CN107104714B

CN107104714B - MIMO detection method without QR decomposition

Info

Publication number: CN107104714B
Application number: CN201710270033.1A
Authority: CN
Inventors: 韩煜
Original assignee: CETC 36 Research Institute
Current assignee: CETC 36 Research Institute
Priority date: 2017-04-24
Filing date: 2017-04-24
Publication date: 2020-02-21
Anticipated expiration: 2037-04-24
Also published as: CN107104714A

Abstract

The invention relates to a MIMO detection method, which comprises the following steps: 1. the channel matrix is rearranged in a layered mode according to the signal-to-noise ratio, and the symbol layer with the highest signal-to-noise ratio is used as a father node; 2. selecting M father node symbol estimation values, deducing the symbol estimation values of the rest layers except the father node, and determining K optimal father nodes; 3. selecting N nearest neighbor symbols as a child node set after the parent node is expanded aiming at one of the optimal parent nodes; 4. carrying out interference cancellation on interference, carrying out zero forcing detection, and quantizing the zero forcing detection into a constellation point set to obtain symbol estimation values of the rest layers except the interference layer; 5. taking the minimum Euclidean distance between the father node and the N child nodes as the Euclidean distance of the current father node; 6. and (5) repeating the steps 3-5 for each of the K optimal father nodes to obtain K Euclidean distances, and taking the path corresponding to the minimum Euclidean distance as a final detection value. The invention can solve the problems that the prior art is complex in calculation and is not suitable for being applied in occasions with limited resources.

Description

MIMO detection method without QR decomposition

Technical Field

The invention relates to the technical field of signal processing, in particular to a MIMO detection method without QR decomposition.

Background

Mimo (Multiple Input Multiple output), a widely used physical layer technology in the next generation wireless communication standard, can provide higher channel capacity over a limited bandwidth, and has become one of the hot spots in wireless communication research in recent years. From the future perspective, the MIMO technology standard is planned in both LTE standard and 4G/5G standard. With the increase of the number of the receiving and transmitting antennas, the complexity of the MIMO system itself is also rapidly increased, and one of the core problems is a low-complexity and high-performance signal detection method.

Maximum Likelihood (ML) detection is a detection method theoretically having optimal error Rate (SER), but the ML method is very large in computational complexity and difficult to implement real-time processing in engineering. Detection based on Zero Forcing (ZF) and Minimum Mean Square Error (MMSE) has the least computational complexity, but the detection performance is poor, and the actual application requirements are difficult to meet. In recent years, a K-Best detection method based on tree search is widely researched, and the K-Best detection can better balance the contradiction between SER performance and algorithm complexity and is widely concerned. In the K-Best method, when the K value is small, the SER performance is sharply reduced, while the K value is large to increase the computational complexity, and the K-Best detection requires a complicated QR (orthogonal triangular decomposition) decomposition operation. Therefore, for resource-constrained applications with a small number of antennas (e.g., portable and mobile terminal nodes, typically with an antenna size of 4 × 4 or less), the K-Best method still has a large amount of computation, and thus the computation resource and power consumption problems are still hard to bear.

Aiming at the problems, the invention provides a low-complexity MIMO detection method without QR decomposition, which combines zero forcing detection, neighborhood expansion and multipath Euclidean distance calculation, has relatively low calculation complexity when the antenna scale is 4x4 and 3x3, still has SER performance similar to K-Best detection even under high-order constellation mapping, and can be applied to occasions with limited resources.

Disclosure of Invention

In view of the above analysis, the present invention aims to provide a MIMO detection method without QR decomposition, so as to solve the problem that the prior art is not suitable for the application in the resource-limited situation due to the complexity of computation.

The purpose of the invention is mainly realized by the following technical scheme:

the MIMO detection method comprises the following steps:

step 1, channel matrixes are rearranged in a layered mode according to the signal-to-noise ratio, and a symbol layer with the highest signal-to-noise ratio is used as a father node;

step 2, selecting M father node symbol estimation values in the constellation point set, deducing the symbol estimation values of other layers according to the M father node symbol estimation values, and determining K optimal father nodes;

and 3, acquiring Euclidean distances of the K optimal father nodes, solving the minimum Euclidean distance, and determining a path corresponding to the minimum Euclidean distance as a final detection value.

Wherein the content of the first and second substances,

in step 1, the channel matrix is rearranged in a layered manner according to the magnitude of the signal-to-noise ratio, and the method specifically comprises the following steps: and calculating the power sum of the channel matrix according to columns, rearranging the channel matrix according to the ascending order of the power, and sequentially decreasing the signal-to-noise ratio of the corresponding symbol at the moment.

Specifically, the channel matrix calculates the power sum by column by P ═ diag (H ═ H)^T) Calculated, where H represents the channel matrix and diag (.) is a take diagonal element operation.

And step 2, determining K optimal father nodes by calculating Euclidean distances.

Specifically, for antenna sizes of 4 × 4 or more, after M father node symbol estimation values are selected in step 2, the father node symbols are regarded as interference, the rest symbols sequentially perform interference cancellation on the father node symbols, zero forcing detection is performed and quantization is performed on constellation points, and each layer of symbol estimation values are obtained.

The obtaining of the euclidean distances of the K optimal parent nodes in step 3 is further implemented by iteratively calculating the euclidean distance of each parent node of the K optimal parent nodes, and specifically includes:

step 31, aiming at one of the K optimal father nodes, selecting N nearest neighbor symbols of father node symbols in a constellation point set as a child node set after father node expansion;

step 32, the residual layer performs interference cancellation on the interference layer, performs zero forcing detection, and quantizes the zero forcing detection into a constellation point set to obtain a symbol estimation value of the residual layer;

step 33, according to the child node set in the step 3 and the symbol estimation values of the rest layers in the step 4, calculating Euclidean distances of N paths between the father node and the N child nodes, and taking the minimum Euclidean distance as the Euclidean distance of the current father node;

and 34, repeating the steps 31-33 respectively for each father node of the K optimal father nodes to obtain the Euclidean distances of the K optimal father nodes.

Further, in step 32, the remaining layers are the two layers with the lowest signal-to-noise ratio, and the interference layer is the layer other than the remaining layers.

The invention has the following beneficial effects:

the invention relates to a low-complexity MIMO detection method without QR decomposition, which comprises the following technologies: (1) reordering the channel matrix according to the signal-to-noise ratio, and regarding the symbol layer corresponding to the high signal-to-noise ratio as residual symbol interference; (2) after the layer-by-layer interference is counteracted, simple inversion is used for replacing complex QR decomposition to realize zero forcing detection; (3) selecting a nearest symbol neighborhood as an expansion child node to avoid complex tree search operation; (4) and calculating and comparing the Euclidean distance of the path according to the father node, the child node and the residual symbol estimation value. The method has the advantages that QR decomposition is replaced by simple matrix inversion, and the symbol neighborhood expansion subnodes are adopted, so that the method is very suitable for MIMO signal detection under the occasions with small antenna scale and resource limitation, and the error rate performance can be effectively improved with relatively less calculation complexity.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Drawings

The drawings are only for purposes of illustrating particular embodiments and are not to be construed as limiting the invention, wherein like reference numerals are used to designate like parts throughout.

Fig. 1 is a 4x4 antenna scale MIMO detection flow;

fig. 2 is a schematic diagram of a 16-QAM nearest symbol neighborhood expansion subnode.

Detailed Description

The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form a part hereof, and which together with the embodiments of the invention serve to explain the principles of the invention.

The embodiment of the invention discloses a method for detecting MIMO (Multiple Input Multiple output), which takes the scale of 4x4 antennas as an example.

For convenience of description, the baseband equivalent model formula of the antenna scale MIMO system is H · S + w; wherein y is a received signal vector, H is a channel matrix, S is a transmitted signal complex vector, and ω is a gaussian noise vector with independent and identically distributed statistical properties.

The 4x4 antenna-scale MIMO system baseband equivalent model is shown in formula (1), wherein the channel matrix H corresponds to 4 input 4 output antennas; s is a 4 × 1-dimensional complex vector of a transmitted signal corresponding to M Quadrature Amplitude Modulation (QAM) constellation point symbols, and in the embodiment, 16-QAM is specifically adopted; y is a 4x1 dimensional received signal vector, and ω is a gaussian noise vector with independent co-distributed statistical properties. The MIMO baseband equivalent model can be divided into 4 layers according to rows, wherein the first row is taken as the topmost layer, namely layer 4, corresponds to a father node, and the subsequent layers are decreased progressively. And y, H and W are known quantities, and the optimal detection value is obtained by calculating and judging the estimated value of S.

The MIMO detection method provided by the invention comprises the following specific steps.

Step 1, rearranging the channel matrix H according to the size of the signal-to-noise ratio according to the 4x4 antenna scale, wherein the size of the signal-to-noise ratio is decreased from left to right. And taking the symbol layer with the highest signal-to-noise ratio as a father node and taking the symbol layer with the second highest signal-to-noise ratio as a child node.

For the determined constellation point symbol power, the smaller the channel power, the larger the signal-to-noise ratio of the corresponding symbol, and the higher the reliability of the symbol estimation value. Thus, the power sum is calculated column by column for the channel matrix H, the calculation is disclosed as shown in equation (2), where diag (.) is a take diagonal element operation. And rearranging the layers of the channel matrix H according to the ascending order of the power P, and sequentially decreasing the corresponding signal-to-noise ratios after sorting, wherein the signal-to-noise ratio of S1 is the highest, the reliability of the estimated value is the highest, the signal-to-noise ratio of S2 is the second, the signal-to-noise ratio of S4 is the lowest, and the reliability of the estimated value is the lowest.

P＝diag(H*H^T) (2)

The detection process is started from the high signal-to-noise ratio symbol estimation, and the detection process is sequentially processed layer by layer, wherein the symbol layer corresponding to the high signal-to-noise ratio is regarded as residual symbol interference.

Step 2: selecting M father node symbol estimated values in the constellation point set by using the symbol layer with the highest signal-to-noise ratio, namely layer 4 as a father node

And the rest layers sequentially carry out interference cancellation on the symbols of the father nodes, then carry out zero forcing detection and quantize the symbols to constellation points to obtain symbol estimation values of the rest layers.

Specifically, the symbol layer with the highest signal-to-noise ratio, i.e. layer 4, is taken as a father node, and M father node symbol estimation values are selected as

The embodiment takes 16-QAM as an example, and 16 father node symbol estimated values are selected

For the subsequent layers, the father node symbol is regarded as interference, the rest symbols sequentially perform interference cancellation on the father node symbol, and the interference cancellation is performed by adopting the announcement (3).

Corresponding to M QAM constellation point symbols S₁After the cancellation, ZF detection (zero forcing detection) is performed according to the following formula (4), and a 3x1 dimensional vector is obtained

Then, the M residual symbols (S) can be obtained by determining and quantizing the symbols to the constellation points₂，S₃，S₄) Estimated value

Corresponding to the estimated symbol values of layer 3, 2 and 1 nodes, respectively.

The present embodiment is described by taking an antenna size of 4x4 as an example, and when the antenna size is 3x3, the step of interference cancellation on the parent node symbol is not required. When the antenna size is above 4x4, an interference cancellation step is also required.

And step 3: and calculating Euclidean distances and sequencing according to the estimated values of the symbols of each layer deduced by the M father node symbols, and determining K optimal father nodes.

According to M groups of symbol estimation values of each layer derived from M father node symbols, Euclidean distance T is calculated according to the following formula (5)₁And sequencing the M Euclidean distances in an ascending order, and determining the node symbol estimation values corresponding to the first K Euclidean distances as an optimal node set.

And 4, step 4: and aiming at one of the K optimal father nodes, selecting N nearest neighbor symbols of father node symbols in the constellation point set as a child node set after the father nodes are expanded.

Specifically, in the optimal node set, the node symbol of the corresponding layer 3

Selecting a nearest neighbor symbol in the constellation point set and the node symbol as a child node set after the parent node is expanded, and assuming that the nearest neighbors are N, namely N child nodesThe nearest neighbor symbol expansion method is shown in fig. 2. Taking 16-QAM as an example, in the 16-QAM constellation in fig. 2,corresponding to the solid black dots in the figure, there are 3 nearest neighbor sets according to their positions in the constellation, as shown by the dotted boxes in fig. 2, which correspond to 3, 5, and 8 nearest neighbor symbols, respectively, i.e., the white dots in the dotted boxes in fig. 2.

When the upper layer node symbol expands the lower layer child node, the nearest neighbor symbol of the node symbol in the constellation point set is selected as the child node set after the parent node is expanded. The nearest neighbor expansion method is that the neighbor symbol with the minimum distance is sequentially selected according to the position of the father node symbol in the MIMO modulation constellation diagram, and the child node set elements are only determined by the MIMO modulation constellation diagram and have extremely low computation complexity.

And 5: and (3) regarding the father node and the child nodes as interference, removing the rest layers of the father node and the child nodes for interference cancellation, then performing zero forcing detection, and quantizing the zero forcing detection into a constellation point set to obtain symbol estimation values of the rest layers except the father node and the child nodes.

In particular, layer 4 is assigned to a parent node

Sub-node corresponding to layer 3

The remaining layers are considered as interference, and the interference is cancelled again as shown in the following equation (6).

Sequentially and sequentially offsetting, and performing ZF detection as shown in the following formula (7) to obtain a 2x1 dimensional vectorThen, the residual symbols are updated (S) after the decision quantization is carried out on the constellation points₃，S₄) Estimated value

Corresponding to the estimated symbol values of layer 2 and layer 1 nodes, respectively.

In this embodiment, the antenna scale of 4x4 is taken as an example for explanation, and when the antenna scale is 3x3, only the parent node needs to be regarded as interference to perform interference cancellation; when the antenna size is 4x4, both the parent node and the child node need to be considered as interference; when the antenna size is larger than 4x4, the remaining layers are two layers with the lowest signal-to-noise ratio, the interference layer is a layer other than the remaining layers, and the remaining layers perform interference cancellation with respect to the interference layer.

Step 6: and 4, calculating Euclidean distances between the parent node and N paths of the N child nodes according to the path set of the child nodes in the step 4, and taking the minimum Euclidean distance as the Euclidean distance of the current parent node.

Specifically, corresponding to each parent node

The Euclidean distance T is calculated according to the following formula (8)₂According to the N child nodes in step 4

N paths can be determined, and the step 5 is repeated to obtain the corresponding paths

Has an Euclidean distance omega (T) of N paths₂,N)。

And 7: and solving the minimum Euclidean distance of the N paths as the Euclidean distance of the current father node.

Specifically, the minimum Euclidean distance of N paths is obtained as the current father node according to the following formula (9)

Of Euclidean distance T₃The minimum calculation complexity is clearly lower compared to the rank calculation.

T₃＝min(Ω(T₂,N)) (9)

And 8: changing one of the K optimal father nodes aimed at in the step S4, namely respectively aiming at other father nodes of the K optimal father nodes, repeating the steps 4-7 to obtain the Euclidean distance omega (T) of the K father nodes₃,K)。

And step 9: and taking the path corresponding to the minimum Euclidean distance of the K father nodes as 4 receiving antenna layer symbols as a final detection value.

Specifically, K father nodes are obtained according to the following formula (10)

And corresponding to the minimum value of the Euclidean distance set, and determining a corresponding path as 4 receiving antenna layer symbols as a final detection value.

T₃＝min(Ω(T₃,K)) (10)

And 2, taking a high signal-to-noise ratio level as an upper node, sequentially selecting node symbol estimated values in a node symbol set, taking the upper symbol estimated values as interference in a subsequent level, subtracting interference influence from an antenna receiving vector, namely performing zero forcing detection after interference cancellation to obtain subsequent symbol vectors, and obtaining residual symbol estimated values by judging and quantizing the subsequent symbol vectors to constellation points. The 4x4 antenna scale performs interference cancellation and zero forcing detection on layer 4 and layer 3 in sequence, and the 3x3 antenna scale only needs interference cancellation and zero forcing detection on layer 3.

According to the method, the symbol estimation value of the father node is expanded, the symbol estimation value of the child node is expanded, the residual symbol estimation value is obtained by zero forcing detection after interference is counteracted, and multi-path Euclidean distance calculation and comparison are carried out. The detection method comprises two times of Euclidean distance calculation, one time of Euclidean distance sequencing and two times of Euclidean distance minimum value calculation. The calculation and comparison of euclidean distances determines the final computational parallelism and complexity.

The method replaces QR decomposition with simple inversion operation of small antenna scale, the nearest neighbor expansion sub-node of the symbol to be estimated and multi-path Euclidean distance calculation and comparison, and can realize SER detection performance similar to K-Best for low-order 16-QAM and high-order 64-QAM under the antenna scale of 4x4 and 3x3, solve the contradiction between MIMO detection calculation complexity and SER performance, effectively reduce calculation complexity and remarkably improve the SER detection performance.

In summary, the embodiments of the present invention provide a low-complexity MIMO detection method without QR decomposition, and mainly provide main flows of channel matrix rearrangement preprocessing, zero-forcing detection, symbol neighborhood expansion, multipath euclidean distance calculation and comparison, and the like, at a 4 × 4 antenna scale. The method replaces QR decomposition with simple matrix inversion, is very suitable for MIMO signal detection under the occasions with small antenna scale and limited resources, and can effectively improve the error rate performance with relatively less calculation complexity.

Those skilled in the art will appreciate that all or part of the flow of the method implementing the above embodiments may be implemented by a computer program, which is stored in a computer readable storage medium, to instruct related hardware. The computer readable storage medium is a magnetic disk, an optical disk, a read-only memory or a random access memory.

The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention.

Claims

1. A MIMO detection method, comprising the steps of:

step 2, selecting M father node symbol estimated values in the constellation point set, regarding the father node symbols as interference, sequentially performing interference cancellation on the father node symbols by the remaining symbols, and then performing zero forcing detection and quantizing the zero forcing detection on the constellation points to obtain each layer of symbol estimated values; deducing the symbol estimation values of the rest layers according to the symbol estimation values of the M father nodes, and determining K optimal father nodes;

step 3, obtaining Euclidean distances of K optimal father nodes, solving a minimum Euclidean distance, and determining a path corresponding to the minimum Euclidean distance as a final detection value;

step 33, according to the child node set in step 31 and the symbol estimation values of the rest layers in step 32, calculating Euclidean distances of N paths between the parent node and the N child nodes, and taking the minimum Euclidean distance as the Euclidean distance of the current parent node;

2. The MIMO detection method according to claim 1, wherein the step 1 of rearranging the channel matrices hierarchically according to the snr level specifically includes: and calculating the power sum of the channel matrix according to columns, rearranging the channel matrix according to the ascending order of the power, and sequentially decreasing the signal-to-noise ratio of the corresponding symbol at the moment.

3. The MIMO detection method of claim 2, wherein the channel matrix calculates the power sum by column by P ═ diag (H × H)^T) Calculated, where H represents the channel matrix and diag (.) is a take diagonal element operation.

4. The MIMO detection method of claim 1, wherein the K optimal parent nodes are determined in step 2 by calculating euclidean distances.

5. The MIMO detection method of claim 1, wherein in step 32, the remaining layers are two layers with the lowest snr, and the interference layer is a layer other than the remaining layers.