A hybrid spectrum efficiency optimization method and device for a knowledge mismatch scenario

Abstract

The application provides a hybrid spectrum efficiency optimization method and device for a knowledge mismatching scene, adjusts the contribution proportion of semantic communication and bit communication in the total spectrum efficiency for each user in a parallel hybrid semantic-bit communication system through semantic participation degree; fits the mapping relationship of the semantic similarity changing with the signal-to-interference-and-noise ratio under different knowledge matching degrees through a data regression method; fits the influence of the knowledge matching degree on the semantic similarity through a generalized logic function; takes the semantic participation degree and a user power allocation strategy as joint optimization variables, constructs an optimization problem with the maximum system total spectrum efficiency as the target under the base station total transmission power constraint, the user maximum transmission delay constraint and the power allocation fairness constraint condition; decomposes and solves the optimization problem through an alternating optimization strategy and a double-layer meta-heuristic selection optimization algorithm; dynamically cooperates transmission according to the optimal solution; and realizes flexible adaptation to the knowledge mismatching and the time-varying channel environment, and significantly improves the system total spectrum efficiency.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and in particular to a hybrid spectrum efficiency optimization method and apparatus for knowledge mismatch scenarios. Background Technology

[0002] As 6G networks expand into high-density services such as autonomous driving, smart industry, and immersive interaction, the massive number of terminals accessing the network causes wireless data volume to grow exponentially, further exacerbating the scarcity of spectrum resources.

[0003] Meanwhile, traditional communication methods centered on precise bit transmission are approaching the Shannon limit within given bandwidth, making it difficult to further improve spectral efficiency in scenarios with multi-user concurrency and high-redundancy data. As a novel paradigm, semantic communication, by extracting task-related semantic information instead of bit transmission, can significantly improve the effective information rate. However, its performance is highly dependent on the consistency of the background knowledge bases of the communicating parties; knowledge mismatch introduces semantic noise, leading to a decrease in semantic similarity and thus weakening spectral efficiency. More importantly, in high signal-to-noise ratio regions, the spectral efficiency of bit communication can still improve with channel quality, while semantic communication, due to the saturation of semantic similarity, struggles to gain further advantages, thus limiting its advantages.

[0004] Therefore, hybrid semantic-bit communication has become a feasible approach to address the aforementioned problems. However, existing hybrid semantic-bit communication schemes mostly adopt fixed pattern partitioning, which cannot dynamically adjust the ratio of semantic and bit transmission based on knowledge matching degree and channel state. Summary of the Invention

[0005] This invention provides a hybrid spectrum efficiency optimization method for knowledge mismatch scenarios, which addresses the shortcomings of existing hybrid semantic-bit communication schemes that often use fixed pattern partitioning and cannot dynamically adjust the proportion of semantic and bit transmission based on knowledge matching degree and channel state. This method enables flexible adaptation to knowledge mismatch and time-varying channel environments. Through joint optimization of semantic participation degree and power allocation, it achieves synergistic enhancement of the two communication modes under a unified spectrum efficiency target, significantly improving the overall spectrum efficiency of the system.

[0006] This invention provides a hybrid spectrum efficiency optimization method for knowledge mismatch scenarios, comprising the following steps.

[0007] Construct a parallel hybrid semantic-bit communication system; for each user in the parallel hybrid semantic-bit communication system, adjust the contribution ratio of semantic communication and bit communication to the total spectral efficiency of the user by semantic participation degree; We use data regression to fit a quantitative mapping relationship between semantic similarity and signal-to-interference-plus-noise ratio (SINR) under different knowledge matching degrees; and we use a generalized logistic function to fit the impact of knowledge matching degree on semantic similarity. Using semantic participation and user power allocation strategy as joint optimization variables, an optimization problem is constructed with the goal of maximizing the total spectral efficiency of the system under the constraints of total base station transmit power, maximum user transmission delay, and power allocation fairness. The optimization problem is decomposed and solved by using an alternating optimization strategy and a two-level meta-heuristic selection optimization algorithm to obtain the optimal power allocation strategy and the optimal semantic participation value; based on the optimal power allocation strategy and the optimal semantic participation value, dynamic cooperative transmission of semantic-bit communication is carried out.

[0008] According to the present invention, a hybrid spectral efficiency optimization method for knowledge mismatch scenarios is provided, which constructs a parallel hybrid semantic-bit communication system, including: A parallel hybrid semantic-bit communication system is constructed in a single-cell, multi-user, multiple-input, single-output downlink scenario. The parallel hybrid semantic-bit communication system includes a base station equipped with multiple transmit antennas and multiple users. The base station simultaneously transmits semantic communication signals and bit communication signals to each user in parallel.

[0009] According to the present invention, a hybrid spectral efficiency optimization method for knowledge mismatch scenarios decomposes and solves the optimization problem through an alternating optimization strategy and a two-level metaheuristic selection optimization algorithm, obtaining the optimal power allocation strategy and the optimal semantic participation value, including: By employing an alternating optimization strategy, the optimization problem is decomposed into a semantic participation optimization subproblem and a power allocation optimization subproblem, and then solved iteratively. For the power allocation optimization subproblem, a two-level metaheuristic selection optimization algorithm is used to solve it, and the optimal power allocation strategy and the optimal semantic participation value are obtained.

[0010] According to the present invention, a hybrid spectral efficiency optimization method for knowledge mismatch scenarios decomposes the optimization problem into a semantic participation optimization sub-problem and a power allocation optimization sub-problem through an alternating optimization strategy, and solves them iteratively, including: Under the condition of fixed power allocation, solve the semantic participation convex optimization subproblem to obtain the updated semantic participation value; Under the condition of fixed semantic participation, solve the non-convex nonlinear programming subproblem of power allocation to obtain the updated power allocation strategy. Repeat the iteration until semantic participation and power allocation converge, and output the optimal solution by alternating optimization.

[0011] According to the present invention, a hybrid spectral efficiency optimization method for knowledge mismatch scenarios is provided. The two-layer metaheuristic selection optimization algorithm includes a policy heuristic layer and a candidate heuristic layer. For the power allocation optimization subproblem, the two-layer metaheuristic selection optimization algorithm is used to solve the problem, obtaining the optimal power allocation policy and the optimal semantic participation value, including: Through the policy heuristic layer, the ant colony optimization algorithm is used to dynamically update the pheromone based on the historical performance of each candidate heuristic algorithm, and adjust the probability of each candidate heuristic algorithm being selected. Through a candidate heuristic layer, the whale optimization algorithm, genetic algorithm, and sine-cosine algorithm are used to perform power allocation search and update the candidate solution population; By using adaptive scheduling in the policy heuristic layer and collaborative search in the candidate heuristic layer, the optimal power allocation strategy and the optimal semantic participation value are obtained based on alternating optimization of the optimal solution.

[0012] According to the present invention, a hybrid spectrum efficiency optimization method for knowledge mismatch scenarios is provided, which performs dynamic cooperative transmission of semantic-bit communication based on the optimal power allocation strategy and the optimal semantic participation value, including: The optimal semantic engagement value is used as a pattern scheduling variable to determine the proportion of data transmitted by each user through semantic communication and bit communication per unit time. The optimal power allocation strategy is used as a physical layer resource allocation variable to determine the actual transmit power and beam direction of semantic communication and bit communication; By combining the optimal semantic participation value and the optimal power allocation strategy, the transmission compression strength and delay control of the parallel hybrid semantic-bit communication system are achieved.

[0013] This invention also provides a hybrid spectrum efficiency optimization device for knowledge mismatch scenarios, comprising the following modules: The system building module is used to construct a parallel hybrid semantic-bit communication system; for each user in the parallel hybrid semantic-bit communication system, the contribution ratio of semantic communication and bit communication to the total spectral efficiency of the user is adjusted by the semantic participation degree. The fitting module is used to fit a quantitative mapping relationship between semantic similarity and signal-to-interference-plus-noise ratio under different knowledge matching degrees using data regression methods; and to fit the impact of knowledge matching degree on semantic similarity using a generalized logistic function. The optimization problem construction module is used to construct an optimization problem with the goal of maximizing the total spectral efficiency of the system, using semantic participation degree and user power allocation strategy as joint optimization variables, under the constraints of total base station transmit power, maximum user transmission delay and power allocation fairness. The solution and transmission module is used to decompose and solve the optimization problem through an alternating optimization strategy and a two-level metaheuristic selection optimization algorithm to obtain the optimal power allocation strategy and the optimal semantic participation value; and to perform dynamic cooperative transmission of semantic-bit communication based on the optimal power allocation strategy and the optimal semantic participation value.

[0014] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the hybrid spectrum efficiency optimization method for knowledge mismatch scenarios as described above.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the hybrid spectral efficiency optimization method for knowledge mismatch scenarios as described above.

[0016] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the hybrid spectrum efficiency optimization method for knowledge mismatch scenarios as described above.

[0017] This invention provides a hybrid spectrum efficiency optimization method and apparatus for knowledge mismatch scenarios. It constructs a parallel hybrid semantic-bit communication system; for each user in the system, it adjusts the contribution ratio of semantic communication to bit communication in the total spectrum efficiency by adjusting semantic participation; it uses data regression to fit a quantitative mapping relationship between semantic similarity and signal-to-interference-plus-noise ratio (SINR) under different knowledge matching degrees; it uses a generalized logistic function to fit the impact of knowledge matching degree on semantic similarity; using semantic participation degree and user power allocation strategy as joint optimization variables, it constructs an optimization problem with the goal of maximizing the total spectrum efficiency of the system under constraints of base station total transmit power, user maximum transmission delay, and power allocation fairness; and it decomposes and refines the optimization problem using an alternating optimization strategy and a two-level metaheuristic selection optimization algorithm. The optimal power allocation strategy and optimal semantic participation value are obtained through problem-solving. Based on these optimal strategies, dynamic collaborative transmission of semantic-bit communication is performed. By dynamically adjusting the transmission ratio of semantic communication to bit communication through semantic participation, the limitations of fixed mode division are broken, enabling the system to flexibly adapt to knowledge mismatch and time-varying channel environments. A quantitative relationship between knowledge matching degree, signal-to-interference-plus-noise ratio (SINR), and semantic similarity is established through a generalized logic function, achieving accurate perception and quantitative compensation of semantic communication performance. Through joint optimization of semantic participation and power allocation, the synergistic enhancement of the two communication modes is achieved under a unified spectral efficiency objective, significantly improving the overall spectral efficiency of the system. Through alternating optimization and a two-layer meta-heuristic algorithm, computational complexity is controlled while ensuring solution quality, meeting the real-time scheduling requirements of multi-user low-latency scenarios. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating the hybrid spectrum efficiency optimization method for knowledge mismatch scenarios provided by this invention.

[0020] Figure 2 This invention provides a parallel hybrid semantic-bit communication system.

[0021] Figure 3 This is the curve showing the change of semantic similarity with signal-to-noise ratio under different knowledge matching degrees provided by the present invention.

[0022] Figure 4This is a schematic diagram of the hybrid spectrum efficiency optimization device for knowledge mismatch scenarios provided by the present invention.

[0023] Figure 5 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0025] The following is combined with Figures 1 to 5 This invention is described.

[0026] Figure 1 This is a flowchart illustrating the hybrid spectral efficiency optimization method for knowledge mismatch scenarios provided by this invention, as shown below. Figure 1 As shown, the method includes the following: Step 101: Construct a parallel hybrid semantic-bit communication system; for each user in the parallel hybrid semantic-bit communication system, adjust the contribution ratio of semantic communication and bit communication to the total spectral efficiency of the user through semantic participation.

[0027] Optionally, in step 101, constructing a parallel hybrid semantic-bit communication system includes step A1: Step A1: Construct a parallel hybrid semantic-bit communication system in a single-cell, multi-user, multiple-input, single-output downlink scenario; wherein, the parallel hybrid semantic-bit communication system includes a base station equipped with multiple transmit antennas and multiple users; the base station simultaneously transmits semantic communication signals and bit communication signals to each user in parallel.

[0028] In step 101 above, as Figure 2 The diagram shows a parallel hybrid semantic-bit communication system, also known as a knowledge mismatch-driven parallel hybrid semantic-bit communication system. The system operates in a single-cell, multi-user, multiple-input, single-output (MU-MISO) downlink scenario, consisting of one base station (BS) and multiple users.

[0029] Among them, the base station is equipped with The set of transmitting antennas and users is denoted as . , contains There are [number] users. Each user will simultaneously receive two different types of signals: semantic communication signals (Sem-signal) and bit communication signals (Bit-signal). Accordingly, the set of semantic signals for all users is denoted as [details omitted]. The set of bit signals is denoted as The base station's transmitted signal vector Represented as in, These represent the allocation as the first Beamforming vectors for semantic and bit communication designed by individual users; Indicates the first Semantic signals for each user; Indicates the first The signal is a bit signal for each user. Beamforming aims to focus the transmit power towards the target user, reduce multi-user interference, and improve the signal quality of both communication modes.

[0030] This invention employs the Maximum Ratio Transmission (MRT) algorithm to maximize the signal-to-noise ratio of the target link, thereby enhancing the received signal strength. Therefore, the first... The semantic signal beam vector of a user can be represented as: . in, This indicates the transmit power allocated to the semantic signal for that user. Indicates the base station to the Channel vectors of individual user semantic communication links Its conjugate transpose. The beamforming vector of the bit signal is constructed using the same MRT principle. Assume and The mean of all of them is zero, and they satisfy the following conditions: Furthermore, the symbols used by different users are independent of each other. For expectation operator, and Indicates the first The conjugate transpose of the semantic signal and bit signal for each user represents the expression that normalizes the average power of both the semantic signal and bit signal to 1. For any user The semantic signals it receives for in The channel vector represents the semantic communication link. This represents additive white Gaussian noise. The second term in the equation represents multi-user interference from other semantic signals, and the third term represents interference introduced by all bit signals. Therefore, the signal-to-interference-plus-noise ratio (SINR) of the semantic signal can be obtained as follows: Similarly, let If it is a bit communication link channel, then the user The received bit signal is in, The corresponding bit signal SINR is In the proposed HSB system, the user The text data will be transmitted via semantic communication and bit communication in a certain proportion. This proportion is called the Semantic Participation Degree (SPD), denoted as . in For vectors The absolute value of semantic participation determines the proportion of semantic and bit communication contributions to the total spectral efficiency, thus affecting the overall system performance. Simultaneously, it determines the data load distribution between the two types of communication, thereby influencing transmission latency. Furthermore, under the same channel conditions, the value of semantic participation is closely related to the knowledge matching degree; generally, a higher knowledge matching degree corresponds to a greater semantic participation degree.

[0031] Step 102: Using data regression, fit a quantitative mapping relationship between semantic similarity and signal-to-interference-plus-noise ratio (SINR) under different knowledge matching degrees; use a generalized logistic function to fit the impact of knowledge matching degree on semantic similarity.

[0032] In step 102 above, the present invention constructs a semantic communication system based on the Transformer architecture and uses three basic deep neural networks (DNNs) to complete the functions of the sender, channel and receiver respectively.

[0033] At the sending end, the semantic encoder uses the multi-head self-attention mechanism of the Transformer encoding layer to capture inter-word dependencies in the text while filtering out redundant information. Subsequently, the extracted semantic features undergo feature dimensionality reduction. This is achieved by adjusting the downsampling depth parameter. This allows adjustment of the dimension of the semantic features output by the semantic encoder, thereby controlling the number of transmission symbols generated by the channel encoder. Specifically, a larger... This reduces the number of transmitted symbols, thus improving transmission efficiency; conversely, a smaller number of symbols... This increases the number of transmitted symbols, allowing semantic information to be preserved more completely, thereby improving the quality of semantic reconstruction.

[0034] At the receiver, based on the semantic vector recovered by the channel decoder, the semantic decoder utilizes the self-attention mechanism of the Transformer decoding layer to fuse background knowledge and reconstruct the semantic features of the original text. The proposed system employs an end-to-end training framework and introduces mutual information loss. and cross-entropy loss Two types of loss functions. The training process is divided into two stages: the first stage only optimizes... The first stage establishes an optimization benchmark for system capacity; the second stage constructs a total loss function by weighted fusion of the two types of losses. By minimizing The system simultaneously achieves two optimization objectives: minimizing semantic error and maximizing system capacity, thereby significantly enhancing the system's robustness in low SINR scenarios.

[0035] Thanks to the characteristics of semantic communication, it allows for a certain degree of bit-level error. Therefore, semantic communication focuses on the semantic similarity between the sent and received statements, rather than precise bit-by-bit recovery. In text transmission scenarios, semantic similarity can be quantified by the cosine similarity between the vector representations generated by a pre-trained language model, as shown in the following formula: in, This represents the sentence vector generated by the BERT model. The vector transpose is the vector representation of the original sentence after processing by the semantic encoder; the semantic similarity given by the above formula takes values ​​in the interval [0,1], with higher values ​​indicating a higher reconstructed sentence. With the original sentence The more similar they are.

[0036] This invention treats the vocabulary in the Transformer architecture as a shared background knowledge base for both the semantic encoder and decoder. The core function of the vocabulary is to map semantic units to numerical indices. When the vocabulary at the sending and receiving ends is inconsistent, the semantic encoding and decoding process will malfunction. For example, if the sending end's index "0" corresponds to the semantic unit "car," while the receiving end's index "0" corresponds to "airplane," it will lead to semantic misunderstanding. Therefore, this invention uses knowledge matching... This is used to measure the consistency between the two vocabulary lists.

[0037] Since there is no explicit analytical relationship between semantic similarity, knowledge matching degree, and SINR, this invention uses a data regression method to fit the relationship between semantic similarity and SINR under different knowledge matching degrees. The numerical fitting results are as follows: Figure 3 As shown, it can be observed that as the knowledge matching degree increases, the semantic similarity is monotonically non-decreasing, and its growth rate is... The curve rises first and then falls, exhibiting a typical "S" shape. Therefore, it can be expressed using a generalized logic function: Among them, parameters All based on knowledge matching degree Sure.

[0038] In semantic communication, semantic spectral efficiency (S-SE) is determined by both semantic similarity and the number of semantic symbols. It represents the amount of semantic information that can be successfully transmitted per unit bandwidth, and is usually measured in units / s / Hz. The semantic spectral efficiency can be expressed as: in, This represents the expected amount of semantic information contained in each sentence. This indicates the expected word count for each sentence. Because... Depending on the specific corpus, it is usually treated as a constant; This represents the average number of semantic symbols corresponding to each word, and it is related to the downsampling depth. Related. Furthermore, since the symbol rate equals the channel bandwidth in passband transmission, therefore, when the bandwidth is... The total amount of semantic information that can be transmitted on a channel is proportional to the bandwidth. Since the model proposed in this invention can convert sentences into continuous real-valued sequences, the constellation diagram output by the channel encoder can be considered to have an infinite number of constellation points, which is consistent with the performance of transmitting the semantic encoder output using Discrete-Time Analog Transmission (DTAT). Therefore, users... The transmission rate of the semantic signal is: in For channel bandwidth, therefore, user The semantic signal transmission delay can be expressed as: in This represents the average length of each sentence. Indicates the total number of sentences.

[0039] For bit communication, the goal is to achieve precise bit transmission and, as far as possible, error-free bit stream transmission. Its spectral efficiency essentially represents the rate of bits that can be transmitted per unit bandwidth. In the traditional bit rate model, the user... The spectral efficiency of a bit signal can be expressed as: in Indicates user The bit signal SINR. However, since bit spectral efficiency and semantic spectral efficiency are not measured in the same way, in order to better evaluate performance at the semantic level, we convert it to equivalent bit semantic efficiency (EB-SE). The definition of EB-SE is in This represents the average number of bits per word, which is generally considered to be... Because semantic similarity in bit-based communication is affected by the bit error rate, it is usually set to 1 in most cases. Meanwhile, users... The bit signal transmission delay can be expressed as: in This represents the bit transmission rate under the traditional Shannon capacity, satisfying... Compared to semantic communication, bit communication does not involve semantic extraction and compression, resulting in a significantly larger data transmission volume. Therefore, under the same bandwidth conditions, its spectral efficiency is usually lower than that of semantic communication. However, the spectral efficiency of bit communication is more sensitive to channel quality. When the SINR increases and exceeds a certain threshold, it can achieve a capacity gain far exceeding that of semantic communication. Nevertheless, with a substantial increase in data volume, the latency introduced by bit communication may still be higher than that of semantic communication. Therefore, system design requires a reasonable trade-off between semantic communication and bit communication based on channel conditions and service requirements, in order to balance spectral efficiency and transmission latency.

[0040] Step 103: Using semantic participation and user power allocation strategy as joint optimization variables, construct an optimization problem with the goal of maximizing the total spectral efficiency of the system under the constraints of total base station transmit power, maximum user transmission delay, and power allocation fairness.

[0041] In step 103 above, in this invention, the power allocated by the base station to each user is denoted as... Because the base station simultaneously sends semantic signals and bit signals to the user, therefore for the user , And satisfy The transmission delay for each user is denoted as . In other words, its latency is determined by the weighted maximum value of semantic communication and bit communication. To balance overall system performance and user experience, and to avoid service quality degradation for some users due to uneven power allocation, the Jain Fairness Index (JFI) is used to measure the fairness of power allocation, defined as follows: The spectral efficiency of each user can be viewed as a weighted sum of semantic spectral efficiency and equivalent bit spectral efficiency, where semantic participation is used to adjust the proportion of the two communication modes. The optimization objective of the system is to maximize the total spectral efficiency, and its optimization problem can be expressed as: Among them, constraints (17a) limit the total power to not exceed the maximum power generation capacity of the base station; (17b) ensure that the fairness of power allocation is not lower than the threshold; (17c) require that the latency of each user does not exceed the maximum allowable value; and (17d) ensure that the value range of SPD is legal. Due to semantic similarity It's about knowledge matching. H-Signal Interference-to-Noise Ratio The signal-to-interference-plus-noise ratio (SINR) is an S-shaped function, while the SINR is a fractional function of power distribution. Meanwhile, fairness metrics... It is a fractional function concerning power allocation. Therefore, the original optimization problem is inherently nonconvex, thus belonging to the typical nonconvex nonlinear programming problem. This invention provides an efficient alternating optimization algorithm to obtain a high-quality solution to the spectral efficiency maximization problem.

[0042] Step 104: Decompose and solve the optimization problem using an alternating optimization strategy and a two-layer meta-heuristic selection optimization algorithm to obtain the optimal power allocation strategy and the optimal semantic participation value; perform dynamic cooperative transmission of semantic-bit communication based on the optimal power allocation strategy and the optimal semantic participation value.

[0043] Optionally, in step 104, the optimization problem is decomposed and solved using an alternating optimization strategy and a two-level metaheuristic selection optimization algorithm to obtain the optimal power allocation strategy and the optimal semantic participation value, including steps B1 to B2: Step B1: By using an alternating optimization strategy, the optimization problem is decomposed into a semantic participation optimization subproblem and a power allocation optimization subproblem, and then solved iteratively.

[0044] Step B2: For the power allocation optimization subproblem, solve it using a two-level metaheuristic selection optimization algorithm to obtain the optimal power allocation strategy and the optimal semantic participation value.

[0045] Optionally, step B1 includes steps B11 to B13: Step B11: Under the condition of fixed power allocation, solve the semantic participation convex optimization subproblem to obtain the updated semantic participation value.

[0046] Step B12: Under the condition of fixed semantic participation, solve the non-convex nonlinear programming subproblem of power allocation to obtain the updated power allocation strategy.

[0047] Step B13: Repeat the iteration until semantic participation and power allocation converge, and output the optimal solution by alternating optimization.

[0048] In steps B11 to B13 above, this invention proposes an alternating optimization method, which decomposes problem P1 in step 103 into two sub-problems: one of the sub-problems is about... The other is a convex optimization problem; This problem is a nonconvex nonlinear programming problem, and the original problem of maximizing spectral efficiency is solved by alternating optimization iteration. The process of alternating optimization is summarized in Algorithm 1.

[0049] about The convex optimization problem can be expressed as: in On the other hand, regarding power The NLP problem can be represented as: It can be seen that the problem This is a linear programming problem, which can be solved using linear programming tools such as CVXPY. The problem... This remains a non-convex optimization problem. Although the dimensionality of the variables is reduced, its solution still relies on heuristic algorithms. Therefore, we propose a Two-Layer Meta-Heuristic Selection Optimization (TLMSO) algorithm to solve the problem. .

[0050] Optionally, the two-layer metaheuristic selection optimization algorithm includes a policy heuristic layer and a candidate heuristic layer; step B2 includes steps B21 to B23: Step B21: Using the policy heuristic layer, the ant colony optimization algorithm is employed to dynamically update the pheromone based on the historical performance of each candidate heuristic algorithm, thereby adjusting the probability of each candidate heuristic algorithm being selected.

[0051] Step B22: Using a candidate heuristic layer, employ the whale optimization algorithm, genetic algorithm, and sine-cosine algorithm to perform power allocation search and update the candidate solution population.

[0052] Step B23: Through adaptive scheduling of the policy heuristic layer and collaborative search of the candidate heuristic layer, the optimal power allocation strategy and the optimal semantic participation value are obtained based on alternating optimization of the optimal solution.

[0053] In steps B21 to B23 above, this invention proposes a two-layer metaheuristic selection optimization algorithm, TLMSO. This algorithm employs a two-layer structure: the upper layer is the Strategic Heuristic Layer (SHL), responsible for driving the selection of algorithms in the lower Candidate Heuristic Layer (CHL) to perform the power allocation task. TLMSO constructs an optimization framework that integrates efficient local search, robust global exploration, and adaptive scheduling capabilities.

[0054] In SHL, Ant Colony Optimization (ACO) is used as the policy heuristic to determine the selection of algorithms in CHL. The ACO algorithm simulates the cooperative behavior of ants searching for the shortest path during foraging. Ants release pheromones along the paths they traverse, and subsequent ants tend to choose paths with higher pheromone concentrations. Shorter paths accumulate pheromones faster due to their shorter round-trip time, gradually converging the process to the optimal path. Fitness is used to evaluate the quality of each candidate solution in the population; therefore, fitness is abstracted as pheromone, assuming that algorithms with higher fitness in CHL are more likely to be selected in the next iteration, thus being more conducive to finding the optimal solution. In the current problem, since the objective is to maximize the system's spectral efficiency, fitness is defined according to spectral efficiency, and is... The exponential form helps the algorithm better identify optimal solutions. When a user's power allocation scheme violates constraints, to simplify the process, we multiply its fitness by a penalty factor. This guides it to adjust in a direction that satisfies the constraints.

[0055] population any solution in It can be represented as .in, The number of solutions in the population, where each solution contains the power allocation strategy for all users. The power allocation vector is represented as ,in and These represent the power of the semantic signal and the power of the bit signal, respectively. First, all candidate solutions are initialized, and their semantic and bit components are generated as follows: in , and These are the lower and upper bounds of the power, respectively.

[0056] In the initial phase, SHL randomly selects an algorithm from CHL with equal probability and calculates its fitness. Subsequently, the candidate solution population is updated based on the fitness obtained from the selected algorithm, and the updated fitness is fed back to SHL. During this process, the "pheromone association value" of each CHL algorithm is updated according to the following rule: Wherein, if the currently selected algorithm Compared to the previous algorithm If they are different, then the information increment is and These represent the optimal fitness for each algorithm; Here is the volatile coefficient. After updating the pheromone, SHL adjusts the selection probabilities of each algorithm in the next round based on the new pheromone. The algorithm with the higher pheromone has a higher selection probability in the next round, and its selection probability is: in For the number of CHL algorithms, Initially set to 1, to represent the current CHL algorithm. With candidate algorithms Pheromonic relationships. Convert the selection probabilities of all algorithms into intervals. For non-overlapping sub-intervals, algorithms with higher probabilities correspond to wider intervals. To avoid the algorithm getting trapped in local optima and to enhance its exploration ability, random numbers are generated. It is generally believed that algorithms with higher selection probabilities are more likely to be selected. Selected. Introducing randomness helps prevent the algorithm from getting stuck in local optima and enhances its exploration capabilities. The detailed process is shown in Algorithm 2.

[0057] In this invention, CHL includes three typical metaheuristic algorithms: Whale Optimization (WOA), Genetic Algorithm (GA), and Sine-Cosine Algorithm (SCA). These algorithms each possess different optimization advantages, and through synergy, they can more effectively explore the solution space, search for the optimal solution, and dynamically update candidate solutions.

[0058] The WOA algorithm searches for the optimal power allocation scheme by simulating the process of a whale hunting its prey, and consists of three phases. In the initial phase, after the whale spots its prey, it surrounds it from all directions. In the power allocation problem, the "prey" corresponds to the currently found optimal power allocation solution. The remaining candidate solutions will adjust their power values ​​to approach the optimal solution, which can be expressed as: in, and These represent the current solution and the current optimal solution, respectively. This represents the distance between the current solution and the optimal solution. , It is a random vector with values ​​in the range [0,1]. in The convergence factor is linearly reduced from 2 to 0 during iteration, thereby adjusting the algorithm's global exploration and local exploitation capabilities. This method embodies a "shrinking encirclement" strategy by reducing the convergence factor. This causes the candidate solutions to gradually shrink towards the optimal solution.

[0059] In the second stage, the algorithm simulates the spiral ascent and bubble-blowing behavior commonly seen in whales during encirclement. This mechanism allows candidate solutions to approach the current optimal solution along a nonlinear spiral path. Simultaneously, sine and cosine functions are introduced to increase nonlinear perturbation, thereby enhancing local search capabilities. This can be expressed as: in, A constant used to control the shape of the spiral; The value is a random number in the range [-1, 1]. This update mechanism allows the algorithm to explore a wider range of search paths near the optimal solution, helping to escape simple local optima, increasing search diversity, and preventing the algorithm from converging prematurely in a single update mode.

[0060] In the third stage, when the whale needs to find new prey, it will move away from its current area to conduct a global exploration. In the corresponding algorithm, this stage aims to escape the current locally optimal region and search for potentially better power allocation solutions, which can be represented as: in, This represents a random solution in the current set of candidate solutions; This represents the distance between the current solution and a random solution. This mechanism helps the algorithm escape local space, explore unknown regions in the state space, and increase the probability of finding the global optimum.

[0061] By adjusting Size and random probability p The algorithm determines which update strategy to use, and the rules are as follows: This strategy dynamically switches between three behaviors: spiral encirclement, contraction encirclement, and global exploration, in order to balance exploration and development, thereby improving the overall optimization performance of the algorithm.

[0062] Genetic Algorithms (GA) liken optimization problems to biological evolution. Specifically, each candidate power allocation scheme is abstracted as a chromosome, and the specific power value allocated to each user corresponds to a gene on that chromosome. By simulating natural selection, GA prioritizes individuals with higher fitness as parents to generate the next generation, thus achieving a "survival of the fittest" evolutionary process. By simulating sexual reproduction, two parent chromosomes exchange some genes to produce entirely new offspring chromosomes, allowing the algorithm to explore new regions in the solution space. Furthermore, by simulating gene mutation, the value of a gene on a chromosome is randomly changed with a small probability, maintaining population diversity and preventing premature convergence.

[0063] In practical implementation, the power allocation scheme is first encoded as chromosomes. For this problem, a real-number encoding method is used, where power values ​​are directly used as genes. Individuals with high fitness in the current population are selected as parents through a roulette wheel selection process; the probability of selection is proportional to their fitness, i.e.: Subsequently, when random number If the probability is greater than the crossover probability, offspring are generated through crossover. For the parent individual... and It can generate two offspring. and .

[0064] in, The distribution factor is determined by the following formula: when At that time, the offspring's position is exactly the same as the parent's; when When the offspring generation is positioned between the parent and child generations, GA tends to favor localized development; when At this time, offspring will fall outside the scope of their parents, thus enhancing the overall exploration capability. In the formula... A random number in the interval [0,1]. This is the crossover index, used to control the degree to which offspring deviate from their parents. Finally, when randomly generated... When the mutation probability is less than the threshold, gene values ​​are randomly altered through mutation to maintain population diversity and enhance local search capabilities. This applies to parental genes. Its mutated genes Represented as: in and These are the upper and lower bounds of the power value, respectively. For a very small perturbation, it is defined as: This is the variation distribution index, used to adjust the degree to which mutated individuals deviate from the original individuals. Through continuous selection, crossover, and mutation operations, newly generated offspring gradually replace most or even all of the old individuals in the population. This evolutionary process continues to iterate until the optimal power allocation scheme is obtained.

[0065] The sine-cosine algorithm updates the position of individuals in a population by constructing a mathematical model based on sine and cosine functions. Its core idea is to utilize the periodic fluctuations of sine / cosine functions, allowing the population to alternate between global exploration and local exploitation in the solution space, ultimately finding the optimal value of the objective function. Its position update formula can be expressed as: in, Indicates the first The optimal power allocation value for all individuals in the population during round iteration. The convergence factor, used to determine the direction and step size of the next iteration, is defined as follows: , usually take , and decreases linearly from 2 to 0 with the number of iterations. When When the step size is large, the algorithm tends to perform a global search; when... When the step size is small, it tends to perform local searches. This is the direction factor, and its value range is... It is used to determine the fluctuation range of the sine / cosine function, that is, the direction of the next movement. This is a weighting factor, with a value range of [0,2], used to dynamically adjust the optimal solution. The impact on updating the current location. When When the trend is stronger, it tends to move closer to the optimal solution, favoring exploration; when At that time, this trend was weakened and favored development. Therefore, and They collectively influence the dynamic range of step size changes. The switching factor determines whether a sine or cosine function is used for position updates. Although sine and cosine functions have similar waveforms, their phases differ, thus affecting the same position update. Different update values ​​will be generated under these conditions. Alternating between the two can diversify the update behavior and effectively avoid getting trapped in local optima.

[0066] In the TLMSO framework proposed in this invention, three metaheuristic algorithms—WOA, GA, and SCA—are integrated into a candidate heuristic layer. In each iteration, TLMSO adaptively selects one of these algorithms for execution, with the maximum number of iterations denoted as […]. Through the synergistic effect of internal population evolution and external algorithm switching, TLMSO possesses both global exploration and local development capabilities, continuously strengthening the search process and gradually approaching and ultimately obtaining the optimal power allocation scheme.

[0067] Based on Algorithm 2, the computational complexity analysis of the TLMSO algorithm is as follows: Lines 1-3 perform initialization operations, the time cost of which is... Lines 4-16 constitute the main iteration process, where the primary complexity stems from the selection and execution of the CHL algorithm. Since the number of algorithms in CHL is fixed, the CHL algorithm selection step incurs only constant-time overhead. Once a CHL algorithm is selected, its execution determines the complexity of the inner loop. Specifically, whether it's WOA, GA, or SCA, each iteration requires updating all MMM candidate solutions, thus the time complexity of a single iteration is O(n log n). In execution After the inner iterations, the total complexity of each CHL algorithm is... Furthermore, lines 13-14 are used to evaluate and update all candidate solutions, with each outer iteration introducing additional... The overhead. Therefore, the time complexity of the outer loop is: In summary, the overall computational complexity of TLMSO is... That is, the complexity of the algorithm is linearly related to the size of the candidate solutions and the number of inner and outer iterations.

[0068] Optionally, step 104, which involves dynamic cooperative transmission of semantic-bit communication based on the optimal power allocation strategy and the optimal semantic participation value, includes steps C1 to C3: Step C1: Use the optimal semantic engagement value as a pattern scheduling variable to determine the proportion of data transmitted by each user through semantic communication and bit communication per unit time.

[0069] Step C2: Use the optimal power allocation strategy as a physical layer resource configuration variable to determine the actual transmit power and beam direction of semantic communication and bit communication.

[0070] Step C3: Combine the optimal semantic participation value and the optimal power allocation strategy to control the transmission compression strength and delay of the parallel hybrid semantic-bit communication system.

[0071] In step 104 above, the optimal semantic participation value and the optimal power allocation strategy are used together to control the transmission compression intensity and delay of the parallel hybrid semantic-bit communication system, so that the system can adaptively and smoothly switch and proportionally fuse semantic communication and bit communication according to the real-time changes in knowledge matching degree and channel state.

[0072] This invention provides a hybrid spectral efficiency optimization method for knowledge mismatch scenarios. It constructs a parallel hybrid semantic-bit communication system. For each user in the system, the contribution ratio of semantic communication to bit communication in the total spectral efficiency is adjusted by semantic participation. A data regression method is used to fit a quantitative mapping relationship between semantic similarity and signal-to-interference-plus-noise ratio (SIR) under different knowledge matching degrees. A generalized logistic function is used to fit the impact of knowledge matching degree on semantic similarity. Using semantic participation and user power allocation strategy as joint optimization variables, an optimization problem is constructed with the goal of maximizing the total system spectral efficiency under constraints of base station total transmit power, user maximum transmission delay, and power allocation fairness. The optimization problem is decomposed and solved using an alternating optimization strategy and a two-level metaheuristic selection optimization algorithm. The optimal power allocation strategy and optimal semantic participation value are obtained. Based on the optimal power allocation strategy and optimal semantic participation value, dynamic collaborative transmission of semantic-bit communication is carried out. By dynamically adjusting the transmission ratio of semantic communication and bit communication through semantic participation, the limitation of fixed mode division is broken, enabling the system to flexibly adapt to knowledge mismatch and time-varying channel environment. By establishing a quantitative relationship between knowledge matching degree, signal-to-interference-plus-noise ratio and semantic similarity through generalized logic functions, accurate perception and quantitative compensation of semantic communication performance are achieved. Through joint optimization of semantic participation and power allocation, the synergistic enhancement of the two communication modes is achieved under the unified spectral efficiency target, which significantly improves the overall spectral efficiency of the system. Through alternating optimization and two-level meta-heuristic algorithm, the computational complexity is controlled while ensuring the solution quality, meeting the real-time scheduling requirements of multi-user low-latency scenarios.

[0073] The hybrid spectrum efficiency optimization device for knowledge mismatch scenarios provided by the present invention will be described below. The hybrid spectrum efficiency optimization device for knowledge mismatch scenarios described below and the hybrid spectrum efficiency optimization method for knowledge mismatch scenarios described above can be referred to in correspondence with each other.

[0074] Figure 4 This is a flowchart illustrating the hybrid spectrum efficiency optimization device for knowledge mismatch scenarios provided by the present invention, as shown below. Figure 4 As shown, the device includes the following: System construction module 401 is used to construct a parallel hybrid semantic-bit communication system; for each user in the parallel hybrid semantic-bit communication system, the contribution ratio of semantic communication and bit communication in the total spectral efficiency of the user is adjusted by the semantic participation degree. The fitting module 402 is used to fit a quantitative mapping relationship between semantic similarity and signal-to-interference-plus-noise ratio under different knowledge matching degrees using data regression methods; and to fit the impact of knowledge matching degree on semantic similarity using a generalized logistic function. The optimization problem construction module 403 is used to construct an optimization problem with the goal of maximizing the total spectral efficiency of the system, using semantic participation degree and user power allocation strategy as joint optimization variables, under the constraints of total base station transmit power, maximum user transmission delay and power allocation fairness. The solution and transmission module 404 is used to decompose and solve the optimization problem through an alternating optimization strategy and a two-level metaheuristic selection optimization algorithm to obtain the optimal power allocation strategy and the optimal semantic participation value; and to perform dynamic cooperative transmission of semantic-bit communication based on the optimal power allocation strategy and the optimal semantic participation value.

[0075] This invention provides a hybrid spectrum efficiency optimization device for knowledge mismatch scenarios. It constructs a parallel hybrid semantic-bit communication system. For each user in the system, the contribution ratio of semantic communication to bit communication in the total spectrum efficiency is adjusted by semantic participation. A data regression method is used to fit a quantitative mapping relationship between semantic similarity and signal-to-interference-plus-noise ratio (SINR) under different knowledge matching degrees. A generalized logistic function is used to fit the impact of knowledge matching degree on semantic similarity. Using semantic participation and user power allocation strategy as joint optimization variables, an optimization problem is constructed with the goal of maximizing the total spectrum efficiency of the system, under constraints of total base station transmit power, maximum user transmission delay, and power allocation fairness. The optimization problem is decomposed and solved using an alternating optimization strategy and a two-level metaheuristic selection optimization algorithm. The optimal power allocation strategy and optimal semantic participation value are obtained. Based on the optimal power allocation strategy and optimal semantic participation value, dynamic collaborative transmission of semantic-bit communication is carried out. By dynamically adjusting the transmission ratio of semantic communication and bit communication through semantic participation, the limitation of fixed mode division is broken, enabling the system to flexibly adapt to knowledge mismatch and time-varying channel environment. By establishing a quantitative relationship between knowledge matching degree, signal-to-interference-plus-noise ratio and semantic similarity through generalized logic functions, accurate perception and quantitative compensation of semantic communication performance are achieved. Through joint optimization of semantic participation and power allocation, the synergistic enhancement of the two communication modes is achieved under the unified spectral efficiency target, which significantly improves the overall spectral efficiency of the system. Through alternating optimization and two-level meta-heuristic algorithm, the computational complexity is controlled while ensuring the solution quality, meeting the real-time scheduling requirements of multi-user low-latency scenarios.

[0076] Figure 5 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 5 As shown, the electronic device may include a processor 810, a communications interface 820, a memory 830, and a communication bus 840. The processor 810, communications interface 820, and memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions from the memory 830 to execute a hybrid spectral efficiency optimization method for knowledge mismatch scenarios.

[0077] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0078] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the hybrid spectrum efficiency optimization method for knowledge mismatch scenarios provided by the above methods.

[0079] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the hybrid spectral efficiency optimization methods for knowledge mismatch scenarios provided by the above methods.

[0080] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units 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. Those skilled in the art can understand and implement this without any creative effort.

[0081] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0082] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A hybrid spectral efficiency optimization method for knowledge mismatch scenarios, characterized in that, include: Construct a parallel hybrid semantic-bit communication system; For each user in the parallel hybrid semantic-bit communication system, the contribution ratio of semantic communication to bit communication in the total spectral efficiency is adjusted by the semantic participation degree. By using data regression methods, we fit a quantitative mapping relationship between semantic similarity and signal-to-interference-plus-noise ratio under different knowledge matching degrees; The influence of the knowledge matching degree on the semantic similarity is fitted using a generalized logical function; Using the semantic participation degree and user power allocation strategy as joint optimization variables, an optimization problem is constructed with the goal of maximizing the total spectral efficiency of the system under the constraints of total base station transmit power, maximum user transmission delay, and power allocation fairness. The optimization problem is decomposed and solved using an alternating optimization strategy and a two-layer meta-heuristic selection optimization algorithm to obtain the optimal power allocation strategy and the optimal semantic participation value; based on the optimal power allocation strategy and the optimal semantic participation value, dynamic cooperative transmission of semantic-bit communication is performed.

2. The hybrid spectral efficiency optimization method for knowledge mismatch scenarios according to claim 1, characterized in that, The construction of the parallel hybrid semantic-bit communication system includes: A parallel hybrid semantic-bit communication system is constructed in a single-cell, multi-user, multiple-input, single-output downlink scenario; wherein, the parallel hybrid semantic-bit communication system includes a base station equipped with multiple transmit antennas and multiple users; the base station simultaneously transmits semantic communication signals and bit communication signals in parallel for each user.

3. The hybrid spectral efficiency optimization method for knowledge mismatch scenarios according to claim 1, characterized in that, The optimization problem is decomposed and solved using an alternating optimization strategy and a two-level metaheuristic selection optimization algorithm to obtain the optimal power allocation strategy and the optimal semantic participation value, including: By employing an alternating optimization strategy, the optimization problem is decomposed into a semantic participation optimization subproblem and a power allocation optimization subproblem, and then solved iteratively. For the power allocation optimization subproblem, a two-level metaheuristic selection optimization algorithm is used to solve it, and the optimal power allocation strategy and the optimal semantic participation value are obtained.

4. The hybrid spectral efficiency optimization method for knowledge mismatch scenarios according to claim 3, characterized in that, The optimization problem is decomposed into semantic participation optimization sub-problems and power allocation optimization sub-problems through an alternating optimization strategy, and solved iteratively, including: Under the condition of fixed power allocation, solve the semantic participation convex optimization subproblem to obtain the updated semantic participation value; Under the condition of fixed semantic participation, solve the non-convex nonlinear programming subproblem of power allocation to obtain the updated power allocation strategy. Repeat the iteration until semantic participation and power allocation converge, and output the optimal solution by alternating optimization.

5. The hybrid spectral efficiency optimization method for knowledge mismatch scenarios according to claim 4, characterized in that, The two-layer metaheuristic selection optimization algorithm includes a policy heuristic layer and a candidate heuristic layer; the solution to the power allocation optimization subproblem using the two-layer metaheuristic selection optimization algorithm to obtain the optimal power allocation policy and the optimal semantic participation value includes: Through the policy heuristic layer, the ant colony optimization algorithm is used to dynamically update the pheromone based on the historical performance of each candidate heuristic algorithm, and adjust the probability of each candidate heuristic algorithm being selected. Through a candidate heuristic layer, the whale optimization algorithm, genetic algorithm, and sine-cosine algorithm are used to perform power allocation search and update the candidate solution population; By combining adaptive scheduling of the policy heuristic layer and collaborative search of the candidate heuristic layer, the optimal power allocation strategy and the optimal semantic participation value are obtained based on the alternating optimization of the optimal solution.

6. The hybrid spectral efficiency optimization method for knowledge mismatch scenarios according to claim 1, characterized in that, The dynamic cooperative transmission of semantic-bit communication based on the optimal power allocation strategy and the optimal semantic participation value includes: The optimal semantic engagement value is used as a pattern scheduling variable to determine the proportion of data transmitted by each user through semantic communication and bit communication per unit time. The optimal power allocation strategy is used as a physical layer resource configuration variable to determine the actual transmit power and beam direction of semantic communication and bit communication. The optimal semantic participation value and the optimal power allocation strategy work together to control the transmission compression strength and delay of the parallel hybrid semantic-bit communication system.

7. A hybrid spectrum efficiency optimization device for knowledge mismatch scenarios, characterized in that, include: The system building module is used to construct a parallel hybrid semantic-bit communication system; For each user in the parallel hybrid semantic-bit communication system, the contribution ratio of semantic communication to bit communication in the total spectral efficiency is adjusted by the semantic participation degree. The fitting module is used to fit a quantitative mapping relationship between semantic similarity and signal-to-interference-plus-noise ratio under different knowledge matching degrees using data regression methods. The influence of the knowledge matching degree on the semantic similarity is fitted using a generalized logical function; The optimization problem construction module is used to construct an optimization problem with the goal of maximizing the total spectral efficiency of the system, using the semantic participation degree and the user power allocation strategy as joint optimization variables, under the constraints of total base station transmit power, maximum user transmission delay, and power allocation fairness. The solution and transmission module is used to decompose and solve the optimization problem through an alternating optimization strategy and a two-level metaheuristic selection optimization algorithm to obtain the optimal power allocation strategy and the optimal semantic participation value; and to perform dynamic cooperative transmission of semantic-bit communication based on the optimal power allocation strategy and the optimal semantic participation value.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the hybrid spectrum efficiency optimization method for knowledge mismatch scenarios as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the hybrid spectrum efficiency optimization method for knowledge mismatch scenarios as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the hybrid spectrum efficiency optimization method for knowledge mismatch scenarios as described in any one of claims 1 to 6.

Images