Personalized lighting adjustment system based on artificial intelligence

Abstract

The present invention discloses a personalized lighting adjustment system based on artificial intelligence, which relates to the field of lighting adjustment, including: an identity recognition module, which is used to provide user identity registration, assign user IDs, and allow login after identity verification; an attention recognition module, which is used to capture the user's facial image through a camera component, determine the direction of the user's attention, and generate the user's current status data; a behavior prediction module, which is used to collect the user's historical movement path and residence time data in the control area, predict the user's future behavior through a time series prediction algorithm, and generate the user's behavior trajectory map; adjust the lighting in advance to adapt to the user's current needs, reduce manual operations, improve convenience, predict the user's movement path and residence time, and can actively adjust the lighting configuration to save energy, ensure that the user can obtain the best lighting effect where the light is most needed, and the adjustment plan can adapt to complex and changing usage scenarios.

Description

Technical Field

[0001] The present invention relates to the technical field of lighting adjustment, and in particular to a personalized lighting adjustment system based on artificial intelligence. Background Art

[0002] In modern life, personalized lighting experiences are gaining increasing importance. With the development of smart homes and the Internet of Things (IoT), users have higher expectations for environmental comfort and convenience. The quality and adaptability of lighting, particularly in various scenarios like office, study, and leisure, directly impacts the user's concentration, mood, and overall experience. Therefore, AI-based personalized lighting adjustment systems have emerged, aiming to optimize the user's lighting environment through intelligent means to adapt to different activities and situations.

[0003] The existing system's control over lighting permissions is limited to the overall regional level, failing to implement refined lighting-level permission management. This makes it difficult to support the personalized needs of multiple users. Conventional solutions rely solely on simple motion sensing and lack facial orientation recognition and behavioral trajectory prediction technology, making it impossible to predict the lighting needs of users' visual focus areas.

[0004] In addition, traditional prediction algorithms are only based on historical path analysis and do not combine time series prediction to generate dynamic behavior trajectory maps, causing lighting adjustments to lag behind actual behavioral changes, ignoring the dynamic impact of body movements and specific events on lighting parameters, and failing to establish a fusion calculation mechanism for body influence coefficients and event influence coefficients. As a result, special scenarios such as reading, meetings, or resting cannot automatically adapt to the optimal lighting mode. Summary of the Invention

[0005] In view of the above-mentioned shortcomings of the prior art, the present invention provides a personalized lighting adjustment system based on artificial intelligence, which can effectively solve the problems of the prior art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0007] The present invention discloses a personalized lighting adjustment system based on artificial intelligence, comprising:

[0008] The identity recognition module is used to provide user identity registration, assign user IDs, and allow login after identity verification;

[0009] The device management module is used to register and manage the location, start / stop status, and lighting mode of all lamps in the target area, and set the control permissions of different users on the start / stop and lighting mode of lamps;

[0010] The attention recognition module is used to capture the user's facial image through the camera component, determine the direction of the user's attention, and generate the user's current status data;

[0011] The behavior prediction module is used to collect the user's historical movement path and residence time data in the control area, predict the user's future behavior through the time series prediction algorithm, and generate the user's behavior trajectory map;

[0012] The regional analysis module is used to analyze the user's attention time and frequency in each area based on the behavior trajectory predicted by the behavior prediction module, and to prioritize the attention areas;

[0013] The control generation module is used to obtain the future stay order and stay time of each area based on the user's priority areas of attention and predicted behavior trajectory, and accordingly generate the lighting start and stop timing and duration adjustment plan in the future cycle;

[0014] The limb feedback module is used to collect the user's limb movement data in the control area during the current cycle, extract the user's behavioral habit characteristics, establish a limb model, input the behavioral habit characteristics, and output the limb influence coefficient that evaluates the behavioral habit characteristics on the corresponding lighting needs;

[0015] The event analysis module is used to record various events conducted by users in the control area, analyze the impact of different events on lighting needs based on event type and duration, and generate event impact coefficients;

[0016] The strategy generation module is used to integrate and process the limb influence coefficient and the event influence coefficient, and match the integration result with the corresponding lighting mode adjustment solution in the preset strategy library;

[0017] The adjustment implementation module is used to match the lighting start and stop timing of the control generation module with the duration adjustment scheme and configure the corresponding lighting in the specified period.

[0018] Furthermore, in the stage of determining the direction of the user's attention, the attention recognition module obtains the target area that the user is currently focusing on by identifying the eye gaze features and head deflection angles in the user's facial image, aligns the eye gaze features with the head deflection angle data in time series, performs weighted calculation according to preset weights, outputs an attention index, sets an attention state determination threshold, and when the obtained attention index exceeds the preset threshold, triggers a response signal of the target area, matches the focus point of the user's gaze with the placed lamps in the target area of the response signal, and generates the user's current state data.

[0019] Furthermore, the working logic of the time series prediction algorithm run by the behavior prediction module includes:

[0020] Continuously collect historical movement path data sets and residence time data sets of users in the control area. The historical movement path data sets contain the spatial coordinate sequences of users at different timestamps, and the residence time data sets contain the starting time and duration of users' stay at specific coordinate points.

[0021] Perform spatiotemporal correlation processing on the historical movement path data set and the residence time data set to construct a three-dimensional spatiotemporal sequence data set;

[0022] A sliding window mechanism is used to segment the three-dimensional spatiotemporal sequence data set to generate multiple training subsequences of equal length time periods.

[0023] The training subsequences are input into a pre-built recurrent neural network model for temporal pattern learning. The model extracts temporal dependencies through a gating mechanism and builds a user location transition probability matrix and a dwell time distribution function.

[0024] Based on the current user status data, the user location transfer probability matrix and the dwell time distribution function are called to perform multi-step iterative prediction to generate a predicted path sequence within a preset period in the future;

[0025] The predicted path sequence is smoothed and a continuous user behavior trajectory map is generated using the cubic spline interpolation algorithm.

[0026] Furthermore, the construction process of the recurrent neural network model includes: defining a network structure, including an input layer, a hidden layer, and an output layer, wherein the input layer receives the user's historical movement path and residence time characteristics, performs nonlinear transformation through multiple hidden layers, and introduces nonlinear factors through activation functions;

[0027] Setting a loss function to quantify the difference between the predicted output and the actual result, and using the backpropagation algorithm to optimize the parameters and update the network weights by gradient descent to minimize the loss function;

[0028] By batch-inputting historical sequence data for multiple rounds of iterations and adjusting network parameters until the model converges, a recurrent neural network model for capturing the characteristics of time series data is obtained.

[0029] Furthermore, the regional analysis module comprehensively measures the user's attention time and entry frequency in each area by weighted integration of duration and event frequency information. The calculation formula is:

[0030] ;

[0031] Where, Representative area The comprehensive attention score of represents the attention time weight factor, represents the total time period of observation, Represents the indicator function, indicating that at the time point When the user is in the area If the user is at time Located in the area Inside, then The value is 1, if the user is at time Not located in the area Inside, then The value of is 0, Represents time interval The degree of attention contribution of internal users, represents the weight factor of attention frequency, Indicates that the user enters the area during the observation period The number of times, Representative area No. The time difference between the first entry event and the current moment, Representative area No. The time difference between the first entry event and the current moment, represents the time decay coefficient, Indicates that the user enters the area during the observation period The number of times, Represents the total number of regions.

[0032] Furthermore, the operating logic of the control generation module is:

[0033] Based on the user behavior trajectory graph, the time series of each predicted stay area and the corresponding stay time period in the future cycle are extracted. Combined with the priority area sorting results, the pre-activation order and trigger time threshold of the lamps in each area are determined;

[0034] Map the dwell time period to the target duration interval of the corresponding area lamps, and dynamically adjust the lighting early start coefficient and delay attenuation coefficient according to the area priority;

[0035] Based on the mapping relationship between regional lamp identification and physical location, the timing control parameters are converted into a set of lamp start and stop instructions distributed according to spatial dimensions, forming a structured adjustment plan that includes lamp identifiers, activation timestamps, duration and shutdown conditions.

[0036] Furthermore, the operation logic of the limb model in the limb feedback module includes:

[0037] The system collects the user's joint motion trajectory, limb displacement speed, posture holding time, and movement frequency parameters in real time within the control area, and extracts the limb movement direction preference coefficient, periodic swing amplitude, and static posture duration ratio characteristics;

[0038] Based on the training of historical behavior data sets, a feature matching library is generated. The feature matching library contains the correlation mapping relationship between different behavior patterns and light sensitivity. Based on the current extracted feature input index, the behavior prediction of the associated feature and the corresponding light demand tendency value are obtained. The light demand tendency value is quantified as a limb influence coefficient, and the strong, medium and weak lighting influence thresholds are set in stages. The lighting influence thresholds correspond to the lighting mode settings of the matching lamps.

[0039] Furthermore, in the process of generating the event impact coefficient, the event analysis module collects the user's activity type and corresponding timestamp data in the control area in real time through the preset event type weight library, extracts features and classifies events for the activity types, and simultaneously extracts the duration, frequency and time period distribution parameters of each event. The weighted weights are preset, and the impact degree of a single event on the lighting demand is weightedly calculated. The impact degree value is mapped to a standardized event impact coefficient through normalization processing.

[0040] Furthermore, the strategy generation module dynamically allocates the weight ratio of the limb influence coefficient and the event influence coefficient according to the user's historical behavior data, generates a comprehensive influence value through weighted calculation, and compares the comprehensive influence value with the threshold interval stored in the preset strategy library. When the comprehensive influence value falls into a specific threshold interval, the lamp mode adjustment scheme bound to the interval is automatically triggered. The lamp mode adjustment scheme includes a color temperature adjustment gradient, a light intensity change curve and a dynamic fill light strategy, and generates an adjustment instruction set including an execution time window and an execution priority parameter and submits it to the adjustment implementation module.

[0041] Furthermore, the identity recognition module is interactively connected with the device management module and the attention recognition module through a wireless network, the behavior prediction module is interactively connected with the attention recognition module, the area analysis module and the control generation module through a wireless network, and the strategy generation module is interactively connected with the limb feedback module, the event analysis module and the adjustment implementation module through a wireless network.

[0042] Compared with the prior art, the technical solution provided by the present invention has the following beneficial effects:

[0043] 1. By capturing the user's facial image, analyzing the direction of attention, and combining historical data to predict behavioral trajectories, the lighting can be adjusted in advance to adapt to the user's current needs, reducing manual operations and improving convenience. At the same time, through historical movement paths and time series prediction algorithms, the user's movement path and stay time are predicted, and the lighting configuration can be actively adjusted instead of passively reacting. Lamps can be turned on or off in advance to save energy. Prioritization can be performed based on the attention time and frequency of different areas to ensure that users can get the best lighting effect where they need it most.

[0044] 2. By collecting body movements and event records, the system can analyze the impact of different events on lighting needs and comprehensively calculate the influence coefficients of different factors, thereby achieving more accurate lighting adjustment. It integrates the body influence and event influence coefficients to form a lamp mode adjustment plan based on comprehensive analysis, ensuring that the lighting adjustment plan can adapt to complex and changing usage scenarios.

[0045] 3. By having the ability to adapt to multi-user environments, it can support the personalized needs of multiple people at the same time, and can be well adapted whether in home, office or public space. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive effort.

[0047] Figure 1 It is a schematic diagram of the framework of the present invention.

[0048] The numbers in the figure represent: 1. Identity recognition module; 2. Equipment management module; 3. Attention recognition module; 4. Behavior prediction module; 5. Regional analysis module; 6. Control generation module; 7. Physical feedback module; 8. Event analysis module; 9. Strategy generation module; 10. Adjustment implementation module. DETAILED DESCRIPTION

[0049] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0050] The present invention will be further described below with reference to the embodiments.

[0051] Example 1

[0052] The personalized lighting adjustment system based on artificial intelligence in this embodiment, such as Figure 1 Shown, including:

[0053] Identity recognition module 1, used to provide user identity registration, assign user ID, and allow login after identity verification;

[0054] Device Management Module 2 is used to register and manage the location, start / stop status, and lighting mode of all lamps in the target area, assign control permissions to each lamp, and set the control permissions of different users for lamp start / stop and lighting mode. The user permission classification mechanism is as follows: Administrator permissions: can modify lamp parameters globally; assign user permissions and set system rules; Registered user permissions: can independently adjust preset lighting modes within the authorized area; Visitor permissions: only trigger temporary area lighting; dynamic token verification is used to ensure that different users can implement differentiated control over the start / stop status, lighting intensity, color temperature mode and other parameters of the corresponding lamps within a specified time and space range, and at the same time set the priority arbitration rules for permission conflicts: administrator instructions > event-driven strategy > user habit preferences;

[0055] The attention recognition module 3 is used to capture the user's facial image through the camera component, determine the direction of the user's attention, and generate the user's current status data; in the stage of determining the direction of the user's attention, the attention recognition module 3 obtains the target area of the user's current attention by identifying the eye gaze features and head deflection angles in the user's facial image, aligns the eye gaze features with the head deflection angle data in time series, performs weighted calculation according to preset weights, outputs the attention index, sets the attention status judgment threshold, and when the obtained attention index exceeds the preset threshold, triggers the response signal of the target area, matches the focus point of the user's gaze with the placed lamps in the target area of the response signal, and generates the user's current status data.

[0056] Behavior prediction module 4 is used to collect the user's historical movement path and residence time data in the control area, predict the user's future behavior through a time series prediction algorithm, and generate a user behavior trajectory map;

[0057] The regional analysis module 5 is used to analyze the user's attention time and frequency in each area based on the behavior trajectory predicted by the behavior prediction module 4, and to prioritize the attention areas;

[0058] The control generation module 6 is used to obtain the future stay order and stay time of each area according to the user's priority focus area and predicted behavior trajectory, and correspondingly generate the lighting start and stop timing and duration adjustment plan in the future cycle.

[0059] The operation logic of the control generation module 6 is:

[0060] Based on the user behavior trajectory graph, the time series of each predicted stay area and the corresponding stay time period in the future cycle are extracted. Combined with the priority area sorting results, the pre-activation order and trigger time threshold of the lamps in each area are determined;

[0061] Map the dwell time period to the target duration interval of the corresponding area lamps, and dynamically adjust the lighting early start coefficient and delay attenuation coefficient according to the area priority;

[0062] Based on the mapping relationship between regional lamp identification and physical location, the timing control parameters are converted into a set of lamp start and stop instructions distributed according to the spatial dimension, forming a structured adjustment scheme that includes lamp identifiers, activation timestamps, duration and shutdown conditions. The lighting duration of high-priority areas is increased by a preset buffer time based on the predicted value and is prioritized at the front end of the control sequence.

[0063] The limb feedback module 7 is used to collect the user's limb movement data in the control area during the current cycle, extract the user's behavioral habit characteristics, establish a limb model, input the behavioral habit characteristics, and output the limb influence coefficient that evaluates the behavioral habit characteristics on the corresponding lighting requirements. The operation logic of the limb model in the limb feedback module 7 includes:

[0064] The system collects the user's joint motion trajectory, limb displacement speed, posture holding time, and movement frequency parameters in real time within the control area, and extracts the limb movement direction preference coefficient, periodic swing amplitude, and static posture duration ratio characteristics;

[0065] Based on the training of historical behavior data sets, a feature matching library is generated. The feature matching library contains the correlation mapping relationship between different behavior patterns and light sensitivity. Based on the current extracted feature input index, the behavior prediction of the associated feature and the corresponding light demand tendency value are obtained. The light demand tendency value is quantified as a limb influence coefficient, and the strong, medium and weak lighting influence thresholds are set in stages. The lighting influence thresholds correspond to the lighting mode settings of the matching lamps.

[0066] The event analysis module 8 is used to record various events conducted by users in the control area, analyze the impact of different events on lighting needs based on event types and durations, and generate event impact coefficients; in the process of generating event impact coefficients, the event analysis module 8 collects the activity types and corresponding timestamp data of users in the control area in real time through a preset event type weight library, performs feature extraction and event classification on the activity types, and simultaneously extracts the duration, occurrence frequency and time period distribution parameters of each event, presets weighted weights, and weightedly calculates the impact degree value of a single event on lighting needs, wherein the event type weight library contains basic weight factors and time-sensitive coefficients corresponding to different activity types, maps the impact degree value to a standardized event impact coefficient through normalization processing, and establishes a time series association model with historical event impact data to dynamically correct the coefficient parameters.

[0067] The strategy generation module 9 is used to integrate and process the body influence coefficient and the event influence coefficient, and match the integration result with the corresponding lighting mode adjustment scheme in the preset strategy library; the strategy generation module 9 dynamically allocates the weight ratio of the body influence coefficient and the event influence coefficient according to the user's historical behavior data, generates a comprehensive influence value through weighted calculation, and compares the comprehensive influence value with the threshold interval stored in the preset strategy library. When the comprehensive influence value falls into a specific threshold interval, it automatically triggers the lighting mode adjustment scheme bound to the interval. The lighting mode adjustment scheme includes a color temperature adjustment gradient, a light intensity change curve and a dynamic fill light strategy, and generates an adjustment instruction set including an execution time window and an execution priority parameter and submits it to the adjustment implementation module 10.

[0068] The adjustment implementation module 10 is used to match the lamp start and stop timing of the control generation module 6 with the duration adjustment scheme, and configure the corresponding lamps in the specified period.

[0069] The identity recognition module 1 is interactively connected with the device management module 2 and the attention recognition module 3 through a wireless network, the behavior prediction module 4 is interactively connected with the attention recognition module 3, the area analysis module 5 and the control generation module 6 through a wireless network, and the strategy generation module 9 is interactively connected with the limb feedback module 7, the event analysis module 8 and the adjustment implementation module 10 through a wireless network.

[0070] Compared with existing technologies, the system realizes efficient user identity management, behavior prediction and lighting adjustment. It provides hierarchical user authority management to ensure that different users can dynamically adjust the status and mode of lamps according to their authority. The camera captures the user's facial image and body movements, accurately analyzes the user's attention direction and behavioral habits, and thus generates user status data and behavior trajectory maps. Taking into account the user's current status and historical behavior data, the system intelligently predicts the user's future stay area and time, and formulates lamp start-stop and mode adjustment plans accordingly.

[0071] By recording user events within the control area and combining their impact on lighting needs, customized lighting settings are formed. The advantage is that it can provide a more refined and personalized lighting experience, intelligently respond to users' real-time needs, and improve the flexibility and comfort of lighting control. Compared with the traditional fixed lighting control mode, it greatly enhances user interactivity and satisfaction.

[0072] Example 2

[0073] In other aspects, this embodiment also provides another optimization mechanism based on embodiment 1, specifically a working logic of a time series prediction algorithm, including:

[0074] Continuously collect historical movement path data sets and residence time data sets of users in the control area. The historical movement path data sets contain the spatial coordinate sequences of users at different timestamps, and the residence time data sets contain the starting time and duration of users' stay at specific coordinate points.

[0075] Perform spatiotemporal correlation processing on the historical movement path data set and the residence time data set to construct a three-dimensional spatiotemporal sequence data set, and establish a mapping relationship between the user's current location coordinates and the time dimension;

[0076] A sliding window mechanism is used to segment the three-dimensional spatiotemporal sequence data set into multiple training subsequences of equal length. Each training subsequence contains the user's movement path characteristics and stay pattern characteristics within a preset time length.

[0077] The training subsequences are fed into a pre-built recurrent neural network model for temporal pattern learning. The model extracts temporal dependencies through a gating mechanism and builds a user location transition probability matrix and a dwell time distribution function.

[0078] Based on the user status data at the current moment, the user location transfer probability matrix and the dwell time distribution function are called to perform multi-step iterative prediction to generate a predicted path sequence within a preset period in the future. The predicted path sequence contains the expected location coordinate set corresponding to the predicted time node set and the associated confidence parameters;

[0079] The predicted path sequence is smoothed, and a continuous user behavior trajectory map is generated using the cubic spline interpolation algorithm. The user behavior trajectory map includes the temporal and spatial thermal distribution characteristics and the visualization expression of the path confidence interval.

[0080] As a preferred implementation of this embodiment, the process of constructing a recurrent neural network model includes: defining a network structure, including an input layer, a hidden layer, and an output layer, wherein the input layer receives the user's historical movement path and residence time characteristics, performs nonlinear transformation through multiple hidden layers, and introduces nonlinear factors through an activation function;

[0081] A loss function is set to quantify the difference between the predicted output and the actual result, and the backpropagation algorithm is used for parameter optimization. The network weights are updated by the gradient descent method to minimize the loss function.

[0082] By batch-inputting historical sequence data for multiple rounds of iterations and adjusting network parameters until the model converges, a recurrent neural network model for capturing the characteristics of time series data is obtained.

[0083] Compared with existing technologies, the time series learning and multi-step prediction of the recurrent neural network model can accurately predict the user's future location and residence time, improving the accuracy of the prediction. The user's historical path and residence time are temporally and spatially correlated, combined with user behavior characteristics, effectively capturing the temporal and spatial dependencies of user behavior. The predicted path is smoothed through cubic spline interpolation to ensure the continuity and naturalness of the trajectory, improve the visualization of the user behavior trajectory map, and avoid incoherent or unrealistic predictions.

[0084] The generated prediction path sequence not only has high-confidence parameters, but can also be visualized through heat maps and path confidence intervals, making the prediction results more intuitive and easy to understand. The model can be dynamically updated and optimized. As more historical data accumulates, the model's prediction ability continues to increase, enabling more accurate predictions of future user behavior and providing personalized lighting adjustment solutions.

[0085] Example 3

[0086] This embodiment provides a weighted integration of duration and event frequency information, and the calculation formula is:

[0087] ;

[0088] Where, Representative area The comprehensive attention score of represents the attention time weight factor, represents the total time period of observation, Represents the indicator function, indicating that at the time point When the user is in the area If the user is at time Located in the area Inside, then The value is 1, if the user is at time Not located in the area Inside, then The value of is 0, Represents time interval The degree of attention contribution of internal users, represents the weight factor of attention frequency, Indicates that the user enters the area during the observation period The number of times, Representative area No. The time difference between the first entry event and the current moment, Representative area No. The time difference between the first entry event and the current moment, represents the time decay coefficient, Indicates that the user enters the area during the observation period The number of times, Represents the total number of regions.

[0089] The formula in this embodiment can reflect not only the length of time a user stays in an area, but also the user's frequent attention to a certain area, thereby providing a basis for intelligent adjustment of lamp start-stop and lighting mode.

[0090] Working principle: After the system of the present invention is installed, the identity recognition module 1 allows the user to register and verify the identity, thereby obtaining the user ID, the device management module 2 performs system registration and manages the location of the lamps and their control permissions, the attention recognition module 3 captures the user's facial image through the camera component, determines the direction of attention, and generates user status data, the behavior prediction module 4 records the user's historical movement path and stay time, uses the time series prediction algorithm to predict their future behavior, and generates a behavior trajectory map, the regional analysis module 5 analyzes the time and frequency of the user's attention to each area, and performs priority sorting, and the control generation module 6 formulates the start and stop timing and duration adjustment plan of the lamps.

[0091] In addition, the limb feedback module 7 collects the user's limb movement data and evaluates its impact on lighting needs. The event analysis module 8 records the user's events and analyzes the degree of their impact on lighting needs. The strategy generation module 9 integrates the limb influence coefficient and the event influence coefficient. The adjustment implementation module 10 configures the final lighting adjustment plan to the corresponding lamps. Through this process, the system can intelligently adjust the lighting according to the user's real-time needs to achieve a personalized lighting experience.

[0092] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. Personalized lighting adjustment system based on artificial intelligence, characterized by: include: The identity recognition module is used to provide user identity registration, assign user IDs, and allow login after identity verification; The device management module is used to register and manage the location, start / stop status, and lighting mode of all lamps in the target area, and set the control permissions of different users on the start / stop and lighting mode of lamps; The attention recognition module is used to capture the user's facial image through the camera component, determine the direction of the user's attention, and generate the user's current status data; The behavior prediction module is used to collect the user's historical movement path and residence time data in the control area, predict the user's future behavior through the time series prediction algorithm, and generate the user's behavior trajectory map; The regional analysis module is used to analyze the user's attention time and frequency in each area based on the behavior trajectory predicted by the behavior prediction module, and to prioritize the attention areas; The control generation module is used to obtain the future stay order and stay time of each area based on the user's priority areas of attention and predicted behavior trajectory, and accordingly generate the lighting start and stop timing and duration adjustment plan in the future cycle; The limb feedback module is used to collect the user's limb movement data in the control area during the current cycle, extract the user's behavioral habit characteristics, establish a limb model, input the behavioral habit characteristics, and output the limb influence coefficient; The event analysis module is used to record the types and duration of events performed by users in the control area, analyze the impact of different events on lighting needs, and generate event impact coefficients; The strategy generation module is used to integrate and process the limb influence coefficient and the event influence coefficient, and match the integration result with the corresponding lighting mode adjustment solution in the preset strategy library; The adjustment implementation module is used to match the start and stop timing of the lamps with the duration adjustment scheme and configure the corresponding lamps in the specified period.

2. The artificial intelligence-based personalized lighting adjustment system according to claim 1, characterized in that: In the stage where the attention recognition module determines the direction of the user's attention, the target area that the user is currently paying attention to is obtained by identifying the eye gaze features and head deflection angles in the user's facial image, the eye gaze features and head deflection angle data are time-series aligned, weighted calculation is performed according to preset weights, an attention index is output, and an attention state determination threshold is set. When the obtained attention index exceeds the preset threshold, a response signal of the target area is triggered, and the focus point of the user's gaze is matched with the lamps placed in the target area of the response signal to generate the user's current state data.

3. The artificial intelligence-based personalized lighting adjustment system according to claim 1, characterized in that: The working logic of the time series prediction algorithm run by the behavior prediction module includes: Continuously collect historical movement path data sets and residence time data sets of users in the control area. The historical movement path data sets contain the spatial coordinate sequences of users at different timestamps, and the residence time data sets contain the starting time and duration of users' stay at specific coordinate points. Perform spatiotemporal correlation processing on the historical movement path data set and the residence time data set to construct a three-dimensional spatiotemporal sequence data set; A sliding window mechanism is used to segment the three-dimensional spatiotemporal sequence data set to generate multiple training subsequences of equal length time periods. The training subsequences are input into a pre-built recurrent neural network model for temporal pattern learning. The model extracts temporal dependencies through a gating mechanism and builds a user location transition probability matrix and a dwell time distribution function. Based on the current user status data, the user location transfer probability matrix and the dwell time distribution function are called to perform multi-step iterative prediction to generate a predicted path sequence within a preset period in the future; The predicted path sequence is smoothed and a continuous user behavior trajectory map is generated using the cubic spline interpolation algorithm.

4. The artificial intelligence-based personalized lighting adjustment system according to claim 3, characterized in that: The construction process of the recurrent neural network model includes: defining a network structure, including an input layer, a hidden layer, and an output layer, wherein the input layer receives the user's historical movement path and residence time characteristics, performs nonlinear transformation through multiple hidden layers, and introduces nonlinear factors through activation functions; Setting a loss function to quantify the difference between the predicted output and the actual result, and using the backpropagation algorithm to optimize the parameters and update the network weights by gradient descent to minimize the loss function; By batch-inputting historical sequence data for multiple rounds of iterations and adjusting network parameters until the model converges, a recurrent neural network model for capturing the characteristics of time series data is obtained.

5. The artificial intelligence-based personalized lighting adjustment system according to claim 1, characterized in that: The regional analysis module comprehensively measures the user's attention time and entry frequency in each area by weighted integration of duration and event frequency information. The calculation formula is: ； Where, Representative area The comprehensive attention score of represents the attention time weight factor, represents the total time period of observation, Represents the indicator function, indicating that at the time point When the user is in the area If the user is at time Located in the area Inside, then The value is 1, if the user is at time Not located in the area Inside, then The value of is 0, Represents time interval The degree of attention contribution of internal users, represents the weight factor of attention frequency, Indicates that the user enters the area during the observation period The number of times, Representative area No. The time difference between the first entry event and the current moment, Representative area No. The time difference between the first entry event and the current moment, represents the time decay coefficient, Indicates that the user enters the area during the observation period The number of times, Represents the total number of regions.

6. The artificial intelligence-based personalized lighting adjustment system according to claim 1, characterized in that: The operating logic of the control generation module is: Based on the user behavior trajectory graph, the time series of each predicted stay area and the corresponding stay time period in the future cycle are extracted. Combined with the priority area sorting results, the pre-activation order and trigger time threshold of the lamps in each area are determined; Map the dwell time period to the target duration interval of the corresponding area lamps, and dynamically adjust the lighting early start coefficient and delay attenuation coefficient according to the area priority; Based on the mapping relationship between regional lamp identification and physical location, the timing control parameters are converted into a set of lamp start and stop instructions distributed according to spatial dimensions, forming a structured adjustment plan that includes lamp identifiers, activation timestamps, duration and shutdown conditions.

7. The artificial intelligence-based personalized lighting adjustment system according to claim 1, characterized in that: The operating logic of the limb model in the limb feedback module includes: The system collects the user's joint motion trajectory, limb displacement speed, posture holding time, and movement frequency parameters in real time within the control area, and extracts the limb movement direction preference coefficient, periodic swing amplitude, and static posture duration ratio characteristics; Based on the training of historical behavior data sets, a feature matching library is generated. The feature matching library contains the correlation mapping relationship between different behavior patterns and light sensitivity. Based on the current extracted feature input index, the behavior prediction of the associated feature and the corresponding light demand tendency value are obtained. The light demand tendency value is quantified as a limb influence coefficient, and the strong, medium and weak lighting influence thresholds are set in stages. The lighting influence thresholds correspond to the lighting mode settings of the matching lamps.

8. The artificial intelligence-based personalized lighting adjustment system according to claim 1, characterized in that: In the process of generating the event impact coefficient, the event analysis module collects the user's activity type and corresponding timestamp data in the control area in real time through the preset event type weight library, extracts features and classifies events for the activity type, and simultaneously extracts the duration, frequency and time period distribution parameters of each event. The weighted weight is preset, and the impact degree of a single event on the lighting demand is weightedly calculated. The impact degree value is mapped to a standardized event impact coefficient through normalization processing.

9. The artificial intelligence-based personalized lighting adjustment system according to claim 1, characterized in that: The strategy generation module dynamically allocates the weight ratio of the limb influence coefficient and the event influence coefficient according to the user's historical behavior data, generates a comprehensive influence value through weighted calculation, and compares the comprehensive influence value with the threshold interval stored in the preset strategy library. When the comprehensive influence value falls into a specific threshold interval, it automatically triggers the lamp mode adjustment scheme bound to the interval. The lamp mode adjustment scheme includes a color temperature adjustment gradient, a light intensity change curve and a dynamic fill light strategy, and generates an adjustment instruction set including an execution time window and execution priority parameters and submits it to the adjustment implementation module.

10. The artificial intelligence-based personalized lighting adjustment system according to claim 1, characterized in that: The identity recognition module is interactively connected with the device management module and the attention recognition module through a wireless network; the behavior prediction module is interactively connected with the attention recognition module, the area analysis module and the control generation module through a wireless network; and the strategy generation module is interactively connected with the limb feedback module, the event analysis module and the adjustment implementation module through a wireless network.

Images