (mN+k) on these frequencies

$\hat{H}(\mathrm{mN}+k)=\frac{{\hat{p}}_m}{\sqrt{N}\·p_m},0\≤m\≤M-1$

In order to obtain the channel frequency domain characteristic of other frequency, need carry out frequency domain interpolation, adopt the method for transform domain interpolation, be exactly again M channel estimation value to be done the inverse-Fourier transform that M orders to transform to time domain in detail,

$h_n=\frac{1}{M}\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}\hat{H}(\mathrm{mN}+k)\·e^{j\frac{2\mathrm{\π}\·m\·n}{M}},0\≤n\≤M-1,$

It is individual zero to insert M * (N-1) in the time domain channel estimated sequence,

Be formulated:

$U={\left[\begin{array}{cc}I& 0\\ 0& 0\\ M& M\\ 0& 0\\ 0& I\end{array}\right]}_{\mathrm{MN}\×M}$ Wherein the size of unit matrix I is $\frac{M}{2}\×\frac{M}{2},$ Onesize full null matrix 0 (4N-2) altogether is individual.

Obtained N like this _BThe time domain channel of point is estimated Again it is transformed to frequency domain as Fourier transform, has just obtained the channel estimation value of all frequencies,

${\hat{H}}_i=\underset{n=0}{\overset{N_B-1}{\mathrm{\Σ}}}{h^{\′}}_n\·e^{-j\frac{2\mathrm{\π}\·i\·n}{N_B}},0\≤i\≤N_B-1.$

3, frequency domain equalization

Utilize the channel estimating of gained, and the signal that receives is carried out Fourier transform, transform to frequency domain,, select a kind of existing frequency domain equalization algorithm, can realize the frequency domain equalization of received signal again according to designing requirement.

4, extract data-signal

Earlier to balanced data stream y=(y ₀, y ₁, y ₂, y ₃, Λ y _MN-1) ^TDo the deinterleaving operation, finish the processing opposite, earlier data are write in the matrix by row with the transmitting terminal interlace operation, every capable M data, it is capable to write N altogether, reads by row again, obtains z=(z after the deinterleaving ₀, z ₁, z ₂, z ₃, Λ z _MN-1) ^T, be formulated as:

z _m·N+n＝y _n·M+m，(0≤m≤M-1，0≤n≤N-1)。

Remake after the deinterleaving and separate orthogonal transform, use matrix

Inverse matrix (just

Conjugate transpose ) multiply by the data block vector.

$b=\left[\begin{array}{c}{b}_0\\ b_1\\ M\\ b_{M\·N-1}\end{array}\right]={\stackrel{\mathrm{\ρ}}{W}}^H\·z=\left[\begin{array}{cccc}{W}^H& 0& \mathrm{\Λ}& 0\\ 0& W^H& \mathrm{\Λ}& 0\\ M& M& O& M\\ 0& 0& \mathrm{\Λ}& W^H\end{array}\right]\·\left[\begin{array}{c}{z}_0\\ z_1\\ M\\ z_{M\·N-1}\end{array}\right].$

Remove pilot signal at last, just removing among the b integral multiple of N, to add L be that (the locational signal of 0≤m≤M-1 obtains the estimated value of former transmission data-signal to mN+L $\hat{D}=({\hat{d}}_0,{\hat{d}}_1,\mathrm{\Λ}{\hat{d}}_{M-N-M-1}),$ Wherein

The data-signal of gained exported to separate mapping and decoder, thereby the reception of finishing baseband digital signal is handled.

Further specify the present invention below in conjunction with drawings and Examples.

In an embodiment, orthogonal transform matrix adopts Hadamard (Hadamar) matrix W _N, W _NBe 2 ⁿ* 2 ⁿDimension, require N=2 ⁿElement has only 1 and-1, from 1 rank to 2 ⁿThe Hadamard matrix on rank can utilize the Kronecker product computing recursion of matrix to obtain, promptly

$W_N=W_2\⊗W_{\frac{N}{2}},$

Wherein:

$W_2=\frac{1}{\sqrt{2}}\left[\begin{array}{cc}1& 1\\ 1& -1\end{array}\right].$

Hadamard transform is the special case of orthogonal transform parameter k=0, L=0, and wherein k=0 makes x in the channel estimating operation _ne ^{J2 π kn/MN}The weight coefficient of pilot signal is equal to 1, realizes the simplest.

Adopt the dimension N=4 of orthogonal transform among the embodiment, 4 rank Hadamard matrixes are

$W_4=W_2\⊗W_2=\frac{1}{\sqrt{4}}\left[\begin{array}{cccc}1& 1& 1& 1\\ 1& -1& 1& -1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\end{array}\right].$

Data block length N _B=MN=128, M=32 wherein, circulating prefix-length N _g=26.

Concrete pilot tone is inserted with the main process of channel estimation methods as follows:

1, data and pilot tone multiple connection

At first known random signal train X=(x ₀, x ₁, Λ x ₃₁) be inserted into the data sequence D=(d after the coded modulation in order uniformly ₀, d ₁, Λ d ₉₅) in, form a data block s, s=(s ₀, s ₁, s ₂, s ₃, Λ s ₁₂₇).Wherein the 0th pilot signal x ₀Be inserted in the 0th position then

s＝(s ₀，s ₁，s ₂，s ₃，Λs ₁₂₇)＝(x ₀，d ₀，d ₁，d ₂，x ₁，d ₃，d ₄，d ₅，x ₂，d ₆，Λx ₃₁，d ₉₃，d ₉₄，d ₉₅)。

2, N orthogonal dimension conversion

Data block is done the conversion of N orthogonal dimension, and what specifically adopt is 4 rank Hadamard transforms.

$q=\left[\begin{array}{c}{q}_0\\ q_1\\ M\\ q_{127}\end{array}\right]=\stackrel{\mathrm{\ρ}}{W}\·s\left[\begin{array}{cccc}{W}_4& 0& \mathrm{\Λ}& 0\\ 0& W_4& \mathrm{\Λ}& 0\\ M& M& O& M\\ 0& 0& \mathrm{\Λ}& W_4\end{array}\right]\·\left[\begin{array}{c}{s}_0\\ s_1\\ M\\ s_{127}\end{array}\right].$

Wherein $W_4=\frac{1}{2}\left[\begin{array}{cccc}1& 1& 1& 1\\ 1& -1& 1& -1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\end{array}\right]$

3, interweave

Interlace operation changes the order of burst, and as shown in Figure 3, elder generation writes the data order in the matrix by row, and 4 data of every row are write 32 row altogether, read the data block after interweaving by row again: g=(g ₀, g ₁, g ₂, g ₃, Λ g ₁₂₇) ^T=(q ₀, q ₄, q ₈, q ₁₂, Λ q ₁₂₄, q ₁, q ₅, q ₉, q ₁₃, Λ q ₁₂₅, q ₂, q ₆, q ₁₀, Λ q ₁₂₆, q ₃, q ₇, q ₁₁, Λ q ₁₂₇) ^T, g wherein _N32+m=q _M4+n, (0≤m≤31,0≤n≤3).

4, add Cyclic Prefix

Last N with data block _g=26 data copy to its front end, and generating a length that includes Cyclic Prefix is N _s=N _B+ N _g=154 data block f, promptly

f＝(f ₀，f ₁，f ₂，f ₃，Λf ₁₅₃) ^T＝(g ₁₀₂，g ₁₀₃，Λg ₁₂₇，g ₀，g ₁，g ₂，g ₃，Λg ₁₂₇) ^T。

The baseband digital signal of gained is through shaping filter and up-conversion, arrive receiving terminal through channels transmit, receiving terminal adopts down-conversion, matched filtering and sampling, suppose at receiving terminal, comprise carrier wave, code element and sampling clock various all be accurately synchronously, then can obtain corresponding base-band digital received signal v '=(v ' ₀, v ' ₁, v ' ₂, v ' ₃, Λ v ' ₁₅₃) ^T

Received signal series processing process is as follows:

1, removes Cyclic Prefix

$\stackrel{\mathrm{\ρ}}{v}={(v_0,v_1,v_2\mathrm{\Λ}{v}_{127})}^T={({v^{\′}}_{26},{v^{\′}}_{26},{v^{\′}}_{28},\mathrm{\Λ}{v^{\′}}_{153})}^T.$

2, channel estimating

Obtain frequency-region signal by Fourier transform:

$V_i=\underset{m=0}{\overset{127}{\mathrm{\Σ}}}{v}_m\·e^{-j2\mathrm{\π}\·m\·i/128},0\≤i\≤127.$

Because k=0, so the frequency-domain pilot information that sends:

$\left[\begin{array}{c}{p}_0\\ p_1\\ M\\ p_{31}\end{array}\right]=F\·\left[\begin{array}{c}{x}_0\\ x_1\·e^{-j2\mathrm{\πk}/128}\\ M\\ x_{31}\·e^{-j2\mathrm{\πk}31/128}\end{array}\right]F\·\left[\begin{array}{c}{x}_0\\ x_1\\ M\\ x_{31}\end{array}\right],$

$p_i=\underset{m=0}{\overset{31}{\mathrm{\Σ}}}{x}_m\·e^{-j2\mathrm{\π}\·m\·i/32},0\≤i\≤31$

The pilot frequency information that receives only appears on the frequency of mN+k=4m (m is an integer, 0≤m≤31), and there are not these frequency components in data, extract the pilot frequency information on these frequencies, $\left\{{\hat{p}}_m\right|{\hat{p}}_m=V_{4m},m=\mathrm{0,1,2},\mathrm{KM}-1\},$ Utilize division can obtain the least-squares estimation of the channel frequency response on these frequencies,

$\hat{H}\left(4m\right)=\frac{{\hat{p}}_m}{2\·p_m},0\≤m\≤31$

The inverse-Fourier transform of channel estimation value being done at 32 transforms to time domain,

$h_0=\frac{1}{32}\underset{m=0}{\overset{31}{\mathrm{\Σ}}}\hat{H}\left(4m\right)\·e^{j\frac{2\mathrm{\π}\·m\·n}{M}},0\≤n\≤31,$

Insert M * (N-1)=96 zero in the middle of the time domain channel estimated sequence,

Obtained 128 time-domain information like this Again it is transformed to frequency domain as Fourier transform, has just obtained the channel estimation value of all frequencies,

${\hat{H}}_i=\underset{n=0}{\overset{127}{\mathrm{\Σ}}}{h^{\′}}_n\·e^{-j\frac{2\mathrm{\π}\·i\·n}{128}},0\≤i\≤128.$

3, frequency domain equalization

Equilibrium can be adopted zero forcing equalization or least mean-square error equilibrium etc.With the zero forcing equalization is example, and weight coefficient equals the inverse of channel transfer function, and the signal that receives is carried out Fourier transform to frequency domain, carry out frequency domain weighting, inverse transformation is proofreaied and correct the transfer function of whole system with realization to time domain again, makes transmission near distortionless condition.

The data block of removing behind the Cyclic Prefix that receives obtains the frequency domain received signal as Fourier transform:

$V_i=\underset{m=0}{\overset{127}{\mathrm{\Σ}}}{v}_m\·e^{-j2\mathrm{\π}\·m\·i/128},0\≤i\≤127.$

The frequency domain weighting operation:

${\mathrm{\η}}_i=\frac{V_i}{{\hat{H}}_i},0\≤i\≤127$

Inverse-Fourier transform arrives time domain:

$y_n=\frac{1}{128}\underset{n=0}{\overset{127}{\mathrm{\Σ}}}{\mathrm{\η}}_i\·e^{j\frac{2\mathrm{\π}\·i\·n}{128}},0\≤n\≤127$

4, extract data-signal

Earlier balanced data stream is done the deinterleaving operation, finish and the transmitting terminal opposite processing that interweaves, earlier data are write in the matrix by row, 32 data of every row are write 4 row altogether, read by row again, obtain z=(z after the deinterleaving ₀, z ₁, z ₂, z ₃, Λ z ₁₂₇) ^T=(y ₀, y ₃₂, y ₆₄, y ₉₆, y ₁, y ₃₃, y ₆₅, y ₉₇, Λ y ₃₁, y ₆₃, y ₉₅, y ₁₂₇) ^T, z wherein _4m+n=y _32n+m, (0≤m≤31,0≤n≤3).

Remake after the deinterleaving and separate orthogonal transform, use matrix Inverse matrix (just

Conjugate transpose ) multiply by the data block vector.

$b=\left[\begin{array}{c}{b}_0\\ b_1\\ M\\ b_{127}\end{array}\right]={\stackrel{\mathrm{\ρ}}{W}}^H\·z=\left[\begin{array}{cccc}{W_4}^H& 0& \mathrm{\Λ}& 0\\ 0& {W_4}^H& \mathrm{\Λ}& 0\\ M& M& O& M\\ 0& 0& \mathrm{\Λ}& {W_4}^H\end{array}\right]\·\left[\begin{array}{c}{z}_0\\ z_1\\ M\\ z_{127}\end{array}\right].$

Hadamard matrix is symmetrical real number matrix, and conjugate transpose just equals itself,

${W_4}^H=\frac{1}{2}\left[\begin{array}{cccc}1& 1& 1& 1\\ 1& -1& 1& -1\\ 1& 1& -1& -1\\ 1& -1& -1& 1\end{array}\right].$

Remove pilot signal at last, just remove among the b 4 the locational signal of integral multiple, obtain the estimated value of former transmission data-signal $\hat{D}=({\hat{d}}_0,{\hat{d}}_1,\mathrm{\&Lambda;}{\hat{d}}_{95})=(b_1,b_2,b_3,b_5,b_6,{\mathrm{<figure-callout\; id="7"\; label="b"\; filenames="C0313756500162.png"\; state="\{\{state\}\}">b</figure-callout>}}_7,b_9,b_{10},\mathrm{\&Lambda;}{b}_{127}),$ Wherein

${\hat{d}}_{3m+n}=b_{4m+n+1},$ 0≤m≤31 0≤n≤2 wherein.

At last the data-signal of gained is exported to and separated mapping and decoder, thereby the reception of finishing baseband digital signal is handled.

This algorithm can be used computer program emulation, also can implement in various programmable logic devices or application-specific integrated circuit (ASIC).

Pilot tone insertion algorithm among prior art and the present invention adopts pilot frequency information to take a quadrature code channel respectively and takies the method for some frequency on the frequency domain, when multipath channel is transmitted, the former pilot signal can be crosstalked with data, cause the channel estimating of carrying out and balanced precision lower according to pilot signal, the latter has overcome the multipath effect of channel, pilot frequency information and data can not crosstalked, and have obtained higher channel estimated accuracy.

Claims (4)

1. be used for the pilot tone insertion of frequency domain equalization and the method for channel parameter estimation, it is characterized in that: in wireless communication system, at transmitting terminal, each data signal sequence and pilot signal sequence multiple connection after the coded modulation mapping form a data block, after orthogonal transform and interweaving, again each data block is added Cyclic Prefix; At receiving terminal, remove Cyclic Prefix again, to eliminate interference between the data block that the multidiameter delay expansion causes and the interference between frequency-region signal, so that utilize the frequency-domain pilot information that extracts to make channel estimating; Particularly, it contains successively and has the following steps,

Transmitting terminal:

(1) data signal sequence and pilot signal sequence multiple connection after the coded modulation mapping:

Is known random signal train X that the pilot signal sequence of time domain is inserted among the data sequence D after the coded modulation in order uniformly, form a data block s, establishing X length is M, and D length is (MN-M), wherein M and N are positive integer, and then data block s length is N _B=MN, note X=(x ₀, x ₁, Λ x _M-1), D=(d ₀, d ₁, Λ d _MN-M-1), s=(s ₀, s ₁, s ₂, s ₃, Λ s _MN-1), between per two pilot signals N-1 data are arranged, establish the 0th pilot signal x ₀Be inserted in L position, 0≤L≤N-1 then has

(2) signal after the above-mentioned multiple connection is carried out the conversion of N orthogonal dimension;

(3) signal that above-mentioned orthogonal transform is obtained is made interlace operation to change the order of burst, promptly writes in the matrix by row, reads by row again;

(4) data block after above-mentioned interweaving is added Cyclic Prefix;

Receiving terminal:

(1) removes the Cyclic Prefix that adds at transmitting terminal;

(2) channel parameter estimation:

Earlier all frequencies of pilot frequency information are made channel parameter estimation, and then the frequency of data message is made channel parameter estimation, to obtain the channel frequency domain characteristic of all frequencies by interpolation.

2. the pilot tone that is used for frequency domain equalization according to claim 1 is inserted the method with channel parameter estimation, and it is characterized in that: the conversion of described N orthogonal dimension can be expressed as: B _{N * 1}=WA _{N * 1}, wherein A is the N dimensional vector, is made up of N-1 data-signal and pilot signal, B is the signal vector after the orthogonal transform, W is the matrix of N * N of answering of N orthogonal dimension transfer pair, if pilot signal is positioned at L the position of vectorial A, then N element of matrix W L row is satisfied:

$w_{n,L}=\frac{1}{\sqrt{N}}\&CenterDot;e^{j\frac{2\mathrm{\&pi;kn}}{N}},n=\mathrm{0,1,2}\mathrm{\&Lambda;N}-1$

Form, w wherein _{N, L}Be the element of the capable L row of n of matrix W, k is an integer arbitrarily.

3. the pilot tone that is used for frequency domain equalization according to claim 1 is inserted the method with channel parameter estimation, it is characterized in that: described interlace operation changes the order of burst, elder generation's writing in the matrix the data order by row, N data of every row, write the M row altogether, read by row, the data block of establishing before and after interweaving is q=(q again ₀, q ₁, q ₂, q ₃, Λ q _MN-1) ^TAnd g=(g ₀, g ₁, g ₂, g ₃, Λ g _MN-1) ^T, wherein T represents vectorial transposition, then:

g _n·M+m＝q _m·N+n，(0≤m≤M-1，0≤n≤N-1)。

4. the pilot tone that is used for frequency domain equalization according to claim 1 is inserted the method with channel parameter estimation, it is characterized in that: described to add the Cyclic Prefix operation be operation to the data block g after interweaving, and establishing circulating prefix-length is N _g, with the last N of g _gIndividual signal replication is to the front end of data block g, and the length of the data block g after interweaving equals N _B, then generating a length that includes Cyclic Prefix is N _s=N _B+ N _gData block f, promptly

f＝T _CP·g

Wherein

$T_{\mathrm{CP}}=\left[\begin{array}{c}{I}_{\mathrm{CP}}\\ I_{N_B}\end{array}\right]$

Use I _NBExpression N _B* N _BUnit matrix, use I _CPExpression I _NBLast N _gOK, T _CPBe by I _CPAnd I _NBThe matrix of combining and forming.

Images