Spherical decoding detection method and device, electronic equipment and readable storage medium

Abstract

The application discloses a spherical decoding detection method and device, electronic equipment and a readable storage medium, and belongs to the technical field of wireless communication. The application obtains a Q matrix and an R matrix by performing orthogonal triangular QR decomposition on a channel response matrix, wherein each column in the Q matrix is arranged according to the size of channel energy; determines a target equalization signal according to the Q matrix and the R matrix; performs an ML path search process according to the target equalization signal to select N branches with minimum metrics as survivor paths; performs ML complement path search on the survivor paths to obtain ML complement paths; takes the survivor paths with minimum metrics as ML paths, and obtains the likelihood ratio information of each bit of each symbol at each layer according to the ML paths and the ML complement paths. The application can reduce the operation complexity of the spherical decoding algorithm while maintaining the decoding detection performance.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and in particular to a spherical decoding detection method, apparatus, electronic device, and readable storage medium. Background Technology

[0002] The core concept of MIMO (Multiple Input Multiple Output) is to utilize the spatial degrees of freedom provided by multiple transmit and receive antennas to effectively improve the spectral efficiency of wireless communication systems, thereby increasing transmission rates and improving communication quality. Because MIMO can significantly increase system data throughput and transmission distance without increasing bandwidth or total transmit power consumption, this technology has attracted much attention in recent years.

[0003] Among commonly used MIMO receiver detection methods, zero-forcing (ZF) and minimum mean square error (MMSE) have low complexity but the worst performance. Maximum likelihood detection (ML) is a high-performance algorithm among MIMO receiver detection methods, but it has the highest complexity, making it unsuitable for engineering implementations requiring real-time processing.

[0004] Sphere Decoding (SD) detection offers decoding performance close to that of Machine Learning (ML) detection with moderate complexity, making it a relatively ideal signal detection method. Essentially, sphere decoding transforms the maximum likelihood detection problem into a search for the optimal path on a source signal constellation tree, continuously strengthening the constraints during the search process. The working principle of sphere decoding is as follows: First, a sphere centered on the received signal point is predefined in the received signal space. This sphere is then mapped to an ellipsoid in the transmitted signal space, and possible transmitted signal points are searched within the ellipsoid. Once a transmitted signal point is found, the predefined sphere is shrunk by the distance between the mapped point and the received signal, thus allowing subsequent searches to be performed within a smaller range.

[0005] However, current spherical decoding algorithms still have many shortcomings, affecting their practical application. The overall search complexity of existing spherical decoding algorithms remains high. For example, with the increase in the number of support layers (transmit antennas) and modulation order in 4G / 5G, the computational load of spherical decoding algorithms will increase exponentially, which is detrimental to engineering implementation and low-cost requirements. Therefore, how to reduce the computational complexity of spherical decoding algorithms while maintaining decoding and detection performance is an urgent problem to be solved. Summary of the Invention

[0006] The main objective of this application is to provide a spherical decoding detection method, apparatus, electronic device, and readable storage medium, which aims to reduce the computational complexity of the spherical decoding algorithm while maintaining decoding detection performance.

[0007] To achieve the above objectives, this application provides a spherical decoding detection method, comprising:

[0008] After determining that the current system layer number is greater than two, the first channel response matrix is ​​decomposed into an orthogonal triangular QR matrix to obtain the first Q matrix and the first R matrix. The columns of the first Q matrix are arranged according to the channel energy, and the column with the smallest channel energy is placed at the top layer of the SD search.

[0009] Multiply the conjugate transpose of the first Q matrix with the received signal to obtain the first Z matrix. Normalize the first R matrix to obtain the first equalization R matrix. Determine the target equalization signal corresponding to the received signal based on the first Z matrix and the first equalization R matrix.

[0010] The first ML path search process is executed according to the target equalization signal, wherein the first ML path search process is to divide all nodes of the top layer into B Nt-1 There are several blocks, each containing a representative point and several extended points. Path detection is performed on each representative point, and the block with the smallest metric, Ka, is selected. Nt-1 Given N1 representative points, the blocks corresponding to these representative points are taken as surviving blocks. Path detection is then performed on the expansion points near the representative points within each surviving block. The N1 branches with the smallest metrics are selected as surviving paths, where B... Nt-1 Ka Nt-1 Both N1 and B are natural numbers, and B Nt-1 >Ka Nt-1 ;

[0011] Perform an ML complement path search on all surviving paths to obtain the ML complement path;

[0012] The surviving path with the smallest metric is taken as the ML path. Based on the ML path and the ML complement path, the likelihood ratio information of each bit of each symbol in each layer is obtained.

[0013] Furthermore, to achieve the above objectives, this application also provides a spherical decoding detection device, comprising:

[0014] The QR decomposition module is configured to perform orthogonal triangular QR decomposition on the first channel response matrix after determining that the current system layer number is greater than two, to obtain the first Q matrix and the first R matrix. The columns in the first Q matrix are arranged according to the magnitude of the channel energy, and the column with the smallest channel energy is placed at the top layer of the SD search.

[0015] The equalization signal calculation module is configured to multiply the conjugate transpose of the first Q matrix with the received signal to obtain the first Z matrix, normalize the first R matrix to obtain the first equalization R matrix, and determine the target equalization signal corresponding to the received signal based on the first Z matrix and the first equalization R matrix.

[0016] The ML path search module is configured to execute a first ML path search process based on the target equalization signal, wherein the first ML path search process involves dividing all nodes at the top level into B... Nt-1 There are several blocks, each containing a representative point and several extended points. Path detection is performed on each representative point, and the block with the smallest metric, Ka, is selected. Nt-1 Given N1 representative points, the blocks corresponding to these representative points are taken as surviving blocks. Path detection is then performed on the expansion points near the representative points within each surviving block. The N1 branches with the smallest metrics are selected as surviving paths, where B... Nt-1 Ka Nt-1 Both N1 and B are natural numbers, and B Nt-1 >Ka Nt-1 ;

[0017] The ML complement path search module is configured to perform ML complement path search on all surviving paths to obtain the ML complement path;

[0018] The soft value information calculation module is configured to take the surviving path with the smallest metric as the ML path, and derive the likelihood ratio information of each bit of each symbol in each layer based on the ML path and the ML complement path.

[0019] In addition, to achieve the above objectives, this application also provides an electronic device, which includes: a memory, a processor, and a spherical decoding detection program stored in the memory and executable on the processor. When the spherical decoding detection program is executed by the processor, it implements the spherical decoding detection method as described above.

[0020] In addition, to achieve the above objectives, this application also provides a readable storage medium, which is a computer-readable storage medium, and stores a spherical decoding detection program thereon. When the spherical decoding detection program is executed by a processor, it implements the spherical decoding detection method as described above.

[0021] This application proposes a spherical decoding detection method, apparatus, electronic device, and readable storage medium. In the spherical decoding detection method, the technical solution of this application embodiment involves, after determining that the current system layer number is greater than two, performing orthogonal triangular QR decomposition on the first channel response matrix to obtain a first Q matrix and a first R matrix. The columns of the first Q matrix are arranged according to the magnitude of the channel energy, with the column containing the minimum channel energy placed at SD (Sphere Response). This paper proposes a method to search the top layer of the SD search (spherical decoding) to achieve the following: In the QR decomposition stage, when the number of layers is greater than 2, the H matrix is ​​sorted by energy column by column (the number of transmit antennas corresponds to the number of columns), and the layer with the minimum energy is placed at the top layer of the SD search. The first Z matrix is ​​obtained by multiplying the conjugate transpose of the first Q matrix with the received signal. The first R matrix is ​​then normalized to obtain the first equalization R matrix. Based on the first Z matrix and the first equalization R matrix, the target equalization signal corresponding to the received signal is determined. This achieves the following in the preprocessing stage: a normalization method for the R matrix is ​​proposed, which greatly simplifies the computational load and implementation complexity of the search stage. The first ML (Maximum Likelihood) path search process is then executed based on the target equalization signal. The first ML path search process involves dividing all nodes at the top layer into B... Nt-1 There are several blocks, each containing a representative point and several extended points. Path detection is performed on each representative point, and the block with the smallest metric, Ka, is selected. Nt-1 A representative point is selected, and the block corresponding to the selected representative point is taken as the surviving block. Path detection is performed on the expansion points near the representative point in each surviving block, and the N1 branches with the smallest metric are selected as the surviving paths. This allows the ML path search stage in this embodiment to only search a portion of the required constellation points through a top-level block search, rather than searching all constellation points (i.e., all nodes) at the top level. This effectively reduces the search complexity for modulation orders greater than 256QAM (Quadrature Modulation). This invention reduces the computational complexity of the spherical decoding algorithm by simplifying the decoding computation in MIMO (Amplitude Modulation) application scenarios. Furthermore, by performing an ML complement path search on all surviving paths, the SD search process proposed in this application includes both ML path search and ML complement path search stages. This effectively reduces the overall search complexity for application scenarios supporting more than two layers, thereby further reducing the computational complexity of the spherical decoding algorithm. For the SD detection process of MIMO, this application provides a pipelined fusion architecture, effectively reducing hardware implementation resources and power consumption while meeting real-time and throughput requirements. This achieves the goal of maintaining decoding and detection performance while reducing the computational complexity of the spherical decoding algorithm. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating the first embodiment of the spherical decoding detection method of this application;

[0024] Figure 2 This is a flowchart illustrating the second embodiment of the spherical decoding detection method of this application;

[0025] Figure 3 This is a schematic flowchart of the SD detection process for MIMO in one embodiment;

[0026] Figure 4 This is a schematic diagram of QR decomposition in an embodiment of this application when the number of system layers is equal to two;

[0027] Figure 5 This is a schematic diagram illustrating a QR decomposition implementation scenario with different system layers in one embodiment of this application;

[0028] Figure 6 This is a schematic diagram of the preprocessing process in the embodiments of this application;

[0029] Figure 7 This is a schematic diagram of the SD search process in an embodiment of this application;

[0030] Figure 8 This is a diagram illustrating the implementation architecture of the search and soft bit computation in the embodiments of this application.

[0031] Figure 9 This is a schematic diagram of the functional modules of the spherical decoding and detection device according to an embodiment of this application;

[0032] Figure 10 This is a schematic diagram of the structure of a wireless receiving device supporting MIMO-OFDM in one embodiment;

[0033] Figure 11 This is a schematic diagram of the hardware structure of the electronic device involved in the embodiments of this application.

[0034] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0035] It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0036] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0037] It should be noted that all directional indicators (such as up, down, left, right, front, back, etc.) in the embodiments of this application are only used to explain the relative positional relationship and movement of each component in a certain specific posture (as shown in the figure). If the specific posture changes, the directional indicator will also change accordingly.

[0038] In this application, unless otherwise expressly specified and limited, the terms "connection," "fixed," etc., should be interpreted broadly. For example, "fixed" can mean a fixed connection, a detachable connection, or an integral part; it can mean a mechanical connection or an electrical connection; it can mean a direct connection or an indirect connection through an intermediate medium; it can mean the internal communication of two components or the interaction between two components, unless otherwise expressly limited. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0039] Furthermore, the use of terms such as "first" and "second" in this application is for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but only on the basis of being achievable by those skilled in the art. If the combination of technical solutions is contradictory or impossible to implement, such a combination of technical solutions should be considered non-existent and not within the scope of protection claimed in this application.

[0040] Currently, spherical decoding algorithms still have many shortcomings, affecting their practical applications. The overall search complexity of existing spherical decoding algorithms remains high. For example, with the increase in the number of support layers (transmit antennas) and modulation order in 4G / 5G, the computational load of spherical decoding algorithms will increase exponentially, which is detrimental to engineering implementation and low-cost requirements. Therefore, how to reduce the computational complexity of spherical decoding algorithms while maintaining decoding and detection performance is an urgent problem to be solved.

[0041] Based on this, embodiments of this application provide a spherical decoding detection method, referring to... Figure 1 , Figure 1This is a flowchart illustrating an embodiment of a spherical decoding detection method according to this application. In this embodiment, the spherical decoding detection method includes:

[0042] Step S10: After determining that the current system layer number is greater than two, perform orthogonal triangular QR decomposition on the first channel response matrix to obtain the first Q matrix and the first R matrix. The columns in the first Q matrix are arranged according to the magnitude of the channel energy, and the column with the smallest channel energy is placed at the top layer of the SD search.

[0043] In this embodiment, as those skilled in the art will know, the first Q matrix is ​​a unitary matrix and the first R matrix is ​​an upper triangular matrix.

[0044] Step S20: Multiply the conjugate transpose of the first Q matrix with the received signal to obtain the first Z matrix; normalize the first R matrix to obtain the first equalization R matrix; and determine the target equalization signal corresponding to the received signal based on the first Z matrix and the first equalization R matrix.

[0045] In this embodiment, by proposing a normalization method for the first R matrix in the preprocessing stage, the computational load and implementation complexity of the search stage are greatly simplified.

[0046] Specifically, the step of normalizing the first R matrix to obtain the first balanced R matrix includes:

[0047] Step A10: Multiply the first R matrix by the constellation point modulation amplitude factor to obtain the normalized first R matrix, and use the normalized first R matrix as the first equalization R matrix.

[0048] After step S20, step S30 is executed, in which the first ML path search process is performed according to the target equalization signal.

[0049] In this embodiment, the first ML (Maximum Likelihood) path search process involves dividing all nodes at the top level into B... Nt-1 There are several blocks, each containing a representative point and several extended points. Path detection is performed on each representative point, and the block with the smallest metric, Ka, is selected. Nt-1 Given N1 representative points, the blocks corresponding to these representative points are taken as surviving blocks. Path detection is then performed on the expansion points near the representative points within each surviving block. The N1 branches with the smallest metrics are selected as surviving paths, where B... Nt-1 Ka Nt-1 Both N1 and B are natural numbers, and B Nt-1 >Ka Nt-1 .

[0050] It should be noted that the path detection performed on each representative point, selecting the Ka with the smallest metric, is described above. Nt-1 The term "representative points" refers to expanding the paths at each representative point and arranging the generated branches in ascending order of metric, selecting the top Ka points. Nt-1 The representative point of each branch. The step of performing path detection on the expansion points near the representative point in each surviving block and selecting the N1 branches with the smallest metric as surviving paths refers to performing path expansion on the expansion points near the representative point in each surviving block, arranging all generated branches in ascending order of metric, and selecting the first N1 branches as surviving paths.

[0051] In this embodiment, if there are duplicate extended points corresponding to different representative points, the duplicate extended points are only calculated once, thus avoiding duplicate calculations. In one embodiment of the present invention, when dividing blocks, the constellation points within the blocks are divided as evenly as possible according to the constellation diagram.

[0052] Furthermore, the metric is Euclidean distance.

[0053] It should be noted that the metric described in this invention can be characterized not only by Euclidean distance, but also by other simplified Euclidean distance formulas. The calculation results of the simplified Euclidean distance formulas are similar to those of the Euclidean distance formulas, but the square calculation in the Euclidean distance formula is replaced by other calculation methods that are easy to implement in hardware.

[0054] In this embodiment, the step of dividing all nodes at the top level into B Nt-1 The steps for each block include:

[0055] Step B10: Generate a constellation diagram based on the target equalization signal, and divide all nodes of the top layer into B groups according to the constellation diagram. Nt-1 Each block.

[0056] Specifically, the step of performing path detection on the extended points near the representative point in each surviving block includes:

[0057] Step C10: Expand the search for expansion points around the representative point in each surviving block, wherein the expansion points in each surviving block expand uniformly outward from the representative point.

[0058] After step S30, step S40 is executed to perform ML complement path search on all surviving paths to obtain ML complement paths.

[0059] For example, the step of performing ML complement path search on all surviving paths to obtain the ML complement path includes:

[0060] Step D10: For each layer below the top layer, perform path expansion on all the surviving paths, and select the branch with the smallest metric from the branches generated by the path expansion as the ML complement path.

[0061] In this embodiment, those skilled in the art will understand that path expansion means: for nodes above the layer where the complement is found, nodes of surviving paths in path detection can be reused; while for nodes below the layer, the same node selection principle as in finding the ML path is used to expand the path until the leaf node.

[0062] After step S40, step S50 is executed, where the surviving path with the smallest metric is taken as the ML path, and the likelihood ratio information of each bit of each symbol in each layer is obtained based on the ML path and the ML complement path.

[0063] Currently, most publicly available methods support two-layer detection. However, with the increase in the number of supported layers in 4G / 5G from two to four or even eight layers, and the improvement of modulation schemes including QPSK, 16QAM, 64QAM, 256QAM, and even 1024QAM and 4096QAM, the computational load increases exponentially with the number of supported layers, while the computational complexity increases linearly with the increase of modulation schemes. This is not conducive to engineering implementation and low-cost requirements. In other words, the complexity of traditional spherical decoding increases exponentially with antenna configuration and modulation order.

[0064] Therefore, this embodiment of the application, after determining that the current system layer number is greater than two, performs orthogonal triangular QR decomposition on the first channel response matrix to obtain a first Q matrix and a first R matrix. The columns of the first Q matrix are arranged according to the channel energy, with the column of minimum channel energy placed at the top layer of the SD (Sphere Decoding) search. This achieves a method in the QR decomposition stage where, when the layer number is greater than two, the H matrix is ​​sorted by energy column by column (the number of transmit antennas corresponds to the number of columns), placing the layer with the minimum energy at the top layer of the SD search. Furthermore, by multiplying the conjugate transpose of the first Q matrix with the received signal, a first Z matrix is ​​obtained. The first R matrix is ​​normalized to obtain a first equalization R matrix. Based on the first Z matrix and the first equalization R matrix, the target equalization signal corresponding to the received signal is determined. This achieves a normalization method for the R matrix in the preprocessing stage, greatly simplifying the computational load and implementation complexity of the search stage. Finally, a first ML (Maximum Likelihood) path search process is executed based on the target equalization signal. The first ML path search process involves dividing all nodes at the top layer into B... Nt-1 There are several blocks, each containing a representative point and several extended points. Path detection is performed on each representative point, and the block with the smallest metric, Ka, is selected. Nt-1A representative point is selected, and the block corresponding to the selected representative point is taken as the surviving block. Path detection is performed on the expansion points near the representative point in each surviving block, and the N1 branches with the smallest metric are selected as the surviving paths. This allows the ML path search stage in this embodiment to only search for some of the required constellation points through a top-level block search method, instead of searching for all constellation points (i.e., all nodes) at the top level. This effectively reduces the search complexity for modulation orders greater than 256QAM (Quadrature Amplitude). This invention reduces the computational complexity of the spherical decoding algorithm by simplifying the decoding computation in MIMO (quadrature amplitude modulation) application scenarios. Furthermore, by performing ML complement path search on all surviving paths, the ML complement path is obtained. This allows the present invention to propose an SD search process including ML path search and ML complement path search stages during the search phase. This effectively reduces the overall search complexity for application scenarios supporting more than two layers, thereby further reducing the computational complexity of the spherical decoding algorithm. For the SD detection process of MIMO, this invention provides a pipelined fusion architecture, effectively reducing hardware implementation resources and power consumption while meeting real-time and throughput requirements. This achieves the goal of reducing the computational complexity of the spherical decoding algorithm while maintaining decoding and detection performance.

[0065] In one possible implementation, after the step of determining the target equalization signal corresponding to the received signal based on the first Z matrix and the first equalization R matrix, the method further includes:

[0066] Step E10: If it is determined that the modulation mode of the current system is the first modulation mode, then execute the step of performing the first ML path search process according to the target equalization signal, wherein the modulation order of the first modulation mode is greater than or equal to a preset value.

[0067] Step E20: If the modulation mode of the current system is determined to be the second modulation mode, then the second ML path search process is executed according to the target equalization signal. The second ML path search process is to perform path detection on all nodes of the top layer and select the N2 branches with the smallest metric as surviving paths. The modulation order of the second modulation mode is less than a preset value, and N2 is a natural number. Then, the step of performing ML complement path search on all surviving paths to obtain the ML complement path is executed.

[0068] In this embodiment, the modulation order of the first modulation method is greater than or equal to a preset value. The modulation order of the second modulation method is less than a preset value. This preset value can be set by those skilled in the art according to actual conditions, and this embodiment does not impose a specific limitation. In one embodiment, the preset value is 256. In another embodiment, the preset value is 64.

[0069] This embodiment proposes a method for SD (Sphere Array) modulation schemes that, when the current system's modulation scheme is determined to be a first modulation scheme, executes the step of performing a first ML path search process based on the target equalization signal, wherein the modulation order of the first modulation scheme is greater than or equal to a preset value; and when the current system's modulation scheme is determined to be a second modulation scheme, executes a second ML path search process based on the target equalization signal. The spherical decoding search process, when the modulation order is greater than or equal to a preset value (e.g., 256QAM), belongs to a high-order modulation scheme (i.e., the first modulation scheme), and its spherical decoding computation is relatively large. Therefore, the computation is reduced by using a top-level block search. When the modulation order is less than the preset value, it belongs to a low-order modulation scheme (i.e., the second modulation scheme), and its spherical decoding computation is relatively small. Therefore, the top-level block search is not required; that is, all constellation points are searched at the top level without block division. This achieves both a reduction in the computation of spherical decoding and effective assurance of decoding and detection performance. Consequently, the SD search process of this embodiment can approach the theoretical ML detection performance and can cover the detection methods for MIMO under the increasing number of layers and modulation schemes in current and future wireless communication systems, including 5G evolution, Wi-Fi evolution, and current 4G and 5G. At the same time, it has low complexity, which is beneficial for hardware implementation and effectively reduces hardware implementation resources and power consumption.

[0070] In one possible implementation, please refer to Figure 2 , Figure 2 This is a flowchart illustrating a second embodiment of the spherical decoding detection method of this application. The method further includes:

[0071] Step S61: After determining that the current system layer number is equal to two, perform column swapping on the second channel response matrix to obtain the column-swapped second channel response matrix. Then, perform orthogonal triangular QR decomposition on the second channel response matrix before column swapping to obtain the second Q matrix and the second R matrix. Perform orthogonal triangular QR decomposition on the column-swapped second channel response matrix to obtain the second Q matrix. s Matrix and second R s matrix;

[0072] Step S62: Multiply the conjugate transpose of the second Q matrix with the received signal to obtain the second Z matrix; normalize the second R matrix to obtain the second equalization R matrix; and determine the first equalization signal corresponding to the received signal based on the second Z matrix and the second equalization R matrix.

[0073] Step S63, the second Q s Multiplying the conjugate transpose of the matrix by the received signal yields the second Z.s Matrix, for the second R s The matrix is ​​normalized to obtain the second equilibrium R. s Matrix, and according to the second Z s Matrix and Second Equilibrium R s The matrix determines the second equalization signal corresponding to the received signal;

[0074] Step S64: Perform the first exchange ML path search process based on the first equalization signal and the second equalization signal;

[0075] In this embodiment, the first exchange ML path search process is to divide all nodes at the top level into B1 blocks, each block containing a representative point and several extension points, perform path detection on each representative point, select the Ka1 representative points with the smallest metric, take the block corresponding to the selected representative point as the surviving block, perform path detection on the extension points near the representative point in each surviving block, and select the N3 branches with the smallest metric as the ML path, where B1, Ka1 and N3 are all natural numbers, and B1>Ka1.

[0076] Step S65: Based on the ML path, obtain the likelihood ratio information of each bit of each symbol in each layer.

[0077] This embodiment, after determining that the current system layer number is equal to two, performs column swapping on the second channel response matrix to obtain the column-swapped second channel response matrix. Then, it performs orthogonal triangular QR decomposition on the second channel response matrix before column swapping to obtain the second Q matrix and the second R matrix. Finally, it performs orthogonal triangular QR decomposition on the column-swapped second channel response matrix to obtain the second Q matrix. s Matrix and second R s The second Q matrix is ​​then multiplied by its conjugate transpose with the received signal to obtain the second Z matrix. The second R matrix is ​​then normalized to obtain the second equalization R matrix. Based on the second Z matrix and the second equalization R matrix, the first equalization signal corresponding to the received signal is determined. Finally, the second Q matrix is... s Multiplying the conjugate transpose of the matrix by the received signal yields the second Z. s Matrix, for the second R s The matrix is ​​normalized to obtain the second equilibrium R. s Matrix, and according to the second Z s Matrix and Second Equilibrium R sThe matrix is ​​used to determine the second equalization signal corresponding to the received signal. Therefore, a search procedure for SD (Search for Means of Detection) is proposed for different numbers of layers. When the number of system layers is greater than 2, in the QR decomposition, a method is proposed to sort the H matrix column by column (i.e., Nt) according to energy, placing the layer with the minimum energy at the top of the SD search. When the number of system layers is equal to 2, a method is proposed to perform QR decomposition by exchanging the H matrix column by column, enabling ML (Mechanical, Multi-Layer) search for SD in both layers. Furthermore, this embodiment proposes a low-complexity search procedure that meets practical application requirements for different numbers of layers during the search phase. For example, when the number of system layers is greater than 2, a sampled ML search and ML complement search SD search procedure is used. When the number of system layers is equal to 2, a sampled and swapped ML search SD search procedure is used, thereby effectively reducing hardware implementation resources and power consumption, meeting the requirements of real-time performance and throughput. Similarly, the architecture of this embodiment is flexible and easily expandable, further enabling the SD search procedure of this embodiment to approach the theoretical ML detection performance, covering the detection methods for MIMO under the current and future wireless communication systems with continuously increasing layers and modulation schemes.

[0078] Furthermore, in one implementable manner, in the case of the second Z... s Matrix and Second Equilibrium R s After determining the matrix and the second equalization signal corresponding to the received signal, the method further includes:

[0079] Step F10: If it is determined that the modulation mode of the current system is the first modulation mode, then execute the step of performing the first switching ML path search process based on the first equalization signal and the second equalization signal;

[0080] Step F20: If it is determined that the modulation mode of the current system is the second modulation mode, then the second switching ML path search process is executed according to the first equalization signal and the second equalization signal.

[0081] In this embodiment, the second exchange ML path search process is to perform path detection on all nodes of the top layer, select the N4 branches with the smallest metric as ML paths, where N4 is a natural number; and execute the step of obtaining the likelihood ratio information of each bit of each symbol in each layer based on the ML path.

[0082] This embodiment proposes a method for SD (Sphere Array) modulation schemes that, when the current system's modulation scheme is determined to be the first modulation scheme, executes the step of performing a first switching ML path search procedure based on the first and second equalization signals; and when the current system's modulation scheme is determined to be the second modulation scheme, executes a second switching ML path search procedure based on the first and second equalization signals. The spherical decoding search process, when the modulation order is greater than or equal to a preset value (e.g., 256QAM), belongs to a high-order modulation scheme (i.e., the first modulation scheme), and its spherical decoding computation is relatively large. Therefore, the computation is reduced by using a top-level block search. When the modulation order is less than the preset value, it belongs to a low-order modulation scheme (i.e., the second modulation scheme), and its spherical decoding computation is relatively small. Therefore, the top-level block search is not required; that is, all constellation points are searched at the top level without block division. This achieves both a reduction in the computation of spherical decoding and effective assurance of decoding and detection performance. Consequently, the SD search process of this embodiment can approach the theoretical ML detection performance and can cover the detection methods for MIMO under the increasing number of layers and modulation schemes in current and future wireless communication systems, including 5G evolution, Wi-Fi evolution, and current 4G and 5G. At the same time, it has low complexity, which is beneficial for hardware implementation and effectively reduces hardware implementation resources and power consumption.

[0083] To aid in understanding the technical concept or principle of the embodiments of this application, a specific embodiment of a spherical decoding detection method is provided:

[0084] First, refer to Figure 3 As will be understood by those skilled in the art, the general MIMO SD detection process is as follows:

[0085] 1. After the channel estimation process is completed, the output is an Nr*Nt dimensional H matrix (i.e., the channel response matrix), where Nr represents the receiving antenna and Nt represents the transmitting antenna, and the output is No (noise value); Y is the receiving antenna data (i.e., the received signal), which is an Nr*1 matrix;

[0086] 2. The QR decomposition module performs QR decomposition (i.e., orthogonal triangular QR decomposition) on the H matrix, outputting the Q matrix and R matrix;

[0087] 3. The preprocessing module multiplies the Q transpose (i.e., the conjugate transpose of the Q matrix) with Y to output the Z matrix, and multiplies the R matrix with c (constellation point modulation amplitude factor) to output the normalized R matrix (i.e., the equalized R matrix).

[0088] 4. The search module uses the input Z matrix and the normalized R matrix to perform the search and outputs the metric value of each bit.

[0089] 5. The soft bit calculation module generates soft bit information (i.e., the likelihood ratio information for each bit) based on the metric value.

[0090] The method and apparatus flow for MIMO detection in this embodiment also includes: QR decomposition, preprocessing, search, and soft bit calculation;

[0091] In this embodiment, we assume a MIMO-OFDM (Multiple Input Multiple Output-Orthogonal Frequency Division Multiplexing) system with Nt transmit antennas and Nr receive antennas, satisfying Nt≤Nr. The received signal model can then be expressed as: Y=H*X+N, where Y is the Nr*1 received signal matrix; H is the Nr*Nt channel estimation matrix; X is the Nr*1 transmit signal matrix; and N is the noise power value.

[0092] Step 1: QR decomposition (i.e., orthogonal triangular QR decomposition)

[0093] The H matrix of Nr*Nt output from the channel estimation is sorted and decomposed into QR such that H = Q*R, where the Q matrix has the property that the product of the transpose of Q and itself equals the identity matrix.

[0094] In the implementation structure, by Figure 2 The QR decomposition module completes the QR decomposition of the H matrix of Nr*Nt, and outputs the Q matrix and R matrix.

[0095] In actual implementation, the following differences exist for different Nt values:

[0096] When Nt = 2 (i.e., the number of transmit antennas is equal to 2, since the number of transmit antennas corresponds to the number of system layers, i.e., the number of system layers is equal to 2), H (channel response matrix) is a two-column matrix [H0, H1], where H0 and H1 are both column vectors of depth Nr. QR decomposition is performed before and after column swapping to ensure the performance of the searched layer in subsequent searches. QR decomposition does not require sorting by column energy. QR decomposition of [H0, H1] outputs the Q matrix and R matrix, and QR decomposition of [H1, H0] outputs the Qs matrix and Rs matrix. The Q and Qs matrices are Nr*2 matrices, and R and Rs are 2*2 upper triangular matrices, as shown below. Figure 4 As shown.

[0097] When Nt > 2 (i.e., the number of transmit antennas is greater than 2, since the number of transmit antennas corresponds to the number of system layers, i.e., the number of system layers is greater than 2), QR decomposition after column swapping of the H matrix is ​​not required. QR decomposition requires sorting by column energy to ensure the output Q matrix is ​​[Q0, Q1, ..., Q...].Nt-1 At that time, Q Nt-1 The column has the lowest energy; the R matrix is ​​an upper triangular matrix of Nt*Nt.

[0098] Taking the scenario of Nt≤4 as an example, describe the state flow of the QR decomposition module (e.g.) Figure 5 As shown):

[0099] After startup, check Nt. If Nt = 4, enter stage0, sort the energy of the 4 columns of H matrix and swap the columns, calculate the first column of Q matrix and the first row of R matrix, then enter state1, sort the energy of the remaining 3 columns and swap the columns, calculate the second column of Q matrix and the second row of R matrix, and so on, enter stage2 and stage3, and finally return to idle, outputting the final Q matrix and R matrix.

[0100] If Nt = 3, then start directly from stage1 and complete the subsequent steps in sequence;

[0101] If Nt = 2, then proceed to 2L_stage0, calculate the first column of the Q matrix and the first row of the R matrix obtained from the QR decomposition of the H matrix, and simultaneously calculate the first column of the Qs matrix and the first row of the Rs matrix obtained from the QR decomposition of the swapped H matrix; then proceed to 2L_stage1, calculate the second column of the Q matrix and the second row of the R matrix obtained from the QR decomposition of the H matrix, and simultaneously calculate the second column of the Qs matrix and the second row of the sR matrix obtained from the QR decomposition of the swapped H matrix, finally return to idle, and output the final Q matrix, Qs matrix, R matrix, and Rs matrix.

[0102] Step 2: Preprocessing

[0103] After step one, we can obtain the formula Y = Q * R * Xp + N, where Xp is the result of reordering X according to the sorting in the QR decomposition.

[0104] Furthermore, Q T Y = Q T Q*R*Xp+Q T *N, we get Z=R*Xs+W, where Z=Q T Y, W = Q T N, R = R * c, Xp extracts the c factor to obtain Xs.

[0105] The preprocessed structure in the implementation structure is as follows: Figure 6 As shown, Figure 6 In this process, Z (i.e., the Z matrix) is generated to perform the operation of multiplying the transpose of the Q matrix by the Y matrix, and the operation of multiplying the transpose of the Qs matrix by the Ys matrix, and outputting the Z matrix and the Zs matrix.

[0106] R (i.e., R matrix) and Rs (i.e., Rs matrix) are normalized by multiplying each element of the R matrix and Rs matrix by c, where c is the modulation amplitude factor of the constellation point, and outputting the normalized R matrix and Rs matrix.

[0107] R's preprocessing ensures that the subsequent search process is limited to integers, greatly simplifying the computational cost and implementation complexity of the search.

[0108] Step 3: Search + Soft Bit Calculation

[0109] Expanding the matrix further, we get the following formula:

[0110]

[0111] The basic idea of ​​the SD search process is as follows: Given Z and R, Xs takes values ​​within a range defined by the modulation scheme. The goal is to find the Xs that minimizes dst, i.e., for each ||. 2 All are minimum values, where dst is called the metric, such as Figure 7 As shown. Figure 7 The basic search process is given:

[0112] Define xs Nt-1 The total number of constellation points required is K. Nt-1 When xs Nt-2 When searching only for the optimal point, set T=1; when xs Nt-2 When searching for the optimal and near-optimal points, T = 2 is set. The search process is described as follows:

[0113] 1. Traverse K Nt-1 xs Nt-1 Find all the required constellation points and calculate dst. Nt-1 ;

[0114] 2. In each xs Nt-1 Below, the search makes dst Nt-2 xs at its minimum value Nt-2 Constellation point; when T=2, the search makes dst Nt-2 xs when it is the second smallest value Nt-2 Find the constellation points and output dst. Nt-2 ;

[0115] 3. Using xs obtained from item 2 Nt-1 and XS Nt-2 Yes, the search makes dst Nt-3 xs at its minimum value Nt-3 Find the constellation points and output dst. Nt-3 ;

[0116] 4. Continue this process until the constellation point xs0 that minimizes dst0 is found, and then output dst0.

[0117] 5. Ultimately, you will get K. Nt-1 *T metrics,

[0118] Right now and the corresponding K Nt-1 *T Xs.

[0119] In this embodiment, the multiple searches in the SD search described below, such as the ML search and the ML complement search, are all performed according to... Figure 7 The given basic search process is executed.

[0120] It should be noted that, in actual implementation, the SD search process differs for different numbers of layers and modulation schemes in this application embodiment:

[0121] (1) Steps for Nt>2

[0122] When Nt>2, the SD search adopts the ML search + ML complement search process, xs Nt-1 Layer sampling ML search, xs Nt-1 All layers below layer 1 are sampled using ML complement search. This distinguishes between different modulation schemes:

[0123] When the modulation order is ≥256QAM (i.e., the preset value is 256QAM, or in other words, the first modulation method is a modulation method with a modulation order ≥256QAM), ML searches T. ML Set to 1, define the number of constellation points as Qm, Qm≥256, if K is selected Nt-1 =Qm, that is, in xs Nt-1 Searching all constellation points in a layer-wise manner would involve a huge amount of computation. In practice, in xs Nt-1 The layer will select K from Qm. Nt-1 Search by constellation point, K Nt-1 The method for selecting constellation points is as follows:

[0124] xs Nt-1 The constellation points of layer Qm are divided into B in total. Nt-1 There are b blocks, and the number of constellation points within each block is b. Nt-1 From B Nt-1 Ka is selected from each block. Nt-1 One block, Ka Nt-1 xs of the constellation points covered by the block Nt-1 Layer search, i.e., selecting K Nt-1 =Ka Nt-1 *b Nt-1 ;

[0125] Ka Nt-1 Method for selecting individual blocks: in xs Nt-1 Layer from B Nt-1In each block, one constellation point is selected for searching, resulting in B. Nt-1 One metric, from B Nt-1 The smallest Ka is selected from the metric values. Nt-1 Locate the surviving block corresponding to each metric value;

[0126] In XS Nt-1 Layer complete K Nt-1 After searching for each constellation point, we obtain Kt. Nt-1 =K Nt-1 *T ML Each metric, and Kt Nt-1 One Xs. For Xs Nt-1 Each bit of the constellation layer is from Kt Nt-1 The smallest metric value that makes each bit 0 and 1 is retained from the metric values ​​and is used for calculation with subsequent soft bits;

[0127] ML search only for xs Nt-1 The search for the required constellation points for the layer was not performed on xs. Nt-1 The search is performed on the constellation points required for the next layer, even if the ML search yields Kt. Nt-1 There are Xs, where xs Nt-2 Reaching xs0 may not be the final constellation point required, therefore xs... Nt-2 Continue the ML complement search from layer to layer xs0, T MLC Set to 1;

[0128] During the ML search phase, from Kt Nt-1 Among the metric values, the Nitr smallest metric value and its corresponding Nitr Xs are selected and defined as the surviving paths. j represents the layer number of the complement search, j∈[0,NT-2]. During the ML complement search phase, based on the constellation points of the xsj layer and xs stored in the Nitr surviving paths... Nt-1 To x sj+1 The constellation points, according to Figure 7 Search xs j The constellation points near the layer are obtained through a table lookup, and their number is related to xs. j The modulation method of the layer is related to m respectively. j Define xs j The number of constellation points searched using the layer complement set is Kt. j If xs j Kt number of ML complement searches for the layer j =Nitr*m j *T MLC , obtain Kt j One metric, for xs j Layer from Kt jThe smallest metric that makes each bit of the constellation point both 0 and 1 is retained from the metric values ​​and is then used to calculate the metric using subsequent soft bits.

[0129] When the modulation order is 64QAM, 16QAM, or QPSK (i.e., the preset value is 256QAM, or in other words, the second modulation method is a modulation method with a modulation order less than 256QAM), let K... Nt-1 =Qm, where Qm = 64, 16, or 4, that is, in xs Nt-1 A layer-wise search of all constellation points is performed, without the need for block division. ML Set to 2, T MLC Set it to 1, and the rest of the search process is the same as described above.

[0130] (2) Steps for Nt = 2

[0131] When Nt=2, the SD search uses a swapped ML search;

[0132] When the modulation order is ≥256QAM, ML search T ML Setting it to 1 defines the number of constellation points as Qm, where Qm ≥ 256. If K1 = Qm is selected, meaning all constellation points are searched in layer xs1, the computational load would be very high. In practice, layer xs1 will select K1 constellation points from Qm for the search. The method for selecting K1 constellation points is as follows:

[0133] Divide the Qm constellation points in the xs1 layer into B1 blocks, with b1 constellation points in each block. Select Ka1 blocks from the B1 blocks. Search the constellation points covered by the Ka1 blocks in the xs1 layer. That is, select K1 = Ka1 * b1.

[0134] The method for selecting Ka1 blocks is as follows: In the xs1 layer, select one constellation point from each of the B1 blocks to search and obtain B1 metric values. Select the block corresponding to the smallest Ka1 metric value from the B1 metric values ​​to locate the surviving block.

[0135] After completing the search for K1 constellation points in layer xs1, we obtain Kt1 = K1 * T. ML Unlike Nt>2, there is no need to retain Kt1 Xs values. For each bit of the xs1 layer constellation point, the minimum metric value that makes each bit 0 and 1 is retained from Kt1 metric values ​​and calculated using subsequent soft bits;

[0136] Using a similar ML complement search (Nt>2) to search xs0 does not achieve the desired performance. Therefore, the ML search for the xs1 layer is also applied to the ML search for xs0, as described above. This will not be elaborated further here.

[0137] When the modulation order is 64QAM, 16QAM, or QPSK, let K1 = Qm and K0 = Qm, where Qm = 64, 16, or 4. Perform an ML search on all constellation points in layers xs1 and xs0; block division is not required. ML Set it to 1, and the rest of the search process is the same as described above.

[0138] In this embodiment, the soft bit is defined as LLR, and the soft bit is calculated as follows:

[0139]

[0140] Where i≤log2(QM), j≤Nt-1, i and j are both integers; D0[j][i] represents the minimum metric value where the i-th bit of layer j is 0, and D1[j][i] represents the minimum metric value where the i-th bit of layer j is 1.

[0141] In this embodiment, the implementation architecture for search and soft-bit calculation is given according to the above process as follows: Figure 8 As shown. Figure 8 The structure is described as follows:

[0142] Figure 8 In this context, LUT stands for Lookup Table. It's worth noting that... Figure 8 The architecture simultaneously supports ML search + ML complement search required when Nt>2, as well as swap method search when Nt=2.

[0143] The LUT+CAL is the basic search unit. During the search, the constellation points to be searched are read from the LUT and sent to the CAL to output the measurement value. The entire search process is completed by instantiating the basic search unit in the architecture. The LUT pre-stores the constellation points to be searched, and the CAL outputs the measurement value of the searched constellation points.

[0144] ml search phase 1: Complete xs Nt-1 Search between constellation point blocks in the layer, and output the metric value, block number and constellation point to the storage module;

[0145] ml search phase 2: Read K from storage Nt-1 One surviving block, completing xs Nt-1 The search within the surviving blocks of the constellation points is then performed to complete the xs. Nt-1 The layer searches for the constellation points it needs, and outputs the measurement values ​​and constellation points to the storage module.

[0146] ML complement search: Read the preserved Nitr surviving paths from storage, and use the surviving path information (constellation points) to look up the LUT to complete the XS. Nt-2 Perform an ML complement search on the constellation points from layer xs0, and output the metric and constellation points to the storage module.

[0147] The storage module is used to store various types of information during the search process:

[0148] In ml1 search phase 1, compare the K surviving blocks with the smallest retention metric;

[0149] In ml1 search phase 1 and ml1 search phase 2, the information of the surviving Nitr path with the smallest retained metric is compared;

[0150] The minimum metric value when each bit of the constellation points in all layers is 0 and 1 throughout the search phase;

[0151] Soft bit calculation: Read the minimum metric value when each bit at each constellation point is 0 and 1 from the storage, and calculate the soft bit;

[0152] The control module schedules each search unit based on configuration information such as Nt and Qm, and completes the entire search process in a pipeline manner.

[0153] In this embodiment, it should be noted that:

[0154] 1. When Nt>2, the SD search adopts the ML search + ML complement search process:

[0155] When the modulation scheme is ≥256QAM, the ml1 search stage 1 is driven sequentially to complete inter-block search → storage → ml1 search stage 2 intra-block search → storage → ml complement search → storage → soft bit calculation;

[0156] When the modulation scheme is 64QAM, the ml1 search stage 1 is driven sequentially to complete the partial constellation point search → storage → ml1 search stage 2 to search the remaining constellation points → storage → ml complement search → storage → soft bit calculation;

[0157] When the modulation scheme is 16QAM or QPSK, the ml1 search stage 1 is driven sequentially to complete constellation point search → storage → ml complement search → storage → soft bit calculation.

[0158] II. When Nt = 2, the SD search uses the exchange ML search path (search without ML complement):

[0159] When the modulation scheme is ≥256QAM, the ml1 search stage 1 is driven sequentially to complete inter-block search → storage → ml1 search stage 2 intra-block search → storage → soft bit calculation.

[0160] When the modulation scheme is 64QAM, the ml1 search stage 1 is driven sequentially to complete the partial constellation point search → storage → ml1 search stage 2 to search the remaining constellation points → storage → soft bit calculation.

[0161] When the modulation scheme is 16QAM or QPSK, the ml1 search stage 1 is driven sequentially to complete constellation point search → storage → soft bit calculation.

[0162] The implementation structure of this specific embodiment is easily expandable. For example, the search operation volume of ml search stage 2 and ml complement search can be determined according to the required number of layers and modulation method, and the number of instantiated matching search units can be selected, which can be quickly implemented in engineering.

[0163] Furthermore, this specific embodiment proposes an SD search process supporting multiple layers and modulation schemes. This process can approximate the theoretical ML detection performance and can cover MIMO detection methods under the ever-increasing number of layers and modulation schemes in current and future wireless communication systems, including 5G evolution, Wi-Fi (Wireless Fidelity) evolution, and current 4G and 5G. It also features low complexity, which is beneficial for hardware implementation and effectively reduces hardware implementation resources and power consumption. Additionally, this specific embodiment proposes a low-complexity architecture for implementing this MIMO detection method. This architecture has low implementation cost and low power consumption, employing a pipelined fusion architecture that can cover the ever-increasing number of layers and modulation schemes in current and future MIMO systems, meeting the requirements of real-time performance and throughput. Similarly, this architecture is flexible, easily expandable, and allows for rapid engineering implementation.

[0164] It should be noted that the above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of protection of this application. Those skilled in the art can understand that all or part of the processes of the above embodiments can be implemented, and equivalent changes made in accordance with the claims of this invention are still within the scope of this application.

[0165] In other words, the above specific embodiments are only used to help understand the technical principles or concepts of the embodiments of this application, and do not constitute a limitation on this application. Any simple modifications based on this technical concept should be within the protection scope of this application.

[0166] Furthermore, this application also proposes a spherical decoding detection device, referring to... Figure 9 , Figure 9 This is a schematic diagram of the functional modules of an embodiment of a spherical decoding and detection device according to this application.

[0167] In this embodiment, the spherical decoding detection device includes:

[0168] QR decomposition module 10 is configured to perform orthogonal triangular QR decomposition on the first channel response matrix after determining that the current system layer number is greater than two, to obtain the first Q matrix and the first R matrix. The columns in the first Q matrix are arranged according to the magnitude of the channel energy, and the column with the smallest channel energy is set at the top layer of the SD search.

[0169] The equalization signal calculation module 20 is configured to multiply the conjugate transpose of the first Q matrix with the received signal to obtain the first Z matrix, normalize the first R matrix to obtain the first equalization R matrix, and determine the target equalization signal corresponding to the received signal based on the first Z matrix and the first equalization R matrix.

[0170] ML path search module 30 is configured to execute a first ML path search process based on the target equalization signal, wherein the first ML path search process involves dividing all nodes at the top level into B... Nt-1 There are several blocks, each containing a representative point and several extended points. Path detection is performed on each representative point, and the block with the smallest metric, Ka, is selected. Nt-1 Given N1 representative points, the blocks corresponding to these representative points are taken as surviving blocks. Path detection is then performed on the expansion points near the representative points within each surviving block. The N1 branches with the smallest metrics are selected as surviving paths, where B... Nt-1 Ka Nt-1 Both N1 and B are natural numbers, and B Nt-1 >Ka Nt-1 ;

[0171] The ML complement path search module 40 is configured to perform ML complement path search on all surviving paths to obtain the ML complement path;

[0172] The soft value information calculation module 50 is configured to take the surviving path with the smallest metric as the ML path, and to obtain the likelihood ratio information of each bit of each symbol in each layer based on the ML path and the ML complement path.

[0173] In some embodiments, the ML path search module 30 is further configured as follows:

[0174] If the modulation mode of the current system is determined to be the first modulation mode, then the following step is executed: performing the first ML path search process according to the target equalization signal, wherein the modulation order of the first modulation mode is greater than or equal to a preset value;

[0175] If the modulation scheme of the current system is determined to be the second modulation scheme, then the second ML path search process is executed according to the target equalization signal. The second ML path search process involves performing path detection on all nodes of the top layer, selecting the N2 branches with the smallest metric as surviving paths, where the modulation order of the second modulation scheme is less than a preset value, and N2 is a natural number; and then performing the step of performing ML complement path search on all surviving paths to obtain the ML complement path.

[0176] In some embodiments, the QR decomposition module 10 is further configured to:

[0177] After determining that the current system layer number is equal to two, the second channel response matrix is ​​column-swapped to obtain the column-swapped second channel response matrix. Then, the second channel response matrix before column swapping is decomposed into an orthogonal triangular QR decomposition to obtain the second Q matrix and the second R matrix. Finally, the second channel response matrix after column swapping is decomposed into an orthogonal triangular QR decomposition to obtain the second Q matrix. s Matrix and second R s matrix;

[0178] The equalization signal calculation module 20 is also configured as follows:

[0179] Multiply the conjugate transpose of the second Q matrix with the received signal to obtain the second Z matrix. Normalize the second R matrix to obtain the second equalization R matrix. Determine the first equalization signal corresponding to the received signal based on the second Z matrix and the second equalization R matrix.

[0180] The second Q s Multiplying the conjugate transpose of the matrix by the received signal yields the second Z. s Matrix, for the second R s The matrix is ​​normalized to obtain the second equilibrium R. s Matrix, and according to the second Z s Matrix and Second Equilibrium R s The matrix determines the second equalization signal corresponding to the received signal;

[0181] ML path search module 30 is also set as follows:

[0182] The first exchange ML path search process is executed according to the first equalization signal and the second equalization signal. The first exchange ML path search process is as follows: all nodes of the top layer are divided into B1 blocks, each block contains a representative point and several extension points, path detection is performed on each representative point, the Ka1 representative points with the smallest metric are selected, the blocks corresponding to the selected representative points are taken as surviving blocks, path detection is performed on the extension points near the representative points in each surviving block, and the N3 branches with the smallest metric are selected as ML paths. Here, B1, Ka1 and N3 are all natural numbers, and B1>Ka1.

[0183] The soft value information calculation module 50 is also configured as follows:

[0184] Based on the ML path, the likelihood ratio information for each bit of each symbol in each layer is obtained.

[0185] In some embodiments, the ML path search module 30 is further configured as follows:

[0186] If the modulation scheme of the current system is determined to be the first modulation scheme, then the following step is executed: performing the first switching ML path search process based on the first equalization signal and the second equalization signal;

[0187] If the modulation scheme of the current system is determined to be the second modulation scheme, then the second exchange ML path search process is executed according to the first equalization signal and the second equalization signal. The second exchange ML path search process is to perform path detection on all nodes of the top layer, select the N4 branches with the smallest metric as ML paths, where N4 is a natural number; and execute the step of obtaining the likelihood ratio information of each bit of each symbol of each layer according to the ML path.

[0188] In some embodiments, the ML complement path search module 40 is further configured as follows:

[0189] For each layer below the top layer, path expansion is performed on all surviving paths, and the branch with the smallest metric is selected from the branches generated by the path expansion as the ML complement path.

[0190] In some embodiments, the ML path search module 30 is further configured as follows:

[0191] A constellation diagram is generated based on the target equalization signal, and all nodes at the top layer are divided into B groups according to the constellation diagram. Nt-1 One block;

[0192] The step of performing path detection on the extended points near the representative point in each surviving block includes:

[0193] In each surviving block, the search expands to points around the representative point, where the expansion points in each surviving block expand uniformly outward from the representative point.

[0194] In some embodiments, the equalization signal calculation module 20 is further configured as follows:

[0195] Multiply the first R matrix by the constellation point modulation amplitude factor to obtain the normalized first R matrix, and use the normalized first R matrix as the first equalization R matrix.

[0196] The spherical decoding detection device provided in this embodiment belongs to the same inventive concept as the spherical decoding detection method provided in the above embodiments. Technical details not described in detail in this embodiment can be found in the embodiments of the above spherical decoding detection method. Furthermore, this embodiment has the same beneficial effects as the embodiments of the spherical decoding detection method, and will not be repeated here.

[0197] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0198] Commonly used wireless receivers that support MIMO-OFDM, such as Figure 10 As shown, the signal received from the receiving antenna is processed by the digital front-end and then subjected to FFT. The output frequency domain data is sent to the channel estimation module. The channel estimation module completes the processing and outputs the H matrix (channel impulse response matrix) to the MIMO detection module. The MIMO detection module completes the detection and outputs the LLR to the decoder. The decoder processes the data and outputs the decoded bits. The spherical decoding detection method of this application is mainly applied in the MIMO detection module of a wireless receiving device. This application embodiment is applicable to terminal receivers and also to base station receivers. Using this application embodiment can achieve MIMO detection performance close to ML while greatly reducing complexity, making MIMO detection feasible.

[0199] Reference Figure 11 , Figure 11 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Figure 11 As shown, the electronic device may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM), such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0200] Those skilled in the art will understand that Figure 11 The structures shown do not constitute a limitation on the electronic device and may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. Figure 11 As shown, the memory 1005, which serves as a storage medium, may include an operating system, a data storage module, a network communication module, a user interface module, and a spherical decoding detection program.

[0201] exist Figure 11In the electronic device shown, the network interface 1004 is mainly used for data communication with other devices; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in this embodiment can be set in the communication device. The communication device calls the spherical decoding detection program stored in the memory 1005 through the processor 1001 and executes the spherical decoding detection method provided in any of the above embodiments.

[0202] The terminal proposed in this embodiment belongs to the same inventive concept as the spherical decoding detection method proposed in the above embodiments. Technical details not described in detail in this embodiment can be found in any of the above embodiments. Furthermore, this embodiment has the same beneficial effects as the spherical decoding detection method.

[0203] Furthermore, embodiments of this application also propose a readable storage medium, which is a computer-readable storage medium, and the computer-readable storage medium can be a non-volatile computer-readable storage medium. The computer-readable storage medium stores a spherical decoding detection program, which, when executed by a processor, implements the spherical decoding detection method of this application as described above.

[0204] The various embodiments of the electronic device and computer-readable storage medium of this application can be referred to the various embodiments of the spherical decoding detection method of this application, and will not be repeated here.

[0205] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0206] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0207] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause an electronic device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0208] The above are merely preferred embodiments of this application and do not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A spherical decoding detection method, characterized in that, include: After determining that the current system layer number is greater than two, the first channel response matrix is ​​decomposed into an orthogonal triangular QR matrix to obtain the first Q matrix and the first R matrix. The columns of the first Q matrix are arranged according to the channel energy, and the column with the smallest channel energy is placed at the top layer of the SD search. Multiply the conjugate transpose of the first Q matrix with the received signal to obtain the first Z matrix. Normalize the first R matrix to obtain the first equalization R matrix. Determine the target equalization signal corresponding to the received signal based on the first Z matrix and the first equalization R matrix. If the modulation scheme of the current system is determined to be the first modulation scheme, a first ML path search process is executed according to the target equalization signal, wherein the first ML path search process involves dividing all nodes of the top layer into B... Nt-1 There are several blocks, each containing a representative point and several extended points. Path detection is performed on each representative point, and the block with the smallest metric, Ka, is selected. Nt-1 Given N1 representative points, the blocks corresponding to these representative points are taken as surviving blocks. Path detection is then performed on the expansion points near the representative points within each surviving block. The N1 branches with the smallest metrics are selected as surviving paths, where B... Nt-1 Ka Nt-1 Both N1 and B are natural numbers, and B Nt-1 >Ka Nt-1 Wherein, the modulation order of the first modulation method is greater than or equal to a preset value; If the modulation mode of the current system is determined to be the second modulation mode, then the second ML path search process is executed according to the target equalization signal. The second ML path search process is to perform path detection on all nodes of the top layer and select the N2 branches with the smallest metric as surviving paths. The modulation order of the second modulation mode is less than a preset value, and N2 is a natural number. Perform an ML complement path search on all surviving paths to obtain the ML complement path; The surviving path with the smallest metric is taken as the ML path. Based on the ML path and the ML complement path, the likelihood ratio information of each bit of each symbol in each layer is obtained.

2. The spherical decoding detection method as described in claim 1, characterized in that, The method further includes: After determining that the current system layer number is equal to two, the second channel response matrix is ​​column-swapped to obtain the column-swapped second channel response matrix. Then, the second channel response matrix before column swapping is decomposed into an orthogonal triangular QR decomposition to obtain the second Q matrix and the second R matrix. Finally, the second channel response matrix after column swapping is decomposed into an orthogonal triangular QR decomposition to obtain the second Q matrix. s Matrix and second R s matrix; Multiply the conjugate transpose of the second Q matrix with the received signal to obtain the second Z matrix. Normalize the second R matrix to obtain the second equalization R matrix. Determine the first equalization signal corresponding to the received signal based on the second Z matrix and the second equalization R matrix. The second Q s Multiplying the conjugate transpose of the matrix by the received signal yields the second Z. s Matrix, for the second R s The matrix is ​​normalized to obtain the second equilibrium R. s Matrix, and according to the second Z s Matrix and Second Equilibrium R s The matrix determines the second equalization signal corresponding to the received signal; The first exchange ML path search process is executed according to the first equalization signal and the second equalization signal. The first exchange ML path search process is as follows: all nodes of the top layer are divided into B1 blocks, each block contains a representative point and several extension points, path detection is performed on each representative point, the Ka1 representative points with the smallest metric are selected, the blocks corresponding to the selected representative points are taken as surviving blocks, path detection is performed on the extension points near the representative points in each surviving block, and the N3 branches with the smallest metric are selected as ML paths. Here, B1, Ka1 and N3 are all natural numbers, and B1>Ka1. Based on the ML path, the likelihood ratio information for each bit of each symbol in each layer is obtained.

3. The spherical decoding detection method as described in claim 2, characterized in that, According to the second Z s Matrix and Second Equilibrium R s After determining the matrix and the second equalization signal corresponding to the received signal, the method further includes: If the modulation scheme of the current system is determined to be the first modulation scheme, then the following step is executed: performing the first switching ML path search process based on the first equalization signal and the second equalization signal; If the modulation scheme of the current system is determined to be the second modulation scheme, then the second exchange ML path search process is executed according to the first equalization signal and the second equalization signal. The second exchange ML path search process is to perform path detection on all nodes of the top layer, select the N4 branches with the smallest metric as ML paths, where N4 is a natural number; and execute the step of obtaining the likelihood ratio information of each bit of each symbol of each layer according to the ML path.

4. The spherical decoding detection method as described in claim 1, characterized in that, The step of performing ML complement path search on all surviving paths to obtain the ML complement path includes: For each layer below the top layer, path expansion is performed on all surviving paths, and the branch with the smallest metric is selected from the branches generated by the path expansion as the ML complement path.

5. The spherical decoding detection method as described in claim 1, characterized in that, The top-level nodes are divided into B. Nt-1 The steps for each block include: A constellation diagram is generated based on the target equalization signal, and all nodes at the top layer are divided into B groups according to the constellation diagram. Nt-1 One block; The step of performing path detection on the extended points near the representative point in each surviving block includes: In each surviving block, the search expands to points around the representative point, where the expansion points in each surviving block expand uniformly outward from the representative point.

6. The spherical decoding detection method as described in claim 1, characterized in that, The step of normalizing the first R matrix to obtain the first balanced R matrix includes: Multiply the first R matrix by the constellation point modulation amplitude factor to obtain the normalized first R matrix, and use the normalized first R matrix as the first equalization R matrix.

7. A spherical decoding detection device, characterized in that, include: The QR decomposition module is configured to perform orthogonal triangular QR decomposition on the first channel response matrix after determining that the current system layer number is greater than two, to obtain the first Q matrix and the first R matrix. The columns in the first Q matrix are arranged according to the magnitude of the channel energy, and the column with the smallest channel energy is placed at the top layer of the SD search. The equalization signal calculation module is configured to multiply the conjugate transpose of the first Q matrix with the received signal to obtain the first Z matrix, normalize the first R matrix to obtain the first equalization R matrix, and determine the target equalization signal corresponding to the received signal based on the first Z matrix and the first equalization R matrix. The ML path search module is configured to, if the modulation scheme of the current system is determined to be the first modulation scheme, execute a first ML path search procedure based on the target equalization signal, wherein the first ML path search procedure involves dividing all nodes at the top level into B... Nt-1 There are several blocks, each containing a representative point and several extended points. Path detection is performed on each representative point, and the block with the smallest metric, Ka, is selected. Nt-1 Given N1 representative points, the blocks corresponding to these representative points are taken as surviving blocks. Path detection is then performed on the expansion points near the representative points within each surviving block. The N1 branches with the smallest metrics are selected as surviving paths, where B... Nt-1 Ka Nt-1 Both N1 and B are natural numbers, and B Nt-1 >Ka Nt-1 ; wherein, the modulation order of the first modulation method is greater than or equal to a preset value; if it is determined that the modulation method of the current system is the second modulation method, then the second ML path search process is executed according to the target equalization signal, wherein the second ML path search process is to perform path detection on all nodes of the top layer and select the N2 branches with the smallest metric as surviving paths, wherein the modulation order of the second modulation method is less than the preset value, and N2 is a natural number. The ML complement path search module is configured to perform ML complement path search on all surviving paths to obtain the ML complement path; The soft value information calculation module is configured to take the surviving path with the smallest metric as the ML path, and derive the likelihood ratio information of each bit of each symbol in each layer based on the ML path and the ML complement path.

8. An electronic device, characterized in that, include: The system includes a memory, a processor, and a spherical decoding detection program stored in the memory and executable on the processor, wherein the spherical decoding detection program, when executed by the processor, implements the spherical decoding detection method as described in any one of claims 1 to 6.

9. A readable storage medium, characterized in that, The readable storage medium is a computer-readable storage medium, and the computer-readable storage medium stores a spherical decoding detection program, which, when executed by a processor, implements the spherical decoding detection method as described in any one of claims 1 to 6.

Images