CN103647620A - QR decomposition method-based channel pre-coding method in LTE-A network - Google Patents

Abstract

The invention discloses a QR decomposition method-based channel pre-coding method in an LTE-A network. The method comprises the following steps that: a matrix which can reflect channel quality information is constructed according to a matrix obtained through channel estimation and codebook requirements; QR decomposition is performed on the matrix which can reflect channel quality information; and minimum elements on main diagonals of R matrixes under each codebook are compared; and an index corresponding to an R matrix of which the square of the minimum main diagonal element is maximum is selected as a pre-coding matrix. With the QR decomposition method-based channel pre-coding method in the LTE-A network of the invention adopted, the complexity of traditional pre-coding matrix solution can be decreased to a great extent. The QR decomposition method-based channel pre-coding method in the LTE-A network is especially applicable to a situation in which the number of antennas is extended to four or even eight.

Description

Channel method for precoding based on QR decomposition method in LTE-A network

Technical field

The present invention relates to communication network technology field, particularly the channel method for precoding based on QR decomposition method in a kind of LTE-A network.

Background technology

In LTE system, MIMO scheme is supported at most N _t* N _r=4 * 4 antennas.In LTE-A system, transmission mode 9(Transmission Mode9 has been proposed)) support at most N _t* N _r=8 * 8 antenna MIMO transmission.[0003] to feed back optimal method be that SVD decomposes to LTE PMI, because equal the right singular matrix of channel matrix H when pre-coding matrix, just mimo channel can be converted into N _tthe individual parallel channel being independent of each other, has directly eliminated the interference between antenna in community.But SVD decomposition algorithm complexity is very high, be unfavorable for realizing.The general method of asking characteristic value or maximum power that adopts in existing algorithm.When antenna amount is N _t* N _r=2 * 2, general characteristic value or the maximum power method of adopting of PMI feedback in engineering.If antenna mode is increased to N _t* N _r=4 * 4, first two method complexity is too high.For the method for characteristic value, ask PMI need to solve unary biquadratic equation, even when the configuration of LTE-A system antenna is increased to N _t* N _r=8 * 8 need to solve Eight equation of element one.The algorithm complex that monobasic polynomial equation solves is quite high, and iterations is more and accurate not.For the algorithm of maximum power, also need 4 * 4 matrix or 8 * 8 complex matrix to invert, general matrix inversion be real matrix invert and also complexity very high.Based on several defects need to be to existing algorithm improvement above.

Summary of the invention

For above deficiency of the prior art, the object of the present invention is to provide the channel method for precoding based on QR decomposition method in a kind of LTE-A network of simplifying the complexity that traditional pre-coding matrix solves.Technical scheme of the present invention is as follows:

A channel method for precoding based on QR decomposition method in LTE-A network, it comprises the following steps:

101, obtain the current order of LTE-A network channel indication RI, corresponding n the alternative PMI code book W of this order indication RI, and to obtain alternative pre-coding matrix corresponding to this order indication RI be W _i, adopt algorithm for estimating to estimate to draw channel matrix H to LTE-A network channel, wherein the dimension of channel matrix H is N _r* N _t;

102, the alternative pre-coding matrix W obtaining according to step 101 _iand channel matrix H, the weighting channel matrix H W of structure LTE-A network channel;

103, adopt Gram-Schmidt Orthogonal Transformation Method to carry out QR decomposition to the weighting channel matrix H W in step 102, breakdown is:

$\begin{array}{c}{\mathrm{HW}}_i=\mathrm{QR}\\ ={\left(\begin{array}{cccc}{q}_1& q_2& q_3& q_4\end{array}\right)}^T\left[\begin{array}{cccc}{r}_{11}& r_{12}& r_{13}& r_{14}\\ 0& r_{22}& r_{23}& r_{24}\\ 0& 0& r_{33}& r_{34}\\ 0& 0& 0& r_{44}\end{array}\right]\end{array};$

104, the element r on the R matrix leading diagonal after QR decomposes in obtaining step 103 _ii, and calculate

wherein

represent the element r on R matrix leading diagonal _iimould square minimum value;

What 105, under each alternative PMI code book W of comparison, through QR, decomposition obtained

select under each alternative PMI code book W maximum PMI index, and export corresponding PMI, Output rusults according to this PMI index; If two alternative PMI code books

be worth identically, export that corresponding PMI that PMI index is less, Output rusults.

In step 102, LTE-A network weighting channel matrix also comprises covariance matrix $R_{\mathrm{TX}}=\frac{1}{\mathrm{Nkl}}\underset{n=0}{\overset{N}{\mathrm{\Σ}}}\underset{k=0}{\overset{k-1}{\mathrm{\Σ}}}\underset{l=0}{\overset{l-1}{\mathrm{\Σ}}}{\left({\hat{H}}_{k,l}^{n}W\right)}^H\left({\hat{H}}_{k,l}^{n}W\right)$ And maximum power matrix $\mathrm{SINR}\left(k\right)=\frac{1}{{\mathrm{\σ}}^2{[W^H{H}^H\mathrm{HW}+{\mathrm{\σ}}^{2}I]}_{k,k}^{-1}},$ Wherein N represents subframe renewal number once, and (k, l) represents the position of resource particle, and I is unit matrix, σ ²represent variance.

Advantage of the present invention and beneficial effect are as follows:

The present invention has utilized QR to decompose and the observating characteristic of original matrix and the similitude that SVD decomposes and QR decomposes, and the thought of asking the characteristic value of matrix is converted into and asks the QR of matrix to decompose, and obtains a upper triangular matrix R matrix.The main diagonal element of R matrix represents primary user's power, can make the elements in a main diagonal maximum and descending arrangement successively in the hope of an optimum R matrix.Thereby eliminated the interference of other user to primary user.The matrix obtaining according to channel estimating and code book need to be constructed the matrix of a reaction channel quality information, and it is carried out to QR decomposition, under each code book of comparison, minimum element on the leading diagonal of R matrix, selects a square maximum index corresponding to that R matrix for minimum the elements in a main diagonal to be selected pre-coding matrix.This invention has reduced the complexity that traditional pre-coding matrix solves to a great extent, especially goes for antenna amount and expands to the even situation of eight antennas of four antennas.

Accompanying drawing explanation

The system block diagram that Fig. 1 the present invention adopts.

The schematic diagram of the precoding feedback method that Fig. 2 provides for example of the present invention.

The another kind of precoding feedback schematic diagram that Fig. 3 provides for example of the present invention.

Another precoding feedback schematic diagram that Fig. 4 provides for example of the present invention.

Embodiment

The invention will be further elaborated below in conjunction with accompanying drawing, to provide the embodiment of an indefiniteness.

Shown in Fig. 1-Fig. 4, according to system, establishing number of transmit antennas is N _t, reception antenna number is N _r.By system block diagram 1, known, the transmitted signal of a bit stream is formed to M subflow after precoder coding.Suppose that a certain moment transmitting terminal symbolic vector is and meet

s _kthrough precoder and N _tthe pre-coding matrix W of * M _kthe length of generation that multiplies each other is N _tvector

accept vector this moment:

$y_k=\sqrt{\frac{{\mathrm{\ϵ}}_s}{M}}{\mathrm{HW}}_k{s}_k+n_k$

Y wherein _kfor receiving vector, noise n _kmeeting average is that 0 variance is N ₀gaussian Profile.Need to be from codebook set w={W ₁, W ₂k,W _na best code book W of middle selection _kas feedback code book.The average error rate of each subflow is weighed by minimum SNR, and the SNR of k subflow is expressed as follows:

${\mathrm{SNR}}_k^{\left(\mathrm{ZF}\right)}=\frac{{\mathrm{\ϵ}}_n}{{\mathrm{MN}}_0{\left[W^H{H}^H\mathrm{HW}\right]}_{k,k}^{-1}}$

${\mathrm{SNR}}_k^{\mathrm{MMSE}}=\frac{{\mathrm{\ϵ}}_s}{{\mathrm{MN}}_0[W^H{H}^H\mathrm{HW}+({\mathrm{MN}}_0/{\mathrm{\ϵ}}_s)I_m{]}_{k,k}^{-1}}-1$

The minimum value that the SNR of each subflow is corresponding if find must meet:

$\begin{array}{c}{\mathrm{SNR}}_k=\underset{1\≤k\≤M}{\min }{\mathrm{SNR}}_k\\ \≥{\mathrm{\λ}}_{\min }^2\left\{\mathrm{HW}\right\}\frac{{\mathrm{\ϵ}}_s}{{\mathrm{MN}}_0}\end{array}$

λ wherein _minsingular value for HW.

The present invention is to asking PMI feedback index example to improve as follows based on svd algorithm:

Owing to requiring, the singular value algorithm complex of HW is very high, and existing SVD decomposes and needs the iteration by QR, the singular value matrix obtaining is to adopt the method for approaching, and not only algorithm complex is very high but also accuracy is not high yet.

According to QR decomposed to R matrix observing the mould of the main diagonal element of R matrix, it is the projection to vector space Q.The selection of optimum R matrix can be resisted interference effectively, its elements in a main diagonal square be primary user's power square.

QR decomposes the plurality of advantages of bringing, so the present invention adopts QR decomposition to solve Optimal Feedback code book.

It is as follows that example of the present invention improves step, and following instance illustrates all for N _r* N _t=4 * 4 situation is set forth.Suppose the channel matrix that H estimates through LS channel estimation method, its dimension is N _r* N _t.

Step 101: select the alternative PMI code book W that current RI is corresponding.

Suppose that H matrix is N _r* N _t=4 * 4 matrix:

$H=\left[\begin{array}{cccc}{h}_{11}& h_{12}& h_{13}& h_{14}\\ h_{21}& h_{22}& h_{23}& h_{24}\\ h_{31}& h_{32}& h_{33}& h_{34}\\ h_{41}& h_{42}& h_{43}& h_{44}\end{array}\right]$

Suppose that the alternative pre-coding matrix of RI=4 is W _i(i=0...15):

$W_i=\left[\begin{array}{cccc}{w}_{11}& w_{12}& w_{13}& w_{14}\\ w_{21}& w_{22}& w_{23}& w_{24}\\ w_{31}& w_{32}& w_{33}& w_{34}\\ w_{41}& w_{42}& w_{43}& w_{44}\end{array}\right]$

?

${\mathrm{HW}}_i=\left[\begin{array}{cccc}{h}_{11}{w}_{11}+h_{12}{w}_{21}+h_{13}{w}_{31}+h_{14}{w}_{41}& h_{11}{w}_{12}+h_{12}{w}_{22}+h_{13}{w}_{32}+h_{14}{w}_{42}& h_{11}{w}_{13}+h_{12}{w}_{23}+h_{13}{w}_{33}{+h}_{14}{w}_{43}& h_{11}{w}_{14}+h_{12}{w}_{24}+h_{13}{w}_{34}+h_{14}{w}_{44}\\ h_{21}{w}_{11}+h_{22}{w}_{21}+h_{23}{w}_{31}+h_{24}{w}_{41}& h_{21}{w}_{12}+h_{22}{w}_{22}+h_{23}{w}_{32}+h_{24}{w}_{42}& h_{21}{w}_{13}+h_{22}{w}_{23}+h_{23}{w}_{33}+h_{24}{w}_{43}& h_{21}{w}_{14}+h_{22}{w}_{24}+h_{23}{w}_{34}+h_{24}{w}_{44}\\ h_{31}{w}_{11}+h_{32}{w}_{21}+h_{33}{w}_{31}+h_{34}{w}_{41}& h_{31}{w}_{12}+h_{32}{w}_{22}+h_{33}{w}_{32}+h_{34}{w}_{42}& h_{31}{w}_{13}+h_{32}{w}_{23}+h_{33}{w}_{33}+h_{34}{w}_{43}& h_{31}{w}_{14}+h_{32}{w}_{24}+h_{33}{w}_{34}+h_{34}{w}_{44}\\ h_{41}{w}_{11}+h_{42}{w}_{21}+h_{43}{w}_{31}+h_{44}{w}_{41}& h_{41}{w}_{12}+h_{42}{w}_{22}+h_{43}{w}_{32}+h_{44}{w}_{42}& h_{41}{w}_{13}+h_{42}{w}_{23}+h_{43}{w}_{33}+h_{44}{w}_{43}& h_{41}{w}_{14}+h_{42}{w}_{24}+h_{43}{w}_{34}+h_{44}{w}_{44}\end{array}\right]$

Step 102: HW is had to little large arrangement of arriving by power, can find optimum R matrix, the main diagonal element power of R matrix is maximum.

Step 103: utilize Gram-Schmidt orthogonal transform to carry out QR decomposition to HW, Q is unit unitary matrice, and R is upper triangular matrix.Now the main diagonal element of upper triangular matrix R is always maximum

$\begin{array}{c}{\mathrm{HW}}_i=\mathrm{QR}\\ ={\left(\begin{array}{cccc}{q}_1& q_2& q_3& q_4\end{array}\right)}^T\left[\begin{array}{cccc}{r}_{11}& r_{12}& r_{13}& r_{14}\\ 0& r_{22}& r_{23}& r_{24}\\ 0& 0& r_{33}& r_{34}\\ 0& 0& 0& r_{44}\end{array}\right]\end{array}$

Step 104: select under each W

Step 105, what relatively under each alternative code book, QR decomposition obtained

select under each W maximum PMI index is required, if two code books value is identical gets that value of feedback as PMI that PMI index is less.

The another example of the present invention is to utilizing the method for characteristic value to improve:

If antenna mode is increased to N _t* N _r=4 * 4, adopt the method for characteristic value to ask PMI need to solve unary biquadratic equation, if the configuration of LTE-A system antenna is increased to N _t* N _r=8 * 8 need to solve Eight equation of element one.The method that existing monobasic polynomial equation adopts is as newton's down-hill method, and the method based on QR iteration, has all used the means of approaching, and not only the very low while algorithm complex of accuracy is very high, close to O (n ³).

The present invention utilizes and directly carries out the simplicity of QR decomposition algorithm and to characteristic value, ask PMI feedback index to improve the observed result of R matrix.First suppose that H passes through the channel matrix that LS channel estimation method estimates, construct its covariance matrix to be $R_{\mathrm{TX}}=\frac{1}{\mathrm{Nkl}}\underset{n=1}{\overset{N}{\mathrm{\Σ}}}\underset{k=0}{\overset{k-1}{\mathrm{\Σ}}}\underset{l=0}{\overset{l-1}{\mathrm{\Σ}}}{\left({\hat{H}}_{k,l}^{n}W\right)}^H\left({\hat{H}}_{k,l}^{n}W\right),$ The every N of covariance matrix subframe upgraded once, can reflect like this channel long-time statistical characteristic, and resource particle is expressed as (k, l).

Step 201: calculate according to all alternative code books under current RI:

$R_{\mathrm{TX}}=\frac{1}{\mathrm{Nkl}}\underset{n=0}{\overset{N}{\mathrm{\Σ}}}\underset{k=0}{\overset{k-1}{\mathrm{\Σ}}}\underset{l=0}{\overset{l-1}{\mathrm{\Σ}}}{\left({\hat{H}}_{k,l}^{n}W\right)}^H\left({\hat{H}}_{k,l}^{n}W\right)$

Step 202: change R _tXask characteristic value unary biquadratic equation ask method.Directly adopt the method for Gram-Schmidt to matrix R _tXmaking QR decomposes.

Step 204: select under each W

r _iifor matrix R _txrectangular projection under vector matrix Q.Step 205, what relatively under each alternative code book, QR decomposition obtained

select under each W

maximum PMI index is required.If two code books

value is identical gets less that of PMI index as PMI value of feedback.

The another example of the present invention is that the method for maximum power is improved to utilizing maximum SINR:

According to theorem 1: suppose A ∈ ^{m * n}reversible, A can resolve into A=QR wherein Q be unitary matrice, i.e. QQ ^h=E, R is upper triangular matrix.

According to theorem 2 hypothesis R=(R _ij) _{n * n}upper triangular matrix, R _ij=0, when i>j, and R _ji≠ 0,1≤i≤n, R ^-1=(σ _ij) _{n * n}can draw by algorithm below:

(1) σ _ij=0, when i>j, i.e. R ^-1also be upper triangular matrix;

(2) ${\mathrm{\σ}}_{\mathrm{kk}}=R_{\mathrm{kk}}^{-1},1\≤k\≤n$

(3) ${\mathrm{\σ}}_{k,k+m}=-\underset{j=1}{\overset{n}{\mathrm{\Σ}}}{R}_{k,k+j}{\mathrm{\σ}}_{k+j,k+m}/R_{\mathrm{kk}},1\≤k\≤n-m$

Theorem 3: suppose that A is reversible, and A=QR, wherein Q is unitary matrice, Q ^-1=R ^-1q ^h, Q ^hit is the conjugate transpose of Q.

According to the method for inverting above, directly inverting of matrix A is converted into inverting of matrix R, reduced complexity a lot, and can directly obtain the contrary of complex matrix.

The present invention utilizes QR method to ask the step of improving one's methods of precoding index as follows to maximum power:

Step 301: the SINR that obtains k subflow in M subflow according to maximum power algorithm is as follows

$\mathrm{SINR}\left(k\right)=\frac{1}{{\mathrm{\σ}}^2{[W^H{H}^H\mathrm{HW}+{\mathrm{\σ}}^{2}I]}_{k,k}^{-1}},$

Step 302: make A=W ^hh ^hhW+ σ ²i, wherein I is unit matrix, it is that 0 variance is σ that simultaneity factor is introduced average ²white Gaussian noise.Directly utilize theorem 3 to carry out QR decomposition to matrix A, the reverse of matrix A turns to contrary to matrix R.

Step 304: for each code book, select SINR minimum in each subflow.

Step 305: in the minimum SINR selecting in each code book, go maximum SINR for PMI index.If the same get less that of PMI index as PMI value of feedback.

These are only preferred embodiments explanations more of the present invention, QR decomposition method, can reduce algorithm complex effectively in antenna amount increase, and the performance that can guarantee system as throughput simultaneously.The part portion that the R matrix that utilizes QR decomposition to obtain carries out precoding selection is the scope of protection of the invention.All modifications within central idea of the present invention, replace improvement etc., within being all included in protection scope of the present invention.

Claims (2)

1. the channel method for precoding based on QR decomposition method in LTE-A network, is characterized in that comprising the following steps:

101, obtain the current order of LTE-A network channel indication RI, corresponding n the alternative PMI code book W of this order indication RI, and to obtain alternative pre-coding matrix corresponding to this order indication RI be W _i, adopt algorithm for estimating to estimate to draw channel matrix H to LTE-A network channel, wherein the dimension of channel matrix H is N _r* N _t;

102, the alternative pre-coding matrix W obtaining according to step 101 _iand channel matrix H, the weighting channel matrix H W of structure LTE-A network channel;

103, adopt Gram-Schmidt Orthogonal Transformation Method to carry out QR decomposition to the weighting channel matrix H W in step 102, breakdown is:

$\begin{array}{c}{\mathrm{HW}}_i=\mathrm{QR}\\ ={\left(\begin{array}{cccc}{q}_1& q_2& q_3& q_4\end{array}\right)}^T\left[\begin{array}{cccc}{r}_{11}& r_{12}& r_{13}& r_{14}\\ 0& r_{22}& r_{23}& r_{24}\\ 0& 0& r_{33}& r_{34}\\ 0& 0& 0& r_{44}\end{array}\right]\end{array};$

104, the element r on the R matrix leading diagonal after QR decomposes in obtaining step 103 _ii, and calculate

wherein represent the element r on R matrix leading diagonal _iimould square minimum value;

What 105, under each alternative PMI code book W of comparison, through QR, decomposition obtained

select under each alternative PMI code book W

maximum PMI index, and export corresponding PMI, Output rusults according to this PMI index; If two alternative PMI code books

be worth identically, export that corresponding PMI that PMI index is less, Output rusults.

2. the channel method for precoding based on QR decomposition method in LTE-A network according to claim 1, is characterized in that: in step 102, LTE-A network weighting channel matrix also comprises covariance matrix $R_{\mathrm{TX}}=\frac{1}{\mathrm{Nkl}}\underset{n=0}{\overset{N}{\mathrm{\Σ}}}\underset{k=0}{\overset{k-1}{\mathrm{\Σ}}}\underset{l=0}{\overset{l-1}{\mathrm{\Σ}}}{\left({\hat{H}}_{k,l}^{n}W\right)}^H\left({\hat{H}}_{k,l}^{n}W\right)$ And maximum power matrix $\mathrm{SINR}\left(k\right)=\frac{1}{{\mathrm{\σ}}^2{[W^H{H}^H\mathrm{HW}+{\mathrm{\σ}}^{2}I]}_{k,k}^{-1}},$ Wherein N represents subframe renewal number once, and (k, l) represents the position of resource particle, and I is unit matrix, σ ²represent variance.

Images