CN101917368B - Lattice reduction-based multiple input multiple output (MIMO) detection soft output method - Google Patents

Abstract

The invention relates to a lattice reduction-based multiple input multiple output (MIMO) detection soft output method. The method comprises the following steps of: firstly, acquiring an ordered list of transform domain signals aiming at different transformation matrixes and storing the ordered list for inquiring during detection; secondly, when executing the lattice reduction-based MIMO detection at each time, applying the lattice reduction to a channel response matrix to obtain a lattice reduction base and the transformation matrixes and acquiring the corresponding ordered list of the transform domain signals by using the transformation matrixes; thirdly, executing QR Decomposition and M algorithm (QRM) detection by using the lattice reduction base and the ordered list of the transform domain signals to acquire survival vectors and the corresponding weight value of each survival vector; and finally, determining the soft information of each bit by using the survival vectors and the weight value of each survival vector. The method has the advantages of effectively reducing the number of nodes which need to calculate the weight values during the QRM detection by using the ordered list stored in advance, greatly reducing the complexity of the calculation, along with no need of model extension and easy implementation; and compared with a method for acquiring the set of the survival vectors through the disturbance of the optimum survival vector, the method has higher accuracy.

Description

The soft output method that detects based on the MIMO of lattice reduction

Technical field

The present invention relates to a kind of soft output method of the many antenna detection of MIMO based on lattice reduction, belong to the technical field of MIMO communication system.

Background technology

Introducing first many antennas sends and many antenna receptions MIMO(Multiple Input Multiple Output) system model: there is N in hypothesis space multiplexed MIMO system _tTransmit antennas and N _rThe root reception antenna, wireless channel is diffuse scattering and quasi-static flat fading, the decline of experiencing between every transmit antennas and the reception antenna can be regarded mutual independence as, and receiver can carry out the estimation of ideal communication channel, and then the system transmissions model can be expressed as: y=Hx+n; In the formula, receive signal phasor

Its dimension is N _r* 1; Channel response matrix

Its dimension is N _r* N _t, in the formula, the channel gain h between transmitting antenna j and the reception antenna i _IjBe that an average is zero, variance is 1 multiple Gaussian random variable; Vector transmits

Its dimension is N _t* 1, be b with bit vectors corresponding to vector x that transmit, the dimension of b is N _tLog ₂M _c, M _cThe expression order of modulation; White Gauss noise on the reception antenna is

Its average is zero, and variance is σ ²

Lattice reduction LR(lattice reduction) algorithm (referring to " Closest point search in lattices[J] " publish in " IEEE Transactions on Information Theory " 2002,48:2201-2214) can effectively reduce the respectively correlation between the row of matrix, so that respectively being listed as much as possible quadrature and waiting mould of the reduced basis that obtains through lattice reduction namely reduced the conditional number of this matrix.Famous Lenstra, Lenstra, Lovsaz(LLL) lattice reduction implementation algorithm (referring to " Lattice-Reduction-Aided Detectors for MIMO Communication Systems[C] " publish in " IEEE Globecom " 2002:424-428) obtains extensive use because it has binomial low complex degree.But, LLL can only be that the channel matrix of real number carries out lattice reduction to value, for the complex channel matrix, just must carry out model extension, is converted into real number matrix and carries out corresponding concrete lattice reduction implementation algorithm again.And Complexity LLL(CLLL) implementation algorithm (referring to " Complex Lattice Reduction Algorithms for Low-Complexity MIMO Detection[C] " publish in " IEEE Globecom " 2005:2953-2957) can be directly carry out lattice reduction to value for the channel matrix of plural number, its algorithm thinking and LLL are similar, but more are applicable to real system because its complexity is low.

Lattice reduction is through being usually used in carrying out linearity test (being that ZF detects ZF(Zero-Forcing) and least mean-square error detection MMSE(Minimum Mean Square Error as receiving terminal)) or non-linear detection algorithm before pretreatment operation, increased this pretreated detection algorithm and all be called as LRA(lattice reduction aided) detection algorithm.

Referring to Fig. 1, the detection block diagram when introducing receiving terminal and adopting LR to carry out preliminary treatment comprises following 4 steps:

(1) carrying out to received signal suitable translation and equivalence transformation, is complex integers to guarantee equivalent transmitted signal.The processing formula of this moment is: $y^{\′}=y+{\mathrm{\βH}1}_{N_t\×1}=\mathrm{\αH}(1/\mathrm{\α}(x+{\mathrm{\β}1}_{N_t\×1}))+n=\mathrm{\αHs}+n,$ In the formula, suppose that transmitting terminal adopts square QAM modulation, then translation numerical value ${\mathrm{\βH}1}_{N_t\×1}=\sqrt{3/2\left((M_c-1)\right)}(\sqrt{M_c}-1+(\sqrt{M_c}-1)*j){H1}_{N_t\×1},$ The translation coefficient value $\mathrm{\β}=\sqrt{3/2\left((M_c-1)\right)}(\sqrt{M_c}-1+(\sqrt{M_c}-1)*j),$ M _cBe order of modulation; The equivalence transformation coefficient

For other modulation system, need to design α according to the concrete form of planisphere, the value of β is to satisfy the requirement that equivalent transmitted signal is complex integers;

For dimension is N _t* 1, value is 1 column vector;

s _i(natural number i is transmitting antenna port sequence number, and its maximum is N _t) the real part value

With the imaginary part value

All belong to set

(2) channel response matrix is carried out lattice reduction, obtain reduced basis H _RedTransformation matrix T:H with correspondence _Red=HT, like this, the processing formula of above-mentioned steps (1) just equivalence is following formula: y '=α HTT ^-1S+n=α H _RedZ+n; In the formula, establish

Be the inverse matrix T through transformation matrix ^-1Transmitted signal vector in the transform domain after the conversion, its real part and imaginary part value all are complex integers; Use M _iExpression z _iMight value number, z then _iThe set that forms of all possible value

Be the set of i dimension transform domain lattice point.

(3) carry out linearity or non-linear detection according to the system model formula in the above-mentioned steps (2) of reduced basis after to equivalence, just obtain adjudicating vector

Below, with nonlinear QRM(QR Decomposition and M algorithm) detect (also being known as K-Best detects), and survival vector number M=1 is that example describes:

Reduced basis α H after elder generation's parity price conversion _RedCarry out ORTHOGONAL TRIANGULAR shape and decompose (Orthogonal-Triangular Decomposition often is called as QR and decomposes), obtain α H _RedCorresponding unitary matrice Q and upper triangular matrix R:aH _Red=QR; And to the associate matrix Q of the system model formula premultiplication Q in the above-mentioned steps (2) ^H, obtain to detect vector

Then have: ρ=Q ^HY '=Q ^H(α H _RedZ+n)=and Rz+n ', in the formula, upper triangular matrix White Gauss noise n ' after the processing=Q ^HN.

Then, according to following computing formula A: $\left\{\begin{array}{c}{\hat{z}}_{N_t}=Q({\mathrm{\&rho;}}_{N_t}/r_{N_t,N_t})\\ {\hat{z}}_{N_t-1}=Q(({\mathrm{\&rho;}}_{N_t-1}-r_{N_t-1,N_t}{\hat{z}}_{N_t})/r_{N_t-1,N_t-1})\\ \&CenterDot;\\ \&CenterDot;\\ \&CenterDot;\\ {\hat{z}}_1=Q(({\mathrm{\&rho;}}_1-\underset{d=2}{\overset{{\mathrm{<figure-callout\; id="1"\; label="N\; t\; r"\; filenames="HSA00000212449800012.png,HSA00000212449800023.png"\; state="\{\{state\}\}">N</figure-callout>}}_t}{\mathrm{\&Sigma;}}}{r}_{}{}_{1,d}{\hat{z}}_d)/{\mathrm{<figure-callout\; id="1"\; label="r"\; filenames="HSA00000212449800012.png,HSA00000212449800023.png"\; state="\{\{state\}\}">r</figure-callout>}}_{\mathrm{1,1}})\end{array}\right.,$ Interference by iteration is eliminated, and obtains successively respectively tieing up from the bottom to top the estimated value of transform domain transmitted signal; In the formula, Q () expression is chosen respectively its immediate integer with real part and the imaginary part of variable.

(4) with the survival vector

Obtain through transformation matrix T conversion

And right

Implement linear inverse transformation, the transmission symbolic vector finally obtains surviving

And the transmission bit vectors of surviving

Because M=1, thus among the formula A to z _iOnly keep a possible value

The survival that then obtains according to two formula in the step (4) sends symbolic vector

Be exactly transmitted signal vector x firmly declare the result.If obtain the soft result of declaring, can in step (3), keep M(M 〉=2) individual survival vector The set of its composition is called the survival set of vectors

Carry out iteration according to computing formula A, and can regard search procedure (referring to shown in Figure 2) to an inverted tree as every layer of process that keeps M survival vector, the height of tree is number of transmit antennas N _tThe i-1 node layer that is derived by the i node layer is called child node, and the i node layer then is the father node of this child node.For example, the root node root among the figure is the father node of all node of i=4 layer, and all node of i=4 layer all are the child nodes of root.The final purpose of search is to find a node of weights minimum, the survival vector that this node and its each layer father node form at the bottom of tree

It is exactly the estimated value of transform domain transmitted signal vector z.And the weights of every layer of upper each node of tree can represent with form with following adding up: $e_i=\underset{k=i}{\overset{\mathrm{Nt}}{\mathrm{\&Sigma;}}}{|r_{k,k}{\hat{z}}_k-({\mathrm{\&rho;}}_k-\underset{d=k+1}{\overset{\mathrm{Nt}}{\mathrm{\&Sigma;}}}{r}_{i,d}{\hat{z}}_d)|}^2,$ r _{I, d}The i that is upper triangular matrix R is capable, d column element value, ρ _kFor detecting vector

K first number value.

For follow-up description convenience, will

The central point that is called the i layer.By this formula as can be known, central point Value depend on (i+1 ..., N _t) father node of each layer.Hereinafter will carry out list ordering according to the concrete value of central point, no longer explain herein.

Enter the PCT Patent Application Publication text in announcement stage take " Ge Ji " as keyword in the search of Chinese patent board web, obtain following three pieces of patent applications relevant with communication, be analyzed as follows respectively:

The lattice reduction auxiliary detection of algorithm " the using modified Lenstra-Lenstra-Lovasz(LLL) " (CN200780000781.2) introduces a kind of LLL algorithm of carrying out in wireless communication system, in the time of with definite lattice reduction, obtain the lattice reduction inverse of a matrix.It is that the corresponding steps of inserting finding the inverse matrix in the LLL algorithm realizes, also provides a kind of easy method of asking the lattice reduction inverse of a matrix.

" Wireless Telecom Equipment " (CN200780000779.5) is described in the OFDM-MIMO system, and the channel matrix of a certain subcarrier is multiplied each other with the corresponding transformation matrix of the channel matrix of last subcarrier, carries out lattice reduction algorithm again.This algorithm is considered the correlation of adjacent carrier respective channels matrix, can effectively reduce the complexity of lattice reduction.

" in the lattice reduction mimo system soft-decision produce " (CN200780000791.6).In the wireless communication system based on the lattice reduction auxiliary receiver, determine by the following method the soft estimation of institute's transmitting ratio paricular value according to received signal: lattice reduction is applied to the channel response matrix that obtains via channel estimating, and come the balanced reception signal according to the reduced basis channel, generate the vector set of surviving based on optimum survival vector, and to correspondence emission symbolic vector of each vector of surviving, and determine to launch the probability that each bit has been launched based on receiving signal; It obtains the set of vectors of surviving by optimum survival vector is carried out disturbance.

In addition, thesis for the doctorate " the detection algorithm research in the mimo system " (Xian Electronics Science and Technology University, a kind of soft information calculations method that is applicable to the searching class detection algorithm is proposed 2009:71-85), in transform domain, determine approximate transform domain lattice point set according to the boundary value of every layer signal first, then carry out stack and detect.

Analyze discovery for the patent application that above-mentioned several typical LR are relevant, existing patent application relates to content and mainly is divided into two classes: a class is to simplify the LRA testing process, namely simplify the lattice reduction implementation procedure of channel matrix, as simplify the process that obtains reduced basis, or simplify the inversion process of transformation matrix.Another kind of is how to find the solution soft information (the Chinese patent storehouse is only searched out one).

Analyze above-mentioned patent application and paper, existing soft information derivation algorithm respectively has characteristics as can be known again.In order to calculate soft information, key is how to obtain a plurality of survival vectors.Method according to obtaining the survival vector can will have LRA(Lattice Reduction Aided now) the soft information derivation algorithm that detects is divided into two classes:

The first kind, optimum survival vector carry out disturbance and obtain the set of vectors of surviving:

In (CN200780000791.6), for the formula in the described step (2): y '=α H _RedZ+n by aforementioned formula A, obtains optimum survival vector first And pass through optimum survival vector

Carry out the vector disturbance, to obtain other required survival vector.A kind of fairly simple processing method is right successively

The real part of each element and imaginary part carry out respectively ± 1 and ± disturbance of j, obtain new survival vector.Will there be 8 survival vectors after the disturbance in explanation as an example of two transmit antennas example, and altogether 9 the survival vectors that obtain are respectively: ${\hat{z}}_1=\left[\begin{array}{c}{\hat{z}}_1\\ {\hat{z}}_2\end{array}\right],$ ${\hat{z}}_2=\left[\begin{array}{c}{\hat{z}}_1+1\\ {\hat{z}}_2\end{array}\right],$ ${\hat{z}}_3=\left[\begin{array}{c}{\hat{z}}_1-1\\ {\hat{z}}_2\end{array}\right],$ ${\hat{z}}_4=\left[\begin{array}{c}{\hat{z}}_1\\ {\hat{z}}_2+1\end{array}\right],$ ${\hat{z}}_5=\left[\begin{array}{c}{\hat{z}}_1\\ {\hat{z}}_2-1\end{array}\right],$ ${\hat{z}}_6=\left[\begin{array}{c}{\hat{z}}_1+j\\ {\hat{z}}_2\end{array}\right],$ ${\hat{z}}_7=\left[\begin{array}{c}{\hat{z}}_1-j\\ {\hat{z}}_2\end{array}\right],$ ${\hat{z}}_8=\left[\begin{array}{c}{\hat{z}}_1\\ {\hat{z}}_2+j\end{array}\right]$ With ${\hat{z}}_9=\left[\begin{array}{c}{\hat{z}}_1\\ {\hat{z}}_2-j\end{array}\right].$

Then, each survival vector is obtained corresponding survival transmission symbolic vector according to two formula of above-mentioned steps (4) respectively (m=1,2 ... M) and survive to send bit vectors

And then the probability that has been sent out of each bit that obtains sending.

Through top introduction, this soft information algorithm is to obtain the set of vectors of surviving by optimum survival vector is carried out disturbance as can be known, and its operating process is simple, relatively is suitable for the few LRA linearity test of number of transmit antennas.But along with the increase of number of transmit antennas, it is large that error can become.In addition, because this algorithm is when obtaining other survival vector by optimum survival vector, not consider the correlation between the different transmit antennas port signal, carry out soft information when finding the solution when it is used for the LRA non-linear detection, soft information error is very large.

Equations of The Second Kind, obtain the survival vector by signal boundary:

As above the review literary composition is for the searching class non-linear detection algorithm, the stack detection algorithm with soft Output rusults that the lattice reduction that provides is auxiliary.It is that the transform domain lattice point that obtains first to be similar to is gathered, and carries out accordingly stack and detect and minimize every node layer weights, to obtain the survival set of vectors.The accuracy of this soft information calculations method is high.But, when stack detects, all child nodes have all been carried out weights calculating, cause computation complexity very high; In addition, its processing domain is real number field, usually needs to carry out model extension for real system, is not easy to Project Realization.

Therefore, how to seek a kind ofly can solve the soft output method based on the many antenna detection of MIMO of the lattice reduction prior art defective, simple general-purpose, just become the focus that scientific and technical personnel in the industry pay close attention to.

Summary of the invention

In view of this, the purpose of this invention is to provide a kind of soft output method of the many antenna detection of MIMO based on lattice reduction, the method is to consider accuracy and the complexity that soft information is found the solution, the soft output method that a kind of simple, the general MIMO that provides detects; The soft accuracy of information that the method obtains is high, can effectively reduce the node number that needs to calculate weights simultaneously, greatly reduces computation complexity.

In order to achieve the above object, the invention provides a kind of soft output method that detects based on the MIMO of lattice reduction, it is characterized in that: at first obtain the sorted lists of transform-domain signals for different transformation matrixs, and it is stored for the detection query; When each MIMO that implements based on lattice reduction detects, first channel response matrix is carried out lattice reduction and obtain reduced basis and transformation matrix, obtain again the sorted lists of corresponding transform-domain signals by transformation matrix; Then utilize the sorted lists of reduced basis and corresponding transform-domain signals to carry out QRM(QR Decomposition and M algorithm) detect, obtain survival vector and weights corresponding to each survival vector; Utilize at last the weights of survival vector and each survival vector, determine the soft information of each bit.

The present invention is a kind of soft output method of the many antenna detection of MIMO based on lattice reduction, the method considers accuracy and the complexity that soft information is found the solution, provide the soft output method of simple general-purpose in detecting based on the MIMO of lattice reduction: obtain first the sorted lists of transform-domain signals for different transformation matrixs, and it is stored can be for inquiry; When each the detection, by current transformation matrix, select the sorted lists corresponding with this transformation matrix, select the survival vector according to sorted lists again, avoided the weights of parton node to calculate.The present invention can reduce the number of the node that needs to calculate weights in the QRM algorithm compared with prior art effectively, greatly reduces computation complexity; And, do not need to carry out model extension, easily realize; Its algorithm carries out disturbance obtain the to survive method of set of vectors with respect to optimum survival vector, and accuracy is higher.

The key of the inventive method is: in the process that obtains candidate vector, effectively utilize the sorted lists of transform-domain signals, so that this soft output method becomes simple.Certainly, to obtain for different transformation matrixs the sorted lists of transform-domain signals in advance, and it be stored for the detection query.When each the detection, first channel response matrix is carried out lattice reduction and obtain reduced basis and transformation matrix, then obtain corresponding sorted lists by transformation matrix, carry out QRM(QR Decomposition and M algorithm according to reduced basis and corresponding sorted lists at last) detect, to obtain the survival vector.

Description of drawings

Fig. 1 is that LRA detects block diagram.

Fig. 2 is inverted tree (N _t=4) schematic diagram.

The detection operating process block diagram that has soft output in Fig. 3 prior art.

Fig. 4 is the soft output method operational flowchart that the present invention is based on the MIMO detection algorithm of lattice reduction.

Fig. 5 (A), (B) are respectively the transform domain planisphere of the first dimensional signal and the transform domain planisphere of the second dimensional signal.

Fig. 6 (A), (B) are respectively two sub-node sequencing tabulating methods.

Embodiment

For making the purpose, technical solutions and advantages of the present invention clearer, the present invention is described in further detail below in conjunction with drawings and Examples.

For the ease of understanding the innovative point of the present patent application, the soft output method in the prior art paper that provides respectively referring to Fig. 3 and Fig. 4 and the present invention is based on two kinds of testing processes of soft output method of the many antenna detection of MIMO of lattice reduction.Can obviously be found out both difference by these two figure: utilized sorted lists owing to the present invention is based on the soft output method of simple general-purpose in the MIMO detection algorithm of lattice reduction, can greatly reduce algorithm complex.Among the figure,

The survival set of vectors,

It is the set of survival bit vector.

Below, specify four operating procedures of the inventive method:

Step 1, obtain the sorted lists of transform-domain signals for different transformation matrix T, and it is stored, with for the detection query.The preparation method of sorted lists comprises following content of operation:

(11) first for different transformation matrix T, obtain respectively the constellation point set in the transform domain.For example, when adopting the QPSK modulation, be s through the constellation point behind the Pan and Zoom _i∈ [0 1 i 1+i].If transformation matrix $T=\left[\begin{array}{cc}1& -1+j\\ 0& 1\end{array}\right],$ Its inverse matrix then $T^{-1}=\left[\begin{array}{cc}1& 1-j\\ 0& 1\end{array}\right]$ The time, signal z=T in the transform domain then ^-1The first dimensional signal of s and the transform domain constellation point of the second dimensional signal set: δ ₁And δ ₂, referring to Fig. 5 (A) with shown in the stain (B).

(12) in order to save the calculating operation of node weights as far as possible, the present invention utilizes prior sorted tabulation, avoids the calculation times of weights.And the list ordering method is not construed as limiting, and introduces a kind of feasible according to being that the list ordering that the method for optiaml ciriterion is carried out is based on central point to the distance of constellation point: i dimension transform domain lattice point is divided into 4M _iIndividual lattice, in the formula, M _iBe the number of i dimension constellation point, natural number i is the sequence number of transmitting antenna port, and its maximum occurrences is emission fate sum N _tM wherein _iIt is the number of i dimension constellation point.According to the position of central point grid of living in, calculate it to the distance of various constellations point, this is tieed up all constellation point sort, apart from the little front that is positioned at, the back that is positioned at apart from large can obtain sorted lists corresponding to this central point.Travel through again the 4M of central point _iPlant the position, the sorted lists of tieing up all signals to obtain this.The present invention can adopt other list ordering method, is not construed as limiting at this.

The planisphere of Fig. 6 (A) and Fig. 5 (A) is corresponding, and central point is (the position one total 48(4M of central point shown in the stain among the figure ₁) individual, just exemplifying a kind of position among the figure describes), 12 constellation point are divided into 48 grids, determine ordering according to the distance to various constellations point of central point, be exactly this ordering number such as the numeral of the mark in the small circle among Fig. 6 (A), be labeled as 1 be exactly the nearest constellation point of distance center point.Fig. 6 (B) is corresponding with Fig. 5 (B).Like this when detecting, according to the precedence centre point

Solution formula

Just can access central point

Judge again the grid position that it is corresponding, just can directly obtain this layer M by inquiry _iThe ordering of individual child node, and do not need the weights of each node are really calculated.

(13) limited because of the value of transformation matrix T, therefore the memory space of sorted lists can be set for different transformation matrixs in advance, when detecting, just can directly select sorted lists according to the value of T like this, this can need certain memory space certainly.

Step 2, carry out based on lattice reduction LRA(Lattice Reduction Aided at every turn) during detection algorithm, channel response matrix H is carried out lattice reduction, obtain reduced basis H _RedWith transformation matrix T, and from the sorted lists of transform-domain signals corresponding to the different transformation matrix T of storage, obtain this transformation matrix T and obtain the corresponding sorted lists of this transformation matrix T.This step comprises following content of operation:

(21) channel response matrix H is carried out lattice reduction, obtain reduced basis H _RedWith transformation matrix T, and this step is not construed as limiting the specific algorithm of lattice reduction;

(22) from the sorted lists of the corresponding transform-domain signals of different transformation matrixs of storage, according to the value of transformation matrix T, obtain sorted lists corresponding to this transformation matrix T.

Step 3, according to the matrix H after approximately subtracting _RedCarry out QRM with corresponding sorted lists and detect, to obtain M survival vector

And this M survival vector Corresponding weights E _mIn the formula, natural number m represent the to survive sequence number of vector, the maximum of m is M.This step comprises following content of operation:

(31) vector y carries out translation to received signal:

Obtain equivalent received signals vector y ', to guarantee equivalent transmitted signal vector as non-negative plural number, wherein, suppose that transmitting terminal adopts square QAM modulation, translation numerical value is ${\mathrm{\&beta;H}1}_{N_t\&times;1}=\sqrt{3/2\left((M_c-1)\right)}(\sqrt{M_c}-1+(\sqrt{M_c}-1)*j){H1}_{N_t\&times;1},$ The translation coefficient value $\mathrm{\&beta;}=\sqrt{3/2\left((M_c-1)\right)}(\sqrt{M_c}-1+(\sqrt{M_c}-1)*j),$ H is channel response matrix, M _cBe order of modulation,

That dimension is N _t* 1 complete 1 vector, N _tBe emission fate sum;

(32) to equivalent received signal phasor y ' according to the following equation:

$y^{\&prime;}=\mathrm{\&alpha;H}(1/\mathrm{\&alpha;}(x+{\mathrm{\&beta;}1}_{N_t\&times;1}))+n=\mathrm{\&alpha;Hs}+n$ Carry out equivalence transformation, to guarantee that equivalent transmitted signal vector is as non-negative complex integers: in the above-mentioned formula, y is for receiving signal phasor, H is channel response matrix, and x is the transmitted signal vector, and s is that value is the equivalent transmitted signal vector of non-negative complex integers, n is white Gauss noise, the equivalence transformation coefficient value

Translation numerical value is:

${\mathrm{\&beta;H}1}_{N_t\&times;1}=\sqrt{3/2\left((M_c-1)\right)}(\sqrt{M_c}-1+(\sqrt{M_c}-1)*j){H1}_{N_t\&times;1},$ H is channel response matrix, M _cBe order of modulation,

Dimension is N _t* 1 complete 1 vector.

(33) the reduced basis α H after the parity price conversion _RedCarry out QR and decompose (Orthogonal-Triangular Decomposition), obtain unitary matrice Q and upper triangular matrix R, and to the associate matrix Q of equivalent received signal phasor y ' premultiplication Q ^H, obtain detecting vector ρ;

(34) successively carry out QRM according to the detection vector ρ that obtains and detect, obtain the survival vector

And this M survival vector

Corresponding weights E _mThis M survival vector Corresponding weights E _mComputing formula be:

In the formula, ρ is for detecting vector, and R is upper triangular matrix,

The expression vector 2 norms square.

Suppose at i layer M node that remains to be

Wherein,

The vector that each layer former generation node forms with it just is m survival vector of i layer:

And this M keeps node and has carried out ordering from small to large according to its weights.Each keeps node (hereinafter referred to as father node) can have

Individual child node.The value of the corresponding central point of each father node then can be carried out from small to large ordering according to the corresponding sorted lists of center position with father and son's node.Therefore with the child node called after

(first subscript i represents its father node at the i layer, and then child node is at the i-1 layer; Second subscript p represents the ordering of the weights size of this father node; The 3rd subscript q represents that this child node is with the ordering in father and son's node set).

Explained later is how from the i-1 layer Select M of the weights minimum to keep node in the individual child node: to calculate first the first father node

First child node z _{I, 1,1}Weights, if its weights are less than the weights of all the other M-1 father node, then with z _{I, 1,1}Deposit in the survival vector tabulation, otherwise, the weights of the first child node of next father node calculated; After if certain child node is stored into survival vector tabulation, then replaced by its next child node with father node, and the weights comparison is carried out in the representative that represents this family and other family; At every turn relatively after, the child node of weights minimum is deposited in the survival vector tabulation.This process is continued until that length is till the survival vector tabulation of M is filled.These nodes through ordering from small to large after, can rename as

So circulation just can obtain the individual required survival vector of M until the 1st layer:

For convenient, this M required survival vector is expressed as again:

With the survival set of vectors: $\hat{Z}=\left[\begin{array}{cccc}{\hat{z}}_1& {\hat{z}}_2& \&CenterDot;\&CenterDot;\&CenterDot;& {\hat{z}}_M\end{array}\right].$

Step 4, determine each survival vector respectively

Corresponding survival sends symbolic vector

And the transmission symbolic vector of surviving

Corresponding survival sends bit vectors (or claim the set of survival bit vector ), send again bit vectors according to surviving

With the survival vector

Corresponding weights E _mObtain the soft information that each sends bit.

In this step (4), l that finds the solution on the i root transmitting antenna sends bit b _{I, l}Soft information L (b _{I, l}) computing formula be: $L\left(b_{i,l}\right)=\frac{1}{{\mathrm{\&sigma;}}^2}({\min }_{b_{i,l}\&Element;{\hat{b}}_m{(i,l)}_0}{E}_m-{\min }_{b_{i,l}{\&Element;\hat{b}}_m{(i,l)}_1}{E}_m),$ In the formula, σ ²Be the power of white Gauss noise n,

To survive to send bit vectors

The bit value of capable, the l of i row, natural number i is transmitting antenna port sequence number, its maximum is emission fate sum N _t, natural number l is the row sequence number that sends bit, its maximum is order of modulation M _c,

Expression: satisfying

All survival send the corresponding weights E of bit vectors _mIn minimum value,

Expression: satisfying

All survival send the corresponding weights E of bit vectors _mIn minimum value.

The above only is preferred embodiment of the present invention, and is in order to limit the present invention, within the spirit and principles in the present invention not all, any modification of making, is equal to replacement, improvement etc., all should be included within the scope of protection of the invention.

Claims (4)

1. a soft output method that detects based on the MIMO of lattice reduction is characterized in that: at first obtain the sorted lists of transform-domain signals for different transformation matrixs, and it is stored for the detection query; When each MIMO that implements based on lattice reduction detects, first channel response matrix is used lattice reduction and obtain reduced basis and transformation matrix, obtain again the sorted lists of corresponding transform-domain signals by transformation matrix; Then utilize the sorted lists of reduced basis and corresponding transform-domain signals to carry out QRM (QR Decomposition and M algorithm) detection, obtain survival vector and weights corresponding to each survival vector; Utilize at last the weights of survival vector and each survival vector, determine the soft information of each bit; Described method comprises following operating procedure:

(1) obtains respectively the sorted lists of transform-domain signals for different transformation matrixs, and it is stored, with for the detection query; This step comprises following content of operation:

(11) constellation point that obtains respectively in the transform domain for different transformation matrix T is gathered;

(12) according to following be that the method for optiaml ciriterion is carried out list ordering based on central point to the distance of constellation point: i dimension transform domain constellation point is divided into 4M _iIndividual lattice, in the formula, M _iThe number natural number i that is i dimension constellation point is transmitting antenna port sequence number, and its maximum occurrences is emission fate sum N _tAccording to the grid position at central point place, after the distance of various constellations point is put by computer center respectively, again according to distance closely, far this is tieed up all constellation point and sorts, obtain sorted lists corresponding to this central point; Then, travel through the 4M of this central point _iIndividual diverse location just obtains the sorted lists of this dimensional signal;

(13) limited because of the value of transformation matrix T, story arranges first the memory space of sorted lists for different transformation matrixs, in order to directly select sorted lists according to the value of T when detecting;

(2) carry out when detecting based on lattice reduction LRA at every turn, channel response matrix H is used lattice reduction, obtain reduced basis H _RedWith transformation matrix T, and from the sorted lists of the corresponding transform-domain signals of different transformation matrix T of storage, obtain the corresponding sorted lists of this transformation matrix T according to transformation matrix T;

(3) according to reduced basis H _RedCarry out detection with corresponding sorted lists, to obtain M survival vector

And this M survival vector

Corresponding weights E _mIn the formula, natural number subscript m represent the to survive sequence number of vector, the maximum of m is M;

(4) determine respectively each survival vector

Corresponding survival sends symbolic vector

And the transmission symbolic vector of surviving

Corresponding survival sends bit vectors Send again bit vectors according to surviving

With the survival vector

Corresponding weights E _mObtain the soft information that each sends bit.

2. method according to claim 1, it is characterized in that: described step (2) further comprises following content of operation:

(21) channel response matrix H is used lattice reduction, obtain reduced basis H _RedWith transformation matrix T, and this step is not construed as limiting the specific algorithm of lattice reduction;

(22) from the sorted lists of the corresponding transform-domain signals of different transformation matrixs of storage, obtain sorted lists corresponding to this transformation matrix T according to transformation matrix T.

3. method according to claim 1, it is characterized in that: described step (3) further comprises following content of operation:

(31) vector y carries out translation to received signal:

Obtain equivalent received signals vector y ', to guarantee equivalent transmitted signal vector as non-negative plural number, translation numerical value wherein is ${\mathrm{\&beta;H}1}_{N_t\&times;1}=\sqrt{3/2\left((M_c-1)\right)}(\sqrt{M_c}-1+(\sqrt{M_c}-1)*j){H1}_{N_t\&times;1},$ In the formula, the translation coefficient value $\mathrm{\&beta;}=\sqrt{3/2\left((M_c-1)\right)}(\sqrt{M_c}-1+(\sqrt{M_c}-1)*j),$ H is channel response matrix, M _cBe order of modulation,

That dimension is N _t* 1 complete 1 vector, N _tBe emission fate sum;

(32) to equivalent received signal phasor y ' according to the following equation:

$y^{\&prime;}=\mathrm{\&alpha;H}(1/\mathrm{\&alpha;}(x+{\mathrm{\&beta;}1}_{N_t\&times;1}))+n=\mathrm{\&alpha;Hs}+n$ Carry out equivalence transformation, to guarantee that equivalent transmitted signal vector is as non-negative complex integers, in the above-mentioned formula, y is for receiving signal phasor, and H is channel response matrix, and x is the transmitted signal vector, s is that value is the equivalent transmitted signal vector of non-negative complex integers, and n is white Gauss noise, the equivalence transformation coefficient value

Translation numerical value is:

${\mathrm{\&beta;H}1}_{N_t\&times;1}=\sqrt{3/2\left((M_c-1)\right)}(\sqrt{M_c}-1+(\sqrt{M_c}-1)*j){H1}_{N_t\&times;1},$ H is channel response matrix, M _cBe order of modulation,

That dimension is N _t* 1 complete 1 vector;

(33) the reduced basis α H after the parity price conversion _RedCarry out ORTHOGONAL TRIANGULAR QR and decompose, obtain unitary matrice Q and upper triangular matrix R, and to the associate matrix Q of equivalent received signal phasor y ' premultiplication Q ^H, obtain detecting vector ρ;

(34) successively carry out the QRM detection to detecting vector ρ, obtain the survival vector

And this M survival vector

Corresponding weights E _mThis M survival vector

Corresponding weights E _mComputing formula be:

In the formula, ρ is for detecting vector, and R is upper triangular matrix,

The expression vector 2 norms square.

4. method according to claim 1 is characterized in that: in the described step (4), l the computing formula that sends the soft information of bit of finding the solution on the i root transmitting antenna is:

$L\left(b_{i,l}\right)=\frac{1}{{\mathrm{\&sigma;}}^2}({\min }_{b_{i,l}\&Element;{\hat{b}}_m{(i,l)}_0}{E}_m-{\min }_{b_{i,l}\&Element;{\hat{b}}_m{(i,l)}_1}{E}_m),$ Wherein, σ ²Be the power of white Gauss noise n,

To survive to send bit vectors

The bit value of capable, the l of i row, natural number i is transmitting antenna port sequence number, its maximum is emission fate sum N _t, natural number l is the row sequence number that sends bit, its maximum is order of modulation M _c,

Expression: satisfying All survival send the corresponding weights E of bit vectors _mIn minimum value,

Expression: satisfying All survival send the corresponding weights E of bit vectors _mIn minimum value.

Images