Low complexity encoding and decoding method, device and equipment for SCMA-d2d network

Abstract

The application relates to a mixed network coding and decoding technology, and discloses a low-complexity coding and decoding method, device and equipment for an SCMA-D2D network, which comprises the following steps: generating a connection factor matrix according to user information of SCMA cellular users and D2D users and spectrum resource blocks, and generating a connection factor graph according to the connection factor matrix; constructing a code word generator according to the connection factor graph, and calculating code words corresponding to the user information by using the code word generator; generating mixed network signal data according to the code words corresponding to the user information and the spectrum resource blocks; inputting the mixed network signal data after being transmitted through a Gaussian channel into a relay network to obtain relay signal data, and broadcasting the relay signal data to receiving ends corresponding to the SCMA cellular users and the D2D users; and after the relay signal data is received at the receiving ends, decoding the relay signal data by using a multi-user classification decoder to obtain decoding information. The application can improve the precision of multi-user decoding and reduce the complexity of coding and decoding.

Description

Technical Field

[0001] This invention relates to the field of hybrid network encoding and decoding technology, and in particular to a low-complexity encoding and decoding method, apparatus and device for SCMA-D2D networks. Background Technology

[0002] With the continuous development of the Internet of Things (IoT), the massive number of devices and diverse service types place higher demands on network access technologies. Network architectures capable of supporting different types of devices and multiple access technologies with higher spectrum utilization have become research focuses. Sparse Code Multiple Access (SCMA) maps user bit information into specific sparse multidimensional codewords at the transmitting end through low-density spread spectrum and multidimensional modulation coding. Different user codewords are non-orthogonally superimposed on spectrum resources, enabling multi-user access and data transmission. Device-to-Device (D2D) communication is a short-range communication technology that supports direct data transmission between users at close range. Adding D2D user communication to cellular systems allows both users to share spectrum resources, further increasing spectrum utilization and meeting the large-scale, diverse application needs of communication networks. However, in hybrid network systems where SCMA and D2D coexist, resource sharing and the increase in devices can lead to severe inter-user interference, increased encoding / decoding complexity, and decreased BER (Bit Error Rate) accuracy. Meanwhile, smart mobile terminals in the Internet of Things (IoT) often possess characteristics such as miniaturization and lightweight design. These characteristics require that the encoding and decoding algorithms of smart terminals maintain low error rate performance while also having low complexity. Therefore, for application scenarios involving multi-user access and data transmission, improving the accuracy of multi-user decoding while reducing the complexity of encoding and decoding has become an urgent problem to be solved. Summary of the Invention

[0003] This invention provides a low-complexity encoding and decoding method, apparatus, electronic device, and computer-readable storage medium for SCMA-D2D networks. Its main purpose is to solve the problem of not being able to effectively improve the accuracy of multi-user decoding while reducing the complexity of encoding and decoding.

[0004] To achieve the above objectives, this invention provides a low-complexity encoding and decoding method for SCMA-D2D networks, comprising:

[0005] A connection factor matrix is ​​generated based on the user information of SCMA cellular users and D2D users and the preset spectrum resource blocks, and a connection factor graph is generated based on the connection factor matrix.

[0006] A codeword generator is constructed based on the connection factor graph, and the codeword corresponding to the user information is calculated using the codeword generator; wherein, the codeword generator is composed of multiple DNN coding units, and the number of output nodes in the output layer of the DNN coding unit is set based on the mapping relationship of the connection factor graph;

[0007] Generate hybrid network signal data based on the codewords corresponding to the user information and the spectrum resource blocks;

[0008] The hybrid network signal data, after being transmitted through a Gaussian channel, is input into a pre-constructed relay network to obtain relay signal data, and the relay signal data is broadcast to the receiving end corresponding to the SCMA cellular user and the D2D user.

[0009] After receiving the relay signal data, the receiving end uses a pre-built multi-user classification decoder to decode the relay signal data to obtain the decoding information corresponding to the SCMA cellular user and the D2D user; wherein, the multi-user classification decoder includes the DNN decoding unit corresponding to the SCMA cellular user and the D2D user in the receiving end.

[0010] Optionally, generating hybrid network signal data based on the codeword corresponding to the user information and the spectrum resource block includes:

[0011] Determine the target user node connected to the spectrum resource block, and obtain the target codeword of the user information corresponding to the target user node;

[0012] The target codewords are superimposed to obtain the resource block codewords corresponding to each spectrum resource block;

[0013] The resource block codewords are calculated using the following output formula to obtain the hybrid network signal data;

[0014] The output formula is expressed as follows:

[0015]

[0016] in, This is represented as the hybrid network signal data. Represented as the first The resource block codeword corresponding to each spectrum resource block This represents the total number of spectrum resource blocks.

[0017] Optionally, the connection factor matrix is ​​represented as:

[0018]

[0019] in, Represented as a link factor matrix, This represents the number of spectrum resource blocks. This represents the number of SCMA cellular users. This represents the number of D2D users in the user information.

[0020] Optionally, before inputting the hybrid network signal data transmitted via the Gaussian channel into the pre-built relay network, the method further includes:

[0021] Construct the relay sharing layer and initialize the DNN relay unit;

[0022] The number of nodes in the input layer of the initial DNN relay unit is configured according to the dimension of the hybrid network signal data after transmission through the Gaussian channel, thus obtaining the DNN relay unit;

[0023] A relay network is generated based on the relay sharing layer and the DNN relay unit.

[0024] Optionally, constructing the codeword generator based on the connection factor graph includes:

[0025] Configure the corresponding initialization DNN encoding unit according to the user nodes in the connection factor graph;

[0026] A one-hot vector is generated by obtaining the user information corresponding to the user of the user node and performing one-hot encoding.

[0027] The number of nodes in the input layer of the initial DNN encoding unit is set according to the dimension of the single-hot vector, and the number of nodes in the output layer of the initial DNN encoding unit is set according to the number of connection lines of user nodes in the connection factor graph, thus obtaining the DNN encoding unit.

[0028] The codeword generator is obtained by summing up the DNN encoding units corresponding to each user node.

[0029] Optionally, before decoding the relay signal data using a pre-built multi-user classification decoder, the method further includes:

[0030] Configure a corresponding initialization DNN decoding unit for each user at the receiving end;

[0031] The number of nodes in the input layer of the initial DNN decoding unit is set according to the dimension of the relay signal data, and the number of nodes in the output layer of the initial DNN decoding unit is set according to the dimension of the single-hot vector, thus obtaining the DNN decoding unit;

[0032] A multi-user classification decoder is generated based on the initialized decoding shared layer and the DNN decoding unit.

[0033] Optionally, the step of decoding the relay signal data using a pre-built multi-user classification decoder to obtain user information corresponding to the SCMA cellular user and the D2D user includes:

[0034] The relay signal data is decoded using the multi-user classification decoder to obtain SCMA cellular user decoded data and D2D user decoded data;

[0035] Probability calculations are performed on the SCMA cellular user decoding data and the D2D user decoding data to obtain the probability information corresponding to the SCMA cellular user decoding data and the D2D user decoding data;

[0036] The probability information is converted into binary to obtain the decoding information corresponding to the SCMA cellular user and the D2D user.

[0037] Optionally, after obtaining the decoding information corresponding to the SCMA cellular user and the D2D user, the method further includes:

[0038] Calculate the cross-entropy loss based on the user information corresponding to the SCMA cellular user and the D2D user, as well as the corresponding decoding information.

[0039] The codeword generator, relay network, and multi-user classification decoder are jointly optimized based on the cross-entropy loss to obtain an SCMA-D2D hybrid network self-coder model composed of the optimized codeword generator, relay network, and multi-user classification decoder.

[0040] Optionally, the step of calculating the cross-entropy loss based on the user information corresponding to the SCMA cellular user and the D2D user, and the corresponding decoding information, includes:

[0041] The first cross-entropy loss is obtained by calculating the user information and decoding information corresponding to the SCMA cellular user using a preset first cross-entropy loss function.

[0042] The first cross-entropy loss function is expressed as:

[0043]

[0044] in, Represented as the first The cross-entropy of each SCMA cellular user. Represented as the first Decoding information corresponding to each SCMA cellular user. Represented as the first User information corresponding to each SCMA cellular user. This is expressed as the number of SCMA cellular users. This represents the number of spectrum resource blocks. yes The item, yes The item, This is represented by the first cross-entropy loss corresponding to the SCMA cellular user. Represented as , The set, Represented as , A set;

[0045] The second cross-entropy loss is obtained by calculating the user information and decoding information corresponding to the D2D user using a preset second cross-entropy loss function.

[0046] The second cross-entropy loss function is expressed as:

[0047]

[0048] in, Represented as the first Cross-entropy corresponding to each D2D user Represented as the first Decoding information corresponding to each D2D user Represented as the first User information corresponding to each D2D user This is represented by the number of D2D users. This represents the number of spectrum resource blocks. yes The item, yes The item, This is represented as the second cross-entropy loss corresponding to the D2D user. Represented as , The set, Represented as , A set of.

[0049] To address the above problems, the present invention also provides a low-complexity encoding and decoding apparatus for SCMA-D2D networks, the apparatus comprising:

[0050] The connection factor graph generation module is used to generate a connection factor matrix based on the user information of SCMA cellular users and D2D users and a preset spectrum resource block, and to generate a connection factor graph based on the connection factor matrix.

[0051] An encoding module is used to construct a codeword generator based on the connection factor graph, and to calculate the codeword corresponding to the user information using the codeword generator; wherein, the codeword generator is composed of multiple DNN encoding units, and the number of output nodes in the output layer of the DNN encoding unit is set based on the mapping relationship of the connection factor graph;

[0052] The spectrum resource mapping module is used to generate hybrid network signal data based on the codewords corresponding to the user information and the spectrum resource blocks.

[0053] The relay network processing module is used to input the hybrid network signal data after transmission through the Gaussian channel into the pre-constructed relay network to obtain relay signal data, and broadcast the relay signal data to the receiving end corresponding to the SCMA cellular user and the D2D user.

[0054] The decoding module is used to decode the relay signal data using a pre-built multi-user classification decoder after the receiving end receives the relay signal data, to obtain the decoding information corresponding to the SCMA cellular user and the D2D user; wherein the multi-user classification decoder includes the DNN decoding unit corresponding to the SCMA cellular user and the D2D user in the receiving end.

[0055] To address the above problems, the present invention also provides an electronic device, the electronic device comprising:

[0056] At least one processor; and,

[0057] A memory communicatively connected to the at least one processor; wherein,

[0058] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the SCMA-D2D network low-complexity encoding and decoding method described above.

[0059] To address the aforementioned problems, the present invention also provides a computer-readable storage medium storing at least one computer program, which is executed by a processor in an electronic device to implement the aforementioned low-complexity encoding and decoding method for SCMA-D2D networks.

[0060] This invention combines SCMA and D2D encoding / decoding, reducing the number of encoder units based on different user types. The encoder employs a fully connected neural network, resulting in a simpler structure compared to existing CNN-SCMA-D2D hybrid network encoders / decoders based on convolutional neural networks. By constructing a codeword generator using connection factor graphs, the number of output nodes in the DNN units at the encoding end is increased while the number of neural network layers in the DNN units is correspondingly reduced, thus lowering the overall computational complexity of the encoder. Based on the multi-task branch decoding strategy of the DNN units in the relay network, the hybrid network signal is preprocessed by the relay before being broadcast to the multi-user classification decoder for classification decoding. This allows useful information from D2D users and SCMA cellular users to be extracted and shared in the relay, improving the decoding accuracy of user decoders based on hybrid network signals and effectively reducing interference between users. Compared to CNN-SCMA-D2D hybrid network encoders / decoders, this invention has a lower encoder complexity. Therefore, the SCMA-D2D network low-complexity encoding / decoding method, apparatus, and device proposed in this invention can solve the problem of not being able to effectively improve the accuracy of multi-user decoding while reducing encoding / decoding complexity. Attached Figure Description

[0061] Figure 1 This is a flowchart illustrating a low-complexity encoding and decoding method for SCMA-D2D networks provided in an embodiment of the present invention.

[0062] Figure 2 This is a schematic diagram of the structure of a connection factor graph provided in an embodiment of the present invention;

[0063] Figure 3 This is a schematic diagram of the structure of a DNN unit provided in an embodiment of the present invention;

[0064] Figure 4 This is a schematic diagram of the structure of an SCMA-D2D hybrid network self-encoder model provided in an embodiment of the present invention;

[0065] Figure 5 This is a comparison chart of the BER performance of a multi-user classification decoder under different signal-to-noise ratios provided in an embodiment of the present invention (SCMA).

[0066] Figure 6 This is a comparison chart of the D2D user BER performance of a multi-user classification decoder under different signal-to-noise ratios according to an embodiment of the present invention;

[0067] Figure 7 This is a comparison chart of the BER performance of the SCMA-D2D hybrid network self-encoder model under different signal-to-noise ratios provided in an embodiment of the present invention;

[0068] Figure 8This is a comparison chart of the decoding time of a traditional LSD decoding algorithm and a multi-user classification decoder LDS-SCMA-D2D provided in an embodiment of the present invention;

[0069] Figure 9 A comparison chart of the encoding complexity of a CNN-SCMA-D2D encoder and an LDS-SCMA-D2D codeword generator provided in an embodiment of the present invention;

[0070] Figure 10 This is a functional block diagram of a low-complexity encoder-decoder device for SCMA-D2D networks provided in an embodiment of the present invention;

[0071] Figure 11 This is a schematic diagram of the structure of an electronic device that implements the low-complexity encoding and decoding method of the SCMA-D2D network according to an embodiment of the present invention.

[0072] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0073] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0074] This application provides a low-complexity encoding and decoding method for SCMA-D2D networks. The executing entity of the SCMA-D2D network low-complexity encoding and decoding method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application embodiment: a server, a terminal, etc. In other words, the SCMA-D2D network low-complexity encoding and decoding method can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0075] This invention provides an SCMA-D2D hybrid network self-encoder / decoder model for small-sized, lightweight smart mobile terminal devices based on Internet of Things (IoT) applications, thereby achieving low-complexity network encoding and decoding. Specifically, the encoder / decoder model design reduces the number of DNN (Dense Neural Network) units, further simplifying the encoder structure, reducing encoding complexity, and decreasing device power consumption. At the receiver end, a multi-task branch decoder based on DNN units is designed and jointly optimized with the encoder to achieve low-complexity multi-user joint encoding and high-precision multi-user classification decoding in the SCMA-D2D hybrid network system.

[0076] Reference Figure 1 The diagram shown is a flowchart illustrating a low-complexity encoding / decoding method for SCMA-D2D networks according to an embodiment of the present invention. In this embodiment, the low-complexity encoding / decoding method for SCMA-D2D networks includes:

[0077] S1. Generate a connection factor matrix based on the user information of SCMA cellular users and D2D users and the preset spectrum resource blocks, and generate a connection factor graph based on the connection factor matrix.

[0078] In this embodiment of the invention, during multi-user access and data transmission, the SCMA cellular users are loaded onto the spectrum resources using Sparse Code Multiple Access (SCMA) mapping, while device-to-device (D2D) users are loaded onto the spectrum resource blocks they occupy using direct modulation, thereby achieving spectrum resource overload in the system and improving spectrum resource utilization.

[0079] In this embodiment of the invention, the connection factor matrix is ​​represented as follows:

[0080]

[0081] in, Represented as a link factor matrix, This represents the number of spectrum resource blocks. This represents the number of SCMA cellular users in the user information. This represents the number of D2D users in the user information.

[0082] In this embodiment of the invention, the number of SCMA cellular users is set to... The number of D2D users is The number of spectrum resource blocks is , , You can set the occupancy per SCMA cellular user. One spectrum resource block, and satisfying Each D2D user occupies one spectrum resource block.

[0083] In this embodiment of the invention, a matrix is ​​randomly generated. , matrix elements are Two possible values, matrix The columns correspond to users (SCMA cellular users, D2D user transmitters), and the matrix... The rows correspond to spectrum resource blocks, matrix middle The first and last columns represent the spectrum resource blocks occupied by SCMA cellular users in the matrix. middle The first and subsequent columns represent the D2D user's allocation of spectrum resource blocks; the matrix middle The column weights of the column and the preceding column are: ,matrix middle The column weights of column A and subsequent columns are: ,matrix The row weight is taken as: Two scenarios, Take the integer part; specifically, if the first element of the connecting factor matrix is... row and number Column elements This indicates that the user corresponding to this column occupies the first position. The nth resource block, if the matrix's nth... row and number Column elements This indicates that the corresponding user has not occupied any resource blocks.

[0084] In this embodiment of the invention, the connection factor graph is a connection factor graph of an SCMA-D2D hybrid network generated from a connection factor matrix. Each user in the connection factor matrix corresponds to a user node in the connection factor graph, and each spectrum resource block in the connection factor matrix corresponds to a spectrum resource block node in the connection factor graph. If the connection factor matrix's... row and number Column elements This indicates that the first The user and the first The spectrum resource blocks have connectivity relationships, which are represented in the connectivity factor graph as the th _ ... The user and the first There are connecting lines (i.e., edges in the connection factor graph) between the spectrum resource blocks, otherwise it indicates that the... The user and the first There are no connections between the spectrum resource blocks, which is represented in the connection factor graph as the th . The user and the first There are no connecting lines between the spectrum resource blocks.

[0085] In a practical application scenario of this invention, taking a hybrid network structure of 6 SCMA cellular users, 2 D2D users, and 4 spectrum resource blocks as an example, the connection factor matrix... Represented as:

[0086]

[0087] Wherein, the connection factor matrix The first six columns correspond to six SCMA cellular users, and the last two columns correspond to two D2D users, in the connection factor matrix. The four rows in the table correspond to the occupancy status of the four spectrum resource blocks.

[0088] S2. Construct a codeword generator based on the connection factor graph, and use the codeword generator to calculate the codeword corresponding to the user information; wherein, the codeword generator is composed of multiple DNN encoding units, and the number of output nodes in the output layer of the DNN encoding unit is set based on the mapping relationship of the connection factor graph.

[0089] In this embodiment of the invention, the codeword generator at the encoding end is configured according to the connection factor graph. A DNN (Dense Neural Network) unit can be configured for each user node in the connection factor graph. The number of nodes in the output layer of each DNN unit and the connection method with the frequency resource block are set according to the factor graph of the hybrid network. The codeword generator generated by this method simplifies the encoder structure and reduces the encoding complexity. It can be called a low-complexity SCMA-D2D (lower complexity D-SCMA-D2D, abbreviated as LD-SCMA-D2D) encoder. Furthermore, SCMA cellular users and D2D users can adaptively learn the optimal encoding mapping of the connected spectrum resource blocks through their respective DNN units to generate the optimal two-dimensional codeword information.

[0090] In this embodiment of the invention, constructing a codeword generator based on the connection factor graph includes:

[0091] Configure the corresponding initialization DNN encoding unit according to the user nodes in the connection factor graph;

[0092] A one-hot vector is generated by obtaining the user information corresponding to the user of the user node and performing one-hot encoding.

[0093] The number of nodes in the input layer of the initial DNN encoding unit is set according to the dimension of the single-hot vector, and the number of nodes in the output layer of the initial DNN encoding unit is set according to the number of connection lines of user nodes in the connection factor graph, thus obtaining the DNN encoding unit.

[0094] The codeword generator is obtained by summing up the DNN encoding units corresponding to each user node.

[0095] In this embodiment of the invention, each user is configured with a DNN unit as their codeword generator, and this DNN unit is a fully connected DNN unit. Indicates the first Codeword generator for SCMA cellular users , Indicates the first A codeword generator for D2D users. .

[0096] In this embodiment of the invention, the number of nodes in the input layer of each codeword generator is consistent with the dimension of the single-hot vector after one-hot encoding for each user; the number of nodes in the output layer of each codeword generator is determined according to the number of connections corresponding to each user in the connection factor graph, and the number of output layer nodes of the codeword generators for SCMA cellular users and D2D users are different; each user in the SCMA cellular user occupies If there are 1 resource block, and each resource block modulates a complex constellation point, then the number of output layer nodes is 1. Each user in the D2D user group occupies If there are 1 resource block, then the number of output layer nodes for a D2D user is 1. .

[0097] In this embodiment of the invention, before inputting user information into the codeword generator, the user information needs to be one-hot encoded, and the binary data corresponding to each user's user information is encoded according to... Bits are grouped together and converted using one-hot encoding. A one-hot vector of dimension is represented as ;No. User data of SCMA cellular users The corresponding single heat vector is represented as , D2D user data The corresponding single heat vector is represented as , Each single heat vector , Only one element has a value of 1, while the other elements have a value of 0.

[0098] In this embodiment of the invention, the single-hot vector after single-hot encoding is input into the corresponding codeword generator to generate the corresponding codeword. Since each SCMA cellular user occupies... Given a spectrum resource, the codeword generated by an SCMA cellular user can be represented as: , , , , They are not equal, and Because each D2D user occupies Given a spectrum resource, the D2D user represents the generated codeword as follows: , Indicates the first Codeword generator for each SCMA cellular user neural network hyperparameters, Including weights and deviation ; Indicates the first Codeword generator for each D2D user neural network hyperparameters, include and deviation .

[0099] In a practical application scenario of this invention, there are 6 SCMA cellular users, 2 D2D users, and 4 spectrum resource blocks. Each SCMA cellular user occupies... For each spectrum resource block, the codeword generated by the SCMA cellular user is represented as follows: , and Each D2D user occupies There are one spectrum resource block, so the codeword representation generated by the D2D user is: The codeword generator for each SCMA cellular user has 4 output nodes, representing the real and imaginary parts of the complex digital characters loaded on the two spectrum resource blocks for each SCMA cellular user; the codeword generator for each D2D user has 2 output nodes, representing the real and imaginary parts of the complex modulation information loaded on the spectrum resource blocks by the D2D user transmitter.

[0100] See Figure 2 The diagram shows the connection factor of 6 SCMA cellular users, 2 D2D users, and 4 spectrum resource blocks. The mapping codewords of SCMA cellular users and D2D users are non-orthogonally superimposed on the 4 spectrum resources. C1-C6 represent 6 SCMA cellular users, DT1 and DT2 represent 2 D2D users, and K1-K4 represent 4 spectrum resource blocks.

[0101] Furthermore, in this embodiment of the invention, each SCMA cellular user and D2D user is configured with a DNN unit at the encoder end. The output layer of the DNN unit sets and connects the number of output nodes according to the mapping relationship of the connection factor graph. Taking 6 SCMA cellular users and 2 D2D users as an example, the encoder of the existing hybrid network autoencoder scheme requires 14 DNN units, while the encoder of this invention only has 8 DNN units. The number of DNN network units at the encoder end is significantly reduced. Although the number of output nodes of the DNN unit at the encoder end is increased, the overall computational complexity of the encoder in this invention is still effectively reduced because the number of neural network layers of the DNN unit is also reduced accordingly.

[0102] S3. Generate hybrid network signal data based on the codewords corresponding to the user information and the spectrum resource block.

[0103] In this embodiment of the invention, based on the connection between each user node and the spectrum resource block node in the connection factor graph, the complex digital code of the corresponding output node of the codeword generator is modulated onto the connected resource block to achieve carrier modulation.

[0104] In this embodiment of the invention, generating hybrid network signal data based on the codeword corresponding to the user information and the spectrum resource block includes:

[0105] Determine the target user node connected to the spectrum resource block, and obtain the target codeword of the user information corresponding to the target user node;

[0106] The target codewords are superimposed to obtain the resource block codewords corresponding to each spectrum resource block;

[0107] The resource block codewords are calculated using the following output formula to obtain the hybrid network signal data;

[0108] The output formula is expressed as follows:

[0109]

[0110] in, This is represented as the hybrid network signal data. Represented as the first The resource block codeword corresponding to each spectrum resource block This represents the total number of spectrum resource blocks.

[0111] S4. Input the hybrid network signal data after transmission through the Gaussian channel into the pre-constructed relay network to obtain relay signal data, and broadcast the relay signal data to the receiving end corresponding to the SCMA cellular user and the D2D user.

[0112] In this embodiment of the invention, the noise of the Gaussian channel is white Gaussian noise. The mean is 0 and the variance is 0. Hybrid network signal data is transmitted through a Gaussian channel. During transmission, it is affected by Gaussian white noise, and the signal after being corrupted by the noise is represented as follows: .

[0113] In this embodiment of the invention, the relay network is established based on relays. A relay is a connection device that operates at the physical layer and is a device for connecting network lines. It is used to interconnect two networks and is often used for bidirectional forwarding of physical signals between two network nodes. In this invention, it is used to connect the encoder and the decoder and to transmit signal data between the encoder and the decoder.

[0114] In this embodiment of the invention, before inputting the hybrid network signal data transmitted via the Gaussian channel into the pre-constructed relay network, the method further includes:

[0115] Construct the relay sharing layer and initialize the DNN relay unit;

[0116] The number of nodes in the input layer of the initial DNN relay unit is configured according to the dimension of the hybrid network signal data after transmission through the Gaussian channel, thus obtaining the DNN relay unit;

[0117] A relay network is generated based on the relay sharing layer and the DNN relay unit.

[0118] In this embodiment of the invention, the relay configuration includes a fully connected DNN unit, denoted as... , This represents the input of the DNN relay unit; therefore, the number of nodes in the input layer of the DNN relay unit is equal to the signal vector. The number of nodes in the output layer of the relay DNN unit is set through training and debugging. For the hyperparameters of the hidden layer in a neural network with relay sharing layers, Including the weights of the relay sharing layer ,deviation , Indicates the number of layers in the relay sharing layer.

[0119] In this embodiment of the invention, the number of layers in the DNN unit and the number of nodes in the hidden layer can both be set through training and debugging. (See reference...) Figure 3 As shown, each hidden layer contains a regularization layer and an activation layer, which alleviates the gradient vanishing problem to some extent. The last layer of the DNN unit uses an activation function as the activation output, which can be a softmax function or a ReLU function.

[0120] In this embodiment of the invention, the output data (i.e., relay signal data) of the DNN relay unit is broadcast to the receiving ends corresponding to the SCMA cellular users and D2D users, as shown below. In the case of broadcasting relay signal data, the relay signal data can be modulated onto a high-frequency radio frequency to obtain a high-frequency signal. This high-frequency signal is then broadcast wirelessly by taking advantage of the long-distance propagation characteristics of radio waves in the air. After tuning, it is received by the aforementioned receiving end.

[0121] S5. After receiving the relay signal data at the receiving end, the relay signal data is decoded using a pre-built multi-user classification decoder (LDS-SCMA-D2D) to obtain the decoding information corresponding to the SCMA cellular user and the D2D user; wherein, the multi-user classification decoder (LDS-SCMA-D2D) includes the DNN decoding unit corresponding to the SCMA cellular user and the D2D user in the receiving end.

[0122] In this embodiment of the invention, before decoding the relay signal data using a pre-built multi-user classification decoder (LDS-SCMA-D2D), the method further includes:

[0123] Configure a corresponding initialization DNN decoding unit for each user at the receiving end;

[0124] The number of nodes in the input layer of the initial DNN decoding unit is set according to the dimension of the relay signal data, and the number of nodes in the output layer of the initial DNN decoding unit is set according to the dimension of the single-hot vector, thus obtaining the DNN decoding unit;

[0125] A multi-user classification decoder (LDS-SCMA-D2D) is generated based on the initialized shared decoding layer and the DNN decoding unit.

[0126] In this embodiment of the invention, the receiving end includes a base station corresponding to SCMA cellular users and a D2D user terminal corresponding to D2D users; both the base station and the D2D user terminal are configured with a structurally identical fully connected DNN decoding unit, and the number of input layer nodes depends on... The dimensions of the output layer nodes and the single-hot vector. Maintain consistency.

[0127] In this embodiment of the invention, the step of decoding the relay signal data using a pre-built multi-user classification decoder (LDS-SCMA-D2D) to obtain user information corresponding to the SCMA cellular user and the D2D user includes:

[0128] The relay signal data is decoded using the multi-user classification decoder (LDS-SCMA-D2D) to obtain SCMA cellular user decoded data and D2D user decoded data;

[0129] Probability calculations are performed on the SCMA cellular user decoding data and the D2D user decoding data to obtain the probability information corresponding to the SCMA cellular user decoding data and the D2D user decoding data;

[0130] The probability information is converted into binary to obtain the decoding information corresponding to the SCMA cellular user and the D2D user.

[0131] In this embodiment of the invention, after decoding the relay signal data using a multi-user classification decoder (LDS-SCMA-D2D), the first... The decoding results output by each SCMA cellular user are represented as follows: , No. Decoding results output by a D2D user ;in, This represents the SCMA cellular subscriber decoder, whose input is... The neural network hyperparameters of the corresponding decoding shared layer at the base station are: The output SCMA cellular user decoding data is ; This represents the D2D user decoder, where the input to the D2D user decoder is... The neural network hyperparameters of the decoding shared layer corresponding to the D2D user terminal are: The output D2D user-decoded data is .

[0132] In this embodiment of the invention, in the multi-user classification decoder (LDS-SCMA-D2D), the softmax function can be used as the activation function for each user at the decoding end; the softmax function can classify users into classes. Transform a vector of real values ​​into A real-valued vector whose sum is 1. It is any real number, which can then be used to convert the real-valued vector output by the multi-user classification decoder into a normalized probability distribution.

[0133] In this embodiment of the invention, after obtaining the decoded probability information, the decoded output is respectively... 3D probability information , User information converted into one-dimensional binary form with b bits as a group , , This represents the decoding information corresponding to SCMA cellular users. This represents the decoding information corresponding to the D2D user.

[0134] See Figure 4 The diagram shows the structure of the SCMA-D2D hybrid network self-encoder model of the present invention. In the diagram, user information from 6 SCMA cellular users and 2 D2D users is encoded in a single-hot manner and then input into a codeword generator for encoding to obtain the optimal mapping codeword and modulation information. Then, according to... Figure 2 The connection factor diagram shown superimposes the mapped codewords non-orthogonally onto four spectrum resources. The signal is transmitted to the relay network through a Gaussian channel. After the relay process obtains the hybrid network signal, it is broadcast to the base station corresponding to the SCMA cellular user and the D2D user receiver corresponding to the D2D user. The base station and the D2D user receiver perform multi-user classification decoding to obtain the decoded information.

[0135] In this embodiment of the invention, after obtaining the decoding information corresponding to the SCMA cellular user and the D2D user, the method further includes:

[0136] Calculate the cross-entropy loss based on the user information corresponding to the SCMA cellular user and the D2D user, as well as the corresponding decoding information.

[0137] The codeword generator, relay network, and multi-user classification decoder are jointly optimized based on the cross-entropy loss to obtain an SCMA-D2D hybrid network self-coder model composed of the optimized codeword generator, relay network, and multi-user classification decoder.

[0138] In this embodiment of the invention, the cross-entropy loss function can be used as the loss function for optimizing the SCMA-D2D hybrid network self-encoder model to measure the difference between the distribution learned by the model and the true distribution. The larger the cross-entropy, the greater the difference between the two probability distributions. The goal of model optimization is to minimize the value of the loss function so that the model can achieve the best performance. The probability information decoded by the decoder in this invention can first be processed into data between 0 and 1 by softmax, and then sent to log() for calculation.

[0139] In this embodiment of the invention, the step of calculating the cross-entropy loss based on the user information corresponding to the SCMA cellular user and the D2D user, and the corresponding decoding information, includes:

[0140] The first cross-entropy loss is obtained by calculating the user information and decoding information corresponding to the SCMA cellular user using a preset first cross-entropy loss function.

[0141] The second cross-entropy loss is obtained by calculating the user information and decoding information corresponding to the D2D user using a preset second cross-entropy loss function.

[0142] In this embodiment of the invention, the first cross-entropy loss function is expressed as:

[0143]

[0144] in, Represented as the first The cross-entropy of each SCMA cellular user. Represented as the first Decoding information corresponding to each SCMA cellular user. Represented as the first User information corresponding to each SCMA cellular user. This is expressed as the number of SCMA cellular users. This represents the number of spectrum resource blocks. yes The item, yes The item, This is represented by the first cross-entropy loss corresponding to the SCMA cellular user. Represented as , The set, Represented as , A set;

[0145] The second cross-entropy loss function is expressed as:

[0146]

[0147] in, Represented as the first Cross-entropy corresponding to each D2D user Represented as the first Decoding information corresponding to each D2D user Represented as the first User information corresponding to each D2D user This is represented by the number of D2D users. This represents the number of spectrum resource blocks. yes The item, yes The item, This is represented as the second cross-entropy loss corresponding to the D2D user. Represented as , The set, Represented as , A set of.

[0148] In this embodiment of the invention, during the joint optimization process, the hidden layer parameters in each DNN unit of the encoder, relay, and receiver can be used as learning targets. They can be optimized using a pre-set loss function and optimizer to achieve accurate prediction of the model structure, determine the optimal DNN unit structure, size, bias, and weight parameters, thereby improving decoding accuracy and reducing interference between users.

[0149] In another optional embodiment of the present invention, binary input data bits of users (SCMA cellular users or D2D users) can be randomly generated, grouped into b-bit sets, as input data for the SCMA-D2D hybrid network self-coder; then, the BER (bit error rate) performance of the SCMA-D2D hybrid network self-coder under different signal-to-noise ratios is calculated, and the BER is characterized within the normal received signal strength range of the link. Figure 5 , Figure 6 and Figure 7 The figures shown are comparisons of the BER performance of the multi-user classifier decoder SCMA, the multi-user classifier decoder D2D, and the BER performance of the SCMA-D2D hybrid network self-written decoder model under different signal-to-noise ratios. The performance changes of the encoder, decoder, and hybrid network module within the range of 2dB to 14dB allow us to determine the optimal signal-to-noise ratio.

[0150] Compared to traditional LSD-MPA and LSD-ML, the embodiments of the present invention have better anti-interference performance among users and higher decoding accuracy. Compared to D-SCMA-D2D and CNN-SCMA-D2D, the embodiments of the present invention have fewer encoder units, lower complexity, and better anti-interference performance of the decoder.

[0151] In embodiments of the present invention, such as Figure 8 The image shows a comparison of decoding times between the traditional LSD decoding algorithm and the multi-user classification decoder (LDS-SCMA-D2D). Compared to the traditional LSD-MPA iterative decoding algorithm, the decoding time in this embodiment is significantly reduced. Figure 9The figure shows a comparison of the encoding complexity of the CNN-SCMA-D2D encoder and the (LDS-SCMA-D2D) codeword generator. Compared with the hybrid network encoder of CNN-SCMA-D2D which uses 12 encoding units, the encoding computation complexity of the present invention is lower. Among them, MAC (Multiplication and Addition) is used to evaluate the number of multiplication and addition operations performed by the model during runtime.

[0152] In this embodiment of the invention, a D2D user type is added to the deep learning-based SCMA encoding / decoding scheme, and the number of encoder units is reduced, while the implementation still uses a fully connected neural network. Compared with the existing CNN-SCMA-D2D hybrid network encoder / decoder based on convolutional neural networks, the encoder structure of this embodiment is simpler. This embodiment can use offline training and single-stage decoding, which greatly reduces complexity compared to traditional iterative LSD-MPA and LSD-ML multi-user decoding algorithms, and has lower encoder complexity compared to the CNN-SCMA-D2D hybrid network encoder / decoder. Therefore, this embodiment can effectively reduce encoding complexity, and by preprocessing the mixed signal through a relay before broadcasting it to different user classification decoders for classification decoding, it effectively reduces interference between users and improves decoding accuracy.

[0153] This invention combines SCMA and D2D encoding / decoding, reducing the number of encoder units based on different user types. The encoder employs a fully connected neural network, resulting in a simpler structure compared to existing CNN-SCMA-D2D hybrid network encoders / decoders based on convolutional neural networks. By constructing a codeword generator using connection factor graphs, the number of output nodes in the DNN units at the encoder end is increased while the number of neural network layers in the DNN units is correspondingly reduced, thus lowering the overall computational complexity of the encoder. Based on the multi-task branch decoding strategy of the DNN units in the relay network, the hybrid network signal is preprocessed by the relay before being broadcast to the multi-user classification decoder for classification decoding. This allows useful information from D2D users and SCMA cellular users to be extracted and shared in the relay, improving the decoding accuracy of user decoders based on hybrid network signals and effectively reducing interference between users. Compared to CNN-SCMA-D2D hybrid network encoders / decoders, this invention has a lower encoder complexity. Therefore, the proposed low-complexity SCMA-D2D network encoding / decoding method can solve the problem of not being able to effectively improve the accuracy of multi-user decoding while reducing encoding / decoding complexity.

[0154] like Figure 10 The diagram shown is a functional block diagram of a low-complexity encoding and decoding device for SCMA-D2D networks provided in an embodiment of the present invention.

[0155] The SCMA-D2D network low-complexity encoding and decoding device 1000 described in this invention can be installed in an electronic device. Depending on the functions implemented, the SCMA-D2D network low-complexity encoding and decoding device 1000 may include a connection factor graph generation module 1001, an encoding module 1002, a spectrum resource mapping module 1003, a relay network processing module 1004, and a decoding module 1005. The module described in this invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0156] In this embodiment, the functions of each module / unit are as follows:

[0157] The connection factor graph generation module 1001 is used to generate a connection factor matrix based on the user information of SCMA cellular users and D2D users and a preset spectrum resource block, and to generate a connection factor graph based on the connection factor matrix.

[0158] The encoding module 1002 is used to construct a codeword generator based on the connection factor graph, and to calculate the codeword corresponding to the user information using the codeword generator; wherein, the codeword generator is composed of multiple DNN encoding units, and the number of output nodes in the output layer of the DNN encoding unit is set based on the mapping relationship of the connection factor graph;

[0159] The spectrum resource mapping module 1003 is used to generate hybrid network signal data based on the codewords corresponding to the user information and the spectrum resource blocks.

[0160] The relay network processing module 1004 is used to input the hybrid network signal data after transmission through the Gaussian channel into the pre-constructed relay network to obtain relay signal data, and broadcast the relay signal data to the receiving end corresponding to the SCMA cellular user and the D2D user.

[0161] The decoding module 1005 is used to decode the relay signal data using a pre-built multi-user classification decoder after the receiving end receives the relay signal data, to obtain the decoding information corresponding to the SCMA cellular user and the D2D user; wherein, the multi-user classification decoder includes the DNN decoding unit corresponding to the SCMA cellular user and the D2D user in the receiving end.

[0162] In detail, each module in the SCMA-D2D network low-complexity encoding and decoding device 1000 described in this embodiment of the invention uses the same technical means as the SCMA-D2D network low-complexity encoding and decoding method described in the accompanying drawings, and can produce the same technical effect, which will not be repeated here.

[0163] like Figure 11 The diagram shown is a schematic representation of an electronic device that implements a low-complexity encoding and decoding method for SCMA-D2D networks according to an embodiment of the present invention.

[0164] The electronic device 1100 may include a processor 1101, a memory 1102, a communication bus 1103, and a communication interface 1104. It may also include a computer program, such as an SCMA-D2D network low-complexity codec program, stored in the memory 1102 and capable of running on the processor 1101.

[0165] In some embodiments, the processor 1101 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 1101 is the control unit of the electronic device, connecting various components of the entire electronic device through various interfaces and lines. It executes programs or modules stored in the memory 1102 (e.g., executing SCMA-D2D network low-complexity codec programs) and calls data stored in the memory 1102 to perform various functions of the electronic device and process data.

[0166] The memory 1102 includes at least one type of readable storage medium, including flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 1102 can be an internal storage unit of an electronic device, such as a portable hard drive. In other embodiments, the memory 1102 can be an external storage device of the electronic device, such as a plug-in portable hard drive, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc. Furthermore, the memory 1102 can include both internal and external storage units of the electronic device. The memory 1102 can be used not only to store application software and various types of data installed on the electronic device, such as the code of a low-complexity SCMA-D2D network codec, but also to temporarily store data that has been output or will be output.

[0167] The communication bus 1103 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 1102 and at least one processor 1101, etc.

[0168] The communication interface 1104 is used for communication between the aforementioned electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, Bluetooth interface, etc.), typically used to establish communication connections between the electronic device and other electronic devices. The user interface may be a display, an input unit (such as a keyboard), or optionally, a standard wired or wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the electronic device and to display a visual user interface.

[0169] Figure 11 Only electronic devices with components are shown; it will be understood by those skilled in the art that... Figure 11 The structure shown does not constitute a limitation on the electronic device 1100, and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0170] For example, although not shown, the electronic device may also include a power supply (such as a battery) to power the various components. Preferably, the power supply can be logically connected to the at least one processor 1101 through a power management device, thereby enabling functions such as charging management, discharging management, and power consumption management. The power supply may also include one or more DC or AC power supplies, recharging devices, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0171] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.

[0172] The SCMA-D2D network low-complexity encoding and decoding program stored in the memory 1102 of the electronic device 1100 is a combination of multiple instructions. When run in the processor 1101, it can achieve the following:

[0173] A connection factor matrix is ​​generated based on the user information of SCMA cellular users and D2D users and the preset spectrum resource blocks, and a connection factor graph is generated based on the connection factor matrix.

[0174] A codeword generator is constructed based on the connection factor graph, and the codeword corresponding to the user information is calculated using the codeword generator; wherein, the codeword generator is composed of multiple DNN coding units, and the number of output nodes in the output layer of the DNN coding unit is set based on the mapping relationship of the connection factor graph;

[0175] Generate hybrid network signal data based on the codewords corresponding to the user information and the spectrum resource blocks;

[0176] The hybrid network signal data, after being transmitted through a Gaussian channel, is input into a pre-constructed relay network to obtain relay signal data, and the relay signal data is broadcast to the receiving end corresponding to the SCMA cellular user and the D2D user.

[0177] After receiving the relay signal data, the receiver uses a pre-built multi-user classification decoder (LDS-SCMA-D2D) to decode the relay signal data and obtain the decoding information corresponding to the SCMA cellular user and the D2D user; wherein, the multi-user classification decoder (LDS-SCMA-D2D) includes the DNN decoding unit corresponding to the SCMA cellular user and the D2D user in the receiver.

[0178] Specifically, the specific implementation method of the processor 1101 of the above instructions can be referred to the description of the relevant steps in the corresponding embodiment of the accompanying drawings, and will not be repeated here.

[0179] Furthermore, if the modules / units integrated in the electronic device 1100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM).

[0180] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor of an electronic device, can perform the following:

[0181] A connection factor matrix is ​​generated based on the user information of SCMA cellular users and D2D users and the preset spectrum resource blocks, and a connection factor graph is generated based on the connection factor matrix.

[0182] A codeword generator is constructed based on the connection factor graph, and the codeword corresponding to the user information is calculated using the codeword generator; wherein, the codeword generator is composed of multiple DNN coding units, and the number of output nodes in the output layer of the DNN coding unit is set based on the mapping relationship of the connection factor graph;

[0183] Generate hybrid network signal data based on the codewords corresponding to the user information and the spectrum resource blocks;

[0184] The hybrid network signal data, after being transmitted through a Gaussian channel, is input into a pre-constructed relay network to obtain relay signal data, and the relay signal data is broadcast to the receiving end corresponding to the SCMA cellular user and the D2D user.

[0185] After receiving the relay signal data, the receiving end uses a pre-built multi-user classification decoder to decode the relay signal data to obtain the decoding information corresponding to the SCMA cellular user and the D2D user; wherein, the multi-user classification decoder includes the DNN decoding unit corresponding to the SCMA cellular user and the D2D user in the receiving end.

[0186] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0187] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0188] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0189] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0190] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0191] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a system claim may also be implemented by a single unit or device through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any specific order.

[0192] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A low-complexity encoding and decoding method for SCMA-D2D networks, characterized in that, The method includes: A connection factor matrix is ​​generated based on the user information of SCMA cellular users and D2D users and the preset spectrum resource blocks, and a connection factor graph is generated based on the connection factor matrix. A codeword generator is constructed based on the connection factor graph, and the codeword corresponding to the user information is calculated using the codeword generator; wherein, the codeword generator is composed of multiple DNN coding units, and the number of output nodes in the output layer of the DNN coding unit is set based on the mapping relationship of the connection factor graph; Generate hybrid network signal data based on the codewords corresponding to the user information and the spectrum resource blocks; The hybrid network signal data, after being transmitted through a Gaussian channel, is input into a pre-constructed relay network to obtain relay signal data, and the relay signal data is broadcast to the receiving end corresponding to the SCMA cellular user and the D2D user. After receiving the relay signal data, the receiving end uses a pre-built multi-user classification decoder to decode the relay signal data to obtain the decoding information corresponding to the SCMA cellular user and the D2D user; wherein, the multi-user classification decoder includes the DNN decoding unit corresponding to the SCMA cellular user and the D2D user in the receiving end.

2. The low-complexity encoding and decoding method for SCMA-D2D networks according to claim 1, characterized in that, The step of generating hybrid network signal data based on the codeword corresponding to the user information and the spectrum resource block includes: Determine the target user node connected to the spectrum resource block, and obtain the target codeword of the user information corresponding to the target user node; The target codewords are superimposed to obtain the resource block codewords corresponding to each spectrum resource block; The resource block codewords are calculated using the following output formula to obtain the hybrid network signal data; The output formula is expressed as follows: in, This is represented as the hybrid network signal data. Represented as the first The resource block codeword corresponding to each spectrum resource block This represents the total number of spectrum resource blocks.

3. The low-complexity encoding and decoding method for SCMA-D2D networks as described in claim 1, characterized in that, The connection factor matrix is ​​represented as follows: in, Represented as a link factor matrix, This represents the number of spectrum resource blocks. This represents the number of SCMA cellular users. This represents the number of D2D users in the user information.

4. The low-complexity encoding and decoding method for SCMA-D2D networks as described in claim 1, characterized in that, Before inputting the hybrid network signal data transmitted via the Gaussian channel into the pre-built relay network, the method further includes: Construct the relay sharing layer and initialize the DNN relay unit; The number of nodes in the input layer of the initial DNN relay unit is configured according to the dimension of the hybrid network signal data after transmission through the Gaussian channel, thus obtaining the DNN relay unit; A relay network is generated based on the relay sharing layer and the DNN relay unit.

5. The low-complexity encoding and decoding method for SCMA-D2D networks as described in claim 1, characterized in that, The step of constructing a codeword generator based on the connection factor graph includes: Configure the corresponding initialization DNN encoding unit according to the user nodes in the connection factor graph; A one-hot vector is generated by obtaining the user information corresponding to the user of the user node and performing one-hot encoding. The number of nodes in the input layer of the initial DNN encoding unit is set according to the dimension of the single-hot vector, and the number of nodes in the output layer of the initial DNN encoding unit is set according to the number of connection lines of user nodes in the connection factor graph, thus obtaining the DNN encoding unit. The codeword generator is obtained by summing up the DNN encoding units corresponding to each user node.

6. The low-complexity encoding and decoding method for SCMA-D2D networks as described in claim 5, characterized in that, Before decoding the relay signal data using a pre-built multi-user classification decoder, the method further includes: Configure a corresponding initialization DNN decoding unit for each user at the receiving end; The number of nodes in the input layer of the initial DNN decoding unit is set according to the dimension of the relay signal data, and the number of nodes in the output layer of the initial DNN decoding unit is set according to the dimension of the single-hot vector, thus obtaining the DNN decoding unit; A multi-user classification decoder is generated based on the initialized decoding shared layer and the DNN decoding unit.

7. The low-complexity encoding and decoding method for SCMA-D2D networks as described in claim 1, characterized in that, The step of decoding the relay signal data using a pre-built multi-user classification decoder to obtain user information corresponding to the SCMA cellular user and the D2D user includes: The relay signal data is decoded using the multi-user classification decoder to obtain SCMA cellular user decoded data and D2D user decoded data; Probability calculations are performed on the SCMA cellular user decoding data and the D2D user decoding data to obtain the probability information corresponding to the SCMA cellular user decoding data and the D2D user decoding data; The probability information is converted into binary to obtain the decoding information corresponding to the SCMA cellular user and the D2D user.

8. The low-complexity encoding and decoding method for SCMA-D2D networks as described in claim 1, characterized in that, After obtaining the decoding information corresponding to the SCMA cellular user and the D2D user, the process further includes: Calculate the cross-entropy loss based on the user information corresponding to the SCMA cellular user and the D2D user, as well as the corresponding decoding information. The codeword generator, relay network, and multi-user classification decoder are jointly optimized based on the cross-entropy loss to obtain an SCMA-D2D hybrid network self-coder model composed of the optimized codeword generator, relay network, and multi-user classification decoder.

9. The low-complexity encoding and decoding method for SCMA-D2D networks as described in claim 8, characterized in that, The step of calculating cross-entropy loss based on the user information corresponding to the SCMA cellular user and the D2D user, and the corresponding decoding information, includes: The first cross-entropy loss is obtained by calculating the user information and decoding information corresponding to the SCMA cellular user using a preset first cross-entropy loss function. The first cross-entropy loss function is expressed as: in, Represented as the first Cross-entropy corresponding to each SCMA cellular user Represented as the first Decoding information corresponding to each SCMA cellular user. Represented as the first User information corresponding to each SCMA cellular user. This is expressed as the number of SCMA cellular users. This represents the number of spectrum resource blocks. yes The item, yes The item, This is represented by the first cross-entropy loss corresponding to the SCMA cellular user. Represented as , The set, Represented as , A set; The second cross-entropy loss is obtained by calculating the user information and decoding information corresponding to the D2D user using a preset second cross-entropy loss function. The second cross-entropy loss function is expressed as: in, Represented as the first Cross-entropy corresponding to each D2D user Represented as the first Decoding information corresponding to each D2D user Represented as the first User information corresponding to each D2D user This is expressed as the number of D2D users. This represents the number of spectrum resource blocks. yes The item, yes The item, This is represented as the second cross-entropy loss corresponding to the D2D user. Represented as , The set, Represented as , A set of.

10. A low-complexity encoding and decoding device for SCMA-D2D networks, characterized in that, The device includes: The connection factor graph generation module is used to generate a connection factor matrix based on the user information of SCMA cellular users and D2D users and a preset spectrum resource block, and to generate a connection factor graph based on the connection factor matrix. An encoding module is used to construct a codeword generator based on the connection factor graph, and to calculate the codeword corresponding to the user information using the codeword generator; wherein, the codeword generator is composed of multiple DNN encoding units, and the number of output nodes in the output layer of the DNN encoding unit is set based on the mapping relationship of the connection factor graph; The spectrum resource mapping module is used to generate hybrid network signal data based on the codewords corresponding to the user information and the spectrum resource blocks. The relay network processing module is used to input the hybrid network signal data after transmission through the Gaussian channel into the pre-constructed relay network to obtain relay signal data, and broadcast the relay signal data to the receiving end corresponding to the SCMA cellular user and the D2D user. The decoding module is used to decode the relay signal data using a pre-built multi-user classification decoder after the receiving end receives the relay signal data, to obtain the decoding information corresponding to the SCMA cellular user and the D2D user; wherein the multi-user classification decoder includes the DNN decoding unit corresponding to the SCMA cellular user and the D2D user in the receiving end.

Images