Semantic communication method based on intelligent reflecting surface assistance

Abstract

The invention discloses a semantic communication method based on intelligent reflecting surface assistance, which comprises the following steps: sending a semantic signal to STAR-RIS and a legal user in a multicast mode through a base station, and reflecting and transmitting the received signal to the legal user by the STAR-RIS; an eavesdropper eavesdrops the signal of the legal user; measuring the similarity between the original text information and the semantic information by using the bilingual evaluation score to obtain a lower bound of a semantic extraction rate; establishing an optimization problem on the basis of considering power consumption; and decomposing the optimization problem into four sub-problems, carrying out iterative solution by utilizing an alternate optimization iterative algorithm, and carrying out confidential transmission of semantic information at an optimal confidentiality rate obtained under an optimal solution. According to the method, the problem of performance optimization of safety transmission of the semantic communication system assisted by the STAR-RIS technology under the condition that hardware equipment is damaged is researched, the semantic secrecy rate is maximized by utilizing an iterative algorithm of alternate optimization, the safety performance of safety transmission of semantic communication can be effectively improved, and the method has a wide application prospect.

Description

Technical Field

[0001] The present invention relates to the technical field of 6G wireless communication networks, and in particular to a semantic communication method based on the assistance of intelligent reflective surfaces. Background Art

[0002] In recent years, with the rapid growth of the demand for intelligent wireless communications, emerging intelligent services based on wireless communication technologies such as industrial Internet, virtual reality and holographic projection have emerged. However, due to the emergence of these intelligent services, the demand for data transmission rate will reach a new peak, so traditional wireless communication networks are facing serious capacity and bandwidth challenges. Semantic communication is an intelligent communication method. Unlike traditional communication, semantic communication relies on knowledge sharing and semantic understanding between the two parties in communication, which can reduce the amount of transmitted information and reduce communication bandwidth requirements. It is expected to become a promising solution for future wireless communication networks. However, the broadcast and open characteristics of wireless channels bring convenience to information transmission to legitimate users, but also provide illegal users with opportunities to eavesdrop on information, which brings potential security risks to semantic communication transmission. Therefore, the security of information transmission is still a key issue in the field of semantic communication.

[0003] Reconfigurable Intelligent Surface (RIS) is a wireless communication auxiliary technology with high spectrum efficiency, high energy efficiency and low cost, and has broad application prospects in wireless communication transmission security. It can constructively reconstruct signals to enhance the desired signal while reducing eavesdropping capabilities. RIS is a plane composed of a large number of low-cost passive reflective elements, each of which can independently sense the amplitude and phase changes of the incident signal to achieve controllable signal transmission. As a disruptive technology in the sixth generation (6G) wireless communication technology, RIS fundamentally solves the problems of wireless channel fading and interference by manipulating the wireless environment, thereby improving communication performance.

[0004] However, the inherent half-space coverage of traditional RIS limits its application flexibility. The emergence of simultaneous transmission and reflection RIS (STAR-RIS) breaks this limitation. By splitting the incident signal into transmitted and reflected signals, STAR-RIS can realize a full-space intelligent radio environment. At present, most of the research on STAR-RIS-assisted secure communication assumes that the transceiver has perfect hardware. However, for actual communication systems, hardware imperfections are always inevitable, including phase noise, quantization error, amplification noise and nonlinear factors. The distorted noise caused by transceiver damage will interfere with users and cause network performance degradation. Therefore, it is necessary to study the secure transmission of STAR-RIS-assisted semantic communication systems under hardware damage conditions to provide secure and reliable transmission. Summary of the invention

[0005] In view of the above-mentioned deficiencies in the prior art, the present invention provides a semantic communication method based on the assistance of an intelligent reflective surface, which solves the problem of high risk of secure transmission of semantic signals under hardware damage conditions.

[0006] In order to achieve the above-mentioned invention object, the technical solution adopted by the present invention is: a semantic communication method based on the assistance of intelligent reflective surface, comprising the following steps:

[0007] S1, the semantic signal is sent to STAR-RIS, the legitimate user R and the legitimate user T by the base station in a multicast transmission mode; the received semantic signal is reflected to the legitimate user R and transmitted to the legitimate user T by STAR-RIS in a time switching protocol;

[0008] S2, according to the semantic signal received by the legitimate user R, the legitimate user T and the eavesdropper E R , Eavesdropper E T The received semantic signal calculates the semantic confidentiality rate; where the eavesdropper E R is an eavesdropper of the legitimate user R; eavesdropper E T is an eavesdropper of the legitimate user T;

[0009] S3, using the bilingual evaluation substitute score to measure the similarity between the original text information and the semantic information before the base station sends it, and then obtain the lower bound of the semantic extraction rate of the legitimate user;

[0010] S4. On the basis of considering power consumption, an optimization problem is established by jointly optimizing the active beamforming of the base station, the passive beamforming of STAR-RIS, the time switching coefficient and the semantic extraction rate;

[0011] S5. Decompose the optimization problem into four interdependent sub-problems, and use an iterative algorithm based on alternating optimization to solve the four interdependent sub-problems step by step to obtain the optimal solution of the optimization problem. Use the optimal confidentiality rate obtained under the optimal solution to perform confidential transmission of semantic information.

[0012] The beneficial effects of the present invention are:

[0013] 1. The present invention introduces STAR-RIS technology into secure semantic communication. STAR-RIS can transmit and reflect incident signals at the same time to achieve full spatial coverage. By reconfiguring the corresponding transmission and reflection coefficients, the transmission and reflection signals of STAR-RIS elements can be controlled, and the wireless channel environment can be reconstructed to optimize the channel quality, thereby achieving safer information transmission.

[0014] 2. In order to solve the non-convex optimization problem, the present invention also proposes an efficient iterative algorithm based on alternating optimization (AO), and adopts continuous convex approximation, semidefinite relaxation and Gaussian randomization methods to process the corresponding sub-problems to obtain a feasible optimal solution, thereby maximizing the semantic confidentiality rate. The processing process has greater flexibility and can achieve faster convergence.

[0015] 3. The framework structure of the present invention is simple, the algorithm is efficient, and it has rigorous theoretical guarantees and experimental verification. In addition, compared with traditional bit-based information transmission schemes and non-optimized schemes, the proposed scheme can achieve better system performance gains in the secure transmission of semantic information. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a communication system model based on the assistance of intelligent reflective surface;

[0017] Figure 2 A flowchart of a semantic communication method based on intelligent reflective surface assistance proposed by the present invention;

[0018] Figure 3 This is a comparison effect diagram of the relationship between the semantic confidentiality rate and the number of iterations of the solution proposed in the present invention under different numbers of base station antennas and different numbers of STAR-RIS elements;

[0019] Figure 4 This is a comparison diagram of the relationship between the semantic confidentiality rate and the transmission power budget of different schemes under different hardware damage levels;

[0020] Figure 5 This is a comparison effect diagram of the relationship between the semantic confidentiality rate and the transmission power budget of the solution proposed in the present invention under different hardware damage levels;

[0021] Figure 6 This is a comparison diagram of the relationship between the semantic confidentiality rate and the number of base station antennas under different schemes;

[0022] Figure 7 This is a comparison diagram of the relationship between the semantic confidentiality rate and the number of STAR-RIS elements under different schemes. DETAILED DESCRIPTION

[0023] The specific implementation modes of the present invention are described below so that those skilled in the art can understand the present invention. However, it should be clear that the present invention is not limited to the scope of the specific implementation modes. For those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations utilizing the concept of the present invention are protected.

[0024] like Figure 1As shown in the figure, a semantic communication method based on intelligent reflective surface assistance proposed in the present invention is applied to a communication system model based on intelligent reflective surface assistance: including a base station (BS) equipped with M antennas, a passive intelligent omnidirectional surface (STAR-RIS) with N intelligent unit elements capable of synchronous reflection and transmission, and a legitimate user R equipped with a single antenna, an eavesdropper E R , legitimate user T and eavesdropper E T ; Legitimate user R and eavesdropper E R Located on the reflective surface of STAR-RIS; legitimate user T and eavesdropper E T Located on the transmission surface of STAR-RIS. STAR-RIS can dynamically adjust its own transmission coefficient (including amplitude and phase shift) and reflection coefficient (including amplitude and phase shift) to intelligently reconstruct the wireless communication channel environment, thereby realizing a highly flexible full-space intelligent radio environment, which can further enhance the information security transmission of the semantic communication system. In this embodiment, the impact of hardware damage on semantic communication is considered.

[0025] like Figure 2 As shown, the semantic communication method based on intelligent reflective surface assistance proposed in the present invention includes the following steps:

[0026] Step S1: Send the semantic signal to STAR-RIS, the legal user R and the legal user T in a multicast transmission mode through the base station; reflect the received semantic signal to the legal user R and transmit it to the legal user T through STAR-RIS in a time switching protocol;

[0027] The time switching coefficients of the time switching protocol used by STAR-RIS are τ T and τ R , and satisfies τ T +τ R =1; the transmission matrix and reflection matrix of STAR-RIS are and in η represents the set η={1,2,...,N}, N is the total number of elements in STAR-RIS; π represents 180 degrees in radians; diag(.) represents the diag function; under the time switching protocol, the semantic signal sent by the base station to the legitimate user u is represented by x u =w u s u +z u , where s u Represents semantic information and follows s u The expected value of power is 1, that is, E[|s u | 2 ]=1;w uis the beamforming vector; z u is an independent Gaussian transmission distortion noise; the Gaussian transmission distortion noise power at each antenna is proportional to its transmission signal power, that is, z u ~CN(0,κ t Diag(w u w u H )); where κ t is the ratio of the Gaussian transmission distortion noise power to the transmission signal power, i.e., the distortion caused by hardware damage at the base station, κ t ≥0; (·) H Represents the conjugate transpose of a matrix.

[0028] Step S2: Based on the semantic signals received by the legitimate user R and the legitimate user T and the eavesdropper E R , Eavesdropper E T The received semantic signal calculates the semantic confidentiality rate; where the eavesdropper E R is an eavesdropper of the legitimate user R; eavesdropper E T is an eavesdropper of the legitimate user T;

[0029] The semantic signal received by the legitimate user R and the legitimate user T and the eavesdropper E R , Eavesdropper E T The received semantic signal calculates the semantic confidentiality rate, which specifically includes the following sub-steps:

[0030] A1. Get the legitimate user u and the eavesdropper E u The semantic signals received at are expressed as follows:

[0031]

[0032] in, G. They represent the channel from the base station to STAR-RIS, the channel from the base station to the legitimate user u, and the channel from the base station to the eavesdropper E. u Channel; They represent the channel from STAR-RIS to the legitimate user u and the channel from STAR-RIS to the eavesdropper E. u Channel; n u and Represent the legitimate user u and the eavesdropper E respectively u The additive Gaussian white noise at is, and its noise variance is m u and They are legitimate user u and eavesdropper E respectively u The independent zero-mean Gaussian distortion noise at is distributed as follows: and κ r and κ e They are the legitimate user u and the eavesdropper E u The ratio of the distorted noise power to the undistorted received signal power at the location, that is, the power difference between the legitimate user u and the eavesdropper E u The distortion caused by hardware damage, κ r ,κ e ≥0; where Respectively expressed as:

[0033]

[0034] A2. Calculate the semantic extraction rate ρ of the legitimate user u u and semantic transmission rate S u , respectively expressed as:

[0035]

[0036] Where 0≤ρ u ≤1, the function len(·) represents the number of bits of information, c u Indicates the original text information before the base station sends it. u Represents semantic information, semantic information s u It is the original text information c before sending semantic signals u Generated by the autoencoder; γ u is the receiving signal-to-noise ratio of user u; u The expression is:

[0037]

[0038] in is the equivalent reflection coefficient vector, is an equivalent channel;

[0039] Correspondingly, the eavesdropper E u The received signal-to-noise ratio at is:

[0040]

[0041] in,

[0042] A3. Computation time switching protocol between legitimate user u and eavesdropper E u The achievable semantic transmission rate is expressed as:

[0043]

[0044] Where R u =log 2 (1+γu ), SS u is the achievable semantic transmission rate of legitimate user u under the time-switching protocol; is the eavesdropper E in the time-switching protocol u Achievable semantic transfer rate;

[0045] A4, according to the legitimate user u and the eavesdropper E u The achievable semantic transmission rate can be used to obtain the semantic confidentiality rate of the legitimate user u:

[0046]

[0047] Where [a] + =max(a,0);SS u is the achievable semantic transmission rate of legitimate user u under the time-switching protocol; is the eavesdropper E in the time-switching protocol u The achievable semantic transfer rate.

[0048] Step S3: Use the Bilingual Evaluation Underscore (BLEU) to measure the similarity between the original text information before the base station sends it and the semantic information after compression and extraction before sending it, and then obtain the lower bound of the semantic extraction rate of the legitimate user.

[0049] The expression of the bilingual assessment substitute score is:

[0050]

[0051] where s u The length is less than or equal to c u The length of BP is the short penalty factor; w g is the weight corresponding to different g-gram measurement rules; p g is the corresponding precision score under different g-gram measurement rules; w g,u represents the weight corresponding to the use of different g-gram measurement rules by legitimate user u; p g,u represents the accuracy score of the legitimate user u under different g-gram measurement rules; g-gram refers to a sequence of g words that appear consecutively in a text; G represents the number of g-grams generated; e is a natural constant; exp is an exponential with the natural constant as the base;

[0052] According to the bilingual evaluation substitute score of the legitimate user u, the semantic extraction rate ρ of the legitimate user u is u The lower bound It is expressed as:

[0053]

[0054] Where Q is BLEU u The lower limit of , ln represents the natural logarithm.

[0055] Step S4: On the basis of considering power consumption, an optimization problem is established by jointly optimizing the active beamforming of the base station, the passive beamforming of STAR-RIS, the time switching coefficient and the semantic extraction rate.

[0056] The power consumption includes the computational power consumption during the semantic signal transmission process and the total communication power consumption;

[0057] The computational power consumption during the semantic signal transmission process is P comp It is expressed as:

[0058]

[0059] Where F is a factor that converts amplitude to power, Indicates the maximum computing power consumption;

[0060] Total communication power consumption P trans for:

[0061]

[0062] in That is, the base station's transmission power P to the legitimate user u u , P u ≤ξ, ξ is the transmit power budget of the base station to the legitimate user u; Represents the square of the Euclidean norm of the vector. Total communication power consumption P trans It is also limited by the base station's transmit power budget, namely P trans ≤ξ.

[0063] The optimization problem is expressed as:

[0064]

[0065] Among them, P1 represents the optimization problem; st represents the constraint condition, C1.1 is the total communication power consumption budget constraint of the base station within the allocated communication time; C1.2 is the maximum transmission power constraint of the base station; C1.3 is the computational power consumption constraint during the semantic signal transmission process; C1.4 provides the boundary of the semantic extraction rate; C1.5 is the constraint on v u Each entry in imposes a unit modulus constraint, and the set is C1.6 is a constraint condition unique to the time switching protocol mode; C1.7 constrains the user's communication time to ensure that each user can obtain a certain amount of communication time, where τ th Indicates the minimum threshold of communication time.

[0066] Step S5: In order to solve the non-convex optimization problem P1, P1 is decoupled into four sub-problems, and then the four interdependent sub-problems are solved step by step using an iterative algorithm based on alternating optimization to obtain the optimal solution of the optimization problem. The optimal confidentiality rate obtained under the optimal solution is used to perform confidential transmission of semantic information.

[0067] The optimization problem is decomposed into four interdependent sub-problems and solved step by step using an iterative algorithm based on alternating optimization, which includes the following sub-steps:

[0068] F1, initialize the iteration parameter to l = 1;

[0069] F2, in the given In the case of , solve problem P2 and get

[0070] F3, in the given In the case of , solving problem P3, we get

[0071] F4, in the given In the case of , solving problem P4, we get

[0072] F5, in the given In the case of , solving problem P5, we get

[0073] F6, add 1 to the current iteration parameter;

[0074] F7. Judgment Is it less than or equal to the iteration precision ∈ or is it equal to the maximum number of iterations l max , when any one of the judgments is established, the optimal solution finally obtained is substituted into the semantic confidentiality rate calculation expression to obtain the optimal semantic confidentiality rate for confidential transmission of semantic information, otherwise, return to step F2;

[0075] in w u The value of the lth iteration; v u The value of the lth iteration; is τ u The value of the lth iteration; is u The value of the lth iteration; F(·) is the optimization problem P1.

[0076] Each iteration includes the following steps:

[0077] B1, fixed w u 、v u and τ u , optimize ρu :when When given, the optimization problem P1 is transformed into problem P2:

[0078]

[0079] From question P2, we can see that the smaller the semantic extraction rate, the higher the semantic confidentiality rate. Therefore, under the premise of satisfying the C2.2 constraint, the semantic extraction rate of each user should be set as small as possible, which is more efficient. This problem can be solved by the dual method to obtain the optimal semantic extraction rate.

[0080] B2. Fixed ρ u 、v u and τ u , optimize w u :when When given, by introducing the slack variable a u =[a 1,u ,a 2,u ,a 3,u ,a 4,u ] T and b u =[b 1,u ,b 2,u ,b 3,u ,b 4,u ] T , transform the optimization problem P1 into problem P3:

[0081]

[0082] in B 1,u =κ r P u +(1+κ r )κ t Diag(P u ), B 2,u =(1+κ r ) u +(1+κ r )κ t Diag(P u ), R u =a 1,u -a 2,u ≤R u and W u represents the optimal beamforming matrix; W u ≥0 means matrix W uis semi-positive definite; Tr(·) represents the trace of the matrix; Rank(·) represents the rank of the matrix; ^ represents the estimated value; constraint C3.4 is eliminated by the semi-positive definite relaxation method, and problem P3 is further relaxed into a convex semi-definite programming, and the convex optimization tool CVX is used to solve problem P3; after satisfying Rank(W u )≤1 u After that, the eigenvalue decomposition is performed to obtain the optimal beamforming vector

[0083] B3. Fixed ρ u ,w u and τ u , optimize v u :when When given, by defining a new variable And introduce the slack variable c u =[c 1,u ,c 2,u ,c 3,u ,c 4,u ] T and d u =[d 1,u ,d 2,u ,d 3,u ,d 4,u ] T , transform the optimization problem P1 into problem P4:

[0084]

[0085] in The convex optimization tool CVX is used to solve problem P4, and Gaussian randomization technology is used to select the optimal solution V u The suboptimal solution q is obtained u , which can maximize the objective value of problem P1; u Projection onto the constraint set χ u On, that is Get the optimal solution in ∠(x) represents the angle of x, [x] a represents the a-th entry of vector x;

[0086] B4. Fixed ρ u ,w u and v u , optimize τ u :when When given, the optimization problem P1 is transformed into P5:

[0087]

[0088] Obviously, problem P5 is a convex problem. We can use the convex optimization tool CVX to solve problem P5 and obtain the optimal time switching coefficient.

[0089] In general, based on the optimization problem P1, the present invention proposes an alternating optimization iterative algorithm to optimize the four sub-problems of problems P2, P3, P4, and P5, and uses the optimal solution of the sub-problem obtained each time as the input of the next sub-problem to solve the next sub-problem, and finally obtains the optimal solution to problem P1 to obtain the optimal semantic confidentiality rate.

[0090] In one embodiment of the present invention, Figure 3 As shown in Figure 2, under the same transmit power budget, the semantic confidentiality rate corresponding to different numbers of BS antennas and STAR-RIS elements increases rapidly with the increase of iteration number and finally converges. In addition, the proposed method has good convergence, whether increasing the number of BS antennas and STAR-RIS elements or increasing the hardware damage κ i ,i∈{t,r,e}, and can achieve rapid convergence after several iterations.

[0091] Depend on Figure 4 It can be seen that under different hardware damage levels, as the transmit power budget increases, the semantic confidentiality rates of all schemes increase first fast and then slowly. i =0.05, i∈{t,r,e}, as the transmit power budget increases, the semantic confidentiality rate gradually tends to be stable, because the distortion noise caused by hardware damage is proportional to the signal power of the transceiver, so the semantic confidentiality rate has an upper limit. Compared with other benchmark schemes, the beamforming gain of this method makes the proposed scheme have a significant advantage. In addition, as κ i As ,i∈{t,r,e} increases, the user's distortion noise power will also increase, resulting in a decrease in the semantic confidentiality rate. Therefore, the upper limit of the semantic confidentiality rate is determined by the degree of hardware damage. As κ i ,i∈{t,r,e} decreases, the semantic confidentiality rate and its upper limit also increase. It can be seen that whether it is the random phase shift scheme or the scheme without RIS, the semantic confidentiality rate achieved under the same conditions is not as high as that of the proposed scheme, which is the semantic communication method based on the assistance of intelligent reflective surface proposed in the present invention.

[0092] Depend on Figure 5 It can be seen that when the transmit power budget is set to 15dBm, by comparing {κ t =0.02,κ r =0.01,κ e =0.01} and {κ t =0.01,κ r =0.02,κ e=0.01}, it can be seen that the hardware damage at the legitimate user has a greater negative impact on the confidentiality rate than the hardware damage at the BS. This is because the proportion of the received distortion noise power in the total received noise power is much larger than the transmitted distortion noise power.

[0093] Figure 6 The transmit power budget of each scheme is set to 15dBm, and the hardware damage level is set to {κ t =0.01,κ r =0.01,κ e =0.01}, as the number of BS antennas increases, the spatial degrees of freedom also increase, which brings more beam gain to the system, so the semantic confidentiality rate of the system also increases with the increase of M. However, among the three schemes, no matter how the number of BS antennas increases, the scheme with the highest semantic confidentiality rate is still the proposed scheme proposed in this invention.

[0094] Figure 7 The transmit budget power of each scheme is set to 15dBm, and the hardware damage level is set to {κ t =0.01,κ r =0.01,κ e =0.01}, as the number of STAR-RIS elements increases, the semantic confidentiality rate of the scheme without RIS deployment remains unchanged, while the semantic confidentiality rates of the proposed scheme and the random phase shift scheme gradually increase. This is because increasing N will cause more STAR-RIS elements to receive signal energy from the BS, thereby achieving a higher spatial diversity gain. It can be seen that the semantic confidentiality rate achieved by the scheme proposed in the present invention is significantly improved as the number of STAR-RIS elements increases.

[0095] Furthermore, increasing M and N can achieve precise beamforming without incurring additional transmit power overhead, thus significantly improving system performance.

[0096] In summary, compared with other existing technologies, the semantic communication method based on the assistance of intelligent reflective surface proposed in the present invention shows excellent performance in the final semantic confidentiality rate under the influence of hardware damage degree, number of iterations, different numbers of base station antennas and different numbers of STAR-RIS elements. It can realize more secure semantic communication, reduce communication risks, and is more suitable for 6G communication.

Claims

1. A semantic communication method based on intelligent reflective surface assistance, characterized in that: The following steps are involved: S1, the semantic signal is sent to STAR-RIS, the legitimate user R and the legitimate user T by the base station in a multicast transmission mode; the received semantic signal is reflected to the legitimate user R and transmitted to the legitimate user T by STAR-RIS in a time switching protocol; S2, according to the semantic signal received by the legitimate user R, the legitimate user T and the eavesdropper E R , Eavesdropper E T The received semantic signal calculates the semantic confidentiality rate; where the eavesdropper E R is an eavesdropper of the legitimate user R; eavesdropper E T is an eavesdropper of the legitimate user T; S3, using the bilingual evaluation substitute score to measure the similarity between the original text information and the semantic information before the base station sends it, and then obtain the lower bound of the semantic extraction rate of the legitimate user; S4. On the basis of considering power consumption, an optimization problem is established by jointly optimizing the active beamforming of the base station, the passive beamforming of STAR-RIS, the time switching coefficient and the semantic extraction rate; S5. Decompose the optimization problem into four interdependent sub-problems, and use an iterative algorithm based on alternating optimization to solve the four interdependent sub-problems step by step to obtain the optimal solution of the optimization problem. Use the optimal confidentiality rate obtained under the optimal solution to perform confidential transmission of semantic information.

2. The method according to claim 1, characterized in that The time switching coefficients of the time switching protocol adopted by STAR-RIS in step S1 are τ T and τ R , and satisfies τ T +τ R =1; the transmission matrix and reflection matrix of STAR-RIS are and in η represents the set η={1,2,...,N}, N is the total number of elements in STAR-RIS; π represents 180 degrees in radians; j represents an imaginary unit; diag(.) represents the diag function; under the time switching protocol, the semantic signal sent by the base station to the legitimate user u is represented by x u =w u s u +z u , where s u Represents semantic information and follows s u The expected value of power is 1, that is, E[|s u | 2 ]=1;w u is the beamforming vector; z u is an independent Gaussian transmission distortion noise; the Gaussian transmission distortion noise power at each antenna is proportional to its transmission signal power, that is, z u ~CN(0,κ t Diag(w u w u H )); where κ t is the ratio of the Gaussian transmission distortion noise power to the transmission signal power, i.e., the distortion caused by hardware damage at the base station, κ t ≥0; (·) H Represents the conjugate transpose of a matrix.

3. The method according to claim 2, characterized in that In step S2, the semantic signal received by the legitimate user R and the legitimate user T and the eavesdropper E R , Eavesdropper E T The received semantic signal calculates the semantic confidentiality rate, which specifically includes the following sub-steps: A1. Get the legitimate user u and the eavesdropper E u The semantic signals received at are expressed as follows: in, G. They represent the channel from the base station to STAR-RIS, the channel from the base station to the legitimate user u, and the channel from the base station to the eavesdropper E. u Channel; They represent the channel from STAR-RIS to the legitimate user u and the channel from STAR-RIS to the eavesdropper E. u Channel; n u and Represent the legitimate user u and the eavesdropper E respectively u The additive Gaussian white noise at is, and its noise variance is and m u and They are legitimate user u and eavesdropper E respectively u The independent zero-mean Gaussian distortion noise at is distributed as follows: and κ r and κ e They are the legitimate user u and the eavesdropper E u The ratio of the distorted noise power to the undistorted received signal power at the location, that is, the power difference between the legitimate user u and the eavesdropper E u The distortion caused by hardware damage, κ r ,κ e ≥0; where and Respectively expressed as: A2. Calculate the semantic extraction rate ρ of the legitimate user u u and semantic transmission rate S u , respectively expressed as: Where 0≤ρ u ≤1, the function len(·) represents the number of bits of information, c u Indicates the original text information before the base station sends it, s u Represents semantic information, γ u is the received signal-to-noise ratio of the legitimate user u; u The expression is: in is the equivalent reflection coefficient vector, is an equivalent channel; Correspondingly, the eavesdropper E u The received signal-to-noise ratio at is: in, A3. Computation time switching protocol between legitimate user u and eavesdropper E u The achievable semantic transmission rate is expressed as: Where Ru = log2(1+γu), SS u is the achievable semantic transmission rate of legitimate user u under the time-switching protocol; is the eavesdropper E in the time-switching protocol u Achievable semantic transfer rate; A4, according to the legitimate user u and the eavesdropper E u The achievable semantic transmission rate can be used to obtain the semantic confidentiality rate of the legitimate user u: Where [a] + =max(a,0).

4. The method according to claim 3, characterized in that The expression of the bilingual assessment substitute score in step S3 is: where s u The length is less than or equal to c u length, BP is the short penalty factor; w g It is the weight corresponding to different g-gram measurement rules; p g is the corresponding accuracy score under different g-gram measurement rules; w g,u represents the weight corresponding to the use of different g-gram measurement rules by legitimate user u; p g,u represents the accuracy score of the legitimate user u under different g-gram measurement rules; g-gram refers to a sequence of g words that appear consecutively in a text; G represents the number of g-grams generated; e is a natural constant; exp is an exponential with the natural constant as the base; According to the bilingual evaluation substitute score of the legitimate user u, the semantic extraction rate ρ of the legitimate user u is u The lower bound It is expressed as: Where Q is BLEU u The lower limit of , ln represents the natural logarithm.

5. The method according to claim 4, characterized in that The power consumption in step S4 includes the computing power consumption and the total communication power consumption during the semantic signal transmission process; The computational power consumption during the semantic signal transmission process is P comp It is expressed as: Where F is a factor that converts amplitude to power, Indicates the maximum computing power consumption; Total communication power consumption P trans It is expressed as: in That is, the base station's transmission power P to the legitimate user u u , P u ≤ξ, ξ is the transmit power budget of the base station to the legitimate user u; Represents the square of the Euclidean norm of a vector.

6. The method according to claim 5, characterized in that The optimization problem in step S4 is expressed as: s.t.C1.1:P trans ≤ξ C1.6:t T +t R =1 Among them, P1 represents the optimization problem; st represents the constraint condition, C1.1 is the total communication power consumption budget constraint of the base station within the allocated communication time; C1.2 is the maximum transmission power constraint of the base station; C1.3 is the computational power consumption constraint during the semantic signal transmission process; C1.4 provides the boundary of the semantic extraction rate; C1.5 is the constraint on v u Each entry in imposes a unit modulus constraint, and the set is C1.6 is a constraint condition unique to the time switching protocol mode; C1.7 constrains the user's communication time, where τ th Indicates the minimum threshold of communication time.

7. The method according to claim 6, characterized in that In step S5, the iterative algorithm based on alternating optimization is used to solve the four interdependent sub-problems step by step. The specific method for obtaining the optimal solution of the optimization problem includes the following steps: F1, initialize the iteration parameter to l = 1; F2, in the given and In the case of , solve problem P2 and get F3, in the given and In the case of , solving problem P3, we get F4, in the given and In the case of , solve problem P4 and get F5, in the given and In the case of , solving problem P5, we get F6, add 1 to the current iteration parameter; F7. Judgment Is it less than or equal to the iteration precision ∈ or is it equal to the maximum number of iterations l max , when any one of the judgments is established, the optimal solution finally obtained is substituted into the semantic confidentiality rate calculation expression to obtain the optimal semantic confidentiality rate for confidential transmission of semantic information, otherwise, return to step F2; in w u The value of the lth iteration; v u The value of the lth iteration; is τ u The value of the lth iteration; is u The value of the lth iteration; F(·) is the optimization problem P1.

8. The method according to claim 7, characterized in that In each iteration the following steps are included: B1, fixed w u 、v u and τ u , optimize ρ u :when and When given, the optimization problem P1 is transformed into problem P2: Solve problem P2 by dual method and get the optimal semantic extraction rate B2. Fixed ρ u 、v u and τ u , optimize w u :when and When given, by introducing the slack variable a u =[a 1,u ,a 2,u ,a 3,u ,a 4,u ] T and b u =[b 1,u ,b 2,u ,b 3,u ,b 4,u ] T , transform the optimization problem P1 into problem P3: among them B 1,u =k r P u +(1+k r )k t Diag(P u ), B 2,u =(1+k) r )P u +(1+k r )k t Diag(P u ), R u =a 1,u -a 2,u ≤R u and W u represents the optimal beamforming matrix; W u ≥0 means matrix W u is semi-positive definite; Tr(·) represents the trace of the matrix; Rank(·) represents the rank of the matrix; ^ represents the estimated value; constraint C3.4 is eliminated by the semi-positive definite relaxation method, and problem P3 is further relaxed into a convex semi-definite programming, and the convex optimization tool CVX is used to solve problem P3; after satisfying Rank(W u )≤1, and then perform eigenvalue decomposition to obtain the optimal beamforming vector B3. Fixed ρ u ,w u and τ u , optimize v u :when and When given, by defining a new variable And introduce the slack variable c u =[c 1,u ,c 2,u ,c 3,u ,c 4,u ] T and d u =[d 1,u ,d 2,u ,d 3,u ,d 4,u ] T , transform the optimization problem P1 into problem P4: in ′ R u =c 1,u -c 2,u ≤R u 、 The convex optimization tool CVX is used to solve problem P4, and Gaussian randomization technology is used to select the optimal solution V u The suboptimal solution q is obtained u ; q u Projection onto the constraint set χ u On, that is Get the optimal solution in ∠(x) represents the angle of x, [x] a represents the a-th entry of vector x; B4. Fixed ρ u ,w u and v u , optimize τ u :when and When given, the optimization problem P1 is transformed into P5: s.t.C5.1:P trans ≤ξ C5.2:t T +t R =1 Use the convex optimization tool CVX to solve problem P5 and obtain the optimal time switching coefficient

Images