Intelligent live broadcast scene adaptive control system and method

Abstract

The invention discloses an intelligent live broadcast scene adaptive control system, and relates to the technical field of scene control. The system comprises a scene perception module, a data analysis module, a strategy generation module, an execution control module and a feedback evaluation module. The scene sensing module collects multi-dimensional information by using a multi-sensor fusion technology, evaluates integrity by using an information integrity formula, and has a self-calibration function; the data analysis module extracts features by using a deep learning algorithm, and evaluates association through a feature association degree formula; the strategy generation module is combined with an expert system and an MPC to generate a strategy; the execution control module executes a strategy in a distributed architecture mode and adjusts equipment parameters; and the feedback evaluation module evaluates the effect and the system performance by using multiple indexes, and can perform trend analysis and early warning. According to the invention, live broadcast quality is improved, scene perception is accurate, and strategy generation is adaptive; network problems are effectively handled, and smoothness is guaranteed. The efficiency is improved, the cost is reduced, manpower is reduced through automatic control, performance is fed back, evaluated and optimized, expansibility is achieved, and development of the live broadcast industry is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of scene control systems, and in particular to an intelligent live broadcast scene adaptive control system and method. Background Art

[0002] In today's digital age, the live broadcast industry has shown explosive growth and has become an important way for people to obtain information, interact with entertainment, and conduct commercial marketing. From e-commerce sales, game competitions to knowledge sharing and talent shows, the application scenarios of live broadcasts are becoming increasingly rich and diverse. However, with the vigorous development of the live broadcast industry, a series of problems that need to be solved have also been exposed, which have seriously restricted the improvement of live broadcast quality and the optimization of user experience.

[0003] At present, the control of live broadcast scenes mostly relies on manual operation. The anchor or operator needs to adjust the parameters of the live broadcast equipment based on experience and intuition, such as the angle of the camera, the brightness of the light, the volume of the microphone, etc. This method is not only inefficient, but also prone to operational errors, resulting in unstable live broadcast picture quality, poor sound effects and other problems. For example, in some large-scale live broadcast activities, due to the limitations of manual operation, it is difficult to respond to emergencies in a timely and accurate manner, such as light changes, personnel movement, etc., which affects the overall effect of the live broadcast. In addition, different live broadcast scenes have different requirements for equipment parameters and content presentation. In indoor live broadcasts, the lights and cameras need to be adjusted according to the layout and light conditions of the room; while in outdoor live broadcasts, factors such as weather and environmental noise need to be considered. The existing live broadcast system lacks the ability to adapt to different scenes and cannot automatically adjust the equipment parameters and content strategies according to the changes in the scenes, which greatly reduces the quality and attractiveness of the live broadcast content.

[0004] In terms of user experience, existing live broadcast systems often ignore the personalized needs of viewers and anchors. Different users have different interests and viewing habits, and have their own preferences for the style of live broadcast images, the type of sound effects, the way of interaction, etc. However, most of the current live broadcast platforms provide the same live broadcast content, which cannot meet the personalized needs of users, resulting in low user participation and loyalty.

[0005] Network conditions are also an important factor affecting the quality of live broadcasts. During live broadcasts, network bandwidth fluctuations, delays, and packet loss often occur, which can easily lead to live broadcast freezes, blurred images, and sound interruptions. Existing live broadcast systems lack effective optimization strategies when dealing with network problems, and are unable to adjust live broadcast parameters in real time according to network conditions to ensure the smoothness of live broadcasts. Moreover, with the increasingly fierce competition in the live broadcast industry, how to improve the efficiency of live broadcasts and reduce costs is also an important challenge facing live broadcast platforms and anchors. Traditional live broadcast methods require a lot of manpower, material resources, and time for equipment debugging and scene layout, which increases the cost and difficulty of live broadcasts.

[0006] Therefore, it is of great practical significance to develop an intelligent live broadcast scene adaptive control system and method. It can automatically sense the changes in the live broadcast scene, adjust the device parameters and content strategy in real time, meet the personalized needs of users, cope with the fluctuation of network conditions, improve the quality and efficiency of live broadcast, reduce the cost of live broadcast, and thus promote the healthy development of the live broadcast industry. Summary of the invention

[0007] The intelligent live broadcast scene adaptive control system and method proposed in the present invention are used to solve the problems mentioned in the above-mentioned prior art.

[0008] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: an intelligent live scene adaptive control system, comprising: Scene perception module: uses sensor fusion technology to integrate sensors, audio sensors and environmental sensors; collects the picture, sound, light, temperature, humidity and air quality information of the live scene in real time; and uses the information completeness formula to Assess the completeness of collected information, including is the amount of information collected, The total amount of information to be collected; Data analysis module: Use convolutional neural network CNN and recurrent neural network RNN ​​to analyze the collected information; use feature correlation formula Evaluate the degree of association between features, where is the strength of the association relationship of the i-th feature; Strategy generation module: Combines the expert system and the model predictive control (MPC) method to generate control strategies for live broadcast scenarios. The expert system stores knowledge and empirical rules for live broadcast scenarios and provides preliminary strategy recommendations based on the results of the data analysis module. MPC optimizes control strategies by building a dynamic model of the live broadcast scenario and predicting the changing trend of the scenario in the future. Execution control module: adopts a distributed control architecture to transmit control instructions to live broadcast equipment; real-time feedback mechanism to monitor the execution status of the equipment in real time, responsible for executing the generated strategy, and adjusting the parameters of the live broadcast equipment, including the angle, focal length, aperture of the camera, the brightness, color, angle of the light, and the gain and directivity of the microphone; Feedback evaluation module: Construct a comprehensive evaluation feedback mechanism, in addition to the effect evaluation formula In addition to evaluating the effectiveness of the control strategy, user satisfaction and resource utilization indicators are also considered. In order to actually achieve the effect indicators, The target effect indicator.

[0009] Furthermore, the following modules are also included: User preference module: Utilizes natural language processing technology and machine learning algorithms to collect preference information of live broadcast viewers and hosts by analyzing users’ comments, likes, and sharing behaviors on the live broadcast platform, as well as their interaction records with hosts. Assess the degree of preference matching, where is the number of preferences matched, is the total number of preferences.

[0010] Real-time optimization module: It uses real-time data mining and adaptive adjustment algorithms to monitor the network status during live broadcast, including bandwidth B, delay L, packet loss rate P and jitter J. Evaluate network quality, including is the weight coefficient.

[0011] Furthermore, the user preference module also analyzes the changes in user preferences in different time periods and live broadcast topics; through the preference change coefficient formula Assess the degree of preference change, where and are the frequencies of the i-th preference in different time periods or topics, and m is the number of preferences.

[0012] Furthermore, when the network quality index is lower than the set threshold, the execution control module adopts a hierarchical cache strategy; through the cache priority formula Prioritize cached content, where is the weight coefficient, R is the content importance level, I is the immediacy demand, and U is the user attention.

[0013] Furthermore, the strategy generation module adopts a strategy generation algorithm based on reinforcement learning. The algorithm approximates the value function and the strategy function through a deep neural network DNN, and uses the reward function Evaluate the reward value of taking action a in state s and optimize the policy by maximizing the cumulative reward.

[0014] Furthermore, in addition to evaluating the execution effect, the feedback evaluation module also evaluates the accuracy of the scene perception module; through the perception accuracy formula Assessing perceived accuracy, where is the amount of correctly perceived information, is the total amount of perceived information.

[0015] A method for applying the intelligent live scene adaptive control system, wherein the scene perception module uses sensor fusion technology and modal fusion algorithm to collect information of the live scene in real time, uses the information integrity formula to evaluate the integrity of the collected information, and ensures the accuracy of the information through the self-calibration function; the data analysis module uses CNN and RNN to analyze the collected information, extracts scene features, uses the feature correlation formula to evaluate the feature correlation degree, and performs anomaly detection at the same time; the strategy generation module combines the expert system and MPC method to generate a control strategy according to the analysis results and preset rules, and uses the recommendation algorithm combined with user preferences; the execution control module uses a distributed control architecture and a real-time feedback mechanism to execute the strategy, adjust the parameters of the live equipment and the control platform functions; the feedback evaluation module constructs a comprehensive evaluation mechanism to evaluate the execution effect and system performance.

[0016] Furthermore, the user preference module uses natural language processing and machine learning algorithms to collect and analyze user preference information, evaluates the degree of preference matching through the preference matching formula, and makes preference predictions; the strategy generation module considers user preferences when generating strategies, and uses a dynamic programming algorithm to adjust strategies according to preference changes; The real-time optimization module uses real-time data mining and adaptive adjustment algorithms to monitor network conditions, uses the network quality index formula to evaluate network quality, and uses diversion and cache pre-fetching technology to adjust strategies when network quality deteriorates.

[0017] Furthermore, the user preference module uses time series analysis and cluster analysis methods to analyze the changes in user preferences over time and over topics, and evaluates the degree of change through the preference change coefficient formula; when the network quality index is lower than the threshold, the execution control module adopts a hierarchical caching strategy and determines the cache content priority based on the cache priority formula; the feedback evaluation module uses Bayesian reasoning and evidence theory to evaluate the accuracy of the scene perception module. If it is lower than the threshold, a genetic algorithm is used to send adjustment instructions to optimize the acquisition strategy.

[0018] Compared with the prior art, the present invention has the following beneficial effects: In terms of improving the quality of live broadcast, the scene perception module can comprehensively and accurately collect live broadcast scene information through multi-sensor fusion and self-calibration functions, providing a solid foundation for subsequent data analysis and strategy generation. The data analysis module uses deep learning algorithms for in-depth analysis and anomaly detection, so that the strategy generation module can accurately generate control strategies that adapt to different scenarios, effectively improving the clarity of the live broadcast picture, the quality of the sound, and the overall coordination of the scene.

[0019] For user experience, the user preference module can deeply explore the personalized needs of users and provide customized live broadcast content and interactive strategies for different users through personalized recommendation algorithms, which greatly improves user participation and satisfaction. At the same time, the strategy is dynamically adjusted according to changes in user preferences to make the live broadcast more in line with the interests and habits of users.

[0020] In terms of dealing with network problems, the real-time optimization module monitors the network status in real time and uses intelligent diversion, cache pre-fetching and hierarchical cache strategies to effectively deal with network bandwidth fluctuations, delays and packet loss, ensuring the smoothness of live broadcasts. Even in the case of poor network quality, it can prioritize the transmission of important data and quickly restore the quality of live broadcasts.

[0021] From the perspective of efficiency and cost, the system adopts an automated control method to reduce the tediousness and mistakes of manual operation and improve the efficiency of live broadcast. At the same time, through intelligent strategy generation and resource allocation, it reduces the manpower, material and time costs in the live broadcast process.

[0022] The multi-index comprehensive evaluation and trend analysis functions of the feedback evaluation module can timely discover system problems and give early warnings and improvement suggestions, continuously optimizing system performance and stability. Moreover, the system also has good scalability and adaptability, and can be continuously upgraded and improved with the development of the live broadcast industry and changes in user needs, injecting new vitality into the development of the live broadcast industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a schematic block diagram of the intelligent live broadcast scene adaptive control system proposed by the present invention; Figure 2 This is a schematic block diagram of the intelligent live broadcast scene adaptive control method proposed in the present invention. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0025] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0026] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, and it can be the internal connection of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The present invention will be further described in detail below in conjunction with the accompanying drawings.

[0027] Reference Figure 1-2 :An intelligent live broadcast scene adaptive control system, comprising: Scene perception module: It uses multi-sensor fusion technology and integrates high-definition visual sensors, high-sensitivity audio sensors, and high-precision environmental sensors. It uses advanced multi-modal fusion algorithms to deeply fuse the data collected by different sensors to obtain more comprehensive and accurate live scene information. It collects multi-dimensional information such as the picture, sound, light, temperature, humidity, air quality, etc. of the live scene in real time. Assess the completeness of collected information is the amount of information collected, The module also has a self-calibration function, which can automatically adjust the sensor parameters according to historical collection data and environmental changes to ensure the accuracy and stability of the collected information.

[0028] Data analysis module: Use convolutional neural networks (CNN) and recurrent neural networks (RNN) in deep learning to conduct in-depth analysis of the collected information. CNN is used to extract spatial features in the picture, such as character posture, scene layout, etc.; RNN is used to process audio and time series data to analyze the emotional tendency and rhythm changes of the sound. Extract scene features, such as character movements, expressions, scene layout, sound emotions, etc., and use the feature correlation formula ( The module can also detect anomalies by building a feature model of normal scenes and identifying abnormal events in live broadcast scenes, such as sudden image jitters and abnormal sound noises.

[0029] Strategy generation module: Combines expert system and model predictive control (MPC)-based methods to generate control strategies that adapt to live broadcast scenarios. The expert system stores a large amount of live broadcast scenario knowledge and empirical rules, and can provide preliminary strategy recommendations based on the results of the data analysis module. MPC establishes a dynamic model of the live broadcast scenario to predict the changing trend of the scenario in the future and optimize the control strategy accordingly. Based on the analysis results and preset rules, control strategies are generated for aspects such as live broadcast device parameter adjustment, content recommendation, and interactive strategies. This module also has the ability to innovate strategies, and can automatically generate novel and effective control strategies based on emerging scenario characteristics and user needs.

[0030] Execution control module: It adopts a distributed control architecture to quickly and accurately transmit control instructions to each live broadcast device. It has a real-time feedback mechanism to monitor the execution status of the device in real time to ensure the effective execution of control instructions. It is responsible for executing the generated strategy and adjusting the parameters of the live broadcast device, such as the angle, focal length, aperture of the camera, the brightness, color, angle of the light, the gain and directivity of the microphone, etc. At the same time, this module can also control related functions of the live broadcast platform, such as switching the live broadcast screen and publishing interactive messages.

[0031] Feedback evaluation module: A feedback mechanism based on multi-indicator comprehensive evaluation is constructed. In addition to the effect evaluation formula In addition to evaluating the effectiveness of control strategies ( In order to actually achieve the effect indicators, As the target effect indicator), multiple indicators such as user satisfaction and resource utilization are also considered. The execution effect is evaluated, and the performance and stability of the entire system are evaluated at the same time. This module can also perform trend analysis, predict possible problems in the future by mining historical evaluation data, and give early warnings and improvement suggestions.

[0032] The present invention also includes the following modules: User preference module: This module uses natural language processing technology and machine learning algorithms to collect preference information of live broadcast viewers and anchors, such as favorite picture style, sound effect type, live broadcast theme, interaction method, etc., by analyzing users' comments, likes, sharing and other behaviors on the live broadcast platform, as well as their interaction records with anchors. ( is the number of preferences matched, is the total number of preferences) to evaluate the degree of preference matching. At the same time, this module can also predict preferences by establishing a time series model of user preferences to predict possible changes in user preferences in the future. When generating strategies, the strategy generation module combines user preference information and adopts a personalized recommendation algorithm to generate customized live content and interactive strategies for different user groups, thereby improving the attractiveness of live content and user satisfaction.

[0033] Real-time optimization module: This module uses real-time data mining and adaptive adjustment algorithms to monitor the network status during the live broadcast, such as bandwidth B, delay L, packet loss rate P and jitter J, and calculates the network quality index through the network quality index formula. ( As the weight coefficient, the network quality is evaluated. When the network quality decreases, the real-time optimization module uses intelligent diversion and cache pre-fetching technology to adjust the output of the strategy generation module to prioritize the smoothness of the live broadcast. For example, the video stream is divided into multiple sub-streams, and the transmission rate of the sub-streams is dynamically adjusted according to the network conditions; the live broadcast data that may be needed is cached in advance to reduce the impact of network delay. At the same time, the module can also communicate with the network service provider in real time and adjust the network resource allocation according to the network status request.

[0034] In the present invention, the user preference module also analyzes the changes in user preferences in different time periods and live broadcast topics. ( and The degree of preference change is evaluated by using the time series analysis and cluster analysis methods to explore the rules and patterns of user preference changes. The strategy generation module dynamically adjusts the control strategy according to the preference changes, and uses the dynamic programming algorithm to optimize the strategy adjustment process to better meet user needs. For example, different types of live broadcast content are recommended to users in different time periods, and the interaction method and picture style are adjusted according to the changes in the live broadcast theme.

[0035] In the present invention, when the network quality index is lower than the set threshold, the execution control module adopts a hierarchical cache strategy. ( The execution control module also uses an intelligent cache replacement algorithm to dynamically adjust the cache content according to network conditions and data usage frequency to ensure that the live broadcast quality can be quickly restored when the network is restored. At the same time, the module can also be combined with cloud storage services to store part of the cached data in the cloud to expand the cache capacity.

[0036] In the present invention, the strategy generation module adopts a strategy generation algorithm based on deep reinforcement learning. The algorithm uses a deep neural network (DNN) to approximate the value function and the strategy function, and learns the optimal control strategy by continuously interacting with the live broadcast environment to improve the adaptability and effectiveness of the strategy. Evaluate the reward value of taking action a in state s and optimize the strategy by maximizing the cumulative reward. At the same time, the strategy generation module also considers the dynamic changes of the live broadcast scene and uses meta-learning technology to quickly adapt to new scenes and tasks. In the process of strategy generation, uncertainty modeling is introduced to take into account the uncertainty of the environment and the noise of the data to improve the robustness of the strategy.

[0037] In the present invention, the feedback evaluation module not only evaluates the execution effect, but also evaluates the accuracy of the scene perception module. Assessing perceived accuracy ( is the amount of correctly perceived information, is the total amount of perceived information). Bayesian reasoning and evidence theory are used to fuse and verify the perceived data to improve the reliability of the perception accuracy assessment. When the perception accuracy is lower than the threshold, the feedback evaluation module sends adjustment instructions to the scene perception module and uses genetic algorithms to optimize sensor parameters and acquisition strategies to improve the accuracy and stability of scene perception.

[0038] The present invention also discloses a method for an intelligent live broadcast scene adaptive control system. The scene perception module uses multi-sensor fusion technology and multi-modal fusion algorithm to collect multi-dimensional information of the live broadcast scene in real time, uses the information integrity formula to evaluate the integrity of the collected information, and ensures the accuracy of the information through a self-calibration function; the data analysis module uses deep learning CNN and RNN to perform in-depth analysis on the collected information, extract scene features, uses the feature correlation formula to evaluate the degree of feature correlation, and performs anomaly detection at the same time; the strategy generation module combines the expert system and the MPC method to generate a control strategy based on the analysis results and preset rules, and uses a personalized recommendation algorithm combined with user preferences; the execution control module uses a distributed control architecture and a real-time feedback mechanism to execute the strategy, adjust the parameters of the live broadcast equipment and the control platform functions; the feedback evaluation module constructs a multi-indicator comprehensive evaluation mechanism, evaluates the execution effect and system performance, performs trend analysis, and gives warnings and suggestions.

[0039] In the present invention, the user preference module uses natural language processing and machine learning algorithms to collect and analyze user preference information, evaluates the degree of preference matching through the preference matching formula, and makes preference predictions; the strategy generation module considers user preferences when generating strategies, and uses a dynamic programming algorithm to adjust strategies according to preference changes; the real-time optimization module uses real-time data mining and adaptive adjustment algorithms to monitor network conditions, uses the network quality index formula to evaluate network quality, and uses intelligent diversion and cache pre-fetching technology to adjust strategies when network quality decreases, and communicates with network service providers in real time; the user preference module uses natural language processing and machine learning algorithms to accurately and comprehensively collect and analyze user preference information. The preference matching formula is used to evaluate the degree of preference matching and make predictions, just like tailoring a service guide for users. This enables the system to understand user needs in advance. For example, in live content recommendations, programs that meet user interests can be accurately pushed, greatly improving user experience and increasing user stickiness to the platform.

[0040] The strategy generation module takes user preferences into account when generating strategies, and uses a dynamic programming algorithm to adjust strategies based on changes in preferences. This makes the system highly flexible and adaptable, just like a caring smart assistant that can adjust live broadcast strategies in real time as user interests change. For example, when users' preference for a certain type of game live broadcast increases, the system can quickly increase resource investment in related game live broadcasts, optimize live broadcast scheduling, etc., effectively improving the utilization efficiency of live broadcast resources and better meeting users' changing needs.

[0041] The real-time optimization module uses real-time data mining and adaptive adjustment algorithms to monitor network conditions, and uses the network quality index formula to evaluate network quality. When network quality decreases, the diversion and cache pre-fetching technology is used to adjust the strategy, providing a solid guarantee for the smoothness of live broadcasting. It is like an intelligent network guardian that can perceive network changes in real time and respond in a timely manner. When the network is congested, the diversion technology is used to reasonably distribute network traffic, and the cache pre-fetching is used to prepare some content in advance, ensuring that users will not experience freezes and other situations due to network problems when watching live broadcasts, thereby improving the stability of live broadcasts and the comfort of users watching.

[0042] In the present invention, the user preference module uses time series analysis and cluster analysis methods to analyze the changes of user preferences over time and topics, and evaluates the degree of change through the preference change coefficient formula; when the network quality index is lower than the threshold, the execution control module adopts a hierarchical cache strategy, determines the cache content priority according to the cache priority formula, and adopts an intelligent cache replacement algorithm combined with cloud storage; the feedback evaluation module uses Bayesian reasoning and evidence theory to evaluate the accuracy of the scene perception module. If it is lower than the threshold, a genetic algorithm is used to send adjustment instructions to optimize the acquisition strategy; the user preference module uses time series analysis and cluster analysis methods to analyze the changes of user preferences over time and topics, and evaluates the degree of change through the preference change coefficient formula. This enables the system to deeply understand the dynamic evolution of user preferences, just like having a "time capsule" of user interest changes. For example, it can find differences in user preferences for live broadcast topics in different seasons and time periods, so as to plan and recommend content more accurately and further improve user satisfaction.

[0043] When the network quality index is lower than the threshold, the execution control module adopts a hierarchical cache strategy and determines the cache content priority according to the cache priority formula. This mechanism can intelligently ensure the smooth playback of important live content when the network is poor. For example, it prioritizes caching the highlights or key live broadcast links of popular anchors to ensure that users can watch the most valuable content even when the network is unstable, reducing the decline in viewing experience caused by network problems.

[0044] The feedback evaluation module uses Bayesian reasoning and evidence theory to evaluate the accuracy of the scene perception module. If it is below the threshold, a genetic algorithm is used to send adjustment instructions to optimize the acquisition strategy. This forms a closed-loop optimization system, which is like installing an "intelligent brain" on the entire intelligent live scene control method. It can continuously self-check and optimize to ensure that the scene perception module always maintains high accuracy, so that the entire system can perceive the live scene more accurately, provide better quality live services that are more in line with user needs, and enhance the stability and reliability of the system.

[0045] The above are only preferred specific implementation modes of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solutions and inventive concepts of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An intelligent live scene adaptive control system, characterized in that: include: Scene perception module: uses sensor fusion technology to integrate sensors, audio sensors and environmental sensors; Collect the live scene's picture, sound, light, temperature, humidity, and air quality information in real time; use the information completeness formula Assess the completeness of collected information, including is the amount of information collected, The total amount of information to be collected; Data analysis module: Use convolutional neural network CNN and recurrent neural network RNN ​​to analyze the collected information; use feature correlation formula Evaluate the degree of association between features, where is the strength of the association relationship of the i-th feature; Strategy generation module: Combines expert system and MPC-based model predictive control method to generate control strategy for live broadcast scenarios; The expert system stores live broadcast scenario knowledge and empirical rules, and provides preliminary strategic recommendations based on the results of the data analysis module; MPC builds a dynamic model of the live broadcast scene to predict the changing trend of the scene in the future and optimize the control strategy; Execution control module: adopts a distributed control architecture to transmit control instructions to live broadcast equipment; real-time feedback mechanism to monitor the execution status of the equipment in real time, responsible for executing the generated strategy, and adjusting the parameters of the live broadcast equipment, including the angle, focal length, aperture of the camera, the brightness, color, angle of the light, and the gain and directivity of the microphone; Feedback evaluation module: Construct a comprehensive evaluation feedback mechanism, in addition to the effect evaluation formula In addition to evaluating the effectiveness of the control strategy, user satisfaction and resource utilization indicators are also considered. In order to actually achieve the effect indicators, The target effect indicator.

2. The intelligent live scene adaptive control system according to claim 1 is characterized in that: Also includes: User preference module: Utilizes natural language processing technology and machine learning algorithms to collect preference information of live broadcast viewers and hosts by analyzing users’ comments, likes, and sharing behaviors on the live broadcast platform, as well as their interaction records with hosts. Assess the degree of preference matching, where is the number of preferences matched, is the total number of preferences.

3. The intelligent live scene adaptive control system according to claim 1 is characterized in that: Also includes: Real-time optimization module: It uses real-time data mining and adaptive adjustment algorithms to monitor the network status during live broadcast, including bandwidth B, delay L, packet loss rate P and jitter J. Evaluate network quality, including is the weight coefficient.

4. The intelligent live scene adaptive control system according to claim 2 is characterized in that: The user preference module also analyzes the changes in user preferences in different time periods and live broadcast topics; through the preference change coefficient formula Assess the degree of preference change, where and are the frequencies of the i-th preference in different time periods or topics, and m is the number of preferences.

5. The intelligent live scene adaptive control system according to claim 3 is characterized in that: When the network quality index is lower than the set threshold, the execution control module adopts a hierarchical cache strategy; through the cache priority formula Prioritize cached content, where is the weight coefficient, R is the content importance level, I is the immediacy demand, and U is the user attention.

6. The intelligent live scene adaptive control system according to any one of claims 1 to 5, characterized in that: The strategy generation module adopts a strategy generation algorithm based on reinforcement learning. The algorithm approximates the value function and the strategy function through the deep neural network DNN, and uses the reward function Evaluate the reward value of taking action a in state s and optimize the policy by maximizing the cumulative reward.

7. The intelligent live scene adaptive control system according to any one of claims 1 to 5, characterized in that: In addition to evaluating the execution effect, the feedback evaluation module also evaluates the accuracy of the scene perception module; through the perception accuracy formula Assessing perceived accuracy, where is the amount of correctly perceived information, is the total amount of perceived information.

8. A method for applying the intelligent live scene adaptive control system according to any one of claims 1 to 7, characterized in that: The scene perception module uses sensor fusion technology and modal fusion algorithms to collect information about live scenes in real time, uses the information integrity formula to evaluate the integrity of the collected information, and ensures the accuracy of the information through the self-calibration function; The data analysis module uses CNN and RNN to analyze the collected information, extract scene features, use the feature correlation formula to evaluate the degree of feature correlation, and perform anomaly detection at the same time; the strategy generation module combines the expert system and MPC method to generate control strategies based on the analysis results and preset rules, and uses the recommendation algorithm combined with user preferences; the execution control module uses a distributed control architecture and real-time feedback mechanism to execute strategies, adjust the live broadcast equipment parameters and control platform functions; the feedback evaluation module constructs a comprehensive evaluation mechanism to evaluate the execution effect and system performance.

9. The intelligent live scene adaptive control method according to claim 8, characterized in that: The user preference module uses natural language processing and machine learning algorithms to collect and analyze user preference information, evaluates the degree of preference matching through the preference matching formula, and makes preference predictions; The strategy generation module considers user preferences when generating strategies and uses a dynamic programming algorithm to adjust strategies according to preference changes; The real-time optimization module uses real-time data mining and adaptive adjustment algorithms to monitor network conditions, uses the network quality index formula to evaluate network quality, and uses diversion and cache pre-fetching technology to adjust strategies when network quality deteriorates.

10. The intelligent live scene adaptive control method according to claim 9, characterized in that: The user preference module uses time series analysis and cluster analysis methods to analyze the changes in user preferences over time and topics, and evaluates the degree of change through the preference change coefficient formula; when the network quality index is lower than the threshold, the execution control module adopts a hierarchical caching strategy and determines the cache content priority based on the cache priority formula; the feedback evaluation module uses Bayesian reasoning and evidence theory to evaluate the accuracy of the scene perception module. If it is lower than the threshold, a genetic algorithm is used to send adjustment instructions to optimize the acquisition strategy.

Images