CN103326813B - Single tree search List Sphere Decoder method and device - Google Patents

Abstract

The invention provides a kind of single tree search List Sphere Decoder method and device, by the maximum weights in the weights of the present node K paths with acquisitionCompare, i.e. introduce one and can determine the maximum weights obtainedThe benchmark deleted as single tree node, thus under ensureing performance loss acceptable premise, carry out node further and delete, i.e. ensure that the performance of single tree search List Sphere Decoder, further reduce complexity.

Description

Single tree search List Sphere Decoder method and device

Technical field

The invention belongs to mobile communication technology field, relate to signal detection aspect, search for soft ball particularly to one list tree Shape interpretation method and device.

Background technology

The heterogeneous linear detection algorithm of MIMO-OFDM system signal detection algorithm and non-linear detection algorithm two class.Conventional linear Algorithm mainly includes ZF (ZF) and least mean-square error (MMSE), and its complexity is minimum, but performance is worst.Conventional non-linear inspection Method of determining and calculating mainly includes that maximum likelihood (ML), interference eliminate (IC), the detection algorithm decomposed based on QR and globular decoding (SD), its Middle ML is the Optimum Detection in MIMO-OFDM system, but complexity is the highest, is unfavorable for processing in real time；Interference eliminates and base The detection algorithm decomposed in QR is calculated by simplification, it is achieved suboptimum Detection results, obtains preferably power between performance and complexity Weighing apparatus, but relative maximum likelihood detection, still have bigger performance loss, and interference eliminates the phenomenon that there will be error propagation；Spherical Decoding detection algorithm is proposed by Fincke and Pohst the earliest, is used for studying integer least square problem.Nearest globular decoding skill Art research shows: it reaches close can be substantially reduced (becoming polynomial relation with launching antenna number) in complexity on the premise of Or it is equal to the error performance of Maximum Likelihood Detection.Therefore, in actual MIMO-OFDM system, general receiving terminal uses SD to calculate Method approaches the performance of ML, applies some can reduce the strategy of complexity in SD algorithm simultaneously.

Assuming that MIMO-OFDM system model is y=Hs+n, wherein, y is N_RThe reception vector of × 1 dimension, H is N_R×N_TDimension Channel matrix, s is N_T× 1 dimension transmission symbolic vector, n be variance be σ²White Gaussian noise.Here N_R、N_TRepresent respectively and connect Receive antenna number, transmission antenna number.Then ML is wholeOptimal solution is found, such as following formula in space:

${\hat{s}}_{\mathrm{ML}}=\underset{s\∈C^{N_T}}{\arg \min }{\left|\right|y-\mathrm{Hs}\left|\right|}^2$

Globular decoding refer to the thought of Maximum Likelihood Detection, but it is to be for reception vector y, radius at a center Find all possible transmission symbolic vector in the ball of r, i.e. meet inequality:

||y-Hs||²≤r²

Owing to being in about beam radius r rather than wholeFind in space and send symbolic vector, thus decrease Amount of calculation.Here C represents the symbolic number that modulation constellation points comprises,Represent the space sending symbolic vector.

The tree search plan of globular decoding typically has depth-first, breadth First and tolerance preferential.For hard globular decoding, Being provided to find ML solution to reach to reduce complexity and the purpose of memory space as soon as possible, three class search plans have accordingly Corrective measure.To depth-first search, common are dynamically reduce initial radium, fixing each father node lower floor at most accesses joint Point, employing Schnorr Euchner sequence, QR sequence etc.；To BFS, common are fixing every layer of most reserve section Count, Schnorr Euchner sequence, QR sequence etc.；Preferential to tolerance, it is also possible to application sequence and the think of of maximum surviving path Think.But these thoughts can only ensure firmly to sentence the performance of BER, for the LTE system of application Turbo decoding, to have preferably BLER performance, then need to ensure a number of alternative symbolic vector, to produce Soft Inform ation fully the most accurately, i.e. needs to carry out soft Globular decoding.

List Sphere Decoder typically has following three kinds of schemes:

List sphere decoding LSD: this scheme calculates soft bit LLR (b according to search listing (assuming that size is K)_{L, n}) (send symbolic vector s l layer the n-th bit and be designated as b_{L, n}, 1≤l≤N_T, 1≤n≤N_c, N_cRepresent that single transmission symbol is corresponding Number of bits, relevant with modulation system), but when list is imperfect, it may appear that some bit value is empty phenomenon, or Choosing a radius the biggest and produce Soft Inform ation the most accurately, this will cause the highest complexity；Compose into Maximum, this will cause Soft Inform ation inaccurate；Approximate according to weights or the current search radius of symbolic vector in list Estimate that bit value is empty weights.

The hard sphere decoding SoD-SDA (some paper is also referred to as RTS:Repeated TreeSearch) of soft-decision: this Scheme first does once hard globular decoding, finds maximum likelihood solution s as early as possible^ML, corresponding bit is b^ML, then for each(represent rightNegate), try again hard globular decoding.The Soft Inform ation so obtained is the most accurately, but non-without being suspected to have The highest complexity.

Single tree search STS (Single Tree Search): this scheme is that search limit, limit calculates bit weights, is specially First find a solution, be initialized as ML and solve.In search every time afterwards, for non-leaf nodes, solve if ML can not be updated, or It is the node of current ML solution opposite bit weights, abandons this node and the branch below it；Otherwise access this node.For leaf Node, if its weights are less than the weights that current ML solves, then update ML and solves；If otherwise weights solve the power of opposite bit less than current ML Value, current weight to be updated to.Advantage be search limit, limit calculate bit weights, can save a part storage, shortcoming be to All bit values are not the most empty, need bigger initial radium, and the worst situation is intended to search for complete tree.

The STS advantage relative to RTS is through once searching for and just can find maximum likelihood solution s^MLWith each each bit of layerIn the STS advantage relative to LSD is identical initial radium, to reach identical decoding performance, the node that STS accesses Number is more less than LSD, and the memory space needed is the least.

For convenience of describing, remember that current maximum likelihood solution is s^ML, weights are λ^ML, corresponding bit is Represent s^MLThe L layer the n-th bit negates, and the minimum weights of its correspondence ares_kFor the path of current accessed node to root, corresponding weights areb_{L, n}For s_kCorresponding l layer the n-th bit value.According to the λ obtained^MLAndJust can be calculated soft by following formula Information:

$\mathrm{LLR}\left(b_{l,n}\right)=\left\{\begin{array}{c}\frac{1}{{\mathrm{\σ}}^2}({\mathrm{\λ}}^{\mathrm{ML}}-\stackrel{\‾}{{\mathrm{\λ}}_{l,n}^{\mathrm{ML}}}),b_{l,n}=1\\ \frac{1}{{\mathrm{\σ}}^2}(\stackrel{\‾}{{\mathrm{\λ}}_{l,n}^{\mathrm{ML}}}-{\mathrm{\λ}}^{\mathrm{ML}}),b_{l,n}=0\end{array}\right.$

The concrete search procedure of existing STS is as follows:

1) first ML is solved s^MLIt is initialized as sky, weights λ^ML=+∞, by corresponding for each for each layer bitAlso initialize For+∞；

2) scanning for, to present node, (weights are), if k=1, go to 3)；Otherwise go to 4)

3) if current sign vector s₁WeightsShow that can update ML solves.Then need to update as follows:

Right'sValue, is updated to λ^ML；

Then s^ML=s₁,

If current sign vector s₁WeightsAlthough ML then can not be updated to be solved, but following judgement need to be done:

If it is rightHaveThen update

Return to 2)；

4) now needing to judge to want to abandon present node and its subtree, refer to Fig. 1, it is existing STS node Deletion rule schematic diagram, as it is shown in figure 1,

IfMore than or equal to the maximum in set ΨThen explanation search present node and subtree thereof, can not Update s^MLCan not updateTherefore 2 are returned to after abandoning this node and subtree thereof)；Otherwise need to access present node.Wherein, $\mathrm{\&Psi;}=\left\{\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{l,n}^{\mathrm{ML}}}\right|(l\&GreaterEqual;k,n=\mathrm{1,2},...,N_C)I(b_{l,n}=\stackrel{\&OverBar;}{b_{l,n}^{\mathrm{ML}}})\left\}U\right\{\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{l,n}^{\mathrm{ML}}}\left|(l<k,n=\mathrm{1,2},...,N_C)\right\}.$

Although STS has the node of oneself to delete rule, but under conditions of high order modulation multiple antennas sends out receipts, need to access Nodes the biggest, complexity is the highest.The way reducing complexity at present is the maximum by setting soft bit Carrying out node to delete, this maximum sets too small, can not meet performance requirement, if too much, does not reduces complexity Degree.How on the premise of ensureing performance, delete node as much as possible, be the problem that have to solve of STS.

Summary of the invention

It is an object of the invention to provide a kind of single tree search List Sphere Decoder method and device, thus ensureing that performance is damaged Lose under acceptable premise, reduce complexity further.

For solving above-mentioned technical problem, the present invention provides a kind of single tree search List Sphere Decoder method, including:

Scan for, obtain K paths, obtain the maximum weights in this K paths

Residue node is continued search for, by the weights of present node and maximum weightsCompare, if this present node Weights more than or equal to these maximum weightsThen abandon present node and its subtree；If the weights of this present node are little In these maximum weightsThen access present node.

Optionally, in described list tree search List Sphere Decoder method, when the weights of this present node are more than or equal to being somebody's turn to do Maximum weightsAnd this present node is positioned at N_TDuring layer, terminate search.

Optionally, in described list tree search List Sphere Decoder method, when the weights of this present node are less than this maximum WeightsAnd present node is when being positioned at the 1st layer, update described maximum weightsAnd by present node place Path replacementThe K paths that path, place composition is new.

Optionally, in described list tree search List Sphere Decoder method, the maximum weights after renewalFor new K The value of maximum weight in paths.

Optionally, in described list tree search List Sphere Decoder method, after search terminates, if what existence was not updatedThen with more than maximum weightsIn a minimum right value update.

Optionally, in described list tree search List Sphere Decoder method, described K value is selected according to signal detection simulation result Take.

Optionally, in described list tree search List Sphere Decoder method, before obtaining K paths, utilizeCarry out Node is deleted.

The present invention also provides for a kind of single tree search List Sphere Decoder device, including:

First search module, in order to scan for, obtains K paths, obtains the maximum weights in this K paths

Second search module, in order to continue search for residue node, by the weights of present node and maximum weights Relatively, if the weights of this present node are more than or equal to these maximum weightsThen abandon present node and its subtree；If should The weights of present node are less than these maximum weightsThen access present node.

Optionally, in described list tree search List Sphere Decoder device, described second search module is grasped the most as follows Make:

When the weights of this present node are more than or equal to these maximum weightsAnd this present node is positioned at N_TDuring layer, Terminate search.

Optionally, in described list tree search List Sphere Decoder device, described second search module is grasped the most as follows Make:

When the weights of this present node are less than these maximum weightsAnd present node is when being positioned at the 1st layer, update described Maximum weightsAnd the path replacement by present node placeThe K paths that path, place composition is new.

Optionally, in described list tree search List Sphere Decoder device, the maximum weights after renewalFor new K The value of maximum weight in paths.

Optionally, in described list tree search List Sphere Decoder device, described second search module is grasped the most as follows Make:

After search terminates, if what existence was not updatedThen with more than maximum weightsIn a minimum power Value updates.

Optionally, in described list tree search List Sphere Decoder device, described K value is selected according to signal detection simulation result Take.

Optionally, in described list tree search List Sphere Decoder device, also include correcting module, terminate in order to search for After, if what existence was not updatedThen with more than maximum weightsIn a minimum right value update.

In the list tree search List Sphere Decoder method and device that the present invention provides, by weights and the acquisition of present node K paths in maximum weightsCompare, i.e. introduce one and can determine the maximum weights obtainedAs The benchmark that single tree node is deleted, thus under ensureing performance loss acceptable premise, carry out node further and delete, I.e. ensure that the performance of single tree search List Sphere Decoder, further reduce complexity.

Accompanying drawing explanation

Fig. 1 is existing STS knot removal rule schematic diagram；

Fig. 2 is the schematic flow sheet of the list tree search List Sphere Decoder method of the embodiment of the present invention；

Fig. 3 is the module diagram of the list tree search List Sphere Decoder device of the embodiment of the present invention.

Fig. 4 is the list tree schematic diagram of three layers.

Detailed description of the invention

The list tree search List Sphere Decoder method and device provided the present invention below in conjunction with the drawings and specific embodiments is made Further describe.According to following explanation and claims, advantages and features of the invention will be apparent from.It should be noted that It is that accompanying drawing all uses the form simplified very much, only in order to facilitate, to aid in illustrating lucidly the purpose of the embodiment of the present invention.

The core concept of the present invention is, it is provided that a kind of single tree search List Sphere Decoder method and device, by working as prosthomere The weights of point and the maximum weights in the K paths of acquisitionCompare, i.e. introduce one and can determine the maximum obtained WeightsThe benchmark deleted as single tree node, thus avoid whole single tree search in existing method and all pass through to set The maximum of fixed soft bitCarry out node to delete, be limited to thisValue, it is impossible to meet simultaneously and ensure performance and subtracting The problem of little complexity, improves the reliability that single tree node is deleted, i.e. ensure that the performance of single tree search List Sphere Decoder, subtracts Little complexity.

Refer to Fig. 2, its be the embodiment of the present invention list tree search List Sphere Decoder method schematic flow sheet.Such as Fig. 2 Shown in, described single tree search List Sphere Decoder method includes:

S10: scan for, obtains K paths, obtains the maximum weights in this K paths

S11: continue search for residue node, by the weights of present node and maximum weightsRelatively, if this is current The weights of node are more than or equal to these maximum weightsThen abandon present node and its subtree；If the power of this present node Value is less than these maximum weightsThen access present node.

Accordingly, the present embodiment also provides for a kind of single tree search List Sphere Decoder device, refer to Fig. 3, and it is the present invention The module diagram of the list tree search List Sphere Decoder device of embodiment.As it is shown on figure 3, described single tree search List Sphere Decoder dress Put and include:

First search module 20, in order to scan for, obtains K paths, obtains the maximum weights in this K paths

Second search module 21, in order to continue search for residue node, by the weights of present node and maximum weightsRelatively, if the weights of this present node are more than or equal to these maximum weightsThen abandon present node and its son Tree；If the weights of this present node are less than these maximum weightsThen access present node.

Concrete, in the list tree search List Sphere Decoder method and device that the present embodiment provides, for the deletion of node Mainly include that two stages are carried out:

First it is the first stage, utilizes the first search module 20 to perform step S10: to scan for, obtain K paths, obtain Take the maximum weights in this K pathsHere, described K paths i.e. includes K leaf node.Described K paths Searching method may utilize prior art and realizes, the most as described in the background art: first ML is solved s^MLBeing initialized as sky, initial radium sets For+∞, i.e. weights λ^ML=+∞, by corresponding for each for each layer bitAlso+∞ it is initialized as；Then, scan for, to currently (weights are node), if k=1, then enter s^ML、λ^MLAndRenewal, otherwise, then enter judge whether that needs are carried out Deleting of node.Carry out when deleting of node in this first stage, available prior art is more than or equal in set Ψ Big valueCarry out node to delete, wherein, $\mathrm{\&Psi;}=\left\{\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{l,n}^{\mathrm{ML}}}\right|(l\&GreaterEqual;k,n=\mathrm{1,2},...,N_C)I(b_{l,n}=\stackrel{\&OverBar;}{b_{l,n}^{\mathrm{ML}}})\left\}U\right\{\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{l,n}^{\mathrm{ML}}}\left|(l<k,n=\mathrm{1,2},...,N_C)\right\}.$ Described K value can be chosen according to signal detection simulation result, concrete, takes different numerical value for K and carries out certain emulation, example As, K value 3～30 is emulated, it is contemplated that the performance of single tree search List Sphere Decoder, in the receivable scope of performance In, to choose K value, be apparent from, K value is the least, then complexity is the least, and the performance of the most single tree search List Sphere Decoder will be affected； And K value is the biggest, then complexity is the biggest, but the performance of single tree search List Sphere Decoder can be protected.According to different tune The simulation result of mode processed can choose different K values, and the concrete value of K is not limited by the application.

Maximum weights in obtaining K paths and this K pathsAfterwards, perform second stage, utilize second Search module 21 performs step S11, here, due to when present node is leaf node, carries out node and deletes and have little significance, because of This, it is preferred that is when carrying out node and deleting, just for the 2nd to N_TThe node of layer, concrete:

If present node is not positioned at the 1st layer:

Residue node is continued search for, by the weights of present node and maximum weightsCompare, if this present node Weights more than or equal to these maximum weightsThen abandon present node and its subtree；If the weights of this present node are little In these maximum weightsThen access present node.

Here, use the depth-first STS of SE sequence, the weights for present node are more than or equal to these maximum weightsSituation, abandon present node and its subtree, specifically include the following process:

If this present node is in N_TDuring layer, terminate search, i.e. in addition to abandoning present node and its subtree, also Abandon other node do not searched for；

If this present node is in N_TDuring in-1 layer to the 2nd layer arbitrary one layer, then abandon present node and its son Tree, now, if need the node of search without residue, then terminates search, the most then searches for next node, be specially and work as prosthomere The next brother's node subsequent node of present node same layer (i.e. with) of point.

Weights for present node are less than these maximum weightsSituation, access present node, specifically include as More lower process:

If this present node is in N_TWhen layer is in the 2nd layer arbitrary one layer, accesses present node, i.e. obtain this node Information so that continuing to walk downward, i.e. continue search for the subtree under this node；

If present node is positioned at the 1st layer: according to circumstances may update s^ML、λ^MLAndIt is likely to update Wherein s^ML、λ^MLAndUpdate method identical with existing method,Update method as follows:

If its routine weight valueThen current path is replacedThe K bar road that path, place composition is new Footpath, remembers that the maximum of new K paths weights isDo not update.

If this present node is in the 1st layer, now the weights of present node are less than these maximum weightsI.e. illustrate At least can be rightIt is updated, now,The weights in path, present node place may be updated toAlso Likely it is updated to second largest weights in K paths, here, it is the most available only to need to do a comparison step, in a word, now, needs RightIt is updated, after being updatedMeanwhile, originallyCorresponding one in K paths After path also will be updatedCorresponding path is substituted.If here, shouldMinimum for weights in K paths A value, the most also will update λ^ML, i.e. occurOperation, additionally, right'sValue, is updated to thisRenewal process the most now will occur, specifically need which does and update, identical with considering of prior art.It should be noted that Be, application concerns be do not reduce performance in the case of, more, more easily deletion of node, say, that close In s^ML、λ^MLAndDeng the renewal of each several part content, can be accordingly with reference to prior art.

Delete according to the node that above-mentioned single tree search List Sphere Decoder method and device carries out Dan Shu, just can improve single burl The reliability that point is deleted, i.e. ensure that the performance of single tree search List Sphere Decoder, reduces complexity.

Further, in order to ensure singly to set the reliability of search List Sphere Decoder, existence is not also updated by the applicationI.e.Situation for initial value (being typically set at+∞) has given considering.After search terminates, exist and be not updatedThen with more than maximum weightsIn a minimum right value update.Now, search is over, and searches according to above-mentioned Rope, it may be determined that the weights in K paths are K minimum weights, and if what appearance was not updatedSituation, the most then Updated with the K+1 minimum weights.

Can more delete to further illustrate list tree search List Sphere Decoder method and device provided herein Node, subsequently will lift an example comprising concrete numerical value, to become apparent from, the advantage of distinct explanation the application.Refer to figure 4, it is the list tree schematic diagram of three layers.When utilizing existing method to search for this Dan Shu, at initial radium (r²>=1.8) search in the range of Suo Shi, needs to access complete tree, i.e. can not reduce complexity, and the result finally obtained is: $s^{\mathrm{ML}}=\left[\begin{array}{c}\mathrm{0,0}\\ \mathrm{0,0}\\ \mathrm{0,1}\end{array}\right],$ λ^ML=0.7 $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{l,n}^{\mathrm{ML}}}=\left[\begin{array}{c}\mathrm{1.7,1.3}\\ \mathrm{1.3,0.8}\\ \mathrm{0.9,0.9}\end{array}\right].$

And when utilizing the list tree search List Sphere Decoder method and device of the present embodiment to conduct interviews, take K=3, then search for To node v₇Search will be terminated, finally obtain $s^{\mathrm{ML}}=\left[\begin{array}{c}\mathrm{0,0}\\ \mathrm{0,0}\\ \mathrm{0,1}\end{array}\right],$ λ^ML=0.7 $\stackrel{\&OverBar;}{{\mathrm{\&lambda;}}_{l,n}^{\mathrm{ML}}}=\left[\begin{array}{c}\mathrm{1.7,1.3}\\ \mathrm{1.3,0.8}\\ \mathrm{0.9,0.9}\end{array}\right],$ Simultaneously Obtaining the K+1 minimum weights is 1.3.Here, due to terminate search time,It is not updated, thus needs by K+1 The weights approximation of individual minimum obtains, thus can affect the accuracy of this bit soft information, but owing to have foundIndividual alternative symbol In number vectorMinimum K, therefore the least, in tolerance interval on decoding performance impact.But need not access joint Point v₁₄、v₃、v₈、v₁₅、v₉、v₁₆, significantly reduce complexity.

Foregoing description is only the description to present pre-ferred embodiments, not any restriction to the scope of the invention, this Any change that the those of ordinary skill in bright field does according to the disclosure above content, modification, belong to the protection of claims Scope.

Claims (13)

1. a single tree search List Sphere Decoder method, it is characterised in that including:

Scan for, obtain K paths, obtain the maximum weights in this K paths

Residue node is continued search for, by the weights of present node and maximum weightsCompare, if the power of this present node Value is more than or equal to these maximum weightsThen abandon present node and its subtree；If the weights of this present node are less than being somebody's turn to do Maximum weightsThen access present node；Before obtaining K paths, utilizeCarry out node to delete,For collection Close the maximum in Ψ,Wherein,Represent bitCorresponding minimum weights, b_l,nPath s for current accessed node to root_KCorresponding l layer n-th compares Paricular value,It is right to representNegate,Represent maximum likelihood solution s^MLL layer the n-th bit, N_cRepresent single transmission symbol pair The number of bits answered.

2. single tree search List Sphere Decoder method as claimed in claim 1, it is characterised in that when the weights of this present node are big In equal to these maximum weightsAnd this present node is positioned at N_TDuring layer, terminate search, N_TRepresent last layer.

3. single tree search List Sphere Decoder method as claimed in claim 1, it is characterised in that when the weights of this present node are little In these maximum weightsAnd present node is when being positioned at the 1st layer, update described maximum weightsAnd prosthomere will be worked as The path replacement at some placeThe K paths that path, place composition is new.

4. single tree search List Sphere Decoder method as claimed in claim 3, it is characterised in that the maximum weights after renewalFor the value of maximum weight in new K paths.

5. single tree search List Sphere Decoder method as claimed in claim 3, it is characterised in that after search terminates, if existing not It is updatedThen with more than maximum weightsIn a minimum right value update,Represent bitCorresponding Minimum weights.

6. single tree search List Sphere Decoder method as claimed in claim 1, it is characterised in that described K value is according to signal detection Simulation result is chosen.

7. a single tree search List Sphere Decoder device, it is characterised in that including:

First search module, in order to scan for, obtains K paths, obtains the maximum weights in this K paths

Second search module, in order to continue search for residue node, by the weights of present node and maximum weightsRelatively, If the weights of this present node are more than or equal to these maximum weightsThen abandon present node and its subtree；If this is current The weights of node are less than these maximum weightsThen access present node；

Before obtaining K paths, utilizeCarry out node to delete,For gathering the maximum in Ψ,Wherein,Represent ratio SpecialCorresponding minimum weights, b_l,nPath s for current accessed node to root_KCorresponding l layer the n-th bit value,Table It is right to showNegate,Represent maximum likelihood solution s^MLL layer the n-th bit, N_cRepresent the bit that single transmission symbol is corresponding Number.

8. single tree search List Sphere Decoder device as claimed in claim 7, it is characterised in that described second search module also enters The following operation of row:

When the weights of this present node are more than or equal to these maximum weightsAnd this present node is positioned at N_TDuring layer, terminate Search, N_TRepresent last layer.

9. single tree search List Sphere Decoder device as claimed in claim 7, it is characterised in that described second search module also enters The following operation of row:

When the weights of this present node are less than these maximum weightsAnd present node is when being positioned at the 1st layer, update described maximum WeightsAnd the path replacement by present node placeThe K paths that path, place composition is new.

10. single tree search List Sphere Decoder device as claimed in claim 9, it is characterised in that the maximum weights after renewalFor the value of maximum weight in new K paths.

11. single tree search List Sphere Decoder devices as claimed in claim 9, it is characterised in that described second search module is also Proceed as follows:

After search terminates, if what existence was not updatedThen with more than maximum weightsIn minimum weights more Newly,Represent bitCorresponding minimum weights.

12. single tree search List Sphere Decoder devices as claimed in claim 7, it is characterised in that described K value is according to signal detection Simulation result is chosen.

List tree search List Sphere Decoder device as described in any one in 13. such as claim 7 to 10 or 12, it is characterised in that Also including correcting module, after searching for and terminating, not being updated if existingThen with more than maximum weightsIn A minimum right value update,Represent bitCorresponding minimum weights.