TW201014235A - Multiple-input multiple-output detection method and detector - Google Patents

Abstract

This invention provides a multiple-input multiple-output (MIMO) detection method and a detector which can be in accordance with various modulation techniques and have low computation complexity. The multiple-input multiple-output (MIMO) detection method comprises the following steps: (A) performing the following sub-steps to every path arriving a current level: (A1) to the current level, according to a received signal corresponding to the processed received signal vector, calculating an interference-removing received signal in which the interference of the previous level is removed; and (A2) to each of the subsets of a first target number divided by a constellation diagram, performing the following step: finding out a constellation point in the subset closest to the interference-removing received signal, extending a branch representing the constellation point from the path to the next level, and calculating the length of the branch and a corresponding path length.

Description

201014235 IX. Description of the invention: [Technical field of the invention] The present invention relates to a detector and a detection method thereof, and more particularly to a multi-input multi-output detector and a detection method thereof. [Prior Art] Multiple-input multiple-output (技术) technology has been widely incorporated into various wireless communication standards to improve spectral efficiency. In a MIMO system, the transmitter cuts the data into a multi-symbol stream, and uses a plurality of antennas to transmit the stream at the same time and in the same frequency band; the receiving end receives these by using most antennas. The sum of the stream of characters is separated and the streams are separated to obtain the characters originally transmitted by the transmitting end. It is currently accepted that the receiver based on the most likely panoramic detection (MLD) has the best performance (ie, the lowest error rate), but its computational complexity is very high. In 2002, Kim et al. proposed a combination of QR decomposition (QRD) and M algorithm detection techniques (hereinafter referred to as QRD-M detection), which can be found through tree search. The most approximate approximation solution, while reducing the computational complexity, still retains the best performance that is most likely to be detected. - I. Signal Model r In a MIMO system with one transmit antenna and one receive antenna, it is assumed that the far-reaching between the transmit and receive antennas is a flat fading channel. The relationship between the characters transmitted by the antenna and the signals received by the receiving antenna can be expressed as: 201014235

Less 2 h2X

klN, ' • V '«1 " ^2N, • + »2 hNrN, _ • XNt _ Jhr_ Equation (1) or abbreviated as y=Hx+n, where x is a transfer character vector, ~ is the 7th The character transmitted by the transmitting antenna is a constellation point on a constellation; y is a received signal vector, 妁 is the signal received by the receiving antenna; Η is a channel matrix, Is the channel response between the yth transmit antenna and the /receive antenna, and can be assumed to be independent and identically distributed 'complex Gaussian rand〇m variable', eight averaging The value is 0, the variogram is 1; n is a noise matrix, ~ is the noise received by the / receiving antenna and can be assumed to be independent and identically distributed complex Gaussian random variables, the average of which is 〇, The number of variograms is II.QR decomposition 藉 by QR decomposition 'Tongzhi Tuo Ling η ή γ main-from the hard matrix Η can not be H = QR, where Q size is NrxNt and has orthogonal rows of single-matrix ) (satisfying QQ=i, superscript, conjugate, complex transposition, β small, NtxNt and diagonal elements The upper triangular moment drop of the real number. The common matrix transposed matrix Q" of a single matrix is multiplied by the received signal vector y - processed - or, or detailed as: β ^ ' 2Γ. f «- — ❿ Z2 z3 11 r'2 Ί3 0 ^ 0 〇^33 :i : 0 〇 o

VriN, riN, rN, Nt_ XX X2 Xl > v2 w3L' (2) 6 201014235 where ' w = Q% and has the same probability density function as noise to fn (ie w ~ CN (G, coffee) Therefore, the most approximate solution can be expressed as: , fly, _ call 2 =, + _ call 2, formula (7) where II is the Euclidean distance. According to this, the definition-cost function (eGstfunetiGn) ): /=1 Ν· (4) ΖΝ, -Μ~ m=N "i+i ΠΙ. Tree-shaped circle. If the constellation used in the input multi-output system includes iq constellation points, then The graph includes a tree diagram with a depth of W.: A tree root with a squaring level (branches (br_h) to a lower level - a level of the roots of the Qth level) (4) Actions of repeated extensions = reaching the next level Each of the first level (five), ... Λ〇 & &" 叙, 层. Therefore, the number of paths to 益· ',) is 丨q4', and for each of these roads = 7. Article (wind, howl): The branches of the layer represent the possible transmission ~ the order c. υ... The y-th constellation in the two-to-seat diagram reaches every line of the first/level has! Transfer The tuple k, ... XI and represents a neat and ancient goods gossip, ~ + ιί. Therefore, the entire tree diagram has a total of | C | M with a knives path, each meta-vector X And this |cf", stop g彳,: - possible transfer word W happens to be a sub-table transfer character vector x of the |c| kinds of possibilities. For example: the target 7 ^ quantity x of the antenna and make 帛BPSK modulation彡*, —Material Antenna and Three Receiver, where the thick line path:! The tree diagram formed by the multi-output system represents the possible transfer character vector 7 201014235 {x3 =-l,x2 =l,x丨=1 }. IV. Tree search

In the tree diagram, for any path that reaches the first level (-I I, hua), its state, state 胄 is the transfer character group k represented by this path, ..,%,,}. For the root of the tree, the branch (>〇, 丨) 疋 group to the next level of the branch (representing the _ / constellation point in the constellation diagram 2 branch length (BM) is defined as: Knife BAf = (5) where the yth branch from the root of the tree to the next level is separated. For the arrival of the 帛W class (Bu U) #任—Path, its =·/条υ=ο, ..,,|(:Η) S extends to the next level of branching (the branch length of the yth constellation point representing the constellation diagram is defined as:

BM -W -% 2 Equation (6) ^ where -i+1 is related to the state in which the path reaches the first level, < π «+1 7 and the path from the first level to the next level Branch related. A political _ _ path is defined as the sum of the eight knives of the branches on this path. For example, for the thick line path in Figure 1, the length of the long-term ― 77 branches is |z3-~(-l)T, |e2-r23(-l)-r22(l)|2 and 201014235 /=2,·..,^) corresponds to the branch length of the branch that reaches the ί-l level in the tree diagram and extends to the branch of the next level. Obviously, the path of any path with a release branch is the cost of the transfer character vector X represented by this path. Therefore, through the tree search to find the one with the shortest path length in the path with the conditional branch, the most approximate solution gold muscle can be obtained. The V.M algorithm needs to calculate the path length of the icr path according to the above tree search. When the number of transmission antennas and the number of constellation points included in the constellation diagram increase |C|, the tree diagram becomes larger. This in turn increases the branch length of the required calculations, which increases computational complexity. In order to reduce the computational complexity, by the Μ algorithm, in the i-th hierarchy (b1,...,#rl) in the tree diagram, only the best path of the Μ·· strip is retained (ie, the path with the shortest path length) Branches are branched to the next level with their respective extensions |C|, and the remaining paths are deleted to reduce the branch length required for calculation. These reserved paths are called survivor paths. The Μ of each level in the tree diagram can be adjusted according to the corresponding channel conditions and path length (this is called the adaptive Μ algorithm). When the channel condition is good, the slave can be lowered to reduce the computational complexity. degree. VI. Improvement of QRD-M algorithm by QRD-M algorithm, in the Μ, # residual path each extension | C 分 branch to the next level, still need to calculate the branch length of the wind 丨 0 分 branch To preserve the Ml+1 best path from the M|C| path. Higuchi et al., IEEE GLOBECOM 2004, No. 2480-2486, "Adaptive Selection of Surviving Symbol Replica Candidates Based on Maximum Reliability in QRM-MLD for OFCDM ΜΙΜΟ Multiplexing", 201014235, proposes an improved method, which extends each from the residual path of the strip. When branching to the next-level, use multiple quadrant detection (multiple qUadrant detect - to rank the branches of the sun, and calculate the score according to this. Branch branches are long (corresponding to the structure of the q road to retain the lust The best path from +1) can reduce the branch length required for the QRD_M algorithm. The 'multiple quadrant bond test only applies to checkerboard constellations, for example: 16-QAM constellation, therefore, This improved method does not apply to multi-input multi-round systems that modulate Xiao Cai (M>4).

Nagay et al. proposed another improvement method in the IEEE VTC_2〇〇6 touch, ΑρΓ〇ρ〇Μ 〇f ❹ QRM-MLD for Reduced Complexity of MLD to detect ΜΙΜΟ signals in Fading Environment When the surviving path is extended, the quadrant detection is used to cause each path to extend only the |C|« (<|C|) branches to the next level, and calculate the branch length of the branch, from the M, The .|C|a path retains the 路径!+ι best path, which reduces the branch length required for the QRD-M algorithm. However, |C|a and M in this paper are fixed. Constants. If you want to violate the most probable detection performance, you must decide |cu and 〇 according to the worst channel conditions. However, the |C| „ and the stipulations are usually not small. Therefore, the degree of reduction in computational complexity will be very limited. SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide a multi-input multi-output <synthesis method that can be combined with a variety of modulation techniques and that has a low computational complexity. - Thus, the multiple input multiple output detection method of the present invention comprises a tree search for use in a multiple input multiple output system utilizing QR decomposition, and comprising 10 201014235 the following steps: (A) for each of the current classes Path, performing the following sub-step (A1) for the current hierarchy, calculating a de-interfering received signal with the previous layer interference removed according to the corresponding received signal in the processed received signal vector; and (A2) for a constellation diagram Each of the subsets of the first target number divided into: one out of a constellation point in the subset that is closest to the de-interference received signal, and a branch extending from the path representing the constellation point to the next - Level, and calculate the branch length of the branch and - (4) the path length of the path. . Yet another object of the present invention is to provide a MIMO device that can be combined with a variety of modulation techniques and whose computational complexity is low.

Therefore, the multiple input multiple output detector of the present invention is suitable for tree search in a multiple input multiple output system using QR decomposition, and includes a channel estimator, a QR decomposition unit, a conversion unit, and a detection unit. The channel estimator performs channel estimation based on a received signal vector to obtain a channel moment = 13⁄4 QR | solution unit performs channel decomposition on the channel matrix to obtain a - single-matrix and an upper triangular matrix. The conversion unit multiplies the conjugate complex transposed matrix of the single matrix by the received signal vector according to the single matrix to obtain a processed received signal vector d. The detecting unit according to the upper triangular matrix, for the arrival-current hierarchy Each path, the next action: for the current class, according to the received signal corresponding to the phase 201014235 in the processed received signal vector, the received signal is removed after removing the previous layer interference; and for a constellation diagram Dividing each of the subsets of the first target number into a subset: finding a constellation point in the subset that is closest to the de-interference received signal, and extending a branch representing the constellation point from the path to the next The hierarchy calculates the branch length of the branch and the path length of a corresponding path. The above and other technical contents, features, and advantages of the present invention will be apparent from the following detailed description of the preferred embodiments. Referring to FIG. 2, a preferred embodiment of the multi-input multi-output (four) device of the present invention is suitable for performing tree search in a QR-based multiple input multiple output system, and receiving from n transmit antennas through Nr receive antennas 25. The # of the figure (not shown) is used to obtain a received signal vector y. The multi-input multi-output detector of the embodiment comprises a channel estimator 21, a QR decomposition unit 22, a conversion unit 23 and a detection unit 24. The channel_estimator 21 performs channel estimation based on the received signal vector y to obtain a channel matrix Η» QR decomposition unit 22 performs qr decomposition on the channel matrix H' to obtain a single matrix Q and an upper triangular matrix R, and H =QRe conversion list " element 23 according to the single-matrix Q, multiply the complex matrix transposed matrix of a single matrix Q " multiplied by the received signal vector y to get - the processed received signal vector z 'so 'z=Qwy Or detailed as formula (7). Referring to FIG. 3 and FIG. 4, the multi-input multi-output 12 used by the detecting unit 24 • 1 201014235 detecting method includes the following steps: Step 31 疋 For the third layer, the 丨 C| constellation is determined according to the corresponding channel condition. The constellation of points @divided into sub-collections - the first target number A (14 postal), and according to the first target number & split constellation diagram (detailed, field decision method and segmentation method are left to be explained after the whole method is introduced) . Step 32: According to the upper triangle to R, the following substeps are performed for the root of the qth level: Substep 321 is to perform the following substeps for each of the & subsets divided in step 31: Substep 3211 is to find the distance received signal in the sub-set ~ the nearest - the constellation point detailed search method is left to be explained after the whole method is introduced). Sub-step 3212 is a path length extending from the root of the tree - representing the constellation point branching to the first level, and counting the branch length of the branch and the corresponding path of the _ (representing a transmission character group 4: seven}) The mode is as follows: Equation (7) ΡΜ^νν')=Βμ(χ^) 〇' Equation (8) Step 33 is for the 0th hierarchy, according to the corresponding channel condition and path length, it is decided to extend to the basin evaluation in step 32.之一 One of the paths left in the S path of the first level is the second target number from W. < Step 34 is for the strip path extending to the 1st level in step 32, leaving the shortest path from the strip Its _ path. In step (4) and step 3", you can use the appropriate algorithm to determine the number of third targets. 适应 适应 适应 此处 此处 此处 ^ ^ ^ ^ ^ ^ ^ 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 13 .=h Step 42 is to determine the sub-Hua 2 of the constellation circle for the channel condition of the Mth level, according to the corresponding (峨, and according to the number of targets - the number of targets & the method of setting and the way of dividing the entire method After the introduction (detailed step 43 is to de-interfere the received signal W for the following sub-step φ for reaching the first (representative-transfer character with -Γ1 path in each path: word such as one seven, ~2}) The calculation method is as follows: the following level interference ^-<+1 = zn, ~m - wide i+2 Equation (9) Sub-step 432 is for each of the & sub-set commands divided in step hui 42 Step: ❹ Substep 4321 is to find the distance de-interference received signal &+] in the subset: the nearest-constellation point' (detailed search mode (10) after the introduction of the whole financial method). Sub-step 4322 is extended from the path - representative The branch of the constellation points ^ to the 1st class (ie, the lower-level) And calculate the branch length of the branch and the path of the corresponding path (generation * a transfer character group I = ~), the calculation is as follows: tear (% +,) = ^ Guang ^ | 2, formula ( 10) Μ<:~Μ)= ΜΚ:- ί+2)+ΒΜ(χ^+ι) 〇(11) 14 201014235 Step 44 is for the Mth class' based on the corresponding channel conditions and length In (4) 43, the path of the path to be retained in the "H (four) path: the number of targets from ... is like the step 45 is to extend to the first step in step 43, the path of the class, the M, the shortest path is deleted The rest of the path. In step 44 and step, the existing various adaptations can be utilized

The algorithm is used to determine the number of targets - and the shortest path from the strip, which will not be explained here. Step 46: Determine if ζ is equal to % (ie, whether the next-level is a final level if no 'skip to step 47, if yes, skip to step private. Step 47 is to add /1' and jump to step 42. Step 48选取 Selecting a shortest path from the path to the first-level hierarchy, the transmission character group represented by the '(4) path is the _ outgoing transmission character vector x = {^,...,'}. It is only expected that the branches of the & branch branch only need to calculate the branch length of MMW=2,,..,w branches in step 43, therefore, when at least one of ~& is less than 丨C丨'It can reduce the branch length required for the qrd_m algorithm. And it is adjusted according to the channel conditions, and the computational complexity is greatly reduced, while still achieving the performance closest to the most likely detection. Constellation ffi combination In step and step 42+, the constellation map is segmented into & subsets using the Ungerboeck's set partiti〇ning (the first target number & must be a power of 2). : See Figure '5: 15 201014235 16-QAM constellation map can be divided into packages A sub-set A of sixteen constellation points, two sub-sets each including eight constellation points and repelled, and A, four sub-sets Cg~a each including four constellation points and repulsive, each including two The eight sub-sets Dq~D? that are constellation points and repelled, or ten, each sub-rejected with a constellation point (as shown in E^E!5 in Figure 5); see Figure 6, 8_PSK The constellation diagram can be divided into a sub-set f〇 including eight constellation points, two sub-sets each including four constellation points and repulsive, and Gi, four sub-sets Ho~H3 each including two constellation points and repelling each other. , or eight sub-sets IG~1 each including a constellation point and repelled? ^Wengerberg's set partitioning makes the sub-set distance of the sub-sets the largest among all the partitions, which can reduce the error spread ( Error propagation. The recent singularity of the constellation points is in substep 3211 and substep 4321, when 5, <|C| (ie, subset)

The number of constellation points included in the port is greater than 1), and the decision interval of the majority of the constellation points is cut in the subset by using at least one threshold TH, and the receiving 彳§ number and the de-interference receiving signal % ^ (below) The signal ζ is collectively located in which decision interval to find the nearest constellation point of the distance signal 〇〇 see Figure 5, an example is given. When &=4, the 丨6_qam constellation map can be divided into four sub-sets c〇~C3 each including four constellation points, as shown in Table 1: Table 1 Included constellation points Sub-collection

Co 16 201014235 C, ———::______________*·*>--- One C〇' Cs 'ClO ' C15 C2 Cl ' C4 ' C\\ x c14 C3 "一. ^1 Cl ' Cfi ' ' c12 The critical value 711 of each sub-set C〇~C:3 is shown in Table 2: Table 2

Sub-set threshold TH Co - 0.4+H) Cl + 〇.5(nf+X〇.5i/_) c2 C3 —---0.5i/min^ -/(0-5(/^) where 'βηών · = ^ "丨~+丨^咖, is the minimum spacing between constellation points of the received constellation diagram' is the minimum spacing between constellation points of the transmitted constellation. Each threshold is 相对 in the corresponding sub Four decision intervals are cut out in the set c〇~C3, wherein each decision interval covers a constellation point.

The rules are shown in Table 3: Table 3. Subsets z> Re(TH) z> Im(TH) Satisfying constellation points C〇C〇 Satisfy Satisfaction Satisfaction Not satisfied ------ Δ -____ C-ι Satisfy Co does not satisfy not satisfying a 0 C x η C, satisfies satisfying -~-£.13 \ ·. Co satisfies dissatisfaction dissatisfaction __ c5 satisfies --c>〇17 201014235 does not satisfy unsatisfied C15 C2 satisfies C Satisfy not satisfied c4 Not satisfied Satisfied C\\ Not satisfied Not satisfied C14 C3 Satisfied C3 Satisfied Satisfied Not satisfied Satisfied C9 Not satisfied Not satisfied Cl2 Signal z as shown in x in Fig. 5. Since in the sub-set C〇, the signal z satisfies z>Re(TH), it also satisfies z>Im(TH) 'so the constellation point found is c2; since in the sub-set Ci, the signal z does not satisfy z>Re (TH), but satisfying z>Im(TH), so the constellation point found is c10; since in the sub-set C2, the signal z does not satisfy 2>Re(TH) ' but satisfies z>Im(TH), The constellation point found is cn; since the signal z satisfies z>Re(TH) and satisfies z>Im(TH) in the sub-set C3, the constellation point found is C3. Referring to Figure 6, another example is illustrated. When 5^=2, the 8-PSK constellation map can be divided into two sub-sets G〇, G! each including four constellation points, wherein the sub-set G〇 can easily find the distance from a threshold value τη The signal z is the nearest constellation point, while the subset 'G! is rotated 45. After that, it is also easy to find the closest constellation point of the distance signal z with a threshold TH. It is worth noting that this embodiment can also be combined with other QAM modulation techniques, other psk modulation techniques, and other modulation techniques. Coordination, not limited to 16_qam 18 201014235 and 8-PSK. The decision on the number of cover areas assumes that in the process of selecting branches, the correct path (representing the correct transfer character vector) is not deleted. 1 This sister is subtracted from the μth level (ί=1,...,Λ 〇 Down-level extension, the signal: as follows: Z = rN, -M, Nl-MX^li+l+WN "M, Equation (12)

Among them, ‘is the correct transfer character, %★ is the noise, and is a complex Gaussian function. Referring to Figure 7, in the case of noise power ( (shown by the circle in Figure 7) 'If the channel attenuation is more severe (ie, smaller), the constellation points of the received constellation (as in Figure 7) The spacing between the circular points will become smaller 'noise ~, 彳 can cause the signal point ~ the centroid %1 is crossed over several constellation points and offset to the signal z (as shown by the diamond point in Figure 7). ), therefore 'need to use a larger first target number & so that the nearest & constellation points of the distance signal Z can cover the signal point χ; ^_+ι. As shown in Figure 7(a), if the channel attenuation is more serious, the circle covers more constellation points (ie, the first target number & larger), as shown in Figure 7(b), if the channel attenuation is slight. The circle covers fewer constellation points (ie, the first item is lighter than the smaller number). As can be seen from the above, the first target number & and channel conditions (ie channel attenuation and noise power) are related. In the tree search, for the first level (Bu 1, ..., M), the appropriate first target number & should be determined according to the corresponding channel conditions. When extending to the Zth level, it is usually desirable to decide - The appropriate number of first targets &, the probability that the correct branch is deleted p... is less than a 19 201014235 target probability. Obviously the 'number of first targets & the larger j ^ fnissy^i) ^ small, but the higher the computational complexity. Therefore, it is necessary to find the minimum number of first targets & that can satisfy the above conditions to reduce the computational complexity. Equivalent to the probability of event £: the branch length of the correctly transmitted character is not within the shortest branch length of the & Since the calculation process of finding the answer of the household according to the probability of the event is very complicated, it is necessary to calculate the branch length of 丨C!, and the present embodiment uses the following method to obtain an approximate solution to reduce the computational complexity. Referring to Fig. 8, in the constellation diagram, a square of culverts & constellation points is defined centering on the signal z. Event 5 can be approximated as event j: Correct The transfer character falls outside the square. According to Forney's continuous approximation, the side length of a square can be approximated as A = V^_·. According to equation (12) and >ιν,_ί+1 ~CN(〇,σ"), the correct branch is deleted m&(5)) as follows:

PmJS^P(A) «1-φβ(^,+1)<| and |^,+1)<|^ ,

Equation (13) jfi e 2«/m. The user can get a period of 20 201014235 (14): where 2 (one) 疋 0 (.) inverse function, the heart is the maximum number of first targets that can be tolerated. If the heart is "not restricted by software and hardware, $ can be set to the number iq of constellation points included in the constellation diagram, for example, 'for b QAM constellation' can be set to 16.

In order to be able to utilize Wenger's set partitioning, the first target number & must be a power of 2, where the horn is the nearest power but not less than the power of 2. In summary, when at least one of the embodiments is less than 丨c丨, the branch length calculated by the QRD-M algorithm can be reduced, and the embodiment is coordinated with various modulation technologies such as QAM and PSK. Therefore, it is indeed possible to achieve the object of the present invention.

Sr =min^ 4 / ΓPt^et ^ \2"> rr ρ ΐ 4 J / , household max, but the above is only a preferred embodiment of the present invention, and the invention may not be limited thereto. The scope, that is, the simple equivalent changes and modifications made by the invention in the scope of the invention and the scope of the invention are still within the scope of the invention. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a tree diagram illustrating a tree search in a multiple input multiple output system; FIG. 2 is a block diagram illustrating a preferred embodiment of the multiple input multiple output detector of the present invention; FIG. And FIG. 4 is a flow chart illustrating the detection method used in the preferred embodiment; 21 201014235 FIG. 5 is a schematic circle illustrating how the set partitioning of Wengebo is applied to the 16-QAM constellation diagram; FIG. 6 is a schematic diagram. Explain how the Wengebo's set partitioning is applied to the 8-PSK constellation; Figure 7 is a schematic diagram illustrating the relationship between the number of constellation points required and the channel conditions; and Figure 8 is a schematic diagram showing how this embodiment illustrates Decide on the number of constellation points you need. ❹ 22 201014235 [Description of main component symbols] 21 .... ...··Channel estimator 24... Detecting unit 22·························································· ...conversion unit 31~48· ...step

twenty three

Claims (1)

201014235 X. Patent application scope: The multi-input multi-output detection method is suitable for tree search in a multi-input multi-output system using decomposition, and includes the following steps. (4) For each path of the arrival-current hierarchy, Performing the following substep (A1) for the current hierarchy, calculating a deinterfering received signal with the previous layer interference removed according to the corresponding received signal in the processed received signal vector; and (A2) for segmentation by a constellation Each of the subset of the number of target-targets is performed: finding the subset of the constellation that is closest to the de-interference received signal, and extending a branch representing the constellation point from the path to the next Level, and calculate the length of the branch of the branch and the path length of a corresponding path. The multiple input multiple output detection method described in the first item of the patent scope further comprises the following steps: (B) If the next level is not a final level, the shell layer, and jumps to step (A). The first step according to the scope of application of the patent application includes the following steps: / then method (C) for extending to the next level in step (A), retaining a second target number The shortest path and the deletion of the remaining 々 have a path according to the multiple input, and abbreviated as described in item 3 of the patent application scope. ‘More includes the following steps: 』Wei Wei method 24 4, 201014235 (D) For the current class, the second target number is determined according to the corresponding channel condition and path length. 5. According to the multi-input multi-output detection method described in item 4 of the patent scope, in step (C) and step (D), an adaptive Μ algorithm is used to determine the second target number and retain the The shortest path of the second target number. 6. According to the multi-input multi-output detection method described in the third paragraph of the patent (4), the following steps are further included: (8) for the current hierarchy, according to the corresponding channel condition, the number of the target is determined, and according to the The first target number divides the constellation diagram 〇 7. According to the multi-input multi-output detection method described in item 6 of the patent scope, in step (8), the constellation map is segmented into pieces by using Wengbo's set partitioning. A subset of the first target number. 8. In accordance with the multi-input multi-output detection method described in item 6 of the patent scope, in step (8), a probability is obtained by finding the probability that an event is less than:: solution, and according to the The expected number is two such a target number 'where' the event means that the correct transmission character falls outside the square, which encompasses a number of constellation points in the constellation circle. The multi-input multi-round detection method described in item 8 of the range is the power of the first target number which is the closest to the number of hopes. In the multi-input and multi-output detection method described in Patent Publication No. 1, in the sub-step (Α2), when the number of constellation points included in the sub-set is greater than 25 201014235, at least one is utilized. The threshold cuts a majority of the decision intervals covering the constellation points in the subset, and determines in which decision interval the de-interference received signal is located to find the closest to the de-interference received signal in the subset Constellation point. H. - A multi-input multi-output detector for tree search in a multi-input multi-output system using qr decomposition, and comprising: a channel estimator for channel estimation based on a received signal vector' Obtaining a channel matrix; a QR decomposition unit, performing channel decomposition on the channel matrix to obtain a single matrix and an upper triangular matrix; a conversion matrix, according to the single matrix, transposing the conjugate complex of the single matrix The matrix is multiplied by the received signal vector to obtain a processed received signal vector; and a detecting unit performs, according to the upper triangular matrix, for each path reaching a current level: for the current level, according to the Calculating a de-interfering received signal that removes the previous level of interference from the corresponding received 彳s number in the processed received signal vector; and for each of a subset of the first target number segmented by a constellation Performing: finding a constellation point in the subset that is closest to the descrambled received signal, and extending one from the path to represent the star Branching point to the next hierarchy, and calculates the path length of the branch, and a branching path corresponding long. 12. According to the patent, please refer to the multi-input multi-output detector described in item u of the patent, 26 201014235, where the tM test list keeps the shortest number of second targets for all paths extending to the lower-level Path and delete the remaining paths. 13.^ In the patent scope 帛12, the multi-input multi-output detector, the track stop = test: yuan for the current class, according to the corresponding pass condition and path length, determine the second target number. Lu 14: The multi-input multi-output detector described in the scope of patent application 帛13, the detection unit is to use the adaptive Μ algorithm to determine the number of the second directory and the shortest number of the second target path. 15: The multi-input multi-output detector according to claim 11, wherein the detecting unit further determines, according to the current level, the first target number according to the corresponding pass=star, and according to the first target number Knife cut the multi-input and multi-input (four) detectors described in the 15th article of the patent (4), 1 (1), and the test list 70 is to use the set division of Wengebo to make the constellation figure j into the first target. A subset of the number. The range is as described in the multi-input multi-round detector, "the test 70 is obtained by finding a solution whose probability is less than a probability, to obtain a period, the first...:_目" and according to The expected number is obtained by the party Lrr, which means that a correctly transmitted character falls on the -positive H-square. The number of the target-targets is covered in the constellation: 18: ΠΓ range, the multi-input multi-output detection described in item 17 The number of squares of the 2, the number of squares is the closest but not less than the expected number of 27 201014235 19. According to the multi-input and multi-output detector of claim 11 of the patent application, where the detection unit is in the sub-man When the number of constellation points included in the road is greater than 1, 'is to use at least one threshold to cut a majority of the decision intervals covering the constellation points in the subset, and determine which decision interval the de-interference received signal is in, To find the constellation point in the subset that is closest to the de-interference received signal. 28