JP2014165646A - Signal separation method in mimo system and receiving device - Google Patents

Abstract

A signal separation method for improving decoding characteristics is provided.
In order to reduce the amount of calculation of the LLR value in the MLD method, depending on the selection method of the signal point set to be left as the signal point set included in the final list in narrowing down the candidate signal point set, Decoding characteristics in the decoding unit that performs the deterioration. In the signal separation using the PCA method of the encoded MIMO system, the decoding characteristic is improved by adding a signal point having the closest square Euclidean distance to the searched signal point. Furthermore, the previous additional signal point is replaced with a new additional signal point obtained by further traversing the tree structure in the reverse direction from the lowest node of the estimated signal point set obtained from the signal point set for which the search has been completed, Alternatively, by adding to the previous additional signal point, a trade-off between the reduction effect of the calculation amount and the decoding characteristic is achieved at a better level.
[Selection] Figure 13

Description

  The present invention relates to wireless communication technology. More specifically, the present invention relates to an encoded multiple-input multiple-output (MIMO) system that uses a plurality of antennas on the transmission side and the reception side, respectively.

  Spatial Multiplexing, which is a representative MIMO communication technology, is known as a technology that uses multiple transmission antennas and independently transmits signals from each antenna, thereby increasing the transmission rate without increasing the frequency band. It has been. It has been attracting attention since the latter half of the 1990s, and is currently adopted in IEEE 802.16m, which is a standard applied to WiMAX2, and LTE systems (ETSI TS 136 211 etc.), and will continue to be a highly important technology.

  FIG. 1 is a diagram showing an outline of the configuration of a MIMO communication system. The MIMO system 1 communicates with the transmission side device 2 and the reception side device 3 via a radio wave propagation path 4 using a radio frequency signal. In the MIMO system 1, the transmission-side device 2 uses a plurality of antennas. Furthermore, the receiving side device 3 also uses a plurality of antennas. In FIG. 1, the number of antennas is four in both the transmission system and the reception system, but is not limited to this. Further, even when the number of antennas in the transmission system and the reception system is different, the effect is known when the number of antennas in the reception system is larger.

  The transmission signal sequence 5 is subjected to encoding such as error correction coding and interleaving by a modulation unit not shown in the figure, and then converted into the number of transmission antennas by a serial / parallel conversion unit (S / P conversion). Divided into a plurality of corresponding signal sequences. Thereafter, each signal sequence is modulated into a radio signal in the same frequency band and transmitted from the corresponding transmitting antenna.

  The radio signal transmitted from each antenna travels along the propagation path 4 while receiving independent fading, and reaches each antenna of the receiving-side device 3. In FIG. 1, the state of radio waves is indicated by four arrows, but radio signals output from one transmission antenna are received by all reception antennas. Therefore, if attention is paid to one receiving antenna, radio signals from all transmitting antennas are received. Note that since radio signals of the same frequency band are used, so to speak, a plurality of different radio signals are received in a state of interference. That is, the transmission rate can be improved by simultaneously transmitting different information sequences using the same frequency resource at the same time. Here, in the MIMO system, the received signals in the mixed state received by the four receiving antennas are separated and detected using a signal separation technique described later. That is, the receiving side apparatus 3 estimates a plurality of signal sequences transmitted from the respective antennas of the transmitting side apparatus 2 and performs parallel / serial conversion (P / S conversion) to obtain the estimated transmission data series 6.

  FIG. 2 is a diagram showing an outline of a more detailed configuration of the MIMO communication system. The MIMO communication system 1 includes a transmission side device 2 drawn on the upper side of FIG. 2 and a reception side device 3 drawn on the lower side of FIG. The wireless propagation path is indicated by a matrix H4 indicated by bold characters. The matrix H is a channel matrix represented by fading fluctuation between the transmission antenna and the reception antenna, which is generally known by the MIMO technique.

  The transmission side device 2 includes an input data sequence 101, an encoding unit 102, an interleaver 103, a modulation unit 104, a MIMO mapping unit 105, and a fast inverse Fourier transform unit 106. The receiving side device 3 includes a fast Fourier transform unit 108, a MIMO detection unit 109, a deinterleaver 110, a decoding unit 111, and an output data sequence 112. The encoding unit 102 performs error correction encoding and the like. Modulation section 104 modulates a bit string into a symbol, and MIMO mapping section 105 arranges a modulated signal for each transmission antenna. On the receiving device 3 side, signal separation is performed based on a plurality of reception signals received by a plurality of reception antennas in the MIMO detection unit 109. In MIMO detection section 109, the most probable received data sequence is estimated and passed to decoding section 111 together with the likelihood value. The present invention relates to the detection of a MIMO signal in the MIMO detection unit 109. The prior art MIMO detection method required to describe the present invention is described below.

FIG. 3 is a diagram for explaining signal points in the multilevel modulation method. In a digital modulation system for transmitting (changing) information of a data signal as a change in phase or amplitude of a carrier wave, a signal point on a signal space defined by two orthogonal I-axis and Q-axis, and data And correspond. For example, FIG. 3A is a signal space diagram showing signal points of quadrature amplitude modulation (4QAM), and four signals 31 having different signal spaces defined by the I axis and the Q axis, Each of 32, 33, and 34 can be associated with 2-bit information of 00, 01, 11, and 10. Therefore, when 4QAM is used, information of 2 bits can be carried by one signal point. The number of bits n = 2 corresponds to 2 ² = 4 signal points.

FIG. 3B shows a signal space diagram of 16QAM. In 16QAM, there are four types of amplitude values that can be taken on each of the I axis and the Q axis. Therefore, a total of 4 × 4 = 16 signal points are obtained, and a plurality of bit strings can be associated with these 16 signal points. For example, 4 bits 00/00 are assigned to the signal point 51, 10/01 is assigned to the signal point 52, and 10/10 is assigned to the signal point 53. Therefore, in 16QAM, information of 4 bits can be carried by one signal point. The number of bits n = 4 corresponds to 2 ⁴ = 16 signal points.

Similarly, if there are eight types of amplitude values that can be taken on the I axis and the Q axis, a 64 QAM capable of carrying information for 6 bits per signal point can be constructed, and the number of bits n = 6 is 2 ^6. = 64 corresponding signal points. In this way, the number of bits that can be carried by one signal point can be increased by using a multi-level modulation technique (high-order modulation technique) that multi-values the amplitude levels that can be taken on the I axis and the Q axis. it can. One signal point in the signal space diagram is associated with one modulation symbol. In 4QAM, 2 bits can be carried by one symbol, and in 64QAM, 6 bits can be carried by one symbol. The duration of a symbol is called symbol time. If the information transmission rate (bit rate) is kept constant, the symbol rate (modulation speed) can be lowered. Conversely, if the symbol rate (modulation speed) is constant, the information transmission rate (bit rate) is increased by the higher-order modulation method.

  (C) of FIG. 3 is a figure explaining the information transmission which is possible when the modulation scheme is QAM and the number of antennas is m = 4. If four antennas receive different radio signals independently (not MIMO), if the signal points 35, 36, 37, and 38 can be detected, the bit strings 39, 40, 41, and 42 are reproduced from the respective antennas. can do. Furthermore, the bit string 43 can be reproduced by synthesizing these bit strings. However, in the MIMO system, all signals from a plurality of transmission antennas are received by each reception antenna in a state of interference and interference with each other.

  FIG. 4 is a diagram for explaining a received signal at one receiving antenna of the receiving apparatus of the MIMO system. In the MIMO system, for example, radio signals are transmitted independently from m = 4 antennas. Therefore, the radio signal 21 reflecting the information of the signal point 14 (bit string 00) is received by the receiving antenna 20 from the first antenna 10. At the same time, the radio signal 22 reflecting the information on the signal point 15 (bit string 11) is received by the receiving antenna 20 from the second antenna 11. Further, the radio signal 23 reflecting the information of the signal point 16 (bit string 00) is received from the third antenna 12 by the receiving antenna 20, and the information of the signal point 17 (bit string 01) is received from the fourth antenna 13. The reflected radio signal 24 is also received by the receiving antenna 20.

  Therefore, if a signal received by the first antenna 20 on the receiving side is shown in a signal space diagram, the radio signals 21 to 24 are synthesized. In FIG. 4, for the sake of simplicity, only the signal point 18 corresponding to the received signal from the first transmitting antenna 10 and the signal point 19 corresponding to the received signal from the second transmitting antenna 11 were received independently. Shown as a thing. Therefore, in practice, the phase and amplitude of the four received signal points are observed as synthesized signal points. Since each of the radio signals 21 to 24 is fading independently in the propagation path, the amplitude and phase fluctuations received by the fading of each received signal point from the four transmitting antennas are also independent. The same situation as that of the first receiving antenna 20 also occurs in the other second to fourth receiving antennas. In this way, a plurality of different signals transmitted from a plurality of antennas on the same frequency band interfere with each other and are received by each of a plurality of separate receiving antennas in a state of interfering with each other. In such a case, signal separation is used to reproduce the information of each of a plurality of different signals.

  One of the most important technologies for a MIMO spatial multiplexing system is separation of signal distribution technology for receiving signals in which a plurality of transmission streams interfere with each other at a receiver. Many signal separation methods in prior art MIMO spatial multiplexing systems have been proposed. For example, linear equalization method (Non-patent document 1), maximum likelihood detection method (Non-patent document 2), M algorithm (Non-patent documents 1 and 2), SD method (Non-patent document 3), FSD method (Non-patent document) 4, 5), PCA method (non-patent document 6), and the like. As will be described later, the present invention improves the PCA (Parallel Candidate Adding) method in the MIMO detection section of FIG. 2, thereby reducing the amount of computation and suppressing the degradation of decoding performance due to signal separation errors. I will provide a. In the following description, the outline of the signal separation method will be briefly described first.

  The received signal y in the MIMO spatial multiplexing system is expressed by the following equation. In the following equation, italic letters represent vectors or matrices.

y = Hx + z Equation (1)
In the above equation, x represents a transmission signal vector, y represents a reception signal vector, z represents a white Gaussian noise vector, and H represents a block Rayleigh fading channel matrix. Each of these elements is expressed in more detail as follows: Here, N _t is the number of transmitting antennas, and N _r is the number of receiving antennas.

The most typical method of signal separation is maximum likelihood detection (MLD). The transmission vector in Expression (2) is given as a combination of transmission signals of a transmission signal sequence transmitted from each transmission antenna. Specifically, for example, when N _t = 4 shown in FIG. 4, the transmission vector is given as a combination of signal points 14, 15, 16, and 17. When the modulation method is 4QAM shown in FIG. 4, there are four types of transmission signals (signal points) that can be taken by each antenna. Therefore, if there is a N _t antennas, possible combinations of signal points it becomes 4 ^Nt pieces. When there are four transmission antennas, 256 signal point combinations are possible.

In the MLD method, the most probable “signal point combination” is selected from all possible “signal point combinations” using the square Euclidean distance to the actual received signal point. This is a method of using the combination as an estimated transmission signal (optimal solution) for subsequent decoding processing. However, as can be easily understood, as the number of antennas and the number of signal points in the signal space increase, the number of signal point combinations increases exponentially. For example, when the number of antennas is four (N _t = 4) and 64QAM, the total number of “signal point combinations” is 16,777,216. In order to obtain the optimum combination of signal points, the square Euclidean distance is calculated as many times as the total number of combinations. Therefore, a huge amount of calculation is required for estimation of the received signal. As a conventional technique, a signal estimation method that reduces the amount of calculation as described above has been studied. Next, in order to explain the PCA method that is the basis of the present invention, the overall likelihood estimation method will be briefly described. First, as a premise of the MLD method, a Fixed-Complexity Sphere Decoding method (hereinafter referred to as FSD method), which is an algorithm for obtaining a suboptimal solution, will be described. Then, the PCA method that is the basis of the present invention will be described, and the problems in the prior art will be clarified.

  Each signal separation method described below does not calculate a square Euclidean distance for all “signal point combinations” (signal point set candidates) in the MLD method, but a limited number of “signal point sets”. This is a technique related to how to select and reduce the amount of calculation. Therefore, it should be noted that how to select a signal point set is a central issue in order to reduce the amount of computation while maintaining the accuracy of signal separation. In the following description, for the sake of simplicity, each “combination of signal points” of a plurality of antennas is also referred to as a “signal point set”. Moreover, selecting a signal point candidate and searching for a signal point are used in the same meaning.

  The SD (Sphere Decoding) method can always obtain the same optimal solution as the MLD method. However, the amount of computation is not constant because it depends on the magnitude of noise and fading. In an actual communication system, since there is a demand for a constant data rate, it is desirable that the amount of calculation required for signal search is constant. Therefore, in the FSD method, by determining the number of signal point search candidates in advance, it is possible to obtain a sub-optimal solution with a constant amount of computation without depending on the magnitude of noise and fading. The FSD method is divided into two algorithms: a pre-process for rearranging channel matrices (Pre-Process) and a core process (Core-Process) for searching for a signal that is a suboptimal solution using a tree structure. Further, the core process performs full enumeration (Full-Enumerate: hereinafter referred to as FE) for searching all signal point candidates in the search layer and single enumeration (Single-Enumerate: hereinafter referred to as SE) for searching only one signal point. And the following two enumeration methods. The outline of the FSD algorithm will be described below.

FIG. 6 is a diagram for explaining the outline of the FSD method. First, pre-processing will be described. In the pre-process, the columns of the channel matrix H are rearranged. That is, association is performed between each of the four physical transmission antenna ports 300 exemplarily shown in FIG. 6 and the layers in the signal search tree structure 302. Therefore, the number of physical transmission antennas and the number of layers are basically the same. In the case of FIG. 6, the tree-structured layers include the highest layer 1 to the lowest layer 4. In the core process to be described later, when each layer in signal search is associated with each physical transmission antenna, transmission that gives a signal sequence with the highest emphasis on receiver noise (lowest signal-to-noise ratio) in FE. The antenna is selected, and in the case of SE, it is planned to select a transmission antenna that gives a signal sequence with a smaller receiver noise enhancement (a better SN ratio). That is, the column rearrangement 301 of the channel matrix H in the pre-process is a process of determining which physical transmission antenna of a plurality of transmission antennas is associated with which layer of the tree structure. The pre-processing algorithm is shown below. i = 1,. . . The N _t, the layer number in the tree structure 302 of the signal search. The layer number is as shown in the tree structure 302 of FIG.

STEP1 (Determination of inverse matrix)
First, a pseudo inverse matrix expressed by the following equation is obtained.

Here, H _Nt = H. H _i is a matrix obtained by nulling the columns selected at the previous iteration in the sorting. Nulling means that only the real part of the row of the selected signal point (k selected in STEP 2) in the selected column is set to 1, and the values of the other columns are set to 0. This makes it possible to remove already selected signal points.

STEP2 (Determination of search signal)
The column is selected by the following formula.

K represents the index of the signal.

By repeating STEP1 and STEP2 N _t times (the number of transmission antennas), the columns are exchanged. As an example, when the selection is finally made in the order of k = 2, 1, 4, 3, the tree structure is searched in the order of the next transmission signal sequence.

  Note that the above-described column replacement is determined using channel information such as a signal-to-noise ratio (S / N ratio) of a reception signal from each transmission antenna. As described above, the channel matrix replacement in the pre-process is performed by associating between the physical antenna and the tree structure layer, and based on the quality of the transmission signals from the plurality of transmission antennas. This is a signal point search step for associating a plurality of layers in a tree structure for estimating a signal point of a transmission antenna with a plurality of transmission antennas. As the quality of the transmission signal, for example, the S / N ratio of each transmission signal from the transmission antenna received by the reception antenna can be used. However, other quality parameters can be used and are not limited thereto. Not.

  Referring to FIG. 6, the association is performed between the physical antenna 300 and the tree structure 302 layer as follows. This means that the physical transmission antenna Ant2 is associated with the highest layer Layer1, the physical transmission antenna Ant1 is associated with the second Layer2, the physical transmission antenna Ant4 is associated with the third Layer3, and the physical transmission antenna Ant3 is associated with the lowest Layer4. Thus, it should be noted that one layer in the tree structure corresponds to one physical transmit antenna. A node (point ●) in the tree structure corresponds to a signal point, and at the same time, corresponds to a specific bit string corresponding to the modulation method. In the tree structure, the search for signal point candidates is basically performed in order from the highest layer Layer 1 to the lower layer.

  In the case of the FE of the core process to be described later, since all signal point candidates are selected at least in the highest layer 1 and signal point candidates are not reduced, all likelihood values that should be originally included in the MLD method are all. Will be included. For this reason, even if a signal sequence (physical transmission antenna) having the largest receiver noise is selected as this Layer 1, characteristic deterioration does not occur. On the other hand, in the case of the SE of the core process, a signal (physical transmission antenna) having relatively smaller receiver noise needs to be selected as a higher layer in the SE. If a higher layer (for example, Layer 2) is selected and a signal point candidate selection error occurs in that layer, the influence of the selection error propagates to the lower layer likelihood value calculation result. is there.

Next, the algorithm of the core process is shown. As already described, in the FE, all signal point candidates of the layer to be searched are selected and survived. Here, surviving means that the signal points are not deleted and maintained as candidates. On the other hand, in SE, only one child node (signal point) is selected and survived for each parent node (signal point) of the layer above the layer to be searched. Layer number Layer _SE applying a layer number Layer _FE and SE apply the FE is obtained by the following equation.

As an example, when the modulation scheme is 4QAM and the number of transmitting antennas N _t = 4, Layer _FE = 1 and Layer _SE = 3. The signal search range in the FSD method in this example is each signal point shown in FIG. In FIG. 6, FE is adopted for the highest layer 1 and all four signal points that can be taken by 4QAM are selected as signal point candidates. Bit lines corresponding to signal points such as 00, 01, 10, and 11 are displayed at the four signal points of Layer 1 in FIG. 6, and the signal points 31, 32, and 33 described with reference to FIG. , 34.

  SE is adopted for the three layers from the second Layer 2 to the lowest Layer 4, and one signal point is selected as a signal point candidate in each layer. For example, in Layer2, one signal point 303-2 is selected for the signal point 303-1 corresponding to the bit string 01 in Layer1. At this time, on the condition that the signal point corresponding to the bit string 01 is selected as the signal point of Layer 1 (physical antenna is Ant 2), Layer 2 (physical antenna is Ant 1) determined to be closest to the actual reception signal point. A signal point is selected. Similarly, in Layer 3, one signal point 303-3 is selected with respect to the previously selected signal point 303-2 of Layer2. At this time, the condition that the signal point 303-1 corresponding to the bit string 01 is selected as the signal point of Layer1 (physical antenna is Ant2) and the signal point 303-2 of Layer2 (physical antenna is Ant1) is selected. Thus, the signal point of Layer 3 (physical antenna is Ant 4) determined to be closest to the actual reception signal point is selected. Similarly, in Layer 4, one signal point 303-4 is selected with respect to the previously selected signal point 303-3 of Layer3. At this time, the signal point 303-1 corresponding to the bit string 01 column is selected as the signal point of Layer1 (physical antenna is Ant2), the signal point 303-2 of Layer2 (physical antenna is Ant1) is selected, and Layer 3 (physical antenna is Ant4) signal point 303-3 is selected under the condition that Layer3 (physical antenna is Ant4) signal point 303-3 is selected, and Layer4 (physical antenna is Ant3) signal point 303-4 is selected. The

  The determination of the closest signal point is simply obtained from the relationship between the actual received signal position and the possible signal point positions in the signal space.

In FIG. 6, the four signal points 303-1 to 303-4 connected between the four layers correspond to one “signal point combination” 305 and form a signal point set 305. As apparent from FIG. 6, when the modulation scheme is 4QAM and the number of transmission antennas N _t = 4, four signal point sets are selected as candidates in the FSD method. When the modulation scheme is 4QAM and the number of transmission antennas N _t = 4, the number of possible signal point sets (“signal point combinations”) is 256. In the MLD method, all 256 signal point sets are used. A likelihood value is obtained for. On the other hand, in the FSD method, four of 256 signals are selected as a signal point set used for calculation of likelihood values. Based on the tree structure 302 shown in FIG. 6, the four signal point sets selected by the FSD method (in the case of 4QAM, N _t = 4) are the basic signal point sets in the PCA method described later. Please note that. Next, the likelihood calculation method will be briefly described.

In an encoded MIMO system, it is necessary to calculate a log-likelihood ratio (LLR) value for each code bit. When the estimated signal of the m-th antenna, that is, the n-th bit of the bit string corresponding to the signal point is b _{m, n} , the LLR is expressed by the following equation.

  However, when the signal separation method is actually implemented based on Expression (9), Expression (9) has a large amount of calculation. Therefore, the equation (9) is approximated as the following equation.

  In the above formula

Represents a set of symbols whose bits b _{m, n} are −1 in the signal point candidates to be searched. Also,

Represents a set of symbols whose bits b _{m, n} are +1 in the signal point candidates to be searched. Bits −1 and +1 are for bipolar encoding and have the same meaning as bits 0 and +1 for unipolar encoding. As already described in FIG. 6, one node in the tree structure corresponds to one signal point and further corresponds to a bit string. Further, the signal points have a one-to-one correspondence with the symbols, and the set of signal points is synonymous with the set of symbols. In the implementation in the actual signal separation method, only “a set of selected signal points” for finally calculating the LLR as a result of the search need be considered. For this reason, Formula (10) can be further approximated as the following formula.

In the above equation (11), Λ represents a list (List) of signal point sets that are finally determined after searching for signal point candidates. The difference between Expression (10) and Expression (11) is whether or not a list is used for calculation of likelihood in order to reduce the amount of calculation. When the likelihood value L (b _{m, n} ) cannot be calculated or the likelihood value L (b _{m, n} ) is too large depending on the calculation amount reduction method, decoding based on the likelihood value obtained as a result of the signal separation method As a result, the decoding characteristics deteriorate. It is necessary to use clipping for this problem. In the present invention, if the LLR cannot be calculated or is too large, the clipping value L _maz = ± 8 is used.

Kenichi Higuchi, Hidekazu Taoka, et al., “Multi-antenna wireless transmission technology, 3 Signal separation technology in MIMO multiplexing”, NTT DoCoMo Technical Journal, Vol. 14, No. 1, April 2004 K. J. Kim and J. Yue, “Joint channel estimation and data detection algorithms for MIMO-OFDM systems,” in Proc. Thirty-Sixth Asilomar nference on Signals, Systems and Computers, vol.2, pp.1857-1861, Nov. 2002 E. Viterbo and J. Boutros, “A universal lattice code decoder for fading channels,” IEEE Trans. Inform Theory, vol. 45, no.5, pp. 1639-1642, July. 1999 LG Barbero and JS Thompson, “A Fixed-Complexity MIMO Detector Based on the Complex Sphere Decoder,” in 7th IEEE Workshop on Signal Processing Advances in Wireless Communications (SPAWC'06), Cannes, France, pp.1-5, July. 2006. L. G. Barbero and J. S. Thompson, “Fixing the Complexity of the Sphere Decoder for MIMO Detection,” IEEE Trans. Wireless Commun, vol. 7, no. 6, pp. 2131-2142, June 2008 Xiang Wu and S. Thompson, “A Fixed-Complexity Soft-MIMO Detector via Parallel Candidate Adding Scheme and its FPGA Implementation” IEEE Commun., Vol. 15, no.2, pp. 241-243, Feb. 2011

  When performing signal separation on the basis of the MLD method, there is a Parallel Candidate Adding (PCA) algorithm as a promising method for reducing the amount of calculation. The present invention is an improvement of the PCA method, which is a signal separation method using the PCA algorithm. Hereinafter, the PCA method as the basis of the present invention will be described.

  FIG. 5 is a block diagram of a configuration of a MIMO detection unit that performs signal separation including the PCA method. The MIMO detection unit 200 corresponds to the MIMO detection unit 109 of the MIMO communication system in FIG. Since each element of the MIMO detection unit 200 in FIG. 5 represents a conceptual functional block, the flow of signals and information is rough. In addition, the functions performed by each element may be performed by another block, and some of the functions shown in FIG. 5 may be performed by substantially the same block. Therefore, it should be noted that the configuration of the functional block is not limited to that shown in FIG. At least a part of each element of the MIMO detection unit 200 is performed by hardware, by software information processing using a CPU, memory, logic circuit, digital signal processor (DSP), or the like, or a combination of hardware and software information processing Can be implemented. The configuration and function of each functional block of the MIMO detection unit shown in FIG. 5 will be understood from the following description of the PCA method.

  The FSD method already described is a method for reducing the amount of calculation in the process until obtaining the suboptimal solution, and does not consider the likelihood of each bit of the bit string corresponding to the signal point in the encoded MIMO. On the other hand, the PCA method generates the bit likelihoods sequentially by calculating the cumulative square Euclidean distance of the parent node, that is, the partial likelihood in addition to the signal point search in the FSD method. This is a method of adding signal point candidates. In other words, the PCA method provides a signal point candidate selection method that has better decoding characteristics and a sufficiently reduced amount of computation than the FSD method by further adding signal point candidates. The PCA method performs exactly the same pre-process as described for the FSD method. Therefore, the difference from the FSD method lies in the method for selecting signal point candidates in the core process. The algorithm outline of the core process of the PCA method will be described below. The core process of the PCA method is composed of FE (complete enumeration method) and SE (single enumeration method) as in the FSD method.

At FE As in the FSD method, all possible signal point candidates in the layer to which the signal point candidate is to be added are searched.

At SE Select the node with the highest partial likelihood (the node with the smallest square Euclidean distance) of the node (parent node) one level higher than the layer to which you want to add the signal point candidate, and try to add the signal point The node having the highest partial likelihood is searched for among the nodes (child nodes) of the existing layer. Note that a node has the same meaning as a signal point or symbol in signal space in the context of signal point search. Next, ω signal points obtained by changing only one bit are added as signal point candidates to the bit string corresponding to the selected node (child node) of the layer to which the signal point is to be added. For example, when 16QAM is used as the modulation method, if the bit string corresponding to the child node is “0000”, the bit string of the signal point candidate to be added is “1000”, in which only the first bit is changed to 1. 2 There are four types: “0100” in which only the bit is changed to 1, “0010” in which only the third bit is changed to 1, and “0001” in which only the fourth bit is changed to 1. These additional candidates are simple to select and configure (although not necessarily signal points with a minimum Euclidean distance for each bit). Except for the parent node (Layer 1) having the largest partial likelihood, only the node having the largest partial likelihood is searched for among the child nodes in the same manner as the FSD method.

FIG. 7 is a diagram for explaining addition of signal point candidates in the PCA method. In this example, for simplicity, the tree structure 400 is shown when the modulation scheme is 4QAM and the number of transmission antennas N _t = 4, and the search range of signal points by the PCA method is shown. As can be seen by comparing the tree structure 400 shown in FIG. 7 and the tree structure 302 of FIG. 6 in the FSD method, four signal point sets arranged immediately below each of the four signal points in Layer 1 in FIG. The configuration is the same as the configuration of the tree structure 302 in FIG. Although details will be described later, nodes 410, 411, and 412 surrounded by dotted lines indicate nodes that are newly added in the PCA method when compared with the FSD method. In addition, nodes 401, 404, and 407 surrounded by circles represent nodes having the highest partial likelihood in each layer. The notation such as 10/01/11/00 indicates a bit string of signal points corresponding to each node in the order of layers. The process of determining signal point set candidates corresponds to finding the tree structure of FIG. 7 and identifying the signal point set that ultimately calculates the likelihood value LLR. The tree structure search is performed in order from the upper layer.

  FIG. 8 is a flowchart showing an outline of signal separation by the PCA method. In the following description, the addition of signal points in the PCA method will be described more specifically using the tree structure of FIG. 7 with reference to the flowchart of FIG.

  Signal separation by the PCA method starts at step 100 (hereinafter, step is denoted as S) in FIG. Signal separation is started for a certain symbol time (S100). In S102, the actual reception signal point of each reception antenna branch is determined based on the channel information. Next, in S104, the signal point candidate determination process is started and the PCA method is started. In the next S106, a pre-process similar to the FSD method is executed. That is, the channel matrix H is exchanged to associate the physical transmission antenna with the layer in the tree structure for searching for the signal point candidate. In S106, the physical transmission antenna selected as Layer 1 of the highest layer is determined.

  Here, referring to FIG. 7, the signal point in Layer 1 of the top layer of the tree structure is selected. Thereafter, in S108 in FIG. 8, the core process in the PCA method is completed by repeating the stage of the number Nt of transmission antennas, and the search list is determined as shown in S110. In the example of FIG. 7, since Nt = 4, signal points are searched in four stages for four antennas. Hereinafter, addition of signal point candidates in the core process of the PCA method corresponding to S108 and S110 will be described. The flow of the core process of the PCA method will be described with reference to the more detailed flow diagrams of FIGS.

  FIG. 9 is a first flowchart for explaining signal point addition in the PCA method, and FIG. 10 is a second flowchart for explaining signal point addition in the PCA method. In the following description, the corresponding step (S ***) in the flow of FIGS. 9 and 10 is described in parentheses.

  Referring to FIG. 7 again, partial likelihoods are calculated for the four signal points selected by Layer 1 (parent node) (S203, S401). At this time, the most probable signal point 401 is identified as the transmission signal from the physical antenna corresponding to Layer 1, that is, the first selection antenna. This is the first stage.

Next, as a second stage, signal points for the second selection antenna (Layer 2) are selected (S205 to S209). First, partial likelihood maximum, i.e. the squared Euclidean distance is minimum signal point _{A 1} is added to each Layer2 (child node) (S205). That is, each signal point Layer2 drawn immediately below each of the four signal points Layer1 is _{A 1.} Add the _{A 1} corresponds to the SE in the FSD process. For simplicity, a signal point selected by the FSD method below the signal points 401, 404, and 407 having the maximum partial likelihood in each layer is referred to as a “reference signal point”. That is, A ₁ , A ₂ , A ₃ , and A 4 are reference signal points including those described later.

The PCA method, in addition to the partial likelihood maximum signal point _{A 1} (reference signal point), with respect to the signal point 401 having the maximum likelihood (bit string 00), the signal point of Layer2 _{A 1} (reference signal Add signal points _{B 1} obtained by inverting one bit of the bit sequence (00/00) corresponding to the point) (S207). That is, signal point 402 (bit string is 00/01) and signal point 403 (bit string is 00/10) are added to Layer2. Thus the PCA method, adds two signal points _{B 1} in Layer2, a total of six signal point candidates are selected in Layer2. Here, the partial likelihood (square Euclidean distance) is calculated for the signal point candidate selected by Layer 2 by cumulatively considering a predetermined signal point, that is, the signal point of the higher Layer 1 (S209, S403). . Then, the signal point 404 having the minimum square Euclidean distance is specified.

Further, as a third stage, signal points for the third selection antenna (Layer 3) are selected (S213 to S217). First, a child node in the second stage is regarded as a parent node, and signal point selection similar to that in the second stage is performed. That is, the signal point A ₂ (reference signal point) having the maximum partial likelihood is added to each signal point of Layer 2 (S213). Further, with respect to A ₂ (reference signal point), the signal string 404 having the maximum likelihood (bit string is 01/00), the bit string (01/00/00) corresponding to the signal point A _{2 of} Layer 3 Add signal point _{B 2} obtained by inverting the bit of the (S215). That is, signal point 405 (bit string is 01/00/10) and signal point 403 (bit string is 01/00/01) are added to Layer3. Thus the PCA method, two signal points _{B 2} in Layer3 is newly added, a total of eight signal point candidates are selected by the Layer3. Here, the partial likelihood (square Euclidean distance) is calculated for the signal point candidates selected in Layer 3 by cumulatively considering the predetermined signal points, that is, the higher Layer 1 and Layer 2 signal points (S217, S405). Then, the signal point 407 having the minimum square Euclidean distance is specified.

Finally, as the fourth stage, signal point selection for the fourth selection antenna (Layer 4) is similarly performed. In Layer 4, two signal points are newly added to the reference signal point, and in Layer 4, a total of 10 signal point candidates are selected. At this stage, partial likelihoods (square Euclidean distance) are calculated by cumulatively considering predetermined signal points, that is, signal points of the higher Layer1, Layer2, and Layer3 (S217, S405).
In FIG. 10, four stages corresponding to the four layers are described, and it is understood that the partial likelihood is calculated by adding the signal points of each layer in each stage. In step S406, when the signal points of the fourth layer Layer 4 are added, the tree structure shown in FIG. 7 is completed, and all signal point sets selected by the PCA method are determined. That is, at the end of S406, a list of signal point sets for finally calculating the LLR value is determined. This list corresponds to the list used when calculating the LLR using equation (11).

  As can be seen from the tree structure 400 in FIG. 7, in the PCA method, in addition to the four signal point sets given by the FSD method, six signal point sets are newly added, and a total of 10 signal point sets are specified. Is done. The PCA method adds only one bit in the bit string corresponding to the signal points 401, 404, and 407 specified based on the partial likelihood from the parent node higher in the layer in each layer. Signal point groups 410, 411, and 412 having changed bit strings. A code in which only one bit is changed from a certain value to an adjacent value is known as a Gray code. The PCA method can be said to be a method of adding signal points corresponding to the Gray code.

When the list of signal point candidate sets (search list) is determined, the process returns to the overall flow chart of FIG. 8 again, and in S112, candidate signal point sets in the list, that is, “signal point combinations” in the list. An LLR value is calculated for the candidate. That is, the likelihood value L (b _{m, n} ) of each bit b _{m, n} is calculated using equation (11), and the most likely signal point set that is most likely to be an estimated signal is determined in S114. The Finally, in S116, the likelihood value information of the bit string corresponding to each antenna is passed to the subsequent decoding block (decoding unit 111 in FIG. 2), and signal separation for the next symbol time is started. (Return to S102).

  Here, returning to the block diagram of FIG. 5 again, the signal point addition process (S401 to S408) by the PCA method described in FIG. 9 and FIG. 10 is performed in the signal point candidate addition block 204. That is, in the signal point candidate addition block 204, the signal points are generated while sequentially generating each bit likelihood by calculating the cumulative square Euclidean distance, that is, the partial likelihood while considering the information of the already selected upper node. Add candidates.

In the PCA method, the number of search nodes μ _PCA and the number of lists N _PCA are each expressed by the following equations.

The maximum likelihood detection (MLD) method, which is the most representative method of signal separation, uses a square Euclidean distance from an actual received signal point out of all possible combinations of signal points. In this method, an LLR value is calculated and used for subsequent decoding processing. Search node number _{mu MLD} in the MLD method, the number of transmission antennas is the case of _{N t} and MQAM modulation scheme, the following equation.

  Comparing Equation (12) and Equation (14), the number of nodes to be searched is greatly reduced, but when the signal points newly added by the PCA method are viewed on the signal space diagram, the problem in the PCA method is recognized. Is done.

  FIG. 11 is a diagram showing the positions of signal points added in the PCA method according to the positions of reference signal points. FIG. 11A shows three regions 500, 501, and 502 of the position of the reference signal point that is the base point of the signal point to be added. (B) to (d) of FIG. 11 show the positions of the additional signal points in each region for each region.

  As shown in FIG. 11A, when the position of the reference signal point is in the region 500 at the position of the four corners of the signal space diagram, in the case of the outer region 501 excluding the four corners, The area 502 is divided into the following cases. FIG. 11B shows the positions of the additional signal points 504-1 to 504-4 when the reference signal point 503 is in the region 500. FIG. 11C shows the positions of the additional signal points 506-1 to 506-4 when the reference signal point 505 is in the region 501. Similarly, (d) of FIG. 11 shows the positions of the additional signal points 508-1 to 508-4 when the reference signal point 507 is in the region 502.

  When the reference signal points are in the areas 500 and 501, in addition to adding signal points adjacent to the reference signal points, signal points 504-3, 504-4, and 506-4 not adjacent to the adjacent signal points are added. ing. The reference signal points are signal points 401, 404, and 407 shown in FIG. 7, and represent the nodes having the highest partial likelihood in each layer. Therefore, it is the most probable estimate of the signal point of the transmission signal. Nevertheless, the additional signal points based on the Gray code in the PCA method include signal points 504-3, 504-4, and 506-4 that are significantly separated from the reference signal point.

  The inventors considered that there is room for further improvement of the PCA method in this regard. In order to reduce the amount of calculation of the LLR value in the MLD method, depending on the selection method of the signal point set included in the final list, the decoding characteristic in the subsequent decoding unit is deteriorated. Resulting in. If the signal point set that should originally remain in the list is not selected, the estimated signal point set may contain errors. Compared with the PCA method of the prior art, there is a high need for new improvements to achieve a trade-off between the effect of reducing the amount of calculation and the error rate characteristics at a better level.

  The present invention has been made in view of such problems, and the object of the present invention is to improve the problems of the conventional PCA method in the signal separation method of the coded MIMO system, and to reduce the amount of calculation. It is an object of the present invention to provide a new signal separation method (a method for estimating a signal point of a transmission signal) and a receiving apparatus that achieve a trade-off between a reduction effect and a decoding characteristic at a better level.

  In order to achieve the above object, according to the present invention, a plurality of transmission signals corresponding to different transmission sequences transmitted simultaneously from each of a plurality of transmission antennas are transmitted from a plurality of reception antennas. In a receiving apparatus for receiving each of the plurality of transmission signals, a method of estimating each signal point of the plurality of transmission signals from a plurality of signal points associated with a plurality of bit strings in each signal space of the plurality of transmission signals. A first signal point search step associating a plurality of layers in a tree structure for estimating the signal points of the plurality of transmission antennas with the plurality of transmission antennas based on the received quality of the transmission signals; , A ray that selects a current signal point candidate out of the plurality of layers based on likelihood information regarding the plurality of bits associated with the signal point. In the first selection step of selecting signal point candidates, and for each of the signal point candidates selected in the layer in which the current signal point candidate is selected, the signal points that have been selected below the highest layer On the condition that the signal point candidate including the candidate is selected, the partial signal level in the lower layer is determined based on the cumulative likelihood information for the signal point in the lower layer than the layer in which the current signal point candidate is selected. A second selection step of selecting a reference signal point having the maximum likelihood; and a third selection of selecting a signal point candidate having a minimum square Euclidean distance with respect to the reference signal point in the lower layer. And the first selection step to the third selection step until the signal point candidate of the lowest layer is selected. A second signal point search step composed of a step repeatedly executed on the upper layer, and a combination of signal point candidates determined by the tree structure obtained from the second signal point search step. Calculating a log-likelihood ratio for each of the bit strings corresponding to the signal points of the plurality of antennas from the signal point combinations in the list based on the list; and based on the log-likelihood ratio, Estimating the most probable combination of signal points with respect to the transmission signal of.

  The quality of the transmission signal is not limited to this. For example, the S / N ratio of the transmission signal received by the reception antenna can be used. An example of the first signal point search step may be a pre-process. Also, the second signal point search step described above. It can be a core process. Further, the signal point candidate having the minimum square Euclidean distance corresponds to the NN node in the embodiment.

  The invention of claim 2 is a method of estimating a signal point according to claim 1, wherein the lowest layer of the most probable signal point combination is followed by the step of estimating the most probable signal point combination. Tracing the tree structure from the lowest layer to the highest layer with the signal point of the base point as the base point, specifying the reference signal point in the reverse direction, and square Euclidean for the reference signal point in the reverse direction A backward signal point comprising a step of selecting a signal point candidate having a minimum distance, a step of identifying the reference signal point in the backward direction, and a step of sequentially repeating the step of selecting until reaching the highest layer. A combination of signal point candidates newly added by the search step and the backward signal point search step is a signal obtained by the tree structure. Re-calculating the log-likelihood ratio based on the new list added to the combination of point candidates, and the most probable signal for the plurality of transmission signals based on the re-calculated log-likelihood ratio The method further includes the step of estimating a combination of points.

  A third aspect of the invention is a method for estimating a signal point according to the first aspect, wherein two or more reference signal points are selected in the second selection step of selecting the reference signal point.

  According to a fourth aspect of the present invention, there is provided a method of estimating a signal point according to the second aspect, wherein two or more reference signal points are selected in the step of specifying the reference signal point in the reverse direction.

  The invention according to claim 5 receives a plurality of transmission signals corresponding to different transmission sequences transmitted simultaneously from each of a plurality of transmission antennas, and corresponds to a plurality of bit strings in the signal space of each of the plurality of transmission signals. In the receiving device that estimates each signal point of the plurality of transmission signals from among the plurality of signal points attached, a plurality of reception antennas that receive the plurality of transmission signals, and each reception from the plurality of reception antennas A detection unit for estimating each signal point of the plurality of transmission signals based on a signal, and a tree structure for estimating the signal points of the plurality of transmission antennas based on the quality of the received transmission signal A first signal point search step for associating the plurality of layers with the plurality of transmission antennas, and likelihood information on the plurality of bits associated with the signal point A first selection step of selecting a signal point candidate in a layer of the plurality of layers in which the current signal point candidate is selected, and a layer selected in the layer of selecting the current signal point candidate. For each of the signal point candidates, the signal point candidates including the signal point candidates that have been selected below the highest layer are selected. A second selection step for selecting a reference signal point having a maximum partial likelihood in the lower layer based on the cumulative likelihood information for the signal point; and for the reference signal point in the lower layer A third selection step of selecting a signal point candidate that minimizes the squared Euclidean distance, and from the first selection step to the third selection step. A second signal point search step comprising the steps of repeatedly executing the next lower layer of each layer until a lower layer signal point candidate is selected; and the second signal point search step Based on a list of signal point candidate combinations determined by the tree structure obtained from the logarithmic likelihood for each of the bit strings corresponding to the signal points of the plurality of antennas from the signal point combinations in the list A signal point detector that implements a method comprising: calculating a ratio; and estimating a most likely signal point combination for the plurality of transmission signals based on the log likelihood ratio; and the signal point detection And a decoding unit that decodes the plurality of bit strings of the plurality of transmission signals based on the log likelihood ratio information output from the unit. This is a receiver.

  The invention according to claim 6 is the receiving apparatus according to claim 5, wherein the method implemented in the signal point detector is configured to estimate the most likely signal combination following the step of estimating the most likely signal point combination. Tracing the tree structure from the lowest layer signal point of the combination of points to the highest layer, specifying the reference signal point in the reverse direction, and the reference in the reverse direction A step of selecting a signal point candidate having a minimum square Euclidean distance, a step of specifying the reference signal point in the reverse direction and a step of selecting the signal point are sequentially repeated until reaching the highest layer. A backward signal point search step comprising: a signal point candidate newly added by the backward signal point search step Re-calculating the log-likelihood ratio based on the new list added to the combination of signal point candidates obtained by the tree structure, and based on the re-calculated log-likelihood ratio, The step of estimating the most probable combination of signal points for the plurality of transmission signals is further performed.

  According to a seventh aspect of the present invention, in a receiving apparatus that receives, at each of a plurality of receiving antennas, a plurality of transmission signals corresponding to different transmission sequences transmitted simultaneously from each of the plurality of transmitting antennas, In the method of estimating each signal point of the plurality of transmission signals from among a plurality of signal points associated with a plurality of bit strings in each signal space, the plurality of the plurality of signal points are based on the quality of the received transmission signal. A signal point search step for associating a plurality of layers in a tree structure for estimating the signal point of the transmission antenna with the plurality of transmission antennas, and likelihood information regarding the plurality of bits associated with the signal point A forward signal point search step for selecting signal point candidates in each of the plurality of layers, and the forward signal point search Determining a list of signal point candidate combinations determined by the tree structure obtained from the step, and using the signal point of the lowest layer of the most probable signal point combinations in the list as a base point As described above, a reverse signal point search step for traversing the tree structure from the lowest layer to the highest layer and selecting signal point candidates in the reverse direction, and a signal newly added by the reverse signal point search step The combination of point candidates is replaced with the combination of signal point candidates from the tree structure obtained in the forward signal point search step, or from the tree structure obtained in the previous forward signal point search step. In addition to the signal point candidate combination, a step of calculating a log likelihood ratio based on the updated list, and based on the calculated log likelihood ratio Te is the estimated method characterized by comprising the step of estimating a combination of the most likely signal point for said plurality of transmission signals.

  The invention of claim 8 is a method of estimating a signal point of claim 7, wherein the backward signal point searching step includes the step of searching the lowest layer of the most probable signal point combinations in the list. Tracing the tree structure from the lowest layer to the highest layer using the signal point as a base point, specifying a reference signal point in the reverse direction, and a square Euclidean distance with respect to the reference signal point in the reverse direction Selecting the smallest signal point candidate, and identifying the reference signal point in the reverse direction and repeating the selecting step until reaching the highest layer in order.

  The invention of claim 9 receives a plurality of transmission signals corresponding to different transmission sequences transmitted simultaneously from each of a plurality of transmission antennas, and corresponds to a plurality of bit strings in the signal space of each of the plurality of transmission signals. In the receiving device that estimates each signal point of the plurality of transmission signals from among the plurality of signal points attached, a plurality of reception antennas that receive the plurality of transmission signals, and each reception from the plurality of reception antennas A detection unit for estimating each signal point of the plurality of transmission signals based on a signal, and a tree structure for estimating the signal points of the plurality of transmission antennas based on the quality of the received transmission signal A signal point search step for associating the plurality of layers with the plurality of transmission antennas, and likelihood information regarding the plurality of bits associated with the signal point, A forward signal point search step for selecting signal point candidates in each of the plurality of layers, and a list of signal point candidate combinations determined by the tree structure obtained from the forward signal point search step And, following the tree structure from the lowest layer to the highest layer, using the signal point of the lowest layer among the most probable signal point combinations in the list, The tree structure obtained in the forward signal point search step is a combination of a reverse signal point search step for selecting a signal point candidate in the direction and a signal point candidate newly added by the reverse signal point search step. Or a combination of signal point candidates from the tree structure obtained in the previous forward signal point search step. In addition, a step of calculating a log likelihood ratio based on the updated list, and a most probable combination of signal points for the plurality of transmission signals based on the calculated log likelihood ratio. A signal point detector that implements a method including an estimating step, and a decoder that decodes the plurality of bit strings of the plurality of transmission signals based on the log likelihood ratio information output from the signal point detector. It is provided with the receiving apparatus characterized by the above-mentioned.

  A tenth aspect of the present invention is the receiving apparatus according to the ninth aspect, wherein the backward signal point searching step in the method implemented in the signal point detection unit is the most probable signal point in the list. A step of identifying the reference signal point in the reverse direction by tracing the tree structure from the signal point of the lowest layer in the combination to the highest layer from the lowest layer, and the reference signal in the reverse direction Selecting a signal point candidate having a minimum square Euclidean distance for a point, repeating the step of identifying the reference signal point in the reverse direction and the step of selecting until the highest layer is reached in sequence It is characterized by including.

  As described above, the signal separation method of the present invention (method for estimating the signal point of the transmission signal) and the receiving apparatus achieve a trade-off between the reduction effect of the calculation amount and the decoding characteristic at a better level. Can do.

FIG. 1 is a diagram showing an outline of the configuration of a MIMO system. FIG. 2 is a diagram showing an outline of a more detailed configuration of the MIMO communication system. FIG. 3 is a diagram for explaining signal points in the multilevel modulation method. FIG. 4 is a diagram for explaining a received signal in one antenna of the MIMO system. FIG. 5 is a block diagram of a configuration of a MIMO detection unit that performs signal separation including the conventional PCA method. FIG. 6 is a diagram for explaining the outline of the FSD method. FIG. 7 is a diagram for explaining addition of signal point candidates in the conventional PCA method. FIG. 8 is a flowchart showing an outline of signal separation by the conventional PCA method. FIG. 9 is a first flowchart for explaining signal point addition in the conventional PCA method. FIG. 10 is a second flowchart for explaining signal point addition in the conventional PCA method. FIG. 11 is a diagram showing the positions of signal points added in the conventional PCA method according to the area of the reference signal points. FIG. 12 is a diagram showing the selection position of the NN node in the PCA-NN method of the present invention in comparison with the prior art. FIG. 13 is a flowchart for explaining signal point addition in the PCA-NN method of the present invention. FIG. 14 is a diagram for explaining additional signal points when the PCA-NN method of the present invention is applied to a 64QAM modulation system. FIG. 15 is a diagram showing the selection position of the NN node in the PCA-NN method of the present invention in comparison with the prior art. FIG. 16 is a diagram for explaining a signal point addition method by the extended PCA method, which is a modification of the PCA-NN method of the present invention. FIG. 17 is a block diagram of a configuration of a MIMO detection unit using the BCA method of the present invention. FIG. 18 is a flowchart for explaining the addition of signal points by the BCA method of the present invention. FIG. 19 is a diagram for explaining addition of signal points by the BCA method of the present invention. FIG. 20 is a diagram showing a table in which the number of search nodes and the number of lists are compared between the present invention and the related art for 16QAM. FIG. 21 is a diagram showing a table in which the number of search nodes and the number of lists are compared between the present invention and the prior art for 64QAM. FIG. 22 is a diagram showing a BER characteristic (16QAM) according to the PCA-NN method of the present invention using an LDPC code. FIG. 23 is a diagram showing a BER characteristic (64QAM) according to the PCA-NN method of the present invention using an LDPC code. FIG. 24 is a diagram showing a BER characteristic (16QAM) according to the E-PCA method of the present invention using an LDPC code. FIG. 25 is a diagram showing a comparison of BER characteristics (16QAM) by each algorithm using an LDPC code. FIG. 26 is a diagram showing a comparison of BER characteristics (64QAM) by each algorithm using an LDPC code.

  In the signal separation using the PCA method of the encoded MIMO system, the present invention adds a signal point having the nearest square Euclidean distance to a signal point (reference signal point) searched by the FSD method, thereby decoding characteristics. To improve. Further, the previous signal is obtained by a new additional signal point candidate obtained by further traversing the tree structure in the reverse direction from the lowest node of the best estimated signal point set obtained from the set of signal point candidates for which the search has been completed. By replacing the point candidates or adding them to the previous signal point candidates, the trade-off between the reduction effect of the calculation amount and the decoding characteristics is achieved at a better level.

First Embodiment of the Present Invention The signal separation method of the present invention improves the method of adding signal point candidates in the PCA method. The signal point addition block 204 in FIG. 5 or the core of S108 shown in FIG.・ It is about process improvement. In the present invention, a nearest node (hereinafter referred to as NN node) is introduced as a signal point (additional node) to be added to the reference signal point. As additional signal point candidates used in the PCA method, signal point candidates obtained by changing one bit from the bit string corresponding to the child node (reference signal point) having the largest partial likelihood are added. In order to select a signal point corresponding to a bit string in which only one bit called a gray code is changed, no additional calculation or the like is required, and the signal point adding method is simple.

  On the other hand, in the present invention, signal points corresponding to the Gray code are not selected, but signal point candidates and NN nodes that minimize the squared Euclidean distance are added to each bit. Hereinafter, the PCA method using the NN node according to the present invention is referred to as a PCA-NN method.

  FIG. 12 is a diagram showing the selection position of the NN node in the PCA-NN method of the present invention in comparison with the prior art. 12A shows the case where the reference signal point is in the area 500 shown in FIG. 11, FIG. 12B shows the case where the reference signal point is in the area 501 shown in FIG. 11, and FIG. A case where there is a reference signal point in the indicated region 502 is shown. In either case, the modulation method is 16QAM, and is shown in contrast to the additional signal points in the conventional PCA method. For example, referring to (a), signal points 601-3 and 601-4 that are greatly separated from the reference signal point 600 are added in the prior art. On the other hand, when an NN node is selected by the PCA-NN method of the present invention, signal points 602-3 and 602-4 closer to the reference signal point are selected. The same applies to (b).

  Instead of adding signal points in the conventional PCA method, in the PCA-NN method of the present invention, a signal point closer to the reference signal point is added by adding an NN node. This is more rational than adding a signal point far away from the selected reference signal point as a signal point candidate, and signal separation accuracy is expected. This effect will be described later. In the PCA-NN method, which is the signal separation method of the present invention, only the signal point candidate selection method to be added is changed as compared with the PCA method. For this reason, only a part of the first flowchart for explaining signal point addition in the conventional PCA method shown in FIG. 9 is changed.

FIG. 13 is a first flowchart for explaining signal point addition in the PCA-NN method of the present invention. Compared to the prior art PCA method shown in FIG. 8, in S507, for the most part the likelihood of a large parent node (Layer1), in addition to the signal point _{A 1,} the signal point _{C 1} of the minimum squared Euclidean distance In addition to the signal point A ₂ in addition to the signal point A ₂ , the signal point C ₂ having the least square Euclidean distance is added to the parent node (Layer 2) having the largest partial likelihood in S515. It is different from PCA method.

  FIG. 14 is a diagram for explaining additional signal points when the PCA-NN method of the present invention is applied to a 64QAM modulation system. FIG. 14 shows three regions 700, 701, and 702 that can be taken by reference signal points serving as a reference for added signal points on a 64QAM signal space diagram.

  FIG. 15 is a diagram showing the selection position of the NN node in the PCA-NN method of the present invention in comparison with the prior art. 15A shows the case where the reference signal point is in the area 700 shown in FIG. 14, FIG. 15B shows the case where the reference signal point is in the area 701 shown in FIG. 14, and FIG. A case where there is a reference signal point in the indicated region 702 is shown. In either case, the modulation method is 16QAM, and is shown in contrast to the additional signal points in the conventional PCA method. For example, referring to the case where the reference signal points in (a) are at the four corners of the signal space, the signal points 705-3, 705-4, and 705 that are far apart from the reference signal point 704 in the PCA method of the prior art. −5, 705-6 were added. On the other hand, when an NN node is selected by the PCA-NN method of the present invention, signal points 706-3, 706-4, 706-5, and 706-6 closer to the reference signal point are selected. The same applies to the case where the reference signal point (b) is in the peripheral portion excluding the four corners of the signal space diagram. Further, even when the reference signal point (c) is in the inner region, the signal point 712-6 closer to the reference signal point 710 is selected instead of the signal point 711-6 according to the conventional technique.

  Therefore, the PCA-NN method of this embodiment can be described as the following method. That is, in a receiving apparatus that receives, at each of a plurality of receiving antennas, a plurality of transmission signals corresponding to different transmission sequences transmitted simultaneously from each of a plurality of transmitting antennas, in each signal space of the plurality of transmitting signals The method for estimating each signal point of the plurality of transmission signals from the plurality of signal points associated with the plurality of bit strings includes the following steps.

That is, a first signal point search step for associating a plurality of layers in a tree structure for estimating the signal points of the plurality of transmission antennas with the plurality of transmission antennas based on the received quality of the transmission signal And selecting a signal point candidate in a layer in which the current signal point candidate is selected among the plurality of layers based on likelihood information regarding the plurality of bits associated with the signal point. For each of the signal point candidates selected in the selection step and the layer where the current signal point candidate is selected, the signal point candidates including the signal point candidates already selected in the uppermost layer or lower are selected. Condition, the lower signal level is based on the cumulative likelihood information for signal points in a layer lower than the layer in which the current signal point candidate is selected. A second selection step of selecting a reference signal point that maximizes the partial likelihood, and a signal point candidate that minimizes the squared Euclidean distance with respect to the reference signal point in the lower layer. Repeat the third selection step and the first selection step to the third selection step for the next lower layer of each layer until the signal point candidate of the lowest layer is selected. Based on a list of combinations of signal point candidates determined by the tree structure obtained from the second signal point search step and the second signal point search step. Calculating a log-likelihood ratio for each of the bit strings corresponding to the signal points of the plurality of antennas from the signal point combinations of Based on the estimated method characterized by comprising the steps of estimating a combination of the most likely signal point for said plurality of transmission signals.

  The quality of the transmission signal is not limited to this. For example, the S / N ratio of the transmission signal received by the reception antenna can be used.

  Further, the PCA-NN method of the present embodiment receives a plurality of transmission signals corresponding to different transmission sequences transmitted simultaneously from each of a plurality of transmission antennas, and in each signal space of the plurality of transmission signals, The present invention can also be applied to a receiving apparatus that estimates signal points of the plurality of transmission signals from a plurality of signal points associated with a plurality of bit strings.

  The receiving apparatus includes a plurality of reception antennas that receive the plurality of transmission signals, and a detection unit that estimates signal points of the plurality of transmission signals based on the reception signals from the plurality of reception antennas.

  The detection unit is configured to associate a plurality of layers in a tree structure for estimating the signal points of the plurality of transmission antennas with the plurality of transmission antennas based on the received quality of the transmission signals. Based on the point search step and the likelihood information regarding the plurality of bits associated with the signal point, the signal point candidate is selected in the layer where the current signal point candidate is selected from among the plurality of layers. For each of the signal point candidates selected in the first selection step and the layer in which the current signal point candidate is selected, the signal point candidates including the signal point candidates already selected below the highest layer are selected. On the basis of the cumulative likelihood information for signal points in a layer lower than the layer in which the current signal point candidate is selected. A second selection step of selecting a reference signal point having the maximum partial likelihood in the layer, and a signal point candidate having a minimum square Euclidean distance with respect to the reference signal point in the lower layer; Repeat the third selection step and the first selection step to the third selection step for the next lower layer of each layer until the signal point candidate of the lowest layer is selected. Based on a list of combinations of signal point candidates determined by the tree structure obtained from the second signal point search step and the second signal point search step. Calculating a log-likelihood ratio for each of the corresponding bit strings of the signal points of the plurality of antennas from the signal point combinations of Based on the degree ratio, implementing a method comprising the steps of estimating a combination of the most likely signal point for said plurality of transmission signals.

  The receiving apparatus further includes a decoding unit that decodes the plurality of bit strings of the plurality of transmission signals based on the log likelihood ratio information output from the signal point detection unit.

  Needless to say, the receiving device includes a transmitting / receiving device having a function of a transmitting device, a wireless communication terminal device, a base station device, and the like.

  As described above, in the PCA-NN method of the present invention, by adding a signal point having a least square Euclidean distance to a reference signal point having a maximum partial likelihood in the tree structure search layer, The signal points closer to each other are added as signal point candidates. The difference between the selected additional signal point and the prior art becomes more prominent as the modulation method has a larger number of signal points. The addition of the signal point candidates by the PCA-NN method of the present invention can be further modified by combining another selection method of additional signal candidate points. As an example, a method for increasing the number of reference signal points as compared with the first embodiment will be described below.

Second Embodiment of the Present Invention FIG. 16 is a diagram for explaining a signal point adding method by the extended PCA method obtained by modifying the PCA-NN method of the present invention. A tree structure 800 determined as a signal point candidate by the extended PCA method is shown. The outline of the extended PCA method will be described below.

  In the conventional PCA method, only one parent node (reference signal point) having the highest partial likelihood is selected, and signal point candidates are added. On the other hand, in the E-PCA method, signal point candidates are added in consideration of one or more upper P pieces having a large partial likelihood. In the E-PCA method, all nodes can be searched in parallel. The algorithm of the E-PCA method is shown below.

When FE: Similar to the FSD method, all signal point candidates are searched.

When SE 2, a child node is selected for each of the top P nodes (signal points) having a large partial likelihood of the parent node, and a node (reference signal point) having the largest partial likelihood among the child nodes. Explore. Next, an NN node is added to each of the selected P child nodes. For the parent node that has not been selected, only the child node having the largest partial likelihood is selected as in the FSD method.

Returning to FIG. 16 again, a signal point adding method by the E-PCA method will be described more specifically. FIG. 16 shows the case where the number of extended reference signal points P = 2, the modulation scheme is 4QAM, and the number of transmission antennas N _t = 4. In the E-PCA method, first, two reference signal points surrounded by circles (P = 2) are selected in Layer 1. In the conventional PCA, for example, only the signal point 801 (bit string 00) is selected. In the E-PCA method, a signal point 802 (bit string 10) is further added.

  In Layer 2, a signal point having a large partial likelihood (a square Euclidean distance is small) with respect to the signal point of Layer 1 is selected. Further, an NN node is added from each of the signal points 801 and 802 selected as the reference signal points in Layer1. That is, signal points 803 and 804 of Layer 2 from the reference signal point 801 are selected as NN nodes. Similarly, signal points 805 and 806 of Layer 2 from the reference signal point 802 are selected as NN nodes. In Layer 2, two signal points having the largest partial likelihood (small square Euclidean distance) including the added four signal points are selected. That is, the signal points 807 and 808 are the reference signal points of Layer 2 this time. An NN node is added from each of the reference signal points 807 and 808. That is, signal points 809 and 810 of Layer 3 from the reference signal point 807 are selected as NN nodes. Similarly, Layer 3 signal points 811 and 812 are selected as the NN node from the reference signal point 808. Thereafter, all signal point candidates are selected in the same manner, and a signal point set is determined.

  Therefore, the invention of this embodiment is characterized in that two or more reference signal points are selected in the second selection step of selecting a reference signal point in the signal point estimation method according to the first embodiment. In the receiving device according to the first embodiment, in the step of specifying the reference signal point in the reverse direction, two or more reference signal points are selected.

The number of search nodes μ _EPCA in the E-PCA method is expressed by the following equation.

The list number N _EPCA is expressed by the following equation.

  In the E-PCA method, a total of 16 “signal point combinations” are selected as shown in FIG. That is, the signal point set list includes 16 sets of “signal point combinations”, and LLR values are calculated for these signal point sets. A signal point set including signal points 819 to 824 surrounded by a dotted line is selected by the E-PCA method. According to the E-PCA method, although the amount of calculation increases as the number of signal points increases, the decoding characteristics can be improved.

  The basic idea of the E-PCA method is applicable to the PCA method of the prior art. By combining the PCA-NN method and the E-PCA method of the present invention described as the first embodiment, signal separation that achieves a trade-off between the reduction effect of calculation amount and the error rate characteristic at a better level is achieved. realizable.

★ Third Embodiment of the Present Invention The signal point candidate adding method by the PCA-NN method of the present invention can be combined with another additional signal point selecting method. Here, a backward candidate adding method (Backward Candidate Adding: BCA method) algorithm will be described. The BCA method is not only combined with the PCA-NN method of the present invention, but also widely applicable to other general signal point search methods. The BCA method is a method of re-searching signal point candidates based on the search result after performing a signal search in a tree structure. The NN node in the first embodiment can be used in the re-search.

  FIG. 17 is a block diagram of a configuration of a MIMO detection unit using the BCA method of the present invention. The basic configuration of the MIMO detection unit 1000 using the BCA method is almost the same as that of the MIMO detection unit 200 using the PCA method shown in FIG. Therefore, only the differences will be described. A MIMO detection unit 1000 using the BCA method includes a BCA processing unit 1009. As will be described later, the BCA method is applied after completing the other signal point search methods implemented in the FSD processing block 1003 and the signal point candidate addition block 1004 and determining an optimal signal point set.

  FIG. 18 is a flowchart for explaining the addition of signal points by the BCA method of the present invention. The overall signal separation flow by the BCA method is also substantially the same as the signal separation flow by the PCA method shown in FIG. The pre-process of S106 in FIG. 8 is common, and the core process of S108 corresponds to the search for signal point candidates in the forward direction of S120 in FIG. Here, “forward direction” means that signal point candidates are searched from Layer 1 to Layer 4 in the tree structure 400 of the PCA method shown in FIG. 7, for example. In the BCA method, after the signal point search in the forward direction is finished and the most powerful signal point set is determined from the first signal point set obtained as a result of the search in the forward direction (S122), The point is characterized in that a new signal point search is performed in the reverse direction using the lowest signal point of this most powerful signal point set as a base point (S124). The second signal point set obtained as a result of the signal point search in the reverse direction is replaced with the first signal set as a new signal point set, or added to the first signal set, and finally A search list is determined (S110).

Here, an example in which the BCA method is applied to the FSD method will be described as a representative example. In the FE in the FSD method or the PCA method, each bit likelihood of a signal point is always generated, and therefore no re-search is performed. The algorithm of the BCA method is shown below.
STEP1 (Search in FSD method, corresponding to S106 in FIG. 18)
Similar to the FSD method, all signal point candidates are searched.
STEP2 (Research for search candidates by BCA method)
At each signal point obtained from the signal point search of the FSD method, the NN node is searched again in the reverse direction.

  Next, the addition of signal point candidates in the BCA method will be described more specifically using a tree structure.

FIG. 19 is a diagram for explaining addition of signal points by the BCA method of the present invention. The tree structure 900 shown in FIG. 19 is obtained as a result of further applying the BCA method after applying the FSD method. Therefore, the signal point on the right half of the tree structure 900 is the same as that of the tree structure 302 shown in FIG. On the other hand, the left half signal point in the tree structure 900 is obtained as a result of applying the BCA method with the signal point 901 as a base point. FIG. 19 shows a signal point search range by the BCA method when the modulation scheme is 4QAM and the number of transmission antennas N _t = 4. Nodes surrounded by dotted lines are nodes re-searched by the BCA method. A circled signal point 901 is a signal point set having the highest likelihood finally obtained by the forward FSD method. That is, the signal point set composed of the signal points 901, 2902, 905, and 910 is the most probable and most powerful signal point set among the signal point sets determined by the FSD method. In the BCA method, this most powerful signal point set is searched again while traversing in the reverse direction from the lowest layer to the highest layer.

  Specifically, the tree structure is returned upward to Layer 3 from the signal point 901, and NN nodes 903 and 904 are added to the signal point 902. At this time, the NN nodes 903 and 904 are signal points that are at least square Euclidean distance from the signal point 901.

  Next, the tree structure is further returned upward to Layer 2 from the signal point 902 of Layer 3, and NN nodes 906 and 907 are added to the signal point 905. The NN nodes 906 and 907 at this time are signal points that are at least square Euclidean distance from the signal point 902. At this time, the bit string of the signal point 908 of Layer 4 below the NN node 906 is set to the bit string 00 of the signal point 901 serving as the base point. Therefore, the bit string of the signal point 908 is 00/00 / ** / 00. The same applies to the signal point 909 of Layer 4 below the NN node 907.

  Further, the tree structure is returned upward to Layer 1 from the signal point 905 of Layer 2, and NN nodes 911 and 912 are added to the signal point 910. The NN nodes 911 and 912 at this time are signal points that are at least square Euclidean distance from the signal point 905. At this time, the bit string of the signal points of Layer 3 / Layer 4 below the NN node 911 is the bit string 00/00 of the signal points 901 and 902 serving as the base points. Therefore, the bit string of the Layer 3 signal point below the signal point 911 is 00 / ** / 00, and the bit string of the Layer 4 signal point is 00 / ** / 00/00. The same applies to the two signal points of Layer 3 / Layer 4 below the NN node 912.

  Therefore, the BCA method of this embodiment can be described as the following method. That is, in a receiving apparatus that receives a plurality of transmission signals corresponding to different transmission sequences transmitted from each of a plurality of transmission antennas at each of a plurality of reception antennas, in each signal space of the plurality of transmission signals, It can be described as a method of estimating each signal point of the plurality of transmission signals from a plurality of signal points associated with a plurality of bit strings.

  The method includes a signal point search step for associating a plurality of layers in a tree structure for estimating the signal points of the plurality of transmission antennas with the plurality of transmission antennas based on the received quality of the transmission signals. A forward signal point search step of selecting a signal point candidate in each of the plurality of layers based on likelihood information relating to the plurality of bits associated with the signal point; and the forward signal point search Determining a list of signal point candidate combinations determined by the tree structure obtained from the step, and using the signal point of the lowest layer in the most probable signal point combinations in the list as a base point As shown below, the tree structure is traced from the lowest layer to the highest layer, and the reverse signal point search is performed to select signal point candidates in the reverse direction. A signal point candidate newly added by the step and the backward signal point search step, instead of the signal point candidate combination from the tree structure obtained in the forward signal point search step, or In addition to the signal point candidate combination from the tree structure obtained in the previous forward signal point search step, a step of calculating a log likelihood ratio based on the updated list, and the calculated logarithm Estimating the most probable combination of signal points for the plurality of transmission signals based on a likelihood ratio.

  Here, when using the NN node, the backward signal point searching step uses the signal point of the lowest layer among the most probable signal point combinations in the list as a base point, Tracing the tree structure from the top to the top layer, specifying a reference signal point in the reverse direction, and selecting a signal point candidate having a minimum square Euclidean distance with respect to the reference signal point in the reverse direction And the step of identifying the reference signal point in the reverse direction and the step of selecting are sequentially repeated until reaching the highest layer.

  Further, the BCA method of the present embodiment receives a plurality of transmission signals corresponding to different transmission sequences transmitted simultaneously from each of a plurality of transmission antennas, and a plurality of transmission signals in each signal space of the plurality of transmission signals are received. The present invention can be applied to a receiving apparatus that estimates each signal point of the plurality of transmission signals from a plurality of signal points associated with a bit string.

  The receiving apparatus includes a plurality of reception antennas that receive the plurality of transmission signals, and a detection unit that estimates signal points of the plurality of transmission signals based on the reception signals from the plurality of reception antennas.

  The reception unit is a signal point search step for associating a plurality of layers in a tree structure for estimating the signal points of the plurality of transmission antennas and the plurality of transmission antennas based on the received quality of the transmission signals. A forward signal point search step of selecting a signal point candidate in each of the plurality of layers based on likelihood information regarding the plurality of bits associated with the signal point; and the forward signal point Determining a list of signal point candidate combinations determined by the tree structure obtained from the search step; and selecting signal points of the lowest layer in the most probable signal point combinations in the list As a base point, follow the tree structure from the lowest layer to the highest layer and select a signal point candidate in the reverse direction. A search step and a combination of signal point candidates newly added by the backward signal point search step, instead of a combination of signal point candidates from the tree structure obtained in the forward signal point search step, or Adding to the combination of signal point candidates from the tree structure obtained in the previous forward signal point search step and calculating a log likelihood ratio based on the updated list; and A method is implemented that includes estimating a most likely combination of signal points for the plurality of transmission signals based on a log likelihood ratio.

  The receiving apparatus further includes a decoding unit that decodes the plurality of bit strings of the plurality of transmission signals based on the log likelihood ratio information output from the signal point detection unit. Needless to say, the receiving device includes a transmitting / receiving device having a function of a transmitting device, a wireless communication terminal device, a base station device, and the like.

  In the BCA method, signal point candidates are newly added with the signal point of the lowest layer of the most powerful signal point set as a base point. For this reason, signal point candidates are added based on the most probable set of signal points. Therefore, by further applying the BCA method to the first signal point candidate set obtained by the FSD method and adding the second signal point set obtained by the BCA method of adding the NN node as described above, The LLR value is calculated while maintaining the same number of search nodes and lists as in the PCA method.

The number of search nodes μ _BCA is expressed by the following equation.

The list number N _BCA is expressed by the following equation.

  FIG. 19 shows an example in which re-search is performed in the reverse direction while adding NN nodes. However, the essence of the BCA method is based on the most probable set of signal points obtained by the forward signal point search. The point is to search again in the direction. Therefore, in the backward search, signal point candidates can be added by the PCA method that does not use the NN node.

★ Fourth Embodiment of the Present Invention As described with reference to FIG. 17, the BCA method of the present invention described above can be combined with all signal point search methods, and can also be combined with the PCA method. Next, a PCA-BCA algorithm using both the PCA method and the BCA method will be described. In the PCA method, signal point candidates that are re-searched by the BCA method are not necessarily searched. That is, the first signal point set determined by the PCA method for searching forward signal points and the second signal point set determined by the BCA method for re-searching signal points in the reverse direction are: Does not necessarily match. Therefore, the decoding characteristics can be improved by using both the PCA method and the BCA method. In the PCA-BCA algorithm, the NN node described in the first embodiment of the present invention is used instead of a signal obtained by changing one bit as a candidate for an additional signal point in the PCA method. This is because in order to calculate an accurate LLR value, the calculation must be performed on signal point candidates that are closer to the actually received signal points. The algorithm of the PCA-BCA method is shown below.
STEP1 (Search by PCA method)
Search for PCA method. At this time, the additional signal point candidate is an NN node, not a candidate obtained by changing one bit as in the prior art.
STEP2 (Search by BCA method)
After the search for the PCA method is completed, signal point candidates are searched again in the reverse direction by the BCA method.
The number of search nodes μ _PCA-BCA is expressed by the following equation.

The list number N _PCA-BCA is expressed by the following equation.

  As problems of the PCA-BCA method, there is a possibility that search nodes of the PCA method and the BCA method overlap, and a search delay occurs because the search of the PCA method and the BCA method cannot be performed simultaneously. There can be. However, the first signal point set and the second signal point set can be combined, and diversity can be added to the method of determining the final list (S110 in FIG. 18). Combining the first signal point set and the second signal point set to determine the final list of signal points can be done in various ways. That is, in the list determination block 1006 of FIG. 17, a signal point set can be selected and combined from the first signal point set and the second signal point set on a predetermined basis. By using two kinds of signal point sets, various means can be provided in order to balance the trade-off between the calculation amount and the decoding performance.

  The table below shows a comparison between the number of search nodes and the number of lists for the signal point addition method (search method) of the prior art and the signal point addition method (search method) of the present invention described so far. As already described, the number of search nodes is the number of all nodes included in the tree structure formed by each signal point addition method, that is, the number of signal points. The number of lists is the number of signal point set candidates included in the tree structure formed by each signal point addition method. The signal point set means a “signal point combination” of signals transmitted from all the transmitting antennas.

  FIG. 20 is a diagram showing a table comparing the number of search nodes and the number of lists between each method of the present invention and the prior art method for 16QAM. A comparison of the number of search nodes and the number of lists when the number of transmitting antennas Nt = 4 and the number of modulation signal points M = 16 is shown. In the table of FIG. 20, NN node signal points specific to the present invention are added in any of the E-PCA method, BCA method, and PCA-BCA method. Although the E-PCA method and the PCA-BCA method, which are signal point addition methods of the present invention, have a larger number of search nodes than the PCA method of the prior art, the number of search nodes and the list when compared with the QRM method and LFSD method Any of the numbers can be reduced. Moreover, when compared with the MLD method, the number of search nodes can be reduced by 99% or more.

  FIG. 21 is a diagram showing a table in which the number of search nodes and the number of lists are compared between each method of the present invention and the prior art method for 64QAM. A comparison of the number of search nodes and the number of lists when the number of transmission antennas Nt = 4 and the number of modulation signal points M = 64 is shown. The E-PCA method and the PCA-BCA method of the present invention increase the number of search nodes and the number of lists as compared with the conventional PCA method. However, the number of search nodes can be reduced as compared with the QRM method and the LFSD method.

The proposed signal separation method was evaluated by computer simulation. Assume that the channel matrix fading coefficients h _{i, j} , i = 1,... N _r , j = 1,..., N _t are all independent and follow a complex Gaussian distribution with mean 0 and variance 1. The number of transmission / reception antennas is 4, and the number of subcarriers for OFDM modulation is 64. Also, an LDPC code having a code length N = 3072, a coding rate r _c = 1/2, a sum-product decoding repetition count T = 40, and a clipping value L _max = 8 was used.

  FIG. 22 is a diagram showing a BER characteristic (16QAM) by the PCA-NN method of the present invention using an LDPC code. The PCA method of the prior art and the PCA-NN method of the present invention showed almost the same BER characteristics. This is because, in the case of 16QAM, the signal point candidate difference is small between the PCA method of the prior art and the addition by the NN node of the present invention.

FIG. 23 is a diagram showing a BER characteristic (64QAM) by the PCA-NN method of the present invention using an LDPC code. When the E _b / N _o value is larger than 11 dB, the PCA-NN method of the present invention shows more excellent characteristics than the PCA method of the prior art. This is because when the modulation method is 64QAM, the difference between the additional signal point of the prior art and the additional signal point of the present invention becomes more conspicuous compared to the case of 16QAM.

  FIG. 24 is a diagram showing a BER characteristic (16QAM) according to the E-PCA method of the present invention using an LDPC code. The BER characteristics are improved as P increases. This is because the number of search nodes is increasing. When P = 3 or more, there is no difference in BER characteristics.

FIG. 25 is a diagram showing a comparison of BER characteristics (16QAM) by each algorithm using an LDPC code. The number of surviving nodes in the QRM method = 16, the search node n in the LFSD method = (1, 2, 2, 16) / 64, and the number of parent nodes P = 2 in the E-PCA method. The LFSD method, PCA method, E-PCA method and PCA-BCA method of the present invention show similar characteristics. Among them, the E-PCA method of the present invention exhibits the most excellent characteristics and has the least amount of calculation. There is a difference of about 0.28 dB compared to the MLD method at BER = 10 ⁻⁵ .

  FIG. 26 is a diagram showing a comparison of the BER characteristics (64QAM) by each algorithm using the LDPC code = 30, n = (1, 2, 2, 64) / 64, and P = 2. The LFSD method and the E-PCA method of the present invention show very close characteristics. The E-PCA method of the present invention can reduce the number of search nodes by about 53% compared to the LFSD method.

  By using the likelihood estimation method according to the present invention in combination with the prior art likelihood estimation method, it is possible to perform likelihood estimation that realizes a better trade-off between the reduction effect of the calculation amount and the error rate characteristic. Thereby, since the estimated likelihood approaches the disappearance likelihood in the MLD method, the characteristics can be improved. From the simulation results, it was also confirmed that the characteristics close to the MLD method can be achieved with a smaller increase in the amount of calculation compared with the conventional method than the likelihood estimation method of the prior art.

  As described above in detail, in the signal separation using the PCA method of the encoded MIMO system, the present invention is a signal having the nearest square Euclidean distance to the signal point (reference signal point) searched in each layer. By adding a point (NN node), decoding characteristics can be improved. Further, instead of selecting only one parent node (reference signal point) having the highest partial likelihood, signal point candidates are considered by the E-PCA method in consideration of one or more upper P pieces having a large partial likelihood. The decoding characteristics can be further improved by increasing the signal points by adding, or by further combining with the addition of NN nodes.

  Further, new signal point candidates are obtained by further traversing the tree structure in the reverse direction while adding NN nodes from the lowest node of the best estimated signal point set obtained from the signal point set for which the search has been completed. By replacing the previously added signal point candidate with this new signal point candidate, or by adding it to the previously added signal point candidate, the trade-off between the reduction effect of the amount of computation and the error rate characteristic can be further increased. Achieve at a good level. By applying the signal separation method of the present invention to a receiving apparatus, it is possible to provide an apparatus that realizes a better trade-off between the calculation amount reduction effect and the error rate characteristics than the conventional technique.

  The present invention is generally applicable to a wireless communication system. In particular, it can be used for a MIMO communication system.

DESCRIPTION OF SYMBOLS 1 MIMO communication system 2 Transmission side device 3 Reception side device 4 Propagation path 5 Transmission signal sequence 6 Reception signal sequence 10-13 Transmission antenna 102 Encoder 103 Interleaver 105 MIMO mapping unit 109, 200, 1000 MIMO detection unit 111 Decoder 204, 1004 Signal point candidate addition block 206, 1006 LLR calculation block 207, 1007 Signal point determination block

Claims (10)

In a receiving apparatus that receives a plurality of transmission signals corresponding to different transmission sequences transmitted from each of a plurality of transmission antennas at each of a plurality of reception antennas, a plurality of signals in each signal space of the plurality of transmission signals In the method for estimating each signal point of the plurality of transmission signals from among a plurality of signal points associated with the bit string of
A first signal point search step for associating a plurality of layers in a tree structure for estimating the signal points of the plurality of transmission antennas with the plurality of transmission antennas based on the received quality of the transmission signals;
A first selection step of selecting a signal point candidate in a layer in which the current signal point candidate is selected from among the plurality of layers based on likelihood information regarding the plurality of bits associated with the signal point. When,
For each of the signal point candidates selected in the layer where the current signal point candidate is selected, on the condition that the signal point candidate including the signal point candidates already selected in the uppermost layer or lower is selected. Second selection for selecting a reference signal point having a maximum partial likelihood in the lower layer based on cumulative likelihood information for signal points in a layer lower than the layer in which the current signal point candidate is selected Steps,
A third selection step of selecting a signal point candidate having a minimum square Euclidean distance with respect to the reference signal point in the lower layer;
Repeatedly executing the first selection step to the third selection step on the next lower layer of each layer until the signal point candidate of the lowest layer is selected. A second signal point search step;
Based on a list of combinations of signal point candidates determined by the tree structure obtained from the second signal point search step, corresponding signal points of the plurality of antennas are obtained from signal point combinations in the list. Calculating a log likelihood ratio for each of the bit sequences;
Estimating the most probable combination of signal points for the plurality of transmission signals based on the log-likelihood ratio. Subsequent to the step of estimating the most probable signal point combination, the tree structure is formed from the signal layer of the lowest layer in the most probable signal point combination toward the highest layer from the lowest layer. To identify the reference signal point in the opposite direction,
Selecting a signal point candidate having a minimum squared Euclidean distance with respect to the reference signal point in the reverse direction;
A reverse signal point search step comprising: identifying the reference signal point in the reverse direction and selecting the step until the top layer is sequentially reached;
The log likelihood ratio is re-established based on a new list obtained by adding the signal point candidate combination newly added by the backward signal point search step to the signal point candidate combination obtained by the tree structure. A calculating step;
The estimation method according to claim 1, further comprising: estimating a most probable combination of signal points for the plurality of transmission signals based on the recalculated log likelihood ratio.   The estimation method according to claim 1, wherein two or more reference signal points are selected in the second selection step of selecting the reference signal points.   3. The estimation method according to claim 2, wherein two or more reference signal points are selected in the step of specifying the reference signal points in the reverse direction. A plurality of transmission points corresponding to different transmission sequences, transmitted simultaneously from each of a plurality of transmission antennas, and a plurality of signal points associated with a plurality of bit strings in each signal space of the plurality of transmission signals In the receiving device for estimating each signal point of the plurality of transmission signals,
A plurality of receiving antennas for receiving the plurality of transmission signals;
A detection unit configured to estimate each signal point of the plurality of transmission signals based on each reception signal from the plurality of reception antennas;
A first signal point search step for associating a plurality of layers in a tree structure for estimating the signal points of the plurality of transmission antennas with the plurality of transmission antennas based on the received quality of the transmission signals;
A first selection step of selecting a signal point candidate in a layer in which the current signal point candidate is selected from among the plurality of layers based on likelihood information regarding the plurality of bits associated with the signal point. When,
For each of the signal point candidates selected in the layer where the current signal point candidate is selected, on the condition that the signal point candidate including the signal point candidates already selected in the uppermost layer or lower is selected. Second selection for selecting a reference signal point having a maximum partial likelihood in the lower layer based on cumulative likelihood information for signal points in a layer lower than the layer in which the current signal point candidate is selected Steps,
A third selection step of selecting a signal point candidate having a minimum square Euclidean distance with respect to the reference signal point in the lower layer;
Repeatedly executing the first selection step to the third selection step on the next lower layer of each layer until the signal point candidate of the lowest layer is selected. A second signal point search step;
Based on a list of combinations of signal point candidates determined by the tree structure obtained from the second signal point search step, corresponding signal points of the plurality of antennas are obtained from signal point combinations in the list. Calculating a log likelihood ratio for each of the bit sequences;
Estimating a most probable combination of signal points for the plurality of transmission signals based on the log-likelihood ratio;
A receiving apparatus comprising: a decoding unit that decodes the plurality of bit strings of the plurality of transmission signals based on the log likelihood ratio information output from the signal point detection unit. The method implemented in the signal point detection unit is:
Subsequent to the step of estimating the most probable signal point combination, the tree structure is formed from the signal layer of the lowest layer in the most probable signal point combination toward the highest layer from the lowest layer. To identify the reference signal point in the opposite direction,
Selecting a signal point candidate having a minimum squared Euclidean distance with respect to the reference signal point in the reverse direction;
A reverse signal point search step comprising: identifying the reference signal point in the reverse direction and selecting the step until the top layer is sequentially reached;
The log likelihood ratio is re-established based on a new list obtained by adding the signal point candidate combination newly added by the backward signal point search step to the signal point candidate combination obtained by the tree structure. A calculating step;
The receiving apparatus according to claim 5, further comprising: estimating a most probable combination of signal points for the plurality of transmission signals based on the recalculated log likelihood ratio. In a receiving apparatus that receives a plurality of transmission signals corresponding to different transmission sequences transmitted from each of a plurality of transmission antennas at each of a plurality of reception antennas, a plurality of signals in each signal space of the plurality of transmission signals In the method for estimating each signal point of the plurality of transmission signals from among a plurality of signal points associated with the bit string of
A signal point search step for associating a plurality of layers in a tree structure for estimating the signal points of the plurality of transmission antennas with the plurality of transmission antennas based on the quality of the received transmission signal;
A forward signal point search step of selecting a signal point candidate in each of the plurality of layers based on likelihood information regarding the plurality of bits associated with the signal point;
Determining a list of signal point candidate combinations determined by the tree structure obtained from the forward signal point search step;
In the list, following the tree structure from the lowest layer signal point of the most probable signal point combination to the highest layer, the signal point candidates are found in the opposite direction. A reverse signal point search step to select;
The combination of signal point candidates newly added by the backward signal point search step is replaced with the combination of signal point candidates from the tree structure obtained in the forward signal point search step, or the previous order In addition to the signal point candidate combination from the tree structure obtained in the direction signal point search step, calculating a log likelihood ratio based on the updated list;
Estimating the most probable combination of signal points for the plurality of transmission signals based on the calculated log likelihood ratio. The backward signal point search step includes:
In the list, following the tree structure from the lowest layer signal point to the highest layer in the most probable signal point combination, the reference signal point in the opposite direction is Identifying steps;
Selecting a signal point candidate having a minimum squared Euclidean distance with respect to the reference signal point in the reverse direction;
The method according to claim 7, further comprising: repeating the step of identifying the reference signal point in the reverse direction and the step of selecting until the highest layer is reached sequentially. A plurality of transmission points corresponding to different transmission sequences, transmitted simultaneously from each of a plurality of transmission antennas, and a plurality of signal points associated with a plurality of bit strings in each signal space of the plurality of transmission signals In the receiving device for estimating each signal point of the plurality of transmission signals,
A plurality of receiving antennas for receiving the plurality of transmission signals;
A detection unit configured to estimate each signal point of the plurality of transmission signals based on each reception signal from the plurality of reception antennas;
A signal point search step for associating a plurality of layers in a tree structure for estimating the signal points of the plurality of transmission antennas with the plurality of transmission antennas based on the quality of the received transmission signal;
A forward signal point search step of selecting a signal point candidate in each of the plurality of layers based on likelihood information regarding the plurality of bits associated with the signal point;
Determining a list of signal point candidate combinations determined by the tree structure obtained from the forward signal point search step;
In the list, following the tree structure from the lowest layer signal point of the most probable signal point combination to the highest layer, the signal point candidates are found in the opposite direction. A reverse signal point search step to select;
The combination of signal point candidates newly added by the backward signal point search step is replaced with the combination of signal point candidates from the tree structure obtained in the forward signal point search step, or the previous order In addition to the signal point candidate combination from the tree structure obtained in the direction signal point search step, calculating a log likelihood ratio based on the updated list;
A signal point detector that implements a method comprising estimating a most probable combination of signal points for the plurality of transmission signals based on the calculated log likelihood ratio;
A receiving apparatus comprising: a decoding unit that decodes the plurality of bit strings of the plurality of transmission signals based on the log likelihood ratio information output from the signal point detection unit. The backward signal point search step in the method performed in the signal point detection unit includes:
In the list, following the tree structure from the lowest layer signal point to the highest layer in the most probable signal point combination, the reference signal point in the opposite direction is Identifying steps;
Selecting a signal point candidate having a minimum squared Euclidean distance with respect to the reference signal point in the reverse direction;
The receiving device according to claim 9, further comprising: repeating the step of identifying the reference signal point in the reverse direction and the step of selecting until the highest layer is sequentially reached.

Images