Sky-wave massive MIMO beam structure turbo receiving method and receiver

Abstract

The application discloses a sky-wave massive MIMO beam structure turbo receiving method and a receiver. The application realizes uplink multi-user iterative detection decoding by using a beam structure turbo receiver. The beam structure turbo receiver comprises a beam structure soft input soft output detector, a soft input soft output decoder, an interleaver and a deinterleaver. The beam structure soft input soft output detector comprises a windowing module, a beam transformation module, a beam selection module, a beam domain signal detection module and the like. The extrinsic information of the code bits sent by each user is calculated through a base station receiving signal, and the extrinsic information is fed back to the beam structure soft input soft output detector after deinterleaving, decoding and interleaving, and is used to update the prior information of the sending symbols to perform the next round of iterative detection decoding. The application utilizes the sparse characteristics of the sky-wave massive MIMO beam domain channel and a turbo receiving framework to design the receiver, and can effectively improve the sky-wave massive MIMO uplink receiving performance and reduce the calculation complexity.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of communication, and particularly relates to a sky wave massive MIMO beam structure turbo receiving method and a receiver. BACKGROUND

[0002] The communication frequency band of sky wave communication is generally 3 to 30 MHz. Electromagnetic waves in the frequency band can be reflected by the ionosphere and the ground to achieve over-the-horizon transmission, and is an important long-distance communication method. However, due to limited spectrum resources and complex ionospheric environment, the communication rate of traditional sky wave communication is usually relatively low. Massive MIMO can significantly improve the communication rate and spectrum efficiency by deploying a large number of antennas at the base station to serve multiple users at the same time. Therefore, it is of great significance to introduce massive MIMO technology into sky wave communication.

[0003] For the uplink reception problem of massive MIMO, widely used linear receivers such as zero-forcing (ZF) and minimum mean-squared error (MMSE) receivers all face the problem of matrix inversion and multiplication with large dimensions, and the computational complexity is difficult to accept in actual systems. Some existing low-complexity signal detection methods that combine the sparsity of the beam domain channel of sky wave massive MIMO, such as beam structure signal detection and signal detection based on Slepian transform. Since these methods are approximations to MMSE detection, their performance is usually inferior to MMSE detection, and when the number of users is large, the increase in inter-user interference will cause the detection performance to drop significantly.

[0004] Turbo receivers improve system performance by iteratively passing soft information about coded bits between soft-input soft-output detectors and soft-input soft-output decoders, but when applied to massive MIMO systems, the soft-input soft-output detector in the conventional turbo receiver has higher computational complexity. Therefore, by taking advantage of the sparsity of the beam domain channel of sky wave massive MIMO, the high complexity problem of turbo receivers is solved, and the near-optimal receiving performance is obtained with low computational complexity, which is very meaningful for uplink reception of sky wave massive MIMO systems. SUMMARY

[0005] The application aims to solve the problems of the prior art. The application provides a sky wave massive MIMO beam structure turbo receiving method and a receiver, which takes advantage of the sparsity of the beam domain channel and combines the turbo receiving framework to effectively improve the detection and decoding performance while reducing the computational complexity.

[0006] TECHNICAL SOLUTION To solve the above technical problems, the application provides the following technical solutions:

[0007] The skywave massive MIMO beam structure turbo receiving method comprises the following steps:

[0008] The beam structure soft input and soft output detector is based on the beam-based channel model and uses the interleaved bit soft information of the received signal and the decoder feedback to calculate the external information of the code bits sent by each user; the beam structure soft input and soft output detector includes a windowing module, a beam transformation module, a beam selection module, a beam domain signal detection module, an external information calculation module, a priori information update module and a signal reconstruction module; the windowing module performs windowing processing on the de-meaned received signal, and the windowing processing is to perform a point multiplication of the window function vector and the de-meaned received signal; the beam selection module extracts the beam domain received signal to obtain multiple groups of reduced-dimensional beam domain signals; the beam domain signal detection module uses multiple groups of reduced-dimensional beam domain signals and the priori information of the symbols sent by each user, combined with the beam domain channel state information, to calculate and output the beam domain signal detector of each group and the signal detection result;

[0009] The external information is deinterleaved, decoded, and interleaved through a deinterleaver, a decoder, and an interleaver, and then fed back to the beam structure soft input and soft output detector to update the prior information of the transmitted symbol for the next round of iterative detection and decoding;

[0010] After several rounds of iterations, the soft-input soft-output decoder finally outputs the decoding results of the information bits sent by each user.

[0011] Furthermore, the beam-based channel model models the spatial domain channel as a beam matrix multiplied by a beam domain channel vector, wherein the beam domain channel vector has significant sparsity, and the beam matrix is ​​a matrix composed of direction vectors corresponding to a set of selected spatial direction cosine sampling grid points, expressed as a discrete Fourier transform matrix with phase shift, cyclic shift, row and column extraction, and normalization.

[0012] Furthermore, the window function vector adopts a classical window function or an energy concentration window function, and the classical window function includes a Hanning window and a Caesar window; the energy concentration window function is obtained by maximizing the energy concentration of the diffuse beam domain channel matrix, wherein the diffuse beam domain channel matrix is ​​the result of beam transformation of the windowed spatial domain channel matrix.

[0013] Furthermore, the beam selection module extracts a reduced-dimensionality beam domain signal set obtained by combining the non-zero beam sets of the users in the group. The non-zero beam set of each user is a set consisting of beam position indices where the non-zero elements of the beam domain channels of each user are located.

[0014] Furthermore, the beam domain signal detection module calculates and outputs the beam domain signal detector and signal detection results of each group according to a minimum mean square error criterion.

[0015] Furthermore, the extrinsic information calculation module calculates and outputs the extrinsic information of each user code bit using the beam domain signal detector of each group of signals and the signal detection result.

[0016] Furthermore, the prior information update module uses the interleaved code bit soft information fed back by the decoder to update the prior mean and prior variance of the symbols sent by each user, where the prior mean is used for signal reconstruction and beam domain signal detection, and the prior variance is used for beam domain signal detection.

[0017] Furthermore, the signal reconstruction module calculates and outputs a reconstructed signal using the mean value of each user's transmitted symbols and beam domain channel state information.

[0018] Furthermore, the signal reconstruction module, windowing module and beam transformation module utilize the sparsity of the beam domain channel and the fast Fourier transform to reduce the implementation complexity; the beam domain signal detection module combines the energy concentration property of the window function to reduce the implementation complexity, including the approximate calculation of the diffuse beam domain channel matrix and the approximate calculation of the beam domain signal detector.

[0019] The skywave massive MIMO beam structure turbo receiver includes: a beam structure soft input and soft output detector, a soft input and soft output decoder, an interleaver and a deinterleaver; the beam structure soft input and soft output detector includes a windowing module, a beam transformation module, a beam selection module, a beam domain signal detection module, an external information calculation module, a priori information update module and a signal reconstruction module, which is used to calculate the external information of the code bits sent by each user based on the beam basis channel model and the interleaved bit soft information fed back by the received signal and the decoder; the windowing module is used to perform windowing processing on the de-meaned received signal, and the windowing processing is to perform a point multiplication of the window function vector and the de-meaned received signal. ; The beam selection module is used to extract the beam domain received signal to obtain multiple groups of reduced-dimensional beam domain signals; the beam domain signal detection module is used to use multiple groups of reduced-dimensional beam domain signals and the prior information of each user's sent symbols, combined with the beam domain channel state information, to calculate and output each group of beam domain signal detectors and signal detection results; the soft input soft output decoder, interleaver and deinterleaver are used to decode, interleave and deinterleave external information respectively, and the external information is deinterleaved, decoded and interleaved and then fed back to the beam structure soft input soft output detector for updating the sent symbol prior information for the next round of iterative detection and decoding; the soft input soft output decoder outputs the decoding result of the information bits sent by each user.

[0020] Beneficial effects: Compared with the existing technology, the present invention utilizes the sparse characteristics of the skywave massive MIMO beam domain channel and the turbo reception framework to design the receiver. Combined with the windowing method, the sparse beam domain channel of each user is used to perform grouped low-dimensional beam domain detection. While significantly improving the performance compared to linear reception, it effectively reduces the computational complexity and improves the overall transmission efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a block diagram of a beam structure turbo receiving system according to an embodiment of the present invention;

[0022] Figure 2 This is a diagram of a skywave massive MIMO communication scenario according to an embodiment of the present invention;

[0023] Figure 3 This is a comparison chart of the bit error rate (BER) performance between a beam structure turbo receiver according to an embodiment of the present invention and an MMSE receiver under rectangular windows and energy concentration windows;

[0024] Figure 4 This is a comparison chart of the BER performance of the beam structure turbo receiver according to an embodiment of the present invention with that of an MMSE receiver under typical window functions such as an energy concentration window, a Hanning window, and a Kaiser window. DETAILED DESCRIPTION

[0025] In order to better understand the purpose, structure and function of the present invention, the skywave massive MIMO beam structure turbo receiving method of the present invention is further described in detail below with reference to the accompanying drawings.

[0026] The embodiment of the present invention discloses a method for skywave massive MIMO beam structure turbo reception, wherein the system block diagram of the beam structure turbo receiver is shown in FIG. Figure 1 The base station receives the signal and passes it through the beam structure turbo receiver to obtain the decoding results of the code bits sent by each user. Figure 1As shown, the beam-structured turbo receiver consists of a beam-structured soft-input and soft-output detector, a soft-input and soft-output decoder, an interleaver, and a deinterleaver. The beam-structured soft-input and soft-output detector includes a windowing module, a beam transformation module, a beam selection module, a beam-domain signal detection module, an extrinsic information calculation module, a priori information update module, and a signal reconstruction module. Based on the beam-based channel model, it uses the received signal and the interleaved bit soft information fed back by the decoder to calculate the extrinsic information of each user's transmitted code bits. This extrinsic information is deinterleaved, decoded, and interleaved, and then fed back to the beam-structured soft-input and soft-output detector to update the transmitted symbol prior information for the next round of iterative detection and decoding. After several rounds of iterations, the soft-input and soft-output decoder ultimately outputs the decoded results of the code bits sent by each user.

[0027] The beam-based channel model models the spatial domain channel as a beam matrix multiplied by a beam domain channel vector, wherein the beam domain channel vector has significant sparsity. The beam matrix is ​​a matrix composed of direction vectors corresponding to a set of spatial direction cosine sampling grid points, and is represented as a discrete Fourier transform matrix with phase shift, cyclic shift, row and column extraction, and normalization. The windowing module performs windowing processing on the de-meaned received signal. The specific process of windowing processing is to perform point multiplication on the window function vector and the de-meaned received signal. The window function vector involved uses a classic window function or an energy concentration window function, wherein classic window functions include Hanning window, Kaiser window, etc. The energy concentration window function is obtained by maximizing the energy concentration of the diffuse beam domain channel matrix, wherein the diffuse beam domain channel matrix is ​​the result of beam transforming the windowed spatial domain channel matrix. The beam transform module performs beam transform on the windowed de-meaned received signal to obtain a beam domain received signal. The specific process of beam transform is to multiply the conjugate transpose of the beam matrix by the windowed de-meaned received signal. The beam selection module extracts the beam-domain received signal to obtain multiple sets of reduced-dimensionality beam-domain signals. The extracted reduced-dimensionality beam-domain signal sets are obtained by combining the non-zero beam sets of users within the group. The non-zero beam set of each user is a set consisting of the beam position indices of the non-zero elements of each user's beam-domain channel. The beam-domain signal detection module uses the multiple sets of reduced-dimensionality beam-domain signals and prior information about each user's transmitted symbols, combined with beam-domain channel state information, to calculate and output the beam-domain signal detector and signal detection results for each group based on the minimum mean square error criterion. The extrinsic information calculation module uses the beam-domain signal detector and signal detection results of each signal group to calculate and output the extrinsic information of each user's code bits. The prior information update module uses the interleaved code bit soft information fed back by the decoder to update the prior mean and prior variance of each user's transmitted symbols. The prior mean is used for signal reconstruction and beam-domain signal detection, and the prior variance is used for beam-domain signal detection. The signal reconstruction module uses the mean of each user's transmitted symbols and beam-domain channel state information to calculate and output the reconstructed signal.

[0028] In some embodiments, the structural characteristics of the beam matrix and the sparsity of the beam-domain channel can be further combined to reduce the implementation complexity of the signal reconstruction module, the windowed beam transform module, and the beam-domain signal detection module. For example, the signal reconstruction module and the windowed beam transform module can be implemented with low complexity by leveraging the sparsity of the beam-domain channel and the fast Fourier transform. The beam-domain signal detection module can be implemented with low complexity by combining the energy concentration property of the window function, including approximate calculations of the diffuse beam-domain channel matrix and the detector.

[0029] See also Figure 2, an embodiment of the present invention discloses a skywave massive MIMO communication scenario diagram, in which data symbols sent by users are reflected by the ionosphere and the ground and received at the base station.

[0030] The method of the present invention is primarily applicable to skywave massive MIMO systems equipped with large-scale antenna arrays in base stations to simultaneously serve multiple users. The specific implementation of the uplink reception method of the present invention is described in detail below, using a specific communication system example. It should be noted that the present invention is applicable not only to the specific system model described in this embodiment, but also to system models with other configurations.

[0031] 1. System Model

[0032] Consider a skywave massive MIMO system operating in Orthogonal Frequency Division Multiplexing (OFDM). The base station is equipped with a uniform linear array (ULA) with M antennas, serving U single-antenna users simultaneously. Define f c is the carrier frequency, f o The highest system operating frequency. The distance between adjacent base station antennas is set to d = λ o / 2, where λ o is the wavelength and c is the speed of light.

[0033] During uplink transmission, the complex symbols sent by U users pass through the channel and are received as M-dimensional signals at the base station. For each subcarrier of each OFDM symbol, the received signal vector can be expressed as

[0034] y=Hx+z (1)

[0035] The OFDM symbol subscript and subcarrier subscript are omitted for convenience of expression. Receive signals for base stations, is the channel matrix, is the transmitted signal vector of U users and represents the set of constellation points, is additive white Gaussian noise.

[0036] According to the multipath channel model, the channel vector h of user u is u It can be expressed as

[0037]

[0038] Among them, P u is the number of multipaths from user u to the base station, α u,p and Ω u,p are the complex gain and direction cosine from the pth path of user u respectively. The direction vector pointing to Ω Defined as a conjugate symmetric structure, its mth element is

[0039]

[0040] where Δ τ =d / c.

[0041] In order to characterize the channel in the beam domain, the direction cosines are uniformly sampled with a sampling interval of

[0042]

[0043] Where F is the refinement factor. Considering the uniform division of the direction cosine range [-1,1), we can get mutually disjoint subsets, where the ath subset is defined as

[0044]

[0045] in definition is a set of direction cosines of all paths of user u, and belongs to S a The direction cosines are mapped to the sampling grid Ω a .

[0046] After uniform sampling of the direction cosines, the spatial domain channel h u can be approximated as

[0047]

[0048] in is the sampling direction vector and is the beam matrix, is the beam domain channel vector and By definition, the beam matrix can be expressed as

[0049]

[0050] Where S = FM, represents the S-dimensional discrete Fourier matrix, Depend on The first N1 columns (N1≤N2) or The first N2 rows (N1>N2) of e i is the i-th column of the identity matrix, the phase shift matrix

[0051] From (7), we can see that the beam matrix is ​​a matrix composed of direction vectors corresponding to a set of spatial direction cosine sampling grid points, which is expressed as a discrete Fourier transform matrix with phase shift, cyclic shift, row and column extraction and normalization.

[0052] Each sampling direction vector in the channel representation (6) represents a physical beam in the spatial domain, and all users share the same set of sampling direction vectors. Therefore, (6) is called the beam-based channel model, which can be written in a more compact form

[0053] H=VG(8) is the beam domain channel matrix.

[0054] Due to the small angular spread and limited transmission path of the skywave channel, the beam domain channel exhibits significant sparsity, i.e., g u There are very few non-zero elements in . The non-zero beam set of user u is defined as And A u =|A u |, which can be achieved through A u ={a|[ω u ] a ≠0}, where the beam domain channel coupling vector is defined as

[0055]

[0056] Beam-domain channel coupling vector ω u It is usually used to represent the information of the beam domain communication channel, which has the property of slowly changing with time. Therefore, ω is used u Go get A u In this embodiment, it is assumed that the beam domain instantaneous channel state information G and the beam domain statistical channel state information ω u All are known on the base station side.

[0057] 2. Beam Structure Turbo Reception Method

[0058] Turbo receivers are usually used in bit interleaved coded modulation systems. The original information bit stream of each user is coded and interleaved. Then, the interleaved bit stream is divided into several N-dimensional vectors and mapped to The constellation points in The corresponding bit sequence constellation points, and K = 2 N is the modulation order, from which the transmitted symbol is generated. Mapped to the transmitted symbol x u The bit sequence is represented as in And q u ∈{b1,…,b K}. Naturally, it can be assumed that all users’ transmitted symbols are independently and uniformly and randomly taken values ​​in And the transmission power of U users is normalized to the unit power, that is, At the base station, the received signal is represented as (1) and a turbo receiver is used. During each turbo iteration, soft information is transferred between a soft-input soft-output detector and a soft-input soft-output decoder. In the final iteration, the decoder outputs the decoded information bits. The soft-input soft-output detector primarily involves signal detection using prior information, external information calculation, and prior information update.

[0059] During the turbo iteration process, the prior mean and variance of the transmitted signal x are defined as

[0060]

[0061] The values ​​of μ and Σ are initialized to 0 and I respectively U , and is updated in the iterative process. Using the prior mean and prior variance, the signal detection goal of user u is to approximate p(y|x u =s k ) to facilitate external information calculation, where p(·) represents the probability density function of the continuous random variable.

[0062] In order to design a low-complexity signal detection method, we consider linear detection combined with the beam-based channel model. First, divide U users into L mutually exclusive user groups N1,…,N L And N l =|N l |, and define the user selection matrix The selection matrix And T={n1,…,n T}, n1<…<n T , T = |T|. So the detection of x can be decoupled into L groups, where the transmitted symbol of group l is The detection is

[0063]

[0064] Detect x l The mean square error is By minimizing the mean square error, the optimal R l and n l It can be obtained by solving the following optimization problem

[0065]

[0066] Solving (13) yields the MMSE detector using prior information

[0067]

[0068] For massive MIMO systems, when the number of base station antennas and users is large, the computational complexity of (14) will become very high due to the large-dimensional matrix inversion and multiplication operations. Therefore, we further explore the internal structure of the optimal solution to problem (13).

[0069] The beam set and beam selection matrix of group l are defined as and Among them B l =|B l Assuming that the number of users is finite and the multipath direction cosines of different user groups do not overlap, it can be proved that when M→∞, the optimal solution of problem (13) has the following structure

[0070]

[0071] in and

[0072]

[0073] Therefore, in terms of minimizing the mean square error, the asymptotically optimal spatial domain detector has a beam structure. Although the number of base station antennas is limited in practice, this idea can still guide the design of the detector when the number of base station antennas is large enough. Specifically, for the lth group of users, the spatial domain detector R l Restricted to beam structure

[0074] R l =VB l W l (18) is the beam domain detector of group l. In order to further suppress interference, the beam structure detector is applied to the received signal Λy after the window function, where is the window function. So the detection of group l is

[0075]

[0076] in So the mean square error of the detection is By minimizing Available

[0077]

[0078] in and

[0079]

[0080] Where D is the diffuse beam domain channel matrix, which can be understood as the result of beam transformation on the windowed spatial domain channel matrix. Therefore, the detection of the lth group of users can be written as

[0081]

[0082] in Extracting the signal in the beam domain

[0083]

[0084] in To reconstruct the signal.

[0085] In the beam structure soft input and soft output detector of this embodiment, the windowing module calculates The beam transformation module performs the calculation of (26), and the beam selection module calculates The beam domain signal detection module performs the calculations of (22), (20) and (25), and the signal reconstruction module calculates

[0086] In practical systems, to achieve satisfactory transmission performance with low complexity, appropriate user scheduling and grouping strategies need to be applied so that the beam set for each user group is relatively small. Given the sparsity of the beam-domain channel and the user grouping scheme, both the beam-domain signal detector design (20) and the beam-domain signal detection (25) involve only low-dimensional matrix-vector operations and thus have low computational complexity.

[0087] After completing the signal detection using the prior information, it is necessary to calculate the extrinsic information of each user's transmitted code bits based on the detection results. u,i The outer log-likelihood ratio (LLR) is

[0088]

[0089] where P(·) represents the probability of a discrete random variable, To s k The corresponding i-th bit b k,j The set of symbols mapped to b. Bit probability By prior LLR L′ a (q u,i ) is calculated and can be expressed as In addition, p(y|x u =s k ) is difficult to calculate directly, so we use the test results to approximate it. Assume that x obeys a cyclic symmetric complex Gaussian distribution with mean μ and variance Σ, then

[0090]

[0091] where x uThe outer mean and outer variance are

[0092]

[0093] where x u The posterior mean and posterior variance are

[0094]

[0095] Put p(y|x u =s k ) is approximately p(y l |x u ) and substitute into (27), then the external LLR can be calculated as follows

[0096]

[0097] Furthermore, after deinterleaving, L e (q u,i ) is mapped to L a (c u,i ), where c u,i =Π-1(q u,i ) and Π(·) is an interleaver. L a (c u,i ) is the prior soft information input to the decoder, and after decoding, L′ is obtained e (c u,i ), further interweave to obtain L′ a (q u,i ), used for next iterative detection. Based on L′ a (q u,i ), the updated probability of sending symbols is

[0098]

[0099] Therefore, the prior mean and prior variance of the transmitted symbols used for the next round of iterative detection are updated as

[0100]

[0101] From the above, the external information calculation module performs the calculations of (31), (32), (29), (30) and (33), and the prior information update module performs the calculations of (34), (35) and (36).

[0102] Based on the above discussion, beam structure turbo reception includes the following steps:

[0103] Step 1: Calculate the prior information of the transmitted symbol according to (34), (35) and (36). For the first iteration, the prior LLR is initialized to L′ a (qu,i )=0;

[0104] Step 2: Use (26) to perform windowed beam transform on the de-averaged signal to obtain

[0105] Step 3: Perform beam selection to obtain

[0106] Step 4: Design a beam domain detector according to (20) and perform beam domain signal detection according to (25) to obtain W1,…,W L ,

[0107] Step 5: Calculate the outer LLR according to (33);

[0108] Step 6: The outer LLR is deinterleaved and input into the decoder. If the maximum number of iterations is reached, the decoding result is output. Otherwise, the decoded output soft information is interleaved to obtain the prior LLR, and then go to step 1.

[0109] 3. Energy Concentration Window Function Design

[0110] In order to suppress inter-user interference and reduce computational complexity, the embodiment of the present invention designs a window function that can maximize the energy concentration of the diffuse beam domain channel matrix D. To this end, due to the linear characteristics of multipath superposition, considering the single-path channel Where α is the complex gain and Ω is the direction cosine. Further definition

[0111] d(Ω,η)=V H Λh(Ω)=αV H diag{η}v(Ω) (37)

[0112] The goal of window function design is to maximize the energy of d(Ω,η) and concentrate it near the sampling point a(Ω). The index set of expected maximum energy concentration is defined as Where c is a preset integer. For d(Ω,η), the total energy and the energy within the range C(Ω) are defined as

[0113]

[0114] In order to design a unified window function for different channels, it is assumed that the direction cosines of all users are uniformly distributed in [-Ω′,Ω′] in a fixed sector, where 0<Ω′≤1, and the average energy ratio is further defined as

[0115]

[0116] Therefore, the window function that maximizes the average energy ratio can be obtained by solving the following optimization problem

[0117]

[0118] It can be shown that the optimal solution to problem (41) is the generalized eigenvector corresponding to the maximum generalized eigenvalue of the matrix pair {Φ,Ξ}, and can be constructed as a real vector with a centrosymmetric structure, where

[0119] [Φ] m,m′ =sinc(π(mm′) / S)D c (2π(mm′) / S) (42)

[0120]

[0121] and

[0122]

[0123] are the sinc function and Dirichlet kernel respectively.

[0124] Therefore, the energy concentration window function η o It can be obtained by maximizing the energy concentration of the diffuse beam domain channel matrix. Specifically, it is obtained by calculating the real symmetric generalized eigenvector of the matrix pair {Φ,Ξ}. Therefore, it can be calculated offline and used for beam-structured turbo reception.

[0125] 4. Low Complexity Implementation

[0126] In order to effectively implement beam-structured turbo reception, real symmetric window functions are considered, such as the energy concentration window proposed in the embodiment of the present invention, as well as typical window functions such as the Hanning window and the Kaiser window. The implementation complexity of beam-structured turbo reception mainly comes from signal reconstruction, windowed beam transformation, and beam-domain signal detection. Signal reconstruction and windowed beam transformation can further utilize the sparsity of the beam-domain channel and the low complexity of the fast Fourier transform, including and Beam domain signal detection can be implemented with low complexity by combining the energy concentration property of the window function, including the approximate calculation of the diffuse beam domain channel matrix D and the detector W l Approximate calculation of .

[0127] First discuss and Calculation.

[0128] and It needs to be calculated at each turbo iteration, combined with the structure of the beam matrix (7), and Can be written as

[0129]

[0130] therefore and The calculation of can be effectively implemented by combining the sparsity of the beam domain channel and the fast Fourier transform.

[0131] Next, we discuss the approximate calculation of the diffuse beam domain channel matrix D and the detector W l Approximate calculation of .

[0132] D is calculated every time the channel changes. Combined with (7), D can be rewritten as in

[0133]

[0134] It can be proved that when the window function is scaled to satisfy tr{Λ}=M, has a real number structure and can be expressed as

[0135]

[0136] in and It can be found that when an appropriate window function is selected, γ k There are a lot of small values ​​in . In order to reduce the computational complexity of D, we ignore This allows for approximate computation, where the filter set And ζ is a threshold. So D can be approximated as

[0137]

[0138] Among them G d It can be regarded as the diffusion of the beam domain channel. According to (49), the approximate calculation of D comes from G d The calculation involves only the multiplication of sparse complex matrices and real scalars, as well as several permutation, sign reversal, decimation, and summation operations.

[0139] When an appropriate window function is selected, the inter-group interference can be effectively suppressed. To this end, the set of interfering users in group l is defined as and So W l can be approximated as

[0140]

[0141] in is a decimation matrix, and

[0142]

[0143] Since the matrix U is fixed, U -1 can be pre-calculated. In addition, since T l and K l It is independent of the sign prior information and can be calculated before the turbo iteration. At each iteration, the matrix inversion term ( Σ l K l +σ z I) -1 The computational complexity of is much lower than the matrix inversion term in (20) because the interfering user set The size of the beam set B is usually smaller than l size.

[0144] In summary, combining (45) and (46) can realize signal reconstruction and windowed beam transformation with low complexity, and combining (49) and (50) can realize beam domain signal detection with low complexity, thereby significantly reducing the computational complexity of beam structure turbo reception.

[0145] VI. Implementation Effect

[0146] In order to enable those skilled in the art to better understand the solution of the present invention, performance results of the uplink receiving method in this embodiment under specific configurations are given below.

[0147] Considering the skywave massive MIMO-OFDM communication system, the system parameters are configured as follows: carrier frequency f c =16MHz, subcarrier spacing Δf=250Hz, number of subcarriers N c =2048, effective subcarrier number N v =1536, the number of base station antennas M = 256, the base station antenna spacing d = 9m, the number of users U = 72, the beam domain refinement factor F = 2, the modulation scheme is 16-order orthogonal amplitude modulation, the channel coding uses low-density parity-check code with a code length of 2112 and a code rate of 3 / 4, and a row-column interleaver is used. To measure the uplink reception performance, the BER performance of a beam-structured turbo receiver and an MMSE receiver is compared under different signal-to-noise ratios (SNRs), where SNR refers to the received SNR.

[0148] Figure 3The BER performance comparison of the method of this embodiment with that of the MMSE receiver under the rectangular window and the energy concentration window is given. Applying a rectangular window is equivalent to not applying windowing processing, and the parameters are set to ζ = 0. The parameters of the energy concentration window are set to c = 3, Ω′ = 1, and ζ = 0.001. It can be seen from the figure that in the first iteration, the BER performance of the proposed embodiment method is not as good as that of the MMSE receiver, especially in the high SNR area. However, after subsequent iterations, the BER performance of the proposed embodiment method is significantly better than that of the MMSE receiver. In addition, there is a certain gap between the BER performance of applying the energy concentration window and the BER performance of applying the rectangular window, which is mainly due to the performance loss caused by some approximate calculations in the low-complexity implementation. Nevertheless, approximate calculations can greatly reduce the implementation complexity of beam structure turbo reception, and its performance gap is significantly narrowed in subsequent iterations.

[0149] Figure 4 The BER performance comparison of the method of this embodiment with that of the MMSE receiver under the energy concentration window and typical window functions such as the Hanning window and the Caesar window is given. The parameters of the energy concentration window are set to c=3, Ω′=1, ζ=0.001, and the shape parameter of the Caesar window is 10. For the sake of fairness, the corresponding thresholds are set for different window functions so that Q=27. At this time, the beam structure turbo reception under different window functions has the same implementation complexity. As can be seen from the figure, after the third iteration, the BER performance of the method of this embodiment under different window functions is better than that of the MMSE receiver, and the BER performance under the energy concentration window function is better than the BER performance under other window functions. This confirms the effectiveness of the method of the proposed embodiment.

[0150] Based on the same inventive concept, the embodiment of the present invention discloses a skywave massive MIMO beam structure turbo receiver, comprising: a beam structure soft input soft output detector, a soft input soft output decoder, an interleaver and a deinterleaver; the beam structure soft input soft output detector comprises a windowing module, a beam transformation module, a beam selection module, a beam domain signal detection module, an external information calculation module, a priori information update module and a signal reconstruction module, which is used to calculate the external information of the code bits sent by each user based on the beam-based channel model and the interleaved bit soft information fed back by the received signal and the decoder; the windowing module is used to perform windowing processing on the de-averaged received signal, and the windowing processing is to combine the window function vector with the de-averaged signal vector. The received signal is point-multiplied by the value; the beam selection module is used to extract the beam domain received signal to obtain multiple groups of reduced-dimensional beam domain signals; the beam domain signal detection module is used to use multiple groups of reduced-dimensional beam domain signals and the prior information of each user's transmitted symbols, combined with the beam domain channel state information, to calculate and output the beam domain signal detector and signal detection results of each group; the soft input soft output decoder, interleaver and deinterleaver are respectively used to decode, interleave and deinterleave the external information. After deinterleaving, decoding and interleaving, the external information is fed back to the beam structure soft input soft output detector to update the prior information of the transmitted symbol for the next round of iterative detection and decoding; the soft input soft output decoder outputs the decoding result of the information bit sent by each user. Detailed implementation examples of each module can be found in the aforementioned method embodiments and will not be repeated here.

[0151] It will be understood that the present invention is described by way of some embodiments, and it will be appreciated by those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present invention. In addition, under the teachings of the present invention, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are intended to be protected by the present invention.

Claims

1. A skywave massive MIMO beam structure turbo reception method, characterized in that: The steps include: The beam structure soft input and soft output detector is based on the beam-based channel model and uses the received signal and the interleaved bit soft information fed back by the decoder to calculate the external information of the code bits sent by each user; the beam structure soft input and soft output detector includes a windowing module, a beam transformation module, a beam selection module, a beam domain signal detection module, an external information calculation module, a priori information update module and a signal reconstruction module; the windowing module performs windowing processing on the de-meaned received signal, and the windowing processing is to perform a point multiplication of the window function vector and the de-meaned received signal; the beam selection module extracts the beam domain received signal to obtain multiple groups of reduced-dimensional beam domain signals; the beam domain signal detection module uses multiple groups of reduced-dimensional beam domain signals and the priori information of the symbols sent by each user, combined with the beam domain channel state information, to calculate and output the beam domain signal detector of each group and the signal detection result; The external information is deinterleaved, decoded, and interleaved through a deinterleaver, a decoder, and an interleaver, and then fed back to the beam structure soft input and soft output detector to update the prior information of the transmitted symbol for the next round of iterative detection and decoding; After several rounds of iterations, the soft-input soft-output decoder finally outputs the decoding results of the information bits sent by each user.

2. The skywave massive MIMO beam structure turbo receiving method according to claim 1, characterized in that: The beam-based channel model models the spatial domain channel as a beam matrix multiplied by a beam domain channel vector, wherein the beam domain channel vector has significant sparsity, and the beam matrix is ​​a matrix composed of direction vectors corresponding to a set of selected spatial direction cosine sampling grid points, expressed as a discrete Fourier transform matrix with phase shift, cyclic shift, row and column extraction, and normalization.

3. The skywave massive MIMO beam structure turbo receiving method according to claim 1, characterized in that: The window function vector adopts a classical window function or an energy concentration window function, and the classical window function includes a Hanning window and a Caesar window; the energy concentration window function is obtained by maximizing the energy concentration of a diffuse beam domain channel matrix, wherein the diffuse beam domain channel matrix is ​​the result of beam transformation of the windowed spatial domain channel matrix.

4. The skywave massive MIMO beam structure turbo receiving method according to claim 1, characterized in that: The beam selection module extracts a reduced-dimensional beam domain signal set obtained by combining the non-zero beam sets of users in the group. The non-zero beam set of each user is a set consisting of beam position indices where the non-zero elements of the beam domain channels of each user are located.

5. The skywave massive MIMO beam structure turbo receiving method according to claim 1, characterized in that: The beam domain signal detection module calculates and outputs the beam domain signal detector and signal detection results of each group according to the minimum mean square error criterion.

6. The skywave massive MIMO beam structure turbo receiving method according to claim 1, characterized in that: The extrinsic information calculation module calculates and outputs the extrinsic information of each user code bit by using the beam domain signal detector of each group of signals and the signal detection results.

7. The skywave massive MIMO beam structure turbo receiving method according to claim 1, characterized in that: The prior information update module uses the interleaved code bit soft information fed back by the decoder to update the prior mean and prior variance of the symbols sent by each user, wherein the prior mean is used for signal reconstruction and beam domain signal detection, and the prior variance is used for beam domain signal detection.

8. The skywave massive MIMO beam structure turbo receiving method according to claim 1, characterized in that: The signal reconstruction module calculates and outputs a reconstructed signal using the mean value of each user's transmitted symbols and beam domain channel state information.

9. The skywave massive MIMO beam structure turbo receiving method according to claim 1, characterized in that: The signal reconstruction module, windowing module and beam transformation module utilize the sparsity of the beam domain channel and the fast Fourier transform to reduce the implementation complexity; the beam domain signal detection module combines the energy concentration property of the window function to reduce the implementation complexity, including the approximate calculation of the diffuse beam domain channel matrix and the approximate calculation of the beam domain signal detector.

10. Skywave massive MIMO beam structure turbo receiver, characterized by: include: Beam structure soft input soft output detector, soft input soft output decoder, interleaver and deinterleaver; The beam structure soft input and soft output detector includes a windowing module, a beam transformation module, a beam selection module, a beam domain signal detection module, an external information calculation module, a priori information updating module and a signal reconstruction module, which is used to calculate the external information of the code bits sent by each user based on the beam-based channel model and the interleaved bit soft information fed back by the received signal and the decoder; the windowing module is used to perform windowing processing on the de-meaned received signal, and the windowing processing is to perform a point multiplication of the window function vector and the de-meaned received signal; the beam selection module is used to extract the beam domain received signal to obtain multiple groups of reduced-dimensional beam domain signals; the beam domain signal detection module is used to use the multiple groups of reduced-dimensional beam domain signals and the priori information of the symbols sent by each user, combined with the beam domain channel state information, to calculate and output the beam domain signal detector of each group and the signal detection result; The soft input soft output decoder, interleaver and deinterleaver are respectively used to decode, interleave and deinterleave external information. After deinterleaving, decoding and interleaving, the external information is fed back to the beam structure soft input soft output detector to update the prior information of the transmitted symbols for the next round of iterative detection and decoding; the soft input soft output decoder outputs the decoding results of the information bits sent by each user.

Images