Iot-based rgb-mini led adaptive dimming and color adjusting method and system

Abstract

This invention relates to an IoT-based RGB-MiniLED adaptive dimming and color adjustment method and system, belonging to the field of RGB-MiniLED adjustment technology. This invention utilizes a large language model to output the user's current activity intention icon and scene atmosphere requirements, thereby collecting personalized preference information, energy-saving optimization information, physiological rhythm adjustment information, and scene adaptation information. Through a dialogue framework based on the large language model, it completes benefit negotiation, generates a globally coordinated dimming and color adjustment strategy, and finally outputs the target color temperature and brightness based on this strategy, controlling the system according to the target color temperature and brightness. This solution achieves an intelligent balance between meeting personalized lighting needs and scene matching through a collaborative negotiation mechanism. Compared to existing solutions, it can coordinate the lighting preferences of different users in a multi-user coexisting space and continuously optimize the control strategy without increasing user intervention, significantly improving the system's intelligence level, personalization adaptability, and energy efficiency.

Description

Technical Field

[0001] This invention relates to the field of RGB-MiniLED adjustment technology, and in particular to an RGB-MiniLED adaptive dimming and color adjustment method and system based on the Internet of Things. Background Technology

[0002] RGB-MiniLEDs face serious challenges in practical applications, including color shift degradation and environmental adaptation. Research shows that red, green, and blue LEDs exhibit significantly different emission peak shifts and full width at half maximum (FWHM) broadening patterns with varying temperatures and injected current densities. Blue and green LEDs show a blue shift as injected current density increases, while red LEDs exhibit a red shift. This electrothermal coupling effect causes the chromaticity coordinates of the three primary colors to continuously deviate from their initial values ​​under varying operating conditions, resulting in chromaticity deviations and color imbalances. Simultaneously, factors such as different home lighting conditions and varying living room lighting at different times of day place dynamic and personalized demands on the color and brightness output of display devices. However, current RGB-MiniLED products lack the ability to actively sense and adaptively adjust to environmental changes in practical use. They cannot perform real-time image quality compensation based on dynamic changes in ambient light and the characteristics of the image content, resulting in inconsistent image quality when viewed in varying environments such as day and night, and natural and artificial lighting.

[0003] Existing intelligent lighting systems typically employ a centralized control architecture based on a single algorithm. Their dimming and color tuning strategies rely on preset rules or limited machine learning models, making it difficult to understand the deep semantic relationships between complex user intentions, dynamic activity scenarios, and changes in the natural environment. More importantly, existing systems lack collaborative mechanisms between multiple agents, hindering intelligent trade-offs and dynamic decision-making among complex multi-dimensional objectives (personalized lighting preferences, group needs, activity scenario matching, and physiological rhythm regulation). Current lighting control lacks hierarchy and coordination, limiting display quality and personalized experiences. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides an RGB-MiniLED adaptive dimming and color adjustment method and system based on the Internet of Things.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: The first aspect of this invention provides an RGB-MiniLED adaptive dimming and color adjustment method based on the Internet of Things, comprising: After user authorization processing, ambient lighting parameters, user micro-movement status, and ambient acoustic parameters in the target area are collected to form multimodal sensing data; Using a large language model as the core, multimodal perception data is transformed into semantic scene descriptions. The large language model is used to output user current activity icons, scene atmosphere requirements descriptions, and lighting preference conflicts among multiple users. Collect personalized preference information, energy-saving optimization information, physiological rhythm regulation information, and scene adaptation information; Interest negotiation is completed within a dialogue framework based on a large language model. A globally coordinated dimming and color adjustment strategy is generated based on preset multi-objective weight constraints and dynamic priority algorithms. The target color temperature and brightness are output based on the globally coordinated dimming and color tuning strategy, and controlled according to the target color temperature and brightness.

[0006] Furthermore, in the IoT-based RGB-MiniLED adaptive dimming and color adjustment method, after user authorization processing, ambient lighting parameters, user micro-motion status, and ambient acoustic parameters in the target area are collected to form multimodal sensing data, specifically: Hyperspectral sensors, millimeter-wave radar human body micro-motion sensors, temperature and humidity sensors, air quality sensors, and sound feature sensors are deployed in various lighting areas to form a miniature sensor array; The aforementioned hyperspectral sensor, millimeter-wave radar human body micro-motion sensor, temperature and humidity sensor, air quality sensor, and sound characteristic sensor are used to collect ambient light parameters, user micro-motion status, and ambient acoustic parameters, respectively. Multimodal sensing data is constructed based on the ambient lighting parameters, user micro-movement status, and ambient acoustic parameters.

[0007] Furthermore, in the IoT-based RGB-MiniLED adaptive dimming and color adjustment method, a large language model is used as the core to transform multimodal perception data into semantic scene descriptions, specifically: Using a 500ms sliding window, mean, variance, peak value, and slope features are extracted from the sensor data within each window. Continuous numerical values ​​are mapped to language labels, and the feature vectors corresponding to the multimodal sensing data are converted into natural language fragments. If multiple users are detected, multiple independent personnel description segments are generated and personnel are distinguished. The microphone signal is converted into event labels through a pre-trained audio event recognition model and semantic fragments are inserted. Semantic fragments are concatenated in chronological order to form a complete large language model input prompt, and system instruction prefixes are added.

[0008] Furthermore, in the IoT-based RGB-MiniLED adaptive dimming and color adjustment method, a large language model is used to output the user's current activity icon, scene atmosphere requirement description, and lighting preference conflict representation among multiple users, specifically: Construct training datasets of simulated and real-world scenarios and label pairs, and use these datasets to train a large language model. The training loss function is cross-entropy plus a regularization term. Add stepwise reasoning process annotations to the training samples to guide the model to output the reasoning chain. Using the Mind Chain Distillation technique, the complex reasoning process is compressed into end-to-end lightweight model weights, the latest semantic fragments are appended to the historical prompts, and the key-value pairs of the previous reasoning are cached. The model weights are quantized using INT4, and the activation values ​​are FP16. After inference is completed, the large language model outputs a structured intent object, through which the user's current activity intent tag and scene atmosphere requirement description are obtained.

[0009] Furthermore, in the IoT-based RGB-MiniLED adaptive dimming and color adjustment method, personalized preference information and scene adaptation information are collected, specifically: Collect users' historical lighting data records, and construct user layer nodes, scene layer nodes, and lighting parameter layer nodes based on the user's historical lighting data records. Connect the user layer nodes, scene layer nodes, and lighting parameter layer nodes through directed edges to form a preference knowledge graph structure. The personalized feature vector of the current user is obtained using the aforementioned preference knowledge graph structure, and the predicted preference score is calculated based on the personalized feature vector of the current user using a Bayesian algorithm. Statistical prediction of preference scores is performed to obtain personalized color temperature and brightness preference score data under different scenarios and time periods. A preference score matrix is ​​constructed based on the personalized color temperature and brightness preferences under different scenarios and time periods.

[0010] Furthermore, the IoT-based RGB-MiniLED adaptive dimming and color tuning method also includes: A lighting scene template library is constructed using activity type and atmosphere requirement tags output by a large language model. Each activity-atmosphere combination corresponds to a set of template parameters and dynamic rules. Use the lighting scene template library to match the most suitable template for the current activity type and atmosphere requirements. If multiple matches exist, calculate the similarity with personalized preference information. When the similarity is lower than the preset similarity, the best matching template is fine-tuned. When the similarity is not lower than the preset similarity, the color temperature offset, dynamic change rate and color saturation parameters in the whole lighting parameters are output.

[0011] Furthermore, in the IoT-based RGB-MiniLED adaptive dimming and color adjustment method, interest negotiation is completed within a dialogue framework based on a large language model to generate a globally coordinated dimming and color adjustment strategy, specifically including: Negotiation is performed based on personalized preference information and scenario adaptation information in the large language model. The negotiation process takes place on the blackboard of the large language model, and the large language model writes the intent structure onto the blackboard. The system obtains the optimal output of personalized preference information, personalized physiological rhythm regulation information, and scene adaptation information, and generates a bid package, which includes a recommended lighting parameter vector, a bid satisfaction score, and the percentage of satisfaction that the agent is willing to sacrifice in each subsequent negotiation round. The blackboard collects all bid packages and checks for parameter conflicts. If there are no conflicts, it directly uses a weighted average. If there are conflicts, it performs round-by-round negotiation. During the negotiation, the large language model sends the conflict reasons and the current compromise suggestions. At the same time, based on the percentage of satisfaction that is willing to be sacrificed in each subsequent negotiation round, a decision is made on whether to lower the requirements, generate a new bid package, repeat conflict detection, and if conflicts still exist, a voting mechanism is initiated to select the most relevant parameter as the final execution parameter based on the dominant dimension of the current scenario. After arbitration, a globally coordinated dimming and color grading strategy is generated based on the final execution parameters.

[0012] A second aspect of the present invention provides an RGB-MiniLED adaptive dimming and color adjustment system based on the Internet of Things (IoT), including a memory and a processor. The memory includes an RGB-MiniLED adaptive dimming and color adjustment method program based on the IoT. When the processor executes the RGB-MiniLED adaptive dimming and color adjustment method program based on the IoT, it implements the steps of any of the RGB-MiniLED adaptive dimming and color adjustment methods based on the IoT described in the present invention.

[0013] This invention addresses the shortcomings of the prior art and has the following beneficial effects: This solution overcomes the limitations of existing single-algorithm lighting control systems. By leveraging a large language model for semantic understanding of multimodal perception data, it significantly improves the accuracy of recognizing complex user intentions. Through a collaborative negotiation mechanism, it achieves an intelligent balance between meeting personalized lighting needs and scene matching. Compared to existing solutions, this solution can coordinate the lighting preferences of different users in spaces where multiple people coexist, and continuously optimize the control strategy without increasing user intervention, significantly improving the system's intelligence level, personalization adaptability, and energy efficiency. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0015] Figure 1 A flowchart of the overall process for an IoT-based RGB-MiniLED adaptive dimming and color adjustment method is shown. Figure 2 A system block diagram of an IoT-based RGB-MiniLED adaptive dimming and color adjustment system is shown. Detailed Implementation

[0016] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0017] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0018] like Figure 1 As shown, the first aspect of the present invention provides an RGB-MiniLED adaptive dimming and color adjustment method based on the Internet of Things, comprising: After user authorization processing, ambient lighting parameters, user micro-movement status, and ambient acoustic parameters in the target area are collected to form multimodal sensing data; With a large language model as the core, multimodal perception data is transformed into semantic scene descriptions, and the large language model is used to output the user's current activity icon and scene atmosphere requirement description. Collect personalized preference information and scene adaptation information, complete interest negotiation in a dialogue framework based on a large language model, and generate a globally coordinated dimming and color adjustment strategy. The target color temperature and brightness are output based on a globally coordinated dimming and color adjustment strategy, and are controlled according to the target color temperature and brightness.

[0019] It should be noted that this solution overcomes the limitations of existing single-algorithm lighting control systems. By leveraging a large language model for semantic understanding of multimodal perception data, it significantly improves the accuracy of recognizing complex user intentions. Through a collaborative negotiation mechanism, it achieves an intelligent balance between meeting personalized lighting needs and scene matching. Compared to existing solutions, this solution can coordinate the lighting preferences of different users in spaces where multiple people coexist, and continuously optimize the control strategy without increasing user intervention, significantly improving the system's intelligence level, personalization adaptability, and energy efficiency.

[0020] Furthermore, in the IoT-based RGB-MiniLED adaptive dimming and color adjustment method, after user authorization processing, ambient lighting parameters, user micro-motion status, and ambient acoustic parameters in the target area are collected to form multimodal sensing data, specifically: Hyperspectral sensors, millimeter-wave radar human body micro-motion sensors, temperature and humidity sensors, air quality sensors, and sound feature sensors are deployed in various lighting areas to form a miniature sensor array; The system utilizes a hyperspectral sensor, a millimeter-wave radar human body micro-motion sensor, a temperature and humidity sensor, an air quality sensor, and a sound characteristic sensor to collect ambient light parameters, user micro-motion status, and ambient acoustic parameters, respectively. Multimodal sensing data is constructed based on ambient lighting parameters, user micro-movement status, and ambient acoustic parameters.

[0021] It should be noted that ambient lighting parameters include, but are not limited to, color temperature, illuminance, and spectral power distribution; user micro-movement states include, but are not limited to, breathing frequency and body movement amplitude. Ambient acoustic parameters include sound pressure level, equivalent continuous sound level, A-weighted sound level, time characteristic parameters, and frequency characteristic parameters.

[0022] Furthermore, in the IoT-based RGB-MiniLED adaptive dimming and color adjustment method, a large language model is used as the core to transform multimodal perception data into semantic scene descriptions, specifically: Using a 500ms sliding window, mean, variance, peak value, and slope features are extracted from the sensor data within each window. Continuous numerical values ​​are mapped to language labels, and the feature vectors corresponding to the multimodal sensing data are converted into natural language fragments. For example, continuous numerical values ​​are mapped to 5 levels of language labels (very low / lower / moderate / higher / very high). For instance, illuminance <50 lx is mapped to "very dark", 50~200 lx to "lower", 200~500 lx to "moderate", 500~1000 lx to "brighter", and >1000 lx to "very bright".

[0023] Use a preset template to convert the feature vector into a natural language fragment, for example: [Timestamp: T+0.5s] Area A: Illumination level "dark", color temperature 3800K, blue light component of spectrum 32%; Human activity: Radar detected 1 adult male, respiratory rate 16 breaths / minute, trunk micro-movement amplitude 0.3cm (static sitting posture). Environment: Temperature 24.5℃, humidity 45%, background noise 42dB (quiet conversation environment).

[0024] If multiple users are detected, multiple independent personnel description segments are generated and personnel are distinguished. The microphone signal is converted into event labels through a pre-trained audio event recognition model and semantic fragments are inserted. If multiple users are detected, multiple independent personnel description segments are generated and distinguished by "Person 1 / Person 2". The microphone signal is converted into event labels (such as "page turning", "keyboard typing", "footsteps approaching") through a pre-trained audio event recognition model and inserted into semantic segments.

[0025] Semantic fragments are concatenated in chronological order to form a complete large language model input prompt, and system instruction prefixes are added.

[0026] It should be noted that the most recent 30 semantic segments (corresponding to 15 seconds of duration) are concatenated in chronological order to form a complete large language model input prompt, with the addition of system instruction prefixes, such as: the main activity type (selected from [reading, dining, resting, socializing, video watching, fine crafts, sports, learning]); the main user's lighting intent (if there are multiple users, output the intent and priority of each user); the scene atmosphere requirements (emotional tone: calm / warm / active / focused / romantic / neutral); and any abnormal or conflict prompts.

[0027] Furthermore, in the IoT-based RGB-MiniLED adaptive dimming and color adjustment method, a large language model is used to output the user's current activity icon, scene atmosphere requirement description, and lighting preference conflict representation among multiple users, specifically: Construct training datasets of simulated and real-world scenarios and label pairs, and use these datasets to train a large language model. The training loss function is cross-entropy plus a regularization term. Add stepwise reasoning process annotations to the training samples to guide the model to output the reasoning chain. It should be noted that a training dataset containing tens of thousands of simulated and real-world scene-label pairs was constructed. Each sample consisted of: [semantic fragment sequence] → [activity type, intent list, atmosphere label, conflict marker]. "Step-by-step reasoning" process annotations were added to the training samples to guide the model in outputting the reasoning chain. For example: Inference: The user's breathing is steady, the micro-movements are small, and there is no sound of speaking -> the user may be in a static activity; There is the sound of turning pages and the illuminance is below the recommended reading illuminance threshold -> the user may have an implicit need to increase the brightness; A second person was detected approaching; their historical preference is warm light -> potential conflict.

[0028] Output: Activity type = Reading, User 1 intent = Brighten to 500lx / 4500K, User 2 preference = 2700K, Conflict = Color temperature preference difference.

[0029] Using the Mind Chain Distillation technique, the complex reasoning process is compressed into end-to-end lightweight model weights, the latest semantic fragments are appended to the historical prompts, and the key-value pairs of the previous reasoning are cached. The model weights are quantized using INT4, and the activation values ​​are FP16. Since there is more than 90% overlap between semantic segments in consecutive time windows, only the latest semantic segment is appended to the historical hints, and the key-value pairs of the previous inference are cached to avoid duplicate calculations.

[0030] After inference is completed, the large language model outputs a structured intent object, through which the user's current activity intent tag and scene atmosphere requirement description are obtained.

[0031] Furthermore, in the IoT-based RGB-MiniLED adaptive dimming and color adjustment method, personalized preference information and scene adaptation information are collected, specifically: Collect users' historical lighting data records, and construct user layer nodes, scene layer nodes, and lighting parameter layer nodes based on the user's historical lighting data records. Connect the user layer nodes, scene layer nodes, and lighting parameter layer nodes through directed edges to form a preference knowledge graph structure. The personalized feature vector of the current user is obtained by using the preference knowledge graph structure, and the predicted preference score is calculated based on the personalized feature vector of the current user using the Bayesian algorithm. The system statistically predicts preference scores, obtains personalized color temperature and brightness preference score data for different scenarios and time periods, constructs a preference score matrix based on personalized color temperature and brightness preferences for different scenarios and time periods, and outputs a list of optimal lighting parameter candidates, such as personalized color temperature and brightness parameters.

[0032] It should be noted that the user layer nodes include user ID, age tag, lighting preference vector, etc., while the scene layer nodes include activity type, time granularity (e.g., morning / daytime / evening / nighttime), and season. The lighting parameter layer nodes include color temperature range, illuminance range, and color preference. For user u in scene s with unobserved lighting parameter combinations (cct, lux), predict their preference score: This represents the global average preference score, the baseline preference value for all users, all scenes, and all combinations of lighting parameters. For example, if statistical analysis of all historical data reveals that the overall average user score for lighting is 3.5 (assuming a rating range of 1-5), then... . This represents the inherent bias of user X, reflecting whether the user generally prefers high or low ratings. For example, if a user is very picky about all lighting conditions and has an average rating of only 2.0, then... Another user was very satisfied with most of the lighting, with an average rating of 4.5. . This is an inherent bias of scene s. It reflects the impact of a specific activity or environmental scene itself on the lighting evaluation. For example, a "reading" scene generally requires high illuminance, and users tend to give it a higher rating (even if the lighting parameters are average). The value is positive; in the "rest" scenario, even if the lighting parameters are perfectly appropriate, users may not give an extremely high score (because perception is not sharp in a relaxed state). Close to zero or slightly negative. For user-scenario interaction items, Let be the latent vector of the user, and each user is represented as A point in 3D space captures the user's lighting preference patterns, such as a preference for warm light and low illuminance, or cool light and high illuminance, or a liking for dynamic changes. These are the latent vectors of the scene. Similarly... Dimension, representing the feature orientation corresponding to the "ideal lighting parameters" in this scene. Inner product The inner product measures the degree of fit between the user and the scene. If the user's preferences align with the scene's needs (large inner product), the user is more likely to achieve high satisfaction in that scene; conversely, if they don't align (small or even negative inner product), the user will still be dissatisfied even if the lighting parameters themselves are reasonable. This represents the user's individual preferences regarding lighting parameters. Even without considering the specific scenario, users have preferences for certain... Combinations have a natural preference tendency—an underlying preference unrelated to the scene (such as someone naturally preferring warm light). User-scene interaction items, on the other hand, capture scene-dependent preferences.

[0033] Furthermore, the IoT-based RGB-MiniLED adaptive dimming and color tuning method also includes: A lighting scene template library is constructed using activity type and atmosphere requirement tags output by a large language model. Each activity-atmosphere combination corresponds to a set of template parameters and dynamic rules. For example, in one rule for a reading scenario, the atmosphere label is "focused," color temperature (K) is 4500-5000, illuminance (lx) is 500-60, dynamic rate of change effect is static and flicker-free, and RGB auxiliary is a slight increase in the G channel to enhance contrast. In another rule for a reading scenario, the atmosphere label is "calm," color temperature (K) is 3000-3500, illuminance (lx) is 300-400, dynamic rate of change effect is static, and RGB auxiliary is absent. In yet another rule for a dining scenario, the atmosphere label is "gentle," color temperature (K) is 2700-3000, illuminance (lx) is 150-250, dynamic rate of change effect is a slow fluctuation period of 2 minutes, amplitude ±30lx, and RGB auxiliary is slightly red (5% intensity).

[0034] Use the lighting scene template library to match the most suitable template for the current activity type and atmosphere requirements. If multiple matches exist, calculate the similarity with personalized preference information. When the similarity is lower than the preset similarity, the best matching template is fine-tuned. When the similarity is not lower than the preset similarity, the color temperature offset, dynamic change rate and color saturation parameters in the whole lighting parameters are output.

[0035] It should be noted that this method combines the similarity with personalized preference information to fine-tune the best matching template, thereby outputting the color temperature offset, dynamic change rate, and color saturation parameters in the overall lighting parameters.

[0036] Furthermore, in the IoT-based RGB-MiniLED adaptive dimming and color adjustment method, interest negotiation is completed within a dialogue framework based on a large language model to generate a globally coordinated dimming and color adjustment strategy, specifically including: Negotiation is performed based on personalized preference information and scenario adaptation information in the large language model. The negotiation process takes place on the blackboard of the large language model, and the large language model writes the intent structure onto the blackboard. The optimal output of personalized preference information and scene adaptation information is obtained, and a bid package is generated. The bid package includes the recommended lighting parameter vector, the bid satisfaction score, and the percentage of satisfaction that is willing to be sacrificed in each subsequent negotiation round. The blackboard collects all bid packages and checks for parameter conflicts. If there are no conflicts, it directly uses a weighted average. If there are conflicts, it performs round-by-round negotiation. During the negotiation, the large language model sends the conflict reasons and the current compromise suggestions. At the same time, based on the percentage of satisfaction that is willing to be sacrificed in each subsequent negotiation round, a decision is made on whether to lower the requirements, generate a new bid package, repeat conflict detection, and if conflicts still exist, a voting mechanism is initiated to select the most relevant parameter as the final execution parameter based on the dominant dimension of the current scenario. After arbitration, a globally coordinated dimming and color grading strategy is generated based on the final execution parameters.

[0037] It should be noted that the blackboard collects all bid packages and checks for parameter conflicts. A conflict is defined as: recommended... Deviation > 1000K, or Bias > 200 lx. If there is no conflict, use the weighted average directly (the weights are the domain weights of each agent in this scenario, preset values: preference 0.6, scenario 0.4).

[0038] If a conflict exists, a negotiation round is executed (maximum 3 rounds): Round 1: The large language model sends the "reason for conflict" and "current compromise suggestion." For example: the preference parameter requires 5000K, the scenario recommendation parameter requires 2700K, and it is suggested that both parties compromise to 3850K. Based on the percentage of satisfaction that is willing to be sacrificed in each subsequent negotiation round, it is decided whether to lower the requirement. For example, the preference parameter may not be willing to lower the energy-saving target, but the scenario recommendation parameter can compromise by 20%. This generates a new bid package, and conflict detection is repeated. If a conflict still exists after 3 rounds, a voting mechanism is initiated: based on the dominant dimension of the current scenario, such as increasing the preference weight in daytime work scenarios and increasing the weight in late-night scenarios, the most relevant parameter is selected as the final execution parameter.

[0039] It should be noted that resolving conflicts through standardized negotiation protocols avoids the masking effect of traditional weighted sum methods. The system dynamically adjusts its negotiation behavior based on the urgency of its objectives, making it more adaptable and robust. The large language model not only handles intent understanding but also acts as a neutral coordinator in the negotiation process, using its natural language generation capabilities to offer understandable compromise suggestions to each agent, thus enhancing the interpretability and transparency of the entire decision-making process.

[0040] like Figure 2 As shown, the second aspect of the present invention provides an RGB-MiniLED adaptive dimming and color adjustment system based on the Internet of Things (IoT), including a memory and a processor. The memory includes an RGB-MiniLED adaptive dimming and color adjustment method program based on the IoT. When the RGB-MiniLED adaptive dimming and color adjustment method program based on the IoT is executed by the processor, it implements the steps of any one of the RGB-MiniLED adaptive dimming and color adjustment methods based on the IoT.

[0041] In addition, this method also includes: By acquiring the spectral characteristics of the user's facial reflection by spectral sensors set up in the target area, when multiple users coexist in the same lighting space, the spatial position, head orientation and movement trajectory of each user are tracked in real time, and the spectral characteristics are associated with the spatial position to build a real-time profile for each user; Based on the user profile, when a single user occupies the dominant activity area, the dimming and color adjustment are based on the user's personalized lighting preferences. When multiple users are active at the same time, a comprehensive weighted decision is made based on the spatial distance weight of each user, the priority of the activity type, and the weight of the user's preset preferences, and the interaction characteristics between users and the environment are continuously monitored. By analyzing the user's head orientation and gaze point, the system determines the area where the user's attention is focused, enhances the lighting brightness of the target area, adjusts the lighting parameters of the forward area to eliminate control delay by predicting the user's movement trajectory, and adjusts the lighting output to achieve a smooth transition by predicting the trend of environmental parameter changes.

[0042] It should be noted that when a single user dominates the activity area, the system adjusts the brightness and color temperature based on that user's personalized lighting preferences. When multiple users are active simultaneously, a comprehensive weighted decision is made based on the spatial distance weights of each user (the user closest to the lighting fixture has the highest weight), the priority of the activity type (detailed visual tasks > general activities > background crowds), and the user's preset preference weights. The system continuously monitors the interaction characteristics between users and the environment, including: user head orientation and gaze point analysis to determine the user's attention focus area and automatically enhance the lighting brightness of the target area; user movement trajectory prediction to adjust the lighting parameters of the forward area in advance to eliminate control delays; and prediction of environmental parameter change trends (such as the natural light color temperature gradually warming with sunset) to adjust the lighting output in advance for a smooth transition. This method reduces the user's operational burden through seamless interaction, and the multi-user preference fusion algorithm solves the problem of lighting fairness in group spaces.

[0043] In addition, this method also includes: Set the optical parameter test indicators and test environment for the RGB-MiniLED lamps, and test the RGB-MiniLED lamps based on the optical parameter test indicators and test environment to obtain the actual optical parameter characteristic information of the RGB-MiniLED lamps; Calculate the ratio between the actual optical parameter characteristics of the RGB-MiniLED lamp and the optical parameter test index of the RGB-MiniLED lamp (the ratio is less than 1, and the smaller the ratio, the higher the degree of aging), define it as the aging coefficient of the RGB-MiniLED lamp, and collect the aging coefficient change characteristics of the RGB-MiniLED lamp within a preset time. An aging coefficient prediction model for RGB-MiniLED lamps is constructed based on LSTM. The aging coefficient change characteristics of RGB-MiniLED lamps within the preset time are used to train the aging coefficient prediction model of RGB-MiniLED lamps to capture the relationship between time and aging coefficient. The aging coefficient change characteristics of the RGB-MiniLED lamps within the previous preset time are obtained. The aging coefficient change characteristics of the RGB-MiniLED lamps within the previous preset time are input into the aging coefficient prediction model of the RGB-MiniLED lamps for prediction. The aging coefficient of the RGB-MiniLED lamps at each position in the current timestamp is obtained. Obtain the theoretical optical parameter characteristic information of RGB-MiniLED lamps at each location, and calculate the actual optical parameter characteristic information of RGB-MiniLED lamps at each location based on the aging coefficient of RGB-MiniLED lamps at each location in the current timestamp and the theoretical optical parameter information of RGB-MiniLED lamps. The deviation between the actual optical parameter characteristics of the RGB-MiniLED lamps at each position and the globally coordinated dimming and color adjustment strategy is calculated. When the deviation is greater than a preset deviation threshold, the working power of the RGB-MiniLED lamps at each position is adjusted.

[0044] It should be noted that RGB-MiniLED lamps will age during use, causing the theoretical illumination characteristic parameters (such as light intensity and color temperature) to differ from the actual display. When the deviation between the actual optical parameter characteristic information of the RGB-MiniLED lamps at each position and the globally coordinated dimming and color adjustment strategy exceeds a preset deviation threshold, the working power of the RGB-MiniLED lamps at each position will be adjusted to further improve the control accuracy of the RGB-MiniLED lamps.

[0045] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0046] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0047] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0048] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0049] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0050] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An RGB-MiniLED adaptive dimming and color adjustment method based on the Internet of Things, characterized in that, include: After user authorization processing, ambient lighting parameters, user micro-movement status, and ambient acoustic parameters in the target area are collected to form multimodal sensing data; Using a large language model as the core, multimodal perception data is transformed into semantic scene descriptions. The large language model is used to output user current activity icons, scene atmosphere requirements descriptions, and lighting preference conflicts among multiple users. Collect personalized preference information and scene adaptation information, complete interest negotiation in a dialogue framework based on a large language model, and generate a globally coordinated dimming and color adjustment strategy. The target color temperature and brightness are output based on the globally coordinated dimming and color tuning strategy, and controlled according to the target color temperature and brightness.

2. The IoT-based RGB-MiniLED adaptive dimming and color adjustment method according to claim 1, characterized in that, After user authorization processing, ambient lighting parameters, user micro-movement status, and ambient acoustic parameters in the target area are collected to form multimodal sensing data, specifically: Hyperspectral sensors, millimeter-wave radar human body micro-motion sensors, temperature and humidity sensors, air quality sensors, and sound feature sensors are deployed in various lighting areas to form a miniature sensor array; The aforementioned hyperspectral sensor, millimeter-wave radar human body micro-motion sensor, temperature and humidity sensor, air quality sensor, and sound feature sensor are used to collect ambient light parameters, user micro-motion status, and ambient acoustic parameters, respectively. Multimodal sensing data is constructed based on the ambient lighting parameters, user micro-movement status, and ambient acoustic parameters.

3. The IoT-based RGB-MiniLED adaptive dimming and color adjustment method according to claim 1, characterized in that, Using a large language model as the core, multimodal perception data is transformed into semantic scene descriptions, specifically: Using a 500ms sliding window, mean, variance, peak value, and slope features are extracted from the sensor data within each window. Continuous numerical values ​​are mapped to language labels, and the feature vectors corresponding to the multimodal sensing data are converted into natural language fragments. If multiple users are detected, multiple independent personnel description segments are generated and personnel are distinguished. The microphone signal is converted into event labels through a pre-trained audio event recognition model and semantic fragments are inserted. Semantic fragments are concatenated in chronological order to form a complete large language model input prompt, and system instruction prefixes are added.

4. The IoT-based RGB-MiniLED adaptive dimming and color adjustment method according to claim 1, characterized in that, The large language model is used to output user current activity icons, scene atmosphere requirements descriptions, and lighting preference conflicts among multiple users, specifically: Construct training datasets of simulated and real-world scenarios and label pairs, and use these datasets to train a large language model. The training loss function is cross-entropy plus a regularization term. Add stepwise reasoning process annotations to the training samples to guide the model to output the reasoning chain. Using the Mind Chain Distillation technique, the complex reasoning process is compressed into end-to-end lightweight model weights, the latest semantic fragments are appended to the historical prompts, and the key-value pairs of the previous reasoning are cached. The model weights are quantized using INT4, and the activation values ​​are FP16. After inference is completed, the large language model outputs a structured intent object, through which the user's current activity intent tag and scene atmosphere requirement description are obtained.

5. The IoT-based RGB-MiniLED adaptive dimming and color adjustment method according to claim 1, characterized in that, Collect personalized preference information and scenario adaptation information, specifically: Collect users' historical lighting data records, and construct user layer nodes, scene layer nodes, and lighting parameter layer nodes based on the user's historical lighting data records. Connect the user layer nodes, scene layer nodes, and lighting parameter layer nodes through directed edges to form a preference knowledge graph structure. The personalized feature vector of the current user is obtained using the aforementioned preference knowledge graph structure, and the predicted preference score is calculated based on the personalized feature vector of the current user using a Bayesian algorithm. Statistical prediction of preference scores is performed to obtain personalized color temperature and brightness preference score data under different scenarios and time periods. A preference score matrix is ​​constructed based on the personalized color temperature and brightness preferences under different scenarios and time periods.

6. The IoT-based RGB-MiniLED adaptive dimming and color adjustment method according to claim 5, characterized in that, Also includes: A lighting scene template library is constructed using activity type and atmosphere requirement tags output by a large language model. Each activity-atmosphere combination corresponds to a set of template parameters and dynamic rules. Use the lighting scene template library to match the most suitable template for the current activity type and atmosphere requirements. If multiple matches exist, calculate the similarity with personalized preference information. When the similarity is lower than the preset similarity, the best matching template is fine-tuned. When the similarity is not lower than the preset similarity, the color temperature offset, dynamic change rate and color saturation parameters in the whole lighting parameters are output.

7. The IoT-based RGB-MiniLED adaptive dimming and color adjustment method according to claim 5, characterized in that, Within a dialogue framework based on a large language model, interest negotiation is completed to generate a globally coordinated dimming and color grading strategy, specifically including: Negotiation is performed based on personalized preference information and scenario adaptation information in the large language model. The negotiation process takes place on the blackboard of the large language model, and the large language model writes the intent structure onto the blackboard. The optimal output of personalized preference information, personalized physiological rhythm regulation information, and scene adaptation information is obtained, and a bid package is generated. The bid package includes a recommended lighting parameter vector, a bid satisfaction score, and the percentage of satisfaction that is willing to be sacrificed in each subsequent negotiation round. The blackboard collects all bid packages and checks for parameter conflicts. If there are no conflicts, it directly uses a weighted average. If there are conflicts, it performs round-by-round negotiation. During the negotiation, the large language model sends the conflict reasons and the current compromise suggestions. At the same time, based on the percentage of satisfaction that is willing to be sacrificed in each subsequent round of negotiation, it is decided whether to lower the requirements, generate a new bid package, repeat conflict detection, and if conflicts still exist, initiate a voting mechanism to select the most relevant parameter as the final execution parameter based on the dominant dimension of the current scenario. After arbitration, a globally coordinated dimming and color grading strategy is generated based on the final execution parameters.

8. An RGB-MiniLED adaptive dimming and color adjustment system based on the Internet of Things, characterized in that, The device includes a memory and a processor. The memory includes a program for an IoT-based RGB-MiniLED adaptive dimming and color tuning method. When the processor executes the IoT-based RGB-MiniLED adaptive dimming and color tuning method program, it implements the steps of the IoT-based RGB-MiniLED adaptive dimming and color tuning method as described in any one of claims 1-8.

Images