CN102281091B - Reception method for multi-antenna communication system - Google Patents

Abstract

The invention discloses a reception method for a multi-antenna communication system and belongs to the field of the transmission of radio signals. The reception method is characterized by comprising the following steps of: ordering the Euclidean distance increments of only the top two layers of signals in a search tree of a detection algorithm so that only K sub-nodes with smaller Euclidean distance increments are kept in each two layers; then calculating the accumulated Euclidean distances of the (2N-1) layer in the top two layers, wherein N represents the number of the antenna; and sequentially selecting the (2N-2) layer and the (2N-3) layer from the (2N-2) layer, calculating the Euclidean distance increment of each layer and adding the calculated Euclidean distance increments with K accumulated Euclidean distances in the first period to obtain the accumulated Euclidean distance of the (2N-3) layer, and performing the calculation in the same manner till the accumulated Euclidean distance of the first layer is calculated. According to the invention, two adjacent layers of signals can be simultaneously treated so that the ordering, selecting, calculating and searching times are reduced, the system time relay and the calculation complexity are decreased, and lots of hardware resources are saved.

Description

Receiving method for multi-antenna communication system

Technical Field

The invention belongs to the field of wireless information transmission, in particular to a wireless local area network, a broadband wireless access system, a mobile communication system and other systems and standards adopting a multi-antenna technology.

Background

Multiple antennas are used at both a signal transmitting end and a signal receiving end by a multiple-input multiple-output (MIMO) technology, a plurality of data streams are simultaneously transmitted in the same frequency band through a spatial multiplexing mode, higher spectral efficiency than that of a single antenna system (SISO) can be provided, and the MIMO technology is currently applied to new-generation broadband wireless communication standards such as 3GPP LTE, LTE-Advanced, ieee802.11n and the like. The best hard decision detection mode for MIMO wireless systems is the Maximum Likelihood (ML) detection method, whose complexity of direct implementation increases exponentially with the increase in the number of antennas and modulation orders, so that ASICs or FPGAs can only be used for low-order modulation schemes for a few antennas. Subsequently, K-Best detection algorithms have been proposed that can maintain Bit Error Rate (BER) performance comparable to that of the optimal ML detector, while significantly reducing computational complexity, and have received much attention.

In general, the K-Best algorithm decomposes a system model into real numbers, and reduces the complexity of hardware implementation by decoding in a real number domain, and the principle of the algorithm is described below. MIMO communication system baseband signal equivalent model considering that both the number of transmitting antennas and the number of receiving antennas are N

y=Hx+n (1)

WhereinH is an NxN channel matrix; x = [ x =₁,x₂,…,x_N]Transmitting signals for the N dimension; y = [ y₁,y₂,…,y_N]Receiving signals for N dimensions; n is N-dimensional additive white Gaussian noise. To avoid extra hardware overhead caused by complex operations, the system model (1) can be decomposed into the following real numbers:

$\left[\begin{array}{c}\mathrm{Re}\left(y\right)\\ \mathrm{Im}\left(y\right)\end{array}\right]=\left[\begin{array}{cc}\mathrm{Re}\left(H\right)& -\mathrm{Im}\left(H\right)\\ \mathrm{Im}\left(H\right)& \mathrm{Re}\left(H\right)\end{array}\right]\left[\begin{array}{c}\mathrm{Re}\left(x\right)\\ \mathrm{Im}\left(x\right)\end{array}\right]+\left[\begin{array}{c}\mathrm{Re}\left(n\right)\\ \mathrm{Im}\left(n\right)\end{array}\right]---\left(2\right)$

where Re (×) and Im (×) represent the real and imaginary parts of the complex (×), respectively. Real-valued channel matrixObtained by QR decompositionSolving (1) using the maximum likelihood criterion yields:

<math> <mrow> <mrow> <mover> <mi>x</mi> <mo>^</mo> </mover> <mo>=</mo> <munder> <mrow> <mi>arg</mi> <mi>min</mi> </mrow> <mrow> <mi>x</mi> <mo>&Element;</mo> <msup> <mi>&Omega;</mi> <mrow> <mn>2</mn> <mi>N</mi> </mrow> </msup> </mrow> </munder> <munderover> <mi>&Sigma;</mi> <mrow> <mi>i</mi> <mo>=</mo> <mn>1</mn> </mrow> <mrow> <mn>2</mn> <mi>N</mi> </mrow> </munderover> <mo>|</mo> <mo>|</mo> <mover> <msub> <mi>y</mi> <mi>i</mi> </msub> <mrow> <mo>~</mo> <mi>T</mi> </mrow> </mover> <mo>-</mo> <munderover> <mi>&Sigma;</mi> <mrow> <mi>j</mi> <mo>=</mo> <mi>i</mi> <mo>+</mo> <mn>1</mn> </mrow> <mrow> <mn>2</mn> <mi>N</mi> </mrow> </munderover> <msub> <mi>R</mi> <mi>ij</mi> </msub> <msub> <mi>x</mi> <mi>j</mi> </msub> <mo>-</mo> <msub> <mi>R</mi> <mi>ii</mi> </msub> <msub> <mi>x</mi> <mi>i</mi> </msub> <mo>|</mo> <mo>|</mo> </mrow> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>3</mn> <mo>)</mo> </mrow> </mrow> </math>

wherein,representing the x estimation value which satisfies the minimum Euclidean measurement distance; argmin represents a mathematical formula for solving the minimum value;representing a Euclidean distance of an i-th layer signal; q represents: a 2 Nx 2N-dimensional orthogonal matrix; r represents: a 2 Nx 2N dimensional upper triangular matrix;is a 2N-dimensional vector and is a vector,the signal is received for the real number domain.

According to equation (3), the tree search structure of the general K-Best MIMO detection algorithm is as shown in fig. 1 (2 × 2 antennas, 16-QAM modulation, K = 4), and the MIMO detectorStarting detection from the highest layer, calculating Euclidean distance Increment (INC) of each layer, adding K accumulated Euclidean distances (PEDs) reserved by the upper layer to obtain PEDs of the layer, and selecting K minimum PEDs reserved by the layer and corresponding nodes (survivor nodes) of the PEDs through sorting operation. For the K-Best detector tree search structure with the modulation order of M, the number of PEDs to be calculated and sequenced in each layer isWhen K and M increase, the computational complexity increases exponentially, becoming a bottleneck in hardware implementation.

The hardware structure of a 2 x 2MIMO detector employing the above K-Best algorithm is shown in fig. 2. The detector consists of a QR decomposition module, a balancing module, a to-be-selected generation module, a 4-layer K-Best module and a final judgment module. Each K-Best module comprises a Euclidean distance calculation unit (DCM) for calculating the Euclidean distance of the nodes in the current layer and a sorting unit (SSM) for sorting and selecting K nodes. The number of layers to be detected in the K-Best detector is doubled along with the real number decomposition of a channel model, and the detector detects 2N layers of signals in 2N different stages respectively.

It can be seen that the sorting selection unit of the K-Best MIMO detector in the conventional structure also occupies a large amount of hardware resources. Therefore, the invention provides a novel K-Best MIMO detector structure which can simultaneously detect two layers of signals, reduces the times of sequencing and tree search and has slightly better BER performance than the traditional K-Best detector.

Disclosure of Invention

The invention aims to provide a detection method of a low-complexity MIMO system.

The idea of the invention is that: by adopting a new tree search structure, only the Euclidean distance increment of the highest two layers of signals is sequenced, the adjacent two layers of signals can be processed simultaneously in parallel, and the sequencing selection operation and the tree search times are reduced.

The invention has the following implementation steps:

step (1), initializing a MIMO communication system with the number of transmitting antennas and the number of receiving antennas both being N, wherein a baseband signal equivalent model of the MIMO communication system is expressed as: y = Hx + n, where y = [ y =₁,y₂,…,y_N]For N-dimensional received signals, x = [ x ]₁,x₂,…,x_N]Sending signals for N dimensions, wherein H is an NxN channel matrix, and N is N-dimensional additive white Gaussian noise;

step (2), a QR decomposition module is utilized to carry out real number on the channel matrix to obtain an NxN dimensional real number channel matrixAnd (3) decomposing:wherein Q is a unitary matrix, and R is an upper triangular matrix;

step (3), the equalizing processing module utilizes the unitary matrix Q to carry out the real-domain part of the N-dimensional receiving signalPerforming equalization processing to obtain 2N-dimensional vectorT is a transposed symbol;

and (4) generating a candidate block R for calculating Euclidean distance increment of each layer of signals in the detection algorithm search tree from the upper triangular matrix R by using a candidate generation module_2N,2Nx，n＝1,2,…,2N-1,2N；

And (5) the first-stage K-Best module calculates K smaller cumulative Euclidean distances PED of the 2N-1 layer according to the following steps in sequence_2N-1(x)；

Step (5.1) of calculating Euclidean distance increment INC of the 2N layer as the highest layer according to the following formula_2N(x) Searching for a start in a tree in the K-Best modeThe node is the lowest layerWherein,for receiving signals in the real number domain of layer 2N, R_2N,2NIs a 2N layer upper triangular matrix R_2NThe element with the sequence number of 2N in the sequence number,

step (5.2) of calculating Euclidean distance increment INC of the 2N-1 th layer as the next higher layer according to the following formula_2N-1(x)， ${\mathrm{INC}}_{2N-1}\left(x\right)={|}_{y_{2N-1}}^{~T}-R_{2N-\mathrm{1,2}N-1}x|,$

Step (5.3) of calculating M cumulative Euclidean distances PED of the 2N-1 th layer according to the following formula_2N-1(x) M is the modulation order, PED_2N-1(x)＝INC_2N(x)+INC_2N-1(x)；

And (5.4) sequencing all Euclidean distance increments in the step (5.3) from small to large, wherein each layer in the highest two layers only keeps L accumulated Euclidean distance increments with small Euclidean distance increments, L is less than K, and L of each layer is the same. The K cumulative Euclidean distances of the 2N-1 th layer are arranged and combined in L Euclidean distance increments of the 2N and 2N-1 th layers, and PED is used_2N-1(x) Representing that each father node in the 2N-1 layer respectively reserves a child node with the minimum Euclidean distance increment, and the total number of the child nodes is K;

and (6) the second-level K-Best module calculates K Euclidean distance increments of two layers from 2N-2 to 2N-3, and then adds the K Euclidean distance increments to the K accumulated Euclidean distances of the 2N-1 layer calculated by the previous-level K-Best module obtained in the step (5) to obtain K accumulated Euclidean distances PED of the 2N-3 layer reserved by two layers of the 2N-2 layer and the 2N-3 layer_2N-3(x)；

Step (7), repeating step (6) until the layer 1 is calculated;

and (8) selecting a group of nodes with the minimum accumulated Euclidean increment distance from the reserved K nodes by using a judgment module as the output of the detector.

Unlike conventional K-Best detectors, the proposed detector contains only two K-Best modules. And each K-Best module simultaneously calculates Euclidean distance increments of two adjacent layers of signals, and adds each K Euclidean distance increments reserved by the two layers to K accumulated Euclidean distances in the previous stage to obtain K accumulated Euclidean distances reserved by the stage and corresponding nodes thereof. The decision module selects a group of nodes with the minimum accumulated Euclidean distance path from K groups of nodes reserved in each layer as decoding output. The detector designed by the invention reduces the times of sequencing selection operation and tree search, can simultaneously detect two layers of signals in parallel, and greatly reduces the system time delay and the hardware resource occupation compared with the traditional K-Best structure.

Drawings

Fig. 12 x 216-QAM conventional K-Best (K = 4) detection algorithm tree search structure.

Fig. 2 a conventional K-Best signal detector architecture.

Fig. 3 is a novel K-Best signal detector structure.

Fig. 4 detection algorithm flow.

Figure 52 x2 antenna, QPSK modulation K-Best algorithm BER performance.

Detailed Description

In the detector, the processing steps for the received signal are as described in fig. 4, as follows:

first, for the channel matrixAnd performing QR decomposition to obtain a unitary matrix Q and an upper triangular matrix R.

And secondly, equalizing the received signal by using the unitary matrix Q.

And thirdly, calculating the Euclidean distance increment by utilizing R to generate a candidate generation module, so that the occupation of hardware resources for calculating the Euclidean distance increment can be reduced.

Fourthly, in the first K-Best module, the Euclidean distance increment of the (2N, 2N-1) th layer is calculated firstly, and only the Best L (L) is reserved in each layer after the Euclidean distance increments are sorted<K) Then, the cumulative Euclidean distance PED is calculated_2N-1(x)。

Fifthly, other K-Best modules calculate Euclidean distance increments from the 2N-2 th layer to the 1 st layer, and then the K Euclidean distance increments are added with K accumulated Euclidean distances in the previous stage to obtain K accumulated Euclidean distances reserved in the current stage.

And sixthly, selecting a group of nodes with the minimum accumulated Euclidean distance path from the reserved K groups of nodes by the judgment module as the output of the detector.

The MIMO detector provided by the invention mainly comprises a QR decomposition module, a balancing module, a to-be-selected generation module, a two-layer K-Best module and a judgment module. Each K-Best module comprises a Euclidean distance calculation unit (DCM) for calculating the Euclidean distance of the nodes in the current layer and a sorting unit (SSM) for sorting and selecting K nodes. For example, for a 2 x2 system, the hardware structure of the proposed K-Best MIMO detector is as shown in fig. 3.

To analyze the performance of the proposed MIMO detector, point-to-point MIMO link with 2 × 2 antenna, QPSK, 16/64-QAM modulation, and no channel coding was simulated under the flat fading channel model, and compared with the performance of the conventional K-Best algorithm and the Maximum Likelihood (ML) algorithm, respectively, and the Bit Error Rate (BER) performance thereof is shown in fig. 5. Simulation results show that the proposed structure has slightly better BER performance than the conventional K-Best detector as K increases.

The following describes the signal detection implementation procedure by taking a MIMO system composed of two transmitting antennas and two receiving antennas as an example.

The first step is as follows: carrying out QR decomposition on the channel matrix to obtain

$Q=\left[\begin{array}{cccc}{Q}_{11}& Q_{12}& Q_{13}& Q_{14}\\ Q_{21}& Q_{22}& Q_{23}& Q_{24}\\ Q_{31}& Q_{32}& Q_{33}& Q_{34}\\ Q_{41}& Q_{42}& Q_{43}& Q_{44}\end{array}\right]R=\left[\begin{array}{cccc}{R}_{11}& R_{12}& R_{13}& R_{14}\\ 0& R_{22}& R_{23}& R_{24}\\ 0& 0& R_{33}& R_{34}\\ 0& 0& 0& R_{44}\end{array}\right]$

The second step is that: equalizing the received signal by using the unitary matrix Q to obtain

The third step: and generating a candidate generation module by using R, wherein the generation module is R x (+ -3) and R x (+ -1) for 16QAM modulation.

The fourth step: in the first K-Best module, Euclidean distance increments of the third (third, fourth) layer are calculated firstly, and only the Best L (L) is reserved in each layer after the Euclidean distance increments are sorted<K) Two layers L are identical, and then the cumulative Euclidean distance PED is calculated_2N-1(x) It is the sum of 2N-1, N (N = 2) layer euclidean distance increments. As in y of FIG. 3₁～y₄Is the real and imaginary parts, z, of the two sets of data for the two sets of antennas₃～z₄The nodes reserved in three or four layers are summed, sorted, selected and summed for output.

The fifth step: in other K-Best modules, Euclidean distance increments from the 2N-2 th layer to the 1 st layer are calculated, and then each K Euclidean distance increments are compared with K accumulated Euclidean distance increments in the previous stageAnd adding the distances to obtain K accumulated Euclidean distances reserved at the stage. As in z in FIG. 3₁～z₄Representing four levels of reserved nodes.

And a sixth step: and the decision module selects a group of nodes with the minimum accumulated Euclidean distance path from the reserved K groups of nodes as the output of the detector.

In order to verify that the proposed K-Best detector structure has low hardware implementation complexity, the proposed structure is comprehensively simulated by using a Xilinx Virtex-4(XC4VLX200) platform. Table 1 shows the hardware resource occupation comparison of the 2 × 2 antenna, 16-QAM modulation, and two structure detectors. It can be seen that the proposed K-Best detector saves a lot of hardware resources compared to the conventional K-Best detector.

TABLE 1 Detector resource occupancy

Claims (1)

1. A receiving method for a multi-antenna communication system, characterized by the following implementation steps in sequence:

step (1), initializing a MIMO communication system with the number of transmitting antennas and the number of receiving antennas both being N, wherein a baseband signal equivalent model of the MIMO communication system is expressed as:

y is Hx + s, where y is [ y ═ y₁,y₂,…,y_N]For N-dimensional received signals, x ═ x₁,x₂,…,x_N]Sending signals for N dimensions, wherein H is an NxN channel matrix, and s is N-dimensional additive white Gaussian noise;

step (2), a channel matrix obtained by real-number processing the NxN channel matrix by using a QR decomposition moduleAnd (3) decomposing:wherein Q is a unitary matrix, and R is an upper triangular matrix;

step (3), the equalizing processing module utilizes the unitary matrix Q to carry out the real-domain part of the N-dimensional receiving signalPerforming equalization processing to obtain 2N-dimensional vectorT is a transposed symbol;

and (4) generating a candidate block R for calculating Euclidean distance increment of each layer of signal in detection algorithm tree search from the upper triangular matrix R by using a candidate generation module_n,nx, N is 1,2, …,2N-1,2N, N is the mark number of the row and column of the upper triangular matrix R;

and (5) the first-stage K-Best module calculates K smaller cumulative Euclidean distances PED of the 2N-1 layer according to the following steps in sequence_2N-1(x)；

Step (5.1): the Euclidean distance increment INC of the 2N layer as the highest layer is calculated as follows_2N(x) The initial node of the search tree in the K-Best algorithm is the lowest layerWhereinFor receiving signals in the real number domain of layer 2N, R_2N,2NIs a 2N layer upper triangular matrix R_2NThe middle serial number is 2N;

step (5.2): the euclidean distance increment of the 2N-1 layer as the next highest layer is calculated as follows,

${\mathrm{INC}}_{2N-1}\left(x\right)=|{\tilde{y}}_{2N-1}^T-R_{2N-\mathrm{1,2}N-1}x|;$

step (5.3): the M cumulative Euclidean distances PED of the 2N-1 th layer are calculated according to the following formula_2N-1(x) M is the modulation order, PED_2N-1(x)＝INC_2N(x)+INC_2N-1(x)；

Step (5.4): sorting all Euclidean distance increments in the step (5.3) from small to large, wherein each layer in the highest two layers only keeps L accumulated Euclidean distance increments with small Euclidean distance increments, and L<K, each layer L is the same, so that the K cumulative Euclidean distances of the 2N-1 th layer are the permutation and combination of L Euclidean distance increments of the 2N-1 nd layer and the 2N-1 th layer, and PED is used_2N-1(x) Representing that each father node in the 2N-1 layer respectively reserves a child node with the minimum Euclidean distance increment, and the total number of the child nodes is K;

and (6) the second-level K-Best module calculates K Euclidean distance increments of two layers from 2N-2 to 2N-3, and then adds the K Euclidean distance increments with the K accumulated Euclidean distances of the 2N-1 layer calculated by the previous-level K-Best module obtained in the step (5) to obtain K accumulated Euclidean distances PED of the 2N-3 layer reserved by two layers of the 2N-2 layer and the 2N-3 layer_2N-3(x)；

Step (7), repeating step (6) until the layer 1 is calculated;

and (8) selecting a group of nodes with the minimum accumulated Euclidean distance increment from the reserved K nodes by using a judgment module as the output of the detector.