LED parameter adaptive control system and method based on neural network

Abstract

The invention belongs to the technical field of adaptive control, and discloses an LED parameter adaptive control system and method based on a neural network. The method comprises the steps of collecting sleep quality data; performing sleep quality analysis based on the sleep quality data to obtain sleep quality feature data; acquiring space environment data; analyzing and processing the space environment based on the space environment data to obtain space environment feature data; collecting sober key data after the user is sober; performing feature extraction on the sober key data to obtain sober feature data; performing intelligent self-adaptive control on LED parameters based on the sleep quality characteristic data, the space environment characteristic data and the waking characteristic data; according to the scheme, intelligent self-adaptive regulation and control of the LED parameters are carried out according to different physiological states of the user, and particularly, a suitable illumination environment is provided in a just waking state, so that personalized self-adaptive illumination regulation is realized, and the comfort and adaptability of the user are improved.

Description

Technical Field

[0001] This invention relates to the field of adaptive control technology, and more specifically, to an LED parameter adaptive control system and method based on neural networks. Background Technology

[0002] Intelligent lighting not only provides basic illumination but also adjusts lighting parameters based on environmental and user behavior to enhance user comfort and optimize energy management. Especially during sleep and wakefulness, lighting significantly impacts the body's circadian rhythms and visual comfort. Therefore, providing suitable lighting upon waking, avoiding discomfort caused by excessively high brightness, and gradually adjusting lighting parameters as the user's vision adapts to maintain a comfortable lighting environment, has become a crucial research direction for intelligent lighting systems.

[0003] However, existing intelligent lighting systems are mainly based on timed switches, manual adjustments, or simple light sensor control, making it difficult to accurately adjust lighting for users who have just woken up. While some intelligent lights can adjust brightness and color temperature, the adjustment methods are fixed and cannot dynamically adjust lighting parameters based on the user's adaptation. Furthermore, existing systems lack awareness of the user's physiological state and do not fully utilize ambient light data and the user's physiological state to dynamically adjust lighting parameters. This results in a mechanized lighting parameter adjustment process with poor generalization of lighting adjustment strategies. Brightness may be increased or color temperature switched too quickly before the user has fully adapted to the strong light, impacting the user experience and failing to achieve truly personalized and comfortable intelligent control, thus failing to meet user needs.

[0004] Therefore, there is an urgent need for a lighting control method that combines environmental perception, user physiological feedback, and intelligent prediction to achieve more natural, gentle, and personalized wake-up lighting adjustment, improve user comfort, and optimize the adaptability of intelligent lighting systems. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: an LED parameter adaptive control method based on a neural network, comprising:

[0006] Collect sleep quality data; the sleep quality data includes total deep sleep time, total light sleep time, total REM sleep time, number of awakenings, and the awakening time corresponding to each awakening;

[0007] Sleep quality analysis is performed based on sleep quality data to obtain sleep quality characteristic data, which includes the proportion of deep sleep, the proportion of light sleep, the proportion of rapid eye movement (REM) sleep, sleep stability index, circadian rhythm matching index, and sleep quality assessment score.

[0008] Collect spatial environment data; the spatial environment data includes wall material, wall color, floor material, ground color, ambient temperature, ambient humidity, and ambient brightness at N sampling points;

[0009] Based on spatial environment data, spatial environment characteristic data is obtained by analyzing and processing the spatial environment data; the spatial environment characteristic data includes the uniformity of illumination environment, wall reflectivity, wall absorptivity, floor reflectivity, floor absorptivity, ambient temperature and ambient humidity;

[0010] Collect key data on the user's wakefulness after they have regained consciousness; the key data on wakefulness includes the user's sleep state when awake, the moment of awakening, and the moment the lights were turned on.

[0011] Feature extraction is performed on key data of wakefulness to obtain wakefulness feature data; the wakefulness feature data includes sleep state values ​​and light response delay during wakefulness.

[0012] Based on sleep quality characteristic data, spatial environment characteristic data, and wakefulness characteristic data, corresponding LED parameters are adaptively adjusted for wakefulness during deep sleep, light sleep, and REM sleep.

[0013] Furthermore, the methods for adaptively adjusting corresponding LED parameters for wakefulness during deep sleep, light sleep, and REM sleep include:

[0014] Extract sleep state values ​​and light response delay during wakefulness from wakefulness feature data;

[0015] If the sleep state value during wakefulness corresponds to the wakefulness state during deep sleep, then the sleep state value during wakefulness is recorded as the wakefulness value during deep sleep. Based on the wakefulness value during deep sleep, light response delay, spatial environment characteristic data, and sleep quality characteristic data, the light parameters are adaptively adjusted for wakefulness during deep sleep.

[0016] If the sleep state value during wakefulness corresponds to the wakefulness stage of light sleep, then the sleep state value during wakefulness is recorded as the wakefulness value during light sleep. Based on the wakefulness value during light sleep, light response delay, spatial environment characteristic data, and sleep quality characteristic data, the light parameters are adaptively adjusted for wakefulness during light sleep.

[0017] If the sleep state value during wakefulness corresponds to the REM sleep state, then the sleep state value during wakefulness is recorded as the REM sleep state value. Based on the REM sleep state value, light response delay, spatial environment characteristic data, and sleep quality characteristic data, the light parameters are adaptively adjusted for REM sleep state.

[0018] Furthermore, methods for adaptively adjusting light parameters to address wakefulness during deep sleep include:

[0019] S100: Input the wakefulness value during deep sleep, light response delay, spatial environment characteristic data and sleep quality characteristic data into the deep sleep parameter setting model to obtain deep sleep adjustment parameters. The deep sleep adjustment parameters include the initial value of light intensity during wakefulness in deep sleep, the target value of light intensity, the initial value of color temperature, the target value of color temperature, the total adjustment time, the adjustment interval duration, the brightness adjustment rate and the color temperature adjustment rate.

[0020] S101: Calculate the total adjustment time during deep sleep and wakefulness by dividing the adjustment interval to obtain the corresponding number of adaptive adjustments, denoted as... The initial value of ycs is set to 1, and the range of ycs values ​​is from 1 to... ;

[0021] S102: The ycs-th adaptively adjusted illuminance is calculated based on the initial value of illuminance during deep sleep, the target value of illuminance, and the brightness adjustment rate, and is denoted as the first adjusted illuminance.

[0022] The color temperature of the ycsth adaptive adjustment is calculated based on the initial color temperature value, target color temperature value, and color temperature adjustment rate during the deep sleep stage. This is denoted as the first adjustment color temperature.

[0023] S103: When the adjustment interval between deep sleep and wakefulness is reached, perform the ycs-th adaptive adjustment, adjust the illuminance to the first adjusted illuminance, and adjust the color temperature to the first adjusted color temperature;

[0024] S104: Let ycs = ycs + 1. If ycs is less than or equal to 1... If ycs is greater than 0, then continue executing S102 to S103. Then, the adaptive adjustment of light parameters for awakening during deep sleep is completed, and the current process ends.

[0025] Furthermore, methods for adaptively adjusting light parameters to address wakefulness during light sleep include:

[0026] S200: Input the wakefulness value during the light sleep stage, the light response delay, the spatial environment characteristic data, and the sleep quality characteristic data into the light sleep parameter setting model to obtain the light sleep adjustment parameters. The light sleep adjustment parameters include the initial value of light intensity during the wakefulness stage of light sleep, the target value of light intensity, the initial value of color temperature, the target value of color temperature, the total adjustment time, the adjustment interval duration, the brightness adjustment rate, and the color temperature adjustment rate.

[0027] S201: Calculate the total adjustment time during the light sleep stage by dividing the adjustment interval, and obtain the corresponding number of adaptive adjustments, denoted as... The default initial value of ECS is 1, and the value of ECS ranges from 1 to... ;

[0028] S202: The light intensity of the first adaptive adjustment is calculated based on the initial light intensity value, the target light intensity value and the brightness adjustment rate during the light sleep stage, and is denoted as the second adjusted light intensity.

[0029] The color temperature of the first adaptive adjustment is calculated based on the initial color temperature value, target color temperature value, and color temperature adjustment rate during the light sleep stage, and is denoted as the second adjustment color temperature.

[0030] S203: When the adjustment interval for awakening during the light sleep stage is reached, the first ECS adaptive adjustment is performed, adjusting the illuminance to the second adjusted illuminance and the color temperature to the second adjusted color temperature;

[0031] S204: Let ecs = ecs + 1. If ecs is less than or equal to 1, then... If S202 to S203 are executed, then if ECS is greater than... Then, the adaptive adjustment of light parameters for awakening during the light sleep stage is completed, and the current process ends.

[0032] Furthermore, methods for adaptively adjusting illumination parameters during REM (Rapid Eye Movement) awakening include:

[0033] S300: Input the wakefulness value, light response delay, spatial environment characteristic data and sleep quality characteristic data of the rapid eye movement (REM) stage into the REM parameter setting model to obtain the REM accommodation parameters. The REM accommodation parameters include the initial value of light intensity, the target value of light intensity, the initial value of color temperature, the target value of color temperature, the total accommodation time, the accommodation interval duration, the brightness adjustment rate and the color temperature adjustment rate during the REM stage wakefulness.

[0034] S301: The total accommodation time during the REM (Rapid Eye Movement) awake phase is divided by the accommodation interval to obtain the corresponding number of adaptive accommodation cycles, denoted as... The default initial value of scs is 1, and the value range of scs is from 1 to 1. ;

[0035] S302: The ssth adaptive adjustment of the illumination is calculated based on the initial value of illumination during the rapid eye movement phase, the target value of illumination, and the brightness adjustment rate. This is denoted as the third adjustment of illumination.

[0036] The color temperature of the scs-th adaptive adjustment is calculated based on the initial color temperature value, target color temperature value, and color temperature adjustment rate during the REM (Rapid Eye Movement) phase of wakefulness, and is denoted as the third adjustment color temperature.

[0037] S303: When the accommodation interval of the REM phase of wakefulness is reached, the scs-th adaptive adjustment is performed, adjusting the illumination to the third adjustment illumination and the color temperature to the third adjustment color temperature;

[0038] S304: Let scs = scs + 1. If scs is less than or equal to 1, then... If S302 to S303 are executed, then continue execution. If scs is greater than... Then, the adaptive adjustment of illumination parameters for the REM (Rapid Eye Movement) awake phase is completed, and the current process ends.

[0039] Furthermore, the method for obtaining the sleep quality characteristic data includes:

[0040] Obtain the total deep sleep time, total light sleep time, total REM sleep time, number of awakenings, and the corresponding awakening time for each awakening from sleep quality data;

[0041] The total sleep time is calculated based on the total deep sleep time, total light sleep time, total REM sleep time, number of awakenings and the awakening time corresponding to each awakening.

[0042] The ratio of deep sleep time to total sleep duration is calculated by dividing the total deep sleep time by the total sleep duration; the ratio of light sleep time to total sleep duration is calculated by dividing the total light sleep time by the total sleep duration; and the ratio of REM sleep time to total REM sleep duration is calculated by dividing the total REM sleep time by the total sleep duration.

[0043] The corresponding sleep stability index is calculated based on the total deep sleep time, total light sleep time, total REM sleep time, number of awakenings and the awakening time corresponding to each awakening.

[0044] The total sleep time is obtained by summing the total deep sleep time, the total light sleep time, and the total REM sleep time.

[0045] The circadian rhythm matching index is calculated based on the total deep sleep time, total light sleep time, total REM sleep time, and total sleep time.

[0046] The proportion of deep sleep, light sleep, rapid eye movement, sleep stability index, and circadian rhythm matching index are input into the sleep quality assessment model to obtain the corresponding sleep quality assessment score.

[0047] The proportion of deep sleep, light sleep, REM sleep, sleep stability index, circadian rhythm matching index, and sleep quality assessment score were used to construct sleep quality characteristic data.

[0048] Furthermore, the method for acquiring the spatial environment feature data includes:

[0049] Obtain wall material, wall color, floor material, ground color, ambient temperature, ambient humidity, and ambient brightness at N sampling points from the spatial environment data;

[0050] The uniformity of the lighting environment is calculated based on the ambient brightness of N sampling points.

[0051] Pre-build a material property data table; obtain the corresponding reflectivity and absorptivity from the material property data table based on the wall material and wall color, and record them as wall reflectivity and wall absorptivity, respectively; obtain the corresponding reflectivity and absorptivity from the material property data table based on the floor material and floor color, and record them as floor reflectivity and floor absorptivity, respectively;

[0052] Spatial environmental characteristic data are constructed by considering factors such as light environment uniformity, wall reflectivity, wall absorptivity, floor reflectivity, floor absorptivity, ambient temperature, and ambient humidity.

[0053] Furthermore, the method for obtaining the sobriety feature data includes:

[0054] Obtain key data on wakefulness, including the user's sleep state when awake, the time of wakefulness, and the time when the lights are turned on.

[0055] Pre-build a matching table of user's sleep state while awake and user's sleep state values ​​while awake; obtain the corresponding sleep state values ​​while awake from the matching table of user's sleep state while awake.

[0056] The light response delay is obtained by subtracting the time when the light is turned on from the time when the person is awake.

[0057] The sleep state values ​​during wakefulness and the light response delay are used to construct wakefulness feature data.

[0058] Furthermore, the training method for the deep sleep parameter setting model includes:

[0059] A deep sleep parameter setting dataset is pre-collected, which includes B group deep sleep parameter setting data and the deep sleep regulation parameters corresponding to the B group deep sleep parameter setting data. The deep sleep parameter setting data includes wakefulness values ​​during deep sleep, light response delay, spatial environment characteristic data, and sleep quality characteristic data. The deep sleep parameter setting dataset is divided into a training set and a validation set, wherein the training set is used to train the deep sleep parameter setting model, and the validation set is used to evaluate the generalization performance of the deep sleep parameter setting model.

[0060] During the training of the deep sleep parameter setting model, minimizing the cross-entropy loss function is used as the optimization objective. An early stopping strategy is used to monitor the performance on the validation set. The model performance is optimized by continuously adjusting the network parameters. Training is stopped when the prediction accuracy on the validation set reaches the expected accuracy. The deep sleep parameter setting model is trained using a deep neural network based on a multilayer perceptron.

[0061] The deep sleep parameter setting data is converted into feature vectors; the input layer of the deep sleep parameter setting model receives the feature vectors, extracts the nonlinear relationships in the data through the hidden layer, and finally the output layer of the deep sleep parameter setting model calculates the probability distribution of the deep sleep regulation parameters through the softmax activation function, and outputs the deep sleep regulation parameter corresponding to the highest probability as the final prediction result.

[0062] Using the same training method as the deep sleep parameter setting model, train the light sleep parameter setting model and the rapid eye movement parameter setting model.

[0063] A neural network-based adaptive control system for LED parameters, used to implement the neural network-based adaptive control method for LED parameters, includes:

[0064] The first acquisition module is used to collect sleep quality data;

[0065] The first processing module performs sleep quality analysis based on sleep quality data to obtain sleep quality characteristic data.

[0066] The second acquisition module is used to acquire space environment data;

[0067] The second processing module analyzes and processes the space environment based on the space environment data to obtain space environment characteristic data;

[0068] The third data acquisition module is used to collect key data on the user's consciousness after they regain consciousness.

[0069] The third processing module is used to extract features from the key data of the conscious state to obtain conscious feature data;

[0070] The intelligent control module, based on sleep quality characteristic data, spatial environment characteristic data, and wakefulness characteristic data, performs corresponding adaptive adjustments to LED parameters for wakefulness during deep sleep, light sleep, and REM sleep.

[0071] Compared with existing technologies, the technical effects and advantages of the LED parameter adaptive control system and method based on neural networks in this invention are as follows:

[0072] This invention provides an LED parameter adaptive control system and method based on neural networks. By combining data on the user's sleep state when awake, light response delay, sleep quality characteristics, and spatial environment characteristics, the system intelligently and adaptively adjusts the light for different physiological states of the user. In particular, it provides a suitable light environment when the user has just woken up, thereby achieving personalized adaptive light adjustment and improving the user's comfort and adaptability.

[0073] Specifically, this invention acquires a user's sleep state while awake, including deep sleep, light sleep, and REM sleep, and analyzes the time interval from wakefulness to light switching on by combining this with light response delay, thereby determining the user's tolerance and adaptation rhythm to light stimulation. Furthermore, this invention combines sleep quality characteristic data and spatial environment characteristic data to construct an adaptive LED parameter adjustment strategy for different wakefulness states. This matches the changes in LED parameters with the user's neurophysiological adaptability, reduces the disturbance of physiological rhythms caused by sudden light exposure, avoids discomfort to the user's eyes from light stimulation, and improves the naturalness and comfort of the wakefulness transition, thereby achieving customized light adjustment based on the individual user's state.

[0074] Compared to traditional LED lighting solutions, this invention not only enables more precise control of LED parameters, but also dynamically optimizes lighting based on the user's physiological needs, improving lighting comfort and health, and has significant technical advantages and application value. Attached Figure Description

[0075] Figure 1 This is a schematic diagram of the LED parameter adaptive control system based on a neural network according to Embodiment 1 of the present invention;

[0076] Figure 2 This is a flowchart of the LED parameter adaptive control method based on neural networks according to Embodiment 3 of the present invention;

[0077] Figure 3 This is a schematic diagram of the LED parameter adaptive control system based on neural networks according to Embodiment 2 of the present invention;

[0078] Figure 4 Flowchart of the method for adaptive adjustment of LED parameters;

[0079] Figure 5 A flowchart illustrating a method for adaptively adjusting light parameters to address wakefulness during deep sleep.

[0080] Figure 6 A diagram illustrating sleep quality data acquired by a smart wearable device. Detailed Implementation

[0081] The technical solutions of the embodiments of the present invention will be described in detail, clearly, and completely below with reference to the accompanying drawings. It should be particularly noted that the specific embodiments described below are only for better illustrating and explaining the technical solutions of the present invention, and are intended to enable those skilled in the art to better understand and implement the present invention, and should not be construed as limiting the scope of protection of the present invention. Without departing from the spirit and substance of the present invention, those skilled in the art can modify, adjust, or make equivalent substitutions based on the content disclosed in the present invention, and these should all be considered within the scope of protection of the present invention.

[0082] Example 1:

[0083] Please see Figure 1 As shown, this embodiment provides an LED parameter adaptive control system based on a neural network, including a first acquisition module, a first processing module, a second acquisition module, a second processing module, a third acquisition module, a third processing module, and an intelligent control module. Each module is connected via wired and / or wireless means to achieve data transmission.

[0084] The first data acquisition module is used to collect sleep quality data, including total deep sleep time, total light sleep time, total REM sleep time, number of awakenings, and the duration of each awakening. This sleep quality data is acquired through a smart wearable device, such as a smart bracelet / watch. The sleep quality data acquired by the smart wearable device is as follows: Figure 6 As shown.

[0085] It's important to note that a person's light sensitivity upon waking is closely related to the quality of their sleep the previous night. This phenomenon is closely related to the regulatory mechanisms of intrinsic photoreceptor ganglion cells (ipRGCs) in the retina. ipRGCs are primarily responsible for regulating circadian rhythms and light adaptation, and their response to light is influenced by melatonin levels. When deep sleep is sufficient, the parasympathetic nervous system is dominant, resulting in greater pupil constriction, reduced light entering the eyes, and a faster decline in melatonin levels. This leads to decreased light sensitivity of ipRGCs, faster pupil adaptation to light, and less discomfort from light stimulation. Therefore, under these conditions, light regulation can transition to normal brightness more quickly, without a prolonged buffer period.

[0086] However, when sleep quality is poor, especially with a high proportion of light sleep, insufficient deep sleep, or frequent awakenings, the sympathetic nervous system is more active, leading to increased pupil dilation and a greater amount of light entering the eyes. Simultaneously, melatonin levels remain high, causing increased photosensitivity of iPRGCs and slower pupil constriction, making people more sensitive to light upon waking and more prone to eye strain. In this situation, rapid changes in light intensity or unsuitable color temperature can cause visual discomfort, temporary blurred vision, or even dizziness, affecting the user experience. Therefore, a gentler lighting transition is needed, gradually increasing brightness and using warm-toned light to reduce stimulation, allowing the eyes sufficient time to adapt to changes in light.

[0087] This solution addresses this physiological phenomenon by optimizing the intelligent light regulation process using sleep monitoring data, ensuring that adaptive light adjustments meet the user's physiological needs. The system collects key sleep parameters such as deep sleep time, light sleep time, REM sleep time, number of awakenings, and awakening time, and dynamically adjusts the rate of light change, initial brightness, and color temperature based on this data. This provides users with a more natural and comfortable lighting experience while reducing the interference of light on the body's circadian rhythms.

[0088] The first processing module performs sleep quality analysis based on sleep quality data to obtain sleep quality characteristic data.

[0089] The method for obtaining the sleep quality characteristic data includes:

[0090] Obtain the total deep sleep time, total light sleep time, total REM sleep time, number of awakenings, and the corresponding awakening time for each awakening from sleep quality data;

[0091] The total sleep time is calculated based on the total deep sleep time, total light sleep time, total REM sleep time, number of awakenings and the awakening time corresponding to each awakening.

[0092] The ratio of deep sleep time to total sleep duration is calculated by dividing the total deep sleep time by the total sleep duration; the ratio of light sleep time to total sleep duration is calculated by dividing the total light sleep time by the total sleep duration; and the ratio of REM sleep time to total REM sleep duration is calculated by dividing the total REM sleep time by the total sleep duration.

[0093] The corresponding sleep stability index is calculated based on the total deep sleep time, total light sleep time, total REM sleep time, number of awakenings and the awakening time corresponding to each awakening.

[0094] The total sleep time is obtained by summing the total deep sleep time, the total light sleep time, and the total REM sleep time.

[0095] The circadian rhythm matching index is calculated based on the total deep sleep time, total light sleep time, total REM sleep time, and total sleep time.

[0096] The proportion of deep sleep, light sleep, rapid eye movement, sleep stability index, and circadian rhythm matching index are input into the sleep quality assessment model to obtain the corresponding sleep quality assessment score.

[0097] The proportion of deep sleep, light sleep, REM sleep, sleep stability index, circadian rhythm matching index, and sleep quality assessment score were used to construct sleep quality characteristic data.

[0098] The method for calculating the total sleep time includes:

[0099] ;

[0100] in, The total sleep time. Total deep sleep time This represents the total time spent in light sleep. The total duration of rapid eye movement (REM) For the number of times one is awake, Let be the awakening time corresponding to the i-th awakening.

[0101] The method for calculating the deep sleep ratio includes:

[0102] ;

[0103] in, The percentage of deep sleep.

[0104] The method for calculating the proportion of light sleep includes:

[0105] ;

[0106] in, The percentage of deep sleep.

[0107] The method for calculating the rapid eye movement ratio includes:

[0108] ;

[0109] in, For rapid eye movement ratio.

[0110] The method for calculating the sleep stability index includes:

[0111] ;

[0112] in, As a sleep stability index, This represents the awakening time for the i-th awakening, raised to the power of i. It emphasizes that the more awakenings one experiences, the greater the impact of the corresponding awakening time on the sleep stability index; that is, the more awakenings and the longer the awakening time, the lower the corresponding sleep stability index. A higher sleep stability index indicates better sleep stability, with the index being highest when the number of awakenings is zero.

[0113] The method for calculating the total sleep time includes:

[0114] ;

[0115] in, This represents the total sleep time.

[0116] The method for obtaining the physiological rhythm matching index includes:

[0117] ;

[0118] in, This is the physiological rhythm matching index. To achieve the target percentage of deep sleep, The target percentage of light sleep. The target proportion for rapid eye movement. Preferably, in the technical solution of the present invention, The value range is [0.20, 0.25]. The value range is [0.50, 0.55]. The value range is [0.20, 0.25].

[0119] It should be noted that the construction method of the circadian rhythm matching index firstly eliminates the influence of differences in total sleep duration among different users by normalizing the time of each stage, so that the circadian rhythm matching index only reflects the relative distribution of sleep structure, thus having good comparability; secondly, it introduces a target proportion determined based on medical experience. , and This method quantifies the "theoretical normal structure" into a vector reference standard, allowing the circadian rhythm matching index to directly characterize the degree to which the current sleep structure deviates from the normal structure. Furthermore, by performing absolute value calculations and summing the three sleep ratio deviations, it effectively measures the distance between the "actual sleep structure vector" and the "target sleep structure vector." This avoids the cancellation of "high" and "low" deviations and offers advantages such as simple calculation, monotonic values, and ease of threshold classification. Therefore, a JLZS closer to 0 indicates a sleep structure closer to the normal circadian rhythm, while a larger JLZS indicates a more unbalanced sleep structure.

[0120] In this invention, the circadian rhythm matching index, as one of the core features of the sleep quality assessment model, is used in conjunction with indicators such as the sleep stability index to characterize the user's overall sleep quality. By introducing the circadian rhythm matching index, this invention can identify users who, although their total sleep duration is sufficient, have sleep structures that significantly deviate from the normal circadian rhythm during the parameter setting stage. This allows for targeted adjustments to light source parameters such as brightness, color temperature, and adjustment rhythm in subsequent deep sleep, light sleep, and rapid eye movement (REM) parameter setting models, making the adaptive light source control strategy more aligned with the individual user's circadian rhythm needs.

[0121] This leads to the realization of intelligent LED parameter control based on quantitative rhythm matching indicators. Compared with traditional solutions that rely solely on total sleep time or the duration of a single stage, this approach can more accurately match the user's actual sleep state, improve the effectiveness and reliability of light intervention in improving sleep quality, and thus provide important theoretical basis and data support for LED parameter adaptive control methods.

[0122] The training method for the sleep quality assessment model includes:

[0123] A pre-constructed sleep quality assessment dataset is provided, comprising Group A sleep quality assessment data and corresponding sleep quality assessment scores, where A is a positive integer greater than 0. The sleep quality assessment data includes the proportion of deep sleep, the proportion of light sleep, the proportion of rapid eye movement (REM) sleep, the sleep stability index, and the circadian rhythm matching index. The sleep quality assessment scores are obtained by multiple individuals skilled in the art based on the proportion of deep sleep, the proportion of light sleep, the proportion of REM sleep, the sleep stability index, and the circadian rhythm matching index. The sleep quality assessment dataset is divided into a training set and a validation set. The training set is used for parameter learning of the sleep quality assessment model, while the validation set is used for real-time evaluation of the generalization ability of the sleep quality assessment model.

[0124] During the training of the sleep quality assessment model, a deep neural network structure based on multilayer perceptrons is adopted. The sleep quality assessment data is converted into feature vectors as input, and nonlinear features in the data are extracted through hidden layers. Finally, the softmax activation function is used in the output layer to generate the probability distribution of sleep quality assessment scores, and the sleep quality assessment score corresponding to the highest probability is output as the final prediction result. The training process aims to minimize the cross-entropy loss function, and an early stopping strategy is introduced to monitor the performance of the sleep quality assessment data validation set. When the prediction accuracy on the sleep quality assessment data validation set reaches the preset accuracy, the sleep quality assessment model is determined to have converged, and the training stops immediately.

[0125] For example, in the technical solution of the present invention, the proportion of deep sleep is 0.23, the proportion of light sleep is 0.52, the proportion of rapid eye movement is 0.25, the sleep stability index is 0.88, the physiological rhythm matching index is 0.05, and the sleep quality assessment score output by the sleep quality assessment model is 88 points, which indicates that the sleep quality is good.

[0126] For example, the proportion of deep sleep was 0.18, the proportion of light sleep was 0.57, the proportion of REM sleep was 0.25, the sleep stability index was 0.65, the circadian rhythm matching index was 0.13, and the sleep quality assessment score output by the sleep quality assessment model was 72, which indicates moderate sleep quality.

[0127] For example, the proportion of deep sleep was 0.12, the proportion of light sleep was 0.63, the proportion of REM sleep was 0.25, the sleep stability index was 0.38, the circadian rhythm matching index was 0.27, and the sleep quality assessment model output a sleep quality assessment score of 48, indicating poor sleep quality.

[0128] It should be noted that the sleep quality assessment model in this invention uses objective sleep structure and circadian rhythm indicators as input. It combines parameters such as the proportion of deep sleep, light sleep, rapid eye movement (REM) sleep, sleep stability index, and circadian rhythm matching index into a feature vector, comprehensively depicting the stage distribution and degree of deviation from the rhythm pattern during a user's entire night's sleep. During the model training phase, multiple individuals skilled in the art provide sleep quality assessment scores based on the aforementioned indicators. The indicators and scores form a pair as a supervision signal, and a deep neural network based on a multilayer perceptron is used to learn the nonlinear mapping relationship between the feature vector and the sleep quality assessment score. The sleep quality assessment score output by the sleep quality assessment model can stably and quantitatively reflect the user's sleep status, providing reliable sleep quality feature input for subsequent deep sleep parameter setting models, light sleep parameter setting models, and REM parameter setting models, thereby supporting the overall technical solution of the LED parameter adaptive control of this invention.

[0129] The second acquisition module is used to collect spatial environment data, including wall material, wall color, floor material, ground color, ambient temperature, ambient humidity, and ambient brightness at N sampling points. Ambient brightness, ambient temperature, and ambient humidity are acquired through corresponding sensors, while wall material, wall color, floor material, and ground color are obtained through pre-collected data and then input into the adaptive control system by the user.

[0130] It should be noted that collecting spatial environmental data aims to obtain key parameters affecting lighting regulation in order to achieve precise and adaptive intelligent lighting control. Ambient brightness reflects the current light intensity in the space and can be used to adjust the output power of LED light sources to avoid over-illumination or under-illumination. Wall materials and colors affect light reflectivity and absorptivity; different materials have different diffuse and specular reflection characteristics, which in turn determine lighting uniformity and glare control strategies. Floor materials and colors also affect light diffusion and visual comfort; for example, high-reflectivity tile floors may cause glare, while dark wood floors may absorb too much light.

[0131] Ambient temperature and humidity further affect the comfort of the human eye in response to lighting. At higher temperatures, cool white light helps reduce feelings of stuffiness; cool white light, for example, is 5000K-6500K (color temperature, measured in Kelvin). Conversely, in low-temperature environments, warm light enhances the feeling of warmth; warm light, for example, is 2700K-3500K. At higher humidity, increased air scattering of light may lead to a decrease in perceived brightness under the same light source, necessitating adjustments to lighting parameters. By collecting this environmental data, intelligent lighting systems can comprehensively analyze spatial characteristics and environmental changes, adaptively adjusting light intensity, color temperature, and distribution to optimize lighting effects and improve visual comfort.

[0132] The second processing module analyzes and processes the space environment based on the space environment data to obtain space environment characteristic data.

[0133] The method for obtaining the spatial environment feature data includes:

[0134] Obtain wall material, wall color, floor material, ground color, ambient temperature, ambient humidity, and ambient brightness at N sampling points from the spatial environment data;

[0135] The uniformity of the lighting environment is calculated based on the ambient brightness of N sampling points.

[0136] Pre-build a material property data table; obtain the corresponding reflectivity and absorptivity from the material property data table based on the wall material and wall color, and record them as wall reflectivity and wall absorptivity, respectively; obtain the corresponding reflectivity and absorptivity from the material property data table based on the floor material and floor color, and record them as floor reflectivity and floor absorptivity, respectively;

[0137] Spatial environmental characteristic data are constructed by considering factors such as light environment uniformity, wall reflectivity, wall absorptivity, floor reflectivity, floor absorptivity, ambient temperature, and ambient humidity.

[0138] The method for calculating the uniformity of the lighting environment includes:

[0139] ;

[0140] in, For the balance of the lighting environment, The maximum ambient brightness among the N sampling points. The minimum ambient brightness among the N sampling points. Indicates the average ambient brightness. Let represent the ambient brightness at the j-th sampling point. The closer the ambient brightness uniformity is to 1, the more uniform the ambient brightness distribution in the space. The further the ambient brightness uniformity is from 1, the more uneven the ambient brightness distribution in the space. If only the average ambient brightness is calculated, it cannot reflect the local unevenness of illumination.

[0141] The material properties data are shown in Table 1:

[0142] Table 1 Material Property Data Table

[0143]

[0144] The third acquisition module is used to collect key data on the user's wakefulness after waking up. The key data on wakefulness includes the user's sleep state when awake, the time of awakening, and the time when the lights are turned on. The user's sleep state when awake includes awakening during deep sleep, awakening during light sleep, and awakening during REM sleep.

[0145] It should be noted that collecting data on a user's sleep state while awake helps the intelligent lighting system to make adaptive adjustments.

[0146] During REM sleep, users exhibit higher levels of neural activity and are more sensitive to external light, especially blue light. Therefore, in this situation, lighting adjustments should prioritize optimizing color temperature parameters to avoid stimulating the retina with high color temperature light sources. At the same time, dynamic brightness adjustment should be combined to ensure a smooth transition of light.

[0147] By collecting data on users' sleep patterns while awake and combining this data with the physiological characteristics of different awake stages, the system can adaptively optimize lighting parameters, including brightness, color temperature, and adjustment rate. This can effectively reduce discomfort from morning light, improve users' morning comfort, and enhance the ergonomic adaptability of intelligent lighting systems, thereby providing a more intelligent and personalized light environment management solution.

[0148] The third processing module is used to extract features from the key data of the conscious state to obtain conscious feature data.

[0149] The method for obtaining the conscious feature data includes:

[0150] Obtain key data on wakefulness, including the user's sleep state when awake, the time of wakefulness, and the time when the lights are turned on.

[0151] Pre-build a matching table of user's sleep state while awake and user's sleep state values ​​while awake; obtain the corresponding sleep state values ​​while awake from the matching table of user's sleep state while awake.

[0152] The light response delay is obtained by subtracting the time when the light is turned on from the time when the person is awake.

[0153] The sleep state values ​​during wakefulness and the light response delay are used to construct wakefulness feature data.

[0154] Table 2 shows the matching table of user's sleep state while awake and the user's sleep state values ​​while awake.

[0155] Table 2. User Sleep Status During Awake - Value Matching Table for User Sleep Status During Awake

[0156] Sleep state when user is awake Sleep state value when user is awake Deep sleep stage awake 1 Light sleep stage awake 2 Rapid eye movement stage awake 3

[0157] It's important to note that turning on the lights immediately after waking up results in a shorter light response time compared to turning them on after a longer buffer period. This leads to a more intense stimulation of the visual system and circadian rhythms. This phenomenon is primarily due to the combined effects of delayed pupillary accommodation, the sensitivity of retinal photoreceptor neurons, and melatonin metabolism. Upon waking, the body is still in the transition from sleep to wakefulness, with pupils initially larger to adapt to low-light environments. If high-intensity light is immediately introduced, the pupils must constrict rapidly, causing momentary glare, visual discomfort, and temporary blurred vision. Furthermore, the intrinsic photoreceptor ganglion cells in the retina are more sensitive to blue light in the morning. Premature exposure to high color temperature light sources may excessively suppress melatonin secretion, accelerating the waking process while potentially disrupting morning circadian rhythms and affecting overall comfort.

[0158] If the user is awake and waits a while before turning on the lights, meaning the light response delay is longer, the visual system has already undergone some dark adaptation to light adaptation. At this time, the light stimulus is more gradual, and the pupil adjustment and nervous system adaptation processes are more natural, reducing discomfort caused by sudden bright light. Therefore, in an intelligent light control system, light parameters can be adaptively adjusted according to different light response delays, thereby optimizing the user's lighting experience.

[0159] The intelligent control module, based on sleep quality characteristic data, spatial environment characteristic data, and wakefulness characteristic data, performs corresponding adaptive adjustments to LED parameters for wakefulness during deep sleep, light sleep, and REM sleep.

[0160] like Figure 4 As shown, the methods for adaptively adjusting LED parameters for wakefulness during deep sleep, light sleep, and REM sleep include:

[0161] Extract sleep state values ​​and light response delay during wakefulness from wakefulness feature data;

[0162] If the sleep state value during wakefulness corresponds to the wakefulness state during deep sleep, then the sleep state value during wakefulness is recorded as the wakefulness value during deep sleep. Based on the wakefulness value during deep sleep, light response delay, spatial environment characteristic data, and sleep quality characteristic data, the light parameters are adaptively adjusted for wakefulness during deep sleep.

[0163] If the sleep state value during wakefulness corresponds to the wakefulness stage of light sleep, then the sleep state value during wakefulness is recorded as the wakefulness value during light sleep. Based on the wakefulness value during light sleep, light response delay, spatial environment characteristic data, and sleep quality characteristic data, the light parameters are adaptively adjusted for wakefulness during light sleep.

[0164] If the sleep state value during wakefulness corresponds to the REM sleep state, then the sleep state value during wakefulness is recorded as the REM sleep state value. Based on the REM sleep state value, light response delay, spatial environment characteristic data, and sleep quality characteristic data, the light parameters are adaptively adjusted for REM sleep state.

[0165] like Figure 5 As shown, methods for adaptively adjusting light parameters to address wakefulness during deep sleep include:

[0166] S100: Input the wakefulness value during deep sleep, light response delay, spatial environment characteristic data and sleep quality characteristic data into the deep sleep parameter setting model to obtain deep sleep adjustment parameters. The deep sleep adjustment parameters include the initial value of light intensity during wakefulness in deep sleep, the target value of light intensity, the initial value of color temperature, the target value of color temperature, the total adjustment time, the adjustment interval duration, the brightness adjustment rate and the color temperature adjustment rate.

[0167] S101: Calculate the total adjustment time during deep sleep and wakefulness by dividing the adjustment interval to obtain the corresponding number of adaptive adjustments, denoted as... The initial value of ycs is set to 1, and the range of ycs values ​​is from 1 to... ;

[0168] S102: The ycs-th adaptively adjusted illuminance is calculated based on the initial value of illuminance during deep sleep, the target value of illuminance, and the brightness adjustment rate, and is denoted as the first adjusted illuminance.

[0169] The color temperature of the ycsth adaptive adjustment is calculated based on the initial color temperature value, target color temperature value, and color temperature adjustment rate during the deep sleep stage. This is denoted as the first adjustment color temperature.

[0170] S103: When the adjustment interval between deep sleep and wakefulness is reached, perform the ycs-th adaptive adjustment, adjust the illuminance to the first adjusted illuminance, and adjust the color temperature to the first adjusted color temperature;

[0171] S104: Let ycs = ycs + 1. If ycs is less than or equal to 1... If ycs is greater than 0, then continue executing S102 to S103. Then, the adaptive adjustment of light parameters for awakening during deep sleep is completed, and the current process ends.

[0172] The first method for calculating adjusted illumination includes:

[0173] ;

[0174] in, The first adjustment of illumination intensity is the adaptive adjustment for the ycs-th time. The first adjustment of illumination is the adaptive adjustment for the ycs-1th time. The target value for light intensity during deep sleep is the level of light when the person is awake. This represents the brightness regulation rate during deep sleep; when ycs=1, ycs-1 is 0. This represents the initial value of the illumination. It has a dynamic adjustment function; as the number of adjustments (ycs) increases, Increase accordingly, Correspondingly, the brightness is reduced to achieve a smooth and gradual adaptation process for waking from deep sleep, avoiding abrupt changes in light that may cause discomfort to the user. This results in high adaptability and improves user comfort of the intelligent lighting control system. Waking from deep sleep is usually accompanied by high sleep inertia, and users have a lower ability to adapt to the external environment after waking up during this stage. The visual system is more sensitive to changes in light. If the light intensity changes too quickly or the brightness is too high, it may cause visual discomfort or even temporary glare. Therefore, when the system detects that a user is waking up from deep sleep, a gradual brightness increase strategy should be adopted, with a low initial brightness and a slow light enhancement process to guide the user to adapt.

[0175] The first method for calculating the adjusted color temperature includes:

[0176] ;

[0177] in, This is the first color temperature adjustment for the ycs-th adaptive adjustment. This is the first color temperature adjustment for the ycs-1th adaptive adjustment. The target color temperature value for wakefulness during deep sleep. This represents the color temperature adjustment rate during wakefulness in the deep sleep stage; when ycs=1, ycs-1 is 0, at which point... This indicates the initial color temperature value. It has a dynamic adjustment function; as the number of adjustments (ycs) increases, Increase accordingly, Correspondingly, this reduces the color temperature, thus achieving a smooth and gradual adaptation process for waking up during deep sleep, avoiding abrupt color temperature changes that may cause discomfort to the user, thereby possessing high adaptability and improving the user comfort of the intelligent lighting control system.

[0178] The training method for the deep sleep parameter setting model includes:

[0179] A deep sleep parameter setting dataset is pre-collected. This dataset includes B sets of deep sleep parameter setting data and corresponding deep sleep regulation parameters, where B is a positive integer greater than 0. The deep sleep parameter setting data includes wakefulness values ​​during deep sleep, light response delay, spatial environment characteristic data, and sleep quality characteristic data. The deep sleep regulation parameters are set by those skilled in the art based on the wakefulness values ​​during deep sleep, light response delay, spatial environment characteristic data, and sleep quality characteristic data. The deep sleep parameter setting dataset is divided into a training set and a validation set. The training set is used to train the deep sleep parameter setting model, and the validation set is used to evaluate the generalization performance of the deep sleep parameter setting model.

[0180] During the training of the deep sleep parameter setting model, minimizing the cross-entropy loss function is used as the optimization objective. An early stopping strategy is used to monitor the performance of the validation set. The model performance is optimized by continuously adjusting the network parameters. When the prediction accuracy on the validation set reaches the expected accuracy, it is determined that the deep sleep parameter setting model has converged and training is stopped. The deep sleep parameter setting model is trained using a deep neural network based on a multilayer perceptron.

[0181] The deep sleep parameter setting data is converted into feature vectors. The input layer of the deep sleep parameter setting model receives the feature vectors and extracts the nonlinear relationships in the data through the hidden layer. Finally, the output layer of the deep sleep parameter setting model calculates the probability distribution of the deep sleep regulation parameters through the softmax activation function and outputs the deep sleep regulation parameter corresponding to the highest probability as the final prediction result.

[0182] For example, in the technical solution of the present invention, in the wakefulness characteristic data, the wakefulness value during deep sleep is 1, and the light response delay is 10s; in the spatial environment characteristic data, the light environment uniformity is 0.90, the wall reflectivity is 0.87, the wall absorptivity is 0.23, the floor reflectivity is 0.55, the floor absorptivity is 0.45, the ambient temperature is 24℃, and the ambient humidity is 50%; in the sleep quality characteristic data, the deep sleep ratio is 0.24, the light sleep ratio is 0.51, the REM sleep ratio is 0.25, the sleep stability index is 0.90, the circadian rhythm matching index is 0.04, and the sleep quality assessment score is 90.

[0183] The deep sleep parameter setting model outputs the following deep sleep adjustment parameters: initial illuminance value is 15% of rated brightness, target illuminance value is 80% of rated brightness, initial color temperature value is 2700K, target color temperature value is 5200K, total adjustment time is 1200s, adjustment interval is 60s, brightness adjustment rate is 0.18, and color temperature adjustment rate is 0.12.

[0184] It should be noted that the deep sleep parameter setting model in this invention is used to automatically match a set of gentle and gradual light adjustment schemes when a user is awakened from deep sleep. Specifically, the awakening value during deep sleep, light response delay, spatial environment feature data, and sleep quality feature data are combined into a deep sleep parameter setting feature vector input to the model. The spatial environment feature data is used to characterize the current optical and thermal environment conditions of the bedroom, and the sleep quality feature data is used to reflect information such as the proportion of deep sleep and sleep stability the previous night, thereby comprehensively characterizing the user's physiological sensitivity and adaptability when awakened. The deep sleep parameter setting model uses a deep neural network based on a multilayer perceptron. The above feature vector is input into the input layer of the model, and the nonlinear coupling relationship between deep sleep state, spatial environment, and light comfort is extracted through multiple hidden layers. The output layer obtains the probability distribution of different deep sleep adjustment parameter combinations through a softmax activation function, and selects the set of initial light intensity, target light intensity, initial color temperature, target color temperature, total adjustment time, adjustment interval duration, and brightness and color temperature adjustment rate with the highest probability as the target deep sleep adjustment parameters for the current scenario.

[0185] In practical applications, when a user is detected to be awakened during deep sleep and has low sleep quality, the model tends to output a combination of parameters with a lower initial brightness value, a longer total adjustment time, and a smaller adjustment rate, so that the illumination increases slowly over a longer period of time and gradually increases the color temperature. When the user has good sleep quality and a shorter illumination response delay, the model can output a combination of brightness and color temperature increases slightly faster, thereby accelerating the user's awakening state without producing strong glare or circadian rhythm disturbances, and achieving personalized adaptive control of LED parameters for awakening scenarios during deep sleep.

[0186] Methods for adaptively adjusting light parameters to address wakefulness during light sleep include:

[0187] S200: Input the wakefulness value during the light sleep stage, the light response delay, the spatial environment characteristic data, and the sleep quality characteristic data into the light sleep parameter setting model to obtain the light sleep adjustment parameters. The light sleep adjustment parameters include the initial value of light intensity during the wakefulness stage of light sleep, the target value of light intensity, the initial value of color temperature, the target value of color temperature, the total adjustment time, the adjustment interval duration, the brightness adjustment rate, and the color temperature adjustment rate.

[0188] S201: Calculate the total adjustment time during the light sleep stage by dividing the adjustment interval, and obtain the corresponding number of adaptive adjustments, denoted as... The default initial value of ECS is 1, and the value of ECS ranges from 1 to... ;

[0189] S202: The light intensity of the first adaptive adjustment is calculated based on the initial light intensity value, the target light intensity value and the brightness adjustment rate during the light sleep stage, and is denoted as the second adjusted light intensity.

[0190] The color temperature of the first adaptive adjustment is calculated based on the initial color temperature value, target color temperature value, and color temperature adjustment rate during the light sleep stage, and is denoted as the second adjustment color temperature.

[0191] S203: When the adjustment interval for awakening during the light sleep stage is reached, the first ECS adaptive adjustment is performed, adjusting the illuminance to the second adjusted illuminance and the color temperature to the second adjusted color temperature;

[0192] S204: Let ecs = ecs + 1. If ecs is less than or equal to 1, then... If S202 to S203 are executed, then if ECS is greater than... Then, the adaptive adjustment of light parameters for awakening during the light sleep stage is completed, and the current process ends.

[0193] The second method for calculating adjusted light intensity includes:

[0194] ;

[0195] in, The second adjustment of illumination is the adaptive adjustment of the ecs-th time. This represents the initial light intensity value for wakefulness during the light sleep stage. The target light intensity value for wakefulness during the light sleep stage. The brightness adjustment rate during wakefulness in the light sleep stage; It has a linear regulating effect; as the number of adjustments (ECS) increases, The light intensity increases linearly, allowing for a smooth and gradual adaptation process to light levels during the light sleep phase. This avoids abrupt changes in light that could cause discomfort to the user, thus providing high adaptability and improving user comfort of the intelligent lighting control system. Waking up during the light sleep phase typically means the user's physiological state is close to wakefulness, with less sleep inertia. At this time, the user's ability to adapt to light is stronger, allowing for linear adjustment of light intensity. The light adjustment rate can be appropriately increased to provide sufficient ambient brightness, helping the user quickly enter a state of wakefulness.

[0196] The second method for calculating color temperature adjustment includes:

[0197] ;

[0198] in, This is the second color temperature adjustment in the ecs adaptive adjustment. This represents the initial color temperature value during the light sleep stage when the person is awake. The target color temperature value for wakefulness during the light sleep stage. The rate of color temperature adjustment during wakefulness in the light sleep stage; It has a linear regulating effect; as the number of adjustments (ECS) increases, The color temperature increases linearly, thus achieving a smooth and gradual adaptation process for awakening during light sleep, avoiding abrupt color temperature changes that may cause discomfort to the user, thereby possessing high adaptability and improving the user comfort of the intelligent lighting control system.

[0199] The training method for the light sleep parameter setting model includes:

[0200] A light sleep parameter setting dataset is pre-collected. This dataset includes C sets of light sleep parameter setting data and corresponding light sleep regulation parameters, where C is a positive integer greater than 0. The light sleep parameter setting data includes wakefulness values ​​during the light sleep stage, light response delay, spatial environment characteristic data, and sleep quality characteristic data. The light sleep regulation parameters are set by those skilled in the art based on the wakefulness values ​​during the light sleep stage, light response delay, spatial environment characteristic data, and sleep quality characteristic data. The light sleep parameter setting dataset is divided into a training set and a validation set. The training set is used to train the light sleep parameter setting model, and the validation set is used to evaluate the generalization performance of the light sleep parameter setting model.

[0201] During the training of the shallow sleep parameter setting model, minimizing the cross-entropy loss function is used as the optimization objective. An early stopping strategy is used to monitor the performance on the validation set. The model performance is optimized by continuously adjusting the network parameters. When the prediction accuracy on the validation set reaches the expected accuracy, it is determined that the shallow sleep parameter setting model has converged and training is stopped. The shallow sleep parameter setting model is trained using a deep neural network based on a multilayer perceptron.

[0202] The light sleep parameter setting data is converted into a feature vector. The input layer of the light sleep parameter setting model receives the feature vector and extracts the nonlinear relationship in the data through the hidden layer. Finally, the output layer of the light sleep parameter setting model calculates the probability distribution of the light sleep adjustment parameter through the softmax activation function and outputs the light sleep adjustment parameter corresponding to the highest probability as the final prediction result.

[0203] For example, in the technical solution of the present invention, in the wakefulness characteristic data, the wakefulness value during the light sleep stage is 2, and the light response delay is 5s, indicating that the user turns on the lights almost as soon as they wake up; in the spatial environment characteristic data, the light environment uniformity is 0.80, the wall reflectivity is 0.75, the wall absorptivity is 0.25, the floor reflectivity is 0.40, the floor absorptivity is 0.60, the ambient temperature is 26℃, and the ambient humidity is 60%; in the sleep quality characteristic data, the deep sleep ratio is 0.18, the light sleep ratio is 0.57, the REM sleep ratio is 0.25, the sleep stability index is 0.65, the circadian rhythm matching index is 0.15, and the sleep quality assessment score is 72.

[0204] The example of the light sleep adjustment parameters output by the light sleep parameter setting model is as follows: the initial value of the illuminance is 30% of the rated illuminance, the target value of the illuminance is 90% of the rated illuminance, the initial value of the color temperature is 3000K, the target value of the color temperature is 6000K, the total adjustment time is 600s, the adjustment interval is 30s, the brightness adjustment rate is 0.30, and the color temperature adjustment rate is 0.25.

[0205] It should be noted that the light sleep parameter setting model in this invention is used to automatically match a set of stable and relatively fast light adjustment schemes when the user is detected to be awake during a light sleep stage. Specifically, the light sleep stage awakening value, light response delay and spatial environment characteristic data, and sleep quality characteristic data are combined into a light sleep parameter setting feature vector, which is then input into the light sleep parameter setting model. The model outputs the initial light intensity value, target light intensity value, initial color temperature value, target color temperature value, total adjustment time, adjustment interval duration, and light sleep adjustment parameters such as brightness and color temperature adjustment rate, which are used to drive the subsequent adaptive light control process.

[0206] In practical applications, when a user awakens during a light sleep phase, the initial brightness value, target brightness value, and adjustment rate output by the light sleep parameter setting model are used to construct a brightness change curve that increases linearly with the number of adjustments. Simultaneously, the color temperature also gradually increases linearly, causing the second adjusted brightness and second adjusted color temperature to gradually approach the target values ​​with each adaptive adjustment, achieving a smooth and gradual brightness and color temperature adaptation process. Considering that awakening during a light sleep phase usually means the user's physiological state is close to wakefulness and sleep inertia is relatively small, the light sleep parameter setting model tends to select a combination with a shorter total adjustment time and a appropriately increased adjustment rate in the parameter space. This is to provide sufficient ambient brightness more quickly while avoiding abrupt changes in illumination and visual discomfort, helping the user smoothly and comfortably transition to a waking state. This effectively supports the LED parameter adaptive control technology solution of this invention for light sleep awakening scenarios.

[0207] Methods for adaptively adjusting illumination parameters during REM (Rapid Eye Movement) awake states include:

[0208] S300: Input the wakefulness value, light response delay, spatial environment characteristic data and sleep quality characteristic data of the rapid eye movement (REM) stage into the REM parameter setting model to obtain the REM accommodation parameters. The REM accommodation parameters include the initial value of light intensity, the target value of light intensity, the initial value of color temperature, the target value of color temperature, the total accommodation time, the accommodation interval duration, the brightness adjustment rate and the color temperature adjustment rate during the REM stage wakefulness.

[0209] S301: The total accommodation time during the REM (Rapid Eye Movement) awake phase is divided by the accommodation interval to obtain the corresponding number of adaptive accommodation cycles, denoted as... The default initial value of scs is 1, and the value range of scs is from 1 to 1. ;

[0210] S302: The ssth adaptive adjustment of the illumination is calculated based on the initial value of illumination during the rapid eye movement phase, the target value of illumination, and the brightness adjustment rate. This is denoted as the third adjustment of illumination.

[0211] The color temperature of the scs-th adaptive adjustment is calculated based on the initial color temperature value, target color temperature value, and color temperature adjustment rate during the REM (Rapid Eye Movement) phase of wakefulness, and is denoted as the third adjustment color temperature.

[0212] S303: When the accommodation interval of the REM phase of wakefulness is reached, the scs-th adaptive adjustment is performed, adjusting the illumination to the third adjustment illumination and the color temperature to the third adjustment color temperature;

[0213] S304: Let scs = scs + 1. If scs is less than or equal to 1, then... If S302 to S303 are executed, then continue execution. If scs is greater than... Then, the adaptive adjustment of illumination parameters for the REM (Rapid Eye Movement) awake phase is completed, and the current process ends.

[0214] The third method for calculating the adjusted illumination intensity includes:

[0215] ;

[0216] in, The third adjustment of illumination intensity is the scs-th adaptive adjustment. This represents the initial light intensity value during the REM (Rapid Eye Movement) phase when the person is awake. The target brightness value for awake light during the REM (Rapid Eye Movement) phase. The rate of brightness accommodation during the REM (Rapid Eye Movement) phase of wakefulness. It is a constant; as the number of adjustments increases, The light intensity increases exponentially, ensuring that the change in light intensity is gradual at the initial moment, and then gradually approaches the target light intensity value in an exponential manner, thereby achieving gradual light adaptation and reducing the sensitivity to sudden light stimulation during the REM sleep phase.

[0217] The calculation method for the third color temperature adjustment includes:

[0218] ;

[0219] in, This is the third color temperature adjustment in the scs-th adaptive adjustment. This is the initial color temperature value for wakefulness during the REM (Rapid Eye Movement) phase. The target color temperature value for wakefulness during REM (Rapid Eye Movement) phase. The rate of color temperature adjustment during the rapid eye movement (REM) phase of wakefulness. The adaptive accommodation number during the rapid eye movement (REM) phase is calculated. As the number of adjustments (scs) increases, the formula for the third adjustment of color temperature increases exponentially, ensuring that the initial color temperature adjustment is gradual to reduce the interference of spectral abrupt changes on vision and physiological rhythms. Subsequently, the adjustment accelerates over time to reach the target color temperature value, achieving a smooth transition of color temperature.

[0220] It's important to note that wakefulness during REM (Rapid Eye Movement) sleep is typically accompanied by high neural excitability, making users highly sensitive to external stimuli. Therefore, direct exposure to strong light during this wakeful state can cause discomfort and even affect emotional stability. In terms of light intensity adjustment strategies, the initial brightness should be set at a low level, with a gradually increasing light pattern to reduce the impact of sudden light stimulation on the retina and nervous system. Regarding color temperature adjustment, warm white light close to natural morning light, such as 3500K-4500K, should be prioritized to reduce the inhibitory effect of blue light on melatonin secretion. After adaptation, a gradual transition to neutral white light, such as 5000K-5500K, should be initiated to promote a smooth transition to wakefulness. This effectively reduces photosensitivity after wakefulness during REM sleep, improves individual comfort with the light environment, and optimizes the naturalness of the wakefulness transition process.

[0221] The training method for the fast eye-tracking parameter setting model includes:

[0222] A fast eye movement (FEM) parameter setting dataset is pre-collected. This dataset includes D sets of FEM parameter setting data and corresponding FEM accommodation parameters, where D is a positive integer greater than 0. The FEM parameter setting data includes wakefulness values, light response delay, spatial environment characteristics, and sleep quality characteristics during the FEM phase. The FEM accommodation parameters are set by those skilled in the art based on these data. The FEM parameter setting dataset is divided into a training set and a validation set. The training set is used to train the FEM parameter setting model, and the validation set is used to evaluate the generalization performance of the FEM parameter setting model.

[0223] During the training of the fast eye-tracking parameter setting model, minimizing the cross-entropy loss function is used as the optimization objective. An early stopping strategy is used to monitor the performance on the validation set. The model performance is optimized by continuously adjusting the network parameters. When the prediction accuracy on the validation set reaches the expected accuracy, the fast eye-tracking parameter setting model is considered to have converged and training is stopped. The fast eye-tracking parameter setting model is trained using a deep neural network based on a multilayer perceptron.

[0224] The fast eye movement (FEM) parameter setting data is converted into feature vectors. The input layer of the FEM parameter setting model receives the feature vectors and extracts the nonlinear relationships in the data through the hidden layer. Finally, the output layer of the FEM parameter setting model calculates the probability distribution of the FEM accommodation parameters through the softmax activation function and outputs the FEM accommodation parameter corresponding to the highest probability as the final prediction result.

[0225] For example, in the technical solution of the present invention, in the wakefulness characteristic data, the wakefulness value during the rapid eye movement (REM) phase is 3, and the light response delay is 25s, indicating that the user wakes up and then waits for a buffer period before turning on the lights; in the spatial environment characteristic data, the light environment uniformity is 0.88, the wall reflectivity is 0.30, the wall absorptivity is 0.70, the floor reflectivity is 0.75, the floor absorptivity is 0.25, the ambient temperature is 22℃, and the ambient humidity is 55%; in the sleep quality characteristic data, the deep sleep ratio is 0.16, the light sleep ratio is 0.49, the REM sleep ratio is 0.35, the sleep stability index is 0.55, the circadian rhythm matching index is 0.22, and the sleep quality assessment score is 65.

[0226] Examples of fast eye-tracking accommodation parameters output by the fast eye-tracking parameter setting model are as follows: initial illumination value is 20% of rated brightness, target illumination value is 60% of rated brightness, initial color temperature value is 3000K, target color temperature value is 4500K, total accommodation time is 900s, accommodation interval is 45s, brightness adjustment rate is 0.22, and color temperature adjustment rate is 0.15.

[0227] It should be noted that the REM (Rapid Eye Movement) parameter setting model in this invention is mainly used to generate a set of illumination adjustment parameters that focus more on suppressing glare stimulation and rhythm disturbances when the user is detected to be awake during the REM phase. Specifically, the awakening value during the REM phase, the illumination response delay, and spatial environment feature data are combined with sleep quality feature data to form a REM parameter setting feature vector, which is input into the REM parameter setting model. The model outputs the initial illumination value, target illumination value, initial color temperature value, target color temperature value, total adjustment time, adjustment interval duration, and light sleep adjustment parameters such as brightness and color temperature adjustment rate during the awakening phase of the REM phase, which are used to drive the subsequent adaptive illumination control process.

[0228] In practical applications, when a user is detected to be awake during REM sleep with a high proportion of REM activity and poor sleep stability, the REM parameter setting model tends to output a combination of parameters with a low initial brightness value, moderate target brightness, color temperature controlled in the low to medium range, longer total adjustment time, and a smaller adjustment rate. By slowly increasing brightness and gently adjusting color temperature in small steps, the model reduces the stimulation of the retina and circadian rhythm caused by high color temperature blue light and instantaneous changes in strong light. When sleep quality is relatively good and response latency is short, the REM parameter setting model can moderately increase the adjustment rate and target brightness to accelerate the awakening process without causing significant discomfort. This enables personalized adaptive control of LED parameters for awakening scenarios during REM sleep, supporting the implementation of the overall technical solution of this invention.

[0229] It should be noted that wakefulness during deep sleep, light sleep, and REM sleep corresponds to different neurophysiological characteristics and sensitivities to light stimulation. Therefore, to achieve precise light environment regulation tailored to an individual's wakefulness state, it is necessary to comprehensively consider the light response delay, spatial environment characteristics, and sleep quality characteristics corresponding to the sleep state during wakefulness, and construct adaptive light regulation strategies for different wakefulness states. This ensures that the changes in light parameters match the user's neurophysiological adaptability, reduces the disturbance of physiological rhythms caused by sudden light exposure, and improves the naturalness and comfort of the wakefulness transition. This enables customized light regulation based on the individual user's state, improving user comfort in intelligent light control systems.

[0230] Example 2:

[0231] Please see Figure 3 As shown, this embodiment provides an LED parameter adaptive control system based on a neural network, which also includes:

[0232] The fourth data acquisition module is used to collect the user's normal daily wake-up time window; for example, the user's normal daily wake-up time window is from 7:00 am to 8:00 am.

[0233] The fourth processing module processes the awakening scenario based on the awakening time and the user's normal daily wake-up time window to obtain awakening scenario categories, which include waking up in the middle of the night and waking up at normal rest time.

[0234] The methods for obtaining the sober scene category include:

[0235] If the time of wakefulness falls within the user's normal daily wake-up time window, then the wakefulness scenario category is wake-up at normal rest time;

[0236] If the time when a user is awake does not fall within the user's normal wake-up time window, then the awakening scenario is categorized as waking up in the middle of the night.

[0237] It should be noted that the waking scenario categories can be represented in digital form, for example, marking waking up in the middle of the night as 1 and waking up at the normal sleep schedule as 2.

[0238] The model optimization module is used to take the awake scene category as a supplementary input feature for the deep sleep parameter setting model, the light sleep parameter setting model, and the rapid eye movement parameter setting model, and to train and optimize the deep sleep parameter setting model, the light sleep parameter setting model, and the rapid eye movement parameter setting model based on the awake scene category, thereby enhancing the model's adaptability to the light requirements of different awake states.

[0239] To improve the LED parameter adaptive control system's ability to accurately match individual physiological states, and to achieve more intelligent and personalized adaptive adjustment of the light environment.

[0240] It should be noted that by introducing wakefulness scenario categories, the model can further distinguish the different impacts of waking up in the middle of the night versus waking up at normal sleep times on lighting adjustment strategies, making the lighting parameter settings more targeted during deep sleep, light sleep, and REM sleep stages. This further optimizes the adaptive capability of the LED parameter adaptive control system, ensuring that lighting adjustment strategies in different scenarios meet the user's physiological adaptation needs, and improving the adaptability and adjustment accuracy of the intelligent lighting control system to different wakefulness states. This solution not only optimizes lighting comfort when waking up at night but also provides more targeted lighting adjustment schemes when waking up in the morning, promoting a faster awakening state for users, improving the matching degree of the light environment to the human body's biological rhythm, thereby enhancing the overall user experience and the intelligence level of the lighting system.

[0241] Example 3:

[0242] Please see Figure 2 As shown, this embodiment provides an adaptive control method for LED parameters based on neural networks, including:

[0243] Collect sleep quality data; the sleep quality data includes total deep sleep time, total light sleep time, total REM sleep time, number of awakenings, and the awakening time corresponding to each awakening;

[0244] Sleep quality analysis is performed based on sleep quality data to obtain sleep quality characteristic data, which includes the proportion of deep sleep, the proportion of light sleep, the proportion of rapid eye movement (REM) sleep, sleep stability index, circadian rhythm matching index, and sleep quality assessment score.

[0245] Collect spatial environment data; the spatial environment data includes wall material, wall color, floor material, ground color, ambient temperature, ambient humidity, and ambient brightness at N sampling points;

[0246] Based on spatial environment data, spatial environment characteristic data is obtained by analyzing and processing the spatial environment data; the spatial environment characteristic data includes the uniformity of illumination environment, wall reflectivity, wall absorptivity, floor reflectivity, floor absorptivity, ambient temperature and ambient humidity;

[0247] Collect key data on the user's wakefulness after they have regained consciousness; the key data on wakefulness includes the user's sleep state when awake, the moment of awakening, and the moment the lights were turned on.

[0248] Feature extraction is performed on key data of wakefulness to obtain wakefulness feature data; the wakefulness feature data includes sleep state values ​​and light response delay during wakefulness.

[0249] Based on sleep quality characteristic data, spatial environment characteristic data, and wakefulness characteristic data, corresponding LED parameters are adaptively adjusted for wakefulness during deep sleep, light sleep, and REM sleep.

[0250] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0251] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A neural network-based adaptive control method for LED parameters, characterized in that, include: Collect sleep quality data; The sleep quality data includes total deep sleep time, total light sleep time, total REM sleep time, number of awakenings, and the duration of each awakening. Sleep quality analysis is performed based on sleep quality data to obtain sleep quality characteristic data; The sleep quality characteristics data include the proportion of deep sleep, the proportion of light sleep, the proportion of rapid eye movement (REM) sleep, the sleep stability index, the circadian rhythm matching index, and the sleep quality assessment score. Collect spatial environment data; the spatial environment data includes wall material, wall color, floor material, ground color, ambient temperature, ambient humidity, and ambient brightness at N sampling points; Based on spatial environment data, spatial environment characteristic data is obtained by analyzing and processing the spatial environment data; the spatial environment characteristic data includes the uniformity of illumination environment, wall reflectivity, wall absorptivity, floor reflectivity, floor absorptivity, ambient temperature and ambient humidity; Collect key data on the user's wakefulness after they have regained consciousness; the key data on wakefulness includes the user's sleep state when awake, the moment of awakening, and the moment the lights were turned on. Feature extraction is performed on key data of wakefulness to obtain wakefulness feature data; the wakefulness feature data includes sleep state values ​​and light response delay during wakefulness. Based on sleep quality characteristic data, spatial environment characteristic data, and wakefulness characteristic data, corresponding LED parameters are adaptively adjusted for wakefulness during deep sleep, light sleep, and REM sleep.

2. The LED parameter adaptive control method based on neural networks according to claim 1, characterized in that, The methods for adaptively adjusting LED parameters to address wakefulness during deep sleep, light sleep, and REM sleep stages include: Extract sleep state values ​​and light response delay during wakefulness from wakefulness feature data; If the sleep state value during wakefulness corresponds to the wakefulness state during deep sleep, then the sleep state value during wakefulness is recorded as the wakefulness value during deep sleep. Based on the wakefulness value during deep sleep, light response delay, spatial environment characteristic data, and sleep quality characteristic data, the light parameters are adaptively adjusted for wakefulness during deep sleep. If the sleep state value during wakefulness corresponds to the wakefulness stage of light sleep, then the sleep state value during wakefulness is recorded as the wakefulness value during light sleep. Based on the wakefulness value during light sleep, light response delay, spatial environment characteristic data, and sleep quality characteristic data, the light parameters are adaptively adjusted for wakefulness during light sleep. If the sleep state value during wakefulness corresponds to the REM sleep state, then the sleep state value during wakefulness is recorded as the REM sleep state value. Based on the REM sleep state value, light response delay, spatial environment characteristic data, and sleep quality characteristic data, the light parameters are adaptively adjusted for REM sleep state.

3. The LED parameter adaptive control method based on neural networks according to claim 2, characterized in that, Methods for adaptively adjusting light parameters to address wakefulness during deep sleep include: S100: Input the wakefulness value during deep sleep, light response delay, spatial environment characteristic data and sleep quality characteristic data into the deep sleep parameter setting model to obtain deep sleep adjustment parameters. The deep sleep adjustment parameters include the initial value of light intensity during wakefulness in deep sleep, the target value of light intensity, the initial value of color temperature, the target value of color temperature, the total adjustment time, the adjustment interval duration, the brightness adjustment rate and the color temperature adjustment rate. S101: Calculate the total adjustment time during deep sleep and wakefulness by dividing the adjustment interval to obtain the corresponding number of adaptive adjustments, denoted as... The initial value of ycs is set to 1, and the range of ycs values ​​is from 1 to... ; S102: The ycs-th adaptively adjusted illuminance is calculated based on the initial value of illuminance during deep sleep, the target value of illuminance, and the brightness adjustment rate, and is denoted as the first adjusted illuminance. The color temperature of the ycsth adaptive adjustment is calculated based on the initial color temperature value, target color temperature value, and color temperature adjustment rate during the deep sleep stage. This is denoted as the first adjustment color temperature. S103: When the adjustment interval between deep sleep and wakefulness is reached, perform the ycs-th adaptive adjustment, adjust the illuminance to the first adjusted illuminance, and adjust the color temperature to the first adjusted color temperature; S104: Let ycs = ycs + 1. If ycs is less than or equal to 1... If ycs is greater than 0, then continue executing S102 to S103. Then, the adaptive adjustment of light parameters for awakening during deep sleep is completed, and the current process ends.

4. The LED parameter adaptive control method based on neural networks according to claim 2, characterized in that, Methods for adaptively adjusting light parameters to address wakefulness during light sleep include: S200: Input the wakefulness value during the light sleep stage, the light response delay, the spatial environment characteristic data, and the sleep quality characteristic data into the light sleep parameter setting model to obtain the light sleep adjustment parameters. The light sleep adjustment parameters include the initial value of light intensity during the wakefulness stage of light sleep, the target value of light intensity, the initial value of color temperature, the target value of color temperature, the total adjustment time, the adjustment interval duration, the brightness adjustment rate, and the color temperature adjustment rate. S201: Calculate the total adjustment time during the light sleep stage by dividing the adjustment interval, and obtain the corresponding number of adaptive adjustments, denoted as... The default initial value of ECS is 1, and the value of ECS ranges from 1 to... ; S202: The light intensity of the first adaptive adjustment is calculated based on the initial light intensity value, the target light intensity value and the brightness adjustment rate during the light sleep stage, and is denoted as the second adjusted light intensity. The color temperature of the first adaptive adjustment is calculated based on the initial color temperature value, target color temperature value, and color temperature adjustment rate during the light sleep stage, and is denoted as the second adjustment color temperature. S203: When the adjustment interval for awakening during the light sleep stage is reached, the first ECS adaptive adjustment is performed, adjusting the illuminance to the second adjusted illuminance and the color temperature to the second adjusted color temperature; S204: Let ecs = ecs + 1. If ecs is less than or equal to 1, then... If S202 to S203 are executed, then if ECS is greater than... Then, the adaptive adjustment of light parameters for awakening during the light sleep stage is completed, and the current process ends.

5. The LED parameter adaptive control method based on neural networks according to claim 2, characterized in that, Methods for adaptively adjusting illumination parameters during REM (Rapid Eye Movement) awake states include: S300: Input the wakefulness value, light response delay, spatial environment characteristic data and sleep quality characteristic data of the rapid eye movement (REM) stage into the REM parameter setting model to obtain the REM accommodation parameters. The REM accommodation parameters include the initial value of light intensity, the target value of light intensity, the initial value of color temperature, the target value of color temperature, the total accommodation time, the accommodation interval duration, the brightness adjustment rate and the color temperature adjustment rate during the REM stage wakefulness. S301: The total accommodation time during the REM (Rapid Eye Movement) awake phase is divided by the accommodation interval to obtain the corresponding number of adaptive accommodation cycles, denoted as... The default initial value of scs is 1, and the value range of scs is from 1 to 1. ; S302: The ssth adaptive adjustment of the illumination is calculated based on the initial value of illumination during the rapid eye movement phase, the target value of illumination, and the brightness adjustment rate. This is denoted as the third adjustment of illumination. The color temperature of the scs-th adaptive adjustment is calculated based on the initial color temperature value, target color temperature value, and color temperature adjustment rate during the REM (Rapid Eye Movement) phase of wakefulness, and is denoted as the third adjustment color temperature. S303: When the accommodation interval of the REM phase of wakefulness is reached, the scs-th adaptive adjustment is performed, adjusting the illumination to the third adjustment illumination and the color temperature to the third adjustment color temperature; S304: Let scs = scs + 1. If scs is less than or equal to 1, then... If S302 to S303 are executed, then continue execution. If scs is greater than... Then, the adaptive adjustment of illumination parameters for the REM (Rapid Eye Movement) awake phase is completed, and the current process ends.

6. The LED parameter adaptive control method based on neural networks according to claim 1, characterized in that, The method for obtaining the sleep quality characteristic data includes: Obtain the total deep sleep time, total light sleep time, total REM sleep time, number of awakenings, and the corresponding awakening time for each awakening from sleep quality data; The total sleep time is calculated based on the total deep sleep time, total light sleep time, total REM sleep time, number of awakenings and the awakening time corresponding to each awakening. The ratio of deep sleep time to total sleep duration is calculated by dividing the total deep sleep time by the total sleep duration; the ratio of light sleep time to total sleep duration is calculated by dividing the total light sleep time by the total sleep duration; and the ratio of REM sleep time to total REM sleep duration is calculated by dividing the total REM sleep time by the total sleep duration. The corresponding sleep stability index is calculated based on the total deep sleep time, total light sleep time, total REM sleep time, number of awakenings and the awakening time corresponding to each awakening. The total sleep time is obtained by summing the total deep sleep time, the total light sleep time, and the total REM sleep time. The circadian rhythm matching index is calculated based on the total deep sleep time, total light sleep time, total REM sleep time, and total sleep time. The proportion of deep sleep, light sleep, rapid eye movement, sleep stability index, and circadian rhythm matching index are input into the sleep quality assessment model to obtain the corresponding sleep quality assessment score. The proportion of deep sleep, light sleep, REM sleep, sleep stability index, circadian rhythm matching index, and sleep quality assessment score were used to construct sleep quality characteristic data.

7. The LED parameter adaptive control method based on neural networks according to claim 1, characterized in that, The method for obtaining the spatial environment feature data includes: Obtain wall material, wall color, floor material, ground color, ambient temperature, ambient humidity, and ambient brightness at N sampling points from the spatial environment data; The uniformity of the lighting environment is calculated based on the ambient brightness of N sampling points. Pre-build a material property data table; obtain the corresponding reflectivity and absorptivity from the material property data table based on the wall material and wall color, and record them as wall reflectivity and wall absorptivity, respectively; obtain the corresponding reflectivity and absorptivity from the material property data table based on the floor material and floor color, and record them as floor reflectivity and floor absorptivity, respectively; Spatial environmental characteristic data are constructed by considering factors such as light environment uniformity, wall reflectivity, wall absorptivity, floor reflectivity, floor absorptivity, ambient temperature, and ambient humidity.

8. The LED parameter adaptive control method based on neural networks according to claim 1, characterized in that, The method for obtaining the conscious feature data includes: Obtain key data on wakefulness, including the user's sleep state when awake, the time of wakefulness, and the time when the lights are turned on. Pre-build a matching table of user's sleep state while awake and user's sleep state values ​​while awake; obtain the corresponding sleep state values ​​while awake from the matching table of user's sleep state while awake. The light response delay is obtained by subtracting the time when the light is turned on from the time when the person is awake. The sleep state values ​​during wakefulness and the light response delay are used to construct wakefulness feature data.

9. The LED parameter adaptive control method based on neural networks according to claim 3, characterized in that, The training method for the deep sleep parameter setting model includes: A deep sleep parameter setting dataset is pre-collected, which includes B group deep sleep parameter setting data and the deep sleep regulation parameters corresponding to the B group deep sleep parameter setting data. The deep sleep parameter setting data includes wakefulness values ​​during deep sleep, light response delay, spatial environment characteristic data, and sleep quality characteristic data. The deep sleep parameter setting dataset is divided into a training set and a validation set, wherein the training set is used to train the deep sleep parameter setting model, and the validation set is used to evaluate the generalization performance of the deep sleep parameter setting model. During the training of the deep sleep parameter setting model, minimizing the cross-entropy loss function is used as the optimization objective. An early stopping strategy is used to monitor the performance on the validation set. The model performance is optimized by continuously adjusting the network parameters. Training is stopped when the prediction accuracy on the validation set reaches the expected accuracy. The deep sleep parameter setting model is trained using a deep neural network based on a multilayer perceptron. The deep sleep parameter setting data is converted into feature vectors; the input layer of the deep sleep parameter setting model receives the feature vectors, extracts the nonlinear relationships in the data through the hidden layer, and finally the output layer of the deep sleep parameter setting model calculates the probability distribution of the deep sleep regulation parameters through the softmax activation function, and outputs the deep sleep regulation parameter corresponding to the highest probability as the final prediction result. Using the same training method as the deep sleep parameter setting model, train the light sleep parameter setting model and the rapid eye movement parameter setting model.

10. A neural network-based adaptive control system for LED parameters, implementing the neural network-based adaptive control method for LED parameters as described in any one of claims 1-9, characterized in that, include: The first acquisition module is used to collect sleep quality data; The first processing module performs sleep quality analysis based on sleep quality data to obtain sleep quality characteristic data. The second acquisition module is used to acquire space environment data; The second processing module analyzes and processes the space environment based on the space environment data to obtain space environment characteristic data; The third data acquisition module is used to collect key data on the user's consciousness after they regain consciousness. The third processing module is used to extract features from the key data of the conscious state to obtain conscious feature data; The intelligent control module, based on sleep quality characteristic data, spatial environment characteristic data, and wakefulness characteristic data, performs corresponding adaptive adjustments to LED parameters for wakefulness during deep sleep, light sleep, and REM sleep.

Images