CN101373975A - Spherical decoding method - Google Patents
Spherical decoding method Download PDFInfo
- Publication number
- CN101373975A CN101373975A CNA2007101430736A CN200710143073A CN101373975A CN 101373975 A CN101373975 A CN 101373975A CN A2007101430736 A CNA2007101430736 A CN A2007101430736A CN 200710143073 A CN200710143073 A CN 200710143073A CN 101373975 A CN101373975 A CN 101373975A
- Authority
- CN
- China
- Prior art keywords
- layer
- search
- value
- metric
- branch
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Error Detection And Correction (AREA)
- Radio Transmission System (AREA)
Abstract
The invention discloses a sphere decoding method, which comprises the steps of performing sequential decomposition of a channel matrix of a signal to be decoded, determining level number of a search space based on the row number of a matrix R, and determining Fano profile of each level; determining constellation point mapping value and metric value of a received signal on each level by using the elements in the matrix R, searching the search space from the last level of the search space with consideration of the influence of the Fano profile so as to effectively reduce number of searching branches, using the branch with minimum final metric value as a searching result, and using the constellation point mapping value corresponding to the branch to be the sphere decoding result. The inventive technical proposal can effectively reduce computational complexity of sphere decoding, and can balance performance and complexity by adjusting the size of the Fano profile.
Description
Technical field
The present invention relates to the signal processing technology of wireless communication field, relate in particular to a kind of globular decoding method.
Background technology
Progressively ripe along with wireless communication technology, it is long that mobile subscriber's quantity is being the index multiplication, and mobile service expanded to multimedia service from traditional speech business, like this frequency spectrum resource growing tension that just seems.At this situation, can realize that multiple-input and multiple-output (MIMO) system of high efficient coding, modulation and signal processing arises at the historic moment.In mimo system, in order to improve the reliability of information source information transmission, at transmitting terminal, signal to be transmitted is at first through providing the chnnel coding of error correcting capability, carry out space-time/space-frequency/space-time frequency coding again, then send simultaneously or according to the regular hour order by several transmitting antennas.For receiving terminal, come from the signal of transmitting terminal simultaneously or according to regular hour order reception by multi-amplitude receiver antenna, and carry out space-time/space-frequency/empty time-frequency decoding and channel decoding successively, thereby with the raw information of decode results as signal to be sent.
The channel decoding that receiving terminal carried out comes down to detection to received signal, promptly detects Optimal Signals from received signal, as decode results.ZF mode that at present can be by linear ZF mode, linear minimum mean-squared error mode, Interference Cancellation, least mean-square error mode, maximum-likelihood decoding mode and the globular decoding mode of Interference Cancellation wait and finish input.Practice shows, the maximum-likelihood decoding mode has preferable performance, but exponent function relation between the number of its computation complexity and order of modulation and antenna, for example: adopt 16QAM (16 rank quadrature amplitude modulation) when 5 width of cloth antennas are modulated, had to mode, computation complexity is 16
5The globular decoding mode has the performance identical with the maximum-likelihood decoding mode, but is cubic relationship between common its computation complexity and number of antennas.Therefore, the globular decoding mode is one of signal detecting mode the most commonly used at present.
The basic thought of globular decoding is that the search volume is expressed as the grid that can reflect channel response and/or coded system, is that hypersphere is set up at the center with the received signal, this hypersphere with interior grid in the best estimate of search to transmitting.With the S-E globular decoding is example, at first carrying out QR by corresponding to received signal channel matrix decomposes, according to the diagonal entry of channel matrix and matrix R received signal is mapped at least on one deck grid, wherein the number of plies of grid is identical with the line number of upper triangular matrix R; When searching for, at first calculate the metric of each layer in this search according to the element among the matrix R, the lattice point place branch of selectance value minimum begins to layer search second from the bottom from last one deck then, obtain the metric on all directions, continue the branch that search has the minimum degree value.The branch of selecting to have the minimum degree value at layer third from the bottom continues to the search of fourth from the last layer, until searching ground floor.This moment this branch the minimum degree value as the current radius of a ball, and preserve the estimated value of each layer lattice point correspondence, seek lattice point in other branch of the second layer, if do not find then not proceed to the 3rd layer of search, by that analogy less than the current radius of a ball.If find the branch of tolerance, then continue the branch that search n-1 layer has the minimum degree value, until ground floor less than current radius at the n layer.If less than current radius value, then upgrade current radius value, and preserve the value of each layer lattice point correspondence at the ground floor metric, search again; If find to have the branch of minimum metric greater than current radius value at the m layer, then proceed to the branch of (m+1) layer search metric less than current radius, at last when can not find than current separate metric littler separate the time, export current separating, this separates corresponding maximum likelihood and separates, i.e. best estimate to transmitting.
Fig. 1 is the schematic diagram of S-E globular decoding grid.Referring to Fig. 1, in shown globular decoding grid, include two-grid altogether, wherein comprise at least two branches in the second layer, each branch all includes 1 grid in ground floor.The number of these branches is determined by employed modulation system, for example under the 16QAM mode, is co-existed in 4 branches in the second layer.Suppose that search radius is infinite (∝), when searching for, the branch of metric minimum begins search from the second layer, promptly 1.2 place branches are searched for to ground floor from the second layer, and the estimated value of second layer branch correspondence is-1, and the estimated value of ground floor minimum degree value correspondence is 3, metric is 1.7, then preserve and separate (1,3), upgrading radius value is 1.7, to the second layer search lattice point littler than this tolerance, three lattice point tolerance are arranged less than 1.7, the lattice point that search tolerance earlier is minimum, this lattice point respective value is 1, tolerance is 1.3, downward search ground floor, finding tolerance is 1.5 lattice point, its respective value is 1, because 1.5＜1.7, upgrading the radius of a ball is 1.5, preserves to separate to be (1,1).To second layer search metric less than 1.5 lattice point, branch's 1 metric is 1.49, less than 1.5, searches for to ground floor, finding metric is 2.1 lattice point, since 2.1〉1.5, abandon this and separate, proceed to second layer search, there is not to find the little lattice point of ratio 1.5 of not search, finish search, the output maximum likelihood is separated (1,1).
From foregoing description as seen, the diagonal entry of matrix R is one of key factor that makes up each layer, and these diagonal entries are big more, shows that the signal to noise ratio of received signal of this element correspondence is big more.Because the diagonal entry of conventional QR split-matrix R is unordered, be the lower right corner to each element in the upper left corner not descending, will make that like this search first is not the branch from the signal to noise ratio maximum, might find in search procedure that so the probability that preceding Search Results several times is defined as correctly separating is lower.This means that the probability that direction of search error diffusion occurs is higher, promptly adjust the increase that the direction of search can cause computation complexity repeatedly.In addition, in the practical application, the number of plies of search volume is the twice of reception antenna number, and like this, computation complexity is owing to the increase of antenna increases substantially, thereby brings bigger burden for the receiving terminal of carrying out globular decoding.
Summary of the invention
In view of this, the invention provides a kind of globular decoding method, can reduce the computation complexity in the decode procedure.
In globular decoding method of the present invention, may further comprise the steps:
Treat the QR that the channel matrix of decoded signal sorts and decompose, determine the number of plies of search volume according to the line number of matrix R, and determine the Fano base of each layer;
Utilize the element among the matrix R to determine constellation point mapping value and the metric of received signal on each layer, begin the search volume is searched for from last one deck of search volume, and when searching the i layer, with the current search radius be updated to described current search radius and (i+1) layer to the end one deck Fano base and difference, i is the positive integer more than or equal to 1;
The branch of final metric value minimum as Search Results, and is consisted of the globular decoding result with the constellation point mapping value of this branch's correspondence.
Preferably, the described QR that channel matrix is sorted is decomposed into:
Conventional QR decomposed squared absolute value and minimum column permutation decompose for carry out QR when the prostatitis in the Q matrix, among the matrix R lower right corner diagonal entry greater than the probability of upper left corner diagonal entry greater than 0.5.
Preferably, described last one deck is the 2M layer, and wherein M is the number of transmitting antenna, and the described search volume is searched for is:
B1. search for successively since the 2M layer, search ground floor and obtain final metric value, with the current search radius be updated to this metric and the 2nd layer to 2M layer Fano base and difference, current the separating that preservation is made up of the constellation point mapping value, and make that current layer number i=2, i are the positive integer between 1 to 2M;
B2. carry out search according to the current search radius, search for the i layer, judge that whether current searched branched measurement value is less than the current search radius, if, i is updated to (i-1), if the i after upgrading equals 1, execution in step B3, if the i after upgrading, then returns the operation of carrying out the i of search described in this step layer greater than 1; Otherwise, i is updated to (i+1), as current searched branch, if i=2M, then execution in step B4 carries out the operation of searching for the i layer in this step otherwise return with other branch except that current searched branch in this layer;
B3. when searching ground floor, if the final metric value of current searched branch is less than the current search radius, with the current search radius be updated to the final metric value of current searched branch and the 2nd layer to 2M layer Fano base and difference, i is updated to (i+1) and preserves current separating, execution in step B2, otherwise, i is updated to (i+1) and execution in step B2;
The current branch that separates corresponding branch as the final metric value minimum that B4. will preserve.
Preferably, further comprise in the described search: when searching the i layer, with the current search radius be updated to search radius and predetermined (i+1) layer to 2M layer Fano base and difference.
Preferably, described each layer Fano base that pre-determines is:
The Fano base of i layer is defined as equaling
R wherein
I, jBe i the diagonal entry of described matrix R,
Be noise power, δ adjusts parameter for the performance of coming out by simulated measurement, and i is the positive integer in 1 to the 2M interval;
Perhaps, the Fano base with each layer is defined as fixed value;
Perhaps, the Fano base with the i layer is defined as and noise power
Linear or non-linear relation.
Preferably, set in advance the metric thresholding of each layer correspondence in the search first, step B1 is described since the search of 2M layer, obtains final metric value and is:
B11. with the i value for equaling 2M;
B12. the i layer is searched for as working as anterior layer, calculate when the received signal estimated value of anterior layer and the constellation point mapping value of this received signal, calculate the metric of this layer according to determined mapping value, judge that whether the metric of working as anterior layer is smaller or equal to predetermined metric thresholding when anterior layer, if, execution in step B13 then; Otherwise, execution in step B14;
B13. i is updated to (i-1), and judges and work as whether anterior layer is the 1st layer,, finish search first if the metric that then will work as anterior layer is defined as final metric value; Otherwise, return execution in step B12;
B14. when anterior layer is not the 2M layer, i is updated to (i+1), and returns execution in step B12.
Preferably, further comprise among the described step B14:
When anterior layer is the 2M layer, strengthens the metric thresholding of each layer, and return execution in step B12.
Preferably, the QR that the described channel matrix for the treatment of decoded signal sorts further comprises: utilize noise parameter that channel matrix is carried out preliminary treatment before decomposing.
Preferably, describedly utilize noise parameter that channel matrix is carried out preliminary treatment to be: channel matrix transformation is treated to
Wherein
Be channel matrix, σ
nBe noise parameter,
M is a number of transmit antennas, and SNR is a signal to noise ratio, and I is a unit matrix.
Use the present invention, can reduce the computation complexity of globular decoding effectively.
Particularly, at first, what carry out among the present invention is the QR decomposition of ordering, and in the matrix R that obtains, the diagonal entry in its lower right corner is higher greater than the probability of the diagonal entry in the upper left corner.And these diagonal entries are relevant with the signal to noise ratio of corresponding received signal, be that diagonal entry is big more, signal to noise ratio is also big more, and the signal to noise ratio of the received signal of matrix R lower right corner diagonal entry correspondence is higher greater than the probability of the signal to noise ratio of the received signal of lower left corner diagonal entry correspondence among the present invention so.Since in the globular decoding first the estimated value of search be that lower right corner element by matrix R calculates and gets, begin to carry out that to search for correct probability first also higher like this from the received signal of signal to noise ratio maximum.Therefore, the present invention can reduce the occurrence probability of direction of search error diffusion greatly, reduces the number of times of adjusting the direction of search, thereby can reduce the computation complexity of globular decoding effectively.And the present invention carries out after obtaining initial solution again in the search procedure, and to 2M layer direction search the time, every process one deck all bounces back search radius.Like this, the convergence rate of search radius is faster among the present invention, can some invalid branch of faster eliminating, find optimal solution, thereby computation complexity reduces further, and this point is particularly outstanding under the more situation of antenna.
Also have, the present invention can also be provided with metric thresholding, the metric thresholding difference of each layer to each layer when searching for first.In this case, when searching for first, at first since the grid of 2M layer moderate value minimum, when searching (2M-1) layer, have only the metric thresholding of the metric of this layer, just continue search less than correspondence, by that analogy, proceed to till the 1st layer up to search first.Like this,, can find wrong path in advance and jump to other layer search, can further improve the probability of seeking optimal solution, thereby reduce computation complexity effectively by the setting of metric thresholding.
Description of drawings
To make clearer above-mentioned and other feature and advantage of the present invention of those of ordinary skill in the art by describe exemplary embodiment of the present invention in detail with reference to accompanying drawing below, in the accompanying drawing:
Fig. 1 is the schematic diagram of S-E globular decoding grid;
Fig. 2 is the exemplary process diagram of globular decoding method among the present invention;
Fig. 3 is the exemplary process diagram of globular decoding method in the embodiment of the invention 1;
Fig. 4 is the exemplary process diagram of globular decoding method in the embodiment of the invention 2;
Fig. 5 is the schematic diagram of another kind of S-E globular decoding grid;
Fig. 6 is the flow chart of searching for first under the consideration metric thresholding situation in the embodiment of the invention 2.
Embodiment
For making purpose of the present invention, technical scheme clearer, below with reference to the accompanying drawing embodiment that develops simultaneously, the present invention is described in further detail.
The present invention is a kind of globular decoding method, its basic thought is: in the globular decoding process, the QR that channel matrix is carried out ordering decomposes, make that the lower right corner element on the diagonal is higher greater than the probability of upper left corner element among the matrix R, for example greater than 0.5, and when search, consider the Fano base of each layer.
Fig. 2 shows the exemplary process diagram of globular decoding among the present invention.Referring to Fig. 2, this method comprises:
In step 201, the QR that channel matrix is sorted decomposes, and determines the number of plies of search volume according to the line number of matrix R;
In step 202, utilize the element among the matrix R to determine constellation point mapping value and the metric of received signal on each layer, search for from last one deck of search volume, and when searching the i layer, with the current search radius be updated to described current search radius and (i+1) layer to the end one deck Fano base and difference, i is the positive integer more than or equal to 1;
In step 203, the branch of final metric value minimum as Search Results, and is consisted of the globular decoding result with the constellation point mapping value of this branch's correspondence.
From above-mentioned flow process as seen, what carry out among the present invention is the QR decomposition of ordering, and in the matrix R that obtains, the diagonal entry in its lower right corner is higher greater than the probability of the diagonal entry in the upper left corner.And these diagonal entries are relevant with the signal to noise ratio of corresponding received signal, be that diagonal entry is big more, signal to noise ratio is also big more, and the signal to noise ratio of the received signal of matrix R lower right corner diagonal entry correspondence is higher greater than the probability of the signal to noise ratio of the received signal of lower left corner diagonal entry correspondence among the present invention so.Since in the globular decoding first the estimated value of search be that lower right corner element by matrix R calculates and gets, begin to carry out that to search for correct probability first also higher like this from the received signal of signal to noise ratio maximum.And the present invention carries out after obtaining initial solution again in the search procedure, and to 2M layer direction search the time, every process one deck all bounces back search radius.Like this, the convergence rate of search radius is faster among the present invention, can some invalid branch of faster eliminating, find optimal solution, thereby computation complexity reduces further, and this point is particularly outstanding under the more situation of antenna.As seen, the present invention can reduce the occurrence probability of direction of search error diffusion greatly, reduces the number of times of adjusting the direction of search, thereby can reduce the computation complexity of globular decoding effectively.
In addition, can also before carrying out the QR decomposition, carry out least mean-square error (MMSE) preliminary treatment among the present invention, noise effect be considered in the channel matrix, and the QR that sorts through the pretreated channel matrix of MMSE is decomposed channel matrix.Above-mentioned MMSE preliminary treatment can reduce the noise component(s) that has been exaggerated in the channel matrix effectively, therefore can further reduce the computation complexity of globular decoding.
Below by two embodiment globular decoding method among the present invention is elaborated.
Globular decoding process in the present embodiment mainly comprises: parts such as MMSE preliminary treatment, QR decomposition and search optimal solution.Wherein, can adopt S-E mode the most commonly used to carry out the operation of search optimal solution.
Fig. 3 shows the flow chart of globular decoding method in the present embodiment.Referring to Fig. 3, this method comprises the steps:
In step 301, channel matrix is carried out the MMSE preliminary treatment.
In this step,, before carrying out follow-up globular decoding operation, earlier noise is joined newly in matrix in order to consider the noise effect in the signals transmission.Specifically, suppose that channel parameter is H,, then utilize channel parameter H to make up channel matrix because globular decoding only can be handled the one-dimensional space
Promptly
Wherein Rea1 (H) and Imag (H) are respectively real part and the imaginary part of channel parameter H.When carrying out the MMSE preliminary treatment, at channel matrix
The middle noise component(s) that adds, so pretreated channel matrix is
σ wherein
nBe noise parameter,
SNR is a signal to noise ratio, and I is a unit matrix.
In step 302, the QR that pretreated channel matrix is sorted decomposes, and determines the number of plies of search volume according to the line number of matrix R.
The QR that sorts in the present embodiment decomposes and is meant and makes that lower right corner diagonal entry decomposes greater than the higher QR of the probability of upper left corner diagonal entry among the matrix R.Decompose with the QR of routine and to compare, increased sorting operation in Pai Xu the QR decomposable process here, be about among the matrix Q that conventional QR decomposes squared absolute value and minimum column permutation and handle for working as the prostatitis.
The matrix R that obtains in this step is the upper triangular matrix of 2M * 2M, and M wherein is the twice of reception antenna number.In the present embodiment in the globular decoding search volume the included number of plies identical with the line number of matrix R, promptly total 2M layer.
In step 303～305, determine the received signal estimated value and the mapping value of this received signal on constellation point of each layer according to matrix R and received signal, and calculate the metric of each layer according to determined mapping value, begin to search for according to this metric last one deck then from the search volume, as Search Results, utilize the constellation point mapping value of each layer in the branch of Search Results place to form the received signal set branch of final metric value minimum.
In signals transmission, the received signal scope that every kind of modulation system is all corresponding certain, the signal that transmitting terminal sends can only be certain the signaling point in this scope, wherein each signaling point is all represented a constellation point.For example under the 16QAM, the value of constellation point has only 1 ,-1,3 and-3 four kind.Can therefore to estimate to received signal owing to all multifactor influences change behind the wireless channel of received signal process transmitting terminal and receiving terminal, and utilize the mapping value of estimated value searching on constellation point that obtains.
In the present embodiment, the received signal estimated value of i layer is: when i=2M,
When i＜2M,
Wherein
For the i of received signal is capable,
R
I, jBe the capable i column element of i of matrix R, symbol int represents to round operation.After obtaining the received signal estimated value, from the layer of place, select and the immediate constellation point of this estimated value the not mapped constellation point, the value of this constellation point constellation point mapping value as this received signal.Correspondingly, when calculating the metric of i layer, when i=2M,
When i＜2M,
As seen the metric of i layer has comprised the metric sum of (i+1) layer to the 2M layer in the present embodiment, the 1st layer metric can be called final metric value like this.
Here the path from 1 layer on 2M layer to the at every turn in will searching for is referred to as a branch, will be called forward towards the direction of search to the 1st layer from the 2M layer, and the 1st layer of direction of search towards the 2M layer is called oppositely.The search procedure here can comprise:
Step 2, carry out reverse search according to the current search radius, search for the i layer, whether judge current searched branched measurement value less than the current search radius, if i is updated to (i-1), if the i after upgrading equals 1, execution in step 3 is if the i after upgrading, then returns the operation of carrying out the i of search described in this step layer greater than 1; Otherwise, i is updated to (i+1), as current searched branch, if i=2M, then execution in step 4 with other branch except that current searched branch in this layer, otherwise return the operation of carrying out search i layer in this step;
Step 3, when searching ground floor, if the final metric value of current searched branch is less than the current search radius, the current search radius is updated to the final metric value of current searched branch, i is updated to (i+1) and preserves current separating, execution in step 2, otherwise i is updated to (i+1) and execution in step 2;
Step 4 is with the current branch that separates corresponding branch as the final metric value minimum that has preserved.
For instance, when seeking the branch of final metric value minimum, according to S-E globular decoding mode, at first in search first, calculate the received signal estimated value of 2M layer, for example equal-0.9, if that adopts is the modulation system of 16QAM and is to search for first, is-1 with the immediate constellation point mapping value of this estimated value so, utilize constellation point mapping value-1 to calculate the metric of 2M layer in this search again.Then calculate estimated value, constellation point mapping value and metric on (2M-1) layer in the search first, and the like, till the 1st layer.The 1st layer final metric value utilizes constellation point mapping value composition the current of received signal on each layer to separate as search radius in will searching for first this moment.After finishing search first, seek the metric lattice point littler to the 2nd layer than current search radius, if do not find, again to the 3rd layer of search, if get back to the 2M layer, received signal has been calculated in the estimated value-0.9 of this layer, from this layer, select and the immediate constellation point of this estimated value the not mapped constellation point so, with the value of selecteed constellation point as the constellation point mapping value in this search, for example be 1, and then utilize the constellation point mapping value in this search to calculate the metric of 2M layer in this search.If the metric that calculates, then calculates the constellation point mapping value and the metric of (2M-1) layer smaller or equal to search radius, and compares with search radius once more.Until reaching the 1st layer,, then utilize this metric to upgrade search radius if the 1st layer final metric value is less than search radius at this moment; If the great-than search radius then abandons this branch.After this, repeat said process again, finish all search, and will be on the 1st layer the branch of final metric value minimum as Search Results, the constellation point mapping value of each layer in this branch is formed optimal solution, gather as received signal.
Because when determining the constellation point mapping value at every turn, all be to select and the most approaching constellation point of estimated value the constellation point not mapped from the layer of place, therefore as can be seen, in the present embodiment for the 2M layer, preceding once the search branch on metric necessarily less than after once the search branch on metric; And, in the present embodiment among the matrix R lower right corner diagonal entry higher greater than the probability of upper left corner diagonal entry, therefore find that the probability of correct path is higher in can guaranteeing to search for first, thereby can reduce searching times, reduce computation complexity.
Above search procedure is identical with the search procedure in the existing S-E globular decoding.
So far, finish globular decoding process in the present embodiment.
Embodiment 2
In the present embodiment, when carrying out globular decoding, outside the MMSE preliminary treatment among the processing execution embodiment 1 and the QR of ordering decompose, also in each search procedure, adjust search radius, make this search radius consider the influence of Fano base than the corresponding search radius among the embodiment 1 according to noise parameter.
Fig. 4 shows the flow chart of globular decoding method in the present embodiment.Referring to Fig. 4, the globular decoding method in the present embodiment comprises:
In step 401～402, channel matrix is carried out the MMSE preliminary treatment, the QR that pretreated channel matrix is sorted decomposes, and determines the number of plies of search volume according to the line number of matrix R.
Two steps herein are identical to 302 operation with the step 301 among the embodiment 1.
In step 403～405, determine the received signal estimated value and the mapping value of this received signal on constellation point of each layer according to matrix R and received signal, and calculate the metric of each layer according to determined mapping value, begin search according to this metric and noise power from last one deck of search volume then, as Search Results, utilize the constellation point mapping value of each layer in the branch of Search Results place to form the received signal set branch of final metric value minimum.
Step 303 among three steps here and the embodiment 1 has been considered the Fano base to 305 comparatively similar when just determining search radius in the present embodiment in search procedure.Specifically, determined by forward lookup first current separate with final metric value after, in the process of reverse search, every search one deck, all search radius is updated to final metric value therewith before the layer each layer Fano base and difference, that is, final metric value and (i+1) layer to 2M layer Fano base and difference.Here the Fano base of i layer is:
R wherein
I, iBe i the diagonal entry of matrix R,
Be noise power, δ adjusts parameter for the performance of coming out by simulated measurement, and i is the positive integer in 1 to the 2M interval.And the value of δ is variable, when the δ value hour, can obtain preferable performance, but the complexity height; When the δ value is big, poor-performing, but complexity is lower.Can be by δ being adjusted the compromise that obtains preferable performance/complexity.In addition, the Fano base also can be predetermined fixed value or noise power
Multiple etc.
Be example with Fig. 1 still, the Fano base of supposing the 1st layer is 0.2.In search first, at first from second layer metric be 2 beginnings of 1.2 branch to the 1st layer of search, search metric and be 1.7 branch, the final metric value of this moment is 1.7, current separating is (1,3); During the search second layer, search radius is defined as equaling the Fano base that final metric value deducts the 1st layer, be 1.7-0.2=1.5, again according to of the 2nd layer search of this search radius to this branch, find the branch 3 littler than current search radius 1.5, this branch is 1.3 at the 2nd layer metric, is 1.5 to the 1st layer of minimum final metric value that searches this branch, during the search second layer, this moment, search radius was 1.5-0.2=1.3; Search in the 2nd layer according to 1.3 such search radius then, do not exist the branch littler than this search radius this moment, so optimal solution is the array that the constellation point mapping value of branch's 3 correspondences is formed, i.e. (1,1).Branch 1 tolerance of the second layer is 1.49 will be not searched, and conventional algorithm can be searched for this branch.Can reduce computation complexity like this.
Fig. 5 shows the grid schematic diagram of another kind of globular decoding.Include the four-layer network lattice among Fig. 5 altogether, suppose that the 1st to 4 layer Fano base is respectively 0.2,0.3,0.1 and 0.4.When searching for first, on the 4th layer, begin to search the 1st depth value 3.5 to the 1st layer from branch 2, current the separating that obtains this moment is (1,3 ,-1,3), search radius is 3.5; Then before carrying out reverse search, the Fano base of ground floor earlier bounces back search radius, equal 3.5-0.2=3.3, according to this search radius other lattice points on the 2nd layer of branch 2 are searched for, owing to there is not lattice point less than this search radius, therefore prepare to the 3rd layer of reverse search, make bounce back the 2nd layer Fano base of search radius this moment again, equal 3, and, get back to the 4th layer, and bounce back the 3rd layer Fano base of search radius at the 3rd layer of lattice point that does not also exist less than this search radius, equal 2.9, this moment, search finished first, current separating still for (1,3,-1,3).And then the startup forward lookup second time, every through 1 layer, all add the Fano base of this layer correspondence, promptly in branch 3, the 3rd layer search radius is 2.9+0.1=3, the 2nd layer search radius is 3+0.3=3.3, and the 1st layer search radius is 3.3+0.2=3.5, searches metric so and be 3.1 branch, current separating is (1,1,1,1), and when reverse search, every all making through one deck do not find the bounce back Fano base of this layer correspondence of search radius than 3.1 littler metrics in this branch, and 3.1 less than the final metric value 3.5 that obtains after the search first, therefore the optimal solution of this moment is (1,1,1,1).Searching for the 4th depth value according to mode as above afterwards is 1.49 branch 1, does not have more excellent separating in this branch.The search radius of this moment still is 2.9, because the 4th layer metric is 3.6 in the branch 4, and the great-than search radius, therefore all search finishes, and optimal solution is separate (1,1,1,1) that obtains in the branch 2.
So far, the globular decoding process in the end present embodiment.
From foregoing description as seen, after searching the 1st layer of each branch, when 2M layer direction search, whenever all search radius is bounced back in the present embodiment through one deck.Like this, the convergence rate of search radius is faster in the present embodiment, and comparing with embodiment 1 can some invalid branch of faster eliminating, find optimal solution more apace, thereby computation complexity reduces further, and this point is particularly outstanding under the more situation of antenna.
For the above embodiments 1 and embodiment 2, can also when searching for first, metric thresholding, the metric thresholding difference of each layer all be set to each layer.In this case, when searching for first, from the 2M layer time, have only the metric thresholding of the metric of this layer, just continue search, by that analogy, proceed to till the 1st layer up to search first less than correspondence.Fig. 6 shows the figure of search routine first that considers in the present embodiment under the metric thresholding situation, and referring to Fig. 6, the search first here comprises the steps:
Step 602～603, the i layer is searched for as working as anterior layer, calculate received signal estimated value and the mapping value of this received signal on constellation point when anterior layer, calculate the metric of this layer according to determined mapping value, judge that whether the metric of working as anterior layer is smaller or equal to predetermined metric thresholding when anterior layer, if then execution in step 604; Otherwise, execution in step 607.
Step 604～606 are updated to i (i-1), and judge and work as whether anterior layer is the 1st layer, if the metric that then will work as anterior layer is defined as final metric value, finish search first; Otherwise, return execution in step 602.
Step 607～609 are judged and are worked as whether anterior layer is the 2M layer, if then the search radius with each layer strengthens, and returns execution in step 602; Otherwise, i is updated to (i+1), return execution in step 602.
Here can adopt multiple mode to realize increasing, for example on the basis of former search radius, multiply by and strengthen the factor etc. search radius.
By the setting of metric thresholding, can find wrong path in advance and jump to other layer search, can further improve the probability of seeking optimal solution, thereby reduce computation complexity effectively.
Still be example with the grid among Fig. 5, suppose that the 1st, 2,3,4 layer metric thresholding is respectively 3.4,3,2.7 and 1.45.When searching for first, be 2 search of 1.2 branch at first since the 4th layer of moderate value, every search one deck all compares with the metric thresholding of this layer, when searching the 2nd layer since the metric 3.2 of this layer greater than the metric thresholding of correspondence, therefore can directly branch 2 be foreclosed.
Have, above-mentioned only is to be the explanation that example is carried out with the way of search in the S-E globular decoding again.Step 303 among the embodiment 1 to 305 and embodiment 2 in step 403 to 405 can also adopt in the F-Pohst globular decoding commonly used etc. way of search, repeat no more herein.
The above only is preferred embodiment of the present invention, and is in order to restriction the present invention, within the spirit and principles in the present invention not all, any modification of being made, is equal to replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (8)
1. a globular decoding method is characterized in that, this method comprises:
Treat the QR that the channel matrix of decoded signal sorts and decompose, determine the number of plies of search volume according to the line number of matrix R, and determine the Fano base of each layer;
Utilize the element among the matrix R to determine constellation point mapping value and the metric of received signal on each layer, begin the search volume is searched for from last one deck of search volume, and when searching the i layer, with the current search radius be updated to described current search radius and (i+1) layer to the end one deck Fano base and difference, i is the positive integer more than or equal to 1;
The branch of final metric value minimum as Search Results, and is consisted of the globular decoding result with the constellation point mapping value of this branch's correspondence.
2. the method for claim 1 is characterized in that, the described QR that channel matrix is sorted is decomposed into:
Squared absolute value among the matrix Q and minimum column permutation are decomposed for carry out QR when the prostatitis, among the matrix R lower right corner diagonal entry greater than the probability of upper left corner diagonal entry greater than 0.5.
3. the method for claim 1 is characterized in that, described last one deck is the 2M layer, and wherein M is the number of transmitting antenna, and the described search volume is searched for is:
B1. search for successively since the 2M layer, search ground floor and obtain final metric value, with the current search radius be updated to this metric and the 2nd layer to 2M layer Fano base and difference, current the separating that preservation is made up of the constellation point mapping value, and make that current layer number i=2, i are the positive integer between 1 to 2M;
B2. carry out search according to the current search radius, search for the i layer, judge that whether current searched branched measurement value is less than the current search radius, if, i is updated to (i-1), if the i after upgrading equals 1, execution in step B3, if the i after upgrading, then returns the operation of carrying out the i of search described in this step layer greater than 1; Otherwise, i is updated to (i+1), as current searched branch, if i=2M, then execution in step B4 carries out the operation of searching for the i layer in this step otherwise return with other branch except that current searched branch in this layer;
B3. when searching ground floor, if the final metric value of current searched branch is less than the current search radius, with the current search radius be updated to the final metric value of current searched branch and the 2nd layer to 2M layer Fano base and difference, i is updated to (i+1) and preserves current separating, execution in step B2, otherwise, i is updated to (i+1) and execution in step B2;
The current branch that separates corresponding branch as the final metric value minimum that B4. will preserve.
4. the method for claim 1 is characterized in that, the Fano base of described definite each layer is:
The Fano base of i layer is defined as equaling
R wherein
I, iBe i the diagonal entry of described matrix R,
Be noise power, δ adjusts parameter for the performance of coming out by simulated measurement, and i is the positive integer in 1 to the 2M interval;
Perhaps, the Fano base with each layer is defined as fixed value;
Perhaps, the Fano base with the i layer is defined as and noise power
Linear or non-linear relation.
5. as claim 3 or 4 described methods, it is characterized in that, set in advance the metric thresholding of each layer correspondence in the search first, step B1 is described since the search of 2M layer, obtains final metric value and is:
B11. with the i value for equaling 2M;
B12. the i layer is searched for as working as anterior layer, calculate when the received signal estimated value of anterior layer and the constellation point mapping value of this received signal, calculate the metric of this layer according to determined mapping value, judge that whether the metric of working as anterior layer is smaller or equal to predetermined metric thresholding when anterior layer, if, execution in step B13 then; Otherwise, execution in step B14;
B13. i is updated to (i-1), and judges and work as whether anterior layer is the 1st layer,, finish search first if the metric that then will work as anterior layer is defined as final metric value; Otherwise, return execution in step B12;
B14. when anterior layer is not the 2M layer, i is updated to (i+1), and returns execution in step B12.
6. method as claimed in claim 5 is characterized in that, further comprises among the described step B14:
When anterior layer is the 2M layer, strengthens the metric thresholding of each layer, and return execution in step B12.
7. the method for claim 1 is characterized in that, the QR that the described channel matrix for the treatment of decoded signal sorts further comprises: utilize noise parameter that channel matrix is carried out preliminary treatment before decomposing.
8. method as claimed in claim 7 is characterized in that, describedly utilizes noise parameter that channel matrix is carried out preliminary treatment to be: channel matrix transformation is treated to
Wherein
Be channel matrix, σ
nBe noise parameter,
M is a number of transmit antennas, and SNR is a signal to noise ratio, and I is a unit matrix.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007101430736A CN101373975A (en) | 2007-08-22 | 2007-08-22 | Spherical decoding method |
| JP2008207261A JP2009049997A (en) | 2007-08-22 | 2008-08-11 | Sphere decoding method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNA2007101430736A CN101373975A (en) | 2007-08-22 | 2007-08-22 | Spherical decoding method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101373975A true CN101373975A (en) | 2009-02-25 |
Family
ID=40447946
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNA2007101430736A Pending CN101373975A (en) | 2007-08-22 | 2007-08-22 | Spherical decoding method |
Country Status (2)
| Country | Link |
|---|---|
| JP (1) | JP2009049997A (en) |
| CN (1) | CN101373975A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101777967A (en) * | 2010-03-12 | 2010-07-14 | 北京天碁科技有限公司 | Method and device for selecting reserved constellation point and sphere decoding method and device |
| CN102882815A (en) * | 2012-09-25 | 2013-01-16 | 电信科学技术研究院 | Multi-input and multi-output data detection method and multi-input and multi-output data detection device |
| WO2014082487A1 (en) * | 2012-11-29 | 2014-06-05 | 中兴通讯股份有限公司 | Method and apparatus for soft output fixed complexity sphere decoding detection |
| CN106452686A (en) * | 2016-08-24 | 2017-02-22 | 重庆大学 | Radius update method based on sphere decoding algorithm and device |
-
2007
- 2007-08-22 CN CNA2007101430736A patent/CN101373975A/en active Pending
-
2008
- 2008-08-11 JP JP2008207261A patent/JP2009049997A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101777967A (en) * | 2010-03-12 | 2010-07-14 | 北京天碁科技有限公司 | Method and device for selecting reserved constellation point and sphere decoding method and device |
| CN101777967B (en) * | 2010-03-12 | 2012-12-05 | 北京天碁科技有限公司 | Method and device for selecting reserved constellation point and sphere decoding method and device |
| CN102882815A (en) * | 2012-09-25 | 2013-01-16 | 电信科学技术研究院 | Multi-input and multi-output data detection method and multi-input and multi-output data detection device |
| CN102882815B (en) * | 2012-09-25 | 2015-07-01 | 电信科学技术研究院 | Multi-input and multi-output data detection method and multi-input and multi-output data detection device |
| WO2014082487A1 (en) * | 2012-11-29 | 2014-06-05 | 中兴通讯股份有限公司 | Method and apparatus for soft output fixed complexity sphere decoding detection |
| US9356733B2 (en) | 2012-11-29 | 2016-05-31 | Zte Corporation | Method and apparatus for soft output fixed complexity sphere decoding detection |
| CN106452686A (en) * | 2016-08-24 | 2017-02-22 | 重庆大学 | Radius update method based on sphere decoding algorithm and device |
Also Published As
| Publication number | Publication date |
|---|---|
| JP2009049997A (en) | 2009-03-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | Multiple feedback successive interference cancellation detection for multiuser MIMO systems | |
| JP5243411B2 (en) | Method, system and computer program for determining signal vectors | |
| US8000422B2 (en) | Apparatus and method for detecting signal in multiple-input multiple-output (MIMO) wireless communication system | |
| JP4373439B2 (en) | Signal detection using sphere decoding technology | |
| KR100930522B1 (en) | Receiving device and method in multi-input wireless communication system | |
| US20070127589A1 (en) | Log-likelihood ration (LLR) generating apparatus and method in MIMO antenna communication system | |
| US8130877B2 (en) | Apparatus and method for detecting signal in multi-antenna system | |
| US8488721B2 (en) | Adaptive QRD-M algorithm based signal detecting method by using constellation set grouping in spatial multiplexing multiple-input multiple-output system | |
| Yang et al. | An improved matched-filter based detection algorithm for space-time shift keying systems | |
| CN107455001B (en) | Decoding based on tree search | |
| CN102723975B (en) | Signal detection method and device of MIMO (multiple input multiple output) system | |
| KR101923782B1 (en) | Weighted sequential decoding | |
| US8391378B2 (en) | Metric computation for lowering complexity of MIMO detection algorithms | |
| KR20070024753A (en) | Apparatus and method for detecting a signal in a mobile communication system using multiple input multiple output scheme | |
| KR101508700B1 (en) | Apparatus and method for detecting signal in multiple input multiple output wireless communication system | |
| CN101373975A (en) | Spherical decoding method | |
| US7835458B2 (en) | Method of decoding a spatially multiplexed signal and its corresponding receiver | |
| US8908806B2 (en) | Method and system for communicating data wirelessly using probabilistic data association | |
| US9066247B2 (en) | Communication devices and methods for signal detection | |
| Al-Semari et al. | Performance analysis of coherent TCM systems with diversity reception in slow Rayleigh fading | |
| US20040101066A1 (en) | Apparatus and method for transmitting and receiving signals using multiple antennas in mobile communication systems | |
| WO2006121296A1 (en) | Detector for multi-level modulated signal and detection method using the same, and iterative receiver for multi-level modulated signal and iteratively receiving method using the same | |
| CN116938394B (en) | MIMO detection method, receiving method and system thereof | |
| KR102441253B1 (en) | Apparatus and method for receiving signal in wireless communication system | |
| Sharma et al. | Performance Analysis of Space Time Trellis Codes Over Nakagami Fading & Rayleigh Fading Channels: With Comparative Analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20090225 |