Distributed multimedia publishing system based on environment interaction control

Abstract

The invention belongs to the technical field of multimedia publishing management and control, and particularly relates to a distributed multimedia publishing system based on environment interaction control, which comprises an environment multi-dimensional perception fusion module, a dynamic content decision module, a distributed resource scheduling module, a multi-mode interaction feedback module, a publishing quality monitoring module and a supervision center, according to the method, the dynamic environment feature vector is generated by collecting and fusing the multi-source environment data in real time, the content publishing strategy is dynamically generated based on the decision tree algorithm and the user behavior data, real-time linkage of the content and the environment is achieved to reduce manual intervention, and through distributed resource scheduling and intelligent load balancing, the user experience is improved. The resource utilization rate is optimized, energy consumption is reduced, multi-dimensional interaction such as voice, gestures and touch screens is supported to remarkably improve the user participation degree, service continuity is guaranteed by monitoring key indexes in real time and triggering a self-adaptive repairing mechanism, and closed-loop optimization of environment perception, dynamic decision making, resource scheduling and interaction feedback is achieved.

Description

Technical Field

[0001] This invention relates to the field of multimedia publishing and management technology, specifically a distributed multimedia publishing system based on environmental interaction control. Background Technology

[0002] Multimedia publishing refers to the use of digital technology to distribute content that integrates multiple media formats such as text, images, audio, video, and animation to target terminals according to specific strategies, so as to realize the dynamic and intelligent display and interaction of information. With the rapid development of digital technology, multimedia content is increasingly widely used in scenarios such as commercial advertising, public information release, and digital exhibition halls. Chinese invention patent CN102916931A discloses a multimedia information publishing system and method. The multimedia information publishing system uses a camera device to wirelessly transmit acquired images or videos to a predetermined storage space in an information publishing server. When the information publishing server detects that there are unpublished images or videos in the predetermined storage space, it pushes the unpublished images or videos to a multimedia information publishing terminal. After receiving the unpublished images or videos, the multimedia information publishing terminal displays them, thus realizing the instant, real-time and rapid publishing of multimedia information. However, as shown in the above-mentioned invention, the existing multimedia publishing system only has a simple real-time publishing function. It has problems such as weak environmental perception, static decision-making mechanism, inefficient resource scheduling, poor interactive experience and lagging quality monitoring. In the application scenarios, it is difficult to take into account the accurate adaptation of content, efficient use of resources and optimization of user experience. Furthermore, it is impossible to reasonably analyze the stability of multimedia publishing and interactive performance and provide timely warnings, which is not conducive to ensuring the effect of multimedia publishing and user experience. To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0003] The purpose of this invention is to provide a distributed multimedia publishing system based on environmental interaction control, which solves the problems of existing technologies in the application scenarios that make it difficult to simultaneously achieve accurate content adaptation, efficient resource utilization and user experience optimization, and also make it impossible to reasonably analyze and provide timely warnings on the stability and interactive performance of multimedia publishing, which is not conducive to ensuring the effect of multimedia publishing and user experience.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The distributed multimedia publishing system based on environmental interaction control includes an environmental multi-dimensional perception and fusion module, a dynamic content decision module, a distributed resource scheduling module, a multimodal interaction feedback module, a publishing quality monitoring module, and a monitoring center; the system collects and fuses multi-source environmental data in real time, generates dynamic environmental feature vectors, and sends them to the dynamic content decision module. The dynamic content decision module dynamically generates content publishing strategies based on environmental feature vectors and user behavior data using a decision tree algorithm. The distributed resource scheduling module dynamically allocates computing, storage, and bandwidth resources based on content decision instructions, supporting cross-node content caching and load balancing. The multimodal interaction feedback module supports multimodal interaction including voice, gesture and touch screen, analyzes user intent in real time and generates feedback, records interaction logs and stores them in the database; the publishing quality monitoring module monitors the publishing quality of multimedia content in real time, triggers an adaptive repair mechanism, and transmits publishing quality monitoring information to the supervision center and the distributed resource scheduling module.

[0005] Furthermore, the environmental multidimensional perception fusion module consists of a data acquisition layer, a data preprocessing layer, a feature fusion layer, and a data output layer. The data acquisition layer collects raw data through IoT sensors deployed in the scene. The data preprocessing layer is used for noise reduction, processing, and normalization. The sliding window averaging method is used to filter the sensor jitter data to eliminate instantaneous noise, and the data of different dimensions are mapped to the [0,1] interval to generate standardized data packets. The feature fusion layer is used to associate multi-source data, establish data associations based on sensor spatial locations, and use a weighted voting mechanism to assign weights to conflicting data to generate a comprehensive environmental feature vector. The data output layer encrypts the feature vector and pushes it to the dynamic content decision module via the MQTT protocol.

[0006] Furthermore, the operation of the dynamic content decision module includes loading the strategy rule base, real-time environment matching, user behavior association, and strategy generation and push: in real-time environment matching, it receives feature vectors sent by the multi-dimensional environment perception fusion module and matches the optimal rule through the decision tree algorithm. In user behavior association, user behavior data is analyzed through edge computing nodes to dynamically adjust content priority; in strategy generation and push, decision instructions containing content ID, publishing node and playback sequence are generated and pushed to the distributed resource scheduling module via HTTP REST API.

[0007] Furthermore, the specific operation process of the distributed resource scheduling module is as follows: Resource topology modeling: Construct a node resource graph and record the relevant attributes of each node, including CPU / GPU computing power, storage capacity, and network bandwidth; Task decomposition and prioritization: The content publishing task is broken down into sub-tasks, and priorities are assigned to the sub-tasks according to the priority of the decision instructions. Intelligent scheduling algorithm: Optimizes task allocation using ant colony optimization; Execution monitoring and dynamic adjustment: Track task progress in real time. If the load on a node exceeds the threshold, trigger load migration to migrate some tasks to a backup node, update the resource graph, and recalculate the optimal path.

[0008] Furthermore, the operation process of the multimodal interactive feedback module includes: Interactive data acquisition: Voice commands are acquired via microphone array, gestures are acquired via camera, and click events are acquired via touchscreen; Intent recognition: For voice commands, the BERT model is used for semantic parsing to generate structured instructions; for gestures, OpenPose is used to identify key points and match them with a predefined gesture library; for touch events, the relationship between click coordinates and page elements is analyzed. Feedback generation and execution: Generate feedback instructions based on user intent; Interaction log recording: Associate interaction events with content ID and user ID and store them in the database.

[0009] Furthermore, the specific operation process of the quality monitoring module is as follows: Quality metrics collection: Deploy probes on the release node to collect metrics data including video frame rate, packet loss rate, and number of rendering errors; Anomaly detection: The LSTM model predicts the quality decline trend. The input is a historical index sequence, and the output is the quality prediction value for the next 5 minutes. If the predicted value is lower than the set threshold, it is marked as an anomaly. Repair strategy execution: Trigger repair actions based on the type of anomaly. For network congestion, switch to a backup CDN node; for node failure, restart the node or migrate the task to a healthy node; for content errors, roll back to the previous version and notify operations and maintenance.

[0010] Furthermore, the release quality monitoring module communicates with the release quality stability assessment module. The release quality monitoring module transmits release quality monitoring information to the release quality stability assessment module. The release quality stability assessment module is used to set the release period, evaluate and analyze the release quality stability during the release period, and generate a quality stability normal signal or a quality stability abnormal signal through analysis. The quality stability normal signal or quality stability non-conforming signal is sent to the supervision center. When the supervision center receives the quality stability non-conforming signal, it issues a corresponding warning.

[0011] Furthermore, the specific analysis process for the quality stability assessment module is as follows: All abnormal information that occurs during the release period is obtained, the number of times abnormal information is generated is marked as the abnormal value, and the abnormal value is compared with the preset abnormal threshold. If the abnormal value exceeds the preset abnormal threshold, a quality stability failure signal is generated. If the published abnormal value does not exceed the preset abnormal value threshold, the repair time for the corresponding abnormal information will be marked as the abnormal repair feature value. The abnormal repair feature value will be compared with the corresponding preset abnormal repair feature threshold. If the abnormal repair feature value exceeds the corresponding preset abnormal repair feature threshold, the corresponding abnormal repair feature value will be marked as an unreasonable repair value. The system obtains the number of unreasonable values ​​repaired during the release period and marks them as re-detection values. It then calculates the ratio of the abnormal repair feature value to the corresponding preset abnormal repair feature threshold to obtain the abnormal re-detection value. The system calculates the average value of all abnormal re-detection values ​​within the distribution period to obtain the re-time average value. The re-detection value and the re-time average value are compared with the preset re-detection threshold and the preset re-time average value threshold, respectively. If the re-detection value or the re-time average value exceeds the corresponding preset threshold, a quality stability failure signal is generated. If neither the re-detection value nor the re-time average value exceeds the corresponding preset threshold, a quality stability normal signal is generated.

[0012] Furthermore, the quality stability assessment module communicates with the interactive tracking and judgment module. The quality stability assessment module sends a quality stability normal signal to the interactive tracking and judgment module. When the interactive tracking and judgment module receives the quality stability normal signal, it analyzes the interaction performance of the multimodal interactive feedback module during the distribution period. Through analysis, it generates an interactive normal signal or an interactive non-compliance signal and sends the interactive normal signal or interactive non-compliance signal to the regulatory center. When the regulatory center receives the interactive non-compliance signal, it issues a corresponding warning.

[0013] Furthermore, the specific analysis process of the interaction tracking and judgment module is as follows: When a user makes a correct interaction using voice, gestures, or touchscreen, if the multimodal interaction feedback module fails to recognize the feedback, the corresponding interaction operation is marked as an unresponsive operation; otherwise, the corresponding interaction operation is marked as an abnormal operation. The number of unresponsive operations during the release period is collected and the ratio of this number to the total number of interaction operations is calculated to obtain the unresponsive detection value. The unresponsive detection value is then compared with a preset unresponsive detection threshold. If the unresponsive detection value exceeds the preset unresponsive detection threshold, an interaction failure signal is generated. If the no-response detection value does not exceed the preset no-response detection threshold, the feedback duration of the multimodal interaction feedback module for all non-abnormal operations is obtained, and the corresponding feedback duration is compared with the corresponding preset feedback duration threshold. If the feedback duration exceeds the corresponding preset feedback duration threshold, the corresponding feedback duration is marked as non-rapid response duration. The non-rapid response time is obtained by comparing the number of non-rapid response times during the release period with the total number of feedback times. The non-rapid response detection value is obtained by comparing the feedback time with the corresponding preset feedback time threshold. The feedback time anomaly value is obtained by averaging all feedback time anomaly values ​​during the release period. The interaction comprehensive judgment coefficient is obtained by weighted summing of the no-response detection value, the non-rapid response detection value, and the feedback time anomaly value. The interaction comprehensive judgment coefficient is compared with the preset interaction comprehensive judgment coefficient threshold. If the interaction comprehensive judgment coefficient exceeds the preset interaction comprehensive judgment coefficient threshold, an interaction failure signal is generated. If the interaction comprehensive judgment coefficient does not exceed the preset interaction comprehensive judgment coefficient threshold, an interaction normal signal is generated.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. In this invention, by collecting and fusing multi-source environmental data in real time, a content publishing strategy is dynamically generated based on decision tree algorithm and user behavior data, thereby achieving real-time linkage between content and environment to reduce manual intervention. Through distributed resource scheduling and intelligent load balancing, resource utilization is optimized and energy consumption is reduced. Furthermore, multi-dimensional interaction such as voice, gesture, and touch screen is supported to significantly improve user engagement. In addition, by monitoring key indicators in real time and triggering an adaptive repair mechanism, service continuity is ensured. This achieves closed-loop optimization of environmental perception, dynamic decision-making, resource scheduling, and interactive feedback, significantly improving the effect of multimedia content publishing and user experience. 2. In this invention, the stability of the release quality during the release period is evaluated and analyzed by the release quality stability assessment module. When a quality stability failure signal is generated, the cause is investigated and reasonable improvement measures are taken to ensure the stability of multimedia release. When a quality stability normal signal is generated, the interaction performance of the interaction based on the multimodal interaction feedback module during the distribution period is analyzed. When an interaction failure signal is generated, the process is checked and optimized to further improve the user experience and reduce the difficulty of multimedia release management. Attached Figure Description

[0015] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a system block diagram of Embodiment 1 of the present invention; Figure 2 This is a system block diagram of Embodiments 2 and 3 of the present invention. Detailed Implementation

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0017] Example 1: As Figure 1 As shown, the distributed multimedia publishing system based on environmental interaction control proposed in this invention includes an environmental multi-dimensional perception and fusion module, a dynamic content decision-making module, a distributed resource scheduling module, a multimodal interactive feedback module, a publishing quality monitoring module, and a monitoring center. The system collects and fuses multi-source environmental data (such as light, pedestrian flow, noise, etc.) in real time, generates dynamic environmental feature vectors, and sends them to the dynamic content decision-making module. This eliminates sensor data silos, improves the accuracy of environmental perception, and the dynamic feature vectors provide a precise basis for subsequent content adaptation. It should be noted that the environmental multi-dimensional perception fusion module consists of a data acquisition layer, a data preprocessing layer, a feature fusion layer, and a data output layer. The data acquisition layer collects raw data through IoT sensors deployed in the scene (such as infrared thermal imaging cameras, microphone arrays, and temperature and humidity sensors). The data format is sensor ID + timestamp + raw value. The data preprocessing layer is used for denoising, processing and normalization. It uses the sliding window averaging method to filter the sensor jitter data to eliminate instantaneous noise, and maps data of different dimensions (such as illumination unit Lux and noise unit dB) to the [0,1] interval to generate standardized data packets. The feature fusion layer is used to associate multi-source data, establish data association based on the spatial location of sensors (such as cameras and microphones in the same area), and use a weighted voting mechanism to assign weights to conflicting data (such as differences in pedestrian counts from different sensors) (e.g., 0.6 for cameras and 0.4 for infrared sensors) to generate a comprehensive environmental feature vector (e.g., a label for "high pedestrian flow + low light + medium noise"). The data output layer encrypts the feature vector and pushes it to the dynamic content decision module via the MQTT protocol.

[0018] The dynamic content decision-making module dynamically generates content publishing strategies (such as switching video versions and adjusting playback order) based on environmental feature vectors and user behavior data using a decision tree algorithm. This enables real-time interaction between content and the environment, improves user experience, and optimizes content strategies through a closed-loop user behavior mechanism, reducing manual intervention. The operation process of the dynamic content decision-making module is as follows: I. Loading the policy rule base: Example of a predefined rule: If "lighting < 0.3 and pedestrian flow > 0.7", then enable high-contrast video; if "noise > 0.8", then switch to silent mode. II. Real-time Environment Matching: Receives feature vectors sent by the multi-dimensional environmental perception fusion module and matches the optimal rule using a decision tree algorithm; Example: Input features [0.8, 0.3, 0.5] (high pedestrian traffic, low lighting, medium noise), matching rule "reduce video brightness and enable subtitles"; 3. User Behavior Correlation: Analyze user behavior data (such as click-through rate and dwell time) through edge computing nodes to dynamically adjust content priority; Example: If the click-through rate of an advertisement video is lower than the threshold for three consecutive times, its playback priority will be reduced; IV. Strategy Generation and Push: Generate decision instructions containing content ID, publishing node, and playback timing, and push them to the distributed resource scheduling module via HTTP REST API.

[0019] The distributed resource scheduling module dynamically allocates computing, storage, and bandwidth resources based on content decision instructions. It supports cross-node content caching and load balancing, reduces system energy consumption, improves resource utilization, and enables automatic migration of failed nodes to ensure service continuity. The specific operation process of the distributed resource scheduling module is as follows: S1. Resource topology modeling: Construct a node resource graph and record attributes such as CPU / GPU computing power, storage capacity and network bandwidth of each node; S2. Task Decomposition and Prioritization: The content publishing task is broken down into sub-tasks (such as video transcoding, cache synchronization, and content verification). Among them, video transcoding converts the original video to different resolutions (such as 1080p and 720p) to adapt to different devices; cache synchronization preloads the content to edge nodes to reduce user access latency; and content verification verifies file integrity to avoid transmission errors. And assign priorities to subtasks based on the priority of the decision instructions; example of priority rules: Emergency content (such as disaster warnings): set to the highest priority (P0) and force resource allocation; High-value content (such as advertising): set to high priority (P1) and prioritize the allocation of remaining resources; Regular content (such as background videos): set to low priority (P2) and execute when resources are idle. S3. Intelligent Scheduling Algorithm: Ant colony optimization is used to optimize task allocation; refer to the following method: Initialization: Each "ant" randomly selects a path (node ​​combination); Pheromone update: Update path weights based on task completion time and resource consumption; Convergence: Select the optimal path and assign tasks; S4. Execution monitoring and dynamic adjustment: Track task progress in real time. If the load of a node exceeds the threshold (e.g., CPU > 90%), load migration is triggered to migrate some tasks to the backup node, update the resource graph and recalculate the optimal path.

[0020] The multimodal interaction feedback module supports multimodal interaction including voice, gestures, and touchscreen. It analyzes user intent in real time and generates feedback, records interaction logs and stores them in a database, improving user engagement and user experience. Furthermore, the interaction logs provide a basis for continuous optimization of the content decision model. The operation process of the multimodal interaction feedback module includes: Interactive data acquisition: Voice commands are acquired via microphone array, gestures are acquired via camera, and click events are acquired via touchscreen; Intent recognition: For voice commands, the BERT model is used for semantic parsing to generate structured commands (e.g., {action: "switch", target: "3D mode"}); for gesture actions, OpenPose is used to identify key points and match them with a predefined gesture library (e.g., "five fingers spread" corresponds to "pause"); for touch screen events, the relationship between click coordinates and page elements is parsed. Feedback generation and execution: Generate feedback instructions based on user intent; for example: voice command "switch to 3D mode" → push the instruction to the content publishing node; gesture "pause" → pause the current video playback; Interaction log recording: Associate interaction events with content ID and user ID and store them in the database.

[0021] The release quality monitoring module monitors the quality of multimedia content releases in real time (such as latency, stuttering, rendering errors, etc.), triggers an adaptive repair mechanism, and transmits the release quality monitoring information to the monitoring center and the distributed resource scheduling module. This significantly reduces the multimedia release failure rate and achieves automated repair to reduce manual intervention costs. The specific operation process of the release quality monitoring module is as follows: Quality metrics collection: Deploy probes on the release node to collect metrics data including video frame rate (FPS), packet loss rate, and number of rendering errors; Anomaly detection: The LSTM model predicts the quality decline trend. The input is a historical index sequence, and the output is the quality prediction value for the next 5 minutes. If the predicted value is lower than the set threshold (e.g., FPS < 20), it is marked as an anomaly. Repair strategy execution: Trigger repair actions based on the type of anomaly. For network congestion, switch to a backup CDN node; for node failure, restart the node or migrate the task to a healthy node; for content errors, roll back to the previous version and notify operations and maintenance.

[0022] Example 2: Figure 2As shown, the difference between this embodiment and embodiment one is that the release quality monitoring module is connected to the release quality stability assessment module. The release quality monitoring module transmits the release quality monitoring information to the release quality stability assessment module. The release quality stability assessment module is used to set the release period, evaluate and analyze the release quality stability during the release period, and generate a quality stability normal signal or a quality stability abnormal signal through analysis. Furthermore, the system sends either a normal or unsatisfactory quality stability signal to the monitoring center. Upon receiving an unsatisfactory quality stability signal, the monitoring center issues a corresponding warning to remind management personnel to investigate the cause and take reasonable improvement measures, ensuring the stability of multimedia publishing and enhancing the user experience. The specific analysis process of the publishing quality stability assessment module is as follows: All abnormal publishing information that occurs during the publishing period is obtained, and the number of times abnormal information is generated is marked as the abnormal publishing value. The abnormal publishing value is compared with the preset abnormal publishing threshold. If the abnormal publishing value exceeds the preset abnormal publishing threshold, it indicates that the stability of multimedia publishing during the publishing period is poor and needs to be optimized and improved in time. In this case, a quality stability failure signal is generated. If the published anomaly value does not exceed the preset anomaly threshold, the repair time for the corresponding anomaly information will be marked as an anomaly repair feature value. It should be noted that the larger the anomaly repair feature value, the slower the repair process, which is less conducive to ensuring the publishing effect. The anomaly repair feature value will be compared with the corresponding preset anomaly repair feature threshold. If the anomaly repair feature value exceeds the corresponding preset anomaly repair feature threshold, it indicates that the corresponding repair process is slower, and the corresponding anomaly repair feature value will be marked as an unreasonable repair value. The number of unreasonable values ​​that were repaired during the release period was obtained and marked as repeated non-detection values. The ratio of the abnormal repair feature value to the corresponding preset abnormal repair feature threshold was calculated to obtain the abnormal retest value. The average value of all abnormal retest values ​​during the distribution period was calculated to obtain the retest value. The repeated non-detection value and the retest value were compared with the preset repeated non-detection threshold and the preset retest threshold, respectively. If the non-detection value or the time-averaged value exceeds the corresponding preset threshold, it indicates that the multimedia release stability during the release period is poor and needs to be optimized and improved in a timely manner, and a quality stability failure signal is generated; if neither the non-detection value nor the time-averaged value exceeds the corresponding preset threshold, it indicates that the multimedia release stability during the release period is good, and a quality stability normal signal is generated.

[0023] Example 3: Figure 2As shown, the difference between this embodiment and Embodiment 1 and Embodiment 2 is that the quality stability assessment module communicates with the interactive tracking and judgment module, the quality stability assessment module sends the quality stability normal signal to the interactive tracking and judgment module, and when the interactive tracking and judgment module receives the quality stability normal signal, it analyzes the interaction performance of the multimodal interactive feedback module during the distribution period, and generates an interactive normal signal or an interactive unqualified signal through analysis. Furthermore, it sends normal or non-compliant interaction signals to the monitoring center. Upon receiving a non-compliant interaction signal, the monitoring center issues a corresponding warning to remind management personnel to check and optimize the multimodal interaction feedback module, ensuring its operational performance and further improving the user experience. The specific analysis process of the interaction tracking and judgment module is as follows: When a user makes a correct interaction using voice, gestures, or touchscreen, if the multimodal interaction feedback module fails to recognize the feedback, the corresponding interaction operation is marked as an unresponsive operation; otherwise, the corresponding interaction operation is marked as an abnormal operation. The number of unresponsive operations during the release period is collected and the ratio of this number to the total number of interaction operations is calculated to obtain the unresponsive detection value. The unresponsive detection value is compared with a preset unresponsive detection threshold. If the unresponsive detection value exceeds the preset unresponsive detection threshold, it indicates that the interaction performance during the release period is poor, and an interaction failure signal is generated.

[0024] Furthermore, if the no-response detection value does not exceed the preset no-response detection threshold, the feedback duration of the multimodal interaction feedback module for all non-abnormal operations is obtained, and the corresponding feedback duration is compared with the corresponding preset feedback duration threshold. If the feedback duration exceeds the corresponding preset feedback duration threshold, it indicates that the feedback for the corresponding interaction operation is slow, and the corresponding feedback duration is marked as non-rapid response duration. The non-rapid response duration is obtained by calculating the ratio of the non-rapid response duration to the total number of feedback durations during the release period. The feedback duration is then calculated by calculating the ratio of the feedback duration to the corresponding preset feedback duration threshold to obtain the feedback time feature value. Finally, the average of all feedback time feature values ​​during the release period is calculated to obtain the feedback time anomaly value. The interaction comprehensive judgment coefficient is calculated by weighting and summing the no-response detection value, the non-rapid response detection value, and the feedback time difference value. Specifically, a corresponding preset weight coefficient is assigned to the no-response detection value, the non-rapid response detection value, and the feedback time difference value, and then each of these three values ​​is multiplied by its corresponding preset weight coefficient. The sum of these three products is then marked as the interaction comprehensive judgment coefficient. It should be noted that the larger the value of the interaction comprehensive judgment coefficient, the worse the overall interaction performance during the release period. The interaction comprehensive judgment coefficient is compared with the preset interaction comprehensive judgment coefficient threshold. If the interaction comprehensive judgment coefficient exceeds the preset interaction comprehensive judgment coefficient threshold, it indicates that the interaction performance during the release period is poor overall, and an interaction failure signal is generated. If the interaction comprehensive judgment coefficient does not exceed the preset interaction comprehensive judgment coefficient threshold, it indicates that the interaction performance during the release period is good overall, and an interaction normal signal is generated.

[0025] The working principle of this invention is as follows: In use, it generates dynamic environmental feature vectors by real-time collection and fusion of multi-source environmental data, providing precise input for content decision-making. Based on decision tree algorithms and user behavior data, it dynamically generates content publishing strategies, achieving real-time linkage between content and the environment to reduce manual intervention. Through distributed resource scheduling and intelligent load balancing, it dynamically allocates computing, storage, and bandwidth resources according to task priority, optimizing resource utilization and reducing energy consumption. It also supports multi-dimensional interaction such as voice, gestures, and touchscreens, real-time parsing of user intent and generation of feedback to significantly improve user engagement. Furthermore, by real-time monitoring of key indicators and triggering adaptive repair mechanisms, it effectively reduces publishing failure rates and ensures service continuity. This achieves closed-loop optimization of environmental perception, dynamic decision-making, resource scheduling, and interactive feedback, significantly improving the effect of multimedia content publishing and user experience, demonstrating significant technological advancement and commercial value.

[0026] In this invention, the threshold, preset value, or preset range settings are for result comparison and analysis to determine whether the result is good or bad. The magnitude of these values ​​is determined by a combination of large-scale model analysis of sample data and human experience, and can also be appropriately adjusted based on seasonal or common-sense influence conditions. Similarly, the preset weight coefficients and influence factors are assigned specific values ​​based on the magnitude of each parameter's influence on the result, ultimately reflecting the impact on the result. These settings are also determined by a combination of large-scale model analysis of sample data and human experience, and can also be appropriately adjusted based on seasonal or common-sense influence conditions.

[0027] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, enabling those skilled in the art to better understand and utilize it. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A distributed multimedia publishing system based on environmental interaction control, characterized in that, It includes an environmental multi-dimensional perception and fusion module, a dynamic content decision-making module, a distributed resource scheduling module, a multimodal interaction feedback module, a release quality monitoring module, and a monitoring center; the environment collects and fuses multi-source environmental data in real time to generate dynamic environmental feature vectors; the dynamic content decision-making module dynamically generates content release strategies based on environmental feature vectors and user behavior data through a decision tree algorithm; the distributed resource scheduling module dynamically allocates computing, storage, and bandwidth resources according to content decision instructions, and supports cross-node content caching and load balancing; The multimodal interaction feedback module supports multimodal interaction including voice, gesture and touch screen, analyzes user intent in real time and generates feedback, records interaction logs and stores them in the database; The publishing quality monitoring module monitors the publishing quality of multimedia content in real time, triggers an adaptive repair mechanism, and transmits the publishing quality monitoring information to the supervision center and the distributed resource scheduling module.

2. The distributed multimedia publishing system based on environmental interaction control according to claim 1, characterized in that, The environmental multidimensional perception fusion module consists of a data acquisition layer, a data preprocessing layer, a feature fusion layer, and a data output layer.

3. The distributed multimedia publishing system based on environmental interaction control according to claim 2, characterized in that, The dynamic content decision-making module's operation process includes loading the policy rule base, real-time environment matching, user behavior correlation, and policy generation and push.

4. The distributed multimedia publishing system based on environmental interaction control according to claim 3, characterized in that, The specific operation process of the distributed resource scheduling module is as follows: Resource topology modeling: Construct a node resource graph and record the relevant attributes of each node; Task decomposition and priority ranking: Decompose the content publishing task into sub-tasks and assign priorities to the sub-tasks according to the priority of the decision instructions. Intelligent scheduling algorithm: ant colony algorithm is used to optimize task allocation; execution monitoring and dynamic adjustment: task progress is tracked in real time. If the load of a node exceeds the threshold, load migration is triggered, some tasks are migrated to backup nodes, the resource graph is updated and the optimal path is recalculated.

5. The distributed multimedia publishing system based on environmental interaction control according to claim 1, characterized in that, The operation of the multimodal interaction feedback module includes: interaction data collection, intent recognition, feedback generation and execution, and interaction log recording.

6. The distributed multimedia publishing system based on environmental interaction control according to claim 1, characterized in that, The specific operation process of the quality monitoring module is as follows: Quality metric collection: Deploy probes at the release node to collect metric data including video frame rate, packet loss rate, and number of rendering errors; Anomaly detection: Predict the quality decline trend using an LSTM model, input historical metric sequences, and output the predicted quality value for the next 5 minutes. If the predicted value is lower than the set threshold, it is marked as an anomaly. Repair strategy execution: Trigger repair actions based on the anomaly type. For network congestion, switch to a backup CDN node; for node failure, restart the node or migrate the task to a healthy node. If there is an error, revert to the previous version and notify the operations and maintenance team.

7. The distributed multimedia publishing system based on environmental interaction control according to claim 6, characterized in that, The release quality monitoring module communicates with the release quality stability assessment module. The release quality monitoring module transmits release quality monitoring information to the release quality stability assessment module. The release quality stability assessment module is used to set the release period, evaluate and analyze the release quality stability within the release period, and send the quality stability normal signal or quality stability unqualified signal to the supervision center.

8. The distributed multimedia publishing system based on environmental interaction control according to claim 7, characterized in that, The specific analysis process for the release quality stability assessment module is as follows: All abnormal release information that occurs during the release period is obtained, and the number of times abnormal information is generated is marked as the abnormal release value. If the abnormal release value exceeds the preset abnormal release threshold, a quality stability failure signal is generated. If the published abnormal value does not exceed the preset published abnormal value threshold, the non-detection value and the time-averaged value are compared with the preset non-detection threshold and the preset time-averaged value threshold respectively. If the non-detection value or the time-averaged value exceeds the corresponding preset threshold, a quality stability failure signal is generated; otherwise, a quality stability normal signal is generated.

9. The distributed multimedia publishing system based on environmental interaction control according to claim 7, characterized in that, The quality stability assessment module is connected to the interactive tracking and judgment module. When the interactive tracking and judgment module receives a quality stability normal signal, it analyzes the interactive performance of the multimodal interactive feedback module during the distribution period and sends the normal or unqualified interactive signal to the regulatory center.

10. The distributed multimedia publishing system based on environmental interaction control according to claim 9, characterized in that, The specific analysis process of the interaction tracking and judgment module is as follows: if the no-response detection value exceeds the preset no-response detection threshold, an interaction failure signal is generated; if the no-response detection value does not exceed the preset no-response detection threshold, the interaction comprehensive judgment coefficient is calculated by weighted summation of the no-response detection value, the non-rapid response detection value, and the feedback time difference value; if the interaction comprehensive judgment coefficient exceeds the preset interaction comprehensive judgment coefficient threshold, an interaction failure signal is generated; otherwise, an interaction normal signal is generated.

Images