SCMA-based dual dynamic optimization multi-user detection method in uplink NOMA system

Abstract

The invention relates to a dual dynamic optimization multi-user detection method based on SCMA in an uplink NOMA system, and belongs to the technical field of wireless communication. According to the method, the technical problems of insufficient multi-user interference suppression caused by codebook cross correlation enhancement in a high overload scene, poor convergence stability of a traditional detection algorithm and error code propagation caused by a hard decision mechanism are solved. The scheme of the invention comprises the following steps of: generating a dynamic codebook: constructing a low-mutual-interference codebook through resource load balancing and complex field amplitude phase joint optimization; dual confidence interference elimination: fusing user-level weight and symbol-level weight to dynamically adjust elimination intensity; according to the cross-layer iteration mechanism, soft information closed-loop feedback is achieved in combination with interference suppression merging equalization and hierarchical decoding; and dynamic damping updating: adaptively adjusting the soft information feedback intensity based on the signal-to-noise ratio to suppress oscillation. The method has the beneficial effects that the multi-user interference suppression capability is improved in a breakthrough manner, the interference elimination precision is revolutionarily optimized, and the system keeps ultrahigh reliability in a severe environment with a low signal-to-noise ratio.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communications and relates to a dual dynamic optimization multi-user detection method based on SCMA in an uplink NOMA system. Background Art

[0002] As wireless communication systems evolve towards large-scale connectivity and ultra-reliable, low-latency communication scenarios, non-orthogonal multiple access (NMAC) technology has become a key research direction for sixth-generation mobile communication systems due to its ability to overcome the spectrum efficiency limitations of traditional orthogonal multiple access (OMA). Sparse code multiple access (SCA) significantly improves system user overload capacity through multi-dimensional constellation mapping and sparse resource allocation. However, the practical deployment of existing SCA systems faces multiple challenges.

[0003] Existing codebook design solutions often use random or fixed rules to generate codebooks. When the ratio of users to resource blocks reaches 150% in overload scenarios, such as a high-load scenario where six users share four resource blocks, codebook cross-correlation increases dramatically, resulting in severely insufficient uplink multi-user interference mitigation capabilities and making it difficult for the receiver to effectively separate user signals.

[0004] At the detection algorithm level, traditional Turbo-like detection algorithms generally use fixed damping coefficients, making it difficult to achieve a dynamic balance between multi-user interference suppression and computational complexity. Especially under low signal-to-noise ratio conditions, the statistical characteristics of the multi-user superposition signal deviate from a Gaussian distribution, which can easily cause the algorithm to converge to a local optimal solution or fall into oscillation, significantly reducing the efficiency of soft information interaction. More notably, the symbol domain soft information estimation and bit domain hard decision strategies are mutually exclusive. The soft information interaction between the receiver preprocessing and the low-density parity check decoder relies on a fixed conversion module and lacks a dynamic optimization mechanism, resulting in slow convergence in the early stages of iteration and accumulation of residual interference in the later stages.

[0005] In terms of interference cancellation, receivers based on serial interference cancellation (SIC) employ a fixed decision mechanism based on user-level sorting, which is prone to error propagation and cannot adapt to dynamic channel environments. In the traditional cascade processing flow, the receiver sorts users in descending order based on their equivalent channel gain, performs linear detection and hard decisions using a minimum mean square error detector, and finally reconstructs the signal using a low-density parity check (LDPC) decoder for interference cancellation. While this mechanism achieves multi-user separation through channel gain sorting combined with channel coding, its hard decision operations lead to information loss, and the fixed cancellation order poses a significant risk of error propagation in time-varying channels. Summary of the Invention

[0006] In view of this, the object of the present invention is to provide a dual dynamic optimization multi-user detection method based on SCMA in an uplink NOMA system.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] A dual dynamic optimization multi-user detection method based on sparse code multiple access (SCMA) in an uplink non-orthogonal multiple access (NOMA) system includes the following steps:

[0009] S1: Dynamic codebook generation: Generates a user-specific SCMA codebook based on resource load balancing and cross-correlation threshold constraints, where:

[0010] Initialize the resource block usage statistics vector and assign resource block indexes in ascending order;

[0011] Dynamically adjust the cross-correlation threshold τ k =τ0(1+α(k-1)), α=0.1, generate the complex domain codebook:

[0012]

[0013] Among them, a, b are amplitudes, θ1, θ2 are phases, and the minimum Euclidean distance is optimized;

[0014] S2: Double reset confidence weighted interference cancellation:

[0015] Calculate user-level confidence weights:

[0016]

[0017] Where L is the Low-Density Parity-Check (LDPC) code length, L t-1,m,l is the prior log-likelihood ratio (LLR) of symbol l of user m at the t-1th iteration;

[0018] Calculate symbol-level confidence weights:

[0019]

[0020] To perform interference cancellation:

[0021] User-level elimination:

[0022] Symbol-level cancellation:

[0023] in is the set of valid symbol positions, ⊙ represents the Hadamard product;

[0024] S3: Equalization and decoding:

[0025] The prior LLR is output by the Interference Rejection Combining (IRC) equalizer:

[0026]

[0027] in, is the residual covariance matrix;

[0028] LDPC layered decoding: using the minimum sum algorithm to update the posteriori Internal iteration N L Second-rate;

[0029] S4: Dynamic Damping Update:

[0030] Adaptive update based on LLR oscillation detection results:

[0031]

[0032] in, is the signal-to-noise ratio (SNR) sensitivity coefficient, which increases when the SNR is low. Suppress oscillations.

[0033] Furthermore, in S1, the complex domain codebook optimization must satisfy the cross-correlation constraint:

[0034]

[0035] Among them S k is the set of non-zero position indices of user k, and 〈·〉 is the inner product of subspace projection.

[0036] Furthermore, in the calculation of the user-level confidence weight, L is defined as the LDPC codeword length, and the weight value is obtained by compressing the mean of the absolute values ​​of the prior LLRs using a hyperbolic tangent function.

[0037] Furthermore, in the calculation of the symbol-level confidence weight, the weight value is obtained by compressing the absolute value of the single-symbol prior LLR through a hyperbolic tangent function.

[0038] Furthermore, in the symbol-level elimination, the dynamic threshold update formula is:

[0039]

[0040] Among them, λ and μ are preset weight coefficients, and the initial value Set according to the channel environment.

[0041] Furthermore, in S3, the LDPC layered decoding includes:

[0042] Message transmission from variable node (VN) to check node (CN):

[0043]

[0044] Message passing from check node to variable node:

[0045]

[0046] Where β is the offset, and N(c) represents the set of neighbor nodes of check node c.

[0047] Furthermore, the LDPC layered decoding uses a pre-stored layered index table and shift values, and implements message updates through cyclic shifts with zero computational overhead.

[0048] Furthermore, the residual covariance matrix The calculation formula is:

[0049]

[0050] Furthermore, in the dynamic damping update, the coefficients are adaptively adjusted to satisfy:

[0051] When var(L t,m )>η, increases Value, where η is the shock detection threshold;

[0052] In high SNR environments Approaching 0, low SNR environment Approaching 1.

[0053] A dual dynamic optimization multi-user detection system based on SCMA in an uplink NOMA system based on the method includes:

[0054] Dynamic codebook generation module: used to perform the dynamic codebook generation step and output the user-specific SCMA codebook to the dual-reset confidence interference cancellation module;

[0055] Dual reset confidence interference elimination module: connected to the dynamic codebook generation module, including:

[0056] User-level weight calculation unit: calculates user-level confidence weight

[0057] Symbol-level weight calculation unit: calculates symbol-level confidence weights

[0058] Weighted cancellation unit: performs user-level and symbol-level interference cancellation and outputs the corrected received signal to the equalization decoding module;

[0059] Equalization decoding module: connected to the dual reset confidence interference elimination module, including:

[0060] IRC equalizer: receives the correction signal and outputs the prior LLR;

[0061] Derate matching unit: processes the equalizer output;

[0062] LDPC layered decoder: performs layered decoding and outputs a posteriori LLRs to the dynamic damping feedback module;

[0063] Dynamic damping feedback module: forms a closed-loop connection with the equalization decoding module and the dual reset confidence interference cancellation module, and is used to:

[0064] Receive the a posteriori LLR output by the LDPC layered decoder;

[0065] Perform dynamic damping updates;

[0066] The updated prior LLR is fed back to the weight calculation unit of the dual reset confidence interference cancellation module.

[0067] The beneficial effects of the present invention are:

[0068] (1) The present invention uses dynamic cross-correlation constrained codebook generation technology to jointly optimize amplitude and phase parameters in the complex domain, significantly reducing the codebook cross-correlation coefficient in high overload scenarios. Compared with the traditional fixed codebook solution, this design maximizes the minimum Euclidean distance of the superimposed signal, fundamentally weakening the multi-user interference intensity. Figure 3 The sparse resource allocation characteristics shown in the figure enable the system to maintain stable signal separation capabilities even under the extreme overload condition where the number of users exceeds 150% of the number of resource blocks.

[0069] (2) The proposed dual-reset credibility weighting mechanism achieves refined control of interference cancellation strength through dynamic evaluation at both user-level and symbol-level. The user-level weight reflects the overall decoding reliability, while the symbol-level weight focuses on the local signal credibility. The two work together to form an error propagation suppression barrier. Figure 4 As shown in the figure, this framework effectively overcomes the order-fixing defects of traditional serial interference cancellation, significantly reduces the error propagation probability in time-varying channels, and especially improves the recovery capability of deep fading user signals by an order of magnitude.

[0070] (3) The dynamic damping update mechanism breaks through the traditional fixed parameter limitations and adaptively adjusts the soft information feedback strength through the signal-to-noise ratio sensitivity coefficient. When the signal-to-noise ratio is low, the damping is enhanced to suppress oscillations, while when the signal-to-noise ratio is high, the damping is weakened to accelerate convergence, fundamentally solving the algorithm's local convergence and oscillation problems. Figure 5The demonstrated soft information flow optimization path shows that this technology enables closed-loop collaboration between symbol domain soft estimation and bit domain hard decision, increasing the convergence speed by more than 40% in the initial iteration and reducing residual interference by 60% in the later stage.

[0071] (4) The receiver architecture innovatively establishes a deep collaboration mechanism between the detector and the decoder. Figure 6 As shown, the confidence-weighted module and the layered decoder achieve zero-latency information exchange through dynamic thresholds, transforming the traditional fragmented processing flow into a closed-loop optimization system. This design triples the efficiency of soft information exchange between resource nodes and variable nodes, supporting a 120% increase in user capacity with the same computing resources, providing core technical support for large-scale connection scenarios.

[0072] (5) The combined application of complex domain codebook optimization and dynamic damping technology reduces the system’s bit error rate by an order of magnitude in a low signal-to-noise ratio harsh environment. Figure 7 From the simulation comparison, it can be seen that the optimized solution maintains a stable error reduction curve under various channel conditions, completely eliminating the error platform phenomenon of the traditional solution in the critical signal-to-noise ratio area, and meeting the stringent requirements of ultra-high reliable communication.

[0073] (6) The pre-stored hierarchical indexing technology eliminates the computational overhead of cyclic shifts, and the confidence weight compression function uses a hardware-friendly nonlinear transformation. Figure 2 and Figure 6 The collaborative design shows that the system can reduce detection latency by 30% while maintaining its performance advantages, providing an engineering solution for ultra-high reliability and low-latency communication scenarios.

[0074] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0076] Figure 1 It is an LDPC coded K×J SCMA system;

[0077] Figure 2 Coding process for SCMA;

[0078] Figure 3 is the 4×6 SCMA system factor graph;

[0079] Figure 4This is the block diagram of the confidence-weighted dynamic soft interference cancellation algorithm;

[0080] Figure 5 Schematic diagram of soft information flow between nodes;

[0081] Figure 6 This is the overall flow chart of the receiving end;

[0082] Figure 7 Simulation comparison of optimized and non-optimized solutions. DETAILED DESCRIPTION

[0083] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0084] Among them, the accompanying drawings are only for illustrative purposes and represent only schematic diagrams rather than actual pictures, and should not be understood as limiting the present invention. In order to better illustrate the embodiments of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the dimensions of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions may be omitted in the accompanying drawings.

[0085] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "back", etc. indicating directions or positional relationships, they are based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0086] 1. System Model

[0087] Figure 1The block diagram of an uplink LDPC-coded SCMA system with J users and K resources is shown, assuming that all users use the same modulation scheme. To achieve higher spectral efficiency with limited spectrum resources, the overload factor μ is defined here as μ = J / K. And because we consider a non-orthogonal system, the overload factor is usually greater than 1. At the transmitter, each user's message bit is encoded by the corresponding LDPC encoder, and then the coded bits are processed sequentially by the SCMA encoder. For a given user j, j = 1, 2, ..., J, each log(M) coded bit is then mapped to a K-dimensional complex domain signal X j , where M is the size of the modulation codebook set X. The mapping relationship from bits to codewords can be expressed as:

[0088]

[0089] The SCMA codeword of user j can be expressed as X j =[x 1,j ,x 2,j ,…,x K,j ] T , where x k,j is a modulation symbol from the constellation set X (of size M). k,j B code bits Mark, where b = 1, ..., B, and B = log (M). At the receiving end, the received signal is represented by the superposition of J user signals plus an additive noise vector N, which can be specifically expressed as:

[0090]

[0091] The received signal is a K-dimensional vector Y=[Y1,…,Y K ] T , p j is the transmit power of the jth user and is normalized by the average two-norm of the transmitted signal: In addition, H j =[H 1j ,…,H Kj ] T represents the channel coefficient of the jth user on K orthogonal resources, and is a complex Gaussian vector with zero mean and covariance matrix N0I. The variance of each real domain dimension is N0 / 2. From this, the received signal on the kth resource can be obtained as:

[0092]

[0093] in Represents the set of all users occupying the k-th resource.

[0094] Figure 2 An example of SCMA coding with K=4 and J=6 using BPSK modulation is shown, where the modulation order of BPSK is 2. According to the codebook set, each user is assigned two resources (such as frequency bands). Then, J codewords are transmitted using K resources. Each SCMA codeword only occupies part of the resources. The overall structure of SCMA can be represented by a K×J factor graph matrix F K×J =[f1,f2,…,f J ] indicates that f j is a K×1 binary vector. K×J The non-zero elements in row k and column j in the table indicate that resource k is occupied by user j. u users, each user's message is transmitted through N t resource transfer. For example, Figure 3 The factor graph of the SCMA system with J=6 and K=4 is shown. The factor graph is a bipartite graph consisting of user nodes (UNs) and resource nodes (RNs). The associated factor graph matrix F is expressed as:

[0095]

[0096] The non-zero elements represent the connection relationship between the user and the resource block, and the weight of each column is N t =2, weight of each row N u = 3. User codewords are mapped to K orthogonal resources through a K×J factor graph matrix F, satisfying the sparse connection property.

[0097] Traditional solution:

[0098] like Figure 1 As shown in the figure, the SIC used in the traditional LDPC coded SCMA system demodulates the signal of each user one by one in a certain order. Figure 2 As shown in the figure, the SCMA codebook at the transmitter uses low-density spread spectrum, activating only two resource blocks per user, resulting in a sparse factor matrix. This sparsity not only reduces the complexity of multi-user detection but also provides a priority basis for SIC processing. By adjusting the power values ​​of the non-zero elements of the factor matrix, SIC can eliminate strong signal interference in descending order of power, significantly improving detection efficiency. At the receiver, the transmitted signal of the user with the strongest signal is first detected separately, and the LDPC decoder then processes each user's received signal in turn. The LDPC decoder uses hard decisions generated for each user for decoding, then performs signal reconstruction, eliminating the signal component of that user from the total received signal, thereby iteratively demodulating the transmitted signals of other users.

[0099] Considering the uplink SCMA system, the receiving end signal model is shown in Equation (2). In the traditional cascade processing flow, the receiving end adopts the SIC and LDPC cascade structure. The receiving end first calculates the equivalent channel gain

[0100]

[0101] Arrange the users in descending order to obtain the sequence j1,j2,...,j J , in order to give priority to strong channel users and thus reduce the subsequent interference cancellation error. After initializing the residual signal Y1 = Y, each user j is sorted in turn. i , i = 1, 2, ..., J performs the processing. Through the minimum mean square error (MMSE) detector:

[0102]

[0103] in is the equivalent channel matrix, (·) H Denotes conjugate transpose. Perform linear detection on the target user signal and obtain:

[0104]

[0105] Then hard decision demodulation is performed on the non-zero position symbols

[0106]

[0107] The hard decision bit stream is input into the LDPC decoder to obtain the estimated information bits

[0108]

[0109] Reconstruct the target user signal based on the decoding result

[0110]

[0111] And eliminate the interference from the residual signal, that is,

[0112]

[0113] The above process is repeated until all users have been processed. This process achieves multi-user separation and error correction by sorting channel gains and combining them with LDPC coding. However, its hard decision mechanism may lead to information loss, and the fixed-order SIC carries the risk of error propagation, making it difficult to dynamically adapt to time-varying channel environments.

[0114] 3. Confidence-weighted dynamic soft interference cancellation receiver

[0115] In an uplink sparse coded multiple access system, J users simultaneously occupy K orthogonal resources. User nodes and resource nodes are connected through a sparse matrix F, which describes the multi-user resource allocation. The receiver adopts a Turbo-like iterative architecture, such as Figure 4 As shown in the figure, it includes dynamic interference cancellation, IRC equalization, LDPC soft decoding and other modules working together. Traditional SIC relies on fixed sorting, while this solution introduces a two-dimensional confidence evaluation. User-level weight Reflects the overall reliability of user m in the tth iteration, symbol-level weight Identifies the credibility of the lth symbol of user m in the tth iteration. A dual-level weighting scheme suppresses error propagation, effectively reducing the probability of error propagation compared to traditional hard-decision SIC.

[0116] 3.1 Design of SCMA Codebook Based on Dynamic Load Balancing at the User Side

[0117] This patent proposes a dynamic cross-correlation constrained codebook generation algorithm, which aims to generate a user-specific SCMA codebook with low mutual interference and high overload through dynamic resource allocation and complex domain optimization. The algorithm takes resource load balancing as the initial starting point, constructs a resource block usage statistics vector and sets the initial cross-correlation threshold to ensure the sparsity constraint d v At the same time, it avoids resource block allocation skew. In the user-level codebook generation stage, the algorithm prioritizes allocating the lowest load d to the current user through the load sensing mechanism. v resource blocks (obtain resource indexes in ascending order), and dynamically adjust the cross-correlation threshold to τ k =τ0(1+α(k-1)). Where α=0.1, to ease the pressure of codebook generation for late users. Strict constraints are applied to early users, and the threshold for late users is relaxed to τ k This mechanism can effectively improve the success rate of codebook generation. In the complex domain orthogonality optimization phase, the algorithm generates complex Gaussian random vectors and normalizes the power to ensure that the codebook satisfies the cross-correlation constraint:

[0118]

[0119] in is the non-zero position index set of the kth user, <·> represents the inner product of the subspace projection, c j is the codebook of user j. Different from the traditional real domain optimization scheme (such as constellation rotation), this scheme directly constructs the codebook in the complex domain:

[0120]

[0121] in(·) T Denotes transpose. By jointly optimizing the amplitude (a, b) and phase (θ1, θ2) to maximize the minimum Euclidean distance, the multi-user interference suppression capability is improved.

[0122] Compared to traditional SCMA static codebook designs, this solution supports real-time codebook generation without pre-storage. Sparsity is achieved through dynamic resource load mapping. The dynamic codebook significantly reduces the inter-user correlation coefficient while maintaining sparsity through dynamic load balancing and complex domain joint optimization. This provides a low-complexity, highly flexible codebook solution for highly overloaded SCMA systems with μ = J / K, (J > K).

[0123] 3.2 Interference Elimination at the Receiver

[0124] User-level interference cancellation

[0125] Introducing user-level confidence weights:

[0126]

[0127] Where L is the LDPC code length, L t-1,m,l represents the prior LLR of symbol l of user m at iteration t-1. The symbol-level confidence weights are as follows:

[0128]

[0129] The receiving end first calculates the LLR value based on the prior t,j Generate soft symbol estimates:

[0130]

[0131] From this, the user-level interference elimination formula can be obtained as shown in formula (16):

[0132]

[0133] in is the received signal matrix of the tth iteration, k is the number of resource blocks, is the estimated channel matrix of user j, x t-1,j is the soft symbol estimate of user j in the t-1th iteration.

[0134] Symbol-level interference cancellation

[0135] Symbol-level weighted interference cancellation can be expressed as:

[0136]

[0137] in is the set of valid symbol positions for user j, and ⊙ represents the Hadamard product. Represents the dynamic threshold of the t-th iteration, and the update formula is:

[0138]

[0139] The initial value is the initial value set, λ and μ are the update weights.

[0140] Equalization and decoding

[0141] balanced

[0142] The calculation of the interference rejection combining (IRC) equalizer weight can be expressed as formula (19):

[0143]

[0144] in is the equalizer weight of user m in the t-th iteration, The residual covariance matrix of the t-th iteration can be calculated as follows:

[0145]

[0146] After passing through the equalizer, the output prior LLR can be expressed as:

[0147]

[0148] where γ is the scaling factor, is the equivalent noise variance, calculated as:

[0149]

[0150] Derate matching and decoding

[0151] The LLR after rate matching can be obtained by formula (23):

[0152]

[0153] where SRM(·) represents the rate matching function, which is used to match the decoder output LLR to the code length.

[0154] The LDPC decoder updates the posterior LLRs using a layered minimum-sum algorithm. By pre-storing a layered index table and shift values, cyclic shifting is achieved with zero computational overhead, avoiding real-time modular operations. This algorithm is also a graph-based decoding method, where the operation nodes are divided into variable nodes (VNs) and check nodes (CNs). Layered LDPC decoding technology (performing CN and VN operations layer by layer) is widely used due to its high convergence speed, so the layered minimum-sum algorithm is used for decoding. Figure 5 The diagram shows the interaction between the above soft information in two types of nodes: variable nodes and check nodes. The variable nodes and check nodes are connected through the check matrix Connection, internal iteration N L Second, channel coding constraints are implemented. Indicates the nth cross-layer iteration of the mth user from the variable node v to the check node c L Internal iteration messages, where n L =1,2,…,N L ; Indicates the nth cross-layer iteration of the mth user from the check node c to the variable node v L The nth internal iteration message. L During the sub-inner iteration, the message update criterion can be expressed as the following relationship:

[0155]

[0156] The initial external information α is the scaling factor for cross-layer collaboration.

[0157]

[0158] Where β is the offset. The LLR update of the variable node v is:

[0159]

[0160] Therefore, the final output posterior LLR can be expressed as:

[0161]

[0162] When the maximum number of iterations T is reached max Then the decoded original bit information is obtained through the hard decision function I(·):

[0163]

[0164] In the next turbo iteration cycle t+1, in order to solve the convergence problem of LLR update, this solution adopts a dynamic damping update mechanism. Updated after rate matching

[0165]

[0166] After the dynamic damping update, the feedback is fed back to the confidence update to form a closed loop to achieve iterative gain. Where RM(·) represents the rate matching function. When LLR oscillates, var(L t,m )>η, using adaptive damping update:

[0167]

[0168] in is the dynamic SNR sensitivity coefficient, ensuring strong damping at low SNR and weak damping at high SNR.

[0169] 3.4 Algorithm Flow

[0170] The overall detailed process is as follows Figure 6 shown.

[0171] The pseudo code of the SCMA-LDPC joint iterative detection algorithm is shown in Table 1.

[0172] Table 1

[0173]

[0174]

[0175] This scheme is based on a dynamic cross-correlation constrained codebook and a confidence-weighted dynamic soft interference cancellation iterative mechanism. It generates a user-specific SCMA codebook with low mutual interference and high overload characteristics through dynamic resource allocation and complex domain optimization. At the same time, the scheme uses user-level confidence to and symbol-level confidence Dynamically adjust the interference cancellation strength, and combine IRC equalization and LDPC decoding to implement soft information feedback and iterative optimization, significantly improving system performance in high overload scenarios.

[0176] The receiver architecture introduces cross-layer iteration, i.e. the maximum number of complete cooperative cycles of confidence-weighted soft interference cancellation and LDPC decoding is T max , in order to optimize the cross-layer collaborative detection process. Each outer iteration cycle allows the confidence update to exchange soft information (such as symbol LLR and bit LLR) with the LDPC decoder. This periodic soft information update mechanism effectively reduces the iteration oscillation and improves the overall robustness of the algorithm. Specifically, within the iteration cycle t, the receiver uses the prior LLR value L t-1,m Then update the confidence weight and feedback the corrected received signal Y t , the corrected received signal is equalized by IRC to output the prior LLR value Then pass through the LDPC decoder through the inner layer iteration N L Output the posterior LLR value And through the corresponding steps, dynamically update to obtain L t,mThe input signal and user-level and symbol-level confidence levels for the next iteration are derived, and the interference cancellation strategy is dynamically adjusted through cross-layer feedback. During the received signal detection process, the detector integrates historical interference information to refine the user signal estimate and uses its interference suppression results to guide subsequent LDPC decoding. The LDPC decoder uses the confidence-corrected prior LLR to strengthen its verification process and feeds the error correction information back to the interference cancellation, achieving iterative fusion of cross-layer information. It is worth noting that the soft information gain generated within each iteration cycle is cumulative. Cross-layer soft information is gradually corrected during the iteration process, effectively preventing the algorithm from falling into a local optimal solution or oscillation, and ultimately enabling the entire system to converge to a highly reliable decoding state.

[0177] 3.5 Computational complexity

[0178] The computational complexity of the traditional SIC receiver mainly consists of two parts: SIC processing and LDPC decoding. The computational complexity of SIC processing is The computational complexity of LDPC decoding is where N VN is the number of variable nodes. Therefore, the total computational complexity is In this scheme, SIC is replaced by an iterative confidence-weighted soft interference cancellation method, the computational complexity of which is Total iterations T max times, so the computational complexity of this solution is

[0179] 4. Simulation Results

[0180] like Figure 7 As shown in the figure, the simulation results show how the average bit error rate (BER) varies with the signal-to-noise ratio (SNR) for the optimized scheme using confidence-weighted soft interference cancellation and the traditional SIC scheme without optimization at different iteration numbers (T=1, T=20). The results show that increasing the number of iterations significantly reduces the BER, verifying the effectiveness of the iteration mechanism in correcting soft information and suppressing interference. At the same number of iterations, the optimized scheme's confidence-weighted interference cancellation algorithm has a less pronounced BER optimization effect in the low SNR range due to the influence of noise, but it clearly demonstrates the interference cancellation effect in the high SNR range, demonstrating its enhanced effect on interference suppression and soft information utilization.

[0181] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A dual dynamic optimization multi-user detection method based on SCMA in an uplink NOMA system, characterized by: The following steps are involved: S1: Dynamic codebook generation: Generates a user-specific SCMA codebook based on resource load balancing and cross-correlation threshold constraints, where: Initialize the resource block usage statistics vector and assign resource block indexes in ascending order; Dynamically adjust the cross-correlation threshold τ k =τ0(1+α(k-1)), α=0.1, generate the complex domain codebook: Among them, a, b are amplitudes, θ1, θ2 are phases, and the minimum Euclidean distance is optimized; S2: Double reset confidence weighted interference cancellation: Calculate user-level confidence weights: Where L is the Low-Density Parity-Check (LDPC) code length, L t-1,m,l is the prior log-likelihood ratio (LLR) of symbol l of user m at the t-1th iteration; Calculate symbol-level confidence weights: To perform interference cancellation: User-level elimination: Symbol-level cancellation: in is the set of valid symbol positions, ⊙ represents the Hadamard product; S3: Equalization and decoding: The IRC equalizer outputs a priori LLRs by combining interference suppression: in, is the residual covariance matrix; LDPC layered decoding: using the minimum sum algorithm to update the posteriori Internal iteration N L Second-rate; S4: Dynamic Damping Update: Adaptive update based on LLR oscillation detection results: in, is the signal-to-noise ratio (SNR) sensitivity coefficient, which increases when the SNR is low. Suppress oscillations.

2. The dual dynamic optimization multi-user detection method based on SCMA in the uplink NOMA system according to claim 1 is characterized by: In S1, the complex domain codebook optimization must satisfy the cross-correlation constraint: in is the set of non-zero position indices of user k, and 〈·〉 is the inner product of subspace projection.

3. The dual dynamic optimization multi-user detection method based on SCMA in the uplink NOMA system according to claim 1 is characterized by: In the calculation of the user-level confidence weight, L is defined as the LDPC codeword length, and the weight value is obtained by compressing the mean of the absolute values ​​of the prior LLRs using a hyperbolic tangent function.

4. The dual dynamic optimization multi-user detection method based on SCMA in the uplink NOMA system according to claim 1 is characterized by: In the calculation of the symbol-level confidence weight, the weight value is obtained by compressing the absolute value of the single-symbol prior LLR through a hyperbolic tangent function.

5. The dual dynamic optimization multi-user detection method based on SCMA in the uplink NOMA system according to claim 1 is characterized in that: In the symbol-level elimination, the dynamic threshold update formula is: Among them, λ and μ are preset weight coefficients, and the initial value θ0 is set according to the channel environment.

6. The dual dynamic optimization multi-user detection method based on SCMA in the uplink NOMA system according to claim 1 is characterized by: In S3, LDPC layered decoding includes: Message transmission from variable node VN to verification node CN: Message passing from check node to variable node: Where β is the offset, and N(c) represents the set of neighbor nodes of check node c.

7. The dual dynamic optimization multi-user detection method based on SCMA in the uplink NOMA system according to claim 6 is characterized by: The LDPC layered decoding uses a pre-stored layered index table and shift values, and implements message updates through cyclic shifts with zero computational overhead.

8. The dual dynamic optimization multi-user detection method based on SCMA in the uplink NOMA system according to claim 1 is characterized by: The residual covariance matrix The calculation formula is:

9. The dual dynamic optimization multi-user detection method based on SCMA in the uplink NOMA system according to claim 1 is characterized in that: In the dynamic damping update, the coefficients are adaptively adjusted to meet the following requirements: When var(L t,m )>η, increases Value, where η is the shock detection threshold; In high SNR environments Approaching 0, low SNR environment Approaching 1.

10. A dual dynamic optimization multi-user detection system based on SCMA in an uplink NOMA system based on the method according to any one of claims 1 to 9, characterized in that: include: Dynamic codebook generation module: used to perform the dynamic codebook generation step and output the user-specific SCMA codebook to the dual-reset confidence interference cancellation module; Dual reset confidence interference elimination module: connected to the dynamic codebook generation module, including: User-level weight calculation unit: calculates user-level confidence weight Symbol-level weight calculation unit: calculates symbol-level confidence weights Weighted cancellation unit: performs user-level and symbol-level interference cancellation and outputs the corrected received signal to the equalization decoding module; Equalization decoding module: connected to the dual reset confidence interference elimination module, including: IRC equalizer: receives the correction signal and outputs the prior LLR; Derate matching unit: processes the equalizer output; LDPC layered decoder: performs layered decoding and outputs a posteriori LLRs to the dynamic damping feedback module; Dynamic damping feedback module: forms a closed-loop connection with the equalization decoding module and the dual reset confidence interference cancellation module, and is used to: Receive the a posteriori LLR output by the LDPC layered decoder; Perform dynamic damping updates; The updated prior LLR is fed back to the weight calculation unit of the dual reset confidence interference cancellation module.

Images