CN101582750A - Sphere decoding detection method based on breadth-first search - Google Patents
Sphere decoding detection method based on breadth-first search Download PDFInfo
- Publication number
- CN101582750A CN101582750A CNA200910086118XA CN200910086118A CN101582750A CN 101582750 A CN101582750 A CN 101582750A CN A200910086118X A CNA200910086118X A CN A200910086118XA CN 200910086118 A CN200910086118 A CN 200910086118A CN 101582750 A CN101582750 A CN 101582750A
- Authority
- CN
- China
- Prior art keywords
- node
- search
- weights
- matrix
- layer
- 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.)
- Granted
Links
Images
Landscapes
- Radio Transmission System (AREA)
Abstract
The invention provides a sphere decoding detection method based on breadth-first search, comprising the following steps: step A, carrying out QR decomposition on channel matrix H to obtain Q matrix and R matrix; step B, multiplying conjugate transpose of Q matrix with received signal y to obtain equalizing signal rho of the received signal; step C, setting search nodes of Ki, i=1, 2, ..., N<T>, and N<T> is the number of sending antennae; and step D, carrying out breadth-first search on R matrix and rho, wherein preserving Ki nodes with the lowest weight when carrying out the i layer search. The method of the invention can effectively reduce operating complexity of the sphere decoding and is easy to realize by hardware.
Description
Technical field
The invention belongs to wireless communication field, be particularly related to a kind of method for detecting spherical decode based on breadth-first search that is used for multiple-input and multiple-output (MIMO) system, the present invention also can be applied in OFDM (OFDM) and the mimo system detection to the MIMO signal.
Background technology
In present wireless communication standard and evolution process thereof, the MIMO antenna technology is widely adopted.No matter be that (long term evolution LTE), still in the 802.16 series technique evolution versions, is extensive use of OFDM and MIMO technology as key technology in the 3GPP Long Term Evolution.Compare with traditional single input and output (SISO) system, the reception of mimo system is that overlapping mutually situation is carried out the MIMO input on time and frequency domain, and therefore, MIMO input complexity is much higher than traditional SISO input.
In theory, (Maximum likelihood, ML) detection method detects can to pass through maximum likelihood to the MIMO signal.But Maximum Likelihood Detection needs the planisphere of traversal search to count along with the increase exponentially growth of number of transmit antennas, the modulation system degree of freedom, and more than number of transmit antennas and under the situation of high order modulation, its computational complexity is difficult to bear in real system.Therefore, seek the performance detection method that complexity reduces greatly near ML, just become the key factor that can the MIMO detection technique realize in real system.
So Viterbo etc. have proposed globular decoding (Sphere Decoding, SD) algorithm to the signal with lattice-shaped planisphere on the research basis of Pohst etc.And Damen is generalized to the MIMO input to this algorithm, and (V-BLAST) detects more performance when obtaining than vertical bell laboratories layered space.
Yet there is an adverse factors in traditional globular decoding, and that is exactly different channel condition, signal quality and initial radium, and its algorithm computation complexity is different.Particularly, might cause whole algorithm not restrain, cause the paralysis of system for singular matrix.Therefore, how effectively to control the stability and the robustness of globular decoding algorithm complex, system, for a real-time radio communication system, be even more important.For these reasons, so occurred utilizing the breadth-first algorithm to replace the K-Best algorithm of depth-first algorithm, its core concept is when every layer of search optimal path, only keeps K node of weights minimum, K node from this reservation continues the search to lower floor then, up to the bottom.
But, for the K-Best algorithm, also how exist from K * M
c(M
cBe that modulation is counted) choose optimum K node in the individual node and how under the situation of guaranteed performance and maximum likelihood ML (MaximumLikelihood) performance basically identical, reduce the problem of the complexity of algorithm as far as possible.
Summary of the invention
Technical problem to be solved by this invention provides a kind of method for detecting spherical decode based on breadth-first search, with the computational complexity of reduction globular decoding, and is easy to realize by hardware.
For solving the problems of the technologies described above, it is as follows to the invention provides technical scheme:
A kind of method for detecting spherical decode based on breadth-first search comprises the steps:
A, channel matrix H is carried out QR decompose, obtain Q matrix and R matrix;
B, the conjugate transpose and the received signal y of Q matrix multiplied each other, obtain the equalizing signal rho of received signal;
C, the search node that the i layer is set are counted K
i, i=1,2 ..., N
T, N
TBe number of transmit antennas;
D, carry out breadth-first search, wherein, when carrying out the search of i layer, keep the K of weights minimum according to described R matrix and ρ
iIndividual node.
Above-mentioned method for detecting spherical decode, in the steps A, the QR that described QR is decomposed into ordering decomposes, and makes the mould value of i element on the diagonal of R matrix be not more than the mould value of i+1 element.
Above-mentioned method for detecting spherical decode, among the step C, the search node that described i layer is set according to modulation system, target bit and channel condition information is counted K
i
Above-mentioned method for detecting spherical decode, among the step C, the search node of set i layer is counted K
iThe search node that is not less than the i+1 layer is counted K
I+1
Above-mentioned method for detecting spherical decode, described step D specifically comprises:
D1, the sphere decoding expression determined according to described R matrix and ρ are carried out table-looking-up sequencing to the 1st layer node, keep the K of Euclidean distance minimum
1Individual node, and calculate the K that keeps
1The weights of individual node;
D2, when carrying out the search of i layer, the K that the i-1 layer is kept respectively
I-1Individual node carries out path expansion, and to the child node that the path expansion the obtains ordering by merging of dividing and ruling, keeps the K of weights minimum
iIndividual node;
D3, searched for last one deck after, output decode results.
Above-mentioned method for detecting spherical decode, among the step D2, the described ordering by merging of dividing and ruling comprises:
D21, respectively to described K
I-1The child node of individual node is carried out table-looking-up sequencing according to described R matrix and the definite sphere decoding expression of ρ, and calculates described K
I-1The weights of the child node of individual node;
D22, according to the table-looking-up sequencing result of step D21, the weights of the child node of the weights of the child node of the 1st node of i-1 layer and the 2nd node are compared, select the K of weights minimum
iIndividual child node, then, with the K of the described weights minimum of selecting
iThe weights of the child node of the weights of individual child node and the 3rd node compare, and select the K of weights minimum again
iIndividual child node, the rest may be inferred, with the K of the weights minimum selected at last
iIndividual child node is as the reservation node of i layer.
Above-mentioned method for detecting spherical decode is among the step D21, for described K
I-1Each node in the individual node calculates the weights of the child node of different numbers.
The present invention improves traditional K-best algorithm, comprises that channel matrix is carried out QR to be decomposed and carry out equilibrium treatment to received signal, and self adaptation is provided with the node number of every layer of search, thereby has reduced the computational complexity of globular decoding.The QR that the present invention also further sorts to channel matrix decomposes, and, adopt lookup table mode to carry out the intranodal ordering, on the basis of described intranodal ordering, merge algorithm carries out a layer internal sort based on dividing and ruling, further reduced the computational complexity of globular decoding, and be easy to realize by VLSI hardware.
Description of drawings
Fig. 1 is basic mimo system illustraton of model;
Fig. 2 is the method for detecting spherical decode flow chart based on breadth-first search of the embodiment of the invention.
Embodiment
For making the purpose, technical solutions and advantages of the present invention clearer, describe the present invention below in conjunction with the accompanying drawings and the specific embodiments.
For basic mimo system model as shown in Figure 1, suppose that number of transmit antennas is N
t, the reception antenna number is N
R, its channel is a falt fading channel, then this system can be with shown in the following formula:
y=H·s+n (1)
Wherein y is N
R* 1 received signal vector, s are N
t* 1 emission signal vector, n is N
R* 1 noise vector, its average are 0, and variance is N
0/ 2, H is N
R* N
tThe channel model vector (channel matrix) of dimension.
Adopt the maximal possibility estimation algorithm, expression formula is as follows:
Wherein, Ω is effective modulation signal point.In order to reduce the maximum likelihood algorithm complexity, traditional K-Best algorithm core concept is based on the thought of breadth-first, promptly selects some nodes at every layer, and then carries out the process of path expansion and node searching.
The method for detecting spherical decode based on breadth-first of the embodiment of the invention improves traditional K-Best algorithm.At first, H is carried out the QR decomposition can obtain unitary matrice Q and upper triangular matrix R, obtain following expression according to formula (2):
Because R is a upper triangular matrix, therefore can utilize the method for iteration, obtain following expression formula (4):
Wherein,
r
I, jThe capable element that is listed as with j of representing matrix R i, ()
HExpression is asked conjugate transpose to (),
Be the demodulation vector,
For
J element of vector.
In the present invention, for convenience of description, with the e in the above formula
iThe Euclidean distance that is called node, E
iThe weights that are called node.
With reference to Fig. 2, the method for detecting spherical decode based on breadth-first search of the embodiment of the invention mainly comprises the steps:
Step 201: channel matrix H is carried out QR decompose, obtain Q matrix and R matrix;
Preferably, the QR that described QR is decomposed into ordering decomposes, and obtains Q matrix, R matrix and P matrix (ordering switching matrix).QR by ordering decomposes, and makes the mould value of i element on the diagonal of R matrix be not more than the mould value of i+1 element, that is, the diagonal entry of described R matrix from the upper left corner to the lower right corner according to mould value sequence arrangement from small to large.The QR of ordering decomposes, and is equivalent to search tree is carried out the interlayer ordering.By described interlayer ordering, make that the Euclidean distance of upper level node is relatively large, the node number of every layer of search is set, can reduce the not searched probability that arrives of correct constellation point in conjunction with self adaptation in the subsequent step, thereby the raising search speed, the computational complexity of reduction globular decoding.
Step 202: the conjugate transpose and the received signal y of Q matrix are multiplied each other, obtain the equalizing signal rho of received signal, that is, and ρ=Q
HY;
In the present invention, be with the received signal of ρ as equivalence, with the channel matrix of R matrix, and make up search tree according to R matrix and ρ and carry out globular decoding as equivalence.
Step 203: the search node that the i layer is set is counted K
i, i=1,2 ..., N
T, N
TBe number of transmit antennas;
In traditional K-Best algorithm, the node number of every layer of search is fixed; The present invention improves it, according to different system parameterss and channel parameter every layer K value is set flexibly.
Particularly, can count K according to the search node that modulation system, target bit and channel condition information (CSI) is provided with described i layer
i(this makes that every layer of node number of being searched for is adaptive).Preferably, the search node that the i layer can also be set is counted K
iThe search node that is not less than the i+1 layer is counted K
I+1, that is, the node number of high-rise search is more relatively.The node number of high-rise search is more relatively, can reduce the not searched probability that arrives of correct constellation point.
Step 204: carry out breadth-first search according to described R matrix and ρ, wherein, when carrying out the search of i layer, keep the K of weights minimum
iIndividual node.
D1, the sphere decoding expression determined according to described R matrix and ρ are carried out table-looking-up sequencing to the 1st layer node, keep the K of Euclidean distance minimum
1Individual node, and calculate the K that keeps
1The weights of individual node;
D2, when carrying out the search of i layer, the K that the i-1 layer is kept respectively
I-1Individual node carries out path expansion, and to the child node that the path expansion the obtains ordering by merging of dividing and ruling, keeps the K of weights minimum
iIndividual node;
D3, when searching last one deck, the output decode results.In this step, be that the constellation point mapping value with the path of the bottom node correspondence of final weights minimum consists of the globular decoding result.Need to prove that what carry out is under the QR of the ordering situation of decomposing in step 201, also need sort to described constellation point mapping value that the sequence that ordering is obtained is as final decode results according to described P matrix.
The described ordering by merging of dividing and ruling comprises:
D21, respectively to described K
I-1The child node of individual node is carried out table-looking-up sequencing according to described R matrix and the definite sphere decoding expression of ρ, and calculates described K
I-1The weights of the child node of individual node;
In this step, preferably, for described K
I-1Each node in the individual node can calculate the weights (that is, can only calculate the weights of the parton node of this node) of the child node of different numbers according to actual conditions.In following step D22, only the child node of having calculated weights is compared.
D22, according to the table-looking-up sequencing result of step D21, the weights of the child node of the weights of the child node of the 1st node of i-1 layer and the 2nd node are compared, select the K of weights minimum
iIndividual child node, then, with the K of the described weights minimum of selecting
iThe weights of the child node of the weights of individual child node and the 3rd node compare, and select the K of weights minimum again
iIndividual child node, the rest may be inferred, with the K of the weights minimum selected at last
iIndividual child node is as the reservation node of i layer.
Child node for a node sorts, and conventional method is to calculate the Euclidean distance of each child node earlier, and the size according to Euclidean distance sorts then, sorts the algorithm complex height as adopting methods such as bubbling, insertion.In the present invention, do not need to calculate earlier the Euclidean distance of each child node, but directly carry out table-looking-up sequencing, thereby reach the purpose that reduces algorithm complex and elevator system performance according to the sphere decoding expression of determining by R matrix and ρ (3).
So-called table-looking-up sequencing is meant: carry out iterative according to sphere decoding expression (3), obtain the demodulation vector
Determine
Vector is at the component when anterior layer, and this component is in the position in the source signal planisphere of anterior layer; Relative size according to the distance (Euclidean distance) between the position of each constellation point (node) of working as anterior layer source signal planisphere and described component sorts to described node.Because constellation point is regular distribution in the planisphere, therefore, after determining described position, do not need to calculate Euclidean distance, just can directly determine the relative size of the Euclidean distance between all constellation point and this position according to this regularity of distribution.
After adopting lookup table mode to carry out the intranodal ordering, just can be on the basis of described intranodal ordering, carry out a layer internal sort based on the merge algorithm of dividing and ruling, can reduce the complexity of sort algorithm on the one hand, be convenient to the realization of VLSI hardware on the other hand by look-up method.
Complexity to traditional K-Best algorithm and algorithm of the present invention compares following (suppose the K=8 when anterior layer, modulation system is 64QAM, and traditional algorithm adopts the ordering of bubbling method):
Computation complexity | Tradition K-Best algorithm | Algorithm of the present invention |
The intranodal ordering | 63+62+...+57+56=504 | 0 |
Layer internal sort | 511+510+...+504=4060 | 20×7=140 |
As can be seen from the above table, method of the present invention can reduce computation complexity greatly.
Should be noted that at last, above embodiment is only unrestricted in order to technical scheme of the present invention to be described, those of ordinary skill in the art is to be understood that, can make amendment or be equal to replacement technical scheme of the present invention, and not breaking away from the spiritual scope of technical solution of the present invention, it all should be encompassed in the middle of the claim scope of the present invention.
Claims (7)
1. the method for detecting spherical decode based on breadth-first search is characterized in that, comprises the steps:
A, channel matrix H is carried out QR decompose, obtain Q matrix and R matrix;
B, the conjugate transpose and the received signal y of Q matrix multiplied each other, obtain the equalizing signal rho of received signal;
C, the search node that the i layer is set are counted K
i, i=1,2 ..., N
T, N
TBe number of transmit antennas;
D, carry out breadth-first search, wherein, when carrying out the search of i layer, keep the K of weights minimum according to described R matrix and ρ
iIndividual node.
2. method for detecting spherical decode as claimed in claim 1 is characterized in that:
In the steps A, the QR that described QR is decomposed into ordering decomposes, and makes the mould value of i element on the diagonal of R matrix be not more than the mould value of i+1 element.
3. method for detecting spherical decode as claimed in claim 1 is characterized in that:
Among the step C, the search node that described i layer is set according to modulation system, target bit and channel condition information is counted K
i
4. method for detecting spherical decode as claimed in claim 1 is characterized in that:
Among the step C, the search node of set i layer is counted K
iThe search node that is not less than the i+1 layer is counted K
I+1
5. method for detecting spherical decode as claimed in claim 1 is characterized in that, described step D specifically comprises:
D1, the sphere decoding expression determined according to described R matrix and ρ are carried out table-looking-up sequencing to the 1st layer node, keep the K of Euclidean distance minimum
1Individual node, and calculate the K that keeps
1The weights of individual node;
D2, when carrying out the search of i layer, the K that the i-1 layer is kept respectively
I-1Individual node carries out path expansion, and to the child node that the path expansion the obtains ordering by merging of dividing and ruling, keeps the K of weights minimum
iIndividual node;
D3, searched for last one deck after, output decode results.
6. method for detecting spherical decode as claimed in claim 5 is characterized in that, among the step D2, the described ordering by merging of dividing and ruling comprises:
D21, respectively to described K
I-1The child node of individual node is carried out table-looking-up sequencing according to described R matrix and the definite sphere decoding expression of ρ, and calculates described K
I-1The weights of the child node of individual node;
D22, according to the table-looking-up sequencing result of step D21, the weights of the child node of the weights of the child node of the 1st node of i-1 layer and the 2nd node are compared, select the K of weights minimum
iIndividual child node, then, with the K of the described weights minimum of selecting
iThe weights of the child node of the weights of individual child node and the 3rd node compare, and select the K of weights minimum again
iIndividual child node, the rest may be inferred, with the K of the weights minimum selected at last
iIndividual child node is as the reservation node of i layer.
7. method for detecting spherical decode as claimed in claim 6 is characterized in that:
Among the step D21, for described K
I-1Each node in the individual node calculates the weights of the child node of different numbers.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN200910086118XA CN101582750B (en) | 2009-06-02 | 2009-06-02 | Sphere decoding detection method based on breadth-first search |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN200910086118XA CN101582750B (en) | 2009-06-02 | 2009-06-02 | Sphere decoding detection method based on breadth-first search |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101582750A true CN101582750A (en) | 2009-11-18 |
CN101582750B CN101582750B (en) | 2012-11-21 |
Family
ID=41364736
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN200910086118XA Expired - Fee Related CN101582750B (en) | 2009-06-02 | 2009-06-02 | Sphere decoding detection method based on breadth-first search |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101582750B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101997657A (en) * | 2010-11-03 | 2011-03-30 | 北京邮电大学 | Detection method for breadth-first sphere decoding in MIMO (multiple input multiple output) system |
CN102006148A (en) * | 2010-12-07 | 2011-04-06 | 西安电子科技大学 | Multiple-input multiple-output (MIMO) signal detection method based on breadth-first tree search |
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 |
CN103973602A (en) * | 2013-01-28 | 2014-08-06 | 电信科学技术研究院 | Signal detecting method and device |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101064579B (en) * | 2006-04-25 | 2011-05-25 | 上海无线通信研究中心 | Method for detecting low-complexity globular decoding |
-
2009
- 2009-06-02 CN CN200910086118XA patent/CN101582750B/en not_active Expired - Fee Related
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101997657A (en) * | 2010-11-03 | 2011-03-30 | 北京邮电大学 | Detection method for breadth-first sphere decoding in MIMO (multiple input multiple output) system |
CN101997657B (en) * | 2010-11-03 | 2013-01-09 | 北京邮电大学 | Detection method for breadth-first sphere decoding in MIMO (multiple input multiple output) system |
CN102006148A (en) * | 2010-12-07 | 2011-04-06 | 西安电子科技大学 | Multiple-input multiple-output (MIMO) signal detection method based on breadth-first tree search |
CN102006148B (en) * | 2010-12-07 | 2013-04-17 | 西安电子科技大学 | Multiple-input multiple-output (MIMO) signal detection method based on breadth-first tree search |
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 |
CN103973602A (en) * | 2013-01-28 | 2014-08-06 | 电信科学技术研究院 | Signal detecting method and device |
Also Published As
Publication number | Publication date |
---|---|
CN101582750B (en) | 2012-11-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105554865B (en) | A kind of MIMO-SCMA system down link design method based on STBC | |
CN101136676B (en) | Mimo wireless data transmission system, device and method | |
CN101662342B (en) | Multi-input multi-output signal detection method and device | |
WO2006130804A1 (en) | Sphere decoding aparatus for mimo channel | |
CN107026681A (en) | Signaling method, method of reseptance, transmitter and receiver | |
KR20180004102A (en) | Tree-based decoding | |
CN107070525A (en) | Parametrization order is decoded | |
CN101582750A (en) | Sphere decoding detection method based on breadth-first search | |
US10374772B2 (en) | Method for slicing K-best detection in multiple-input multiple-output wireless communications system | |
CN101562464B (en) | Method for detecting spherical decode based on depth-first search | |
US10084526B2 (en) | Path detection method and device, and sphere decoding detection device | |
CN101557269A (en) | Sphere decoding detection method based on ultra large scale integrated circuit | |
Lai et al. | Efficient channel-adaptive MIMO detection using just-acceptable error rate | |
CN101594202B (en) | Method and device for detecting sphere decoding | |
CN102355295B (en) | High-efficiency reception method for multi-antenna OFDM (Orthogonal Frequency Division Multiplexing) system | |
US9143376B2 (en) | Efficient tree-based MIMO OFDM detection | |
CN108293029A (en) | Utilize stack rearrangement carry out sequence decoding | |
Turhan et al. | Deep learning aided generalized frequency division multiplexing | |
US10615912B2 (en) | Apparatuses, computer-readable media, and methods for single dimension maximal likelihood symbol detection | |
KR101348557B1 (en) | Method for detecting signal using mimo-ofdm system and apparatus thereof | |
Park et al. | The Reinforcement Learning based Interference Whitening Scheme for 5G | |
Al-Nahhal et al. | Reduced complexity K-best sphere decoding algorithms for ill-conditioned MIMO channels | |
Hui et al. | MIMO detection techniques based on low complexity adaptive QR-decomposition with M-algorithm for 3GPP LTE systems | |
Liu et al. | A lower complexity QRD-M algorithm based on VBLAST for MIMO-OFDM systems | |
CN105978616B (en) | In conjunction with the extensive mimo system LAS signal detecting method of hereditary property |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20121121 Termination date: 20180602 |
|
CF01 | Termination of patent right due to non-payment of annual fee |