Soft decision detection method and system for multi-stream data stream combination

Abstract

The application discloses a soft decision detection method and system of a multiple stream data integral combination, and belongs to the fields of communication and information system, information theory and coding, and signal and information processing. The specific implementation steps are as follows: step one, sending a signal step; step two, receiving a signal; step three, defining the integral combination of the multiple stream data; and step four, calculating the posterior probability of the integral combination of the multiple stream data. The method has low complexity, possesses a parallel processing architecture and low processing delay, and can make the decoding performance close to the limit. The application is particularly suitable for a non-cellular network, and can make the non-cellular network obtain better use efficiency of an air interface and a backhaul link.

Description

[Technical Field]

[0001] This invention relates to a soft-decision detection method and system for multi-stream data integer combinations, belonging to the fields of communication and information systems, information theory and coding, and signal and information processing. [Background Technology]

[0002] The core issue in wireless communication is the handling of interference between multiple users, between multiple beams, between symbols, and between carriers.

[0003] For multi-user uplinks, the optimal detection method is maximum likelihood (ML) detection. However, the complexity of ML detection is exponentially related to the number of data streams, making it infeasible when the number of streams is large. Existing linear filtering-based methods, such as zero-forcing (ZF) and minimum mean square error (MMSE), suffer from significant performance losses [D. Tse and P. Viswanath, “Fundamentals of wireless communication,” Cambridge University Press, 2005]. Iterative detection and decoding (IDD) can significantly improve the performance of linear detection, but it suffers from high complexity, high processing latency, high memory consumption, and convergence problems caused by the mismatch between the detector and the channel coding decoder [Q. Chen, F. Yu, T. Yang, and R. Liu, “Gaussian and fading multiple access using linear physical-layer network coding,” IEEE Trans. Wireless Comm., May, 2023. ]. For the downlink, precoding based on ZF and MMSE suffers from significant rate loss, while the smudge-paper coding scheme is complex and suffers from high latency due to serial processing at the transmitter.

[0004] Existing research has shown that lattice reduction (LR) detection and integer-forcing (IF) linear receivers can achieve efficient interference handling [B. Nazer and M. Gastpar, “Compute-and-forward: Harnessing interference through structured codes,” IEEE Trans. Inf. Theory, vol. 57, no. 10, pp. 6463–6486, Oct. 2011.] and [J. Zhan, B. Nazer, U. Erez, and M. Gastpar, “Integer-forcing linear receivers,” IEEE Trans. Inf. Theory, vol. 60, no. 12, pp. 7661–7685, Dec. 2014. “Integer-forcing linear receivers,” IEEE Journal of Information Theory, December 2014.]. The core idea is to utilize the algebraic properties of lattice to relax the detection and decoding of each stream of data into the detection and decoding of "integer combinations of multiple streams of data", substantially improving the efficiency of interference handling. Compared with ZF and MMSE detection, LR and IF can achieve "full diversity gain" and support "overload transmission" (the number of streams K is greater than the number of receiving antennas N). LR and IF do not require receiver iterative detection, avoiding a series of implementation problems of IDD. For multi-user downlinks, precoding based on LR and IF can achieve a combined rate close to the channel capacity with low cost and low processing delay [D. Silva, G. Pivaro, G. Fraidenraich, and B. Aazhang, "On integer-forcing precoding for the Gaussian MIMO broadcast channel," IEEE Tran. Wireless Comm., vol. 16, no. 7, pp. 4476–4488, 2017. ]Essentially, the LR and IF processing methods are based on the concept of solving integer combinations of multi-stream data in network information theory, specifically in lattice codes, computation-forward (CF), and physical-layer network coding (PNC) [B. Nazer and M. Gastpar, “Compute-and-forward: Harnessing interference through structured codes,” IEEE Trans. Inf. Theory, vol. 57, no. 10, pp. 6463–6486, Oct. 2011. IEEE Information Theory Journal, October 2011.].

[0005] In communication systems, channel coding is a core element. It not only supports near-limit spectral efficiency and bit error rate performance but also provides system stability and reliability, aligning the communication system with information theory. Mainstream channel coding methods include low-density parity-check (LDPC) codes, polar codes, and their various variants in the 5G NR standard. In particular, soft-decision decoding based on posterior probability is essential for achieving near-limit decoding performance, offering several decibels (dB) improvement over hard-decision decoding [S. Lin and D.J. Costello, “Error control coding, 2nd edition,” Pearson, 2004].

[0006] However, prior to this invention, the calculation of integer combinations of multi-stream data in LR, PNC, CF, and IF was based on hard-decision detection, and a high-efficiency, low-complexity soft-decision detection method was lacking. This resulted in a performance gap of at least several decibels from its limits, and poor reliability and stability. Therefore, there is an urgent need to provide a soft-decision detection method for integer combinations of multi-stream data, which can accurately calculate the posterior probability of integer combinations with reasonable complexity, and fully realize the gains of the PNC, CF, and IF concepts in actual downlink multi-user transmission and non-cellular networks.

[0007] Based on the above situation, this invention addresses the interference problems between users, beams, symbols, and carriers in wireless communication. It considers the concepts of lattice, physical layer network coding, or computation and transmission, and achieves efficient interference processing by solving for integer combinations of multi-stream data. Specifically, for channel-coded data streams, this invention proposes a soft-decision detection method for integer combinations of multi-stream data. This method accurately calculates the posterior probability of integer combinations of multi-stream data, bringing decoding performance close to the limit, and the complexity is linearly related to the number of streams. Furthermore, based on this method, this invention proposes two new systems: a lattice-based downlink MIMO broadcast system and a lattice-based non-cellular MIMO system, achieving significant improvements in functionality and performance. [Summary of the Invention]

[0008] (I) Purpose of the invention:

[0009] To address the communication and interference issues of channel-coded multi-stream data, this invention provides a soft-decision detection method for integer combinations of multi-stream data. This method accurately calculates the posterior probability of integer combinations of multi-stream data, bringing decoding performance close to its limits. The complexity of this soft-decision detection method is linearly related to the number of data streams, features a parallel processing architecture, and has low processing latency. This invention applies advanced concepts from information theory and coding theory, such as lattice coding, computation and transmission, and physical layer network coding, to practical communication systems. It can be used for high-performance multi-user detection based on lattice, precoding, distributed base station processing in cellular networks, and inter-carrier interference handling.

[0010] Furthermore, another object of the present invention is the application of the aforementioned soft-decision detection method in a lattice-code multiple-access (LCMA) system.

[0011] Furthermore, another objective of the present invention is to provide a lattice-based downlink MIMO broadcast system that applies the soft-decision detection method, applicable to flat channels and frequency-selective channel models, thereby improving the system's functionality and performance.

[0012] Furthermore, another objective of the present invention is to provide a lattice-based non-cellular MIMO system for the aforementioned ICB soft-decision detection, which has better FER (frame error rate) performance.

[0013] (II) Technical Solution:

[0014] The soft-decision detection method for multi-stream data integer combinations of the present invention is implemented in the following steps:

[0015] Step 1: Send signal

[0016] Consider K-stream message data, using row vector b1 T ,…,b K T Let the row vector c be represented. i T Let c represent the i-th data stream after channel coding, where i = 1, 2, ..., K, and the data stream length is n. i [t] indicates Let the t-th sign bit be t, where t = 1, ..., n. Let the column vector c[t] = [c1[t], ..., c K [t] T This represents the t-th sign bit of all K-stream data.

[0017] Consider 2 m Meta-channel coding, m = 1, 2, ... Thus, c i [t]∈{0,…,2 m -1}, meaning its elements are no greater than 2. m -1 is a non-negative integer. The channel-coded data stream sequence is symbolically mapped to 2... m The PAM modulated signal sequence is as follows:

[0018]

[0019] Where γ is the normalization factor, ensuring that sequence x i T The average energy is 1. Here, x i T All elements are integers divided by γ. All K-stream signals are transmitted simultaneously.

[0020] For the complex model, two independent coding and modulation methods are used, transmitted separately in the in-phase and quadrature parts, forming a 2:1 I / Q ratio. 2m -QAM modulation. This conforms to the 2QAM modulation widely used in mainstream communication systems. m -PAM and 2 2m -QAM modulation.

[0021] Step 2: Receiving Signals

[0022] Consider a receiver with a received signal spatial dimension of N. (For example, the receiver is equipped with N antennas, each antenna providing one observation. Alternatively, the system's spread spectrum sequence length is N, with each chip-level signal providing one observation.)

[0023] For the real number model, the received signal is represented as:

[0024]

[0025] Where hi H represents the channel vector from the i-th stream signal to the N observations at the receiver; H = [h1, ..., h K [x1, ..., x2] represents the channel matrix, containing the channel vectors corresponding to all streaming signals; matrix X = [x1, ..., x3] K ] T Let Y represent all K-stream signal sequences, with the i-th row being the i-th stream signal; Z represents the additive white Gaussian noise (AWGN) matrix, where each element is an independent and identically distributed zero-mean, unit-variance Gaussian noise; ρ represents the average energy of each stream signal, which is equivalent to the signal-to-noise ratio. Here, Y = [y[1],…,y[n]], y[t] is the received signal vector of the t-th symbol bit.

[0026] A complex model can be represented by a real model of two dimensions, that is:

[0027]

[0028] For clarity, this invention is described using a real number model.

[0029] Step 3: Definition of Integer Combinations for Multistream Data

[0030] Consider a vector of length K (not all zero) integer coefficients. Another one in The integer combination (ICB) of c[t] is represented as follows:

[0031]

[0032] Here mod(·,2) m ) represents modulo 2 m The range of values ​​for the operation and the combination of integers is:

[0033] Generally, L-path ICB is represented as:

[0034]

[0035] This represents the integer coefficient vector corresponding to the l-th ICB.

[0036] This invention applies to any coefficient vector a1 T ,…,a L T The optimal a1 T ,…,a L TThe selection method is not the focus of this invention. For the sake of completeness, step three of Embodiment 1 is briefly described below, which involves a1. T ,…,a L T The optimized selection method.

[0037] Step 4: Calculate the posterior probability of integer combinations of multi-stream data (the core algorithm of this invention)

[0038] The receiver calculates the L-path ICB based on the received signal Y = [y[1],…,y[n]], drawing on the ideas of lattice, PNC, and IF. For channel-coded systems, this invention provides an efficient algorithm for accurately solving the posterior probability of the L-path ICB, forming a soft-decision detection result for the ICB.

[0039] Because we use a sign-by-sign operation, we can omit the sign bit index "[t]" below. Recall that the range of ICB values ​​is... Thus, the operation for calculating the posterior probability of ICB is represented as:

[0040]

[0041] Given a vector of L integer coefficients a1 T ,…,a L T The operation of the above formula (6) is as follows:

[0042] a) Linear filtering

[0043] Let W be a linear filter matrix of size L×N, where each element is a real number. Let ||w| represent the l-th row of W, and normalize it to ||w|. l || 2 =1. After filtering, L-channel signals are formed:

[0044]

[0045] in, The (real-valued) equivalent gain, noise term The variance is 1.

[0046] b) Signal representation

[0047] In order to calculate Given the posterior probability of the received signal, we make the following equivalent expression for equation (7). Let... Collect a l The position of the non-zero term, let Indicates its complement. Let Indicates al The number of non-zero terms. Thus, formula (7) can be expressed as:

[0048]

[0049] here, This item indicates a l ω(a) with a non-zero coefficient l The superposition of signals from 10 users is the useful signal component for calculating ICB; Including the rest of K-ω(a) l The user's signal, its corresponding a l The coefficient is zero, indicating that it is not correlated with the ICB. It is considered as equivalent noise, and it is uncorrelated with the useful signal portion. For a sufficiently large K, It's also large enough. Applying the central limit theorem, the equivalent noise ξ... l It follows a Gaussian distribution with a mean of 0 and a variance of .

[0050] Using formula (1) x i With c i bijective relation, i.e. Further simplification of formula (8) yields:

[0051]

[0052] here Independent of the signal, it resembles a DC component, and its purpose is to convert the signal from {-1, +1} to {0, 1} for processing. This is achieved through... After compensation, we get:

[0053]

[0054] After the above simplification, only a remains in the signal part of equation (10). l The signal of the user whose bit is not zero. The operation of calculating the posterior probability of ICB is expressed as:

[0055]

[0056] c) Exact calculation of the likelihood function of ICB

[0057] When calculating the posterior probability, i.e., formula (11), the likelihood function is required. The calculation method is as follows. Let the vector Only contains a l The non-zero elements of the vector, and let the vector Only includes the corresponding a in c l The part of non-zero elements (belonging to) (part of the text). and The length is Applying the law of total probability:

[0058]

[0059] From equation (10), we obtain:

[0060]

[0061] If we directly calculate the likelihood function (12) here, we need to calculate... Candidates vector The value is so that its complexity is O(n). Order of magnitude. Below, this invention provides a method for efficient computation (12).

[0062] d) Calculation of low-complexity likelihood function based on Gaussian approximation

[0063] consider and There exists a many-to-one mapping. Here we first calculate... likelihood function It can then be transformed into

[0064] Let set Collect and satisfy of The candidate sequence. For a given The conditional mean is:

[0065]

[0066] The conditional variance is:

[0067]

[0068] Thus, if the transmitted signal satisfies The received signal can be represented as:

[0069]

[0070] When K is large enough, for a given It can be approximated as the mean. variance is The Gaussian distribution is denoted by . Thus, the likelihood function can be expressed as:

[0071]

[0072] Then, using the law of total probability, we obtain the likelihood function of ICB:

[0073]

[0074] e) Calculate the posterior probability of ICB

[0075] Based on the likelihood function of ICB, i.e., formula (18), and using Bayes' theorem, the posterior probability of ICB is:

[0076]

[0077] Here, η is a normalization factor, which ensures that the calculated soft decision terms are normalized. Adding them together gives 1. The second step of equation (19) utilizes the equal probability property of integer combinations, that is... The calculation result of the posterior probability of the ICB (i.e., soft decision information) is passed to the channel coding decoder for decoding to obtain the decision of the integer combination of multi-stream message data.

[0078] The following process illustrates that the method for efficiently calculating the likelihood function (12) proposed in this invention can greatly reduce complexity.

[0079] For the l-th ICB, define Called a l The "weight". The complexity of the ICB soft-decision detection of this invention is O(ω). H (a l (2) m -1)+1) level. This is far lower than what is required to directly execute (12). The complexity is at level 1.

[0080] For all L-way ICBs, the complexity is:

[0081]

[0082] Where E a (ω H (a) represents the average weight of the coefficient vector. The average complexity for each user is O(2^3). m E a (ω H (a))), which is only E(ω) of the single-user detection complexity. H (a) times. Typically, E a (ω H (a) is much smaller than the number of data streams K. For example, in a system where K=32 and N=32, E a (ω H The value of (a) is not higher than 4.

[0083] This invention further provides an application of the soft-decision detection method in a gridcode multiple access system, specifically:

[0084] This study considers a K-stream user single-cell uplink multiple access model, where communication within each cell is unaffected by interference from other cells. For clarity and simplicity, the following settings are made: each user is equipped with a single antenna, and the base station receiver has N antennas. There is no inter-symbol interference in the model, which can be guaranteed using Orthogonal Frequency Division Multiplexing (OFDM). A flat fading model is considered, meaning the channel coefficients of each coding block remain constant. Following the convention for studying uplink multi-user systems, an open-loop system is considered, in which the base station receiver does not provide a feedback link to provide the transmitter with channel state information (CSI) or adaptive coding modulation (ACM) information. Each user transmits at the same target rate. The base station has known the channel state information H.

[0085] a) Channel coding and modulation

[0086] Let user i's 2 m Meta-message data sequences are represented by row vector b. i T ∈{0,1,…,2 m -1} k Let i = 1, 2, ..., K, and k be the length of the message sequence. The message data for all K users can be represented by matrix B = [b1, ..., bk]. K ] T This indicates that the size is K×k. The present invention uses 2... m The ring code encodes the sequence of user message data as follows:

[0087]

[0088] The channel coding operation is described in step a) of Example 2 in the specific implementation. Then, 2 is formed using formula (1). m -PAM symbol. All users transmit simultaneously on the same frequency band.

[0089] Note that this encoding and modulation belong to lattice codes, which possess the algebraic properties of lattice codes. Therefore, this multiple access scheme is also called "lattice code multiple access".

[0090] b) Received signal

[0091] The base station receiver receives the signal as shown in equation (2). Based on the channel state information H from the receiver, the base station uses the method provided in step three of the soft-decision detection method to select L = K linearly independent integer coefficient vectors a1. T ,…,a K T . Let A=[a1,…,a K ] T , is called an integer coefficient matrix, which is in Full rank. Define the integer combination of message data as:

[0092]

[0093] The receiving end wants to first calculate the K-way integer combination u1,…,u K Afterwards, restore all users' message data B = [b1,…,b K ] T u l The calculation is shown in steps c) and d) below, and its derivation is shown in Specific Implementation Example 2.

[0094] c) ICB Soft Judgment Detection

[0095] For the l-th ICB, the receiver uses the aforementioned ICB soft-decision method to calculate the posterior probability of integer combinations of the channel-coded K-stream data bit by bit:

[0096]

[0097] Then, pass the posterior probability to 2. m A decoder for meta-channel coding.

[0098] d) Decoding of channel coding

[0099] Decoder output:

[0100]

[0101] The judgment is as follows:

[0102]

[0103] If the judgment is correct, then the l-th integer combination of the K user message data is obtained:

[0104]

[0105] and u l T =[u l [1],…,u l [k]]. For details on decoder operation, please refer to step d of Example 2 of the specific implementation method.

[0106] e) User Data Recovery

[0107] The soft-decision detection and decoding operations of the K-way ICB are performed in parallel, resulting in:

[0108]

[0109] Because A is A matrix of full rank has a unique inverse. Accessible via:

[0110]

[0111] Operation restores all user message data B.

[0112] f) Simulation and Performance Evaluation

[0113] Considering m=1 and m=2, corresponding to BPSK and 4-PAM respectively, simulations were performed, and the frame error rate (FER) results were recorded. The results were then compared with the baseline scheme of iterative MMSE detection and decoding. Using the ICB soft-decision detection and decoding of this invention, this lattice-code-based multiple access scheme significantly outperforms the baseline scheme. Furthermore, the method of this invention has low complexity, a parallel processing architecture, and low processing latency, and does not suffer from the convergence problem caused by detector-decoder mismatch in iterative detection and decoding schemes.

[0114] Furthermore, this invention proposes a lattice-based downlink MIMO broadcast system applying this soft-decision detection method, as detailed below:

[0115] A lattice-based MIMO broadcasting (LBC) system utilizes the aforementioned ICB soft-decision detection method. Consider a base station transmitting its data streams to K users. The base station is equipped with N antennas; considering each user has a single antenna, it can be easily extended to multiple antennas at the user end. OFDM modulation is considered, eliminating inter-symbol interference. The base station has known channel state information. The LBC system and processing method of this invention are applicable to flat channels and frequency-selective channel models. Here, we use a frequency-selective channel model, where if the interval between t' and t is greater than the coherent bandwidth, then H[t] ≠ H[t'].

[0116] The block diagram of this system can be found here. Figure 4The main modules include a channel encoder, a codeword-level pre-encoder, a PAM modulator, a signal-level pre-encoder, an integer combination soft-decision detector, and a decoder. The channel encoder, codeword-level pre-encoder, PAM modulator, and signal-level pre-encoder are located at the base station, while the integer combination soft-decision detector and decoder are located at the user end. The specific functions of each module are explained below:

[0117] a) Channel encoder, used to encode each message sequence.

[0118] Let the message data sequence of user i be represented by row vector b. i T Let i = 1, 2, ..., K, where k is the length of the message sequence. Generally, for multivariate b... i T Channel coding can be achieved using formula (21).

[0119] In the actual implementation of the system, b can be made i T The data stream is binary, and it is encoded using mainstream binary LDPC or polar codes. The output codeword sequence is mapped to {0, 1, ..., 2} using an "m-to-1" mapping. m The elements within {-1} are represented as c. i T ∈{0,1,…,2 m -1} n Let i = 1, 2, ..., K. Let the column vector c[t] = [c1[t], ..., c K [t] T This indicates that the t-th sign bit of all K-stream codeword sequences can be encoded using encoders at different rates in the downlink system.

[0120] b) Codeword-level pre-encoder, used to pre-encode the column vector c[t] obtained from the channel encoder to obtain the pre-encoded codeword sequence.

[0121] The base station of the LBC system, based on the channel state information H[t] from the receiver, uses the method provided in step three of the soft-decision detection method to select K linearly independent integer coefficient vectors a1 for each signal sequence within the coherent bandwidth. T [t],…,a K T [t]. Let the integer coefficient matrix A[t] = [a1[t],…,a K [t] T Note that, considering the frequency-selective channel, if the interval between t' and t is greater than the coherence bandwidth, then H[t] ≠ H[t'], therefore A[t] ≠ A[t']. LBC systems require A[t] to be in the frequency-selective channel. A matrix of full rank has a unique inverse.

[0122] In the LBC system, A is used -1 [t] performs codeword-level precoding on c[t] to obtain the precoded codeword sequence:

[0123]

[0124] Here, v[t] = [v1[t],…,v K [t] T Let v l T =[v l [1],…,v l [n]],l=1,…,K, is called the codeword sequence after precoding the l-th path.

[0125] c) PAM modulator:

[0126] By mapping them one by one to 2 using equation (1). m -PAM; symbol sequence x l T =[x l [1],…,x l [n]], l=1,…,K. Let the column vector x[t]=[x1[t],…,x K [t] T It represents the t-th symbol bit of all K-way symbol sequences.

[0127] d) Signal-level pre-encoder, used to pre-encode the codeword sequence obtained after pre-coding at the signal level to generate the transmission signal.

[0128] The LBC system uses an integer-forced precoding matrix for signal-level precoding. The precoding matrix is:

[0129]

[0130] The precoding operation at the base station generates a transmission signal, represented as follows:

[0131] s[t]=P[t]x[t],t=1,…,n, (31)

[0132] Transmitted via multiple antennas at the base station.

[0133] e) An integer combination soft-decision detector, used to calculate, sign-by-sign, the posterior probability of integer combinations of codeword sequences v[t] pre-encoded by the codeword-level pre-encoder.

[0134] The signal received by K users is represented as follows:

[0135] y[t]=H[t]s[t]+z[t]=H[t]P[t]x[t]+z[t], t=1,...,n. (32)

[0136] Wherein, the i-th element y of the column vector y[t] i [t] represents the signal received by the i-th user at time t. Let y i T =[y i [1],…,y i [n]], i = 1, ..., K represents the signal sequence received by the i-th user.

[0137] Consider that the receiver of user i is informed of the coefficient vector a i T [t]. Using the method in step four of the soft-decision detection method described in this invention, the posterior probability of integer combinations of (pre-encoded codeword) v[t] is calculated bit-by-bit, that is:

[0138]

[0139] For the specific calculation steps of the posterior probability, please refer to step four of the soft decision detection method. Only c[t] in step four needs to be replaced with v[t], and the other operations are the same.

[0140] Because codeword-level precoding (29) has been performed in advance, here we have:

[0141]

[0142] Therefore, the posterior probability of the integer combination of v[t] is the codeword c. i The posterior probability of [t], i.e.:

[0143]

[0144] Note that even if the channel changes on each symbol, we still obtain the posterior probability of each user's codeword. Therefore, the method of this invention is applicable to frequency-selective channels.

[0145] f) Decoder, used to perform hard decisions on the posterior probabilities obtained from the integer combination soft-decision detector to obtain the decoding result of the desired message sequence.

[0146] The posterior probability is passed to the channel-coded decoder. Each user performs decoding once, and the decoder output for user i is:

[0147] p(b i [t]), t=1,…,k。 (36)

[0148] The desired message sequence is obtained through hard decision.

[0149] In practice, iterative BP decoding is used for LDPC codes, and serial decoding or serial list decoding is used for polar codes.

[0150] Furthermore, this invention proposes a lattice-based cellular-free MIMO system for ICB soft-decision detection, as detailed below:

[0151] Consider a K-stream user non-cellular MIMO (cf-MIMO) network model, with a total of N... BS There are three distributed base station units (DUs), each connected to the central processing unit (CU) via a backhaul (BH) link. The capacity of the BH link is limited, on the same order of magnitude as the air interface capacity. We still assume each user has a single antenna, and the base station receiver has N antennas.

[0152] The block diagram of the cellular-free MIMO system is as follows: Figure 8 As shown, it includes the following modules: channel coding and modulator, non-cellular network channel, integer combination soft decision detector, channel coding decoder, and CU user data decoder.

[0153] a) Channel coding and modulator, used to encode the user message data sequences.

[0154] Let user i's 2 m Meta-message data sequences are represented by row vector b. i T ∈{0,1,…,2 m -1} k Let i = 1, 2, ..., K, where k is the length of the message sequence. The message data for all K (stream) users can be represented by matrix B = [b1, ..., bk]. K ] T This indicates that the size is K×k. The present invention uses 2... m The ring code encodes the sequence of user message data as follows: i = 1, 2, ..., K; then, 2 is formed through formula (1). m -PAM symbol. All users transmit simultaneously on the same frequency band.

[0155] b) No cellular network channel, used to receive signals from various distributed base stations.

[0156] The receiver at base station j receives the same signal as in equation (2), which is expressed here as:

[0157]

[0158] Base station j wants to generate data B = [b1,…,b] for stream K. K ] T L j A combination of integers, L j The larger the value, the better, provided it does not exceed the BH capacity limit. The base station determines the value based on the channel state information H received from the receiver. j Using the method provided by the definition of integer combinations of multi-stream data described in step three, select L j a linearly independent vector of integer coefficients Let A j =[a j,1 ,…,a j,K ] T The integer coefficient matrix selected for base station j.

[0159] c) Integer Combination (ICB) soft-decision detector, used to calculate the posterior probability of integer combinations of channel-coded K-stream data on a symbol-by-symbol basis.

[0160] For its l-th ICB, base station j uses the ICB soft-decision method described in step four of the soft-decision detection method to calculate the posterior probability of integer combinations of channel-coded K-stream data bit by bit:

[0161]

[0162] Then, pass the posterior probability to 2. m A decoder for meta-channel coding.

[0163] d) A channel-coded decoder, used to decode the posterior probability and output it.

[0164] Decoder output

[0165]

[0166] The judgment is

[0167]

[0168] If the judgment is correct, then the l-th integer combination u is obtained. j,l T =[u j,l [1],…,u j,l [k]]. For details on decoder operation, please refer to step five (d) of the specific implementation plan.

[0169] e) The CU's user data decoder is used to generate decisions based on message-level integer combinations.

[0170] L of base station j j The soft-decision detection and decoding operations of the ICB are carried out in parallel to obtain... It is passed to CU via BH.

[0171] Meanwhile, soft decision and decoding operations from other base stations generate CU collects all integer combinations

[0172]

[0173] like exist A matrix of full rank has a unique inverse. CU can be accessed

[0174]

[0175] Operation restores all user message data B.

[0176] Note that the total BH usage for the backhaul links in this solution is... Bits / symbols, which are on the same order of magnitude as the capacity of the air interface.

[0177] Simulations were conducted considering different numbers of distributed base stations, different numbers of users and antennas, and different code rates. The frame error rate (FER) results were recorded and compared with the baseline scheme. It is evident that, using the ICB soft-decision detection and decoding of this invention, the performance of this cf-MIMO scheme is significantly better than the baseline scheme, and it also has a higher BH utilization rate.

[0178] (III) Advantages and Efficacy:

[0179] This invention proposes a soft-decision detection method and system based on multi-stream data integer combinations. It features low complexity, a parallel processing architecture, and low processing latency, enabling decoding performance close to the limits. The method in this invention realizes the theoretical gains of lattice coding, computation and transmission, physical layer network coding, LR, and IF in practical communication systems. This solves the significant performance loss of existing linear detectors and precoders, as well as the non-convergence problem caused by detector-decoder matching in iterative detection. The results of this invention are highly versatile, forming high-efficiency uplink multi-user detection, downlink precoding, and distributed base station processing in cellular networks. It can also be used to handle inter-symbol or inter-carrier interference. In uplink systems, "full diversity gain" can be obtained, supporting overload transmission with K / N > 300%. In downlink systems, spatial "full multiplexing" can be achieved with linear precoding, approaching the performance limits of MIMO broadcast channels. This invention is particularly suitable for cellular networks, enabling better utilization of air interface and backhaul links. [Attached Image Description]

[0180] Figure 1 The diagram shown is a soft decision detection and decoding block diagram for integer combinations of multi-stream data according to Embodiment 1 of the present invention.

[0181] Figure 2 The figure shows the performance curves of ICB soft-decision detection and decoding in an uplink lattice code multiple access system according to Embodiment 2 of the present invention. The horizontal axis represents the signal-to-noise ratio per user, and the vertical axis represents the frame error rate. Binary LDPC coding and BPSK modulation are used, with a code length of k = 480, a code rate of 1 / 2, and a spectral efficiency of 1 / 2 bits / symbol per user. The number of receiver antennas N = 8, and the number of users K = 16, 20, 24, with total spectral efficiencies of 8, 10, and 12 bits / symbol, respectively. The baseline here is iterative MMSE detection. The selection of the coefficient matrix A for lattice code multiple access (LCMA) considers three methods: rank-constrained sphere decoding (RC-SD), HKZ, and LLL algorithms.

[0182] Figure 3 The figure shows the frame error rate (FER) performance curves of the ICB soft-decision detection and decoding method in an uplink lattice code multiple access (LCMA) system according to Embodiment 2 of the present invention. The FER performance is shown for receiver antenna number N=4 and user number K=8. Quad-ary coding and 4-PAM modulation are used, with the channel coding employing a 4-ary LDPC code in [8,9], a code length of k=256, a code rate of 1 / 2, and a single-user spectral efficiency of 1 bit / symbol per user. The total spectral efficiency is 8 bits / symbol. The performance of ICB soft-decision detection method I and II is shown here. The interference-free lowerbound (LB) provides a lower bound on the lowest achievable bit error rate for this system.

[0183] Figure 4 The diagram shown is a block diagram of a lattice-based downlink MIMO broadcast system according to Embodiment 3 of the present invention.

[0184] Figure 5 The figure shows the bit error rate (BER) performance curve of the lattice-based downlink broadcast system under slow fading channel conditions in Embodiment 3 of the present invention, using regularized integer-forcing (RIF) precoding. The receiver antenna count N = 4, and the number of users K = 4. Here, 5-ary IRA coding and 5-PAM modulation are used. During simulation, each signal-to-noise ratio is implemented at least 500 times, with a code length of 50,000 and an average code rate of 1 / 2. This method is applicable to PAM modulation of any order, number of antennas, and number of users. The baseline system considers zero-forcing (ZF) and regularized zero-forcing (RZF) schemes.

[0185] Figure 6The diagram shows the BER performance of the downlink lattice code multiple access system under a slow fading channel in Embodiment 3 of the present invention, and its comparison with the theoretical performance limit. The number of receiver antennas N = 4, and the number of users K = 4. The discrete points marked with asterisks represent the BER performance obtained using the ICB soft-decision algorithm and decoding of the present invention. It can be seen that the theoretical upper bound of the lattice-based downlink MIMO broadcast system is only about 1.2-1.3 dB. Meanwhile, compared with the baseline planning zero-forcing RZF precoding and ZF precoding, the performance improvement is greater than 5 dB.

[0186] Figure 7 The figure shows the BER performance of the downlink receiver under different numbers of users and modulation orders in a fast fading channel according to Embodiment 3 of the present invention. Multi-level IRA coding and the corresponding modulation order are used here, with a code length of k = 50000 and a code rate of 1 / 2. It can be seen that, using the ICB soft-decision detection of the present invention, the theoretical gap between the lattice-based MIMO broadcast system and RIF is less than 1 dB.

[0187] Figure 8 The figure shown is a block diagram of a lattice-based non-cellular MIMO system according to Embodiment 4 of the present invention.

[0188] Figure 9 The figure shows the FER performance of the lattice-based uplink cellular-free MIMO system proposed in Embodiment 4 of this invention. It uses four distributed base stations, each with N=8 antennas, for a total of K=24 users. Binary coding and BPSK modulation are employed. The channel coding uses the LDPC code of the 5G NR standard, with a code length of k=480, a code rate of 1 / 2, a spectral efficiency of 1 / 2 bits / symbol per user, and a total spectral efficiency of 12 bits / symbol. This figure illustrates the outage probability (OP) and frame error rate performance. The baseline scheme is a scalar quantization compression-forward scheme. It is evident that lattice network coding (LNC) offers significant performance gains.

[0189] Figure 10 The figure shows the FER performance of the uplink non-cellular MIMO system based on lattice in Embodiment 4 of the present invention, with 4 distributed base stations, each base station having N=8 antennas and K=12 users. Here, quaternary LDPC coding and 4-PAM modulation are used. The channel coding adopts the 4-ary LDPC code in [8,9], with a code length of k=256, a code rate of 1 / 2, a spectral efficiency of 1 bit / symbol per user, and a total spectral efficiency of 12 bits / symbol.

[0190] Figure 11The embodiment 4 of this invention illustrates a lattice-based uplink non-cellular MIMO system with 1, 2, 4, or 8 distributed base stations, each with N=8 antennas and K=24 users. It employs binary LDPC coding and BPSK modulation, with a code length of k=480, a code rate of 1 / 2, a spectral efficiency of 1 bit / symbol per user, and a total spectral efficiency of 12 bits / symbol.

[0191] Figure 12a Figures 1 and 2 show the performance of the uplink non-cellular MIMO system based on a lattice in Embodiment 4 of the present invention, with 4 distributed base stations, each with N = 32 antennas and K = 32, 40, and 48 users. Quadratic LDPC encoding and 4-PAM modulation are used, with a spectral efficiency of 1 bit / symbol per user and a total spectral efficiency of 32, 40, and 48. The left figure shows the outage probability performance, and the right figure shows the BH consumption. It can be seen that the transmission rate and BH consumption are on the same order of magnitude.

Detailed Implementation Methods

[0192] To gain a better understanding of the principles, methods, features, and performance advantages of this invention, it is described in detail below.

[0193] Example 1: A soft-decision detection method for positive combinations of multi-stream data, as detailed below:

[0194] Step 1: Send signal

[0195] Consider K-stream message data, using row vector b1 T ,…,b K T Let the row vector c be represented. i T Let c represent the i-th data stream after channel coding, where i = 1, 2, ..., K, and the data stream length is n. i [t] indicates Let the t-th sign bit be t, where t = 1, ..., n. Let the column vector c[t] = [c1[t], ..., c K [t] T This represents the t-th sign bit of all K-stream data.

[0196] Consider 2 m Meta-channel coding, m = 1, 2, ... Thus, c i [t]∈{0,…,2 m -1}, meaning its elements are no greater than 2. m -1 is a non-negative integer. The channel-coded data stream sequence is symbolically mapped to 2... m The PAM modulated signal sequence is as follows:

[0197]

[0198] Where γ is the normalization factor, ensuring that sequence x i T The average energy is 1. Here, x i T All elements are integers divided by γ. All K-stream signals are transmitted simultaneously.

[0199] For the complex model, two independent coding and modulation methods are used, transmitted separately in the in-phase and quadrature parts, forming a 2:1 I / Q ratio. 2m -QAM modulation. This conforms to the 2QAM modulation widely used in mainstream communication systems. m -PAM and 2 2m -QAM modulation.

[0200] Step 2: Receiving Signals

[0201] Consider a receiver with a received signal spatial dimension of N. (For example, the receiver is equipped with N antennas, each antenna providing one observation. Alternatively, the system's spread spectrum sequence length is N, with each chip-level signal providing one observation.)

[0202] For the real number model, the received signal is represented as:

[0203]

[0204] Where h i H represents the channel vector from the i-th stream signal to the N observations at the receiver; H = [h1, ..., h K [x1, ..., x2] represents the channel matrix, containing the channel vectors corresponding to all streaming signals; matrix X = [x1, ..., x3] K ] T Let Y represent all K-stream signal sequences, with the i-th row being the i-th stream signal; Z represents the additive white Gaussian noise (AWGN) matrix, where each element is an independent and identically distributed zero-mean, unit-variance Gaussian noise; ρ represents the average energy of each stream signal, which is equivalent to the signal-to-noise ratio. Here, Y = [y[1],…,y[n]], y[t] is the received signal vector of the t-th symbol bit.

[0205] A complex model can be represented by a real model of two dimensions, that is:

[0206]

[0207] For clarity, this invention is described using a real number model.

[0208] Step 3: Definition of Integer Combinations for Multistream Data

[0209] Consider a vector of length K (not all zero) integer coefficients. Another one in The integer combination (ICB) of c[t] is represented as follows:

[0210]

[0211] Here mod(·,2) m ) indicates modulo 2 m The range of values ​​for the operation and the combination of integers is:

[0212] Generally, L-path ICB is represented as:

[0213]

[0214] This represents the integer coefficient vector corresponding to the l-th ICB.

[0215] This invention applies to any coefficient vector a1 T ,…,a L T The optimal a1 T ,…,a L T The method for selecting a1 is not the focus of this invention. The following is a brief description of a1. T ,…,a L T The optimized selection method.

[0216] This step requires identifying the optimal integer coefficient matrix A. The following two methods can be used in implementation.

[0217] Method 1 , Lenstra–Lenstra–Lovász (LLL) lattice reduction

[0218] Considering the channel matrix H, its MMSE matrix (I+ρH) T H) -1 Perform eigenvalue decomposition to obtain

[0219] (I+ρH T H) -1 =ΨΣΨ T ,

[0220] Ψ is the matrix formed by the eigenvectors. The optimal coefficient matrix A is the solution to the following optimization problem.

[0221]

[0222] This optimization problem can be described as: Let Indicated by Σ 1 / 2 Ψ T For all grid points formed by the basis vector set. The goal is to find a set of L grid points with distinct orientations and the shortest maximum length. This optimization problem is NP-hard, but several existing algorithms can find an approximate optimal solution in polynomial time, such as the LLL algorithm.

[0223] LLL reduction basis definition: Let d1,…,d1,…,d2 M It is a set of lattice bases, and the lattice space formed by them is denoted as . d1,…,d M The vector set obtained after Schmitt orthogonalization is If satisfied

[0224] 1. Size-reduce condition: For any m2 < m1 ≤ M, in The Schmidt orthogonalization coefficients are... This is an inner product operation;

[0225] 2. Lovasz condition: For any d m-1 ,d m (m=2,…,M), in

[0226] Then d1,…,d M Therefore, Σ 1 / 2 Ψ T The set of lattice points generated for the basis vectors A set of LLL reduced bases. The size-reduce condition ensures that the vectors in the LLL reduced base are relatively short and approximately orthogonal, while the Lovasz condition roughly orders the basis vectors. Since the LLL reduced base is not... The LLL algorithm finds the strictly shortest basis vectors in the LLL, so the result obtained is not the optimal solution, but this approximate optimal solution is sufficient to achieve better performance.

[0227] The LLL algorithm finds Σ 1 / 2 Ψ T The lattice space formed by column vectors The LLL reduction group is the LLL reduction group. The approximate shortest basis vector in Σ. 1 / 2 Ψ T The linear transformation matrix between the LLL reduced basis and the LLL reduced basis is the desired optimized integer coefficient matrix.

[0228] Algorithm 1: Solving the optimized A using LLL

[0229] Inputs: Channel parameters H, signal-to-noise ratio ρ, Lovasz parameter α;

[0230] Output: Integer coefficient matrix A;

[0231] [Ψ,Σ]=eig((I+ρH T H) -1 ), where Σ is the eigenvalue diagonal matrix and Ψ is the orthogonal matrix.

[0232] D=Σ 1 / 2 Ψ T

[0233] D1 = LLL(D, α)

[0234] A = D -1 D1

[0235] return A

[0236] Where eig(·) is the eigenvalue decomposition function, and LLL(·) is the LLL algorithm in reference [1]. In specific implementation, given the channel parameters H and the signal-to-noise ratio, the optimized coefficient matrix A can be obtained using algorithm 1, thereby obtaining the optimal linear filter matrix W and completing the decoding process. In this invention, α = 0.99 is taken.

[0237] Method 2 spherical decoding

[0238] For ΨΣΨ T Perform Choleski decomposition, i.e., ΨΣΨ T =ΠΠ T Let Π be an upper triangular matrix. Given a radius r centered at the zero point, use a list-based spherical decoding method based on tree search to search for all grid points within this radius, forming a list. Find L vectors in the integer ring from the candidate vectors in the list using a greedy algorithm. an integer coefficient vector a1 with rank L T ,…,a L T This yields the coefficient matrix A. If the rank L cannot be achieved, the radius r is increased appropriately, and the list spherical decoding is run again until the rank condition is met.

[0239] Spherical decoding can provide a coefficient matrix A that is closer to the optimal solution than the LLL algorithm, but its complexity is increased.

[0240] Step 4: (The core algorithm of this invention)

[0241] a) Linear filtering

[0242] The ICB soft-decision detection method proposed in this invention is applicable to any matrix W to implement the linear filtering in equation (7). This paper takes the regularized integer forcing (RIF) method as an example, where the l-th row of the filtering matrix W...

[0243]

[0244] In this embodiment, the N-dimensional received signal is converted into an L-stream single-dimensional signal using RIF filtering. Then, each stream is used to calculate the posterior probability of an ICB. Note that the method proposed in this invention is applicable to any W. In practice, regularized integer forcing (RIF) can be used to form the filtering matrix W.

[0245] b) Signal Expression

[0246] In order to calculate Given the posterior probability of the received signal, we make the following equivalent expression for equation (7). Let... Collect a l The position of the non-zero term, let Indicates its complement. Let Indicates a l The number of non-zero terms. Thus, formula (7) can be expressed as:

[0247]

[0248] here, This item indicates a l ω(a) with a non-zero coefficient l The superposition of signals from 10 users is the useful signal component for calculating ICB; Including the rest of K-ω(a) l The user's signal, its corresponding a l The coefficient is zero, indicating that it is not correlated with the ICB. It is considered as equivalent noise, and it is uncorrelated with the useful signal portion. For a sufficiently large K, It's also large enough. Applying the central limit theorem, the equivalent noise ξ... l It follows a Gaussian distribution with a mean of 0 and a variance of .

[0249] Using formula (1) x i With c i bijective relation, i.e. Further simplification of formula (8) yields:

[0250]

[0251] here Independent of the signal, it resembles a DC component, and its purpose is to convert the signal from {-1, +1} to {0, 1} for processing. This is achieved through... After compensation, we get:

[0252]

[0253] After the above simplification, only a remains in the signal part of equation (10). l The signal of the user whose bit is not zero. The operation of calculating the posterior probability of ICB is expressed as:

[0254]

[0255] c) Exact calculation of the likelihood function of integer combinations

[0256] When calculating the posterior probability, i.e., formula (11), the likelihood function is required. The calculation method is as follows. Let the vector Only contains a l The non-zero elements of the vector, and let the vector Only includes the corresponding a in c l The part of non-zero elements (belonging to) (part of the text). and The length is Applying the law of total probability:

[0257]

[0258] From equation (10), we obtain:

[0259]

[0260] If we directly calculate the likelihood function (12) here, we need to calculate... Candidates vector The value is so that its complexity is O(n). Order of magnitude. Below, this invention provides a method for efficient computation (12).

[0261] d) Calculation of low-complexity likelihood function based on Gaussian approximation

[0262] The high complexity of calculating the likelihood function precisely stems from the fact that it satisfies the linear equation Exhaustive search of the candidate set. The core idea of ​​using Gaussian approximation is that when K is sufficiently large, for a given... The set of channel received values ​​corresponding to the candidate set can be approximated by the mean value being variance is The Gaussian distribution. Determining this Gaussian distribution requires three statistical values: 1) prior probability. 2) Conditional mean 3) Conditional variance Details will be provided below.

[0263] Note that since these statistics only need to be calculated once per channel implementation (n-length sequence) coherence time (or coherence bandwidth), the computational cost of these statistics is negligible if n is large enough.

[0264] For the sake of simplicity, the index 1 of ICB will be omitted below.

[0265] 1) Prior probability Calculation

[0266] Prior probability It can be determined by the distribution of the number of candidate sets. Indicates satisfaction [a1,…,a] k The number of ] . When k=1, it is obvious that This can be obtained by executing the following formulas in sequence.

[0267]

[0268] At the kth layer, Until K′=ω(a) is reached. This requires a total of no more than [a].

[0269]

[0270] This addition operation does not involve multiplication. Prior probability. can be get.

[0271] 2) Conditional mean

[0272] express The sum of the received signals corresponding to the probabilities of each element in the candidate set is called the equal probability sum of the candidate set. The conditional mean is obtained by dividing the equal probability sum by the number of candidate sets. When k=0, it is obvious that... The conditional mean is obtained by performing the following operations sequentially:

[0273]

[0274] Until the layer K′=ω(a) is reached, the conditional mean is obtained from calculate.

[0275] 3) Conditional variance

[0276] Similarly, we define the sum of squares with equal probability. By executing sequentially

[0277]

[0278] Until layer K′ is reached. The conditional variance can be obtained from the following formula:

[0279]

[0280] The above method is called ICB soft decision method one.

[0281] In practice, using RIF simplifies the acquisition of statistical values. The signal can be represented as...

[0282]

[0283] The estimated error term is

[0284]

[0285] The error term e here l With useful signal part It is correlated. This leads to and They should be calculated as (57) and (59) respectively.

[0286] For a sufficiently large K, e can be used. l Applying the central limit theorem, e l Approximately a variance of The Gaussian random variable is denoted by . It can be easily proven that e l The MSE has a closed expression

[0287]

[0288] Furthermore, by disregarding the bias in the estimation error term, e l The average value is approximated as zero. Therefore, the calculation in (17) is further simplified to

[0289]

[0290] This method is called ICB soft decision method two.

[0291] Note the prior probability. The calculations are the same as above, but the calculations of the conditional mean and variance can be greatly simplified. For small K, ICB soft-decision method one is used because the loss of ICB soft-decision method two may be significant. For sufficiently large K, ICB soft-decision method two can be used, and its performance difference from the previously used method is small enough.

[0292] e) Calculate the posterior probability of ICB

[0293] Based on the likelihood function of ICB, i.e., formula (18), and using Bayes' theorem, the posterior probability of ICB is:

[0294]

[0295] Here, η is a normalization factor, which ensures that the calculated soft decision terms are normalized. Adding them together gives 1. The second step of equation (19) utilizes the equal probability property of integer combinations, that is... The calculation result of the posterior probability of the ICB (i.e., soft decision information) is passed to the channel coding decoder for decoding to obtain the decision of the integer combination of multi-stream message data.

[0296] f) Characterization of complexity

[0297] The complexity of the ICB soft-decision detection in this invention mainly comes from the calculation of the likelihood function in (18) and (17), and its complexity depends on the order of magnitude of the calculation. The number of integer values ​​that can be taken. Because The maximum value is Minimum value is therefore,

[0298]

[0299] So, The number of integer values ​​that can be obtained is ω H (a l (2) m -1)+1. In other words, ω needs to be calculated in the likelihood function (18). H (a l (2) m -1)+1 probability values ​​(17). It is proposed that the complexity of the ICB soft-decision detection of this invention is O(ω). H (a l (2) m -1)+1) level, far lower than the level required for direct execution (12). The complexity is at level 1.

[0300] Example 2

[0301] The present invention further provides an application of the soft-decision detection method in a grid code multiple access system, as detailed below:

[0302] a) Channel coding and modulation

[0303] In implementation, for m=1, i.e., binary coding and BPSK, LDPC codes or Polar codes from the 5G NR standard can be used. For m=2,3,... and 2 m -PAM modulation, this invention recommends using 2 m The original integer ring LDPC code or irregular repeat accumulate (IRA) code[2][3]. This can guarantee that the coding modulation belongs to the lattice code, satisfies the properties of lattice codes, and has a bit error rate performance that approaches the limit.

[0304] b) Received signal

[0305] 2 m The ring code (or linear code) has the "superposition property," meaning that when K usable codewords are superimposed in multiples, the resultant codeword is equal to 2^k. m The codeword is still usable after modulo operation, that is...

[0306]

[0307] These are still available codewords in the codebook.

[0308] By utilizing the additivity of codewords, we can obtain

[0309]

[0310] In other words, the codeword-level ICB and the message-level ICB are also linked by the channel coding generation matrix G. Based on this property, decoding can be achieved through the following steps:

[0311] 1. The soft-decision detector calculates the symbol-by-symbol posterior probabilities (APPs) of the codeword-level ICB, i.e., p(v l [t]|y[t]),t=1,…,n. Here v l [t] and y[t] represent v l And the t-th column of Y. See step c) below.

[0312] 2. The decoder takes the codeword-level ICB's APP as input, performs the decoding operation, and outputs a message-level ICB. l The verdict. See step d) below.

[0313] c) ICB Soft Decision Detection (The specific steps are the same as in step four of Example 1, and will not be repeated here.)

[0314] d) Decoding of channel coding

[0315] For m=1, i.e., binary code combined with BPSK, if LDPC encoding is used, then iterative belief propagation (BP) decoding is employed. If polar encoding is used, then serial list decoding is employed. For m=2,3,…, i.e., using 2… m Combining LDPC or IRA ring codes 2 m -PAM modulation, then 2 m Meta-iterative belief propagation (BP) decoding. Through FFT / IFFT transformations, 2... m The computational complexity of the meta-verification node can be further reduced, see [8, 9] for details.

[0316] e) User Data Recovery

[0317] The soft-decision detection and decoding operations for K-way integer combinations are performed in parallel, resulting in:

[0318]

[0319] Because A is A matrix of full rank has a unique inverse matrix A. -1 : Accessible via:

[0320]

[0321] Operation restores all user message data B.

[0322] f) Simulation and Performance Evaluation

[0323] Figure 2 This paper demonstrates the frame error rate (FER) performance of ICB soft-decision detection and decoding in an uplink lattice code multiple access system. Binary LDPC coding and BPSK modulation are used, with a code length of k = 480, a code rate of 1 / 2, and a per-user spectral efficiency of 1 / 2 bits / symbol. The number of receiver antennas N = 8, and the number of users K = 16, 20, and 24, with total spectral efficiencies of 8, 10, and 12 bits / symbol, respectively. It is evident that this method supports a higher number of users and achieves a lower FER compared to the baseline iterative MMSE detection and decoding method. Here, we see that the system performance is highly dependent on the choice of coefficient matrix A. Specifically, the method of finding A using rate-constrained spherical decoding (RC-SD) achieves better FER performance than the LLL method.

[0324] Figure 3The performance of FER using ICB soft-decision detection and decoding in an uplink trellis code multiple access system is demonstrated. The receiver antenna count is N = 4, and the number of users is K = 8. Quad-ary coding and 4-PAM modulation are used. The channel coding uses a 4-ary LDPC code in [8,9] with a code length of k = 256 and a code rate of 1 / 2. The single-user spectral efficiency is 1 bit / symbol per user. The total spectral efficiency is 8 bits / symbol. It can be seen that this method, compared to the baseline iterative MMSE detection and decoding method, can support a higher number of users and has a lower FER. The performance of our method is approximately 2.6 dB below the lower bound of the FER under the assumption of interference-free conditions. The ICB soft-decision method mentioned in step four, using (17) (referred to as method one, detection method 1) and (63) (referred to as method two, detection method II), has a certain performance difference. Although method two eliminates the calculation of conditional mean and conditional variance, the performance loss is not negligible.

[0325] Example 3

[0326] This invention proposes a lattice-based downlink MIMO broadcast system applying this soft-decision detection method, as detailed below:

[0327] A lattice-based MIMO broadcasting (LBC) system utilizes the aforementioned ICB soft-decision detection method. Consider a base station transmitting its data streams to K users. The base station is equipped with N antennas; considering each user has a single antenna, it can be easily extended to multiple antennas at the user end. OFDM modulation is considered, eliminating inter-symbol interference. The base station has known channel state information. The LBC system and processing method of this invention are applicable to flat channels and frequency-selective channel models. Here, we use a frequency-selective channel model, where if the interval between t' and t is greater than the coherent bandwidth, then H[t] ≠ H[t'].

[0328] The block diagram of this system can be found here. Figure 4 The main modules include a channel encoder, a codeword-level pre-encoder, a PAM modulator, a signal-level pre-encoder, an integer combination soft-decision detector, and a decoder. The channel encoder, codeword-level pre-encoder, PAM modulator, and signal-level pre-encoder are located at the base station, while the integer combination soft-decision detector and decoder are located at the user end. The specific functions of each module are explained below:

[0329] a) Channel encoder, used to encode each message sequence.

[0330] Let the message data sequence of user i be represented by row vector b. i T Let i = 1, 2, ..., K, where k is the length of the message sequence, for a multivariate b i T Channel coding uses the formula

[0331] Let b i T The data stream is binary, and it is encoded using binary LDPC or polar codes; the output codeword sequence is mapped to {0,1,…,2} using an "m-to-1" mapping. m The elements within {-1} are represented as c. i T ∈{0,1,…,2 m -1} n Let i = 1, 2, ..., K; let column vector c[t] = [c1[t], ..., c K [t] T This indicates that the t-th sign bit of all K-stream codeword sequences is in the downlink system;

[0332] b) Codeword-level pre-encoder, used to pre-encode the column vector c[t] obtained from the channel encoder to obtain the pre-encoded codeword sequence.

[0333] The base station of the LBC system, based on the channel state information H[t] from the receiver, uses the soft-decision detection method described above to select K linearly independent integer coefficient vectors a1 for each signal sequence within the coherent bandwidth. T [t],…,a K T [t]; Let the integer coefficient matrix A[t] = [a1[t],…,a K [t] T Since frequency-selective channels are considered, if the interval between t' and t is greater than the coherence bandwidth, then H[t] ≠ H[t'], therefore A[t] ≠ A[t']. LBC systems require A[t] to be in... A matrix of full rank has a unique inverse matrix A[t]. -1 :

[0334] In the LBC system, A is used -1 [t] performs codeword-level precoding on c[t] to obtain the precoded codeword sequence:

[0335]

[0336] Here, v[t] = [v1[t],…,v K [t] T Let v lT =[v l [1],…,v l [n]], l=1,…,K, is called the codeword sequence after precoding the l-th path;

[0337] c) PAM modulator

[0338] By mapping them one by one to 2 using equation (1). m -PAM; symbol sequence x l T =[x l [1],…,x l [n]], l=1,…,K. Let the column vector x[t]=[x1[t],…,x K [t] T Represents the t-th sign bit of all K-way sign sequences;

[0339] d) Signal-level pre-encoder, used to pre-encode the codeword sequence obtained after pre-coding at the signal level to generate the transmission signal.

[0340] The LBC system uses an integer-forced precoding matrix for signal-level precoding. The precoding matrix is:

[0341]

[0342] The precoding operation at the base station generates a transmission signal, represented as follows:

[0343] s[t]=P[t]x[t],t=1,…,n, (72)

[0344] Transmitted via multiple antennas at the base station;

[0345] e) An integer combination soft-decision detector, used to calculate, sign-by-sign, the posterior probability of integer combinations of codeword sequences v[t] pre-encoded by the codeword-level pre-encoder.

[0346] The signal received by K users is represented as follows:

[0347] y[t]=H[t]s[t]+z[t]=H[t]P[t]x[t]+z[t],t=1,…,n; (73)

[0348] Wherein, the i-th element y of the column vector y[t] i [t] represents the signal received by the i-th user at time t. Let y i T =[y i [1],…,y i [n]], i = 1, ..., K represents the signal sequence received by the i-th user.

[0349] Consider that the receiver of user i is informed of the coefficient vector a i T [t]; calculate the posterior probability of integer combinations of v[t] bit by bit, i.e.:

[0350]

[0351] Because codeword-level precoding (29) has been performed in advance, here we have:

[0352]

[0353] Therefore, the posterior probability of the integer combination of v[t] is the codeword c. i The posterior probability of [t], i.e.:

[0354]

[0355] f) Decoder, used to perform hard decisions on the posterior probabilities obtained from the integer combination soft-decision detector to obtain the decoding result of the desired message sequence.

[0356] The posterior probability is passed to the channel-coded decoder. Each user performs decoding once, and the decoder output for user i is:

[0357] p(b i [t]), t=1,…,k; (77)

[0358] The desired message sequence is obtained through hard decision.

[0359] g) Simulation and Performance Evaluation

[0360] Simulations were performed considering different modulations, user numbers, and antenna numbers. The bit error rate (BER) results were recorded, and different detection methods were compared. Comparisons were also made with existing MMSE and ZF baseline precoding methods.

[0361] Figure 5 and Figure 6This paper demonstrates the BER performance of the proposed ICB soft-decision and lattice-based downlink MIMO broadcast system in a slow-fading channel (flat channel), and compares it with theoretical and baseline schemes. The receiver antenna count is N=4, the number of users is K=4, 5-ary IRA coding and 5-PAM modulation are used, and at least 500 channel implementations are performed for each signal-to-noise ratio in the simulation. The code length is 50,000, and the average code rate is 1 / 2. It is evident that this system exhibits better BER performance compared to the baseline RZF and ZF precoding methods. The performance gap between our method and the 5-PAM limit rate is approximately 1.3 dB. This gap is partly due to the difference in coding performance and Shannon limit, and partly due to the fact that in a slow-fading channel, the available channel coding rate is not continuous, and in practice, not all channel resources can be fully utilized during coding.

[0362] Figure 7 The BER performance of the proposed method in a fast fading channel (frequency-selective channel) is demonstrated. Here we compare the performance of two detection methods (13) and (17). It can be seen that the theoretical gap between the lattice-based MIMO broadcast system and RIF using the ICB soft-decision detection of this invention is less than 1 dB.

[0363] Example 4

[0364] This invention proposes a lattice-based cellular-free MIMO system for ICB soft-decision detection, as detailed below:

[0365] Consider a K-stream user non-cellular MIMO (cf-MIMO) network model, with a total of N... BS There are three distributed base station units (DUs), each connected to the central processing unit (CU) via a backhaul (BH) link. The capacity of the BH link is limited, on the same order of magnitude as the air interface capacity. We still assume each user has a single antenna, and the base station receiver has N antennas.

[0366] The block diagram of the cellular-free MIMO system is as follows: Figure 8 As shown, it includes the following modules: channel coding and modulator, non-cellular network channel, integer combination soft decision detector, channel coding decoder, and CU user data decoder.

[0367] e) Channel coding and modulator, used to encode the user message data sequences.

[0368] Let user i's 2 m Meta-message data sequences are represented by row vector b. i T ∈{0,1,…,2 m -1}k Let i = 1, 2, ..., K, where k is the length of the message sequence; the message data for all K stream users can be represented by matrix B = [b1, ..., bk]. K ] T It is indicated that the size is K×k; this invention uses 2 m Metaring codes encode the sequence of user message data as follows: Then, 2 is formed through formula (1). m -PAM symbol; all users transmit simultaneously on the same frequency band;

[0369] f) No cellular network channel is used to receive signals from various distributed base stations.

[0370] The signal received by the receiver at base station j is represented as follows:

[0371]

[0372] Base station j wants to generate data B = [b1,…,b] for stream K. K ] T L j A combination of integers, L j The larger the capacity H, the better, provided it does not exceed the BH capacity limit; the base station determines the H based on the channel state information received from the receiver. j Select L j a linearly independent vector of integer coefficients Let A j =[a j,1 ,…,a j,K ] T The integer coefficient matrix selected for base station j;

[0373] g) An integer combination soft-decision detector, used to calculate the posterior probability of integer combinations of channel-coded K-stream data on a symbol-by-symbol basis.

[0374] For its l-th integer combination, base station j uses the integer combination soft decision method to calculate the posterior probability of the integer combination of the channel-coded K-stream data bit by bit:

[0375]

[0376] Then, pass the posterior probability to 2. m A decoder for meta-channel coding;

[0377] h) A channel-coded decoder, used to decode the posterior probability and output it.

[0378] Decoder output

[0379]

[0380] The judgment is

[0381]

[0382] If the judgment is correct, then the l-th integer combination u is obtained. j,l T =[u j,l [1],…,u j,l [k]];

[0383] e) The CU's user data decoder is used to generate decisions based on message-level integer combinations.

[0384] L of base station j j The soft-decision detection and decoding operations for the combination of road integers are performed in parallel to obtain... It is passed to CU via BH;

[0385] Meanwhile, soft decision and decoding operations from other base stations generate CU collects all integer combinations

[0386]

[0387] like exist A matrix of full rank has a unique inverse matrix A. CU -1 : CU can be accessed via:

[0388]

[0389] Operation restores all user message data B.

[0390] The total backhaul link BH usage of this system is Bits / symbols, which are on the same order of magnitude as the capacity of the air interface.

[0391] f) Simulation and Performance Evaluation

[0392] Simulations were conducted considering different numbers of distributed base stations, different numbers of users and antennas, and different code rates. The frame error rate (FER) results were recorded and compared with the baseline scheme. It is evident that, using the ICB soft-decision detection and decoding of this invention, the performance of this cf-MIMO scheme is significantly better than the baseline scheme, and it also has a higher BH utilization rate.

[0393] Figure 9This paper demonstrates the FER performance of an uplink cellular-free MIMO system with 4 DUs (Dedicated Units), each with N=8 antennas and K=24 users. Binary coding and BPSK modulation are employed, with channel coding using LDPC codes from the 5G NR standard. The code length is k=480, the code rate is 1 / 2, and the spectral efficiency is 1 / 2 bits / symbol per user. The baseline scheme uses compression-forward, considering both vector quantization and scalar quantization. It is evident that, under the same BH consumption, the proposed cf-MIMO scheme exhibits better FER performance. Figure 10 The FER performance of four DUs, each with N=8 antennas and K=12 users, is demonstrated. Here, m=2, indicating the use of quaternary coding and 4-PAM modulation. The channel coding employs a quaternary LDPC code in [8,9] with a code length of k=256, a code rate of 1 / 2, and a spectral efficiency of 1 bit / symbol per user. Similarly, under the same BH consumption, the cf-MIMO scheme proposed in this invention exhibits better FER performance.

[0394] Figure 11 The performance of an uplink cellular-free MIMO system is demonstrated with 1, 2, 4, and 8 DUs, each with N=8 antennas and K=24 users. Here, m=1, indicating the use of binary coding and BPSK modulation. The channel coding adopts the LDPC code of the 5G NR standard, with a code length of k=480, a code rate of 1 / 2, and a spectral efficiency of 1 / 2 bits / symbol per user. It is evident that the system's FER performance improves with the increase in the number of DUs. Here, the BH consumption of a single DU decreases with the increase in the number of DUs.

[0395] Figure 12a Figure 1b demonstrates the performance of a cellular-free massive MIMO system with 4 duplexers (DUs), each DU having N=32 antennas, and K=32, 40, and 48 users. Here, m=2, indicating the use of quaternary coding and 4-PAM modulation, with a spectral efficiency of 1 bit / symbol per user. Figure 12a For interrupt probability performance, Figure 12b This refers to backhaul consumption. It's clear that the transmission rate and backhaul consumption are on the same order of magnitude.

Claims

1. A soft-decision detection method for multi-stream data integer combinations, characterized in that: The specific implementation steps are as follows: Step 1: Send signal consider K Message data from streaming users is represented by row vectors. Representation; let row vector Represents the channel-coded first... i Streaming data, The data stream length is n ;make express The t sign bit, Let column vectors Indicates all K The first of streaming data t Sign bit; consider Meta-channel coding, where m is a positive integer; That is, the element is not greater than Non-negative integers; the channel-coded data stream sequence is mapped symbol by symbol to The PAM modulated signal sequence is as follows: (1) in, As a normalization factor, to ensure the sequence The average energy is 1; All elements are integers divided by ;all K Simultaneous transmission of streaming signals; For the complex model, two independent coding and modulation methods are used, transmitted separately in the in-phase and quadrature components, to form the I / Q signal. -QAM modulation; Step 2: Receiving Signals Considering the spatial dimension of the received signal of the receiver is N ; For the real number model, the received signal is represented as: (2)； in, Indicates the first i Streaming signal to receiver N Channel vectors for each observation; The channel matrix represents the channel vectors corresponding to all streaming signals; the matrix... Indicates all K Streaming Signal Sequences, Chapter i Behavior No. i stream signal; This represents an additive white noise matrix, where each element is independent and identically distributed Gaussian noise with zero mean and unit variance. This represents the average energy of each stream signal, and is equivalent to the signal-to-noise ratio; For the first t The received signal vector of the sign bit; A complex model can be represented by a real model of two dimensions, that is: (3)； Step 3: Definition of Integer Combinations for Multistream Data Consider a length of K integer coefficient vector ; another one in The above about The integer combination is represented as: (4)； Among them, mod ( ) represents the modulus The range of values ​​for the operation and the combination of integers is: ; L The combination of integers is represented as: (5) in, Indicates the first l The vector of integer coefficients corresponding to the integer combinations of paths; Step 4: Calculate the posterior probability of integer combinations of multi-stream data The receiver is based on the received signal calculate L Combinations of integers; Review the range of integer combinations. The operation for calculating the posterior probability of integer combinations is represented as: (6) in, ; For a given L Integer coefficient vector The operation on formula (6) is as follows: a) Linear filtering; make For a size of Let be a linear filtering matrix, with each element being a real number; let express The Okay, and normalize it into ; after filtering, the result is L Road signal: (7)； in, The equivalent gain is a real value, and the noise term is... The variance is 1; b) Signal representation; In order to calculate The posterior probability of the received signal, i.e., formula (7), can be expressed as follows: Let collect The position of the non-zero term, let Indicates complement; express The number of non-zero terms; Formula (7) can be expressed as: ；(8) in, This item indicates Coefficient is not zero The superposition of signals from individual users is used to calculate the useful signal portion of integer combinations; Include the rest The user's signal, corresponding The coefficient is zero, and it is unrelated to integer combinations. Considered as equivalent noise, it is uncorrelated with the useful signal portion; for a sufficiently large , It is also large enough; applying the central limit theorem, equivalent noise It follows a Gaussian distribution with a mean of 0 and a variance of . ; Using formula (1) and bijective relation, i.e. Further simplifying formula (8), we get: (9) in, It is unrelated to the signal; for transmission After compensation, we get: (10) In equation (10), only the signal part remains. The signal of the user whose bit is not zero; the operation of calculating the posterior probability of integer combinations is represented as: (11) in, ; c) Exact calculation of the likelihood function for integer combinations; When calculating the posterior probability, i.e., formula (11), the likelihood function is required. The calculation method is as follows: Let vector Only contains The non-zero elements of the vector, and let the vector Only contains Chinese correspondence The part of non-zero elements; and The length is =| |; Applying the law of total probability: (12) From equation (10), we get: (13) d) Calculation of low-complexity likelihood function based on Gaussian approximation; consider and There exists a "many-to-one" mapping; calculate first. likelihood function Later transformed ; Let set Collect and satisfy of The candidate sequence; for a given , The conditional mean is: ；(14) The conditional variance is: (15) If the sent signal satisfies The received signal is represented as: (16) when When large enough, for a given , Approximately the mean is The variance is The Gaussian distribution is denoted by: The likelihood function is expressed as: ；(17) Then, using the law of total probability, we obtain the likelihood function for the combination of integers: ；(18) e) Calculate the posterior probability of the integer combination; Based on the likelihood function of integer combinations, i.e., formula (18), and using Bayes' theorem, the posterior probability of integer combinations is: ；(19) in, This is a normalization factor to ensure that all terms of the calculated soft decision are normalized. The second step of equation (19) utilizes the equal probability property of integer combinations, namely... The calculation result of the posterior probability of the integer combination, i.e. the soft decision information, is passed to the decoder of the channel coding to perform decoding operations and obtain the decision of the integer combination of multi-stream message data.

2. The soft-decision detection method according to claim 1, characterized in that: This method further includes its application in gridcode multiple access systems, the specific process of which is as follows: a) Channel coding and modulation Allow users i of Meta-message data sequences use row vectors express , The length of the message sequence; all K User message data is used in a matrix Indicate; use Metaring codes encode the sequence of user message data as follows: (20) Then, it is formed by formula (1) -PAM symbol; All users transmit simultaneously on the same frequency band; b) Receiving signals The base station receiver receives the signal as shown in equation (2); based on the channel state information of the receiver, the base station uses the soft-decision detection method to select the integer combination of multi-stream data according to the definition of the multi-stream data. L = K a linearly independent vector of integer coefficients ;make , is called an integer coefficient matrix, in Full rank; define the integer combination of message data as: ； (21) The receiver needs to calculate first. K Road Integer Combinations Afterwards, restore all users' message data. ; c) Integer combination soft decision detection For the first l The receiver uses the integer combination soft decision method to calculate the channel-coded integer combination bit by bit. K Posterior probability of integer combinations of streaming data: ； (22) Then, pass the posterior probability to A decoder for meta-channel coding; d) Decoding of channel coding Decoder output: ； (23) The judgment is as follows: ； (24) If the judgment is correct, then [the person / entity] will receive [the reward / benefits]. K User message data l Road integer combinations: ，(25) as well as ; e) User data recovery K The soft-decision detection and decoding operations for the combination of integers are performed in parallel, resulting in: ； (26) Because A is A matrix of full rank has a unique inverse. ,pass: (27) Operation restores all user message data B.

3. A lattice-based downlink MIMO broadcast system, which specifically applies the soft-decision detection method as described in claim 1, as follows: A lattice-based downlink MIMO broadcast system, hereinafter referred to as an LBC system, includes the following modules: channel encoder, codeword-level precoder, PAM modulator, signal-level precoder, integer combination soft-decision detector, and decoder; among which... The channel encoder, codeword-level pre-encoder, PAM modulator, and signal-level pre-encoder are located at the base station, while the integer combination soft-decision detector and decoder are located at the user end. Specifically, the functions of each module are as follows: a) Channel encoder, used to encode each message sequence; Allow users i The message data sequence is represented by row vectors. express , for diversity Channel coding uses the formula ; make The data stream is binary, and it is encoded using binary LDPC or polar codes; the output codeword sequence uses " m Mapping 1 to The elements inside are represented as , Let column vectors Indicates all K The first of the stream codeword sequence t The sign bit is in the downlink system; b) Codeword-level pre-encoder, used to process the column vector obtained from the channel encoder. Perform codeword-level precoding to obtain the precoded codeword sequence; The base station of the LBC system uses the channel state information from the receiver. Using the soft-decision detection method described above, for each signal sequence within a coherent bandwidth, select... K a linearly independent vector of integer coefficients Let the integer coefficient matrix Because frequency-selective channels are considered, i.e., if and If the interval is greater than the coherence bandwidth, then ,so LBC system requirements exist A matrix of full rank has a unique inverse. ; In the LBC system, using right Perform codeword-level precoding to obtain the precoded codeword sequence: ； (28) in, ;make , is called the first l The path is through the pre-encoded codeword sequence; c) PAM modulator: By using equation (1) Mapped one by one to -PAM; symbol sequence Let column vectors Indicates all K The first of the road symbol sequence t Sign bit; d) Signal-level pre-encoder, used to pre-encode the codeword sequence obtained after pre-coding at the signal level to generate the transmission signal; The LBC system uses an integer-forced precoding matrix for signal-level precoding. The precoding matrix is: ；(29) The precoding operation at the base station generates a transmission signal, represented as follows: ，(30) Transmitted via multiple antennas at the base station; e) An integer combination soft-decision detector, used to compute the codeword sequence pre-encoded by the codeword-level precoder on a sign-by-sign basis. The posterior probability of integer combinations; K The signal received by a user is represented as follows: ；(31) Among them, column vectors The i element For the first i Signals received by each user; Considering the user i The receiver is told the coefficient vector ; calculate the information bit by bit. The posterior probability of an integer combination, i.e.: ； (32) Because codeword-level precoding (28) was performed in advance, we have: ；(33) Therefore, the calculation yields The posterior probability of an integer combination is the codeword. The posterior probability, i.e.: ；(34) f) Decoder, used to make hard decisions on the posterior probabilities obtained from the integer combination soft-decision detector to obtain the decoding result of the desired message sequence; The posterior probability is passed to the channel-coded decoder, and each user performs decoding once. i Decoder output: ；(35) The desired message sequence is decoded using hard decision.

4. A lattice-based non-cellular MIMO system for performing the soft-decision detection method for multi-stream data integer combinations as described in claim 1, specifically as follows: consider K Streaming user non-cellular MIMO network model, total There are three distributed base station units (DUs), each connected to the central processing unit (CU) via a backhaul link (BH). The capacity of the BH link is limited, on the same order of magnitude as the air interface capacity. Considering that each user is equipped with a single antenna, the base station receiver is equipped with... N One antenna; The block diagram of the cellular-free MIMO system includes the following modules: channel coding and modulator, cellular-free network channel, integer combination soft decision detector, channel coding decoder, and CU user data decoder; a) Channel coding and modulator, used to encode the sequence of user message data; Allow users i of Meta-message data sequences use row vectors express , The length of the message sequence; all K Streaming user message data uses a matrix Indicate; use Metaring codes encode the sequence of user message data as follows: ; Then, it is formed by formula (1) -PAM symbol; All users transmit simultaneously on the same frequency band; b) No cellular network channel, used to receive signals from various distributed base stations; base station j The signal received by the receiver is represented as: ； (36) base station j To generate about K Streaming message data of A combination of integers, The larger the capacity, the better, provided it does not exceed the BH capacity limit; the base station determines this based on the receiver's channel state information. Select a linearly independent vector of integer coefficients ; make For base stations j The selected integer coefficient matrix; c) An integer combination soft-decision detector, used to compute the channel-coded... K The posterior probability of integer combinations of streaming data; base station j Regarding its first l The integer combination method is used to calculate the channel-coded integer combination bit by bit. K Posterior probability of integer combinations of streaming data: ； (37) Then, pass the posterior probability to A decoder for meta-channel coding; d) A channel-coded decoder for decoding and outputting the posterior probability; Decoder output: ；(38) The judgment is as follows: ； (39) If the judgment is correct, then the [number]th [item] is obtained. l Road Integer Combinations ; e) The CU's user data decoder is used to generate decisions based on message-level integer combinations; base station j of The soft-decision detection and decoding operations for the combination of road integers are performed in parallel to obtain... It is passed to the CU via BH; Meanwhile, soft decision and decoding operations from other base stations generate ;CU collects all integer combinations ； (40) like exist A matrix of full rank has a unique inverse. CU passed: (41) Operation restores all user message data B.

5. The system according to claim 4, characterized in that: The total backhaul link BH usage of this system is Bits / symbols, which are on the same order of magnitude as the capacity of the air interface.

Images