WO2009115042A1 - 一种联合迭代检测译码方法和装置 - Google Patents

一种联合迭代检测译码方法和装置 Download PDF

Info

Publication number
WO2009115042A1
WO2009115042A1 PCT/CN2009/070861 CN2009070861W WO2009115042A1 WO 2009115042 A1 WO2009115042 A1 WO 2009115042A1 CN 2009070861 W CN2009070861 W CN 2009070861W WO 2009115042 A1 WO2009115042 A1 WO 2009115042A1
Authority
WO
WIPO (PCT)
Prior art keywords
constellation points
detection
transmitting
antenna
detection signal
Prior art date
Application number
PCT/CN2009/070861
Other languages
English (en)
French (fr)
Inventor
曹晏波
李洪强
王鑫
乔元新
王映民
孙建勋
董育新
Original Assignee
大唐移动通信设备有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 大唐移动通信设备有限公司 filed Critical 大唐移动通信设备有限公司
Publication of WO2009115042A1 publication Critical patent/WO2009115042A1/zh

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00Arrangements for detecting or preventing errors in the information received
    • H04L1/004Arrangements for detecting or preventing errors in the information received by using forward error control
    • H04L1/0045Arrangements at the receiver end
    • H04L1/0047Decoding adapted to other signal detection operation
    • H04L1/005Iterative decoding, including iteration between signal detection and decoding operation
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L1/00Arrangements for detecting or preventing errors in the information received
    • H04L1/004Arrangements for detecting or preventing errors in the information received by using forward error control
    • H04L1/0045Arrangements at the receiver end
    • H04L1/0047Decoding adapted to other signal detection operation
    • H04L1/0048Decoding adapted to other signal detection operation in conjunction with detection of multiuser or interfering signals, e.g. iteration between CDMA or MIMO detector and FEC decoder

Definitions

  • the present invention relates to signal detection techniques for multiple input multiple output (MIMO) systems, and more particularly to a joint iterative detection decoding method and apparatus.
  • MIMO multiple input multiple output
  • Detection techniques include linear detection, interference cancellation, lattice reduction aid detection, Monte Carlo statistics, probabilistic data joint detection, ball decoding detection and other optimal and sub-optimal detection methods. Usually detection and decoding at the receiving end are performed independently.
  • the Turbo receiver uses iterative processing techniques to improve system performance by approaching channel capacity by external information transfer between the detector and the decoder.
  • a similar iterative technique i.e., a joint iterative detection decoding algorithm
  • a joint iterative detection decoding algorithm cascades MIMO detection and decoding together, and in the existing algorithm, the optimal detection part mostly uses list ball decoding.
  • the detection algorithm preserves the soft information of the detection result of the list ball decoding algorithm by the MAP bit detection algorithm, and improves the detection performance.
  • Step 11 In the inner MIMO detector, the received signal is detected by using a ball decoding detection algorithm to obtain a candidate set list.
  • the candidate set list includes a plurality of selectable detection signal vectors, and each vector is composed of transmission symbols of each transmit antenna.
  • the ball decoding detection algorithm is used to find a plurality of detection signal vectors that are closer to the real transmission symbol vector, and constitute a candidate set list, and the log likelihood ratio of each bit is calculated by the MAP bit detection algorithm.
  • Step 12 In the inner MIMO detector, using the candidate set list and the bit prior information OP60-090015 Calculates the output of information per bit.
  • the out-of-bit information is the difference between the log-log likelihood ratio and its prior information, and is the new information obtained when calculating the log-likelihood ratio per bit.
  • Step 13 De-interleave the external information output by the inner MIMO detector and input it to the decoder.
  • Step 14 The deinterleaving result of the outer information obtained by using step 13 is input as a priori information of the outer soft-in soft-out decoder, and the decoded outer information result (including information bits and check bits) of all bits is output after decoding. ) and the log likelihood ratio of the information bits.
  • Step 15 If the maximum number of iterations is reached, the log likelihood ratio of the information bits output by the decoder is hard-decided to obtain the final desired information bit result, and the iterative detection decoding is ended. Otherwise, go to step 16.
  • Step 16 The external information outputted by the Turbo decoder is re-interleaved, and input as an a priori information to the detector for iterative detection, and the process proceeds to step 12.
  • the progress of the MAP bit detection algorithm depends on the determination of the candidate set list in step 11, so the computational complexity of the candidate set list is also affected by the performance of the entire joint iterative detection decoding algorithm.
  • the determining manner of the candidate set list in the step includes:
  • Step 11a determining a sphere radius according to a channel estimation result
  • Step lib using a radius constraint to perform upper triangulation preprocessing on the channel matrix
  • Step 11c Searching for a possible transmitted symbol vector, that is, a symbol combination on all transmitting antennas, using a depth-first or breadth-first algorithm in a multi-dimensional hypersphere within a constrained radius with the received signal as the center of the sphere;
  • Step lid if a reasonable solution is not found, increase the radius and search again; if a reasonable solution can be found, save it, and calculate a new radius based on the searched symbol combination.
  • OP60-090015 Search again with the new radius until no better solution can be found;
  • Step lie then the best value found is the optimal maximum likelihood solution.
  • the advantage of the above ball decoding algorithm is that it does not have to search all the lattice points in the entire lattice space, but only needs to search within a predetermined limited spherical area.
  • Commonly used ball decoding algorithms are within a certain range of system parameters, such as the appropriate SNR interval, signal constellation size, number of transmitting and receiving antennas, and their complexity is polynomial, which is similar to the complexity of the linear detection method.
  • system parameters such as the appropriate SNR interval, signal constellation size, number of transmitting and receiving antennas, and their complexity is polynomial, which is similar to the complexity of the linear detection method.
  • the computational complexity of the ball decoding algorithm is uncertain. If the above parameters are not selected properly, the computational complexity of the algorithm is also exponential. Due to the diversity of actual channel types and signal-to-noise ratio conditions, it is difficult to optimize the design. It can be seen that this method cannot control resource allocation during hardware implementation, and often there is insufficient or waste of resources. Summary of the invention
  • the present invention provides a joint iterative detection decoding method and apparatus, which can overcome the shortcomings of the ball decoding detection algorithm whose computational complexity is uncontrollable, and realize the detection result of the soft information with a small cost.
  • a joint iterative detection decoding method comprising:
  • a determining the number of optional constellation points corresponding to each transmitting antenna according to the number of detection signal vectors required for joint iterative detection, the modulation order M of the transmitting antenna, and the number of transmitting and receiving antennas, where c ⁇ 2 M , j is the transmit antenna index, and M is the modulation order of the transmit antenna;
  • step b According to the number of optional constellation points corresponding to each antenna determined in step a, among all the detection signal vectors, select f[detection signal vectors closest to the real transmission signal, OP60-090015 constitutes a candidate set list, where m is the number of transmit antennas, and the detected signal vector is a vector formed by any one of the constellation points of each transmit antenna;
  • the log likelihood ratio of each information bit is calculated according to all the detected signal vectors in the candidate set list.
  • a joint iterative detection decoding device includes: a storage unit, a detection signal vector selection unit, and a detection unit;
  • the storage unit is configured to save the number of optional constellation points corresponding to each transmitting antenna determined according to the number of detection signal vectors required for joint iterative detection, the modulation order M of the transmitting antenna, and the number of transmitting and receiving antennas.
  • ⁇ 2 M , j is the transmit antenna index
  • M is the modulation order of the transmit antenna
  • the detection signal vector selection unit is configured to select a distance true in a detection signal vector formed by a constellation point of each transmitting antenna according to the number of optional constellation points corresponding to each antenna stored in the storage unit. The most recent detection signal vector of the transmitted signal,
  • the detecting unit is configured to calculate a log likelihood ratio of each information bit according to all detection signal vectors in the candidate set list provided by the detection signal vector selecting unit by using a MAP bit detection mode.
  • the embodiment of the present invention first, determine the number of optional constellation points corresponding to each transmitting antenna; then, perform rounding on the current channel matrix to obtain a matrix R and a matrix Q, and use the matrix Q to The received signal y is weighted to obtain a weighted received signal
  • the detection signal vector closest to the real transmission signal is selected to form a candidate set list; finally, the MAP bit detection is performed.
  • Algorithm determining the logarithm of each information bit according to all detected signal vectors in the candidate set list OP60-090015 likelihood ratio.
  • the number of optional constellation points of each transmitting antenna is determined to control the computational complexity, and on the other hand, the detection signal vector that is closer to the real transmitting signal is selected for signal detection, thereby ensuring detection performance.
  • FIG. 1 is a system block diagram of a conventional joint detection decoding.
  • Fig. 2 is a schematic diagram of searching for a detection signal in accordance with a tree structure.
  • FIG. 3 is a structural diagram of a joint detection and decoding apparatus according to an embodiment of the present invention.
  • the basic idea of the embodiment of the present invention is: first determining the number of optional constellation points of each transmitting antenna, and then selecting a detection signal vector that is closer to the real transmission signal among all the constellation points, thereby limiting the number of detection signal vectors. At the same time, to ensure detection performance.
  • the detection method of the embodiment of the present invention is based on Hochwald's log-likelihood ratio calculation method based on the MAP bit detection algorithm. Next, the log likelihood ratio calculation method will be first introduced.
  • X is the transmitted symbol vector on the transmitting antenna
  • j is the received signal vector on the receiving antenna
  • H is the transmission channel matrix, "is a Gaussian white noise vector.
  • ⁇ m> is a data bit vector of ⁇ ⁇ x 1 dimension
  • ⁇ ⁇ is the number of bits contained in each constellation symbol
  • is the number of transmitting antennas
  • is the modulation of the bit vector ⁇ m> signal.
  • L D can be recorded as a priori information L A and external information ⁇ OP60-090015
  • the log likelihood ratio for the bit vector can be expressed as:
  • the result L is the log likelihood ratio L, L of all encoded bits (including information bits and parity bits) output by the outer soft-in soft-out decoder, which is the pair of output information bits. Number likelihood ratio.
  • the uncoded data bits are recorded as ⁇ ⁇ , and + is the same vector set as the length of the interleaver OP60-090015
  • equation (8) Using the Max-log approximation method, the outer information of equation (8) can be given by:
  • the received signal: V can be detected by applying the formula, and the outer information of the bit vector bl is obtained, that is, the operation of step 2 described in the background art.
  • the candidate set of the embodiment of the present invention is specifically The manner in which the detected signal vectors are determined in the list is described later.
  • the BCJR algorithm is adopted as the decoding algorithm of the standard Turbo code of 3GPP.
  • the standard decoding algorithm only the log likelihood ratio calculation of the system information bits and the corresponding external information output are performed, and in the iterative detection and decoding algorithm, the outer information output of all coded bits is required to be interleaved as the a priori information of the detector. Therefore, it is also necessary to calculate the external information of the check bits.
  • the information out-of-information information L E (ul ) and the log-likelihood ratio L D (ul ) are calculated according to equations (13) and (14), respectively, that is, the operation of step 4 described in the background art.
  • AW and AW are forward recursive and backward recursive metrics, respectively, and branch metrics of system information bits and check bits, respectively.
  • L £ 4«E s / N. Is the channel value.
  • the MAX_LOG_MAP algorithm can be used to reduce the complexity of the algorithm.
  • the above theoretically derived formula is applied to the joint iterative detection decoding method described in the background art, that is, joint detection can be performed on the MIMO system.
  • the computational complexity is related to the number of detection signal vectors in the candidate set list, and in the current ball decoding detection, in the candidate set list
  • the determination of the detection signal vector is not deterministic. Therefore, when the hardware implements the joint iterative detection decoding method described above, the resource allocation cannot be controlled, and the phenomenon of insufficient or wasted resources often occurs.
  • the candidate set list when determining the candidate set list, first determining the number of selectable constellation points corresponding to each transmit antenna, and then, in all the detected signal vectors, according to each transmit antenna The number of optional constellation points is selected, and the f[detection signal vectors closest to the real transmission signal are selected to form a candidate set list.
  • the number vector is a vector formed by any one of the constellation points of each transmitting antenna.
  • the total number of detection signal vectors in the candidate set list can be limited by the number of optional constellation points, thereby controlling the complexity of detection; on the other hand, when selecting the detection signal vector constituting the candidate set list, The distance of the real transmitted signal is standard, thus ensuring the detection performance.
  • the implementation of the detection signal vector constituting the candidate set list among all the detection signal vectors may be implemented in various manners, for example, the most direct traversal manner.
  • the detection signal is searched according to the tree structure, and corresponding constellation points are selected for each transmitting antenna to form a candidate set list.
  • Detection signal vector by triangulating the channel matrix, the signal detection can be iterated from the last transmitting antenna, and the selection process of the detection signal vector can be simplified. In this process, the selection condition of the aforementioned detection signal vector is required (ie, the detection signal vector closest to the real transmission signal is selected)
  • the selection conditions on which the detection signal vectors constituting the candidate set list are selected are derived.
  • the constellation points, the selected constellation points constitute a detection signal vector in the candidate set list, and are used to perform the MAP bit detection algorithm.
  • the size of I depends on the matrix R and the signal y', and depends on the constellation point values of all transmit antennas.
  • the embodiment selects a constellation point for each transmitting antenna from the last transmitting antenna, and the selection basis is: before the current transmitting antenna Any combination of constellation points selected by the selected transmit antenna selects the constellation point that minimizes £ ; joins the candidate set
  • the embodiment of the present invention starts from the last transmitting antenna, selects constellation points for each transmitting antenna in turn, and forms at least one tree structure by using constellation points selected for all transmitting antennas, and selects the last transmitting antenna for the last transmitting antenna.
  • the constellation point is the root node of the tree structure, the constellation points selected for other transmitting antennas are sequentially arranged as the sub-nodes of the root node, and the constellation points selected for the same transmitting antenna are located in the same layer of the tree structure; Root of structure
  • the node of OP60-090015 is the last layer of the tree structure, and the layer where the leaf node is located is the first layer of the tree structure; the number of tree structures finally formed is equal to the number of constellation points selected by the last antenna.
  • all constellation points included in the path from each leaf node to the root node serve as a detection signal vector in the candidate set list.
  • a constellation point selected for the last transmit antenna is used as a root node; when a j-th layer is formed by selecting a constellation point for the jth transmit antenna, each of the nodes of the j+1th layer One, select the Cj constellation points that make y'j - ⁇ r j,i x i the smallest as the child nodes of the current node of the j+i layer.
  • the complete tree is: the path from all leaf nodes to the root node of the tree includes m nodes, and m is the number of transmit antennas.
  • the QPSK modulation mode is taken as an example.
  • the number of all constellation points of each transmitting antenna is 4, and the predetermined optional constellation points are: the optional constellation corresponding to the 1st and 2nd transmitting antennas.
  • the number of points is one, and the number of optional constellation points corresponding to the third and fourth transmitting antennas is four, as shown in FIG. 2 .
  • the other nodes of the third layer are also selected as the second method as described above.
  • OP60-090015 The constellation point of the root transmitting antenna; similarly, the constellation point is selected for the first transmitting antenna, and the selected constellation point is used as the child node of the layer 2 node, and finally constitutes 4 complete trees, respectively, tree A , tree tree C and tree D. So far, all the constellation points included in the path from the leaf node to the corresponding root node constitute a detection signal vector in the candidate set list, and the four trees include a total of 16 detection signal vectors. Since the constellation points which minimize y are selected each time during the selection process, the detection signal vector formed by the above method is A set of constellation points that make i smaller, thus ensuring detection
  • the breadth-first search may be performed.
  • the mode when the constellation point is selected for the jth antenna, all the nodes for the j+1 layer are selected, and then selected for the subsequent transmit antenna; or, the depth-first search mode may be performed, that is, When the j-th antenna selects the constellation point, after selecting one node of the j+1 layer, the selection of the subsequent transmit antenna is started, and after the selection of a detection signal vector is completed, the root node or other child nodes are returned. Select the subordinate node for the remaining root node or other child nodes.
  • Step 21 Determine the number of optional constellation points corresponding to each transmitting antenna.
  • the number of selectable constellation points that can be used as the detected signal vector in the candidate set list is determined from all constellation points of the transmit antenna.
  • the number of optional constellation points corresponding to each transmitting antenna may be determined according to the number of detection signal vectors required for joint iteration detection, the modulation order of the transmitting antenna, and the number of transmitting and receiving antennas.
  • the transmit antenna uses QPSK tuning.
  • the modulation order is lower, and a larger one can be selected. If the modulation order is higher and the number of transmitting antennas is larger, if it is large, it will cause The computational complexity is unbearable, so there are not many options that are large.
  • the higher the energy of the transmitting antenna the higher the detection accuracy of the transmitting signal on the receiving end; the lower the energy of the transmitting antenna, the lower the detection accuracy of the transmitting signal on the receiving end, therefore,
  • fewer optional constellation points can be selected, that is, the detection performance can be ensured; for a low-energy transmitting antenna, more optional constellation points need to be selected to ensure detection performance.
  • the energy of each transmitting antenna may be further performed. Specifically, the energy of each transmitting antenna is calculated according to the channel matrix in advance.
  • a larger number of selected points is selected on the antenna with lower energy, and a smaller selection is selected on the antenna with higher energy.
  • a set of detection signal vectors closer to the real transmission signal can be searched, and the external information L £ ( IJ) is calculated based on the signal set.
  • Step 22 Perform upper triangulation on the current channel matrix to obtain a matrix R and a unitary matrix Q, and use the transposed Q T of the unitary matrix Q to weight the current received signal y to obtain a weighted received signal y '
  • Step 23 Calculate each of all constellation points corresponding to the last transmit antenna
  • Step 24 Use the last transmit antenna of the last transmit antenna as the current transmit antenna j.
  • Step 25 For each of the nodes of the j+1th layer in all the tree structures, select the smallest constellation point as the child node of the current node of the j+1th layer, where the value has been determined, and Since the search process starts from the root node, before the step, the j+1th layer to the root node are known, that is, ... ⁇ is known, and for each node of the j+1th layer, the current is transmitted. All constellation points of the antenna; ⁇ selects the constellation points that make the norm y the smallest as the child nodes of the current node of the j+i layer. among them,
  • x j+1 ... ⁇ is all constellation points included in the path from the current node to the root node.
  • Step 26 Determine whether the current transmitting antenna is the first transmitting antenna. If yes, all the constellation points included in the path from the leaf node to the corresponding root node form a detection signal vector, join the candidate set list, and perform step 27 Otherwise, the next transmit antenna of the current transmit antenna is taken as the current transmit antenna j, and the process returns to step 25.
  • the above steps 23 to 26 are processes for determining the candidate set list by means of breadth-first search.
  • the number of detection signal vectors in the candidate set list determined by the above manner can reach a preset level, and the detection performance can be guaranteed.
  • Step 27 Determine a log likelihood ratio of information bits on all transmit antennas by using all detected signal vectors in the candidate set list.
  • the detection signal vector in the candidate set list can also be determined by depth-first search. Specifically, in steps 23 to 26, as shown in FIG. 2, each time the constellation point of the current transmitting antenna is determined, only for the current A node of the jth layer in the tree selects a constellation point until a constellation point of the first transmitting antenna is determined; then, returns to the jth layer, selects a constellation point for the other leaf nodes, and so on, until the construct The complete 4 trees shown in Figure 2.
  • the number of optional constellation points is determined in real time according to each parameter.
  • the number of optional constellation points can be set in advance in the hardware device, and the corresponding constellation points can be directly allocated according to the set of optional constellation points in real time during signal detection. For example, taking four transmitting antennas as an example, the number of optional constellation points of all transmitting antennas may be determined in advance as follows: The number of optional constellation points corresponding to four transmitting antennas with high to low energy is 1, 1, 4, respectively.
  • an optional constellation point combination may be set in the hardware device, and when performing signal detection, determining the energy of each transmitting antenna according to the channel matrix, and following the order of energy from high to low Rearranging the columns, the transmitting antennas, and the received signals in the channel matrix, and using the rearranged channel matrix as the current channel matrix, and sorting the connections OP60-090015 Receive signal as the current received signal y, then perform steps 22 ⁇ 26, and when transmitting steps 23 ⁇ 26, the transmitting antennas are all arranged in the sorted order.
  • step 26 After performing step 26, further, the elements in each detection signal vector in the candidate set list are rearranged according to the order of the transmitting antennas before sorting, and then step 27 is performed, and all the rearranged detection signal vectors are used to determine all The log likelihood ratio of the information bits on the transmit antenna. By adding rearrangement operations, the complexity of the entire process is reduced.
  • the value of Cj ranges from 1 ⁇ ⁇ ⁇ 2 ⁇
  • the embodiment of the invention further provides a joint iterative detection and decoding device, and the specific structure is as shown in the figure
  • the apparatus includes a storage unit, a detection signal vector selection unit, and a detection unit.
  • the storage unit is configured to save the number of optional constellation points corresponding to each transmitting antenna determined according to the number of detection signal vectors required for joint iterative detection, the modulation order M of the transmitting antenna, and the number of transmitting and receiving antennas.
  • ⁇ 2 M , j is the transmit antenna index
  • M is the modulation order of the transmit antenna
  • the detection signal vector selection unit is configured to use the number of selectable constellation points corresponding to each antenna stored in the storage unit, In the detection signal vector composed of one constellation point of each transmitting antenna, selecting the nearest detection signal from the real transmitting signal
  • a vector constituting a candidate set list, where m is a number of transmit antennas;
  • a detecting unit configured to calculate, by using a MAP bit detection manner, a pair of each information bit according to all detected signal vectors in the candidate set list provided by the detection signal vector selecting unit Number likelihood ratio.
  • a specific structure of the detection signal vector selection unit including a weighting unit and a constellation point selection unit.
  • the Q T y ; constellation point selection unit is configured to select constellation points for all the transmitting antennas by using a tree search method according to the number of optional constellation points corresponding to each transmitting antenna in the saving unit and the weighted receiving signals provided by the weighting unit.
  • a detection signal vector in the candidate set list is formed.
  • the memory cell and the detection signal vector selection unit in the above apparatus may be in FIG.
  • the OP60-090015 is implemented in an internal MIMO detector in the system, and the detecting unit can be implemented by using the WAP bit detection method of the system shown in FIG.
  • no radius constraint is required for performing the search of the detected signal vector (a typical ball decoding algorithm requires a radius constraint); an optional corresponding to each transmitting antenna may be adopted.
  • Controlling the number of constellation points to control the number of detection signal vectors, thereby controlling the computational complexity; meanwhile, when selecting the detection signal vector, selecting the detection signal vector that is closer to the real transmission signal, and then using these selected detections The signal vector performs signal detection to ensure detection performance.
  • the search process of each constellation point in the detection signal vector can be calculated in parallel, the algorithm complexity is fixed, and the hardware design is easy.
  • the simplified tree search detection algorithm has low complexity, and the complexity is not sensitive to the increase of the antenna or the number of multi-users; the bit log likelihood ratio soft information can be output in parallel, and the hard decision result can also be output; the detector can be utilized An iterative detection decoding algorithm for external information transfer between decoders improves performance.
  • curve 1 (ZF) is expressed as the simulation performance of the zero-forcing detection method
  • curve 2 (ZFLLR) is the simulation performance of the zero-forcing weighted detection method
  • curve 3 (MMSE) is expressed as the simulation performance of the minimum mean square error detection method.
  • Curve 4 (RTSD (Reduced Tree Searching Detection) hard decision result) and curve 5 (RTSD soft decision result) represent simulation performance of the simplified tree search detection method in the embodiment of the present invention
  • curve 6 UIDD Jointed Iterative Detection & Decoding
  • the simulation performance of the joint iterative detection decoding method of the embodiment of the invention has an iteration number of 1.
  • the BER curve of the simplified tree search detection method is faster than that of the zero-forcing and zero-forcing detection algorithm, and when the BER is close to 10, the simplified tree search detection algorithm has a hard decision result than the ZFLLR detection algorithm. well 9dB above, about 3dB better than the MMSE, and the soft decision results in BER close above 10-4 Shi 1.5dB better than the hard decision result.
  • the joint iterative detection decoding algorithm with one iteration has the best performance, and the soft decision result is about 0.8dB. It can be seen that the performance of the simplified tree search detection algorithm is significantly improved compared with the linear algorithm, and is convenient for hardware design and implementation. It is a very effective detection algorithm, and the performance is obtained after joint iterative detection and decoding algorithm based on simplified tree search detection. Can be further improved.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Radio Transmission System (AREA)

Description

OP60-090015 一种联合迭代检测译码方法和装置
技术领域
本发明涉及多输入多输出 (MIMO ) ***的信号检测技术, 特别涉 及一种联合迭代检测译码方法和装置。 发明背景
检测技术包括线性检测、 干扰消除、 格约简辅助检测、 蒙特卡罗统 计法、 概率数据联合检测、 球译码检测等最优和次最优检测方法。 通常 在接收端检测和译码是独立进行。 Turbo接收机采用迭代处理技术通过 在检测器和译码器间的外信息传递来提高***性能逼近信道容量。
对于多用户或多天线***, 同样可以采用类似迭代技术, 即联合迭 代检测译码算法来逼近信道容量。 具体地, 在多用户或多天线***中, 联合迭代检测译码算法将 MIMO检测和译码级联在一起来处理,在已有 算法中, 最优的检测部分多采用的是列表球译码检测算法, 通过 MAP 比特检测算法保留了列表球译码算法检测结果的软信息, 提高了检测性 能。 以 MIMO多天线***为例 (对多用户***同样适用), 其***框图 如图 1所示, 包含以下几个步骤:
步骤 11 , 在内 MIMO检测器中, 利用球译码检测算法对接收信号 进行检测, 得到候选集列表。
其中, 候选集列表中包括多个可选的检测信号矢量, 每个矢量由各 个发射天线的发送符号构成。 本步骤利用球译码检测算法找到距离真实 的发送符号矢量比较接近的多个检测信号矢量, 构成候选集列表, 以通 过 MAP比特检测算法计算每个比特的对数似然比。
步骤 12, 在内 MIMO检测器中, 利用候选集列表和比特先验信息 OP60-090015 计算输出每比特外信息。
在 Turbo译码中, 比特外信息为该比特对数似然比与其先验信息的 差值, 为计算每比特对数似然比时所得到的新的信息。
步骤 13, 对于内 MIMO检测器输出的外信息进行解交织并输入到 译码器。
步骤 14, 利用步骤 13得到的外信息的解交织结果作为外软入软出 译码器的先验信息输入, 译码后输出所有比特的译码后外信息结果(包 括信息比特和校验比特 ) 以及信息比特的对数似然比结果。
步骤 15, 如果达到最大迭代次数, 对译码器输出的信息比特的对数 似然比结果进行硬判决, 得到最终希望的信息比特结果, 并结束迭代检 测译码, 否则转到步骤 16。
步骤 16, 对于 Turbo译码器输出的外信息重新进行交织, 并作为先 验信息输入到检测器进行迭代检测, 转入步骤 12。
通过上述方式进行的迭代译码算法中, 虽然能够通过 MAP比特检 测算法保留球译码的软信息, 理论上可以提高***性能, 但是由于该
MAP比特检测算法的进行依赖于步骤 11中候选集列表的确定, 因此求 取候选集列表的计算复杂度也会影响整个联合迭代检测译码算法的性 能。 具体地, 该步骤中候选集列表的确定方式包括:
步骤 11a, 根据信道估计结果确定球半径;
步骤 lib, 利用半径约束对信道矩阵进行上三角化预处理;
步骤 11c, 在以接收信号为球心的约束半径内的多维超球内, 利用 深度优先或广度优先算法搜索一个可能的发送符号矢量, 也就是所有发 射天线上的符号组合;
步骤 lid, 如果找不到合理的解, 则增加半径, 重新搜索; 如果能 搜出合理的解, 则进行保存, 并根据搜索出的符号组合计算新的半径, OP60-090015 利用新的半径再进行搜索, 直到找不到更优的解为止;
步骤 lie, 随后搜索到的最佳值即是最优的最大似然解。
上述球译码算法的优点在于它不必搜索整个格空间内的所有格点, 而只需要在一个预先设定的有限球形区域内进行搜索。 常用的球译码算 法在一定的***参数范围内, 如在合适的 SNR区间、信号星座大小、发 射和接收天线数目情况下, 其复杂度为多项式级的, 与线性检测方法的 复杂度相近。 但是由于搜索半径及信道条件的影响, 很难确定在以 r为 半径的球内的格点数, 这也就导致了球译码算法的计算复杂度是不确定 的。 如果以上参数选择的不合适, 该算法的计算复杂度也是指数级的。 由于实际信道类型和信噪比条件的多样性, 难以进行优化设计, 由此可 见该方法在硬件实现时无法控制资源分配, 经常会出现资源不足或浪费 的现象。 发明内容
有鉴于此, 本发明中提供一种联合迭代检测译码方法和装置, 能够 克服球译码检测算法计算复杂度不可控的缺点, 并以较小的代价实现软 信息的检测结果。
为实现上述目的, 本发明采用如下的技术方案:
一种联合迭代检测译码方法, 包括:
a、根据联合迭代检测所需的检测信号矢量的个数、发射天线的调制 阶数 M和收发天线个数, 确定每根发射天线对应的可选星座点个数 , 其中, c≤2M, j为发射天线索引, M为发射天线的调制阶数;
b、根据步骤 a确定的每根天线对应的可选星座点个数 ,在所有检 测信号矢量中, 选择距离真实发射信号最近的 f[ 个检测信号矢量, OP60-090015 构成候选集列表, 其中, m为发射天线数目, 所述检测信号矢量为每根 发射天线的任意一个星座点构成的矢量;
c、 利用 MAP比特检测方式, 根据候选集列表中的所有检测信号矢 量计算每个信息比特的对数似然比。
一种联合迭代检测译码装置, 包括: 存储单元、 检测信号矢量选择 单元和检测单元;
所述存储单元, 用于保存根据联合迭代检测所需的检测信号矢量的 个数、发射天线的调制阶数 M和收发天线个数所确定的每根发射天线对 应的可选星座点个数 , 其中, ≤2M, j为发射天线索引, M为发射天 线的调制阶数;
所述检测信号矢量选择单元, 用于根据所述存储单元中保存的每根 天线对应的可选星座点个数 , 在由每根发射天线的一个星座点构成的 检测信号矢量中, 选择距离真实发射信号最近的] 个检测信号矢量,
7=1
构成候选集列表, 其中, m为发射天线数目;
所述检测单元, 用于利用 MAP比特检测方式, 根据所述检测信号 矢量选择单元提供的候选集列表中的所有检测信号矢量计算每个信息 比特的对数似然比。
由上述技术方案可见, 本发明实施例中, 首先确定每根发射天线对 应的可选星座点个数; 然后, 对当前信道矩阵进行上三角化处理得到矩 阵 R和矩阵 Q, 利用矩阵 Q对当前接收信号 y进行加权得到加权接收信号
QTy ; 接下来, 在所有检测信号矢量中, 选择距离真实发射信号最接近 的 个检测信号矢量, 构成候选集列表; 最后, 通过 MAP比特检测
7=1
算法, 根据候选集列表中的所有检测信号矢量确定每个信息比特的对数 OP60-090015 似然比。 通过上述方式, 一方面确定每根发射天线的可选星座点个数, 以控制计算复杂度, 另一方面, 选择距离真实发射信号较近的检测信号 矢量, 用于信号检测, 从而保证检测性能。 附图简要说明
图 1为现有的联合检测译码的***框图。
图 2为按照树形结构搜索检测信号的示意图。
图 3为本发明实施例提供的联合检测译码装置的结构图。
图 4为本发明实施例方法与其他现有检测方法的性能比较仿真结果 示意图。 实施本发明的方式
为使本发明的目的、 技术手段和优点更加清楚明白, 以下结合附图 对本发明做进一步详细说明。
本发明实施例的基本思想是: 首先确定每根发射天线的可选星座点 个数, 然后, 在所有星座点中选择距离真实发射信号较近的检测信号矢 量, 从而在限制检测信号矢量个数的同时, 保证检测性能。
本发明实施例的检测方法是以 Hochwald的基于 MAP比特检测算法 的对数似然比计算方法为框架进行的。 下面, 首先介绍该对数似然比计 算方法。
考虑一个 MIMO信道, 其数学模型为:
y = Hx + n ( 1 )
其中, X为发射天线上的发送符号矢量, j为接收天线上的接收信 号矢量, H为传输信道矩阵, 《为高斯白噪声矢量。 令待检测信号构成 的比特矢量为 6 , 由式(2 )给出: OP60-090015 xm =map(b<m>),m = l,---,M (2)
其中, 6<m>是一个 Με x 1维的数据比特矢量, Με是每个星座符号所 包含的比特数, Μ为发射天线的个数, ^即为对比特矢量 <m>的调制信 号。
^"于数据比特 M'Me -J , 令 =1表示逻辑 1, bk =_1表示 逻辑 0, 当接收信号为 j时的对数似然比为:
P[bk =+11 J]
LD(bk \y) = \n (3)
P[bk = -11 J]
假设 经过了信道编码及交织, 因此 可以认为是统计独立的,那么 根据贝叶斯定理可以得到:
Figure imgf000008_0001
其中, Xk,+1为比特矢量 6的集合, 且其中 ^ =+1, 即
Figure imgf000008_0002
是序号 '的集合;
Jkfi ={j\ j = 0,-,M-Mc -l,j≠k,bj =l} (6)
Figure imgf000008_0003
将分子分母同时乘以 εχρμ 2. χ^ -1 LA (bk ) 可将式(4)记为:
Figure imgf000008_0004
其中, 6^记为矢量 6去除第 个元素 后的子向量, /^^7为对数似 然比 LA去除第 项以后的值。 因此, LD可以记为先验信息 LA和外信息^ OP60-090015
根据上述推导得到的公式(8), 在图 1所示的结构中, 对于比特矢 量 的对数似然比可以表示为:
Ik]
(9)
1
L 4,W 其中, L£i =LDi -LAi , 为比特矢量 的外信息, 初始时 LAi =0, 即 LEi =LDi , 其为图 1中内 MIMO检测器对接收信号: V的检测结果。
式(9) 中外信息为:
Figure imgf000009_0001
^表示将被传输的编码比特, : V为接收向量。 那么式( 8 )就成为了 由外信道编码得到的后验对数似然比。 因此, 图 1中外软入软出译码器 就可以分为一个先验信息和一个外信息, 可以得到下式:
LD, (b,k \LA ) = LA、b2,k) + ( 11)
Figure imgf000009_0002
在式(11 ) 中, 结果 L 为外软入软出译码器输出的编码后所有比 特(包括信息比特和校验比特)的对数似然比 L , L ,为输出的信息比 特的对数似然比。
其中, 未编码数据比特记为 , Χλ,+为与交织器长度相同的向量集 OP60-090015
Figure imgf000010_0001
采用 Max-log近似方法, 式(8) 的外信息可以由下式给出:
Figure imgf000010_0002
在图 1所示的内 MIMO检测器中,应用本公式即可以对接收信号: V 进行检测, 得到比特向量 bl 的外信息, 即背景技术中描述的步骤 2的 操作。 其中, x =map(P ^+1为候选集列表中 bk = +l 时的所有检测信 号矢量, 为候选集列表中 = -1时的所有检测信号矢量。 具体本发 明实施例的候选集列表中检测信号矢量的确定方式在后续进行描述。
对于译码器的外信息输出,采用 BCJR算法作为 3GPP的标准 Turbo 码的解码算法。 对于标准解码算法只有对***信息比特的对数似然比计 算和相应的外信息输出, 而在迭代的检测译码算法中需要所有编码比特 的外信息输出再交织后作为检测器的先验信息, 所以同样需要计算校验 比特的外信息。对于信息比特外信息 LE (ul )和对数似然比 LD (ul )分别按式 (13)和式(14)计算, 即背景技术中描述的步骤 4的操作。
Figure imgf000010_0003
LD (uk s ) = Lcxk s + L(uk s ) + LE (uk s ) (14)
对于校验位的外信息 L£( )按式(15)计 OP60-090015
Figure imgf000011_0001
其中, 和 AW分别为前向递推和后向递推度量, 和 分别为***信息比特和校验比特的分支度量。 是分量译码器 间传递的先验信息, L£ = 4«Es / N。是信道值。 在实现时可采用 MAX_LOG_MAP算法来减小算法复杂度。
将上述理论推导得到的公式应用于背景技术中描述的联合迭代检 测译码方法中, 即可以对 MIMO***进行联合检测。 但是, 正如背景技 术中描述的, 在该联合迭代检测译码方法中, 计算复杂度与候选集列表 中的检测信号矢量的个数有关, 而在目前的球译码检测中, 候选集列表 中检测信号矢量的确定不具备确定性, 因此, 在硬件实现上述联合迭代 检测译码方法时无法控制资源分配, 经常会出现资源不足或浪费的现 象。
本发明实施例中的联合迭代检测译码方法, 在确定候选集列表时, 首先确定每根发射天线对应的可选星座点个数, 然后, 在所有检测信号 矢量中, 根据每根发射天线对应的可选星座点个数 , 选择距离真实发 射信号最近的 f[ 个检测信号矢量, 构成候选集列表。 这里, 检测信
7=1
号矢量是每根发射天线的任意一个星座点所构成的矢量。 这样, 一方面 能够通过可选星座点的个数来限制候选集列表中检测信号矢量的总数, 从而控制检测的复杂度; 另一方面, 在选择构成候选集列表的检测信号 矢量时, 以与真实发射信号的距离为标准, 从而保证检测性能。
具体地, 在所有检测信号矢量中选择构成候选集列表的检测信号矢 量的实现方式可以有多种, 例如最直接的遍历的方式等。 OP60-090015 本发明中, 为进一步降低计算复杂度, 根据各个星座点的可选星座 点个数, 按照树形结构搜索检测信号, 为每根发射天线选择对应的星座 点, 构成候选集列表中的检测信号矢量。 其中, 通过对信道矩阵的三角 化处理, 能够使信号检测从最后一根发射天线开始进行迭代, 也就可以 简化检测信号矢量的选择过程。 在此过程中, 需要对前述检测信号矢量 的选择条件(即选择距离真实发射信号最近的 个检测信号矢量)
7=1
进行相应修改, 下面首先推导在树形结构搜索方式下, 选择构成候选集 列表的检测信号矢量所依据的选择条件。
对/ ixm维信道矩阵 H进行 QR分解, H可以写成:
H = Q Q' ( 16) 其中, R是一个 mxm维的上三角矩阵,且对角元素为整数, Q ^j nxm 的酉阵, β'为 x(/i-m)的酉矩阵。 由式(1 )可知
Figure imgf000012_0001
( 17)
根据式( 17 )可得
Figure imgf000012_0002
上式可以记为
|/_Rx| =||« ( 18)
其中, / = βΓν , ||n|2 =||n||2 - [Q'Y y2 , 如果 m = w, 则《' = «。 由 R的 上三角矩阵的性质, 可以将式(18)展开为: -∑,rj,ixi , i = 1,2,'·' ,m ( 19)
Figure imgf000012_0003
由公式 (19 ) 可见, 只要选择出的各个发射天线上的星座点 OP60-090015
∑ 点构成的检测信号矢量越接近真实的发 射信号。 因此, 以∑ 越小越好为原则, 选择每根发射天线上
Figure imgf000013_0001
的星座点, 将选择的星座点构成候选集列表中的检测信号矢量, 用于进 行 MAP比特检测算法。 另外, 由于 R是上三角矩阵, 因此当 j = m时, £ 的大小
Figure imgf000013_0002
除与矩阵 R和信号 y'有关外,仅取决于最后一根发射天线 m的星座点取 值, 而与其他发射天线的星座点取值无关; 当 j = m-l时, 的大小除与矩阵 R和信号 y'有关外,仅取决于最后一根发射天线 m和倒 数第二根发射天线 m - 1的星座点取值, 而与其他发射天线的星座点取 值无关; ...; 当 j = l时, I: 的大小除与矩阵 R和信号 y'有关 外, 取决于所有发射天线的星座点取值。 根据上述规律, 本发明实施例 在为发射天线选择作为检测信号矢量元素的星座点时, 从最后一根发射 天线开始, 依次为每根发射天线选择星座点, 选择依据为: 针对当前发 射天线之前已经选择过的发射天线所选择星座点的任意一种组合 选择出使得 £ 最小的 个星座点 ;加入候选集列
Figure imgf000013_0003
表中, 构成检测信号矢量。
具体实现时, 本发明实施例从最后一根发射天线开始, 依次为每根 发射天线选择星座点, 并且利用为所有发射天线选择的星座点构成至少 一个树形结构, 为最后一根发射天线选择的星座点为树形结构的根节 点, 为其他发射天线选择的星座点顺序排列为根节点的各级子节点, 且 为同一根发射天线选择的星座点位于树形结构的同一层; 树形结构的根 OP60-090015 节点为该树形结构的最后一层, 叶子节点所在层为该树形结构的第一 层; 最终形成的树形结构的数量与最后一根天线选择的星座点个数相 等。 在形成的所有树形结构中, 从每个叶子节点到根节点的路径所包括 的所有星座点作为候选集列表中的一个检测信号矢量。
具体地, 将为最后一根发射天线选择的 个星座点作为根节点; 在 为第 j根发射天线选择星座点构成树形结构的第 j层时,针对第 j+1层所 有节点中的每一个, 选择使 y'j -∑rj,i xi 最小的 Cj个星座点作为第 j+i层 当前节点的子节点。 直到为所有发射天线选择完星座点形成完整的树, 该完整的树即: 树的所有叶子节点到根节点的路径均包括 m个节点, m 为发射天线数目。
下面以 QPSK调制方式为例进行说明, 其中, 每根发射天线的所有 星座点个数均为 4, 预先确定的可选星座点分别为: 第 1根和第 2根发 射天线对应的可选星座点个数均为 1个, 第 3根和第 4根发射天线对应 的可选星座点个数均为 4个, 如图 2所示。
首先, 从第 4根发射天线开始, 选择 4个星座点作为根节点, 如图 2中的根层对应的节点 1、 节点 2、 节点 3和节点 4; 然后, 为第 3根发 射天线选择星座点时, 针对第 4层(即根层) 的节点 1 , 为第 3根发射 天线选择 4个星座点构成树形结构的第 3层, 并将这 4个星座点 (图 2 中的节点 5、 节点 6、 节点 7和节点 8 )作为节点 1的子节点, 针对第 4 层的其他节点也按照上述方式选择第 3根发射天线的星座点; 接下来, 为第 2根发射天线选择星座点时, 针对第 3层的节点 5 , 为第 2根发射 天线选择使 £ - ί 2 Λ 2的取值最小的 1个星座点 (图 2中的节点 9 )
j=2 1=2 '
作为节点 5的子节点, 针对第 3层的其他节点也按照上述方式选择第 2 OP60-090015 根发射天线的星座点; 同理, 为第 1根发射天线选择星座点, 并将选择 的星座点作为第 2层节点的子节点, 最后构成 4个完整的树, 分别为树 A、 树 树 C和树 D。 至此由叶子节点到相应根节点的任意一条路径 所包括的所有星座点构成候选集列表中的一个检测信号矢量, 则 4个树 中共包括 16 个检测信号矢量。 由于在选择过程中, 每次均选择使 y 最小的 个星座点, 因此由上述方式形成的检测信号矢量是
Figure imgf000015_0001
使得 i 较小的一组星座点构成的集合, 从而能够保证检测性
•6匕
接下来, 以通过树形结构搜索并选择检测信号矢量为例, 详细描述 本发明实施例的具体实施方式, 其中, 在按照上述树形结构进行发射天 线的星座点选择时, 可以按照广度优先搜索方式进行, 即在为第 j根天 线选择星座点时, 针对 j + 1层的所有节点均选择完毕, 然后再为后续的 发射天线选择; 或者, 也可以按照深度优先搜索方式进行, 即在为第 j 根天线选择星座点时, 针对 j + 1层的一个节点选择完毕后, 即开始为后 续的发射天线选择, 待完成一个检测信号矢量的选择后, 再返回到根节 点或其他子节点, 为剩余的根节点或其他子节点选择其下级节点。
下面的实施例以广度优先搜索方式为例进行说明。
步骤 21 , 确定每根发射天线对应的可选星座点个数。
本步骤中需要针对每根发射天线, 从该发射天线的所有星座点中确 定可以作为候选集列表中检测信号矢量的可选星座点的个数。
最基本地, 可以根据联合迭代检测所需的检测信号矢量的个数、 发 射天线的调制阶数和收发天线个数, 确定每根发射天线对应的可选星座 点个数。 例如, 假定 4发 4收的 MIMO***中, 发射天线采用 QPSK调 OP60-090015 制方式, 联合迭代检测所需的检测信号矢量为 16 个, 则发射天线的调 制阶数为 2, 发射天线对应的所有星座点为 4个, 那么 4根发射天线可 以分别对应选择 1、 1、 4、 4个星座点, 从而使得这些星座点构成的检 测信号矢量可以有 1x1x4x4 = 16个。 出于计算复杂度的考虑, 如果发射 天线个数较小调制阶数较低, 可选择较大的 , 而如果调制阶数较高, 发射天线个数较大, 则如果 都很大则会导致计算复杂度不可忍受, 因 此不能很多的 选择的很大。
通常来说, 发射天线的能量越高, 接收端对于该天线上发送信号的 检测准确性越高; 发射天线的能量越低, 接收端对于该天线上发送信号 的检测准确性越低, 因此, 对于能量高的发射天线, 可以选择较少的可 选星座点数, 即可以保证检测性能; 对于能量低的发射天线, 需要选择 较多的可选星座点数, 从而确保检测性能。 基于上述考虑, 为进一步保 证检测性能, 除依据前述因素确定每根发射天线对应的可选星座点个数 外, 还可以进一步根据每根发射天线的能量进行。 具体地, 预先根据信 道矩阵计算每根发射天线的能量, 在确定可选星座点个数时, 在能量低 的天线上选择较大的选点个数, 能量高的天线上选择较小的选点个数, 从而可以搜索到一组距离真实发射信号较近的检测信号矢量集合, 并据 此信号集合来计算外信息 L£ ( I J)。
步骤 22 ,对当前信道矩阵进行上三角化处理得到矩阵 R和酉矩阵 Q , 利用酉矩阵 Q的转置 QT对当前接收信号 y进行加权得到加权接收信号 y '
= QTy。
步骤 23 , 对应最后一根发射天线的所有星座点中的每一个计算
I -^ Λ|2 , 在得到的所有 I - Λ|2中选择最小的 Cm个所对应的星座 点, 将每个星座点作为一个树的根节点。 OP60-090015 步骤 24,将最后一根发射天线的上一根发射天线作为当前发射天线 j。
步骤 25, 针对所有树形结构中的第 j+1层所有节点中的每一个, 选 择使 最小的 个星座点作为第 j+1层当前节点的子节点, 此处 是已经确定的值, 同时由于搜索过程是从根节点开始的, 因 此在本步骤前,第 j+1层到根节点均已知, 即; … ^已知,而针对第 j+1 层的每个节点, 在当前发射天线的所有星座点; ^中选出使得范数 y 最小的 个星座点作为第 j+i层当前节点的子节点。 其中,
Figure imgf000017_0001
xj+1…^为当前节点到根节点的路径中包括的所有星座点。
由于本实施例中是采用广度优先搜索的方式, 因此本步骤会针对所 有树形结构第 j层的每个节点, 分别选择出 个星座点, 如图 2所示, 若 j = 2, 则为 16个第 3层节点分别选择一个星座点, 从而实现为当前 发射天线进行的星座点选择。
步骤 26, 判断当前发射天线是否为第一根发射天线, 若是, 则将由 叶子节点到相应根节点的任意一条路径所包括的所有星座点构成一个 检测信号矢量, 加入候选集列表, 并执行步骤 27; 否则, 将当前发射天 线的下一个发射天线作为当前发射天线 j, 返回步骤 25。
上述步骤 23到步骤 26即为通过广度优先搜索的方式确定候选集列 表的过程, 通过上述方式确定的候选集列表中检测信号矢量的个数可以 达到预先设定的水平, 并且能够保证检测性能。
步骤 27 ,利用候选集列表中的所有检测信号矢量确定所有发射天线 上信息比特的对数似然比。
具体本步骤的操作可以通过背景技术中描述的步骤 12 ~ 16 的操作 OP60-090015 完成, 这里就不再赘述。
至此, 本发明实施例的联合迭代译码检测方法流程结束。
事实上, 还可以通过深度优先搜索的方式确定候选集列表中的检测 信号矢量, 具体在步骤 23到 26中, 如图 2所示, 在每次确定当前发射 天线的星座点时,仅针对当前树中的第 j层的一个节点选择 个星座点, 直到确定完第一根发射天线的 个星座点; 然后, 再返回第 j层, 针对 其他叶子节点选择 个星座点, 依次类推, 直到构造出图 2所示的完整 的 4个树。
在上述实施例的步骤 21 中确定可选星座点个数时, 是根据各项参 数实时地确定可选星座点个数的。 事实上, 为简化处理, 也可以预先根 在信号检测时根据预先确定的可选星座点个数组合, 为每根发射天线分 配对应的可选星座点个数。 这样, 可以预先将可选星座点个数组合设置 在硬件设备中, 在信号检测时实时根据设置的可选星座点组合, 直接分 配对应的可选星座点即可。 例如以四根发射天线为例, 可以预先确定所 有发射天线的可选星座点个数组合为: 能量由高到低的四根发射天线对 应的可选星座点个数分别为 1、 1、 4、 4; 然后, 在信号检测时, 确定每 根发射天线的能量, 再按照该能量, 为相应的发射天线分配对应的可选 星座点个数, 即为最高和次高能量的发射天线分配的可选星座点个数为 1 , 为次低和最低能量的发射天线分配的可选星座点个数为 4。
更具体的在实施上述方式时, 优选地, 可以在硬件设备中设置可选 星座点组合, 在进行信号检测时, 根据信道矩阵确定每根发射天线的能 量, 并按照能量由高到低的顺序对信道矩阵中的各列、 发射天线和接收 信号进行重排, 将重排后的信道矩阵作为当前信道矩阵, 将排序后的接 OP60-090015 收信号作为当前接收信号 y ,然后执行步骤 22 ~ 26,并且在执行步骤 23 ~ 26 时发射天线均为按照排序后的顺序排列的发射天线。 在执行完步骤 26后, 进一步地, 对候选集列表中的每个检测信号矢量中的元素按照排 序前的发射天线顺序进行重排, 然后执行步骤 27, 利用重排后的检测信 号矢量确定所有发射天线上信息比特的对数似然比。 通过加入的重排操 作, 使得整个处理过程中的复杂度降低。
在上述实施例中为每根发射天线选择 个星座点时, Cj的取值范围 为 1≤^≤2Μ, M为发射天线的调制阶数。 如果 = 2Μ ,即第 + i层的每个节点下所对应的第 层的所有星座点 都被选择出来, 从而不需要计算( 19 ) 式, 进而大大减小了计算量。
如果 c = ι , 即第 + i层的每个节点下只选择一个星座点作为该 + i 层节点下的第 层节点, 那么将公式 y) 展开可得:
Figure imgf000019_0001
部、 虚部分别最近的星座点。 因此, 如果如果^ = 1 , 只需计 r,. ,. ,然后再分实部和虚部分别与当前发射天线所有星座符 号的实部和虚部相比较, 实部和虚部分别最近的点即为使得 y 最小的星座点。 OP60-090015 由上述可见, 当 . = 1或 = 2M时计算量大大减小, 因此本发明实施 例建议对于 的选取, 在同样的检测信号矢量个数的前提下, 尽可能取 为 2^或 1为宜, 以减少计算量, 此即为简化的树搜索检测方法。
本发明实施例还提供了一种联合迭代检测译码装置, 具体结构如图
3 所示。 该装置包括存储单元、 检测信号矢量选择单元和检测单元。 具 体地, 存储单元, 用于保存根据联合迭代检测所需的检测信号矢量的个 数、发射天线的调制阶数 M和收发天线个数所确定的每根发射天线对应 的可选星座点个数 , 其中, ≤2M, j为发射天线索引, M为发射天线 的调制阶数; 检测信号矢量选择单元, 用于根据所述存储单元中保存的 每根天线对应的可选星座点个数 , 在由每根发射天线的一个星座点构 成的检测信号矢量中, 选择距离真实发射信号最近的] 个检测信号
7=1
矢量, 构成候选集列表, 其中, m为发射天线数目; 检测单元, 用于利 用 MAP比特检测方式, 根据检测信号矢量选择单元提供的候选集列表 中的所有检测信号矢量计算每个信息比特的对数似然比。
其中, 以树形结构搜索选择检测信号矢量为例, 给出了检测信号矢 量选择单元的一种具体结构, 包括加权单元和星座点选择单元。具体地, 加权单元,用于对当前信道矩阵进行上三角化处理得到矩阵 R和酉矩阵 Q,利用酉矩阵 Q的共轭转置 QT对当前接收信号 y进行加权得到加权接 收信号 y' = QTy ; 星座点选择单元, 用于根据保存单元中每根发射天线 对应的可选星座点个数 和加权单元提供的加权接收信号, 通过树搜索 方式逐级为所有发射天线选择星座点, 形成候选集列表中的检测信号矢 量。
具体上述装置中的存储单元和检测信号矢量选择单元可以在图 1所 OP60-090015 示***中的内 MIMO检测器中实现,检测单元则可以利用图 1所示*** 的 WAP比特检测方法来实现。
按照上述本发明实施例的联合迭代检测译码方法和装置, 在进行检 测信号矢量的搜索时无需半径约束(典型的球译码算法需要半径约束); 可以通过对每根发射天线对应的可选星座点个数的控制, 来实现检测信 号矢量数目的控制, 从而控制计算复杂度; 同时, 在选择检测信号矢量 时, 选择距离真实发射信号较近的检测信号矢量, 再利用这些选择出的 检测信号矢量进行信号检测, 从而保证检测性能。
进一步地, 在应用上述实施例中给出的基于树形结构搜索选择检测 信号矢量的方法和装置中, 该检测信号矢量中各个星座点的搜索过程可 以并行计算, 算法复杂度固定, 易于硬件设计及实现; 简化树搜索检测 算法复杂度低, 且复杂度对天线或多用户数的增加不敏感; 可并行输出 比特对数似然比软信息, 也可输出硬判决结果; 可利用检测器和译码器 间的外信息传递的迭代检测译码算法提高性能。
下面给出应用本发明实施例方法的仿真结果。 考虑 4x4的 MIMO 系 统,采用包括 256个检测信号矢量的候选集,选点组合取为 C = [l;l;16;16] , ***仿真条件如表 1所示, 仿真结果如图 4所示。
表 1 仿真条件设置
Figure imgf000021_0001
OP60-090015
Figure imgf000022_0001
图 3中曲线 1( ZF )表示为迫零检测方法的仿真性能,曲线 2( ZFLLR ) 表示迫零加权检测方法的仿真性能, 曲线 3 ( MMSE )表示为最小均方 误差检测方法的仿真性能, 曲线 4 ( RTSD ( Reduced Tree Searching Detection )硬判决结果)和曲线 5 ( RTSD软判决结果)表示本发明实施 例中简化树搜索检测方法的仿真性能, 曲线 6 UIDD, Jointed Iterative Detection & Decoding )表示本发明实施例的联合迭代检测译码方法的仿 真性能, 迭代次数为 1。
由仿真结果可以看出, 相对于迫零及迫零加权检测算法, 简化树搜 索检测方法的 BER曲线下降的更快, 且当 BER接近 10 时, 简化树搜 索检测算法硬判决结果比 ZFLLR检测算法好 9dB以上, 比 MMSE好约 3dB, 而软判决结果比硬判决结果在 BER接近 10- 4时要好 1.5dB以上。 采用一次迭代的联合迭代检测译码算法的性能最佳, 较软判决结果好约 0.8dB。 可见, 简化树搜索检测算法的性能较线性算法有了明显的提升, 且便于硬件设计及实现, 是一种非常有效的检测算法, 在采用基于简化 树搜索检测的联合迭代检测译码算法后性能可以进一步提高。
以上仅为本发明的较佳实施例而已, 并非用于限定本发明的保护范 围。 凡在本发明的精神和原则之内, 所作的任何修改、 等同替换、 改进 等, 均应包含在本发明的保护范围之内。

Claims

OP60-090015 权利要求书
1、 一种联合迭代检测译码方法, 其特征在于, 该方法包括:
a、根据联合迭代检测所需的检测信号矢量的个数、发射天线的调制 阶数 M和收发天线个数, 确定每根发射天线对应的可选星座点个数 , 其中, c≤2M, j为发射天线索引, M为发射天线的调制阶数;
b、根据步骤 a确定的每根天线对应的可选星座点个数 ,在所有检 测信号矢量中, 选择距离真实发射信号最近的 f[ 个检测信号矢量,
7=1
构成候选集列表, 其中, m为发射天线数目, 所述检测信号矢量为每根 发射天线的任意一个星座点构成的矢量;
c、 利用 MAP比特检测方式, 根据候选集列表中的所有检测信号矢 量计算每个信息比特的对数似然比。
2、 根据权利要求 1所述的方法, 其特征在于, 所述步骤 b包括: bl、 对当前信道矩阵进行上三角化处理得到矩阵 R和酉矩阵 Q, 利 用酉矩阵 Q的共轭转置 QT对当前接收信号 y进行加权得到加权接收信 号 y' = QTy ;
b2、 根据每根发射天线对应的可选星座点个数和所述加权接收信 号, 通过树搜索方式逐级为所有发射天线选择星座点, 形成候选集列表 中的检测信号矢量。
3、 根据权利要求 2所述的方法, 其特征在于, 为除最后一根发射 天线外的其他当前发射天线选择星座点的方式为: 对应每种 { w, . . . , m } 的组合, 选择使 最小的 个星座点, 其中, j 为当前发射天
Figure imgf000023_0001
线的编号, m为发射天线数, 为加权接收信号中当前发射天线对应的 OP60-090015 分量, 为矩阵 R的第 j行第 1列元素, + ,..., 分别是第 j+1根发射 天线到第 m根发射天线选择出的星座点, ·是当前发射天线所有星座点 中的任意一个, 为步骤 a中确定的第 j根发射天线对应的可选星座点 个数。
4、 根据权利要求 2所述的方法, 其特征在于, 为最后一根发射天 线选择星座点的方式为: 对应最后一根发射天线的所有 2M个星座点中 的任意一个 m, 计算 I - Γμ,λ|2的取值, 在得到的所有 I - Γμ,λ|2的取值 中选择最小的 Cm个所对应的星座点作为最后一根发射天线所选择的星 座点。
5、根据权利要求 2所述的方法, 其特征在于, 所述步骤 c中按照深 度或广度优先搜索为所有发射天线选择星座点。
6、 根据权利要求 1 所述的方法, 其特征在于, 进一步根据由信道 矩阵所确定的每根发射天线的能量确定每根发射天线对应的可选星座 点个数 。
7、 根据权利要求 6 所述的方法, 其特征在于, 能量越低的发射天 线对应的可选星座点个数越大, 能量越高的发射天线对应的可选星座点 个数越小。
8、 根据权利要求 1 所述的方法, 其特征在于, 所述确定每根发射 天线对应的可选星座点个数包括:
预先根据联合迭代检测所需的检测信号矢量的个数、 发送天线的调 制阶数和收发天线个数, 确定所有发射天线的可选星座点个数组合;
在进行信号检测时, 根据预先确定的可选星座点个数组合, 为每根 发射天线任意分配对应的可选星座点个数。
9、 根据权利要求 1 所述的方法, 其特征在于, 所述确定每根发射 OP60-090015 天线对应的可选星座点个数包括:
预先根据联合迭代检测所需的检测信号矢量的个数、 发送天线的调 制阶数和收发天线个数, 确定所有发射天线的可选星座点个数组合;
在进行信号检测时, 根据预先确定的可选星座点个数组合, 按照由 信道矩阵确定的每根发射天线的能量, 为每根发射天线分配对应的可选 星座点个数, 能量越低的发射天线对应的可选星座点个数越大, 能量越 高的发射天线对应的可选星座点个数越小。
10、 根据权利要求 9所述的方法, 其特征在于,
在步骤 a前进一步包括: 根据信道矩阵确定每根发射天线的能量, 并按照能量由高到低的顺序对信道矩阵中的各列、 发射天线和接收信号 进行重排, 将重排后的信道矩阵作为当前信道矩阵, 将排序后的接收信 号作为当前接收信号 y;
步骤 c中的发射天线为按照排序后的顺序排列的发射天线;
在步骤 c和 d之间进一步包括: 对候选集列表中的每个检测信号矢 量中的元素按照所述排序前的发射天线顺序进行重排;
步骤 d中的检测信号矢量为重排后的检测信号矢量。
11、 根据权利要求 1所述的方法, 其特征在于, 步骤 a中确定的每 根发射天线对应的可选星座点个数 为 1或 2M
12、 根据权利要求 3或 4所述的方法, 其特征在于, 若第 j根发射 天线 j对应的可选星座点个数 为 1 , 则选择使 y) -∑, rj,i xi 最小的 ;个 星座点包括: 计- 并将计算结果的实部和虚部分别与
Figure imgf000025_0001
所述第 j 根发射天线的每个星座点的实部和虚部进行比较, 选择距离 OP60-090015 、
y) ~ 最近的星座点;
Figure imgf000026_0001
若第 j 根发射天线对应的可选星座点个数 为 2M , 则选择使 最小的 个星座点为: 选择所述第 j根发射天线的所有星座
Figure imgf000026_0002
*。
13、 一种联合迭代检测译码装置, 其特征在于, 该装置包括: 存储 单元、 检测信号矢量选择单元和检测单元;
所述存储单元, 用于保存根据联合迭代检测所需的检测信号矢量的 个数、发射天线的调制阶数 M和收发天线个数所确定的每根发射天线对 应的可选星座点个数 , 其中, ≤2M, j为发射天线索引, M为发射天 线的调制阶数;
所述检测信号矢量选择单元, 用于根据所述存储单元中保存的每根 天线对应的可选星座点个数 , 在由每根发射天线的一个星座点构成的 检测信号矢量中, 选择距离真实发射信号最近的] 个检测信号矢量,
7=1
构成候选集列表, 其中, m为发射天线数目;
所述检测单元, 用于利用 MAP比特检测方式, 根据所述检测信号 矢量选择单元提供的候选集列表中的所有检测信号矢量计算每个信息 比特的对数似然比。
14、 根据权利要求 13 所述的装置, 所述检测信号矢量选择单元包 括加权单元和星座点选择单元;
所述加权单元, 用于对当前信道矩阵进行上三角化处理得到矩阵 R 和酉矩阵 Q, 利用酉矩阵 Q的共轭转置 QT对当前接收信号 y进行加权 得到加权接收信号 y' = QTy ; OP60-090015 所述星座点选择单元, 用于根据所述保存单元中每根发射天线对应 的可选星座点个数 和所述加权单元提供的加权接收信号, 通过树搜索 方式逐级为所有发射天线选择星座点, 形成候选集列表中的检测信号矢 量。
PCT/CN2009/070861 2008-03-18 2009-03-18 一种联合迭代检测译码方法和装置 WO2009115042A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN2008101021584A CN101541023B (zh) 2008-03-18 2008-03-18 一种联合迭代检测译码方法和装置
CN200810102158.4 2008-03-18

Publications (1)

Publication Number Publication Date
WO2009115042A1 true WO2009115042A1 (zh) 2009-09-24

Family

ID=41090508

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2009/070861 WO2009115042A1 (zh) 2008-03-18 2009-03-18 一种联合迭代检测译码方法和装置

Country Status (2)

Country Link
CN (1) CN101541023B (zh)
WO (1) WO2009115042A1 (zh)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101834827B (zh) * 2010-03-29 2012-07-18 大唐联诚信息***技术有限公司 一种多输入多输出***中的信号检测方法和装置
CN102594467B (zh) * 2011-08-15 2014-07-02 上海交通大学 无线多输入多输出***的接收机检测方法
CN104796239B (zh) * 2015-01-30 2019-03-19 苏州恩巨网络有限公司 一种mimo无线通信***及信号检测装置和方法
CN108155972A (zh) * 2017-12-26 2018-06-12 厦门大学 分布式联合信源信道编码***的译码优化方法
CN108494709B (zh) * 2018-03-29 2020-06-16 北京理工大学 一种降低Turbo-MPA接收机复杂度的方法
WO2022094778A1 (zh) * 2020-11-04 2022-05-12 华为技术有限公司 多输入多输出mimo***的检测方法和装置
CN113746776B (zh) * 2021-09-03 2022-12-02 电子科技大学 基于星座点排序和动态树搜索的信号接收方法

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1477793A (zh) * 2002-06-24 2004-02-25 ������������ʽ���� 具有加速球译码的多输入多输出电信***
CN1855795A (zh) * 2005-04-28 2006-11-01 松下电器产业株式会社 多天线传输中的迭代检测方法及其接收终端
CN101059544A (zh) * 2006-04-18 2007-10-24 大唐移动通信设备有限公司 多输入多输出***中实现迭代检测的方法及多天线检测器

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1477793A (zh) * 2002-06-24 2004-02-25 ������������ʽ���� 具有加速球译码的多输入多输出电信***
CN1855795A (zh) * 2005-04-28 2006-11-01 松下电器产业株式会社 多天线传输中的迭代检测方法及其接收终端
CN101059544A (zh) * 2006-04-18 2007-10-24 大唐移动通信设备有限公司 多输入多输出***中实现迭代检测的方法及多天线检测器

Also Published As

Publication number Publication date
CN101541023A (zh) 2009-09-23
CN101541023B (zh) 2012-03-14

Similar Documents

Publication Publication Date Title
RU2303330C1 (ru) Способ приема сигнала в системе связи с несколькими каналами передачи и приема
CN104618061B (zh) 一种大规模多天线***中多用户信号的检测方法
WO2009115042A1 (zh) 一种联合迭代检测译码方法和装置
US8982986B2 (en) Method and device for coded modulation
US20060023636A1 (en) Detector and method for estimating data probability in a multi-channel receiver
Jaldén et al. Parallel implementation of a soft output sphere decoder
JP2006238426A (ja) 逐次モンテカルロによる、複雑さが低減された多入力多出力(mimo)チャネル検出
JP2009527174A (ja) Mimo受信機を提供する装置、方法およびコンピュータ・プログラム
US8045604B2 (en) Estimation of log-likelihood using constrained markov-chain monte carlo simulation
WO2008012807A2 (en) Multi input multi output wireless communication reception method and apparatus
JP2004266838A (ja) 逐次モンテカルロによる準最適多入力多出力(mimo)チャネル検出
CN100589469C (zh) 空时比特交织调制***的迭代译码方法及siso解调器
JP2009207117A (ja) Mimo受信機において受信信号に対する軟出力を生成する方法及び送信機と受信機とを含む送受信機
CN100442062C (zh) 多输入多输出***中实现迭代检测的方法及多天线检测器
US20080225976A1 (en) Method for multiple-input-multiple-output system demodulation
CN105610484B (zh) 大规模mimo低复杂度迭代接收方法
JP6180333B2 (ja) 無線周波数受信機において信号を復号化する方法
EP1609265B1 (en) Signal processing apparatus and method
Ju et al. VLSI implementation of an 855 Mbps high performance soft-output K-Best MIMO detector
US8432989B2 (en) Max-log stack decoder
KR101731723B1 (ko) 다중 안테나 시스템에서 연판정 검출 방법 및 장치
Bhagawat et al. Dynamically reconfigurable soft output MIMO detector
CN108512580A (zh) 适用于低精度量化的大规模多用户mimo迭代检测方法
Nekuii et al. Efficient soft demodulation of MIMO QPSK via semidefinite relaxation
CN101582743B (zh) 一种用于迭代接收机的mimo检测方法及***

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 09723330

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 09723330

Country of ref document: EP

Kind code of ref document: A1