Dynamic adaptation method and device for display content of LED screen and medium

Abstract

The invention discloses a dynamic adaptation method and device for display content of an LED screen and a medium, and the method comprises the steps: collecting original feature data of audiences in a radiation range of the LED screen through a multi-mode biosensor, carrying out the preprocessing of the original feature data, and generating a desensitization feature vector through local anonymization processing; performing real-time analysis on the desensitization feature vector by using a multi-modal fusion model in an edge computing node to generate a primary emotion tag; uploading the primary emotion label and the associated historical behavior data of the audience to a cloud platform so as to predict potential interests of the audience through a preset emotion deep analysis model, and generating a dynamic content adaptation strategy in combination with the potential interests of the audience; on the basis of a dynamic content adaptation strategy, display content matched with the emotion label is called from a pre-stored material library, low-delay rendering is conducted on the display content, and color parameters and the content display sequence of the LED display screen are dynamically adjusted according to the emotion analysis result.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and in particular, to a method, device, and medium for dynamically adapting the display content of an LED screen. Background Art

[0002] Traditional LED display systems mainly adjust display parameters through preset programs or environmental sensors. Their optimization goals are limited to basic physical indicators such as brightness and refresh rate, lacking the ability to perceive and respond to the subjective emotional states of viewers. Such systems cannot dynamically adapt content according to the emotions of viewers, resulting in low advertising conversion efficiency or insufficient public information dissemination effects.

[0003] In existing LED control solutions based on artificial intelligence, attempts are made to introduce single-modal data such as human pose detection to optimize the interaction experience, but the data dimension limitations are strong, making it difficult to accurately identify complex emotional states. In addition, traditional solutions mostly adopt a cloud centralized processing mode, and biometric sensing data needs to be remotely transmitted to the cloud for analysis, which not only introduces significant delays but also poses a risk of leakage of sensitive biometric information, making it difficult to meet the dual requirements of real-time interaction and privacy protection in public places.

[0004] Furthermore, there are obvious deficiencies in the multi-modal data fusion and dynamic decision-making levels in the existing technology. Most systems adopt a fixed-weight data fusion method, which cannot dynamically adjust the contribution degrees of different sensor data according to scene changes or emotional states, resulting in a large deviation between the analysis results and the real emotions. At the same time, the mapping logic between the display content and the emotional state is single, lacking refined control over the color space and content scheduling strategy, making it difficult to effectively guide the attention of viewers. In addition, the detection and response mechanisms for special states such as viewer fatigue in existing solutions are weak, and it is impossible to relieve fatigue or enhance the interaction willingness through display parameter optimization and content strategy adjustment. Summary of the Invention

[0005] Embodiments of this application provide a method, device, and medium for dynamically adapting the display content of an LED screen to solve the above technical problems.

[0006] On the one hand, embodiments of this application provide a method for dynamically adapting the display content of an LED screen, including: Collecting the original feature data of viewers within the radiation range of the LED screen through multi-modal biosensors, preprocessing the original feature data, and generating a desensitized feature vector through local anonymization processing; the original feature data includes facial expression data, voice feature data, and physiological signal data; Using a multi-modal fusion model in an edge computing node to perform real-time analysis on the desensitized feature vector and generate a primary emotion label; Upload the primary emotion tags and associated historical behavior data of the audience to the cloud platform to predict the potential interests of the audience through a preset in-depth emotion analysis model, and generate a dynamic content adaptation strategy in combination with the potential interests of the audience; Based on the dynamic content adaptation strategy, call the display content matching the emotion tags from the pre-stored material library to perform low-latency rendering on the display content, and dynamically adjust the color parameters and content display order of the LED display screen according to the emotion analysis results.

[0007] In one implementation of the present application, the original feature data of the audience within the radiation range of the LED screen is collected through a multi-modal biosensor, specifically including: Based on the high-frame-rate infrared camera and 3D structured light sensor embedded at the border of the LED screen, collect the facial expressions, micro-expressions, and head pose data of the audience within the radiation range of the LED screen; Based on the directional microphone array deployed on the LED screen, collect the voice data of the audience within the radiation range of the LED screen, and extract the intonation features and speech rate features in the voice data through voiceprint separation technology; Detect the heart rate and breathing rate of the audience within the radiation range of the LED screen through a non-contact millimeter-wave radar, and combine with a thermal imaging camera to obtain the body surface temperature distribution data of the audience within the radiation range of the LED screen.

[0008] In one implementation of the present application, predict the potential interests of the audience through a preset in-depth emotion analysis model, and generate a dynamic content adaptation strategy in combination with the potential interests of the audience, specifically including: Construct an in-depth emotion analysis model based on the Transformer architecture, and obtain the historical behavior data within the radiation range of the LED screen; the historical behavior data includes the stay duration and interaction frequency; Through the trained in-depth emotion analysis model, combine the historical behavior data and the primary emotion tags to predict the potential interests of the audience, and determine the emotion type corresponding to the potential interests of the audience; When the emotion type is fatigue, push interactive games or dynamic advertisements to enhance the attention of the audience; When the emotion type is excitement, extend the display duration of the corresponding advertisement, and match the corresponding advertisement with a preset high-saturation color.

[0009] In one implementation of the present application, preprocess the original feature data and generate a desensitized feature vector through local anonymization processing, specifically including: Perform spatio-temporal alignment and noise filtering on the original feature data; Through the federated learning framework, local anonymization processing is performed on the sensitive information in the preprocessed original feature data, and only the desensitized feature vectors related to emotion recognition are retained; the sensitive information includes face features; Upload the desensitized feature vectors to the cloud platform, and automatically destroy the original feature data after storing it locally for a preset duration.

[0010] In an implementation manner of the present application, a multimodal fusion model in the edge computing node is used to perform real-time analysis on the desensitized feature vectors to generate primary emotion labels, specifically including: Based on the multimodal fusion model pre-deployed in the edge computing node, an attention mechanism is adopted, and according to the current environmental light conditions or the audience behavior state, the input weights of the facial expression data and the physiological signal data are adjusted; In an emotional fluctuation scenario, preferentially process the feature fusion between the voice feature data and the heart rate data in the physiological signal data, generate primary emotion labels in real time, and trigger a fast response strategy; the primary emotion labels include excitement, fatigue, and neutral.

[0011] In an implementation manner of the present application, the color parameters and content display order of the LED display screen are dynamically adjusted according to the emotion analysis results, specifically including: Map the primary emotion labels to the HSL color space, and select the corresponding hue, saturation, and brightness combination according to the emotion type in the primary emotion labels; Optimize the advertisement playback order through a reinforcement learning algorithm to maximize the audience stay duration and interaction probability.

[0012] In an implementation manner of the present application, low-latency rendering is performed on the display content, specifically including: Use a hardware accelerator to perform parallel rendering on the display content to ensure that the rendering latency is lower than a preset threshold; During the rendering process, the color engine parameters corresponding to the primary emotion labels are matched in real time to dynamically adjust the screen brightness and color temperature.

[0013] In an implementation manner of the present application, it further includes: Analyze the eye closure frequency, head posture deviation angle, and heart rate variability of the audience to determine the audience fatigue level; Trigger an eye protection mode according to the fatigue level, dynamically reduce the blue light ratio of the screen and simplify the interface information, and at the same time push voice navigation prompts or emergency warning content.

[0014] On the other hand, an embodiment of the present application further provides a device for dynamically adapting the display content of an LED screen, and the device includes: At least one processor; And a memory communicatively connected to the at least one processor; Wherein, the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute a method for dynamically adapting the display content of an LED screen as described above.

[0015] On the other hand, an embodiment of the present application also provides a non-volatile computer storage medium storing computer-executable instructions, and when the computer-executable instructions are executed, a method for dynamically adapting the display content of an LED screen as described above is implemented.

[0016] The embodiment of the present application provides a method, device and medium for dynamically adapting the display content of an LED screen, which at least include the following beneficial effects: By fusing multi-dimensional data such as facial expressions, voice features and physiological signals, and dynamically adjusting the sensor weights in combination with the attention mechanism, the defect that single-modal data is vulnerable to environmental interference is effectively overcome, and the accuracy and robustness of emotion recognition are significantly improved. Secondly, the edge computing node localizes the data and generates primary emotion labels, greatly reducing the decision-making delay caused by the traditional cloud centralized processing mode. At the same time, the biometric data is anonymized through the federated learning framework, and only the desensitized feature vectors are uploaded to the cloud, solving the risk of sensitive information leakage and meeting the real-time interaction requirements and privacy compliance requirements in public places. In addition, based on the dynamic content optimization strategy and hardware acceleration rendering technology, the system can adjust the color space mapping and advertisement playback order in real time according to the emotion analysis results, so as to accurately guide the audience's attention and improve the advertisement conversion rate and information transmission efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The schematic embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings: Figure 1 It is a schematic flowchart of a method for dynamically adapting the display content of an LED screen provided by an embodiment of the present application; Figure 2 It is a schematic internal structure diagram of a device for dynamically adapting the display content of an LED screen provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] To make the objectives, technical solutions, and advantages of this application more clear, the following will clearly and completely describe the technical solutions of this application in combination with specific embodiments of this application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of this application.

[0019] The following will, in combination with the drawings, elaborate on the technical solutions provided by each embodiment of this application.

[0020] Figure 1 It is a schematic flowchart of a method for dynamically adapting the display content of an LED screen provided by an embodiment of this application.

[0021] The implementation of the analysis method involved in the embodiments of this application can be a terminal device or a server, and this application does not make special restrictions on this. For the convenience of understanding and description, the following embodiments will be described in detail taking the server as an example.

[0022] It should be noted that this server can be a single device or a system composed of multiple devices, that is, a distributed server, and this application does not make specific limitations on this.

[0023] As Figure 1 shown, a method for dynamically adapting the display content of an LED screen provided by an embodiment of this application includes: Step 101: Collect the original feature data of the audience within the radiation range of the LED screen through a multimodal biosensor, preprocess the original feature data, and generate a desensitized feature vector through local anonymization processing.

[0024] It should be noted that the original feature data in the embodiments of this application includes facial expression data, voice feature data, and physiological signal data.

[0025] In this embodiment, the multimodal biosensor includes a vision module, a voice module, and a physiological signal module. The vision module is composed of a high-frame-rate infrared camera and a 3D structured light sensor. Exemplarily, the infrared camera is used to capture facial micro-expressions in a low-light environment, such as twitching of the corners of the mouth and pupil dilation, while the 3D structured light sensor projects structured light spots and analyzes the reflected deformation to accurately calculate the head pose angles, such as pitch angle and yaw angle. In a shopping mall scenario, when the audience approaches the LED screen, the vision module can track the changes in their facial expressions in real time, providing basic data for emotion recognition.

[0026] The voice module collects the voice data of the audience through a directional microphone array, uses beamforming technology to filter out environmental noise, and extracts intonation features and speech rate features, such as pitch and speech rate, through a voiceprint separation algorithm. Exemplarily, in a shopping mall scenario, when the audience talks to the screen, the system can distinguish hesitation or excitement in their voices.

[0027] The physiological signal module uses a non-contact millimeter-wave radar to detect heart rate and breathing rate. The principle is to emit millimeter-wave signals and receive the Doppler frequency shift caused by the movement of the human chest to calculate the physiological rhythm. At the same time, a thermal imaging camera generates a temperature distribution map by capturing the infrared radiation on the body surface, such as the temperature difference between the forehead and the hand, for assisting in the judgment of the emotional state.

[0028] In this embodiment, the preprocessing of the original feature data includes spatio-temporal alignment and noise filtering. It can be understood that spatio-temporal alignment can ensure the spatio-temporal consistency of different sensor data through timestamp synchronization and spatial coordinate mapping. For example, aligning the time window of voice data with the frame rate of facial expression capture to avoid temporal deviation during emotion analysis. Noise filtering uses a wavelet transform algorithm to eliminate environmental interference, such as filtering out the interference of the self-light reflection of the LED screen on the camera data.

[0029] The local anonymization process is implemented through a federated learning framework. Specifically, the face features are encoded in a blurred manner, such as only retaining the features of the eye area, and the voiceprint data is spectrally desensitized, such as removing the frequency bands that can identify the timbre, to generate a desensitized vector that only contains the features required for emotion recognition. The original data is automatically destroyed after being stored locally for a preset duration to ensure compliance with privacy protection requirements.

[0030] In this embodiment, a sensor array is embedded in the border of the LED screen, the camera spacing ≤ 10 cm, the radar coverage angle is 120°, and the edge computing node uses NVIDIA Jetson AGX Orin with a computing power of 200 TOPS. When it is detected that the audience enters within 3 meters, multi-modal data collection is started.

[0031] In this embodiment, intelligent information screens in transportation hubs such as airports and high-speed railway stations need to realize the perception of passenger status and accurate information push in complex lighting and crowded environments, and at the same time solve the problem of attention dispersion caused by passenger fatigue.

[0032] The vision module uses a near-infrared wide-angle camera (120° FOV) and a ToF depth sensor, supports synchronous detection of multiple people at a long distance (5 - 10 meters), and eliminates crowd occlusion interference through binocular disparity algorithms. The physiological signal module integrates a 60GHz millimeter-wave radar array, non-contact detects heart rate variability (HRV) and respiratory rate, and combines a thermal imaging camera to monitor the change of the body surface temperature gradient, such as the temperature difference between the forehead and the hand, for fatigue grading, including mild, moderate, and severe. The environmental perception module deploys dust sensors (PM2.5 / PM10) and temperature and humidity sensors around the LED screen to dynamically correct display parameters, such as increasing the contrast in hazy weather.

[0033] Step 102: Use the multi-modal fusion model in the edge computing node to perform real-time analysis on the desensitized feature vectors to generate primary emotion labels.

[0034] In this embodiment, the edge computing node deploys a lightweight multi-modal fusion model (such as MMF-Net), the core of which is the attention mechanism. This mechanism dynamically adjusts the weights of different sensor data according to the current environmental scene (such as light intensity, crowd density) and the viewer's behavior state (such as standing, walking). For example, in a shopping mall environment with sufficient light and sparse crowds, the multi-modal fusion model enhances the weight of the vision module and preferentially analyzes facial micro-expressions and head postures. In a noisy transportation hub scene, the multi-modal fusion model increases the weights of the voice module and the physiological signal module. For example, it combines an elevated heart rate and an accelerated speech rate to determine an anxious emotion.

[0035] Specifically, when it detects that the viewer's speech intonation is rapid and the heart rate fluctuates significantly, the model triggers an excitement label. If the viewer remains stationary for a long time and the body temperature drops, it is determined to be in a fatigue state in combination with a decreased respiratory rate. The generated primary emotion labels include excitement, fatigue, and neutral, which will trigger a fast response strategy, such as immediately switching to a high-contrast screen or simplifying the interface information. It can be understood that the local processing of edge computing significantly reduces the decision-making delay. For example, in the advertising screen scenario, when the viewer shows interest, the edge node can generate emotion labels and trigger content switching within milliseconds, avoiding the response lag caused by traditional cloud processing.

[0036] In this embodiment, the Qualcomm QCS8550 chipset is adopted, which supports parallel processing of multi-modal data (with a computing power of 45 TOPS). A lightweight fatigue recognition model is locally deployed, based on the MobileNetV3+Bi-LSTM architecture, and outputs the fatigue level and emotion labels in real time.

[0037] Step 103: Upload the primary emotion labels and the associated historical behavior data of the viewer to the cloud platform to predict the viewer's potential interest through a preset emotion depth analysis model, and generate a dynamic content adaptation strategy in combination with the viewer's potential interest.

[0038] In this embodiment, the cloud platform constructs an emotion in-depth analysis model based on the Transformer architecture. Its inputs include primary emotion tags and historical behavior data, such as residence duration and interaction frequency. The emotion in-depth analysis model mines the correlation between emotion tags and historical behaviors through the self-attention mechanism to predict the potential interests of the audience. For example, if a certain audience responds positively to technology-related advertisements in multiple interactions, when it is detected that their current emotion is excitement, the emotion in-depth analysis model predicts that they may be interested in the promotion of new products, and then generates a strategy of "extending the display duration of technology advertisements". If the fatigue label appears for a certain audience in the transportation hub scene multiple times, the emotion in-depth analysis model combines its residence location data to generate a strategy of pushing the navigation of the nearby rest area.

[0039] The generation of the dynamic content adaptation strategy also includes scenario-based logic. Specifically, when the model determines that the audience is in a fatigue state, interactive games or high-dynamic advertisements are pushed to enhance attention. If it is in an excited state, the display duration of relevant advertisements is extended and high-saturation colors are matched. In addition, the cloud platform continuously optimizes the strategy through reinforcement learning. For example, the content recommendation priority is dynamically adjusted according to the advertisement click-through rate.

[0040] It can be understood that the collaborative computing mode between the cloud and the edge not only ensures the accuracy of in-depth analysis but also meets the real-time requirements through the fast response of the edge nodes. For example, in the shopping mall scene, the optimization strategy sent by the cloud can update the content library of the edge nodes in real time to ensure that the advertisement content is iterated synchronously with the interests of the audience.

[0041] Step 104: Based on the dynamic content adaptation strategy, call the display content that matches the emotion tag from the pre-stored material library to perform low-latency rendering of the display content, and dynamically adjust the color parameters and content display order of the LED display according to the emotion analysis result.

[0042] In this embodiment, the pre-stored material library contains multiple versions of content, such as videos, images, and texts. Each piece of material is associated with emotion tags and scene metadata, such as time period and geographical location. Exemplarily, for the excited emotion, dynamic advertisements with high-saturation colors, such as high-saturation red-based advertisements, are stored in the material library, and dynamic special effects are combined to enhance the visual impact. For the fatigue emotion, soothing images with cool colors, such as blue-based images, are matched. In addition, the system synthesizes personalized content in real time through Generative Adversarial Networks (GANs). For example, a virtual image interaction interface is generated according to the age of the audience.

[0043] The dynamic adjustment of color parameters is achieved through mapping in the HSL color space. Specifically, emotion labels are mapped to combinations of preset hue (H), saturation (S), and lightness (L). For example, a neutral emotion corresponds to natural hues and medium saturation, while a fatigued emotion is mapped to warm colors with low lightness to relieve visual fatigue.

[0044] The content display order is optimized by a reinforcement learning algorithm such as Q-learning. For example, when it is detected that the viewer is about to leave, the system preferentially plays limited-time discount information to increase the conversion probability. If the viewer stays for a long time, deep interactive content is gradually pushed to extend the stay duration. Finally, the rendered content is output to the LED screen, and the emotional changes of the viewer are continuously monitored to form a closed-loop feedback.

[0045] The display controller is implemented using an FPGA or GPU hardware accelerator to achieve low-latency rendering. It can be understood that the parallel computing architecture of the FPGA can complete complex rendering tasks in an extremely short time, such as real-time synthesis of virtual avatar interaction interfaces. During the rendering process, the color engine dynamically adjusts the picture parameters according to the emotion labels. For example, the contrast is increased in the "excited" state, or the proportion of blue light is reduced in the "fatigued" state.

[0046] Exemplarily, in the scenario of a mall advertising screen, when a viewer stands in front of the screen, the system detects an increase in their heart rate through a millimeter-wave radar, and at the same time the voice module recognizes a brisk intonation. The edge node fuses the data to generate an excited label. The cloud model combines the viewer's historical browsing records (such as frequently clicking on clothing advertisements) to generate a strategy of "extending the display duration of clothing advertisements + high-saturation colors". The FPGA controller completes the rendering within milliseconds, and the screen dynamically switches to the corresponding content, significantly increasing the advertisement click-through rate.

[0047] In this embodiment, the determination of the viewer's fatigue level and the linkage control of the eye protection mode are carried out. It should be noted that the percentage of eye closure (PERCLOS) refers to the proportion of the eye closure duration within a unit time, and the eye movement data is captured in real time through a high-frame-rate camera in the vision module. Exemplarily, in the scenario of a transportation hub, the system locates the viewer's eye area through the YOLOv8 object detection algorithm and calculates the proportion of the area where the eyelid covers the pupil in consecutive frames. When it is detected that PERCLOS exceeds the preset threshold, it is initially determined that the viewer is in a fatigued state.

[0048] The head pose offset angle is calculated by a 3D structured light sensor, specifically including the pitch angle and the yaw angle. For example, when the viewer's head continuously droops (pitch angle greater than the preset range) or frequently shakes irregularly (yaw angle fluctuates violently), the accuracy of fatigue determination can be enhanced by combining the percentage of eye closure.

[0049] The heart rate variability (HRV) is collected by a non-contact millimeter-wave radar. It can be understood that HRV reflects the activity level of the autonomic nervous system, and the ratio of its low-frequency power (LF) to high-frequency power (HF) (LF / HF) can be used to evaluate the fatigue level. For example, when the LF / HF ratio increases, it indicates the activation of the sympathetic nerve, which may be caused by stress fatigue. Specifically, the system extracts HRV features through spectral analysis and performs multimodal fusion with visual and head pose data. For example, in a night environment, the weight of visual data is reduced, and the fatigue grading is preferentially based on physiological signals.

[0050] The trigger of the eye protection mode includes optical parameter adjustment and content strategy optimization. For optical parameter adjustment, the peak wavelength of blue light is shifted from 450 nm to 460 nm, and the 590 nm yellow light band is enhanced, so that the proportion of blue light is reduced to below the preset threshold, thereby reducing retinal stimulation. At the same time, the screen brightness is adaptively adjusted non-linearly according to the ambient light. For example, the brightness is reduced in a dim environment to avoid glare.

[0051] Moreover, non-critical information such as advertising banners is hidden, the font of the navigation logo is enlarged and the line spacing is increased. In the airport scenario, when it is detected that the passenger is moderately fatigued, the screen only highlights the boarding gate information and the countdown, and the rest of the content is dynamically folded. Multilingual voice navigation prompts are pushed through a directional speaker, such as the rest area is on the B2 floor, and combined with AR projection technology, laser arrows are projected on the side of the LED screen to guide the audience to the target area.

[0052] In this embodiment, in the traffic hub information screen scenario, the system determines that it is moderately fatigued by analyzing the audience's eye closure frequency (PERCLOS) and heart rate variability (HRV). The edge node triggers the eye protection mode, reduces the proportion of blue light on the screen and highlights the navigation information, and at the same time guides the audience to the rest area through AR projection. If severe fatigue is detected, a red warning box is displayed full screen, and the broadcast system is linked to play a voice reminder to ensure the safety of passengers.

[0053] The above is the method embodiment proposed in this application. Based on the same inventive concept, the embodiment of this application also provides a display content dynamic adaptation device for an LED screen, and its structure is as Figure 2 shown.

[0054] Figure 2 It is a schematic internal structure diagram of a display content dynamic adaptation device for an LED screen provided by the embodiment of this application. As Figure 2 shown, the device includes: At least one processor; And a memory communicatively connected to at least one processor; Among them, the memory stores instructions executable by at least one processor. The instructions are executed by at least one processor to enable the at least one processor to: Collect the original feature data of the audience within the radiation range of the LED screen through a multimodal biosensor, preprocess the original feature data, and generate a desensitized feature vector through local anonymization processing; the original feature data includes facial expression data, voice feature data, and physiological signal data; Use the multimodal fusion model in the edge computing node to perform real-time analysis on the desensitized feature vector to generate a primary emotion label; Upload the primary emotion label and the associated historical behavior data of the audience to the cloud platform to predict the potential interests of the audience through a preset emotion in-depth analysis model, and generate a dynamic content adaptation strategy in combination with the potential interests of the audience; Based on the dynamic content adaptation strategy, call the display content matching the emotion label from the pre-stored material library to perform low-latency rendering on the display content, and dynamically adjust the color parameters and content display order of the LED display screen according to the emotion analysis results.

[0055] The embodiment of the present application also provides a non-volatile computer storage medium storing computer-executable instructions that, when executed, are capable of: Collect the original feature data of the audience within the radiation range of the LED screen through a multimodal biosensor, preprocess the original feature data, and generate a desensitized feature vector through local anonymization processing; the original feature data includes facial expression data, voice feature data, and physiological signal data; Use the multimodal fusion model in the edge computing node to perform real-time analysis on the desensitized feature vector to generate a primary emotion label; Upload the primary emotion label and the associated historical behavior data of the audience to the cloud platform to predict the potential interests of the audience through a preset emotion in-depth analysis model, and generate a dynamic content adaptation strategy in combination with the potential interests of the audience; Based on the dynamic content adaptation strategy, call the display content matching the emotion label from the pre-stored material library to perform low-latency rendering on the display content, and dynamically adjust the color parameters and content display order of the LED display screen according to the emotion analysis results.

[0056] Each embodiment in the present application is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the key points of each embodiment are the differences from other embodiments. In particular, for the device and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments.

[0057] The devices, media, and methods provided by the embodiments of this application correspond one-to-one. Therefore, the devices and media also have beneficial technical effects similar to those of their corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the devices and media will not be elaborated here.

[0058] Those skilled in the art should understand that the embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0059] This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of this application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more flows and / or Figure 1 blocks.

[0060] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more flows and / or Figure 1 blocks.

[0061] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are performed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more flows and / or Figure 1 blocks.

[0062] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0063] The memory may include non - permanent memory in the form of computer - readable media, random access memory (RAM) and / or non - volatile memory such as read - only memory (ROM) or flash RAM. The memory is an example of computer - readable media.

[0064] Computer - readable media includes permanent and non - permanent, removable and non - removable media that can store information by any method or technology. The information can be computer - readable instructions, data structures, program modules or other data. Examples of the computer's storage media include, but are not limited to, phase - change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read - only memory (ROM), electrically erasable programmable read - only memory (EEPROM), flash memory or other memory technologies, compact disc read - only memory (CD - ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic disk storage or other magnetic storage devices or any other non - transitory media that can be used to store information that can be accessed by a computing device. As defined herein, computer - readable media does not include transitory computer - readable media such as modulated data signals and carrier waves.

[0065] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non - exclusive inclusion, such that a process, method, article or apparatus that comprises a series of elements includes not only those elements but also other elements not expressly listed, or elements that are inherent to such process, method, article or apparatus. Without further limitation, an element defined by the statement "comprising an …" does not exclude the presence of additional identical elements in the process, method, article or apparatus that comprises the element.

[0066] The above are only embodiments of the present application and are not used to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A method for dynamically adapting the display content of an LED screen, characterized in that, The method includes: Collecting original feature data of the audience within the radiation range of the LED screen through a multimodal biosensor, preprocessing the original feature data, and generating a desensitized feature vector through local anonymization processing; the original feature data includes facial expression data, voice feature data, and physiological signal data; Using a multimodal fusion model in the edge computing node to perform real-time analysis on the desensitized feature vector to generate a primary emotion label; Uploading the primary emotion label and associated historical behavior data of the audience to the cloud platform to predict the potential interests of the audience through a preset emotion depth analysis model, and generating a dynamic content adaptation strategy in combination with the potential interests of the audience; Based on the dynamic content adaptation strategy, calling display content matching the emotion label from a pre-stored material library, performing low-latency rendering on the display content, and dynamically adjusting the color parameters and content display order of the LED display screen according to the emotion analysis result.

2. The display content dynamic adaptation method of an LED screen according to claim 1, characterized in that Collecting original feature data of the audience within the radiation range of the LED screen through a multimodal biosensor, specifically including: Based on a high-frame-rate infrared camera and a 3D structured light sensor embedded in the LED screen frame, collecting facial expressions, micro-expressions, and head pose data of the audience within the radiation range of the LED screen; Based on a directional microphone array deployed on the LED screen, collecting voice data of the audience within the radiation range of the LED screen, and extracting intonation features and speech rate features in the voice data through voiceprint separation technology; Detecting the heart rate and breathing rate of the audience within the radiation range of the LED screen through a non-contact millimeter-wave radar, and combining with a thermal imaging camera to obtain the body surface temperature distribution data of the audience within the radiation range of the LED screen.

3. A method for dynamically adapting the display content of an LED screen according to claim 1, characterized in that, Predicting the potential interests of the audience through a preset emotion depth analysis model, and generating a dynamic content adaptation strategy in combination with the potential interests of the audience, specifically including: Constructing an emotion depth analysis model based on the Transformer architecture and obtaining the historical behavior data within the radiation range of the LED screen; the historical behavior data includes residence duration and interaction frequency; Through the trained emotion depth analysis model, combining the historical behavior data and the primary emotion label to predict the potential interests of the audience and determine the emotion type corresponding to the potential interests of the audience; When the emotion type is fatigue, pushing interactive games or dynamic advertisements to improve the attention of the audience; When the emotion type is excitement, extending the display duration of the corresponding advertisement and matching a preset high-saturation color for the corresponding advertisement.

4. A method for dynamically adapting the display content of an LED screen according to claim 1, characterized in that Preprocessing the original feature data and generating a desensitized feature vector through local anonymization processing, specifically including: Performing spatio-temporal alignment and noise filtering on the original feature data; Through the federated learning framework, performing local anonymization processing on the sensitive information in the preprocessed original feature data, and only retaining the desensitized feature vector related to emotion recognition; the sensitive information includes face features; Uploading the desensitized feature vector to the cloud platform, and automatically destroying the original feature data after storing it locally for a preset duration.

5. A method for dynamically adapting the display content of an LED screen according to claim 1, characterized in that, Utilize the multimodal fusion model in the edge computing node to perform real-time analysis on the desensitized feature vector and generate primary emotion labels, specifically including: Based on the multimodal fusion model pre-deployed in the edge computing node, adopt the attention mechanism, and adjust the input weights of facial expression data and physiological signal data according to the current environmental lighting conditions or the audience behavior state; In an emotional fluctuation scenario, prioritize the feature fusion between voice feature data and heart rate data in physiological signal data, generate primary emotion labels in real time, and trigger a fast response strategy; the primary emotion labels include excitement, fatigue, and neutral.

6. The method for dynamically adapting the display content of an LED screen according to claim 1, wherein Dynamically adjust the color parameters and content display order of the LED display according to the emotion analysis result, specifically including: Map the primary emotion label to the HSL color space, and select the corresponding hue, saturation, and brightness combination according to the emotion type in the primary emotion label; Optimize the advertisement playback order through a reinforcement learning algorithm to maximize the audience stay time and interaction probability.

7. A method for dynamically adapting the display content of an LED screen according to claim 1, characterized in that Perform low-latency rendering on the display content, specifically including: Adopt a hardware accelerator to perform parallel rendering on the display content to ensure that the rendering latency is lower than a preset threshold; During the rendering process, match the color engine parameters corresponding to the primary emotion label in real time to dynamically adjust the screen brightness and color temperature.

8. A method for dynamically adapting the display content of an LED screen according to claim 1, characterized in that, The method further includes: Analyze the eye closure frequency, head posture deviation angle, and heart rate variability of the audience to determine the audience fatigue level; Trigger an eye protection mode according to the fatigue level, dynamically reduce the proportion of blue light on the screen and simplify the interface information, and at the same time push voice navigation prompts.

9. A display content dynamic adaptation device for an LED screen, characterized in that, The device includes: At least one processor; And a memory communicatively connected to the at least one processor; Wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute a method for dynamically adapting the display content of an LED screen according to any one of claims 1-8.

10. A non-volatile computer storage medium stores computer-executable instructions, characterized in that, When the computer-executable instructions are executed, a method for dynamically adapting the display content of an LED screen according to any one of claims 1-8 is implemented.

Images