Optimal beam forming design method under semantic communication service quality constraint

Abstract

The invention discloses an optimal beam forming design method under semantic communication service quality constraint, and belongs to the technical field of wireless communication. An optimal beam forming design method is provided for a multi-input single-output digital semantic communication system; an end-to-end semantic performance and compression ratio model is utilized to establish a relational expression of end-to-end semantic performance and a signal-to-noise ratio, and an optimal beam is constructed with the purposes of meeting semantic communication service quality and minimizing transmitting power; dual constraints of transmission time delay and semantic communication service quality are met, and meanwhile, an optimal beam is constructed by minimizing a semantic energy consumption target; through a Dinkelbach method and a matched filtering strategy, the optimal wave beam meeting the requirements of transmission delay and semantic communication service quality at the same time is designed under the target of minimum semantic energy consumption, and the robustness and the resource utilization rate of a semantic communication system under the conditions of high bandwidth, high reliability and limited resources can be remarkably improved.

Description

Technical Field

[0001] This invention relates to an optimal beamforming design method under semantic communication service quality constraints, belonging to the field of wireless communication technology. Background Technology

[0002] Semantic communication, as an emerging intelligent communication paradigm, has gradually attracted widespread attention. Unlike traditional communication methods based on bit-level accurate transmission, semantic communication emphasizes the modeling and extraction of the semantic hierarchy of information. It utilizes artificial intelligence technologies such as deep learning to mine and abstract semantic features of source data, thereby achieving efficient transmission of useful information. Through semantic-level compression and reconstruction, semantic communication can not only effectively reduce the transmission overhead of redundant data but also improve the robustness and intelligence of communication systems in complex environments. Although this type of method improves end-to-end transmission efficiency to some extent, several prominent problems remain. First, this type of method generally lacks explicit guarantees for the quality of service of semantic communication, and cannot guarantee stable and reliable semantic transmission in multi-user and complex channel environments. Second, the black-box nature of deep learning networks makes it difficult to quantify and model the relationship between end-to-end semantic transmission performance and physical layer parameters, thus limiting the feasibility of optimizing the allocation of physical layer resources. In other words, current research has not yet formed a theoretical and methodological framework that can simultaneously achieve physical layer beamforming optimization under the constraint of ensuring the quality of service of semantic communication.

[0003] Prior art, disclosed in publication number CN114091673A, comprises a semantic encoder, a decoder, and a semantic communication framework, belonging to the field of semantic communication. This invention, based on a traditional semantic communication architecture, adds a semantic comparator to the semantic encoder and a semantic inferencer to the semantic decoder. Instead of directly encoding semantic information at the sender, this semantic communication architecture trains a semantic comparator capable of distinguishing the sender's expert reasoning path. The semantic inferencer, through adversarial training against the semantic comparator, learns the reasoning mechanism of the sender's expert reasoning path and can ultimately decode the received information to recover the semantic information. However, this requires training using a neural network, resulting in high costs and computational demands. Summary of the Invention

[0004] To address the shortcomings of existing semantic communication technologies, this invention provides an optimal beamforming design method under semantic communication service quality constraints. It proposes the ABG formula, which describes the relationship between end-to-end semantic performance and signal-to-noise ratio, and designs an optimized beamforming scheme based on this formula. This scheme minimizes transmission power while ensuring semantic communication service quality, thereby reducing system power consumption; it also minimizes semantic communication energy consumption under transmission delay and service quality constraints. Through these solutions, this invention can significantly enhance the reliability and resource utilization efficiency of semantic communication systems, making it suitable for applications requiring high bandwidth and high reliability.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an optimal beamforming design method under semantic communication service quality constraints, comprising the following steps: S1. Using an end-to-end semantic performance and compression rate model, establish the relationship between end-to-end semantic performance and signal-to-noise ratio. S2. Based on the relationship between end-to-end semantic performance and signal-to-noise ratio, construct the optimal beam to meet the quality of semantic communication services and minimize transmission power. S3, to construct the optimal beam to meet the dual constraints of transmission delay and semantic communication service quality, while minimizing semantic energy consumption; S4. Using the Dinkelbach method and matched filtering strategy, we design the optimal beam that minimizes semantic energy consumption while simultaneously meeting the requirements of transmission delay and semantic communication service quality, thereby achieving a balance between energy consumption and transmission reliability in the semantic communication system.

[0006] Furthermore, the theoretical relationship between end-to-end semantic performance and compression ratio can be expressed as follows: , in Corresponding to a given compression ratio The theoretical upper limit of semantic coding performance is reflected in the maximum perceptual quality achievable under different compression levels. These are parameters for the semantic encoder. Semantic compression rate affects the quality of semantic communication services. and These represent the quantization bit count and the original bit count of the image, respectively. The ABG formula for the relationship between end-to-end semantic performance and signal-to-noise ratio is expressed as: , in Represents the set of beamforming vectors. For the first The beamforming vector corresponding to the root antenna. This refers to the number of antennas at the transmitter of a multi-input single-output channel. The upper bound of semantic communication performance, parameter Determined by the semantic codec, obtained through fitting experimental data; signal-to-noise ratio is superscript This indicates the conjugate transpose. This represents the channel vectors at the transmitting and receiving ends, where the vector... The elements of the array are independent and identically distributed, following a zero-mean complex Gaussian distribution with variance . ,and This represents the noise power of the additive white Gaussian noise channel.

[0007] Furthermore, based on the ABG formula relating end-to-end semantic performance and signal-to-noise ratio, an optimal beamforming design optimization problem satisfying semantic communication service quality constraints is constructed: , in, A threshold representing end-to-end semantic performance; The semantic layer's quality of service constraints are transformed into equivalent physical layer signal-to-noise ratio constraints, thus converting the optimal beamforming design optimization problem into: , , In the formula This represents the signal-to-noise ratio threshold required to achieve semantic communication service quality. This represents the noise power under the square root. Based on the matched filtering method, the optimal beamforming vector that satisfies the semantic communication service quality constraints is solved using the following formula. : .

[0008] Furthermore, the transmission delay from the transmitter to the receiver is expressed as: , in Indicates the original number of bits. Represents the compression ratio of a semantic communication system. Indicates transmission bandwidth; Based on the matched filtering method, the beamforming vector can be generally represented as: , in To represent the transmission power, we substitute the general expression for the beamforming vector into the signal-to-noise ratio (SNR) definition. The SNR is then expressed as: ; Based on the ABG formula relating end-to-end semantic performance and signal-to-noise ratio, an optimal energy-efficient beamforming design optimization model is constructed to satisfy the constraints of semantic communication service quality and transmission delay. , in For discrete values, Indicates the upper limit of transmission latency. The semantic transmission quality threshold for end-to-end transmission. Total power; The optimal energy consumption beam problem that satisfies the constraints of semantic communication quality of service and transmission delay is solved using an enumeration method: for determining the compression ratio through an exhaustive search method. For each fixed value, the optimization problem is transformed into variables. Optimization subproblems: , The Dinkelbach method was used to divide the variables. The optimization subproblem is transformed into: , in For non-negative update parameters; by separating variables The constraints are restated into the following linear constraint form: , Finally, the optimal solution for separating the variables was obtained using the primal-dual interior-point method. This allows us to obtain the beamforming vector with optimal semantic energy consumption. .

[0009] The semantic communication system used in the optimal beamforming design method under semantic communication service quality constraints includes a transmitter with multiple transmit antennas and a receiver with a single receive antenna. The transmitter includes a semantic encoder, a quantizer and a beam optimizer connected in sequence, and the receiver includes a channel equalizer, a dequantizer and a semantic decoder connected in sequence. A semantic encoder is used to extract semantic information from the raw data to be sent; A quantizer is used to convert the semantic information of an analog numerical type into a numeric type; The beam optimizer performs beam optimization processing on the quantized semantic information: it constructs the optimal beam with the goal of minimizing transmit power and satisfying the semantic communication service quality constraints, and constructs the optimal beam with the goal of minimizing system energy consumption under the dual constraints of transmission delay and semantic communication service quality; and transmits the beam to the receiver through multiple transmit antennas at the transmitter. A channel equalizer is used to equalize the received signals. A dequantizer is used to restore the equalized signal to the same dimension as the original semantic information, mapping a discrete vector to a continuous feature vector. A semantic decoder is used to reconstruct the original data from signals that have been restored to the same dimensions as the original semantic information.

[0010] Furthermore, the semantic encoder includes linear projection of image patches, positional encoding, a Vision Transformer model, and linear layers. The Vision Transformer model includes multi-layered normalization and multi-head attention structures, which are used to perform semantic linear extraction and joint source-channel coding on the original input image data.

[0011] A working method for a semantic communication system, comprising the following steps: Step 1: The transmitting end uses a semantic encoder to extract the raw data to be transmitted. User semantic information ,in This represents a semantic encoder. This represents the parameters that the semantic encoder can learn; Step 2: Use a quantizer that includes a linear layer and a sign function. Using a quantizer to simulate the semantic information of numerical types Quantized into a digital semantic signal of a specified number of bits , represented as ; Step 3: The transmitting end transmits the quantized digital semantic signal. The signal is transmitted through multiple antennas to a receiver equipped with a single antenna. The signal received by the receiver is: ,in This indicates the conjugate transpose. Indicates channel fading. Represents the beamforming vector. This represents the received additive white Gaussian noise, i.e. ; Step 4: Use the channel state information of the receiver to analyze the received signal. Perform equalization to obtain the equalized signal. ,Right now: ; Step 5: Use a dequantizer consisting of a linear layer. The equalized signal is processed to have the same dimension as the original semantic information, that is: ,in This represents the dequantized semantic information; Step 6: Dequantize the semantic information The input is fed into the semantic decoding module, which is built on the Vision Transformer model. This module performs non-linear mapping and feature reconstruction on the input semantic features to recover the original data information. ,Right now: ,in This represents a semantic decoder. This indicates its learnable parameters.

[0012] Beneficial Effects: This invention addresses the "black box" characteristics of deep learning encoding / decoding structures, weak theoretical foundations for semantic transmission, and difficulties in balancing quality of service (QoS) and physical layer resource optimization in existing semantic communication technologies. It proposes an optimal power and energy consumption beamforming design method under QoS constraints for semantic communication. By constructing an ABG empirical model between end-to-end semantic performance and signal-to-noise ratio (SNR), this invention establishes an analytical relationship between semantic communication performance and physical layer channel characteristics, thereby overcoming the limitation of lacking theoretical interpretability in traditional semantic communication systems.

[0013] The optimal beamforming design method proposed in this invention can minimize system transmission power while meeting the requirements of semantic communication service quality, thereby effectively reducing overall power consumption and improving system energy efficiency and stability. Simultaneously, by introducing the definition of semantic energy consumption and jointly optimizing compression ratio and transmission power, an energy consumption optimization model with transmission delay and semantic service quality as constraints is established. Furthermore, the Dinkelbach method is used to transform the non-convex optimization problem into a quasi-convex form, and an exhaustive search algorithm is combined to solve for the optimal energy consumption beam vector, thus achieving joint optimization of the semantic and physical layers under limited communication resources.

[0014] This invention establishes an explicit relationship between semantic communication performance and signal-to-noise ratio (SNR) using the ABG formula, significantly improving the theoretical interpretability of semantic communication systems and providing a reliable analytical foundation for end-to-end semantic performance optimization. In beamforming design, this invention can simultaneously minimize transmit power and energy consumption while meeting semantic communication service quality constraints, thereby effectively improving system stability and reducing energy consumption. By jointly optimizing semantic compression ratio and transmission power, this invention achieves a rational allocation of energy and resources while ensuring semantic communication service quality and transmission delay requirements, further improving the overall system performance. Furthermore, this invention employs optimization strategies such as the Dinkelbach method and matched filtering closed-form solutions, exhibiting low algorithmic complexity and good convergence, making it easy to implement in practical communication systems. The proposed method is particularly suitable for high-bandwidth, high-reliability applications in sixth-generation mobile communication and higher-generation communication systems, effectively improving the robustness and resource utilization of semantic communication systems. Attached Figure Description

[0015] Figure 1 This is a structural block diagram of a multi-input single-output digital semantic communication system model in an embodiment of the present invention.

[0016] Figure 2 This is a block diagram of the semantic encoder structure based on Vision Transformer in this invention.

[0017] Figure 3 This represents the upper limit of semantic communication performance in this invention, measured by multi-scale structural similarity (MS-SSIM) and peak signal-to-noise ratio (PSNR). With compression ratio The simulated curves and fitted curves show the changes; in the figure, (a) is the fitting for multi-scale structural similarity (MS-SSIM) and (b) is the fitting for peak signal-to-noise ratio (PSNR).

[0018] Figure 4 The figures show the simulation and fitting results of the ABG formula for semantic communication performance as a function of signal-to-noise ratio in this invention; (a) is the MS-SSIM fitting of multi-scale structural similarity based on the ABG formula, and (b) is the PSNR fitting based on the ABG formula.

[0019] Figure 5 The image shows a comparison of reconstructed images of the optimal beamforming, fixed-phase beamforming, and random-phase beamforming schemes proposed in this invention on the CIFAR-10 dataset; (a) is the original image, (b) is the optimal beamforming, (c) is the fixed-phase beamforming, and (d) is the random-phase beamforming.

[0020] Figure 6 This invention compares the cumulative distribution functions of the optimal beamforming design based on the ABG formula with those of fixed-phase and random-phase beamforming schemes.

[0021] Figure 7 This is the experimental result of the optimal energy consumption beamforming design in this invention. Figure 8 The semantic communication system of this invention has different maximum transmission delay thresholds under optimal energy consumption beamforming design. A schematic diagram of the cumulative distribution function of transmission quality. Detailed Implementation

[0022] The implementation of the present invention will be further described below with reference to the accompanying drawings: like Figure 1 As shown, this invention relates to a semantic communication system used in an optimal beamforming design method under semantic communication service quality constraints. The system includes a transmitter with multiple transmit antennas and a receiver with a single receive antenna. The transmitter includes a semantic encoder, a quantizer, and a beam optimizer connected in sequence. The receiver includes a channel equalizer, a dequantizer, and a semantic decoder connected in sequence. The detailed structure of the semantic encoder is shown below. Figure 2 As shown, it consists of linear projection, position encoding, Visiontransformer encoding, and a final linear transformation layer, used to extract semantic features; A semantic encoder is used to extract semantic information from the raw data to be sent; A quantizer is used to convert the semantic information of an analog numerical type into a numeric type; The beam optimizer performs beam optimization processing on the quantized semantic information: it constructs the optimal beam with the goal of minimizing transmit power and satisfying the semantic communication service quality constraints, and constructs the optimal beam with the goal of minimizing system energy consumption under the dual constraints of transmission delay and semantic communication service quality; and transmits the beam to the receiver through multiple transmit antennas at the transmitter. A channel equalizer is used to equalize the received signals. A dequantizer is used to restore the equalized signal to the same dimension as the original semantic information, mapping a discrete vector to a continuous feature vector. A semantic decoder is used to reconstruct the original data from signals that have been restored to the same dimensions as the original semantic information.

[0023] The semantic encoder comprises linear projection of image patches, positional encoding, a Vision Transformer model, and linear layers. The Vision Transformer model includes multi-layered normalization and multi-head attention structures. It is used for semantic linear extraction and joint source-channel encoding of the original input image data. The working method of the semantic communication system is as follows: Step 1: The transmitting end uses a semantic encoder to extract the raw data to be transmitted. User semantic information ,in This represents a semantic encoder. This represents the parameters that the semantic encoder can learn; Step 2: Use a quantizer that includes a linear layer and a sign function. Using a quantizer to simulate the semantic information of numerical types Quantized into a digital semantic signal of a specified number of bits , represented as ; Step 3: The transmitting end transmits the quantized digital semantic signal. The signal is transmitted through multiple antennas to a receiver equipped with a single antenna. The signal received by the receiver is: ,in This indicates the conjugate transpose. Indicates channel fading. Represents the beamforming vector. This represents the received additive white Gaussian noise, i.e. ; Step 4: Use the channel state information of the receiver to analyze the received signal. Perform equalization to obtain the equalized signal. ,Right now: ; Step 5: Use a dequantizer consisting of a linear layer. The equalized signal is processed to have the same dimension as the original semantic information, that is: ,in This represents the dequantized semantic information; Step 6: Dequantize the semantic information The input is fed into the semantic decoding module, which is built on the Vision Transformer model. This module performs non-linear mapping and feature reconstruction on the input semantic features to recover the original data information. ,Right now: ,in This represents a semantic decoder. This indicates its learnable parameters.

[0024] This invention proposes an optimal power and optimal energy consumption beamforming design method under semantic communication service quality constraints. The specific steps are as follows: Step 1: By evaluating the performance of the ViT-Transformer semantic encoder on image reconstruction tasks, experiments show that as the semantic compression ratio increases, the end-to-end semantic communication performance first rises rapidly, then approaches its upper limit, and finally plateaus. Based on this, a theoretical model of the relationship between end-to-end semantic performance and compression ratio is proposed. : , in Corresponding to a given compression ratio The theoretical upper limit of semantic coding performance is reflected in the maximum perceptual quality achievable under different compression levels. These are parameters for the semantic encoder. Semantic compression rate affects the quality of semantic communication services. and These represent the quantized bit count and the original bit count of the image, respectively. Using multi-scale structural similarity and peak signal-to-noise ratio as performance metrics for end-to-end semantic communication, the proposed... The model accurately approximates the test curve under different semantic compression rates, demonstrating excellent fitting ability and stability, such as Figure 3 As shown in (a) and (b) in the figure.

[0025] Step 2: Same as model Similar to the previous approach, experiments show that end-to-end semantic performance initially increases rapidly with increasing signal-to-noise ratio, then gradually levels off. Therefore, the ABG formula is proposed. To model the relationship between end-to-end semantic communication performance and signal-to-noise ratio: , in Represents the set of beamforming vectors. For the first The beamforming vector corresponding to the root antenna. This refers to the number of antennas at the transmitter of a multi-input single-output channel. The upper bound of semantic communication performance, parameter Determined by the semantic codec, obtained through fitting experimental data; signal-to-noise ratio is superscript This indicates the conjugate transpose. This represents the channel vectors at the transmitting and receiving ends, where the vector... The elements of the array are independent and identically distributed, following a zero-mean complex Gaussian distribution with variance . ,and This represents the noise power of the additive white Gaussian noise channel. The fitting results are as follows: Figure 4 As shown in (a) and (b), using multi-scale structural similarity and peak signal-to-noise ratio (PSNR) as performance indicators for end-to-end semantic communication, it can be observed that the proposed ABG model can accurately characterize the performance trend with SNR variation. Under different semantic compression rates, the fitting curves are highly consistent with the test results, further verifying the universality and effectiveness of the model for semantic communication systems.

[0026] Step 3: Construct the optimal beamforming design optimization problem that satisfies the semantic communication service quality constraints based on the ABG formula for the relationship between end-to-end semantic performance and signal-to-noise ratio. , in, The threshold representing end-to-end semantic performance.

[0027] Step 4: Solve the optimization problem in Step 3. The semantic performance constraints in the optimization problem in Step 3 are non-convex. Therefore, by converting the service quality constraints of the semantic layer into equivalent physical layer signal-to-noise ratio constraints, the above optimization problem is transformed into: , In the formula This represents the signal-to-noise ratio threshold required to achieve semantic communication performance. This represents the noise power under the square root.

[0028] Step 5: Find the optimal beamforming vector: Based on the matched filtering method, the optimal beamforming vector that satisfies the semantic communication service quality constraints is solved using the following formula. : .

[0029] In practical semantic communication systems with limited communication resources, this study aims to improve system reliability and reduce energy consumption. Based on the obtained optimal beamforming vector, it compares the results with fixed-phase beamforming and random-phase beamforming schemes. Figure 5 Images (a), (b), (c), and (d) in the figure show the original transmitted images and the reconstruction results corresponding to the three beamforming methods. It is clearly observed that the proposed optimal beamforming method significantly outperforms the other two benchmark schemes in terms of visual quality. Furthermore, Figure 6 The cumulative distribution functions of end-to-end semantic communication performance for three beamforming methods are presented. The results show that the optimal beamforming scheme can stably meet the quality of service requirements of semantic communication, while fixed-phase beamforming and random-phase beamforming cannot achieve the same performance level.

[0030] Step 6: The transmission delay from the transmitter to the receiver is expressed as: , in Indicates the original number of bits. Represents the compression ratio of a semantic communication system. Indicates transmission bandwidth; Based on the matched filtering method, the beamforming vector can be generally represented as: , in To represent the transmission power, we substitute the general expression for the beamforming vector into the signal-to-noise ratio (SNR) definition. The SNR is then expressed as: .

[0031] Step 7: Construct an optimal energy consumption beamforming design optimization model that satisfies the semantic communication service quality and transmission delay constraints based on the ABG formula relating end-to-end semantic performance and signal-to-noise ratio. , in For discrete values, Indicates the upper limit of transmission latency. The semantic transmission quality threshold for end-to-end transmission. Total power; The optimal energy consumption beam problem that satisfies the constraints of semantic communication quality of service and transmission delay is solved using an enumeration method: for determining the compression ratio through an exhaustive search method. For each fixed value, the optimization problem is transformed into variables. Optimization subproblems: , The Dinkelbach method was used to divide the variables. The optimization subproblem is transformed into: , in For non-negative update parameters; by separating variables The constraints are restated into the following linear constraint form: , Finally, the optimal solution for separating the variables was obtained using the primal-dual interior-point method. This allows us to obtain the beamforming vector with optimal semantic energy consumption. . Figure 7 The graphs show the relationship between transmission power and transmission delay under different transmission delay constraints. The results show that transmission power gradually decreases as the allowable transmission delay increases, while the system consistently meets the set delay constraints throughout the optimization process. Figure 8 The cumulative distribution function of the proposed optimal energy-consumption beamforming scheme for end-to-end semantic communication performance is further demonstrated under different transmission delay constraints. It can be seen that the proposed method can stably meet the quality of service requirements for semantic communication under all delay conditions.

[0032] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any minor modifications, equivalent substitutions, and improvements made to the above embodiments based on the technical essence of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for optimal beamforming design under semantic communication quality of service constraints, characterized in that, The steps are as follows: S1, using an end-to-end semantic performance and compression rate model, an end-to-end semantic performance and signal-to-noise ratio relationship is established, S2, based on the end-to-end semantic performance and signal-to-noise ratio relationship, an optimal beam is constructed to meet the semantic communication service quality and minimize the transmission power; S3, an optimal beam is constructed to meet the transmission delay and semantic communication service quality constraints while minimizing the semantic energy consumption; S4, through the Dinkelbach method and the matching filter strategy, an optimal beam is designed to meet the transmission delay and semantic communication service quality requirements under the minimum semantic energy consumption target, and the balance between energy consumption and transmission reliability of the semantic communication system is realized.

2. The method of claim 1, wherein, The theoretical relationship model of end-to-end semantic performance and compression rate is represented as: ， wherein the theoretical limit of semantic coding performance corresponding to a given compression ratio the maximum perceptual quality achievable at different compression levels, are semantic encoder parameters, is a semantic compression rate, affecting the quality of semantic communication services, and are the quantized and original bit counts of the image, respectively; The ABG formula of the relationship between end-to-end semantic performance and signal-to-noise ratio is represented as: ， wherein denotes a set of beamforming vectors, is the beamforming vector corresponding to the antenna pair, is the number of antennas at the multiple-input single-output channel transmitter, is the semantic communication performance upper bound, and the parameter is determined by the semantic codec and obtained by fitting experimental data; the signal-to-noise ratio is , and the superscript denotes the conjugate transpose, denotes the channel vectors of the transmitter and the receiver, wherein the elements of the vector are independent and identically distributed and subject to a zero-mean complex Gaussian distribution with a variance of , and is the additive white Gaussian noise channel noise power.

3. The method of claim 2, wherein, Based on the ABG formula of the relationship between end-to-end semantic performance and signal-to-noise ratio, the optimal beamforming design optimization problem that meets the semantic communication service quality constraint is constructed: ， wherein, represents a threshold value for end-to-end semantic performance; The service quality constraint of the semantic layer is converted into an equivalent physical layer signal-to-noise ratio constraint, and the optimal beamforming design optimization problem is converted into: ， ， wherein denotes a signal-to-noise ratio threshold to reach a quality of service for the semantic communication, denotes the square root of the noise power; Based on the matched filter method, the optimal beamforming vector satisfying the semantic communication quality of service constraint is solved by using the following formula : 。 4. The method of claim 3, wherein, The transmission delay from the transmitting end to the receiving end is represented as: ， wherein denotes the original number of bits, represents the compression rate of the semantic communication system, denotes the transmission bandwidth; Based on the matching filter method, the general representation of the beamforming vector is: ， wherein For the transmission power, the general representation of the beamforming vector is brought into the definition of the signal-to-noise ratio, which is then given by ; Based on the ABG formula of the relationship between end-to-end semantic performance and signal-to-noise ratio, the optimal energy consumption beamforming design optimization model that meets the semantic communication service quality and transmission delay constraints is constructed: ， wherein is a discrete value, denotes an upper limit of the transmission delay, is an end-to-end semantic transmission quality threshold value, is the total power; The optimal energy consumption beam problem that satisfies the constraints of semantic communication quality of service and transmission delay is solved using an enumeration method: for determining the compression ratio through an exhaustive search method. For each fixed value, the optimization problem is transformed into variables. The optimization subproblem: ， The optimization sub-problem of variables is converted to: ， wherein are non-negative update parameters; by separation of variables The constraint condition is reformulated as the following linear constraint form: ， Finally, the optimal solution of the separated variables is obtained by using the primal-dual interior point method , and then the beamforming vector of the optimal semantic energy consumption is obtained .

5. A semantic communication system for use in a method of optimal beamforming design under semantic communication quality of service constraints as claimed in claim 1, characterized by, The transmitting end includes a semantic encoder, a quantizer, and a beam optimizer connected in sequence, and the receiving end includes a channel equalizer, a dequantizer, and a semantic decoder connected in sequence; The semantic encoder is used to extract semantic information from the original data to be transmitted; The quantizer is used to convert the analog numerical type semantic information into a digital type; The beam optimizer performs beam optimization processing on the quantized semantic information: constructs an optimal beam to meet the semantic communication service quality constraint and minimize the transmission power, and constructs an optimal beam to minimize the system energy consumption under the dual constraints of transmission delay and semantic communication service quality; Through the multiple transmitting antennas of the transmitting end, it is sent to the receiving end; The channel equalizer is used to equalize the received signal; The dequantizer is used to restore the equalized signal to the same dimension as the original semantic information, and map the discrete vector to a continuous feature vector; The semantic decoder is used to reconstruct the original data from the signal restored to the same dimension as the original semantic information.

6. The semantic communication system of claim 1, wherein, The semantic encoder includes linear projection, position encoding, Vision Transformer model, and linear layer on image blocks, wherein the Vision Transformer model includes multiple layers of normalization and multi-head attention structure; used for semantic linear extraction and joint source channel coding on original input image data.

7. A method of operating a semantic communication system as claimed in claim 5, characterized by The steps are as follows: Step 1, the sender extracts the user semantic information in the original data to be transmitted by using a semantic encoder wherein denotes the semantic encoder, denotes parameters that the semantic encoder can learn;​ Step 2, using a quantizer comprising a linear layer and a sign function , utilizing the quantizer to quantize semantic information of an analog numeric type into a digital semantic signal of a specified number of bits , represented as ; Step 3, the transmitter sends the quantized digital semantic signal Through multiple antennas to the receiving end equipped with a single antenna, the signal received by the receiving end is: where denotes the conjugate transpose, denotes the channel fading, denotes the beamforming vector, denotes the received additive white Gaussian noise, that is ; Step 4, equalizing the received signal using the channel state information of the receiving end to obtain an equalized signal i.e. ; Step 5, dequantizer consisting of a linear layer The equalized signal is processed to the same dimension as the original semantic information, i.e. wherein denotes the dequantized semantic information; Step 6, the dequantized semantic information to a semantic decoding module, which is constructed based on a Vision Transformer model, for performing nonlinear mapping and feature reconstruction on the input semantic features, so as to recover the original data information , that is, , wherein denotes a semantic decoder, denotes the learnable parameters thereof.

Images