CN103986561A - Detecting algorithm based on planisphere reduction in high order modulation MIMO system - Google Patents

Abstract

The invention provides a detecting algorithm based on planisphere reduction in a high order modulation MIMO system. The problem that a detecting algorithm in the high order modulation MIMO system is high in complexity is solved. According to the detecting algorithm, a planisphere is reduced, tree expansion is performed according to the reduced planisphere, and the planisphere reduction scheme can be selected in a self-adaptive mode according to the state of a channel. The advantages of the detecting algorithm in an actual value model are fully used. The simulation results show that the self-adaptive detecting algorithm based on planisphere reduction is low in BER, greatly reduces the complexity of the detecting algorithm and has the advantages in the aspects of the BER and complexity. Compromise between the BER and the complexity of the algorithm can be achieved by adjusting parameters, and the detecting algorithm is suitable for the high order modulation MIMO system.

Description

Detection algorithm based on planisphere reduction in high order modulated MIMO system

Technical field

The invention belongs to wireless communication technology field, relate to the signal detection algorithm in mimo system, be specifically related to the detection algorithm based on planisphere reduction in a kind of high order modulated MIMO system.

Background technology

Next generation wireless network has proposed more and more higher requirement to the service speed of wireless transmission, and the availability of frequency spectrum that improves mimo system becomes a urgent and important job.In order to improve the availability of frequency spectrum, in mimo system, multiplexed data fluxion and the order of modulation of employing are increasing always.In wireless communication system, detection algorithm is the key that assurance system is brought into play its superiority accurately, therefore about the research of detection algorithm, has received the concern of Chinese scholars.

In various MIMO detection algorithms, ML detection algorithm can reach optimum detection performance, but complexity is higher, can be exponential increase along with the increase of antenna number and order of modulation.The MIMO detection algorithm of low complex degree comprises ZF detection algorithm and MMSE detection algorithm, but in changing various fading channel environment, the performance of these detection algorithms often can not meet the demand of practical application.Class maximum likelihood algorithm can be obtained the performance that approaches maximum likelihood algorithm with lower complexity, be a kind of very promising detection algorithm.Typical class maximum likelihood algorithm has the preferential stack algorithm of Sphere Decoding Algorithm, the measure value of depth-first and the K-best algorithm of breadth-first, these detection algorithms are all the detection algorithms based on tree search, by maximum likelihood algorithm being converted into tree search problem, can adopt various cut operator to reduce the complexity that MIMO detects.The advantage of Sphere Decoding Algorithm is that performance is good, and shortcoming is that complexity is unstable, can change along with the variation of signal to noise ratio.The advantage of stack algorithm is that complexity is lower, and shortcoming is that tree search procedure is complicated, need to frequently recall.The K-best detection algorithm of breadth-first can be obtained the good compromise between performance and complexity.But in the larger mimo system of order of modulation, as having supported 64QAM and 802.11ac, LTE-A supported 256QAM, the complexity of K-best detection algorithm is still higher, so the high problem of detection algorithm complexity in High Order Modulation System that solves becomes a problem demanding prompt solution.

Summary of the invention

The object of the present invention is to provide the detection algorithm based on planisphere reduction in a kind of high order modulated MIMO system, to solve the high problem of detection algorithm complexity in high level MIMO system.

For achieving the above object, the technical solution used in the present invention is:

Detection algorithm based on planisphere reduction in high order modulated MIMO system, comprises the following steps:

1) set up system model:

For one, there is N _tindividual transmitting antenna, N _rthe non-coding mimo system of individual reception antenna, the complex value model that MIMO detects is the corresponding real-valued model of this complex value model is y=Hs+n,

Wherein: n _rthe receiving symbol vector of * 1 dimension; n _r* N _tthe channel matrix of dimension; n _tthe transmission symbolic vector of * 1 dimension; that average is 0, variance is σ ²multiple Gaussian noise vector; Y, H, s, n are respectively real-valued form;

2) preliminary treatment:

Channel matrix H is carried out to QR decomposition, the diagonal element r to matrix R _iisort, select N _aCRindividual have a less r _iilayer, leave in S set et_k, the number of plies k at initialization search place, makes k=2N _t;

Wherein: Q is 2N _r* 2N _tthe orthogonal matrix of dimension; R is 2N _t* 2N _tthe upper triangular matrix of dimension, (i, j) individual element representation of R is r _ij, N _aCRfor the self adaptation constellation reduction number of plies of setting, 0<N _aCR<2N _t, k>0;

3) search:

The search parameter of m for setting;

Work as 2N _tduring-k<m, the front m layer of expanding for tree carries out full constellation point expansion, the child node number M that in now tree search, father node need to be expanded _requal the modulation system specified number or amount in mimo system;

Work as 2N _tduring-k>=m, first planisphere is reduced, reduce the child node number M that in tree search, father node need to be expanded _r, make the child node number M that in tree search, father node need to be expanded _rbe less than the modulation system specified number or amount in mimo system, and then set expansion according to the planisphere after reduction;

4) search of every end one deck, upgrades k=k-1, then repeats step 3), until k=1.

Described m=1 or 2.

By formula (8), calculate the part of Euclidean distance of tree search i layer,

$e\left(s^{\left(i\right)}\right)={|z_i-\underset{j=i+1}{\overset{{2N}_T}{\mathrm{\&Sigma;}}}{r}_{\mathrm{ij}}{s}_j-r_{\mathrm{ii}}{s}_i|}^2={\left|r_{\mathrm{ii}}\right|}^2{|c_i-s_i|}^2---\left(8\right)$

Wherein: e (s ⁽ⁱ⁾) be the part of Euclidean distance of tree search i layer, c _iit is the expansion center of father node; s _ithe point on planisphere, z _ii the element of z, s _jj the element of s, r _iiit is the diagonal entry of R.

By formula (7), calculate the accumulative total Euclidean distance of tree search i layer,

T(s ⁽ⁱ⁾)＝T(s ⁽ⁱ⁺¹⁾)+e(s ⁽ⁱ⁾) (7)

Wherein: T (s ⁽ⁱ⁾) for setting the accumulative total Euclidean distance of search i layer, T (s ⁽ⁱ⁺¹⁾) for setting the accumulative total Euclidean distance of search i+1 layer.

By formula (9), calculate the expansion center of father node,

$c_i=\frac{1}{r_{\mathrm{ii}}}(z_i-\underset{j=i+1}{\overset{{2N}_T}{\mathrm{\&Sigma;}}}{r}_{\mathrm{ij}}{s}_j)---\left(9\right).$

When planisphere is reduced, according to formula (10) reduction constellation collection, thereby reduce the child node number that in tree search, father node need to be expanded,

$\hat{l}=\arg \underset{l\&Element;\left\{\mathrm{1,2,3}\right\}}{\min }|q_l-c_i|---\left(10\right)$

Wherein: q _lthe reduction constellation expressing possibility is concentrated the heart, and l optionally reduces constellation to concentrate heart numbering, that the reduction constellation of selecting is concentrated heart numbering.

With respect to prior art, beneficial effect of the present invention is:

The present invention has considered the advantage of K-best detection algorithm in real-valued model, has proposed the detection algorithm based on planisphere reduction in a kind of high order modulated MIMO system, in order to solve the high problem of detection algorithm complexity in high level MIMO system.In High Order Modulation System, the advantage of K-best detection algorithm in real-valued model mainly comprises 2 points: one is to have lower beta pruning probability, and another is to make full use of the advantage that sequence QR decomposes.These two advantages play an important role on assurance algorithm performance.Based on this point, the present invention proposes the detection algorithm based on planisphere reduction in the high order modulated MIMO system in real-valued model.The method is first set up out system model, then carries out preliminary treatment, then searches for, and in search procedure, planisphere is reduced, and then according to the planisphere after reduction, sets expansion, and can be according to the scheme of the adaptively selected planisphere reduction of channel status.The method takes full advantage of the advantage of detection algorithm in real-valued model, simulation result shows, compare with existing detection algorithm, the detection algorithm that the present invention proposes has lower BER (error rate), greatly reduce again complexity, there is the advantage of BER and complexity two aspects, by regulating parameter simultaneously, can the compromise of implementation algorithm between BER and complexity, be a kind of detection algorithm that is suitable for high order modulated MIMO system.

The measure that the algorithm that this powder proposes reduces complexity is that planisphere is reduced, and reduces the son node number that father node need to be expanded when setting expansion, thereby reduces the complexity of algorithm.Yet, planisphere is reduced and will bring performance loss, so the present invention takes following measures to guarantee that algorithm has lower BER.

First, at the initial stage (being the data Layer that corresponding label is larger) of tree search, carry out full constellation collection expansion, reduce the error that initiation layer detects.

Then, when planisphere is reduced, taked the reduction method less to performance impact.Be specially when carrying out planisphere reduction, consider the part of Euclidean distance of tree search i layer, from formula (8), the constellation point nearer with expansion center has less part of Euclidean distance, is also that measure value increment is less.In child node expansion, should preferentially expand the child node that measure value is less, also preferentially expand the child node nearer with expansion centre distance, so planisphere reduction should retain the constellation point with expansion center with small distance.Thereby derive definite method of reduction constellation collection, use q _lthe reduction constellation expressing possibility is concentrated the heart, and the method for determining reduction constellation collection is to select the reduction constellation nearest with expansion center to concentrate the constellation collection of heart representative, namely according to formula (10), selects .

What is more important, the BER loss bringing in order to reduce planisphere reduction, has proposed self adaptation (ACR) detection algorithm based on planisphere reduction, and the different detection layers that realize at a secondary channel are selected planisphere reduction scheme adaptively.From formula (12), if k+1, k+2 ..., 2N _tthe symbol detection of layer is correct, so r _kkbeing one affects the key factor that k layer detects performance, owing to having less r _kkthe noise of layer larger on the impact of detection algorithm performance, so select N _aCRlayer has minimum r _kklayer carry out the planisphere reduction of less degree, to guarantee that BER is lower.

Accompanying drawing explanation

Fig. 1 is non-coded mimo systems block diagram;

Fig. 2 is the real-valued constellation collection reduction of 64QAM scheme;

Fig. 3 is the real-valued constellation collection reduction of 256QAM scheme;

Fig. 4 is the tree expansion schematic diagram after planisphere reduction;

Fig. 5 compares various constellations figure to reduce the BER of scheme in 4 * 4,64QAM system;

Fig. 6 compares the complexity that various constellations figure reduces scheme in 4 * 4,64QAM system;

Fig. 7 compares the algorithm of the present invention's proposition and the BER of existing algorithm in 4 * 4,64QAM system;

Fig. 8 compares the algorithm of the present invention's proposition and the BER of existing algorithm in 4 * 4,256QAM system;

Fig. 9 compares the algorithm of the present invention's proposition and the BER of existing algorithm in 8 * 8,64QAM system;

Figure 10 compares the algorithm of the present invention's proposition and the BER of existing algorithm in 8 * 8,256QAM system.

Embodiment

Below in conjunction with accompanying drawing, the present invention is described in further detail.

Thought of the present invention is, first utilizes expansion center to reduce planisphere, then according to the planisphere after reduction, set expansion, and the scheme that can reduce according to the adaptively selected planisphere of channel status.By planisphere is reduced, reduced the son node number that father node need to be expanded, thereby reduced the complexity of detection algorithm.And, proposition be the planisphere reduction method in real-valued model, take full advantage of the advantage of detection algorithm in real-valued model.Algorithm in the present invention has the advantage of BER and complexity two aspects, by regulating the parameter can the good compromise of implementation algorithm between BER and complexity, is a kind of detection algorithm of applicable high order modulated MIMO system.

Adaptive detection algorithm based on planisphere reduction in high order modulated MIMO system provided by the invention, specifically comprises the following steps:

1) foundation of system model.Consider one and there is N _tindividual transmitting antenna, N _rthe non-coding mimo system of individual reception antenna.The complex value model that MIMO detects can be expressed as

$\tilde{y}=\tilde{H}\tilde{s}+\tilde{n}$ Formula (1)

Wherein: n _rthe receiving symbol vector of * 1 dimension; n _r* N _tthe channel matrix of dimension; n _tthe transmission symbolic vector of * 1 dimension; that average is 0, variance is σ ²multiple Gaussian noise vector.Complex value model can be converted into real-valued model y=Hs+n, namely

$\left[\begin{array}{c}\mathrm{Re}\left\{\tilde{y}\right\}\\ \mathrm{Im}\left\{\tilde{y}\right\}\end{array}\right]=\left[\begin{array}{cc}\mathrm{Re}\left\{\tilde{H}\right\}& -\mathrm{Im}\left\{\tilde{H}\right\}\\ \mathrm{Im}\left\{\tilde{H}\right\}& \mathrm{Re}\left\{\tilde{H}\right\}\end{array}\right]\left[\begin{array}{c}\mathrm{Re}\left\{\tilde{s}\right\}\\ \mathrm{Im}\left\{\tilde{s}\right\}\end{array}\right]+\left[\begin{array}{c}\mathrm{Re}\left\{\tilde{n}\right\}\\ \mathrm{Im}\left\{\tilde{n}\right\}\end{array}\right]$ Formula (2)

Wherein: Re{} represents the real part of amount of orientation; Im{} represents the imaginary part of amount of orientation, and y, H, s, n are respectively real-valued form.

The target of MIMO detection algorithm is to find a vectorial s, makes itself and the vectorial y receiving have minimum Euclidean distance, namely

$\hat{s}=\arg \underset{s\&Element;{\mathrm{\&Omega;}}^{2N_T}}{\min }{\left|\right|y-\mathrm{Hs}\left|\right|}^2$ Formula (3)

Wherein: for the transmission symbolic vector of selecting, Ω is real-valued constellation collection, and size is M _r, M _rfor the child node number that in tree search, father node need to be expanded.Channel matrix is carried out to QR decomposition, H=QR, Q is 2N _r* 2N _tthe orthogonal matrix of dimension; R is 2N _t* 2N _tthe upper triangular matrix of dimension, (i, j) individual element of R can be expressed as r _ij.Test problems can be expressed as

$\hat{s}=\arg \underset{s\&Element;{\mathrm{\&Omega;}}^{2N_T}}{\min }{\left|\right|Q^{H}y-\mathrm{Rs}\left|\right|}^2$ Formula (4)

Make z=Q ^hy, above formula is expressed as

$\hat{s}=\arg \underset{s\&Element;{\mathrm{\&Omega;}}^{2N_T}}{\min }\underset{i=1}{\overset{{2N}_T}{\mathrm{\&Sigma;}}}{|z_i-\underset{j=i}{\overset{{2N}_T}{\mathrm{\&Sigma;}}}{r}_{\mathrm{ij}}{s}_j|}^2$ Formula (5)

Z wherein _ii the element of z, s _jj the element of s.

Definition e (s ⁽ⁱ⁾) be the part of Euclidean distance of tree search i layer,

${e\left(s^{\left(i\right)}\right)=|z_i-\underset{j=i}{\overset{{2N}_T}{\mathrm{\&Sigma;}}}{r}_{\mathrm{ij}}{s}_j|}^2$ Formula (6)

If the accumulative total Euclidean distance that searches i layer is expressed as to T (s ⁽ⁱ⁾), there is T (s ⁽ⁱ⁾)=T (s ⁽ⁱ⁺¹⁾)+e (s ⁽ⁱ⁾) formula (7), also, the accumulative total Euclidean distance of i layer be the accumulative total Euclidean distance of i+1 layer and the part of Euclidean distance of i layer and.

MIMO test problems can be understood as tree search problem, and the process of finding optimal solution is exactly from root node, to start to set the process of search.Different from exhaustive search, every layer of K-best detection algorithm only retains K best both candidate nodes, when carrying out lower one deck search, is each both candidate nodes expansion M _rindividual child node.Therefore every layer need to be calculated M _rthe part of Euclidean distance of K child node and accumulative total Euclidean distance.In High Order Modulation System, larger due to constellation point number, has brought serious computation burden to detection.

2) planisphere reduction method.The effect of planisphere reduction method is to lose under little prerequisite at BER, reduces the son node number that father node need to be expanded, and the planisphere that is also about to expansion reduces.How in the situation that guaranteeing that BER loss is less, planisphere to be reduced and to become a key issue.

The part of Euclidean distance of tree search i layer, can be rewritten into following this formula

$e\left(s^{\left(i\right)}\right)={|z_i-\underset{j=i+1}{\overset{{2N}_T}{\mathrm{\&Sigma;}}}{r}_{\mathrm{ij}}{s}_j-r_{\mathrm{ii}}{s}_i|}^2={\left|r_{\mathrm{ii}}\right|}^2{|c_i-s_i|}^2---\left(8\right)$ Formula (8)

Wherein: c _iit is the expansion center of father node; s _iit is the point on planisphere.The expansion center of father node can be expressed as

$c_i=\frac{1}{r_{\mathrm{ii}}}(z_i-\underset{j=i+1}{\overset{{2N}_T}{\mathrm{\&Sigma;}}}{r}_{\mathrm{ij}}{s}_j)---\left(9\right).$ Formula (9)

Known, the constellation point nearer with expansion center has less part of Euclidean distance, is also that measure value increment is less.In child node expansion, should preferentially expand the child node that measure value is less, also preferentially expand the child node nearer with expansion centre distance, so planisphere reduction should retain the constellation point with expansion center with small distance.

Introduce definite method of reduction constellation collection below, use q _lthe reduction constellation expressing possibility is concentrated the heart, and the method for determining reduction constellation collection is to select the reduction constellation nearest with expansion center to concentrate the constellation collection of heart representative, namely selects $\hat{l}=\arg \underset{l\&Element;\left\{\mathrm{1,2,3}\right\}}{\min }|q_l-c_i|$ Formula (10),

Wherein: l optionally reduces constellation to concentrate heart numbering, that the reduction constellation of selecting is concentrated heart numbering.By constellation collection is reduced, reduced the child node number that father node need to be expanded, the complexity of K-best detection algorithm has also just reduced.

3) adaptive detection algorithm based on planisphere reduction.The planisphere reduction method proposing above can reduce the complexity of K-best detection algorithm, but can bring BER loss, and the BER loss bringing in order to reduce planisphere reduction, has proposed self adaptation (ACR) detection algorithm based on planisphere reduction.

At the both sides of real-valued detection model y=Hs+n premultiplication Q ^h, there is z=Rs+ η, wherein, z=Q ^hy, η=Q ^hn.K the element of z can be expressed as

$z_k=\underset{j=k}{\overset{{2N}_T}{\mathrm{\&Sigma;}}}{r}_{\mathrm{kj}}{s}_j+{\mathrm{\&eta;}}_k$ Formula (11)

K sends symbol s so _kcan be expressed as

$s_k=Q\left(\frac{z_k-{\mathrm{\&Sigma;}}_{j=k+1}^{2N_T}{r}_{\mathrm{kj}}{s}_j-{\mathrm{\&eta;}}_k}{r_{\mathrm{kk}}}\right)$ Formula (12)

In formula: η _ibe k the element of η, Q (.) realizes parameter quantification to constellation symbol.If k+1, k+2 ..., 2N _tthe symbol detection of layer is correct, so r _kkbeing one affects the key factor that k layer detects performance, because a larger r _kkcan the impact of attenuating noise on symbol detection, detect better performances.In fact this is also the principle that sequence QR decomposes, but the low complex degree often using sequence QR decomposes and can not guarantee r _kkstrict order.R _kkit is the diagonal element of R.

The BER loss bringing in order to reduce planisphere reduction, the different detection layers that propose to realize at a secondary channel are selected planisphere reduction scheme adaptively.Owing to thering is less r _kkthe noise of layer larger on the impact of detection algorithm performance, so this layer is carried out to constellation collection when reduction, select the reduction scheme of the planisphere of less degree, less to guarantee that BER loses.Select N _aCRlayer has minimum r _kklayer carry out the planisphere reduction of less degree.Visible N _aCRan adjustable parameter, parameter N _aCRlarger, BER improves more obvious, and complexity also can correspondingly increase, and is also that ACR detection algorithm can be realized the compromise between the complexity of BER.

Adaptive detection algorithm step based on planisphere reduction is as follows:

1. preliminary treatment:

Channel matrix is carried out to QR decomposition, the diagonal element of matrix R is sorted, select N _aCRindividual have a less r _kklayer, leave in S set et_k the number of plies k=2N at initialization search place in _t; N wherein _aCRfor the self adaptation constellation reduction number of plies of setting, 0<N _aCR<2N _t, k>0;

2. if k>0

If a) 2N _t-k<m

Front m layer for tree expansion carries out full constellation point expansion, and this is because the error of first detection layers can cause on later detection serious impact, so will guarantee the correctness of first detection layers.

B) if 2N _t-k>=m

First planisphere is reduced, then according to the planisphere after reduction, set expansion;

If planisphere is carried out to planisphere reduction largely;

If k ∈ is Set_k, planisphere is carried out to the planisphere reduction of less degree;

C) search of every end one deck, upgrades k=k-1.

About more greatly, the definition of the planisphere of less degree reduction:

For the real-valued model of QAM modulation, seat collection on duty for the week can be expressed as in fact $\mathrm{\&Omega;}=\{-(\sqrt{Q}-1),-(\sqrt{Q}-3),...,-3,-\mathrm{1,1,3},...,(\sqrt{Q}-3),(\sqrt{Q}-1)\},$ Wherein Q is the element number comprising in QAM complex value model constellation collection.Planisphere reduction is that the concentrated constellation point of constellation is divided into a plurality of subsets, and wherein the number of subset represents with variable Z.Can mutual exclusion between each subset, also can comprise identical element.In addition, comprise the constellation point that number is equal in each subset, and represent the constellation point number in each subset with variable W, the value of W should be less than the number of the constellation point in real-valued model constellation point in each subset is counted as an integral body, and uses the barycenter (i.e. the average of each element in this subset) of this subset to concentrate the heart as the reduction constellation that represents this subset.Because the value of W should be less than the number of the constellation point in real-valued model so be called the reduction of planisphere; Accordingly, in the present invention, use CR (W, Z) to represent the reduction feature of planisphere.The constellation reduction of the larger less degree of W representative, and less W represents constellation reduction largely.The feasible reduction example of 64QAM and 256QAM is distinguished as shown in Figures 2 and 3 with corresponding parameter.

Fig. 1 has provided non-coding mimo system block diagram, and input bit transmits by spatial multiplexing mode.

Fig. 2 has provided the real-valued constellation collection of 64QAM reduction scheme, the real-valued constellation collection of known 64QAM can be expressed as Ω=7 ,-5 ,-3 ,-1,1,3,5,7}.According to the expansion center of father node, 8 constellation collection can be reduced to 4 constellation collection or 2 constellation collection, constellation collection reduction scheme can have following four kinds, and (a) CR (4,2) scheme represents that constellation collection can be reduced to 24 possible constellation point set; (b) CR (4,3) scheme represents that constellation collection can be reduced to 34 possible constellation point set; (c) CR (2,4) scheme represents that constellation collection can be reduced to 42 possible constellation point set; (d) CR (2,7) scheme represents that constellation collection can be reduced to 72 possible constellation point set.In CR (2,4) scheme, each father node is only expanded 2 child nodes, so BER loss can be greater than CR (4, the 2) scheme of 4 child nodes of expansion.In four kinds of reduction schemes, (b), (d) scheme has been considered overlapping constellation collection division methods, overlapping constellation collection can be with higher probability packet containing actual transmission symbol, so BER is lower, but determine that the complexity of reduction constellation collection also can increase along with the increase of possibility constellation collection number.

Definite method with CR (4,3) scheme as exemplary introduction reduction constellation collection, uses q _lthe reduction constellation expressing possibility is concentrated the heart, q ₁=-4, q ₂=0, q ₃=4.The method of determining reduction constellation collection is to select the reduction constellation nearest with expansion center to concentrate the constellation collection of heart representative, namely selects in: l is that reduction constellation is concentrated heart numbering.By constellation collection is reduced, reduced the child node number that father node need to be expanded, the complexity of K-best detection algorithm has also just reduced.

Fig. 3 has provided the real-valued constellation collection reduction of 256QAM scheme, according to the expansion center of father node, the real-valued constellation collection of 16 can be reduced to 8 points, or 2 constellation collection at 4, planisphere reduction scheme is, (a) CR (8,2) scheme represents that constellation collection can be reduced to 28 possible constellation point set; (b) CR (8,3) scheme represents that constellation collection can be reduced to 38 possible constellation point set; (c) CR (4,4) scheme represents that constellation collection can be reduced to 44 possible constellation point set; (d) CR (4,7) scheme represents that constellation collection can be reduced to 74 possible constellation point set; (e) CR (2,8) scheme represents that constellation collection can be reduced to 82 possible constellation point set; (f) CR (2,15) scheme represents that constellation collection can be reduced to 15 2 possible constellation point set.In six kinds of reduction schemes, (b), (d) and (f) scheme has been considered overlapping constellation collection division methods, overlapping constellation collection can be with higher probability packet containing actual transmission symbol, so BER is lower, owing to determining that reduction constellation collection need to judge the distance at the concentrated heart of reduction constellation and expansion center, so overlapping constellation collection splitting scheme needs more complexity, and determine that the complexity while reducing constellation collection also can increase along with the increase of possibility constellation collection number.

The BER loss bringing in order to reduce planisphere reduction, has proposed the adaptive detection algorithm based on planisphere reduction.Under 64QAM modulation system, algorithm characteristics is ACR (2,4, N _aCR), m=1 or 2, represents front m layer to carry out full constellation collection expansion, then selects N _aCRlayer carries out the reduction of CR (4,3) scheme, and other layer carries out the reduction of CR (2,4) scheme.The 64QAM modulation system of take is introduced the step of algorithm as example:

1. preliminary treatment:

Channel matrix is carried out to QR decomposition, the diagonal element of matrix R is sorted, select N _aCRindividual have a less r _kklayer, leave in S set et_k the number of plies k=2N at initialization search place in _t;

2. if k>0

If a) 2N _t-k<m

Front m layer for tree expansion carries out full constellation point expansion, and this is because the error of first detection layers can cause on later detection serious impact, so will guarantee the correctness of first detection layers.

B) if 2N _t-k>=m

First planisphere is reduced, then according to the planisphere after reduction, set expansion;

If planisphere is carried out to the reduction of CR (2,4) scheme;

If k ∈ is Set_k, planisphere is carried out to the reduction of CR (4,3) scheme;

C) search of every end one deck, upgrades k=k-1.

In 256QAM modulating system, the feature of the adaptive detection algorithm based on planisphere reduction of proposition can be expressed as ACR (2,8, N _aCR), m=1 or 2, represents front m layer to carry out full constellation collection expansion, then selects N _aCRlayer carries out the reduction of CR (4,7) scheme, and other layer carries out the reduction of CR (2,8) scheme.

Fig. 4 has provided the tree extended mode after planisphere reduction, also only has near the child node in expansion center to be just expanded, thereby has reduced the complexity of algorithm.In figure, K is the search width of the algorithm that proposes, also the i.e. nodes of every layer of reservation in tree search.C _kit is the expansion center of K father node obtaining of formula (9).

In the emulation of this patent, from the performance of BER and complexity two aspect measure algorithms, flops for complexity (real addition number of times and real multiplications number of times sum) is weighed, and simulated channel is set to flat Rayleigh channel.Fig. 5 has provided the BER emulation that planisphere reduction scheme is carried out, abscissa SNR (Signal Noise Ratio) wherein, represent signal to noise ratio, communication system generally uses errored bit (BER) and the relation of signal to noise ratio to weigh the performance of detection algorithm.In figure, algorithm CR and ACR are the instantiations of algorithm of the present invention.Fig. 6 has provided the complexity statistics that planisphere reduction scheme is carried out, emulation setting: 4 * 4MIMO, 64QAM modulation system, K=20, m=1.As can be seen from Figure 5, when compared algorithm is all got identical width K, the K-best detection algorithm in real-valued model has the BER that is better than K-best algorithm in complex value model.Because constellation collection is overlapping, the BER of CR (2,7) scheme is lower than the BER of CR (2,4) scheme, and the BER of CR (4,3) scheme is lower than the BER of CR (4,2) scheme.As can be seen from Figure 6, in real-valued model, K-best algorithm has lower complexity, only need 25% of K-best algorithm complex in complex value model, adopt the K-best algorithm of planisphere reduction scheme only to need 3%-6% of K-best algorithm complex in complex value model.Visible, for K-best detection algorithm, real-valued model has the advantage of BER and complexity two aspects; Planisphere reduction method in the real-valued model proposing can significantly reduce the complexity of K-best algorithm, is a kind of detection algorithm with lower complexity.

Fig. 7～10 are four kinds of simulation configurations, and the adaptive detection algorithm based on planisphere reduction and existing HL-M algorithm and the Mlbm-M algorithm that propose are carried out to BER comparison, and the search width that has HL-M algorithm and Mlbm-M algorithm represents with parameter M.Fig. 7 has compared while adopting 64QAM modulation system in 4 * 4MIMO system, proposes the BER of ACR algorithm and existing HL-M algorithm and Mlbm-M algorithm, and parameter arranges K=M=20, m=1.Fig. 8 has compared while adopting 256QAM modulation system in 4 * 4MIMO system, proposes the BER of algorithm and existing algorithm, and parameter arranges K=M=25, m=1.As can be seen from the figure,, owing to having adopted adaptive constellation reduction scheme, the BER of ACR algorithm is lower than the BER of CR algorithm; Parameter N _aCRlarger, the BER of ACR algorithm is lower; The BER of the ACR algorithm proposing will be lower than the BER of existing HL-M and Mlbm-M algorithm.

Table 1 complexity of each algorithm in 4 * 4MIMO system, and calculate and propose algorithm with respect to the complexity range of decrease of existing algorithm.As can be seen from the table, while comparing with K-best algorithm in real-valued model, the complexity range of decrease of detection algorithm is proposed 40%-70%.When comparing with Mlbm-M algorithm, the complexity range of decrease of the detection algorithm of proposition surpasses 70%.When comparing with HL-M algorithm, the complexity range of decrease of the detection algorithm of proposition surpasses 80%.Wherein ACR and CR scheme are all instantiations of the present invention, and ACR proposes on the basis of CR.

The complexity comparison sheet of each algorithm in table 14 * 4MIMO system

Fig. 9 has compared while adopting 64QAM modulation system in 8 * 8MIMO system, proposes the BER of ACR algorithm and existing HL-M algorithm and Mlbm-M algorithm, and parameter arranges K=M=30.Figure 10 has compared while adopting 256QAM modulation system in 8 * 8MIMO system, proposes the BER of algorithm and existing algorithm, and parameter arranges K=M=30.As can be seen from the figure, in 8 * 8MIMO system, obtained and conclusion similar in 4 * 4MIMO system, different is than 4 * 4 systems, and increasing of antenna number can increase the BER loss that planisphere reduction brings, in order to reduce BER loss, can increase parameter m.

Table 2 complexity of each algorithm in 8 * 8MIMO system, and calculate and propose algorithm with respect to the complexity range of decrease of existing algorithm.As can be seen from the table, while comparing with K-best algorithm in real-valued model, the complexity range of decrease of detection algorithm is proposed 30%-60%.When comparing with Mlbm-M algorithm, the complexity range of decrease of the detection algorithm of proposition surpasses 70%.When comparing with HL-M algorithm, the complexity range of decrease of the detection algorithm of proposition surpasses 80%.

The complexity comparison sheet of each algorithm in table 28 * 8MIMO system

In a word, with existing HL-M algorithm, Mlbm-M algorithm is compared, the adaptive detection algorithm based on planisphere reduction proposing has the advantage of BER and complexity two aspects, it is a kind of MIMO detection algorithm of applicable High Order Modulation System, by regulating parameter, the algorithm of proposition can compromise to meet different actual demands between BER and complexity.

Claims (6)

1. the detection algorithm based on planisphere reduction in high order modulated MIMO system, is characterized in that, comprises the following steps:

1) set up system model:

For one, there is N _tindividual transmitting antenna, N _rthe non-coding mimo system of individual reception antenna, the complex value model that MIMO detects is the corresponding real-valued model of this complex value model is y=Hs+n,

Wherein: n _rthe receiving symbol vector of * 1 dimension; n _r* N _tthe channel matrix of dimension; n _tthe transmission symbolic vector of * 1 dimension; that average is 0, variance is σ ²multiple Gaussian noise vector; Y, H, s, n are respectively real-valued form;

2) preliminary treatment:

Channel matrix H is carried out to QR decomposition, the diagonal element r to matrix R _iisort, select N _aCRindividual have a less r _iilayer, leave in S set et_k, the number of plies k at initialization search place, makes k=2N _t;

Wherein: Q is 2N _r* 2N _tthe orthogonal matrix of dimension; R is 2N _t* 2N _tthe upper triangular matrix of dimension, (i, j) individual element representation of R is ri _j, N _aCRfor the self adaptation constellation reduction number of plies of setting, 0<N _aCR<2N _t, k>0;

3) search:

The search parameter of m for setting;

Work as 2N _tduring-k<m, the front m layer of expanding for tree carries out full constellation point expansion, the child node number M that in now tree search, father node need to be expanded _requal the modulation system specified number or amount in mimo system;

Work as 2N _tduring-k>=m, first planisphere is reduced, reduce the child node number M that in tree search, father node need to be expanded _r, make the child node number M that in tree search, father node need to be expanded _rbe less than the modulation system specified number or amount in mimo system, and then set expansion according to the planisphere after reduction;

4) search of every end one deck, upgrades k=k-1, then repeats step 3), until k=1.

2. the detection algorithm based on planisphere reduction in high order modulated MIMO system according to claim 1, is characterized in that: described m=1 or 2.

3. the detection algorithm reducing based on planisphere in high order modulated MIMO system according to claim 1 and 2, is characterized in that: by formula (8), calculate the part of Euclidean distance of tree search i layer,

$e\left(s^{\left(i\right)}\right)={|z_i-\underset{j=i+1}{\overset{{2N}_T}{\mathrm{\&Sigma;}}}{r}_{\mathrm{ij}}{s}_j-r_{\mathrm{ii}}{s}_i|}^2={\left|r_{\mathrm{ii}}\right|}^2{|c_i-s_i|}^2---\left(8\right)$

Wherein: e (s ⁽ⁱ⁾) be the part of Euclidean distance of tree search i layer, c _iit is the expansion center of father node; s _ithe point on planisphere, z _ii the element of z, s _jj the element of s, r _iiit is the diagonal entry of R.

4. the detection algorithm reducing based on planisphere in high order modulated MIMO system according to claim 3, is characterized in that: by formula (7), calculate the accumulative total Euclidean distance of tree search i layer,

T(s ⁽ⁱ⁾)＝T(s ⁽ⁱ⁺¹⁾)+e(s ⁽ⁱ⁾) (7)

Wherein: T (s ⁽ⁱ⁾) for setting the accumulative total Euclidean distance of search i layer, T (s ⁽ⁱ⁺¹⁾) for setting the accumulative total Euclidean distance of search i+1 layer.

5. the detection algorithm reducing based on planisphere in high order modulated MIMO system according to claim 3, is characterized in that: by formula (9), calculate the expansion center of father node,

$c_i=\frac{1}{r_{\mathrm{ii}}}(z_i-\underset{j=i+1}{\overset{{2N}_T}{\mathrm{\&Sigma;}}}{r}_{\mathrm{ij}}{s}_j)---\left(9\right).$

6. the detection algorithm reducing based on planisphere in high order modulated MIMO system according to claim 3, it is characterized in that: when planisphere is reduced, according to formula (10) reduction constellation collection, thereby reduce the child node number that in tree search, father node need to be expanded

$\hat{l}=\arg \underset{l\&Element;\left\{\mathrm{1,2,3}\right\}}{\min }|q_l-c_i|---\left(10\right)$

Wherein: q _lthe reduction constellation expressing possibility is concentrated the heart, and l optionally reduces constellation to concentrate heart numbering, that the reduction constellation of selecting is concentrated heart numbering.