A priority-based semantic communication optimization method, device and system

Abstract

The application discloses a kind of priority-based semantic communication optimization method, device and system, by parsing the key semantic information of data to be transmitted is obtained, and information priority and merged coding data packet are obtained according to information calculation;Based on network resource allocation algorithm, the first network resource allocation strategy is obtained according to the above two, and the allocation strategy is optimized according to network state index and collaborative optimization algorithm, obtains and transmits data packet according to the first collaborative optimization result, while collecting and adjusting the first collaborative optimization result based on allocation algorithm according to transmission performance index, and transmits data according to the strategy after adjustment.The integrity and transmission reliability of important semantic information in the data to be transmitted are ensured by information priority confirmation during data transmission, and the amount of data transmission is reduced by merging and coding data packet;Resource allocation algorithm and collaborative optimization algorithm improve the practicability of resource allocation strategy, so that the strategy can effectively respond to the change of network environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semantic communication, in particular to a priority-based semantic communication optimization method, device and system. BACKGROUND

[0002] Semantic communication is a technology that first selectively extracts, compresses and transmits the features of the original signal, and then communicates using semantic level information. Its advantage is to understand first and then transmit based on the task, which greatly improves the transmission efficiency and reliability of the communication system. Semantic communication technology is a complete revolution of the design idea and concept of communication system, which brings a new way of thinking and means for future communication and will fundamentally change the traditional information communication technology system.

[0003] With the development of Internet of Things, autonomous driving, remote medical treatment and other applications, the requirements for low delay and high reliability of communication technology are becoming higher and higher. Traditional communication methods are mainly based on transmitting as much data as possible to ensure the integrity of information, but this method often leads to network congestion, increases delay and reduces transmission reliability. How semantic communication determines the priority of key information and how it cooperates with network resource allocation to achieve low delay and high reliability of high-speed data transmission has become a difficult problem in the field. SUMMARY

[0004] The present application provides a priority-based semantic communication optimization method, device and system to achieve the technical effect of improving the reliability and practicality of data transmission while reducing the network delay of data transmission.

[0005] To solve the above technical problems, the present application provides a priority-based semantic communication optimization method, comprising the following steps:

[0006] In response to the data transmission signal, the to-be-transmitted data is received and parsed to obtain key semantic information, and the to-be-transmitted data is prioritized and merged and encoded according to the key semantic information to obtain information priority and merged and encoded data packet;

[0007] Based on the preset network resource allocation algorithm, the first network resource allocation strategy is calculated based on the information priority and the merged and encoded data packet, and then the first network resource allocation strategy is processed based on the network state index and the cooperative optimization algorithm to obtain the first cooperative optimization result, and the merged and encoded data packet is transmitted according to the first cooperative optimization result;

[0008] In the data transmission process, the transmission performance index is collected in real time, and the first cooperative optimization result is adjusted according to the transmission performance index and the network resource allocation algorithm, so that the merged and encoded data packet is transmitted according to the adjusted strategy.

[0009] The semantic communication method provided by this invention, after parsing the data to be transmitted, can calculate the information priority and perform data merging and encoding based on the key semantic information obtained from the parsing, thereby obtaining the information priority of the key semantic information and the merged and encoded data packet. The calculated information priority can be used as reference data for subsequent adjustments to the network resource allocation strategy of the system, that is, the network resource allocation weight of data with higher information priority is increased accordingly, thereby ensuring the integrity and reliability of important semantic information in the data to be transmitted during data transmission. Furthermore, merging and encoding the data to obtain merged and encoded data packets reduces the amount of data transmitted while ensuring data integrity, further improving the reliability and practicality of data transmission.

[0010] Based on the network resource allocation algorithm, a first network resource allocation strategy is calculated according to the calculated information priority and the merged coded data packets. This first strategy considers both the protection priority of key semantic information and the priority of network resource allocation, as well as the number and volume of merged coded data packets, thus improving the practicality and reliability of the resource allocation strategy. Simultaneously, the first network resource allocation strategy is further optimized using a collaborative optimization algorithm based on network state indicators. This optimization result takes into account the real-time network state, further improving the reliability and efficiency of data transmission and enabling the optimized allocation strategy to effectively cope with changes in the network environment.

[0011] After data packet transmission begins, the system will collect transmission performance indicators in real time, update the parameters of the network resource allocation algorithm based on the collected performance indicator data, solve the updated algorithm to obtain the corresponding solution results, and adjust and update the first collaborative optimization result based on the solution results so that data packets are transmitted according to the updated strategy. This achieves adaptive feedback adjustment of the data packet transmission strategy, and adjusts the transmission strategy in real time based on the transmission performance indicators, so that the adjusted transmission strategy is more adaptable to the network state, further improving communication efficiency and the reliability of communication data transmission.

[0012] As a preferred example, the step of receiving and parsing the data to be transmitted to obtain key semantic information, and then performing priority calculation and semantic data merging encoding on the data to be transmitted based on the key semantic information to obtain information priority and merged encoded data packets, specifically involves:

[0013] The system receives and performs noise removal and standardization on the data to be transmitted in sequence, outputs the corresponding processed data, and uses an autoencoder or convolutional neural network to extract features from the processed data to obtain the key semantic information.

[0014] The key semantic information is divided into a training dataset and a test dataset. The training dataset is input into a neural network model for training. The trained model performs feature analysis and semantic importance evaluation on the test dataset to obtain semantic importance. Then, based on the semantic importance, a weighted algorithm is used to calculate the information priority of each key semantic information.

[0015] The merged encoded data packet is obtained by merging and encoding the data corresponding to information whose similarity reaches a similarity threshold in the key semantic information.

[0016] After receiving the data to be transmitted, the system first performs noise reduction and standardization, i.e., data preprocessing, to improve data quality and enhance the adaptability of the processed data to subsequent neural network models. Following preprocessing, the system selects a method for feature extraction, i.e., key semantic information acquisition, based on the processed data. It chooses either an autoencoder or a convolutional neural network for feature extraction, which not only improves the flexibility of selecting key semantic information extraction methods but also makes the selected extraction method more adaptable to the processed data.

[0017] After extracting the key semantic information, the system divides it into training and test datasets. The neural network model is then trained using the training dataset to improve its sensitivity to the processed data. This enhances the accuracy of the model's feature analysis and semantic importance assessment of the test dataset, providing reference data for prioritizing information in subsequent calculations.

[0018] As a preferred example, the step of obtaining the merged encoded data packet by merging and encoding the data corresponding to information whose similarity reaches a similarity threshold in the key semantic information specifically involves:

[0019] Based on the key semantic information, cosine similarity or Jaccard similarity coefficient is selected to measure the semantic similarity of each piece of information in the key semantic information, thereby obtaining multiple corresponding similarities;

[0020] The multiple similarities are compared with the preset similarity threshold. The data items corresponding to the key semantic information whose similarity is greater than or equal to the similarity threshold are clustered to obtain several data groups. The data items in the several data groups are sorted according to the information priority of the corresponding key semantic information.

[0021] By using a preset Huffman coding method, the sorted data items in each data group are merged and encoded to obtain several merged and encoded data packets.

[0022] To improve data transmission efficiency and achieve high-efficiency communication, the semantic communication optimization method provided by this invention will also perform merge encoding processing on data whose key semantic information similarity reaches a threshold when calculating and determining the information priority of the data. First, based on the key semantic information, a similarity calculation method of cosine similarity or Jaccard similarity coefficient is selected to calculate and quantify the similarity between each key piece of information. The calculated similarity is then compared with a preset similarity threshold to determine whether the two data items corresponding to the two key semantic information items with the same similarity can be classified into the same group of data. Then, all key semantic information is traversed, and all data items are clustered according to the determined similarity to obtain several groups of data, thereby reducing the amount of data transmitted and improving data transmission efficiency.

[0023] Before merging and encoding the data items in each data group, the data items in each data group will be sorted according to the information priority of the key semantic information corresponding to each data item. This ensures that the key semantic information with higher priority can be sent and processed faster, improving the responsiveness and efficiency of the system, as well as the timeliness and accuracy of the system's data transmission.

[0024] As a preferred example, the first network resource allocation strategy, calculated using the information priority and the merged encoded data packet based on the preset network resource allocation algorithm, specifically includes:

[0025] The network load is calculated based on the number and amount of the merged encoded data packets, while the average network latency and packet loss rate are monitored and collected in real time. Then, the network congestion index is calculated based on the average network latency and the packet loss rate.

[0026] Based on the network resource allocation algorithm, the first network resource allocation strategy is obtained by calculating and solving according to the network load, the network congestion index, and the information priority.

[0027] After obtaining several merged and encoded data packets, the system will determine the first network resource allocation strategy based on the calculated information priority and the merged and encoded data packets. First, the system will calculate the network load based on the number and amount of data generated by the merged and encoded packets. At the same time, it will also collect the current average network latency and packet loss rate, and then calculate the current network congestion index, i.e., the degree of network congestion, based on the average network latency and packet loss rate.

[0028] Once determined, the network load, network congestion index, and information priority are calculated and solved based on the network resource allocation algorithm to obtain the first network resource allocation strategy. This improves the adaptability of the network allocation strategy to the current network state and also ensures that higher priority data will be transmitted and processed more quickly.

[0029] As a preferred example, the process of processing the first network resource allocation strategy based on network state indicators and a collaborative optimization algorithm to obtain the first collaborative optimization result specifically includes:

[0030] The network latency and network utilization in the network status indicators are collected in real time, and an adjustment factor is calculated based on the network latency and network utilization using a preset activation function.

[0031] The weight factors in the collaborative optimization algorithm are obtained by optimizing the adjustment factors. Then, the collaborative optimization algorithm is updated according to the calculated weight factors. The first network resource allocation strategy is then collaboratively optimized using the updated collaborative optimization algorithm, and the first collaborative optimization result is output.

[0032] After obtaining the first network resource allocation strategy, the system will further optimize and adjust the strategy based on the network status. Adjustment factors are obtained by real-time collection of network latency and utilization, as well as by calculating activation functions. Weight factors in the collaborative optimization algorithm are then calculated based on these adjustment factors. This allows the system to adjust the weight factors and update the collaborative optimization algorithm accordingly, thereby improving its adaptability to the current network status. The updated collaborative optimization algorithm then optimizes the first network resource allocation strategy, outputting the optimized first collaborative optimization result. This further enhances the adaptability of the optimized network resource allocation strategy to the current network status, thus improving the reliability, practicality, and accuracy of data transmission.

[0033] As a preferred example, the real-time acquisition of transmission performance indicators and the adjustment of the first collaborative optimization result based on the transmission performance indicators and the network resource allocation algorithm specifically include:

[0034] During data packet transmission, the bandwidth utilization and transmission power efficiency in the transmission performance indicators are collected in real time, and the adjustment coefficient in the network resource allocation algorithm is obtained by calculating based on the bandwidth utilization and transmission power efficiency through a preset adaptive control algorithm.

[0035] The network resource allocation algorithm is updated with parameters based on the calculated adjustment coefficients, and the updated network resource allocation algorithm is solved. Then, the first collaborative optimization result is adjusted based on the solution result.

[0036] When the system transmits data packets based on the first collaborative optimization result, it will also collect the bandwidth utilization and transmission power efficiency in the network transmission performance indicators in real time. The collected data will be used as input data to the preset adaptive control algorithm for calculation and solution, and the corresponding calculation results will be used as adjustment coefficients in the network resource allocation algorithm to update the algorithm parameters accordingly. This will make the updated network resource allocation algorithm more adaptable to the current network transmission performance, thereby improving the reliability and practicality of the system's data transmission.

[0037] By adjusting the system's data transmission strategy using network transmission performance status data feedback, not only are data transmission efficiency, security, and reliability improved, but dynamic and flexible adjustments to the data transmission strategy and the system's self-adaptive capabilities are also achieved. Furthermore, the stability and response speed of the entire communication system are enhanced.

[0038] Accordingly, this embodiment of the invention also provides a priority-based semantic communication optimization device, which includes a priority calculation module, an allocation strategy transmission module, and an indicator feedback optimization module.

[0039] The priority calculation module is used to respond to the data transmission signal, receive and parse the data to be transmitted, obtain key semantic information, and perform priority calculation and semantic data merging and encoding on the data to be transmitted according to the key semantic information to obtain information priority and merged encoded data packets.

[0040] The allocation strategy transmission module is used to calculate a first network resource allocation strategy based on a preset network resource allocation algorithm, through the information priority and the merged coded data packet, and then retrieve and process the first network resource allocation strategy through a collaborative optimization algorithm to obtain a first collaborative optimization result, and transmit the merged coded data packet according to the first collaborative optimization result.

[0041] The indicator feedback optimization module is used to collect network transmission performance indicators in real time during data transmission, and adjust the first collaborative optimization result according to the network transmission performance indicators through the network resource allocation algorithm, so that the merged encoded data packet is transmitted according to the adjusted strategy.

[0042] As a preferred example, the priority calculation module receives and parses the data to be transmitted to obtain key semantic information, and performs priority calculation and semantic data merging encoding on the data to be transmitted based on the key semantic information to obtain information priority and merged encoded data packets, specifically:

[0043] The system receives and performs noise removal and standardization on the data to be transmitted in sequence, outputs the corresponding processed data, and uses an autoencoder or convolutional neural network to extract features from the processed data to obtain the key semantic information.

[0044] The key semantic information is divided into a training dataset and a test dataset. The training dataset is input into a neural network model for training. The trained model performs feature analysis and semantic importance evaluation on the test dataset to obtain semantic importance. Then, based on the semantic importance, a weighted algorithm is used to calculate the information priority of each key semantic information.

[0045] The merged encoded data packet is obtained by merging and encoding the data corresponding to information whose similarity reaches a similarity threshold in the key semantic information.

[0046] Accordingly, embodiments of the present invention also provide a priority-based semantic communication optimization system, the semantic communication optimization system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the priority-based semantic communication optimization system described above when processing the computer program.

[0047] Accordingly, embodiments of the present invention also provide a storage medium storing a computer program, which is called and executed by a processor to implement a priority-based semantic communication optimization system as described in any of the above embodiments. Attached Figure Description

[0048] Figure 1 : A flowchart illustrating an embodiment of the priority-based semantic communication optimization method provided by the present invention;

[0049] Figure 2 : A schematic diagram of an embodiment of the priority-based semantic communication optimization device provided by the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Example 1

[0052] Please refer to Figure 1This is a flowchart illustrating an embodiment of the priority-based semantic communication optimization method provided by the present invention, including steps 101 to 103, each step being as follows:

[0053] Step 101: In response to the data transmission signal, receive and parse the data to be transmitted, obtain key semantic information, and perform priority calculation and semantic data merging encoding on the data to be transmitted according to the key semantic information to obtain information priority and merged encoded data packets.

[0054] The semantic communication method provided in this invention, after parsing the data to be transmitted, can calculate the information priority and perform data merging and encoding based on the key semantic information obtained from the parsing, thereby obtaining the information priority of the key semantic information and the merged and encoded data packet. The calculated information priority can be used as reference data when the system adjusts the network resource allocation strategy in the future, that is, the network resource allocation weight of data with higher information priority is increased accordingly, so as to ensure the integrity and reliability of important semantic information in the data to be transmitted during data transmission. Furthermore, merging and encoding the data to obtain merged and encoded data packets reduces the amount of data transmitted while ensuring data integrity, further improving the reliability and practicality of data transmission.

[0055] Specifically, in this embodiment, receiving and parsing the data to be transmitted to obtain key semantic information, and performing priority calculation and semantic data merging encoding on the data to be transmitted based on the key semantic information to obtain information priority and merged encoded data packets, specifically:

[0056] The system receives and performs noise removal and standardization on the data to be transmitted in sequence, outputs the corresponding processed data, and uses an autoencoder or convolutional neural network to extract features from the processed data to obtain the key semantic information.

[0057] The key semantic information is divided into a training dataset and a test dataset. The training dataset is input into a neural network model for training. The trained model performs feature analysis and semantic importance evaluation on the test dataset to obtain semantic importance. Then, based on the semantic importance, a weighted algorithm is used to calculate the information priority of each key semantic information.

[0058] The merged encoded data packet is obtained by merging and encoding the data corresponding to information whose similarity reaches a similarity threshold in the key semantic information.

[0059] After receiving the data to be transmitted, the system first performs noise reduction and standardization, i.e., data preprocessing, to improve data quality and enhance the adaptability of the processed data to subsequent neural network models. Following preprocessing, the system selects a method for feature extraction, i.e., key semantic information acquisition, based on the processed data. It chooses either an autoencoder or a convolutional neural network for feature extraction, which not only improves the flexibility of selecting key semantic information extraction methods but also makes the selected extraction method more adaptable to the processed data.

[0060] After extracting the key semantic information, the system divides it into training and test datasets. The neural network model is then trained using the training dataset to improve its sensitivity to the processed data. This enhances the accuracy of the model's feature analysis and semantic importance assessment of the test dataset, providing reference data for prioritizing information in subsequent calculations.

[0061] In this embodiment, the specific steps for training the neural network model using the training dataset are as follows: A Long Short-Term Memory (LSTM) network is used to process time-series data to identify and predict key events and information points in the data stream. The LSTM network structure can be represented as follows:

[0062] f t =σ(W f ·[h t-1 x t ]+b f )

[0063] i t =σ(W i ·[h t-1 x t ]+b i )

[0064]

[0065] o t =σ(W o ·[h t-1 x t ]+b o )

[0066] h t =o t *tanh(C t )

[0067] Where ft, it, and ot are the forget gate, input gate, and output gate, respectively; Ct and ht are the unit state and hidden state, respectively; W and b are the weight and bias parameters, respectively; and σ and tanh are the sigmoid function and hyperbolic tangent activation function, respectively. The model is trained through the above process, and the test dataset is input into the trained model for key semantic information analysis to evaluate the semantic importance of each piece of information. The evaluated semantic importance is then output, and the information priority pi for each piece of information is calculated using a weighted algorithm based on the urgency and impact of the information. The specific calculation formula is shown below:

[0068] p i =w1*urgency i +w2*impact i

[0069] Furthermore, in this embodiment, the method of merging and encoding data corresponding to information whose similarity reaches a similarity threshold in the key semantic information to obtain the merged encoded data packet specifically involves:

[0070] Based on the key semantic information, cosine similarity or Jaccard similarity coefficient is selected to measure the semantic similarity of each piece of information in the key semantic information, thereby obtaining multiple corresponding similarities;

[0071] The multiple similarities are compared with the preset similarity threshold. The data items corresponding to the key semantic information whose similarity is greater than or equal to the similarity threshold are clustered to obtain several data groups. The data items in the several data groups are sorted according to the information priority of the corresponding key semantic information.

[0072] By using a preset Huffman coding method, the sorted data items in each data group are merged and encoded to obtain several merged and encoded data packets.

[0073] To improve data transmission efficiency and achieve high-efficiency communication, the semantic communication optimization method provided in this embodiment of the invention will also perform merge encoding processing on data whose key semantic information similarity reaches a threshold when calculating and determining the information priority of the data. First, based on the key semantic information, a similarity calculation method of cosine similarity or Jaccard similarity coefficient is selected to calculate and quantify the similarity between each key piece of information. The calculated similarity is then compared with a preset similarity threshold to determine whether the two data items corresponding to the two key semantic information items with the same similarity can be classified into the same group of data. Then, all key semantic information is traversed, and all data items are clustered according to the determined similarity to obtain several groups of data, thereby reducing the amount of data transmitted and improving data transmission efficiency.

[0074] Before merging and encoding the data items in each data group, the data items in each data group will be sorted according to the information priority of the key semantic information corresponding to each data item. This ensures that the key semantic information with higher priority can be sent and processed faster, improving the responsiveness and efficiency of the system, as well as the timeliness and accuracy of the system's data transmission.

[0075] In this embodiment, after the system extracts features from the data to be transmitted to obtain the corresponding key semantic information, it will also divide the data to be transmitted into multiple groups C1, C2, ..., Cn according to the similarity of the key semantic information corresponding to the predefined semantic similarity measure, such as cosine similarity, Jaccard similarity and other similarity calculation methods. Each group contains semantically similar data items, that is, the similarity between the data items in each group is greater than or equal to the preset similarity threshold.

[0076] After clustering, the system uses Huffman coding to merge the data items within each group Ck into a single data packet. Specifically, it merges and encodes the data in each group by calculating and sorting the frequency f(x) of each unique data item in Ck. Each time, it selects the two nodes with the lowest frequencies and merges them, creating a new parent node whose frequency is the sum of the frequencies of its two child nodes. This process is repeated until a complete Huffman tree is constructed. Then, starting from the root node of the Huffman tree, it assigns "0" to the left and "1" to the right, generating a unique code for each leaf node (original data item). The generated Huffman codes are used to encode the data in Ck, and all codes are then merged into a single data packet. The specific formula is as follows:

[0077]

[0078] Here, code(x) is the Huffman code of data item x.

[0079] In addition to the above-mentioned merging encoding process, this embodiment also provides another merging encoding method. In its specific implementation steps, after the data items in the data to be transmitted are divided into multiple groups such as C1, C2, ..., Cn according to the similarity of their corresponding key semantic information, the data items in each group will be sorted according to the information priority. Since the data items in each group are semantically similar, each group will also be sorted by priority.

[0080] When merging and encoding the data within a group after sorting, the system calculates a weighted frequency f(x) × px for each data item x in each group Ck, where px is the priority of data item x. The data items in Ck are sorted according to the weighted frequency f(x) × px. Each time, the two nodes with the lowest weighted frequencies are selected and merged to create a new parent node whose frequency is the sum of the frequencies of its two child nodes. This process is repeated until a complete Huffman tree is constructed. Then, starting from the root node of the Huffman tree, "0" is assigned to the left and "1" to the right, generating a unique code for each leaf node. The generated Huffman code is used to encode the data in Ck, and then all codes are merged into a single data packet. The specific formula is as follows:

[0081]

[0082] By incorporating information priority into the frequency calculation of Huffman coding, it is possible to ensure that critical information is sent and processed faster, thereby improving the responsiveness and efficiency of the system. This method is more suitable for application scenarios that have high requirements for the timeliness and accuracy of data transmission.

[0083] Step 102: Based on a preset network resource allocation algorithm, a first network resource allocation strategy is calculated using the information priority and the merged coded data packet. Then, the first network resource allocation strategy is processed based on network status indicators and a collaborative optimization algorithm to obtain a first collaborative optimization result. Finally, the merged coded data packet is transmitted according to the first collaborative optimization result.

[0084] Based on the network resource allocation algorithm, a first network resource allocation strategy is calculated according to the calculated information priority and the merged coded data packets. This first strategy considers both the protection priority of key semantic information and the priority of network resource allocation, as well as the number and volume of merged coded data packets, thus improving the practicality and reliability of the resource allocation strategy. Simultaneously, the first network resource allocation strategy is further optimized using a collaborative optimization algorithm based on network state indicators. This optimization result takes into account the real-time network state, further improving the reliability and efficiency of data transmission and enabling the optimized allocation strategy to effectively cope with changes in the network environment.

[0085] Specifically, the first network resource allocation strategy, calculated based on the information priority and the merged encoded data packet, according to the preset network resource allocation algorithm described in this embodiment, is as follows:

[0086] The network load is calculated based on the number and amount of the merged encoded data packets, while the average network latency and packet loss rate are monitored and collected in real time. Then, the network congestion index is calculated based on the average network latency and the packet loss rate.

[0087] Based on the network resource allocation algorithm, the first network resource allocation strategy is obtained by calculating and solving according to the network load, the network congestion index, and the information priority.

[0088] After obtaining several merged and encoded data packets, the system will determine the first network resource allocation strategy based on the calculated information priority and the merged and encoded data packets. First, the system will calculate the network load based on the number and amount of data generated by the merged and encoded packets. At the same time, it will also collect the current average network latency and packet loss rate, and then calculate the current network congestion index, i.e., the degree of network congestion, based on the average network latency and packet loss rate.

[0089] Once determined, the network load, network congestion index, and information priority are calculated and solved based on the network resource allocation algorithm to obtain the first network resource allocation strategy. This improves the adaptability of the network allocation strategy to the current network state and also ensures that higher priority data will be transmitted and processed more quickly.

[0090] In this embodiment, the network resource allocation algorithm formula is as follows:

[0091] R t =γ×BW(t)+δ×TP(t)

[0092] Where R(t) represents the resource allocation at time t, BW(t) represents the available bandwidth, TP(t) represents the transmission power, and γ and δ are adjustment coefficients. The specific formula for calculating network load based on the data volume and number of merged encoded data packets is as follows:

[0093]

[0094] Where Si(t) is the size of the i-th data packet transmitted at time t, and N is the total number of data packets.

[0095] The network congestion index C(t), which represents the degree of network congestion, is calculated based on network latency and packet loss rate. The specific calculation formula is as follows:

[0096] C t = k1×D(t) + k2×P(t)

[0097] Where D(t) is the average network latency, P(t) is the packet loss rate, and K1 and K2 are weighting factors used to adjust the impact of latency and packet loss rate on the congestion index. Users can adjust the weighting of network latency and packet loss rate on the network congestion index by adjusting K1 and K2.

[0098] The first network resource allocation strategy can be calculated based on the network congestion index, network load, and information priority. This strategy dynamically adjusts bandwidth and transmission power allocation according to the congestion index C(t). If C(t) exceeds a preset threshold Cmax, resource allocation is reduced to alleviate congestion; conversely, if C(t) is less than the preset threshold Cmax, resource allocation is increased to improve network efficiency. This resource adjustment strategy can be expressed as:

[0099]

[0100] Here, ρ and σ are response coefficients for adjusting bandwidth and transmission power. The first network resource allocation strategy can be calculated using the algorithm described above. Furthermore, network resources can be adjusted in real time based on this algorithm to respond to changes in network conditions.

[0101] Furthermore, in this embodiment, the first network resource allocation strategy is processed based on network state indicators and a collaborative optimization algorithm to obtain a first collaborative optimization result, specifically as follows:

[0102] The network latency and network utilization in the network status indicators are collected in real time, and an adjustment factor is calculated based on the network latency and network utilization using a preset activation function.

[0103] The weight factors in the collaborative optimization algorithm are obtained by optimizing the adjustment factors. Then, the collaborative optimization algorithm is updated according to the calculated weight factors. The first network resource allocation strategy is then collaboratively optimized using the updated collaborative optimization algorithm, and the first collaborative optimization result is output.

[0104] After obtaining the first network resource allocation strategy, the system will further optimize and adjust the strategy based on the network status. Adjustment factors are obtained by real-time collection of network latency and utilization, as well as by calculating activation functions. Weight factors in the collaborative optimization algorithm are then calculated based on these adjustment factors. This allows the system to adjust the weight factors and update the collaborative optimization algorithm accordingly, thereby improving its adaptability to the current network status. The updated collaborative optimization algorithm then optimizes the first network resource allocation strategy, outputting the optimized first collaborative optimization result. This further enhances the adaptability of the optimized network resource allocation strategy to the current network status, thus improving the reliability, practicality, and accuracy of data transmission.

[0105] In this embodiment, the objective function of the collaborative optimization algorithm provided by the system includes minimizing latency and maximizing reliability; therefore, its expression is:

[0106] f(D, R) = α × D + β × (1 - R)

[0107] Where D represents latency, R represents reliability, and α and β are weighting factors. This embodiment introduces dynamically adjusted factors to optimize the weights α and β in the collaborative optimization algorithm. Specifically, the influencing factors of network state indicators are first determined as network latency L and network utilization U, and then the weighting factors α and β are adjusted according to these two parameters.

[0108] First, assuming initial weights α0 and β0 exist and are set as the optimal weights under ideal network conditions, two dynamic adjustment factors αadj and βadj are introduced, which are the adjustment factors described in this embodiment. These two dynamic adjustment factors are adjusted based on network latency L and utilization rate U; that is, the dynamic adjustment factors are calculated based on network latency L and utilization rate U, and the weight factors α and β are calculated based on these dynamic adjustment factors. The definitions of the dynamic adjustment factors are as follows:

[0109]

[0110]

[0111] Here, k is a constant that adjusts the steepness of the curve, while Lthreshold and Uthreshold are the thresholds for network latency and network utilization, respectively. The above function is the Sigmoid function, used to smoothly adjust the weighting factors, ensuring that changes in weights do not cause system instability due to abrupt changes. After determining the dynamic adjustment factor, the weighting factors are calculated based on it, using the following formula:

[0112] α=α0*αadj

[0113] β = β0 * βadj,

[0114] As shown in the formula above, when the network latency L is higher than Lthreshold, αadj increases, leading to an increase in α, thus emphasizing latency reduction. Conversely, when the network utilization U is higher than Uthreshold, βadj increases, leading to an increase in β, thus emphasizing network reliability enhancement. Therefore, the adjusted weight factors are used in the optimization algorithm, and the real-time network status is continuously monitored. Based on the monitoring results, the values ​​of k, Lthreshold, and Uthreshold are continuously fine-tuned to adapt to different network environments and requirements, ensuring that the algorithm always performs optimally under various conditions. Simultaneously, the system will generate a user interface to display a visual network status chart with accompanying data streams for more intuitive viewing. Users can manually adjust α and β based on the network status on this interface.

[0115] Step 103: During data transmission, real-time transmission performance indicators are collected, and the first collaborative optimization result is adjusted according to the transmission performance indicators and the network resource allocation algorithm, so that the merged encoded data packets are transmitted according to the adjusted strategy.

[0116] After data packet transmission begins, the system will collect transmission performance indicators in real time, update the parameters of the network resource allocation algorithm based on the collected performance indicator data, solve the updated algorithm to obtain the corresponding solution results, and adjust and update the first collaborative optimization result based on the solution results so that data packets are transmitted according to the updated strategy. This achieves adaptive feedback adjustment of the data packet transmission strategy, and adjusts the transmission strategy in real time based on the transmission performance indicators, so that the adjusted transmission strategy is more adaptable to the network state, further improving communication efficiency and the reliability of communication data transmission.

[0117] Specifically, the real-time acquisition of transmission performance indicators and the adjustment of the first collaborative optimization result based on the transmission performance indicators and the network resource allocation algorithm described in this embodiment are as follows:

[0118] During data packet transmission, the bandwidth utilization and transmission power efficiency in the transmission performance indicators are collected in real time, and the adjustment coefficient in the network resource allocation algorithm is obtained by calculating based on the bandwidth utilization and transmission power efficiency through a preset adaptive control algorithm.

[0119] The network resource allocation algorithm is updated with parameters based on the calculated adjustment coefficients, and the updated network resource allocation algorithm is solved. Then, the first collaborative optimization result is adjusted based on the solution result.

[0120] When the system transmits data packets based on the first collaborative optimization result, it will also collect the bandwidth utilization and transmission power efficiency in the network transmission performance indicators in real time. The collected data will be used as input data to the preset adaptive control algorithm for calculation and solution, and the corresponding calculation results will be used as adjustment coefficients in the network resource allocation algorithm to update the algorithm parameters accordingly. This will make the updated network resource allocation algorithm more adaptable to the current network transmission performance, thereby improving the reliability and practicality of the system's data transmission.

[0121] By adjusting the system's data transmission strategy using network transmission performance status data feedback, not only are data transmission efficiency, security, and reliability improved, but dynamic and flexible adjustments to the data transmission strategy and the system's self-adaptive capabilities are also achieved. Furthermore, the stability and response speed of the entire communication system are enhanced.

[0122] In this embodiment, the formula for the network allocation algorithm is:

[0123] R t =γ×BW(t)+δ×TP(t)

[0124] Therefore, the system will update the allocation strategy by adjusting the adjustment coefficients in the formula. The adaptive adjustment method of the adjustment coefficients provided in this embodiment first requires real-time collection of network transmission performance index data. In this embodiment, the network transmission performance index data is defined as bandwidth utilization rate U and transmission power efficiency E, and target values ​​are set for them as Utarget and Etarget, respectively.

[0125] Once determined, the collected transmission performance index data can be calculated and solved using an adaptive control algorithm based on performance feedback. The specific formula is shown below:

[0126] γ(t+1)=γ(t)+k1·(U target -U(t))

[0127] δ(t+1)=δ(t)+k2·(E target -E(t))

[0128] Here, k1 and k2 are adjustment coefficients, set according to the sensitivity and speed of system performance adjustment. The system continuously monitors real-time network performance indicators such as bandwidth utilization and transmission power efficiency, and automatically adjusts γ and δ based on the monitoring data to optimize resource allocation and meet performance targets. These two adjustment coefficients are also displayed on the user interface with accompanying data streams for a more intuitive viewing experience. Users can also manually adjust the parameters γ and δ on the interface according to network conditions.

[0129] To better illustrate the working principle and steps of the priority-based semantic communication optimization method, apparatus and system of the present invention, please refer to the relevant description above, but not limited to.

[0130] Accordingly, see Figure 2 , Figure 2 This is a schematic diagram of one embodiment of the priority-based semantic communication optimization device provided by the present invention. Figure 2 As shown, the semantic communication optimization device includes a priority calculation module 201, an allocation strategy transmission module 202, and an indicator feedback optimization module 203.

[0131] The priority calculation module 201 is used to receive and parse the data to be transmitted in response to the data transmission signal, obtain key semantic information, and perform priority calculation and semantic data merging encoding on the data to be transmitted based on the key semantic information to obtain information priority and merged encoded data packets.

[0132] Furthermore, the priority calculation module 201 receives and parses the data to be transmitted to obtain key semantic information, and performs priority calculation and semantic data merging encoding on the data to be transmitted based on the key semantic information to obtain information priority and merged encoded data packets, specifically:

[0133] The system receives and sequentially performs noise removal and standardization on the data to be transmitted, outputting the corresponding processed data. Based on the processed data, an autoencoder or convolutional neural network is invoked to extract features from the processed data to obtain the key semantic information. The key semantic information is divided into a training dataset and a test dataset, and the training dataset is input into the neural network model for training. The trained model then performs feature analysis and semantic importance evaluation on the test dataset to obtain semantic importance. Based on the semantic importance, a weighted algorithm is used to calculate the information priority of each key semantic information.

[0134] The merged encoded data packet is obtained by merging and encoding the data corresponding to information whose similarity reaches a similarity threshold in the key semantic information.

[0135] Furthermore, the priority calculation module 201 obtains the merged encoded data packet by merging and encoding the data corresponding to information whose similarity reaches a similarity threshold in the key semantic information, specifically as follows:

[0136] Based on the key semantic information, cosine similarity or Jaccard similarity coefficient is selected to measure the semantic similarity of each piece of information in the key semantic information, thereby obtaining multiple corresponding similarities;

[0137] The multiple similarities are compared with a preset similarity threshold. The data items corresponding to the key semantic information whose similarity is greater than or equal to the similarity threshold are clustered to obtain several data groups. The data items in the several data groups are sorted according to the information priority of the corresponding key semantic information. The sorted data items in each data group are merged and encoded using a preset Huffman coding method to obtain several merged and encoded data packets.

[0138] The allocation strategy transmission module 202 is used to calculate a first network resource allocation strategy based on a preset network resource allocation algorithm, through the information priority and the merged coded data packet, and then retrieve and process the first network resource allocation strategy through a collaborative optimization algorithm to obtain a first collaborative optimization result, and transmit the merged coded data packet according to the first collaborative optimization result.

[0139] Furthermore, the allocation strategy transmission module 202 calculates a first network resource allocation strategy based on a preset network resource allocation algorithm, using the information priority and the merged encoded data packet, specifically as follows:

[0140] The network load is calculated based on the number and amount of the merged encoded data packets. Simultaneously, the average network latency and packet loss rate are monitored and collected in real time. Then, the network congestion index is calculated based on the average network latency and the packet loss rate. Based on the network resource allocation algorithm, the first network resource allocation strategy is obtained by calculating and solving according to the network load, the network congestion index, and the information priority.

[0141] Furthermore, the allocation strategy transmission module 202 processes the first network resource allocation strategy based on network status indicators and a collaborative optimization algorithm to obtain a first collaborative optimization result, specifically:

[0142] The network latency and network utilization in the network status indicators are collected in real time, and an adjustment factor is calculated based on the network latency and network utilization using a preset activation function. The weight factor in the collaborative optimization algorithm is obtained by optimizing the calculation based on the adjustment factor. Then, the collaborative optimization algorithm is updated based on the calculated weight factor, and the first network resource allocation strategy is collaboratively optimized using the updated collaborative optimization algorithm to output the first collaborative optimization result.

[0143] The indicator feedback optimization module 203 is used to collect network transmission performance indicators in real time during data transmission, and adjust the first collaborative optimization result according to the network transmission performance indicators through the network resource allocation algorithm, so that the merged encoded data packet is transmitted according to the adjusted strategy.

[0144] Furthermore, the standard feedback optimization module 203 collects transmission performance indicators in real time, and adjusts the first collaborative optimization result according to the transmission performance indicators and the network resource allocation algorithm, specifically as follows:

[0145] During data packet transmission, the bandwidth utilization and transmission power efficiency in the transmission performance indicators are collected in real time. Based on the bandwidth utilization and transmission power efficiency, an adjustment coefficient in the network resource allocation algorithm is calculated using a preset adaptive control algorithm. The parameters of the network resource allocation algorithm are updated according to the calculated adjustment coefficient, and the updated network resource allocation algorithm is solved. Then, the first collaborative optimization result is adjusted according to the solution result.

[0146] Accordingly, embodiments of the present invention also provide a priority-based semantic communication optimization system, the semantic communication optimization system including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the priority-based semantic communication optimization system described above when processing the computer program.

[0147] Accordingly, embodiments of the present invention also provide a storage medium storing a computer program, which is called and executed by a processor to implement a priority-based semantic communication optimization system as described in any of the above embodiments.

[0148] In summary, this invention also provides a priority-based semantic communication optimization method, apparatus, and system. It obtains key semantic information by parsing the data to be transmitted, and calculates information priorities and merged encoded data packets based on this information. A first network resource allocation strategy is calculated based on the above two factors using a network resource allocation algorithm. This strategy is then optimized using network status indicators and a collaborative optimization algorithm. Data packets are obtained and transmitted according to the first collaborative optimization result. Simultaneously, the first collaborative optimization result is adjusted based on transmission performance indicators using the allocation algorithm, and data is transmitted according to the adjusted strategy. Information priority confirmation ensures the integrity and reliability of important semantic information in the data to be transmitted during data transmission. Merging encoded data packets reduces the amount of data transmitted. The resource allocation algorithm and collaborative optimization algorithm improve the practicality of the resource allocation strategy, enabling it to effectively respond to changes in the network environment.

[0149] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A priority-based semantic communication optimization method, characterized in that, Includes the following steps: In response to a data transmission signal, the system receives and parses the data to be transmitted to obtain key semantic information. Based on this key semantic information, it performs priority calculation and semantic data merging and encoding on the data to be transmitted to obtain information priority and merged encoded data packets. Specifically, it merges and encodes data corresponding to information in the key semantic information whose similarity reaches a similarity threshold to obtain merged encoded data packets. Based on the key semantic information, it selects cosine similarity or Jaccard similarity coefficient to measure the semantic similarity of each piece of information in the key semantic information, obtaining multiple similarities. It compares these multiple similarities with a preset similarity threshold, clusters the data items corresponding to the key semantic information whose similarity is greater than or equal to the similarity threshold, and obtains several data groups. It then sorts the data items in these data groups according to the information priority of the corresponding key semantic information. Finally, it merges and encodes the sorted data items in each data group using a preset Huffman coding method to obtain several merged encoded data packets. Based on a preset network resource allocation algorithm, a first network resource allocation strategy is calculated using the information priority and the merged coded data packet. Then, the first network resource allocation strategy is processed based on network status indicators and a collaborative optimization algorithm to obtain a first collaborative optimization result. Finally, the merged coded data packet is transmitted according to the first collaborative optimization result. During data transmission, transmission performance indicators are collected in real time, and the first collaborative optimization result is adjusted according to the transmission performance indicators and the network resource allocation algorithm, so that the merged encoded data packets are transmitted according to the adjusted strategy.

2. The priority-based semantic communication optimization method as described in claim 1, characterized in that, The process of receiving and parsing the data to be transmitted to obtain key semantic information, and then performing priority calculation and semantic data merging encoding on the data to be transmitted based on the key semantic information to obtain information priority and merged encoded data packets, specifically involves: The system receives and performs noise removal and standardization on the data to be transmitted in sequence, outputs the corresponding processed data, and uses an autoencoder or convolutional neural network to extract features from the processed data to obtain the key semantic information. The key semantic information is divided into a training dataset and a test dataset. The training dataset is then input into a neural network model for training. The trained model performs feature analysis and semantic importance assessment on the test dataset to obtain semantic importance. Based on the semantic importance, a weighted algorithm is used to calculate the information priority of each key semantic information.

3. The priority-based semantic communication optimization method as described in claim 1, characterized in that, The preset network resource allocation algorithm calculates a first network resource allocation strategy based on the information priority and the merged encoded data packets, specifically as follows: The network load is calculated based on the number and amount of the merged encoded data packets, while the average network latency and packet loss rate are monitored and collected in real time. Then, the network congestion index is calculated based on the average network latency and the packet loss rate. Based on the network resource allocation algorithm, the first network resource allocation strategy is obtained by calculating and solving according to the network load, the network congestion index, and the information priority.

4. The priority-based semantic communication optimization method as described in claim 1, characterized in that, The first network resource allocation strategy is processed based on network state indicators and a collaborative optimization algorithm to obtain a first collaborative optimization result, specifically as follows: The network latency and network utilization in the network status indicators are collected in real time, and an adjustment factor is calculated based on the network latency and network utilization using a preset activation function. The weight factors in the collaborative optimization algorithm are obtained by optimizing the adjustment factors. Then, the collaborative optimization algorithm is updated according to the calculated weight factors. The first network resource allocation strategy is then collaboratively optimized using the updated collaborative optimization algorithm, and the first collaborative optimization result is output.

5. The priority-based semantic communication optimization method as described in claim 1, characterized in that, The real-time acquisition of transmission performance indicators, and the adjustment of the first collaborative optimization result based on the transmission performance indicators and the network resource allocation algorithm, specifically includes: During data packet transmission, the bandwidth utilization and transmission power efficiency in the transmission performance indicators are collected in real time, and the adjustment coefficient in the network resource allocation algorithm is obtained by calculating based on the bandwidth utilization and transmission power efficiency through a preset adaptive control algorithm. The network resource allocation algorithm is updated with parameters based on the calculated adjustment coefficients, and the updated network resource allocation algorithm is solved. Then, the first collaborative optimization result is adjusted based on the solution result.

6. A priority-based semantic communication optimization device, characterized in that, The semantic communication optimization device includes a priority calculation module, an allocation strategy transmission module, and an indicator feedback optimization module. The priority calculation module is used to respond to the data transmission signal, receive and parse the data to be transmitted, obtain key semantic information, and perform priority calculation and semantic data merging encoding on the data to be transmitted according to the key semantic information to obtain information priority and merged encoded data packets. Specifically, the merged encoded data packets are obtained by merging and encoding the data corresponding to information in the key semantic information whose similarity reaches a similarity threshold. Cosine similarity or Jaccard similarity coefficient is selected based on the key semantic information to measure the semantic similarity of each piece of information in the key semantic information, obtaining multiple similarities. These multiple similarities are compared with a preset similarity threshold. Data items corresponding to the key semantic information whose similarity is greater than or equal to the similarity threshold are clustered to obtain several data groups. The data items in these data groups are then sorted according to the information priority of the corresponding key semantic information. Finally, the sorted data items in each data group are merged and encoded using a preset Huffman coding method to obtain several merged encoded data packets. The allocation strategy transmission module is used to calculate a first network resource allocation strategy based on a preset network resource allocation algorithm, through the information priority and the merged coded data packet, and then process the first network resource allocation strategy with a cooperative optimization algorithm to obtain a first cooperative optimization result, and transmit the merged coded data packet according to the first cooperative optimization result; The indicator feedback optimization module is used to collect network transmission performance indicators in real time during data transmission, and adjust the first collaborative optimization result according to the network transmission performance indicators and the network resource allocation algorithm, so that the merged encoded data packets are transmitted according to the adjusted strategy.

7. The priority-based semantic communication optimization device as described in claim 6, characterized in that, The priority calculation module receives and parses the data to be transmitted, obtains key semantic information, and performs priority calculation and semantic data merging encoding on the data to be transmitted based on the key semantic information to obtain information priority and merged encoded data packets, specifically: The system receives and performs noise removal and standardization on the data to be transmitted in sequence, outputs the corresponding processed data, and uses an autoencoder or convolutional neural network to extract features from the processed data to obtain the key semantic information. The key semantic information is divided into a training dataset and a test dataset. The training dataset is then input into a neural network model for training. The trained model performs feature analysis and semantic importance assessment on the test dataset to obtain semantic importance. Based on the semantic importance, a weighted algorithm is used to calculate the information priority of each key semantic information.

8. A priority-based semantic communication optimization system, characterized in that, The semantic communication optimization system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor processes the computer program, it implements a priority-based semantic communication optimization system as described in any one of claims 1-5.

9. A storage medium, characterized in that, The storage medium stores a computer program, which is called and executed by a processor to implement a priority-based semantic communication optimization system as described in any one of claims 1-5.

Images