Full life cycle healthy lighting system, method, apparatus, and storage medium

Abstract

The application provides a full life cycle healthy lighting system, method, device and storage medium. The system comprises: a multi-modal sensing module for collecting multi-modal data; a life stage recognition module connected with the multi-modal sensing module, for determining a probability distribution of a specific life stage of a user corresponding to the multi-modal data; a stage-constrained digital twin module connected with the life stage recognition module, for instantiating a stage-constrained digital twin for each user in a current space based on the probability distribution; a multi-objective health intervention module connected with the stage-constrained digital twin module, for determining a target lighting intervention strategy; and a swarm intelligence adaptive module connected with the multi-objective health intervention module, for realizing adaptive healthy lighting in the current space based on the target lighting intervention strategy. The application proposes an automatic recognition mechanism for the life stage and specific physiological state of a user based on multi-modal data, and can obtain an optimal lighting strategy based on the health priority of each user.

Description

Technical Field

[0001] This application belongs to the field of intelligent lighting technology and relates to a whole life cycle healthy lighting system, method, device and storage medium. Background Technology

[0002] With the deep integration of smart home and healthy lighting technologies, modern lighting systems have evolved from traditional single lighting functions towards intelligence, personalization, and health. However, existing smart lighting systems still have significant technical limitations: most systems rely solely on human motion sensing, timing strategies, or changes in ambient light for basic adjustments, and even the few systems that can sense user activity have relatively simple control logic, making it difficult to meet the personalized health needs of different users at complex physiological stages.

[0003] Therefore, in multi-user coexistence scenarios, how to negotiate the optimal lighting strategy based on the health priority of each user has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0004] This application provides a whole-lifecycle health lighting system, method, device, and storage medium for negotiating the optimal lighting strategy based on the health priorities of each user, thereby constructing a truly personalized health lighting system.

[0005] In a first aspect, this application provides a full lifecycle health lighting system, the system comprising: a multimodal perception module for collecting multimodal data of all users in the current space; a life stage identification module connected to the multimodal perception module for determining the probability distribution of the user corresponding to each set of multimodal data being in a specific life stage based on each set of multimodal data; a stage constraint digital twin module connected to the life stage identification module for instantiating a stage constraint digital twin for each user in the current space based on the probability distribution; a multi-objective health intervention module connected to the stage constraint digital twin module for determining a target lighting intervention strategy based on the digital twins of all users in the current space using a multi-objective optimization algorithm; and a swarm intelligence adaptive module connected to the multi-objective health intervention module for aggregating model update parameters corresponding to multiple edge devices in the current space via a federated learning mechanism based on the target lighting intervention strategy to achieve adaptive health lighting in the current space.

[0006] This application proposes for the first time an automatic identification mechanism for user life stages and specific physiological states based on multimodal data, enabling precise anchoring of physiological characteristics from all age groups (infants to the elderly) to special groups (pregnant women, disabled individuals, etc.), and constructing a truly deep and personalized health lighting system; for multi-user coexistence spaces, it dynamically quantifies real-time health risks to establish priority weights, and seeks the optimal solution for a compromise lighting strategy that maximizes global health benefits; relying on a federated learning architecture to aggregate edge model updates, it achieves continuous iteration of the underlying algorithm and distributed self-evolution of the global model, ensuring the long-term vitality of the system.

[0007] In one implementation of the first aspect, the multimodal perception module includes a first camera, a second camera, and a sensor; the first camera is used to acquire user facial image sequences and physiological feature sequences; the second camera, in conjunction with the sensor, extracts user gait feature sequences and behavioral feature sequences through a skeletal point recognition algorithm.

[0008] In one implementation of the first aspect, the life stage identification module incorporates a multimodal fusion neural network, which includes: a feature extraction layer, a cross-modal attention fusion layer, and a classification output layer; the feature extraction layer is used to extract facial feature vectors, gait feature vectors, physiological feature vectors, and behavioral feature vectors based on the multimodal data; the cross-modal attention fusion layer is used to interactively fuse the facial feature vectors, gait feature vectors, physiological feature vectors, and behavioral feature vectors to generate a fused feature vector; and the classification output layer is used to output the probability distribution of the user being in a specific life stage based on the fused feature vector.

[0009] In one implementation of the first aspect, the feature extraction layer includes a facial image sub-network, a gait video sub-network, a physiological time-series data sub-network, and a behavioral heatmap factor network. The facial image sub-network is used to extract the facial feature vector from the facial image sequence, and the facial feature vector is used to encode phenotypes including but not limited to skin elasticity, wrinkle density, and facial muscle activity. The gait video sub-network is used to extract the gait feature vector from the gait video sequence, and the gait feature vector is used to encode dynamic features including but not limited to stride length, cadence, left-right symmetry, and center of gravity shift. The physiological time-series data sub-network is used to extract the physiological feature vector from the physiological indicator time-series data, and the physiological feature vector is used to encode rhythmic features including but not limited to heart rate variability and respiratory rate. The behavioral heatmap factor network is used to extract the behavioral feature vector from the behavioral feature data corresponding to the behavioral heatmap, and the behavioral feature vector is used to encode, including but not limited to, indoor activity intensity, location preference, and sedentary / sleep duration. The multimodal data includes the facial image sequence, the gait video sequence, the behavioral feature data, and the physiological indicator time-series data.

[0010] In one implementation of the first aspect, the digital twin has a built-in physiological baseline model and health risk map corresponding to the specific life stage; the physiological baseline model is a statistical model or a multivariate Gaussian model, which defines the healthy range, circadian rhythm template, and normal fluctuation pattern of at least one physiological indicator under the specific life stage, and the physiological indicator includes heart rate, respiratory rate, and heart rate variability; the health risk map is an association matrix or a decision tree, which defines the mapping relationship between health risk types and lighting intervention directions; the digital twin compares the real-time sensed physiological values ​​with the standard values ​​in the physiological baseline model and calculates the deviation; if the deviation exceeds a threshold, it triggers a query of the health risk map to obtain the health risk type and recommended lighting intervention direction.

[0011] In one implementation of the first aspect, the stage-constrained digital twin module further includes a stage migration unit. When a user is detected migrating from a first life stage to a second life stage, parameter smooth migration is performed. If the first life stage is childhood and the second life stage is adolescence, when a user migrates from the first life stage to the second life stage, parameter smooth migration is performed through a migration trigger judgment rule. The user's childhood digital twin is loaded. The underlying feature extraction parameters of the childhood digital twin are frozen, wherein the underlying feature extraction parameters include basic vital sign parameters and basic voice feature parameters. The top-level feature extraction parameters of the childhood digital twin are extracted, wherein the top-level feature extraction parameters include sleep time parameters and calorie consumption parameters. The user's adolescence adaptation data is collected. The top-level feature extraction parameters are fine-tuned and optimized using the adolescence adaptation data, and combined with the underlying feature extraction parameters, an adolescence digital twin is constructed and generated.

[0012] Secondly, this application provides a control method for a full life-cycle health lighting system, applied to the system described in any of the first aspects. The method includes: collecting multimodal data of all users in the current space; determining the probability distribution of the life stage of the user corresponding to each set of multimodal data based on the multimodal data; constructing or calling a corresponding digital twin for each user based on the probability distribution; predicting the user's current health status and quantifying the risk prediction value based on the digital twin and the user's real-time body data; collecting the digital twins and the risk prediction values ​​of all users in the current space to construct a multi-objective optimization function; using a multi-objective optimization algorithm to solve the multi-objective lighting optimization problem in the current space to obtain a target lighting intervention strategy; and issuing control commands corresponding to the target lighting intervention strategy to each LED lamp in the current space to adjust the lighting information of each LED lamp and achieve precise light output in the current space.

[0013] In one implementation of the second aspect, the method further includes: monitoring the user's multimodal feedback data after intervention; and updating the target lighting intervention strategy based on the multimodal feedback data.

[0014] Thirdly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the full life-cycle healthy lighting system as described in any one of the first aspects of embodiments of this application.

[0015] Fourthly, embodiments of this application provide an electronic device, the electronic device comprising: a memory storing a computer program; and a processor communicatively connected to the memory, which executes the full life-cycle healthy lighting system as described in any one of the first aspects of this application when the computer program is invoked.

[0016] As described above, the whole life cycle healthy lighting system, method, device, and storage medium described in this application have the following beneficial effects:

[0017] 1) This application proposes for the first time an automatic identification mechanism for user life stages and specific physiological states based on multimodal data, which can accurately anchor the physiological characteristics of all age groups (infants to the elderly) and special groups (pregnant women, disabled people, etc.) to build a truly deep personalized health lighting system; facing the coexistence space of multiple users, it dynamically quantifies real-time health risks to establish priority weights and seeks the best solution for the compromise lighting strategy that maximizes global health benefits; relying on the federated learning architecture to aggregate edge model updates, it realizes the continuous iteration of the underlying algorithm and the distributed self-evolution of the global model, ensuring the long-term vitality of the system.

[0018] 2) The life stage recognition module in this application incorporates a multimodal fusion neural network, which includes a feature extraction layer, a cross-modal attention fusion layer, and a classification output layer. The feature extraction layer processes four-dimensional heterogeneous data of face, gait, physiology, and behavior in parallel, breaking the limitations of single-modal perception. By utilizing the complementary effect of information redundancy, it significantly improves the robustness of the system in complex environments such as sudden changes in illumination and partial occlusion. The cross-modal attention fusion layer abandons simple feature splicing and dynamically mines deep semantic relationships between modalities through joint inference and adaptively allocates weights to suppress modal conflicts and noise interference, achieving high-precision anchoring of life stages and special states. The classification output layer uses probability distribution instead of hard labels to quantify the ambiguity and transitional state of life stage evolution, providing a continuous decision basis for the smooth transition of subsequent lighting strategies and avoiding the discomfort caused by illumination intervention due to sudden changes in state.

[0019] 3) In this embodiment, transfer learning fine-tuning is used instead of model reconstruction, so that the physiological baseline model retains the core physiological characteristics of the user's historical stage during the transition, avoids personalized data gaps, and ensures the continuity of the system's cognition of the user; by smoothly transferring the distribution parameters and the mapping parameters in the health risk map, the oscillation of calculation results caused by model updates is avoided, and the sudden changes in lighting strategy and visual discomfort caused by state switching are eliminated from the root of the underlying algorithm. Attached Figure Description

[0020] Figure 1 The diagram shown is a structural diagram of a full life-cycle healthy lighting system provided in an embodiment of this application.

[0021] Figure 2 The diagram shown is a structural diagram of the life stage identification module provided in an embodiment of this application.

[0022] Figure 3 The diagram shown illustrates the workflow of the stage migration unit provided in this application embodiment.

[0023] Figure 4 The diagram shown illustrates the negotiation process of the multi-objective health intervention module provided in this application embodiment.

[0024] Figure 5 The diagram shown is a federated learning architecture diagram of the swarm intelligence evolution module provided in an embodiment of this application.

[0025] Figure 6 The flowchart shown is a control method for a full life-cycle healthy lighting system provided in an embodiment of this application.

[0026] Figure 7 The diagram shown is a structural diagram of an electronic device provided in an embodiment of this application.

[0027] Component designation explanation

[0028] 100 Full life cycle healthy lighting system 130 Phase-constrained digital twin module 110 Multimodal sensing module 131 Phase migration unit 120 Life stage identification module 140 Multi-objective health intervention module 1200 Multimodal fusion neural network 150 Swarm intelligence adaptive module 1210 Feature extraction layer S61~S67 step 1211 Facial Image Subnetwork 70 electronic devices 1212 Gait video subnetwork 71 processor 1213 Physiological time series data subnetwork 72 Non-volatile storage media 1214 Behavioral thermal factor network 73 System bus 1220 Cross-modal attention fusion layer 74 Internal memory 1230 Classification output layer 75 Network interface Detailed Implementation

[0029] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0030] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0031] The technical solutions in the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0032] like Figure 1 As shown in the figure, this application provides a structural diagram of a full life-cycle healthy lighting system. Figure 1 As shown, the full life cycle health lighting system 100 provided in this application embodiment includes a multimodal perception module 110, a life stage identification module 120, a stage-constrained digital twin module 130, a multi-objective health intervention module 140, and a swarm intelligence adaptive module 150.

[0033] The multimodal perception module is used to collect multimodal data from all users in the current space.

[0034] Specifically, the multimodal perception module includes a first camera, a second camera, and a sensor; the first camera is used to acquire user facial image sequences and physiological feature sequences; the second camera, in conjunction with the sensor, extracts user gait feature sequences and behavioral feature sequences through a skeletal point recognition algorithm.

[0035] For example, the first camera may be an RGB camera, which is used to acquire a sequence of facial images of the user and extract physiological feature sequences such as the user's heart rate, heart rate variability and respiratory rate through remote photoplethysmography (rPPG) technology.

[0036] For example, the second camera may be a depth camera, and the gait feature sequence includes stride length, cadence, left-right symmetry, etc., while the behavioral feature sequence includes sitting, standing, lying down, walking, etc.

[0037] The life stage identification module, connected to the multimodal perception module, is used to determine the probability distribution of the user corresponding to each set of multimodal data being in a specific life stage based on each set of multimodal data.

[0038] For example, when the system detects that a user's facial skin elasticity has decreased (wrinkle density feature value in facial feature vector > 0.85), walking speed has slowed (step length feature value in gait feature vector < 0.3 times the standard value), heart rate variability has decreased (SDNN in physiological feature vector < 50ms), and daily activity level has decreased (sedentary time in behavioral feature vector > 8 hours / day), the fused feature vector score Z_7 for the elderly category is significantly higher than that for other categories. After Softmax, the score for the elderly category can reach above 0.85, and the system comprehensively judges that the user is in the "elderly stage".

[0039] The stage constraint digital twin module, connected to the life stage identification module, is used to instantiate a stage constraint digital twin for each user in the current space based on the probability distribution.

[0040] In some embodiments, the digital twin has a built-in physiological baseline model and health risk map corresponding to the specific life stage; the physiological baseline model is a statistical model or a multivariate Gaussian model, which defines the healthy range, circadian rhythm template and normal fluctuation pattern of at least one physiological indicator under the specific life stage, and the physiological indicator includes heart rate, respiratory rate and heart rate variability; the health risk map is an association matrix or a decision tree, which defines the mapping relationship between health risk types and lighting intervention directions.

[0041] The digital twin compares the real-time sensed physiological values ​​with the standard values ​​in the physiological baseline model and calculates the deviation. If the deviation exceeds a threshold, it triggers a query of the health risk map to obtain the health risk type and recommended lighting intervention direction.

[0042] For example, a physiological baseline model is a statistical or multivariate Gaussian model that defines the healthy range, daily rhythm template, and normal fluctuation pattern of various physiological indicators (heart rate, respiratory rate, heart rate variability, etc.) at this life stage; the greater the deviation of the user's real-time data from this baseline, the higher the health risk.

[0043] For example, a health risk map is an association matrix or decision tree that maps specific physiological / behavioral deviation patterns to specific health risks and specifies corresponding lighting intervention directions. For example, map entries include: [Detected: Frequent nighttime activity + decreased heart rate variability] → [Risk: Sleep cycle disorder] → [Lighting intervention: Reduce color temperature to 2200K 1 hour before bedtime, reduce illuminance to <10 lx]; [Detected: Increased gait swaying amplitude + sitting to standing time >2 seconds] → [Risk: Increased risk of falls] → [Lighting intervention: Increase path illuminance to >150 lx, uniformity >0.8].

[0044] For example, the specific calculation method for deviation is as follows:

[0045] (1) Deviation of a single indicator: The standard score (Z-score) formula is used: D k =∣x k −μ k ∣ / σ k , where x k μ is the real-time measurement value of the user's current k-th metric. k Let σ be the normal mean of the user's current k-th metric. k The standard deviation is D. k This represents the standard deviation of the user's current k-th metric, and the judgment criterion is: D k ≤1 is normal, 1 <D k ≤2 indicates a slight deviation, 2 <D k ≤3 indicates moderate deviation, D k >3 indicates a serious deviation.

[0046] (2) Multi-indicator comprehensive deviation: When multiple physiological indicators need to be evaluated simultaneously, a weighted comprehensive deviation is used:

[0047]

[0048] Where m is the number of physiological indicators involved in the evaluation. Let be the weight coefficient of the k-th indicator, satisfying =1, The single-indicator deviation of the k-th indicator. This represents the overall deviation of multiple indicators.

[0049] The weighting of each indicator varies across different life stages. For example, gait stability has the highest weight in old age (w=0.5), while focus duration has the highest weight in adolescence (w=0.4).

[0050] Deviation triggering mechanism: when When the threshold is exceeded (default value is 1.5), the system determines that the user's current physiological state is significantly abnormal, triggers the query of the health risk map, and matches the corresponding health risk type and recommended lighting intervention direction according to the deviation pattern.

[0051] It should be noted that the data listed in the above examples are merely illustrative and this application does not impose any limitations on them.

[0052] A multi-objective health intervention module, connected to the stage-constrained digital twin module, is used to determine the target lighting intervention strategy based on the digital twins of all users in the current space using a multi-objective optimization algorithm.

[0053] The swarm intelligence adaptive module, connected to the multi-target health intervention module, is used to aggregate model update parameters corresponding to multiple edge devices in the current space based on the target lighting intervention strategy via a federated learning mechanism, so as to realize adaptive health lighting in the current space.

[0054] For example, when both an elderly person and an infant are present in the living room, the system uses a multi-objective optimization algorithm to solve for a lighting intervention strategy. The specific steps are as follows:

[0055] 1) Obtain the health needs and risk weights of each user.

[0056] For each user i within the space, the system extracts the following information from that user's digital twin:

[0057] Current health risk value R i The value is calculated by the health risk map based on the deviation between the user's real-time physiological data and the physiological baseline model. The value ranges from 0 to 100, and the higher the value, the greater the health risk.

[0058] Lighting demand preference vector ,in, This represents the optimal illuminance for user i under ideal conditions. This represents the optimal color temperature for user i under ideal conditions. This represents the optimal color rendering index for user i under ideal conditions.

[0059] The risk-light sensitivity function fi(x) represents the increment of the health risk value when the lighting parameter x deviates from the user's preference.

[0060] 2) Construct a multi-objective optimization function.

[0061] Using the lighting control parameters x=[I, CCT, SPD, Δt] as decision variables (representing illuminance, color temperature, spectral weight, and gradation time, respectively), the following multi-objective optimization problem is constructed:

[0062]

[0063] Where N is the total number of users in the current space, and R i (x) represents the expected health risk value of user i under the current lighting parameters x, calculated using the following formula: That is, the baseline risk plus the risk increased due to the deviation of illumination from the optimal value. Baseline health risk value, Let w be the light sensitivity coefficient for user i. i The dynamic priority weight for user i is wi = softmax( This means that users with higher current health risks receive higher weights, ensuring that high-risk users are given priority.

[0064] C(x) is the comprehensive visual comfort function, defined as follows: Wherein, UGR is the Uniform Glare Value (lower is better), and Ra is the Color Rendering Index (higher is better). These represent the comfort weights, by default. =0.6, =0.4, To standardize the glare value.

[0065] 3) Solve for the Pareto optimal solution set:

[0066] The above multi-objective optimization problem is solved using a non-dominated sorting genetic algorithm with an elitist strategy (NSGA-II):

[0067] The population size is set to 50, and the number of iterations is 100 generations;

[0068] Crossover probability 0.9, mutation probability 0.1;

[0069] The algorithm outputs a set of Pareto front nondominated solutions. Each solution is a compromise that cannot further optimize either objective without harming the other objective. Let M represent the Pareto optimal solution set, and M represent the number of Pareto solutions. This is the Mth non-dominated solution in the solution set.

[0070] 4) Select the final lighting strategy:

[0071] The final lighting strategy is selected from the Pareto solution set, and the selection rule is as follows: That is, choosing the solution that minimizes the "maximum weighted health risk" reflects the fairness principle of "minimizing the maximum risk".

[0072] 5) Implement lighting control:

[0073] Select the lighting parameters This translates into specific lighting control commands, including: adjusting the PWM duty cycle of the LED lights to achieve the target illuminance; adjusting the mixing ratio of warm and cool white LEDs to achieve the target color temperature; adjusting the ratio of multi-color LEDs (such as RGBW) to achieve the target spectral distribution; and setting the gradient time to Δt (default 5 minutes) to avoid sudden changes in lighting that could cause user discomfort.

[0074] For example, the swarm intelligence evolution module is connected to the multi-objective health intervention module and deployed on a cloud server. It aggregates model updates from multiple edge devices through a federated learning mechanism, continuously iterating and optimizing the global model for the current space. The edge devices are deployed in each independent lighting space and include: a computing unit using a low-power edge AI chip for real-time lightweight inference of each module; a local storage unit for storing the user's encrypted digital twin copy and local data from the past 30 days; a communication unit supporting Wi-Fi 6 or 5G for encrypted communication with the cloud, where the communication content is only model gradients; and a federated learning client with a built-in federated learning framework that periodically triggers local model training and uploads gradients.

[0075] This application provides a full life-cycle health lighting system. For the first time, this system proposes an automatic identification mechanism based on multimodal data of user life stages and specific physiological states. This enables precise anchoring of physiological characteristics from all age groups (infants to the elderly) to special groups (pregnant women, disabled individuals, etc.), constructing a truly personalized health lighting system. For multi-user coexistence spaces, it dynamically quantifies real-time health risks to establish priority weights, seeking optimal solutions for maximizing global health benefits through compromise lighting strategies. Relying on a federated learning architecture to aggregate edge model updates, it achieves continuous iteration of the underlying algorithm and distributed self-evolution of the global model, ensuring the system's long-term viability.

[0076] like Figure 2 As shown in the figure, this application provides a structural diagram of a life stage identification module, as follows: Figure 2 As shown, the life stage identification module 120 provided in this application embodiment has a built-in multimodal fusion neural network 1200, which includes a feature extraction layer 1210, a cross-modal attention fusion layer 1220 and a classification output layer 1230.

[0077] The feature extraction layer 1210 is used to extract facial feature vectors, gait feature vectors, physiological feature vectors, and behavioral feature vectors based on the multimodal data.

[0078] In some embodiments, the feature extraction layer 1210 includes a facial image subnetwork 1211, a gait video subnetwork 1212, a physiological time-series data subnetwork 1213, and a behavioral thermal factor network 1214.

[0079] The facial image subnetwork is used to extract the facial feature vector from the facial image sequence. The facial feature vector is used to encode related phenotypes including but not limited to skin elasticity, wrinkle density, and facial muscle activity.

[0080] For example, facial image subnetworks include ResNet50, EfficientPhys, etc.

[0081] For example, the facial image subnetwork extracts facial feature vector F from a sequence of facial images. f ∈R256, encoding aging / development-related phenotypes such as skin elasticity, wrinkle density, and facial muscle activity.

[0082] The gait video sub-network is used to extract the gait feature vector from the gait video sequence. The gait feature vector is used to encode dynamic features including but not limited to stride length, stride frequency, left-right symmetry, and center of gravity shift.

[0083] For example, gait video subnetworks include 3D-CNN+LSTM, SlowFast Networks, etc.

[0084] For example, the gait feature subnetwork (3D-CNN+LSTM) extracts gait feature vectors Fg∈R12 from gait video sequences, encoding dynamic features such as stride length, stride frequency, left-right symmetry, and center of gravity shift.

[0085] The physiological time-series data subnetwork is used to extract the physiological feature vector from the physiological indicator time-series data. The physiological feature vector is used to encode rhythmic features including but not limited to heart rate variability and respiratory rate.

[0086] For example, the physiological time-series data subnetwork includes a Transformer encoder, a local feature extractor, etc.

[0087] For example, the physiological feature subnetwork extracts the physiological feature vector F from the time-series data of physiological indicators. p ∈R64, encoding rhythmic features such as heart rate variability and respiratory rate.

[0088] The behavioral heatmap network is used to extract the behavioral feature vector from the behavioral feature data corresponding to the behavioral heatmap. The behavioral feature vector is used to encode, but is not limited to, indoor activity intensity, location preference, and sedentary / sleep duration.

[0089] Examples of behavioral heat factor networks include CNN, ConvLSTM, and Spatiotemporal Transformer.

[0090] For example, the behavior feature subnetwork extracts the behavior feature vector F from the behavior heatmap. b ∈R64, encoding patterns such as indoor activity intensity, location preference, and sedentary / sleep duration.

[0091] The multimodal data includes the facial image sequence, the gait video sequence, the behavioral feature data, and the time-series data of physiological indicators.

[0092] The cross-modal attention fusion layer is used to interactively fuse the facial feature vector, the gait feature vector, the physiological feature vector, and the behavioral feature vector to generate a fused feature vector.

[0093] For example, the cross-modal attention fusion layer is used to interactively fuse the facial feature vector, the gait feature vector, the physiological feature vector, and the behavioral feature vector to generate the expression corresponding to the fused feature vector:

[0094]

[0095] in, , , , .

[0096] in, These represent the query, key, and value matrices, respectively. Let X represent the dimension of the key vector, X represent the stacked input matrix, and W represent the weight matrix.

[0097] Through this mechanism, the model automatically learns the association weights between different modalities. For example, when determining "old age," gait features should be given high attention weight to facial features; when determining "adolescence," behavioral features should be given high attention weight to physiological features. The output of the multi-head attention is then processed through residual connections, layer normalization, and a feedforward neural network to generate a fused feature vector F. fused ∈R4×256, then flattened into an R1024-dimensional vector.

[0098] The classification output layer is used to output the probability distribution of a user's specific life stage based on the fused feature vector.

[0099] For example, fusing feature vector F fusedThe input classification output layer consists of a fully connected network and a softmax activation function. The fully connected network linearly transforms the 1024-dimensional features to a K-dimensional space (K = number of life stage categories + number of special physiological state categories, set to 10), outputting an unnormalized score vector Z ∈ R10. The softmax function transforms the scores into a probability distribution:

[0100]

[0101] in, For the first The original score for each category, where K represents the total number of output categories. For the first Class prediction results.

[0102] Output probability distribution vector P = [P1, P2, ..., P 10 These correspond to the stages of infancy, early childhood, childhood, adolescence, youth, middle age, old age, pregnancy, postpartum, and disability, respectively. Furthermore, the system takes argmax(P) as the final recognition result.

[0103] This application provides a structure for a life stage recognition module. The life stage recognition module incorporates a multimodal fusion neural network, which includes a feature extraction layer, a cross-modal attention fusion layer, and a classification output layer. The feature extraction layer processes four-dimensional heterogeneous data (facial, gait, physiological, and behavioral) in parallel, breaking the limitations of single-modal perception. By utilizing the complementary effect of information redundancy, it significantly improves the system's robustness in complex environments such as sudden changes in illumination and partial occlusion. The cross-modal attention fusion layer abandons simple feature splicing and dynamically mines deep semantic relationships between modalities through joint inference and adaptively allocates weights to suppress modal conflicts and noise interference, achieving high-precision anchoring of life stages and special states. The classification output layer uses probability distribution instead of hard labels, quantifying the ambiguity and transitional states of life stage evolution, providing a continuous decision basis for the smooth transition of subsequent lighting strategies, and avoiding the discomfort caused by illumination intervention due to sudden changes in state.

[0104] like Figure 3 As shown in the figure, this application provides a schematic diagram of the workflow of a stage migration unit. Figure 3As shown, the stage-constrained digital twin module 130 further includes a stage migration unit 131. When a user is detected migrating from the first life stage to the second life stage, parameter smooth migration is performed. If the first life stage is childhood and the second life stage is adolescence, when the user migrates from the first life stage to the second life stage, parameter smooth migration is performed through migration trigger judgment rules. The user's childhood digital twin is loaded; the low-level feature extraction parameters of the childhood digital twin are frozen, wherein the low-level feature extraction parameters include basic vital sign parameters and basic voice feature parameters; the top-level feature extraction parameters of the childhood digital twin are extracted, wherein the top-level feature extraction parameters include sleep time parameters and calorie consumption parameters; the user's adolescence adaptation data is collected; the top-level feature extraction parameters are fine-tuned and optimized using the adolescence adaptation data, and combined with the low-level feature extraction parameters, an adolescence digital twin is constructed and generated.

[0105] For example, when the first life stage is childhood and the second life stage is adolescence, the workflow of the stage migration unit 131 when a user migrates from the first life stage to the second life stage is as follows:

[0106] 1) Migration trigger judgment: If a user is detected to exhibit adolescent characteristics for 7 consecutive days with a confidence level greater than 85%, then migration is triggered.

[0107] 2) Load the source model, using the parameters from the user's original childhood digital twin as the starting point for transfer learning.

[0108] 3) Freeze the underlying data and freeze the underlying feature extraction parameters of the childhood model. Considering that although the heart structure develops from 10 to 14 years old, the baseline morphology and R wave characteristics of the electrocardiogram are similar, and the basic vital signs parameters are still applicable; extract the parameters of the Mel frequency cepstral coefficients of the sound to identify phonemes. Even if the voice changes in teenagers, the basic physical feature extraction method of pronunciation remains unchanged.

[0109] 4) Fine-tune the top-level parameters. Using only the adolescent adaptation data collected in the last 14 days, fine-tune the top-level parameters of the digital twin in childhood. For example, in adolescence, delayed melatonin secretion leads to a shift in the circadian rhythm. Reduce the weight of "reduced light exposure" and increase the weight of the "social stimulation / academic pressure" feature to predict delayed sleep time.

[0110] 5) Generate new models, create digital twins of adolescents, and put them into online use.

[0111] 6) Gradual lighting during the transition period: During the 7-day transition period after the migration is completed, the system adopts a gradual lighting intervention strategy to avoid discomfort caused by sudden changes.

[0112] It should be noted that, based on the above parameter smoothing transfer steps, the distribution parameters (such as mean, variance, and rhythm threshold) in the physiological baseline model and the mapping parameters (such as weight matrix and decision tree node threshold) in the health risk map can be fine-tuned through transfer learning in order to retain some physiological characteristics of the user in the historical stage and adapt to the health needs of the new stage.

[0113] It should be noted that the 7 days, 14 days, etc., listed in the above examples are merely illustrative and this application imposes limitations on them.

[0114] This application provides a workflow for a stage transition unit. In this application, transfer learning fine-tuning is used instead of model reconstruction, so that the physiological baseline model retains the core physiological characteristics of the user's historical stage during the transition, avoids personalized data gaps, and ensures the continuity of the system's understanding of the user. By smoothly migrating the distribution parameters and the mapping parameters in the health risk map, the oscillation of calculation results caused by model updates is avoided, and the sudden changes in lighting strategy and visual discomfort caused by state switching are eliminated from the root of the underlying algorithm.

[0115] like Figure 4 As shown in the figure, this application embodiment provides a schematic diagram of the negotiation process of a multi-objective health intervention module, as follows: Figure 4 As shown in the example, this application uses the presence of elderly people and infants in the living room to illustrate the negotiation process of the multi-objective health intervention module.

[0116] The output requirements for infant twins are: color temperature 2700K, illuminance ≤50 lux, to promote melatonin secretion. The output requirements for elderly twins are: color temperature 4000K, illuminance ≥300 lux, color rendering index Ra≥90, and no glare, to prevent the risk of falls.

[0117] The health risk calculation unit calculates the instantaneous health risk values ​​for the two users under the current lighting parameters.

[0118] A multi-objective optimization solver is used to construct the objective function Minimize [R_infant, R_old], and the Pareto front is solved using the NSGA-II algorithm.

[0119] The strategy generation unit outputs the following compromise strategy: The living room uses basic lighting of 3500K / 150 lux; the crib area is supplemented with directional lighting of 2700K / 30 lux; the elderly corridor is supplemented with accent lighting of 300 lux; all luminaires are designed to be anti-glare.

[0120] like Figure 5 As shown in the figure, this application provides a schematic diagram of a federated learning architecture for a swarm intelligence evolution module, as follows. Figure 5As shown, edge devices are deployed in various homes or locations, containing local data storage and local models. These edge devices periodically use local data to train the model within the current space, calculate model update gradients, and upload the encrypted gradients to a cloud server.

[0121] The cloud server comprises an aggregation unit and a global model library. The aggregation unit receives model update gradients from multiple edge devices, aggregates them using the FedAvg federated averaging algorithm, and updates the global model in the global model library. The model distribution unit distributes the updated global model to each edge device, completing one round of federated learning iteration.

[0122] like Figure 6 As shown in the flowchart, this application provides a control method for a full life-cycle healthy lighting system. Figure 6 As shown, the control method for the whole life cycle healthy lighting system provided in this application embodiment includes the following steps S61 to S67.

[0123] S61 collects multimodal data from all users within the current space.

[0124] S62, determine the probability distribution of the user's life stage corresponding to each set of multimodal data.

[0125] S63, construct or invoke a corresponding digital twin for each user based on the probability distribution.

[0126] S64, Based on the digital twin and the user's real-time body data, predict the user's current health status and quantify the risk prediction value.

[0127] S65, collect the digital twins of all users in the current space and the risk prediction values, and construct a multi-objective optimization function.

[0128] S66, a multi-objective optimization algorithm is used to solve the multi-objective lighting optimization problem in the current space to obtain the target lighting intervention strategy.

[0129] S67, the control command corresponding to the target lighting intervention strategy is sent to each LED lamp in the current space to adjust the lighting information of each LED lamp and achieve precise light output in the current space.

[0130] For example, the lighting information for each LED luminaire includes the PWM duty cycle, drive current, and multi-color LED ratio.

[0131] For example, the specific steps for issuing control commands corresponding to the target lighting intervention strategy to each LED lamp in the current space, adjusting the lighting information of each LED lamp, and achieving precise light output in the current space are as follows:

[0132] I. Control Command Conversion

[0133] Lighting strategies output by the multi-objective health intervention module Converted into control commands:

[0134] Illumination I <![CDATA[PWM=I / I max ×100%, adjust PWM duty cycle]]> Color temperature CCT Refer to the table to obtain the warm and cool white color ratio, and adjust the dual-channel ratio. Spectral weight Directly mapped to the weights of each RGBW channel. Gradual change time Δt Set the gradient step size and total duration; the default is 300 seconds.

[0135] II. PWM Duty Cycle Adjustment

[0136] Pulse width modulation technology is used:

[0137]

[0138] in, This represents the average illuminance (in lx) of the actual output of the LED. This indicates the rated maximum illuminance of an LED at 100% duty cycle (unit: lx). This represents the time the LED is lit within one PWM cycle (unit: ), Indicates PWM period (unit: ), This indicates the proportion of the total time that the light is on.

[0139] in, Right now Implementation parameters: PWM frequency 20KHz (to avoid flicker), resolution 12-bit (0-4095 levels).

[0140] Implementation parameters: PWM frequency 20KHz (to avoid flickering), resolution 12-bit (0-4095 levels).

[0141] III. Color Temperature Adjustment (Dual Channel for Warm and Cool White)

[0142] Determine the weight of the cool white channel based on the target color temperature. Warm white channel weight ;

[0143] ,

[0144] in, This indicates that the PWM value of the cool white channel is calculated (cool white channel duty cycle = cool white weight × total brightness).

[0145] in, This indicates that the PWM value of the warm white channel is calculated (warm white channel duty cycle = warm white weight × total brightness).

[0146] Color temperature-ratio example:

[0147] 2700K 0% 100% Baby sleep 3500K 40% 60% Multiple compromises 4000K 60% 40% Activities for the elderly 5000K 85% 15% Teen Learning

[0148] IV. Gradual Transition Control (Smooth Transition)

[0149] Linear interpolation is used to achieve a smooth gradient, avoiding discomfort caused by abrupt changes.

[0150]

[0151] in, Indicates time Lighting parameters (illuminance, color temperature, etc.). Indicates the current lighting parameters. Indicates the target lighting parameters. Indicates the elapsed time (in seconds) since the start of the gradual change. Indicates the total fading time (unit: seconds, default 300 seconds).

[0152] Execution parameters: Total fade time is 300 seconds by default, and the update frequency is 10Hz (updates once every 0.1 seconds).

[0153] This application provides an embodiment of: living room main light illumination.

[0154] Scenario: The living room main light switches from off (current PWM=0) to 3500K / 150 lx (target state).

[0155] Step 1:

[0156] Target illuminance 150 lx → = 30%

[0157] Target color temperature 3500K → =40%, =60%

[0158] Gradient time = 300 seconds

[0159] Step two is executed as follows:

[0160] (12-bit resolution) → = 4095 × 30% = 1229

[0161] Step 3:

[0162] =40%×1229=492

[0163] =60%×1229=737

[0164] Step four is to be executed:

[0165] PWM(t) = 0 + (492−0)×t / 300 is calculated every 0.1 seconds and transmitted to the lighting fixture via the Zigbee protocol.

[0166] Execution result: Within 300 seconds, the lights smoothly transitioned from off to 3500K / 150 lx without any abrupt changes.

[0167] In some embodiments, the method further includes: monitoring multimodal feedback data of users after intervention; and updating the target lighting intervention strategy based on the multimodal feedback data.

[0168] It should be noted that each step in the method of this application has been described in detail in the above-mentioned whole life cycle healthy lighting system, and will not be repeated here.

[0169] The scope of protection for the whole life cycle healthy lighting method described in this application is not limited to the order of steps listed in this embodiment. Any solution implemented by adding, subtracting, or replacing steps in the prior art based on the principles of this application is included within the scope of protection of this application.

[0170] In the embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, or methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of apparatuses or modules or units may be electrical, mechanical, or other forms.

[0171] The modules / units described as separate components may or may not be physically separate. The components shown as modules / units may or may not be physical modules; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules / units can be selected to achieve the objectives of the embodiments of this application, depending on actual needs. For example, the functional modules / units in the various embodiments of this application may be integrated into one processing module, or each module / unit may exist physically separately, or two or more modules / units may be integrated into one module / unit.

[0172] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0173] This application also provides an electronic device. Figure 7 The diagram shown is a structural schematic of an electronic device 70 in one embodiment of this application. The full life-cycle healthy lighting system provided in this embodiment can be applied to… Figure 7 The electronic device 70 shown is an example, but not limited to it. For example... Figure 7 As shown, the electronic device 70 includes a processor 71, a memory, a system bus 73, and a network interface 75. The memory may include a non-volatile storage medium 72 and internal memory 74.

[0174] The non-volatile storage medium 72 can store an operating system and a computer program. The computer program includes program instructions that, when executed, cause a processor to perform any of the full-lifecycle healthy lighting methods provided in the embodiments of this application.

[0175] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0176] The internal memory 74 provides an environment for the execution of a computer program in a non-volatile storage medium. When the computer program is executed by the processor, it enables the processor to execute any of the full life cycle healthy lighting methods provided in the embodiments of this application.

[0177] This network interface 75 is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0178] It should be understood that processor 71 can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, the general-purpose processor can be a microprocessor or any conventional processor.

[0179] The electronic device 70 in this application embodiment may include terminal devices such as tablet computers, laptop computers, mobile phones, supercomputers, and smart wearable devices. It can also be applied to databases, servers, and service response systems based on terminal artificial intelligence. This application embodiment does not impose any restrictions on the specific type of electronic device.

[0180] For example, electronic devices can be stations (STAION, ST) in WLANs, cellular phones, cordless phones, Session Initiation Protocol (SIP) phones, Wireless Local Loop (WLL) stations, handheld devices with wireless communication capabilities, computing devices or other processing devices connected to a wireless modem, computers, laptops, handheld communication devices, handheld computing devices, and / or other devices for communicating over wireless systems, as well as next-generation communication systems, such as mobile terminals in 5G networks, mobile terminals in future evolved Public Land Mobile Networks (PLMNs), or mobile terminals in future evolved Non-terrestrial Networks (NTNs).

[0181] This application also provides a computer-readable storage medium. Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing a processor. The program can be stored in a computer-readable storage medium, which is a non-transitory medium, such as random access memory, read-only memory, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disk, and any combination thereof. The storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., digital video disc (DVD)), or a semiconductor medium (e.g., solid-state disk (SSD)).

[0182] This application embodiment may also provide a computer program product comprising one or more computer instructions. When the computer instructions are loaded and executed on a computing device, all or part of the processes or functions described in this application embodiment are generated. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means.

[0183] When the computer program product is executed by a computer, the computer performs the method described in the foregoing method embodiments. The computer program product can be a software installation package; when the foregoing method is required, the computer program product can be downloaded and executed on the computer.

[0184] The descriptions of the processes or structures corresponding to the above figures each have their own emphasis. For parts of a process or structure that are not described in detail, please refer to the relevant descriptions of other processes or structures.

[0185] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A whole-life-cycle healthy lighting system, characterized in that, The system includes: The multimodal perception module is used to collect multimodal data from all users within the current space; A life stage identification module, connected to the multimodal perception module, is used to determine the probability distribution of the user corresponding to the multimodal data being in a specific life stage based on each set of multimodal data; A stage constraint digital twin module, connected to the life stage identification module, is used to instantiate a stage constraint digital twin for each user in the current space based on the probability distribution; A multi-objective health intervention module, connected to the stage-constrained digital twin module, is used to determine the target lighting intervention strategy based on the digital twins of all users in the current space using a multi-objective optimization algorithm. The swarm intelligence adaptive module, connected to the multi-target health intervention module, is used to aggregate model update parameters corresponding to multiple edge devices in the current space based on the target lighting intervention strategy via a federated learning mechanism, so as to realize adaptive health lighting in the current space.

2. The system according to claim 1, characterized in that, The multimodal sensing module includes a first camera, a second camera, and sensors; The first camera is used to capture user facial image sequences and physiological feature sequences; The second camera, in conjunction with the sensor, extracts the user's gait feature sequence and behavioral feature sequence through a skeletal point recognition algorithm.

3. The system according to claim 1, characterized in that, The life stage identification module incorporates a multimodal fusion neural network, which includes: a feature extraction layer, a cross-modal attention fusion layer, and a classification output layer; The feature extraction layer is used to extract facial feature vectors, gait feature vectors, physiological feature vectors, and behavioral feature vectors based on the multimodal data; The cross-modal attention fusion layer is used to interactively fuse the facial feature vector, the gait feature vector, the physiological feature vector, and the behavioral feature vector to generate a fused feature vector; The classification output layer is used to output the probability distribution of a user's specific life stage based on the fused feature vector.

4. The system according to claim 3, characterized in that, The feature extraction layer includes a facial image subnetwork, a gait video subnetwork, a physiological time-series data subnetwork, and a behavioral thermal factor network. The facial image subnetwork is used to extract the facial feature vector from the facial image sequence. The facial feature vector is used to encode related phenotypes including but not limited to skin elasticity, wrinkle density, and facial muscle activity. The gait video sub-network is used to extract the gait feature vector from the gait video sequence. The gait feature vector is used to encode dynamic features including but not limited to stride length, stride frequency, left-right symmetry, and center of gravity shift. The physiological time-series data subnetwork is used to extract the physiological feature vector from the physiological indicator time-series data. The physiological feature vector is used to encode rhythmic features including but not limited to heart rate variability and respiratory rate. The behavioral heatmap network is used to extract the behavioral feature vector from the behavioral feature data corresponding to the behavioral heatmap. The behavioral feature vector is used to encode, but is not limited to, indoor activity intensity, location preference, and sedentary / sleep duration. The multimodal data includes the facial image sequence, the gait video sequence, the behavioral feature data, and the time-series data of physiological indicators.

5. The system according to claim 1, characterized in that, The digital twin has a built-in physiological baseline model and health risk map corresponding to the specific life stage; The physiological baseline model is a statistical model or a multivariate Gaussian model. The physiological baseline model defines the healthy range, daily rhythm template and normal fluctuation pattern of at least one physiological indicator under the specific life stage. The physiological indicator includes heart rate, respiratory rate and heart rate variability. The health risk map is an association matrix or a decision tree, and the mapping relationship between health risk types and lighting intervention directions is defined in the health risk map; The digital twin compares the real-time sensed physiological values ​​with the standard values ​​in the physiological baseline model and calculates the deviation. If the deviation exceeds the threshold, the health risk map is queried to obtain the health risk type and recommended lighting intervention direction.

6. The system according to claim 1, characterized in that, The stage-constrained digital twin module also includes a stage migration unit, which performs parameter smooth migration when it detects that a user is migrating from the first life stage to the second life stage. If the first life stage is childhood and the second life stage is adolescence, when a user migrates from the first life stage to the second life stage, the migration trigger judgment rule execution parameters are used to smoothly migrate. Load the user's childhood digital twin; Freeze the low-level feature extraction parameters of the childhood digital twin, wherein the low-level feature extraction parameters include basic vital sign parameters and basic speech feature parameters; Extract the top-level feature extraction parameters of the childhood digital twin, wherein the top-level feature extraction parameters include sleep time parameters and calorie consumption parameters; Collect the user's adolescent adaptation data; The top-level feature extraction parameters are fine-tuned and optimized using the adolescent adaptation data, and combined with the bottom-level feature extraction parameters to construct and generate an adolescent digital twin.

7. A control method for a full life-cycle healthy lighting system, characterized in that, Applied to the system according to any one of claims 1 to 6, the method comprises: Collect multimodal data from all users within the current space; Based on each set of multimodal data, determine the probability distribution of the user's life stage corresponding to the multimodal data; Based on the probability distribution, a corresponding digital twin is constructed or invoked for each user; Based on the digital twin and the user's real-time physical data, the user's current health status is predicted, and the risk prediction value is quantified. Collect the digital twins of all users in the current space and the risk prediction values, and construct a multi-objective optimization function; A multi-objective optimization algorithm is used to solve the multi-objective lighting optimization problem in the current space to obtain the target lighting intervention strategy. The control command corresponding to the target lighting intervention strategy is sent to each LED lamp in the current space to adjust the lighting information of each LED lamp and achieve precise light output in the current space.

8. The method according to claim 7, characterized in that, The method further includes: Multimodal feedback data from users after monitoring and intervention; The target lighting intervention strategy is updated based on the multimodal feedback data.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 7 to 8.

10. An electronic device, characterized in that, The electronic device includes: A memory that stores a computer program; The processor, which is communicatively connected to the memory, executes the method of any one of claims 7 to 8 when the computer program is invoked.

Images