Streaming media transmission optimization system

Abstract

The invention relates to the technical field of streaming media transmission, and discloses a streaming media transmission optimization system, which comprises an information acquisition module used for acquiring transmission environment information and transmission demand information in real time and performing accuracy detection and correction on the acquired information; and the analysis decision module adopts a simulation algorithm based on deep learning to accurately simulate cooperation characteristics among individuals in the microbial community, then generates a group cooperation strategy and an individual adaptive strategy, and performs weighted fusion on the group cooperation strategy and the individual adaptive strategy to obtain a final optimization decision. The optimization decision comprises a comprehensive adjustment scheme for each transmission parameter; the adjustment execution module is used for executing the optimization decision made by the analysis decision module; according to the method, the characteristics of cooperation, self-adaption and the like among individuals in the microbial community are used for reference, so that streaming media transmission can be automatically adjusted and optimized according to the environment and transmission requirements, and the transmission efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of streaming media transmission, and more specifically, it relates to a streaming media transmission optimization system. Background Art

[0002] With the popularization of the Internet and the development of mobile devices, streaming media services (such as video live streaming, online movies and TV shows, etc.) have become an indispensable part of people's daily lives. However, due to factors such as complex and changing network environments and diverse user needs, how to ensure the quality of streaming media transmission has become a difficult problem.

[0003] The technical problems faced mainly include:

[0004] Dynamic nature of the network environment: Parameters such as network bandwidth, latency, and packet loss rate fluctuate over time. Fixed transmission parameter settings are difficult to adapt to these changes, which may lead to problems such as video stuttering and blurred picture quality.

[0005] Diversity of user needs: Different users may be in different network conditions and have different expectations for video quality. For example, some users may value smoothness more, while others may focus more on high definition.

[0006] Accuracy of information collection: Accurately obtaining the current network environment and user demand information is a prerequisite for effective optimization. However, in practice, it is often interfered by various factors, affecting the accuracy of information collection.

[0007] Intelligence of optimization decision-making: How to make intelligent optimization decisions based on the collected information is a challenge. Traditional methods usually rely on manual experience or simple rules, lacking sufficient flexibility and adaptability.

[0008] Timeliness of execution adjustment: Even with optimization decisions, how to quickly and effectively execute these adjustments is also a problem that needs to be solved, especially in the case of high concurrency. Summary of the Invention

[0009] The present invention provides a streaming media transmission optimization system to solve the technical problems in the related art.

[0010] The present invention provides a streaming media transmission optimization system, including:

[0011] An information collection module, which is used to collect transmission environment information and transmission demand information in real time, and perform accuracy detection and correction on the collected information;

[0012] An analysis and decision-making module, which uses a simulation algorithm based on deep learning to accurately simulate the cooperative characteristics among individuals in the microbial community, and then generates a group cooperation strategy and an individual adaptive strategy It should be noted that there seems to be an error in the description of the "analysis and decision-making module" in the original text. It mentions simulating microbial community characteristics instead of relevant to streaming media transmission. The above translation is based on the provided text. Combine the group cooperation strategy and the individual adaptive strategy through weighted fusion to obtain the final optimized decision, which includes a comprehensive adjustment plan for each transmission parameter;

[0013] Adjust the execution module, which is used to execute the optimized decision made by the analysis and decision-making module.

[0014] Furthermore, the acquisition and correction of transmission environment information include:

[0015] Correct the transmission environment information by taking the average value through multiple acquisitions.

[0016] Furthermore, the acquisition and correction of transmission requirement information:

[0017] The video resolution is R, the frame rate is F, the bit rate is M, and the transmission requirement information is represented as a vector Correct the transmission requirement information by taking the average value through multiple acquisitions.

[0018] Furthermore, the analysis and decision-making module specifically executes the following steps:

[0019] Data preprocessing: Normalize the corrected transmission environment information vector and the transmission requirement information vector so that their component values are mapped to the interval [0, 1];

[0020] Feature extraction: Use a convolutional neural network to extract features from the preprocessed and to obtain their high-level feature representations and

[0021] Feature fusion: Fuse the extracted environment features and requirement features to obtain a comprehensive feature representation

[0022] Group cooperation simulation: Use a deep neural network composed of L fully connected layers to simulate the group cooperation behavior in the microbial community;

[0023] Take the comprehensive feature as the input, and after L layers of non-linear transformation, the number of neurons in each layer is H, and output the simulated group cooperation strategy

[0024] which represents an overall adjustment strategy for transmission parameters according to the current network environment and transmission requirements;

[0025] Individual adaptive simulation: Based on the group cooperation strategy , use a deep neural network with L' layers to simulate the adaptive behavior of microbial individuals;

[0026] Take and together as the input, after passing through the non-linear transformation of L' layers, the number of neurons in each layer is H', and output the simulated individual adaptive strategy

[0027] It represents a targeted fine-tuning strategy for individual transmission parameters according to different video content and user terminal characteristic factors on the basis of the overall transmission strategy ;

[0028] Strategy fusion and decision-making: Weightedly fuse the group cooperation strategy and the individual adaptive strategy to obtain the final optimized decision Decision. The fusion weights α and β are obtained through training and learning. The mathematical expression of decision fusion is:

[0029]

[0030] Among them, Decision represents the final transmission optimization strategy obtained by comprehensively considering the two levels of group cooperation and individual adaptation.

[0031] Furthermore, the fusion method adopts concatenation or attention mechanism.

[0032] Furthermore, use the method of reinforcement learning to train and analyze the decision-making module;

[0033] Take the decision-making module as the agent, whose state is the current transmission environment information and the transmission requirement information , and the action is the optimized decision Decision output. The reward is the quantified value of the improvement in transmission efficiency after executing the optimized decision;

[0034] Through the continuous interaction between the agent and the environment, use the reinforcement learning algorithm to optimize the model parameters θ so that it takes the optimal action according to the state to obtain the maximum reward.

[0035] Furthermore, the objective function of training is expressed as:

[0036]

[0037] Among them represents the expectation, π θ is the decision-making strategy function with parameter θ, γ t is the discount factor of the reward, rt is the reward obtained after performing the action at the t-th step. By maximizing the objective function J(θ), the model learns the optimal decision-making strategy, and T is the total number of training steps.

[0038] Further, when adjusting the execution module to execute the optimized decision made by the analysis and decision-making module, a buffering mechanism is adopted to reduce the delay of adjustment execution. Let the buffering time be t, the system state be SystemStatus, and the adjustment execution function be k. Then the adjustment execution process is expressed as:

[0039] AdjustedParams = k(Decision, t, SystemStatus)

[0040] where Decision is the optimized decision made by the analysis and decision-making module, and AdjustedParams are the streaming media transmission parameters after adjustment execution;

[0041] The adjustment execution module first temporarily stores the optimized decision in the buffer, and then quickly executes the parameter adjustment when the buffering time ends and the system state meets the requirements.

[0042] The present invention provides a method for optimizing streaming media transmission, which performs the following steps:

[0043] Collect the transmission environment information and transmission requirement information multiple times, and correct and obtain more accurate and

[0044] The analysis and decision-making module combines the corrected and to simulate the characteristics of the microbial community and make an optimized decision Decision;

[0045] The adjustment execution module temporarily stores the optimized decision Decision in the buffer, and when the buffering time t ends and the system state SystemStatus meets the requirements, executes the parameter adjustment through the function k to obtain the optimized streaming media transmission parameters AdjustedParams.

[0046] The present invention provides a computer-readable storage medium, which stores computer-readable instructions that can execute the aforementioned method for optimizing streaming media transmission when read by a computer.

[0047] The beneficial effect of the present invention is that it draws on the characteristics of cooperation and self-adaptation among individuals in the microbial community, enabling the streaming media transmission to automatically adjust and optimize according to the environment and transmission requirements, thereby improving the transmission efficiency. Brief Description of the Drawings

[0048] Figure 1It is a schematic diagram of the modules of the streaming media transmission optimization system of the present invention;

[0049] Figure 2 It is an example diagram of the transmission environment information and transmission requirement information of the present invention;

[0050] Figure 3 It is an example diagram of the adjustment of transmission parameters of the present invention. Detailed implementation manners

[0051] Now, the subject matter described herein will be discussed with reference to example embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein. Without departing from the scope of protection of the content of this specification, changes can be made to the functions and arrangements of the elements discussed. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described in some examples can also be combined in other examples.

[0052] In at least one embodiment of the present invention, a streaming media transmission optimization system is disclosed, as Figure 1 shown, including an information collection module, an analysis and decision-making module, and an adjustment and execution module, where:

[0053] I. Information collection module

[0054] The information collection module is responsible for collecting the transmission environment information and transmission requirement information in real time, and detecting and correcting the collected information for accuracy.

[0055] The information collection module specifically performs the following steps:

[0056] (1) Collection and correction of transmission environment information

[0057] The network bandwidth is B, the delay is D, and the packet loss rate is L. The transmission environment information is represented as a vector To improve the collection accuracy, the number of collections n is set, and the collected network bandwidth is corrected through the following formula:

[0058]

[0059] where B i is the network bandwidth value collected for the i-th time, and B′ is the corrected network bandwidth value. Similarly, other environmental information such as the delay D and the packet loss rate L is corrected in a similar manner.

[0060] (2) Collection and correction of transmission requirement information

[0061] The video resolution is R, the frame rate is F, and the bit rate is M. The transmission requirement information is represented as a vector Similarly, to improve the acquisition accuracy, the transmission requirement information is corrected by adopting the method of taking the average of multiple acquisitions similar to the transmission environment information.

[0062] II. Analysis and Decision-making Module

[0063] When the analysis and decision-making module conducts analysis and decision-making by drawing on the characteristics of the microbial community, a simulation algorithm based on deep learning is adopted to accurately simulate the characteristics such as the cooperation and self-adaptation among individuals in the microbial community.

[0064] The analysis and decision-making module specifically executes the following steps:

[0065] Data preprocessing: Normalize the corrected transmission environment information vector and the transmission requirement information vector so that the values of their respective components are mapped to the interval [0, 1], which is convenient for subsequent neural network training and prediction. The normalization function can choose linear normalization or non-linear normalization methods.

[0066] Feature extraction: Use a convolutional neural network (CNN) to extract features from the preprocessed and to obtain their high-level feature representations and

[0067] The structure of the CNN is designed as the superposition of N convolutional layers and pooling layers. The number of convolutional kernels in each layer is K, the size of the convolutional kernel is W×W, the pooling method is max pooling, and the size of the pooling kernel is P×P to fully extract the key features of the transmission environment and requirements.

[0068] Feature fusion: Fuse the extracted environment features and requirement features to obtain a comprehensive feature representation

[0069] The fusion method can adopt simple concatenation, or more advanced fusion methods such as the attention mechanism. If the attention mechanism is adopted, M attention heads can be set, and the dimension of each head is d k , and feature fusion is achieved through weighted summation.

[0070] Group cooperation simulation: Use a deep neural network composed of L fully connected layers to simulate the group cooperation behavior in the microbial community.

[0071] Input the comprehensive features As input, it undergoes L layers of non-linear transformation, with the number of neurons in each layer being H. Activation functions such as ReLU can be selected, and the output is a simulated group cooperation strategy.

[0072] It represents an overall adjustment strategy for transmission parameters (such as coding method, resolution, frame rate, etc.) according to the current network environment and transmission requirements.

[0073] Group cooperation strategy What is simulated is the adaptive adjustment made by the entire microbial community to environmental changes, such as changing the ways of material exchange and information transmission within the community to improve the survival ability of the entire community.

[0074] Individual adaptive simulation: Based on the group cooperation strategy On this basis, a deep neural network with L' layers is used to simulate the adaptive behavior of microbial individuals.

[0075] Taking and as inputs, after L' layers of non-linear transformation, with the number of neurons in each layer being H', activation functions such as ReLU can be selected, and the output is a simulated individual adaptive strategy.

[0076] It represents a targeted fine-tuning strategy for individual transmission parameters (such as bit rate, pre-buffering time, etc.) according to factors such as different video contents and user terminal characteristics on the basis of the overall transmission strategy. On this basis, according to different video contents, user terminal characteristics and other factors, a targeted fine-tuning strategy for individual transmission parameters (such as bit rate, pre-buffering time, etc.) is carried out.

[0077] It simulates that microbial individuals fine-tune their behaviors according to their own situations, such as adjusting the metabolic rate, movement mode, etc., to better adapt to the environment.

[0078] Strategy fusion and decision-making: The group cooperation strategy and the individual adaptive strategy are weighted and fused to obtain the final optimized decision Decision. The fusion weights α and β are obtained through training and learning. The mathematical expression of decision fusion is:

[0079]

[0080] Among them, Decision represents the final transmission optimization strategy obtained by comprehensively considering the two levels of group cooperation and individual adaptation.

[0081] It includes a comprehensive adjustment plan for each transmission parameter, such as how to select the coding format and parameters, how many layers to transmit, what resolution and frame rate to use for each layer, how large the buffer is set, etc.

[0082] The values of α and β determine the relative importance of the overall group strategy and the individual fine-tuning strategy in the final decision-making, which can be set through training optimization.

[0083] Model training: Since it is difficult to directly obtain the theoretically optimal transmission strategy under the current environment and requirements, a reinforcement learning method is adopted to train the analysis and decision-making module.

[0084] The analysis and decision-making module is regarded as an agent, and its state is the current transmission environment information and the transmission requirement information The action is the output optimization decision Decision, and the reward is the quantified value of the improvement in transmission efficiency after executing the optimization decision. Through the continuous interaction between the agent and the environment, reinforcement learning algorithms such as Q-learning and policy gradient are used to optimize the model parameters θ, enabling it to learn to take the optimal action according to the state (i.e., make the best optimization decision) to obtain the maximum reward (i.e., maximize the transmission efficiency). The objective function of model training is expressed as:

[0085]

[0086] where π θ is the decision-making policy function with parameter θ, γ is the discount factor of the reward, and r t is the reward obtained after executing the action at the t-th step. By maximizing the objective function J(θ), the model learns the optimal decision-making policy.

[0087] During the training process, the ∈-greedy exploration strategy can be used, that is, randomly select actions with a probability of ∈, and execute actions according to the decisions output by the model with a probability of 1 - ∈, so as to achieve a balance between exploration and exploitation. As the training progresses, gradually reduce the value of ∈, so that the model makes decisions more and more based on the learned strategy.

[0088] Through multiple rounds of interaction between the agent and the environment, continuously optimize the model parameters, and finally obtain a reinforcement learning model that can output the optimal transmission decision according to the transmission environment and requirement information. This training method overcomes the difficulty of obtaining the theoretically optimal decision, enabling the model to master the optimization strategy of streaming media transmission through autonomous learning.

[0089] Online prediction: Deploy the trained analysis and decision-making module into the streaming media transmission system, and receive the and provided by the information acquisition module in real time. After the above steps of calculation, output the optimization decision Decision to guide the subsequent adjustment and execution.

[0090] In summary, the analysis and decision-making module simulates the group collaboration and individual adaptation characteristics of the microbial community through a carefully designed deep learning model and algorithm. It can make intelligent optimization decisions based on the real-time collected transmission environment and demand information, thus achieving self-optimization of streaming media transmission.

[0091] III. Adjustment and Execution Module

[0092] When the adjustment and execution module executes the optimization decision made by the analysis and decision-making module, it adopts a buffering mechanism to reduce the latency of adjustment and execution.

[0093] Let the buffering time be t, the system status be SystemStatus, and the adjustment and execution function be k. Then the adjustment and execution process is expressed as:

[0094] AdjustedParams = k(Decision, t, SystemStatus)

[0095] Among them, Decision is the optimization decision made by the analysis and decision-making module, and AdjustedParams is the streaming media transmission parameter after adjustment and execution.

[0096] The adjustment and execution module first temporarily stores the optimization decision in the buffer. When the buffering time t ends and the system status meets the requirements, it quickly executes the parameter adjustment to reduce the latency of adjustment and execution and improve the optimization timeliness.

[0097] Some embodiments of the present invention provide a method for optimizing streaming media transmission, including the following steps:

[0098] The information acquisition module collects the transmission environment information and transmission demand information multiple times according to the set number of acquisitions n, and corrects and obtains more accurate and

[0099] The analysis and decision-making module uses the model Model trained by deep learning, combines the corrected and and makes an optimization decision Decision by simulating the characteristics of the microbial community through the function h.

[0100] The adjustment and execution module temporarily stores the optimization decision Decision in the buffer. When the buffering time t ends and the system status SystemStatus meets the requirements, it executes the parameter adjustment through the function k to obtain the optimized streaming media transmission parameter AdjustedParams.

[0101] After completing this round of optimization, return to step 1 to re-collect information and start a new round of self-optimization process to achieve continuous self-optimization of streaming media transmission.

[0102] Through the above-mentioned self-optimizing mechanism for streaming media transmission, the efficiency of streaming media transmission can be effectively improved, enabling it to automatically adjust and optimize according to changes in the network environment and transmission requirements. Specifically, it is manifested as smoother transmission, more suitable picture quality and frame rate, etc., and it can operate stably under actual difficulties such as inaccurate information collection, difficult simulation of microbial community characteristics, and delay in adjustment execution.

[0103] The self-optimizing mechanism for streaming media transmission of the present invention well draws on the excellent characteristics of microbial communities, provides a new idea and method for improving the quality of streaming media transmission, and has strong practical value and innovation.

[0104] To more intuitively illustrate the working process and optimization effect of the present invention, the following will be elaborated through an application embodiment in a multi-user complex scenario.

[0105] Suppose a video live streaming platform needs to provide a high-definition live broadcast of a 2-hour sports event to 1000 users simultaneously. The users are distributed in different geographical locations and use various network environments and receiving devices. The initial transmission parameters are set as: encoding format H.265, resolution 1080P, frame rate 50FPS, GOP length 25 frames, number of reference frames 1, and bit rate 8Mbps.

[0106] During the live broadcast, due to the dynamic changes in the network conditions and demands of users, as well as the influence of factors such as sudden network congestion, the live broadcast experience of each user has declined to varying degrees, such as Figure 2 shown:

[0107] In the face of such a complex multi-user transmission scenario, the self-optimizing mechanism for streaming media transmission begins to play a role:

[0108] The information collection module continuously collects transmission environment parameters such as the network bandwidth, latency, and packet loss rate of each user, as well as feedback information such as video stuttering, blurring, and desynchronization. Considering the large number of users, the collection frequency is set to once every 10 seconds, and the average value is taken after continuous collection for 1 minute to obtain the corrected user environment information matrix I env ′ and demand information matrix I req ′.

[0109] The analysis and decision-making module inputs I env ′ and I req ′ into a pre-trained deep reinforcement learning model, and the model outputs online the group cooperation strategy for all users as well as the individual adaptive strategy for different user groups

[0110] Specifically, the group cooperation strategy The encoding format was adjusted to H.264, the frame rate was set to 30 FPS, the GOP length was 50 frames, and the bit rate was 4 Mbps to adapt to the network conditions of most users.

[0111] On this basis, the individual adaptive strategy was further adjusted specifically: for user group A, the resolution was lowered to 720P to reduce lags; for user group B, the number of reference frames was increased to 3 to improve image quality; for user group C, the audio-visual synchronization parameter settings were refined to optimize synchronization; for user group D, the original parameters were kept unchanged.

[0112] Then, the group collaborative strategy and the individual adaptive strategy were weighted and fused according to the following formula to obtain the final optimized decision Decision i :

[0113]

[0114] where α and β i are the fusion weights, and the result parameters obtained from training are α = 0.6, β A = 0.4, β B = 0.3, β C = 0.4, β D = 0.1. While taking into account the global optimization, a certain amount of personalized adjustment space is given to different user groups.

[0115] The optimized decision Decision i after fusion corresponds to the transmission parameter adjustments as Figure 3 shown:

[0116] Figure 3 It shows the optimized adjustment results of the transmission parameters for different user groups after fusing the group collaborative strategy and the individual adaptive strategy. It can be seen that the resolution of user group A is appropriately reduced, the number of reference frames of user group B is appropriately increased, the audio-visual synchronization parameters of user group C are enhanced specifically, and user group D maintains the original parameter settings. These adjustments are further personalized optimizations for the characteristics of different user groups on the basis of global optimization.

[0117] The adjustment execution module first temporarily stores the group collaborative strategy in the global buffer for 5 seconds, and then quickly updates it to all users; then for each user group, it temporarily stores its optimized decision Decision i in the local buffer for 3 seconds and then updates it to the corresponding users. The entire hierarchical adjustment execution process is completed within 10 seconds.

[0118] After the transmission optimization, the user satisfaction has generally improved. The number of lags for user group A has dropped to 1 time per minute, and the satisfaction has increased to 3 points; the picture quality of user group B has been significantly improved, and the satisfaction has increased to 4 points; the audio-video synchronization problem of user group C has been basically solved, and the satisfaction has increased to 4 points; user group D has remained stable, and the satisfaction has remained at 4 points. The overall average satisfaction has increased from 2 points to 3.75 points.

[0119] In subsequent live broadcasts, the self-optimization mechanism of streaming media transmission continuously monitors the network conditions and feedback of multiple users. Global optimization is performed every 2 minutes, and personalized optimization within user groups is performed every 30 seconds. During the entire 2-hour live event, despite the complex and changing environment, a good and stable live broadcast experience has been maintained, receiving wide acclaim from users.

[0120] From this embodiment of the complex multi-user scenario, it can be seen that the self-optimization mechanism of streaming media transmission described in the present invention can make full use of the deep reinforcement learning intelligent algorithm to perform real-time information collection, analysis and decision-making, and adjustment and execution for a large number of users, and output a "one-to-many" hierarchical optimization strategy. While ensuring the global transmission effect, it takes into account the personalized needs of different users, greatly improves the user experience, and demonstrates significant technical advantages and application values. In the actual deployment and application of a streaming media service platform, adopting the self-optimization mechanism of the present invention can cope with complex transmission scenarios with high concurrency and dynamic changes, maximize the service quality, and provide strong support for the intelligent development of the streaming media industry.

[0121] In some embodiments of the present invention, a computer-readable medium is provided, which stores computer-readable instructions that can execute the aforementioned streaming media transmission optimization method when read by a computer.

[0122] The above describes the embodiments of the present invention, but these embodiments are not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative and not restrictive. Under the inspiration of these embodiments, those of ordinary skill in the art can also make more equivalent embodiments in various forms, all of which fall within the protection scope of these embodiments.

Claims

1. A streaming media transmission optimization system, characterized in that: include: An information collection module, which is used to collect transmission environment information and transmission demand information in real time, and to detect and correct the accuracy of the collected information; The analysis and decision-making module uses a deep learning-based simulation algorithm to accurately simulate the collaborative characteristics of individuals in the microbial community and then generate a group collaboration strategy and individual adaptive strategies Group coordination strategy and individual adaptive strategies Perform weighted fusion to obtain the final optimization decision, which includes a comprehensive adjustment plan for each transmission parameter; The adjustment execution module is used to execute the optimization decision made by the analysis decision module.

2. A streaming media transmission optimization system according to claim 1, characterized in that: The collection and correction of transmission environment information includes: The transmission environment information is corrected by taking the average value through multiple collections.

3. A streaming media transmission optimization system according to claim 1, characterized in that: Collection and correction of transmission requirement information: The video resolution is R, the frame rate is F, the bit rate is M, and the transmission requirement information is represented as a vector The transmission demand information is corrected by taking the average value of multiple collections.

4. A streaming media transmission optimization system according to claim 1, characterized in that: The analysis and decision-making module specifically performs the following steps: Data preprocessing: Transmit the corrected transmission environment information vector and the transmission demand information vector Perform normalization processing so that each component value is mapped to the interval [0,1]; feature Extraction: Use convolutional neural network to extract the preprocessed and Perform feature extraction to obtain its high-level feature representation and Feature fusion: The extracted environmental features and demand characteristics Fusion to obtain comprehensive feature representation Group collaboration simulation: Use a deep neural network consisting of L fully connected layers to simulate group collaboration behaviors in microbial communities; The comprehensive features As input, after L layers of nonlinear transformation, the number of neurons in each layer is H, and the output is the simulated group coordination strategy Indicates the overall adjustment strategy of transmission parameters according to the current network environment and transmission requirements; Individual adaptive simulation: In group coordination strategy Based on this, a deep neural network with L′ layers is used to simulate the adaptive behavior of microbial individuals. Will and As input, after L' layers of nonlinear transformation, the number of neurons in each layer is H', and the output is the simulated individual adaptive strategy Indicates the overall transmission strategy Based on the above, we can fine-tune the individual transmission parameters according to different video contents and user terminal characteristics. Strategy Fusion and Decision-making: Group Collaboration Strategy and individual adaptive strategies Perform weighted fusion to obtain the final optimized decision. The fusion weights α and β are obtained through training and learning. The mathematical expression of decision fusion is: Among them, Decision represents the final transmission optimization strategy obtained after comprehensive consideration of both group collaboration and individual adaptation.

5. A streaming media transmission optimization system according to claim 4, characterized in that: The fusion method uses concatenation or attention mechanism.

6. A streaming media transmission optimization system according to claim 1, characterized in that: Use reinforcement learning methods to train analysis and decision-making modules; The analysis and decision module is used as an intelligent agent, and its state is the current transmission environment information. and transmission requirements information The action is the output optimization decision Decision, and the reward is the quantitative value of the transmission efficiency improvement after executing the optimization decision; Through the continuous interaction between the agent and the environment, the reinforcement learning algorithm is used to optimize the model parameters θ so that it can take the optimal action according to the state to obtain the maximum reward.

7. A streaming media transmission optimization system according to claim 6, characterized in that: The objective function of training is expressed as: in represents expectation, π θ is the decision strategy function with parameter θ, γ t is the discount factor of the reward, r t It is the reward obtained after executing the action in the tth step. By maximizing the objective function J(θ), the model learns the optimal decision-making strategy. T is the total number of training steps.

8. A streaming media transmission optimization system according to claim 1, characterized in that: When the adjustment execution module executes the optimization decision made by the analysis and decision module, a buffer mechanism is used to reduce the delay of adjustment execution. Assuming the buffer time is t, the system status is SystemStatus, and the adjustment execution function is k, the adjustment execution process is expressed as: AdjustedParams=k(Decision,t,SystemStatus) Among them, Decision is the optimization decision made by the analysis and decision module, and AdjustedParams is the streaming media transmission parameters after adjustment and execution; The adjustment execution module first temporarily stores the optimization decision in the buffer, and then quickly executes the parameter adjustment when the buffer time ends and the system status meets the requirements.

9. A method for optimizing streaming media transmission, characterized in that: The streaming media transmission optimization system according to any one of claims 1 to 8 performs the following steps: Collect transmission environment information and transmission demand information multiple times, and correct them by taking the average value to get a more accurate and The analysis and decision module is combined with the revised and Simulate the characteristics of microbial communities to make optimal decisions; The adjustment execution module temporarily stores the optimization decision Decision in the buffer. When the buffer time t ends and the system status SystemStatus meets the requirements, the parameter adjustment is performed through the function k to obtain the optimized streaming media transmission parameters AdjustedParams.

10. A computer-readable storage medium, characterized in that: It stores computer-readable instructions, which, when read by a computer, can execute the streaming media transmission optimization method as claimed in claim 9.

Images