Learning space-oriented brain-computer interface light environment parameter regulation method

Abstract

This invention provides a brain-computer interface method for regulating optical environment parameters in learning spaces, belonging to the field of visual cognition technology. The method includes: extracting features from measured light output waveforms acquired from the learning space and multimodal physiological signals acquired from the target object to obtain light-specific transient response features and stable cognitive state features related to the learning task; inputting these features into a visual cognitive performance evaluation model to obtain the visual cognitive performance evaluation result of the target object; determining multiple candidate combinations of optical environment parameters based on a set of safety constraints related to the learning task; and selecting a target combination of optical environment parameters from the multiple candidate combinations based on the visual cognitive performance evaluation result, so as to adjust the light source of the learning space using the target combination of optical environment parameters.

Description

Technical Field

[0001] This invention relates to the field of visual cognition technology, and more specifically, to a method for regulating the optical environment parameters of a brain-computer interface for learning spaces. Background Technology

[0002] The learning space is a crucial setting for prolonged reading, writing, memorizing, visual searching, and screen learning. Ambient lighting conditions directly affect learners' visual clarity, attention span, information processing speed, short-term memory performance, and visual fatigue levels.

[0003] Existing lighting systems for learning spaces typically employ fixed illuminance or preset scene modes, such as reading mode, eye protection mode, and rest mode. These systems mostly rely on environmental time, manual selection by the target object, or simple sensor feedback for control, failing to perceive the learner's current visual cognitive state in real time, nor can they adaptively adjust according to the type of learning task, individual differences, and changes in cognitive performance. Summary of the Invention

[0004] In view of this, the present invention provides a method for regulating the optical environment parameters of a brain-computer interface for learning spaces.

[0005] One aspect of the present invention provides a brain-computer interface method for regulating optical environment parameters in a learning space, comprising: during the execution of a learning task by a target object, responding to a triggering optical environment regulation event in the learning space where the target object is located; extracting features based on measured light output waveforms collected from the learning space and multimodal physiological signals collected from the target object to obtain light-specific transient response features and stable cognitive state features related to the learning task, wherein the optical environment regulation event is triggered when the light source in the learning space is adjusted using the current combination of optical environment parameters; inputting the light-specific transient response features and the stable cognitive state features into a visual cognitive performance evaluation model to obtain the aforementioned... The visual cognitive performance evaluation results of the target object are used; based on the set of safety constraints related to the above learning task, multiple candidate light environment parameter combinations are determined; for each candidate light environment parameter combination, the above candidate light environment parameter combination, the above current light environment parameter combination, the above visual cognitive performance evaluation results, and the task type of the above learning task are input into the light environment response prediction model to obtain the visual cognitive performance prediction results related to the above candidate light environment parameter combination; based on the visual cognitive performance prediction results related to each of the multiple above candidate light environment parameter combinations, a target light environment parameter combination is selected from the multiple above candidate light environment parameter combinations so as to adjust the light source of the above learning space using the above target light environment parameter combination.

[0006] According to an embodiment of the present invention, the above-mentioned feature extraction based on the measured light output waveform acquired from the learning space and the multimodal physiological signals acquired from the target object to obtain light-specific transient response features and stable cognitive state features related to the learning task includes: extracting event feature points corresponding to the light environment regulation event from the measured light output waveform to obtain the event feature point time; performing optical link phase correction on the event feature point time using the trigger time of the light environment regulation event to obtain the corrected light stimulation event time; determining a light-specific response window based on the corrected light stimulation event time; extracting light-specific transient response features related to the learning task from the multimodal physiological signals based on the light-specific response window; determining at least one stable cognitive window based on the stable start time that satisfies the stability criterion in the measured light output waveform; and extracting at least one steady-state sub-feature related to the learning task from the multimodal physiological signals based on at least one of the stable cognitive windows, wherein the stable cognitive state features include at least one of the steady-state sub-features.

[0007] According to an embodiment of the present invention, the method further includes: determining the safety boundary of the light environment parameter regulation behavior in multiple dimensions based on the task type of the learning task, and obtaining multiple sub-constraint sets; and obtaining a set of safety constraints related to the learning task based on the multiple sub-constraint sets.

[0008] According to an embodiment of the present invention, the plurality of sub-constraint sets include a flicker safety constraint subset; wherein, the determination of a plurality of candidate optical environment parameter combinations based on the safety constraint set related to the learning task includes: initializing a plurality of optical environment parameters to be optimized to obtain an initial optical environment parameter combination, wherein the initial optical environment parameter combination includes a plurality of flicker safety control parameters related to the flicker safety constraint subset; when the plurality of flicker safety control parameters satisfy the plurality of flicker safety constraint conditions included in the flicker safety constraint subset, matching the initial optical environment parameter combination with the plurality of safety constraint conditions included in the safety constraint set to obtain a matching result; when the matching result indicates that the initial optical environment parameter combination matches all the plurality of safety constraint conditions included in the safety constraint set, determining the initial optical environment parameter combination as the candidate optical environment parameter combination.

[0009] According to an embodiment of the present invention, the above-mentioned inputting the above-mentioned candidate light environment parameter combinations, the above-mentioned current light environment parameter combinations, the above-mentioned visual cognitive performance evaluation results, and the above-mentioned task type of the learning task into the light environment response prediction model to obtain the visual cognitive performance prediction results related to the above-mentioned candidate light environment parameter combinations includes: generating a candidate scheme input vector based on the above-mentioned visual cognitive performance evaluation results, the above-mentioned current light environment parameter combinations, the above-mentioned candidate light environment parameter combinations, the change between the above-mentioned candidate light environment parameter combinations and the above-mentioned current light environment parameter combinations, the task type encoding of the above-mentioned learning task, the individualized baseline of the above-mentioned target object, and the historical control records of the above-mentioned learning space; inputting the above-mentioned candidate scheme input vector into the above-mentioned light environment response prediction model to obtain the predicted response change; and superimposing the above-mentioned visual cognitive performance evaluation results and the predicted response change to obtain the visual cognitive performance prediction results related to the above-mentioned candidate light environment parameter combinations.

[0010] According to an embodiment of the present invention, the above-mentioned visual cognitive performance evaluation result includes the current state of each of the plurality of control targets, and the above-mentioned predicted response change includes the state change of each of the plurality of control targets; the above-mentioned superposition of the above-mentioned visual cognitive performance evaluation result and the above-mentioned predicted response change to obtain a visual cognitive performance prediction result related to the above-mentioned candidate light environment parameter combination includes: superimposing the current state and state change of each of the plurality of control targets respectively to obtain the predicted state of each of the plurality of control targets; and obtaining a visual cognitive performance prediction result related to the above-mentioned candidate light environment parameter combination based on the predicted state of each of the plurality of control targets.

[0011] According to an embodiment of the present invention, the above-mentioned method of selecting a target light environment parameter combination from multiple candidate light environment parameter combinations based on the visual cognitive performance prediction results associated with each of the multiple candidate light environment parameter combinations includes: obtaining a control score associated with each candidate light environment parameter combination based on the task type of the learning task and the visual cognitive performance prediction results associated with each candidate light environment parameter combination; and selecting the target light environment parameter combination from multiple candidate light environment parameter combinations based on the maximum control score among the control scores associated with each of the multiple candidate light environment parameter combinations.

[0012] According to an embodiment of the present invention, the above-mentioned visual cognitive performance prediction result includes the prediction state of each of a plurality of control targets; wherein, the above-mentioned visual cognitive performance prediction result based on the task type of the above-mentioned learning task and the visual cognitive performance prediction result related to the candidate light environment parameter combination, to obtain the control score related to the candidate light environment parameter combination, includes: determining the control weight of each of the plurality of control targets based on the task type of the above-mentioned learning task; and using the control weight of each of the plurality of control targets, performing a weighted summation of the prediction states of each of the plurality of control targets to obtain the control score related to the candidate light environment parameter combination.

[0013] According to an embodiment of the present invention, the method further includes: in response to triggering a target light environment adjustment event, determining the impact of triggering the target light environment adjustment event on visual cognitive performance and visual fatigue risk based on the target measured light output waveform and target multimodal physiological signals corresponding to the target light environment adjustment event, and the task behavior data of the target object, and obtaining a closed-loop feedback evaluation result, wherein the target light environment adjustment event is triggered when the light source of the learning space is adjusted using the target light environment parameter combination; and generating a new light environment parameter combination when the closed-loop feedback evaluation result indicates a decrease in visual cognitive performance after adjustment, or an increase in visual fatigue risk after adjustment.

[0014] According to an embodiment of the present invention, the method further includes: obtaining a model confidence score based on the visual cognitive performance evaluation results of the target object, wherein the model confidence score represents the credibility of the visual cognitive performance evaluation results; adjusting the light source of the learning space based on a protection strategy and pausing the adjustment of the light environment parameters when the visual cognitive performance evaluation results meet the protection conditions, or when the model confidence score is lower than the confidence score threshold; wherein adjusting the light source of the learning space based on the protection strategy includes at least one of the following: reducing brightness stimulation, switching to a soft color temperature, increasing background fill light, and reducing the rate of change of light environment parameters.

[0015] According to an embodiment of the present invention, measured light output waveforms are acquired from the learning space, and multimodal physiological signals are acquired from the target object, thereby obtaining the visual cognitive performance evaluation results of the target object. Based on the visual cognitive performance evaluation results, a target light environment parameter combination is selected from multiple candidate light environment parameter combinations determined from a set of safety constraints related to the learning task. This allows for the adjustment of the light source in the learning space using the target light environment parameter combination, enabling visual cognitive performance-oriented light environment control and improving the individualized adaptation capability of the learning space lighting. Attached Figure Description

[0016] The above and other objects, features and advantages of the present invention will become clearer from the following description of embodiments of the invention with reference to the accompanying drawings.

[0017] Figure 1 A flowchart of a brain-computer interface optical environment parameter regulation method for learning spaces according to an embodiment of the present invention is shown.

[0018] Figure 2 A schematic diagram illustrating the principle of optical link phase correction and dual-window segmentation according to an embodiment of the present invention is shown.

[0019] Figure 3 A schematic diagram of a brain-computer interface optical environment parameter regulation method for learning spaces according to an embodiment of the present invention is shown.

[0020] Figure 4 A schematic diagram of a brain-computer interface optical environment parameter regulation method for learning spaces according to another embodiment of the present invention is shown.

[0021] Figure 5 A schematic diagram of the architecture of a brain-computer interface optical environment parameter regulation system for learning spaces according to an embodiment of the present invention is shown. Detailed Implementation

[0022] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the invention. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the invention for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0023] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. The terms "comprising," "including," etc., as used herein indicate the presence of the above-described features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0024] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0025] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0026] In the embodiments of this invention, the collection, updating, analysis, processing, use, transmission, provision, disclosure, and storage of data (e.g., including but not limited to target object personal information) comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures are taken to prevent unauthorized access to target object personal information data and to safeguard the security of target object personal information, network security, and national security.

[0027] In embodiments of the present invention, the authorization or consent of the target object is obtained before acquiring or collecting the target object's personal information.

[0028] In recent years, the development of brain-computer interfaces and wearable physiological monitoring technologies has made it possible to infer an individual's attentional state, alertness level, visual fatigue, and cognitive load in real time using signals such as electroencephalography (EEG), eye movement, pupil size, heart rate, and skin conductance. Existing technologies have begun to explore the use of multimodal physiological signals for assessing the impact of ambient light, but most solutions remain at the experimental evaluation or offline analysis level. Their outputs are mostly cognitive performance assessments, and a real-time closed-loop light environment control mechanism for real learning spaces has not yet been established. In other words, existing solutions focus more on "what kind of impact the light environment has on the human body's state," but lack the technical link to further generate light environment parameter control commands based on the state recognition results.

[0029] Especially for groups, the lighting control of learning spaces not only needs to pursue high visual cognitive performance, but also needs to take into account visual comfort, eye safety, flicker control, circadian rhythm effects, and individual adaptation. Existing solutions lack a technical solution that can integrate brain-computer interface models, physiological state recognition, visual cognitive performance prediction, light environment parameter optimization, and safety constraint control into a closed-loop system. Specifically, existing solutions typically do not construct a unified visual cognitive performance objective function for indicators such as visual attention, reading efficiency, visual search efficiency, short-term memory, reaction speed, cognitive load, and visual fatigue risk, nor do they set safety constraints for individual learning spaces, such as illuminance, brightness abrupt changes, flicker, color temperature, short-wavelength components, screen light coordination, and nighttime rhythms, outside of this objective function. Furthermore, even if some existing solutions can generate dimming strategies based on current attentional state, fatigue state, or task performance, they usually only map the current state to preset dimming rules, lacking a predictive mechanism for "what changes in visual cognitive performance will be caused by the execution of candidate light environment parameter combinations." Therefore, existing solutions struggle to provide interpretable benefit evaluations among multiple candidate illuminance, color temperature, spectrum, brightness, screen brightness, and spatial light distribution schemes, and also fail to support truly multi-objective optimization. Furthermore, while existing solutions address attention-based lighting control, learning and work lighting optimization based on work efficiency and EEG data, and human-centered adaptive lighting control based on EEG feedback, they lack a dedicated technical solution for personal learning spaces that uses visual cognitive performance as the control objective and directly applies brain-computer interface state recognition results to closed-loop control of ambient light parameters. Most existing solutions focus on recommending or adjusting lighting parameters based on attention, work efficiency, EEG status, or general comfort needs. They haven't yet constructed a unified visual cognitive performance objective function encompassing indicators such as visual attention, reading efficiency, visual search efficiency, short-term memory, reaction speed, cognitive load, and visual fatigue risk. Furthermore, they haven't fully integrated constraints such as eye safety, flicker limitations, brightness abrupt changes, color temperature variation limits, short-wavelength component limitations, screen light coordination, and nighttime rhythm protection. They also haven't jointly optimized various light environment parameters, including contrast, brightness, correlated color temperature, spectral composition, short-wavelength component ratio, light field spatial distribution, dynamic change curves, and display terminal brightness. In addition, existing solutions typically lack a dual-timescale differentiation mechanism for transient light-specific responses and stable cognitive states after changes in light parameters. This makes it difficult to avoid misinterpreting short-term physiological responses such as pupillary light reflexes and visual evoked responses as decreased attention or increased fatigue, and it's also difficult to form individualized model updates and long-term adaptive regulation based on executive feedback.

[0030] Therefore, it is necessary to propose a novel method for regulating the light environment of learning spaces based on a brain-computer interface (BCI) model. This method enables the system to acquire the learner's visual cognitive state in real time and dynamically adjust the ambient light parameters according to the visual cognitive performance goals, thereby improving the intelligence, personalization, and health-adaptability of the learning space. This method should evolve from simple "state assessment" to a closed-loop control system encompassing "state recognition—light parameter decision-making—illumination execution—feedback update." Furthermore, it is necessary to construct a closed-loop control chain consisting of "actual light output phase correction—dual-timescale state analysis—real-time state estimation via BCI—candidate light parameter generation—light environment response prediction—visual cognitive performance objective function evaluation—safety constraint screening—multi-parameter optimization of the light environment—illumination execution—feedback update" to achieve personalized, safe, and cognitively performance-oriented adaptive adjustment of the learning space's light environment.

[0031] Figure 1 A flowchart of a brain-computer interface optical environment parameter regulation method for learning spaces according to an embodiment of the present invention is shown.

[0032] like Figure 1 As shown, the method for regulating the optical environment parameters of a brain-computer interface for learning spaces includes operations S110~S150.

[0033] In operation S110, during the learning task performed by the target object, a light environment adjustment event is triggered in the learning space where the target object is located. Based on the measured light output waveform collected from the learning space and the multimodal physiological signals collected from the target object, feature extraction is performed to obtain light-specific transient response features and stable cognitive state features related to the learning task. The light environment adjustment event is triggered when the light source in the learning space is adjusted using the current combination of light environment parameters.

[0034] In embodiments of the present invention, an adjustable lighting environment can be established in a personal learning space, and basic information of the learning space, learning task information, and individual baseline information of the target object can be obtained through the human-light-task joint perception and baseline management module.

[0035] Basic information about the learning space includes the location of the study desk, lighting arrangement, display terminal location, natural light entry point, desktop reflection characteristics, wall reflection characteristics, main reading area, writing area, screen learning area, and background supplementary lighting area. Furthermore, a light field zoning model of the learning space can be established based on the above information, dividing the learning space into a desktop accent lighting area, an eye incident light area, a background supplementary lighting area, a screen display area, and a surrounding environment area.

[0036] Learning task information includes the current learning task type, task start time, task duration, task difficulty, task presentation method, and whether it involves screen learning. Learning task types include, but are not limited to, paper reading, written assignments, memorization, mathematical calculations, visual search, screen learning, exam practice, and rest and recovery.

[0037] The target subject's individual baseline information includes brain-computer interface signals, multimodal physiological signals, eye-tracking and pupillary signals, task behavior data, and visual fatigue status under a baseline lighting environment. Basic data can be collected from the target subject during reading, writing, visual search, or short-term memory tasks under a preset baseline lighting environment, establishing baselines for attention, alertness, reading speed, reaction time, short-term memory, EEG frequency bands, pupillary light, blinking, heart rate or heart rate variability, and visual fatigue. This step provides an initial state reference for a specific target subject in a specific learning space, improving the individual adaptability of subsequent state recognition and lighting environment regulation decisions.

[0038] During the learning process, brain-computer interfaces and multimodal physiological acquisition devices can continuously collect human body state data, thereby acquiring multimodal physiological signals of the target object. At the same time, light environment sensing devices can collect actual light environment parameters in the learning space to obtain measured light output waveforms, and task behavior data can be collected through learning task terminals or task recognition units to form a synchronous data stream of "human-light-task".

[0039] Human-side state data includes, but is not limited to, electroencephalography (EEG) signals, eye movement trajectories, fixation points, saccades, number of retrospectives, blink information, pupil diameter, pupillary rate change rate, heart rate, heart rate variability (HRV), photoplethysmography (PPG) signals, electrodermal activity (EDA) signals, head movement signals, sitting stability signals, head-down distance, and other physiological or behavioral data that can reflect the learner's visual attention, alertness level, cognitive load, and risk of visual fatigue.

[0040] Task behavior data includes task type, task presentation time, answer accuracy, reading speed, reaction time, reaction time fluctuation, missed answer rate, wrong answer rate, short-term memory task performance, input rhythm, and screen learning behavior.

[0041] Light environment parameters include desktop illuminance, vertical eye illuminance, field brightness, background brightness, correlated color temperature, spectral composition, proportion of short-wavelength components, flicker frequency, modulation depth, pulse width modulation (PWM) duty cycle, natural light input, screen brightness, screen color temperature, and actual light output waveform.

[0042] All data acquisition devices are preferably synchronized via a unified clock, hardware trigger interface, network time synchronization, or software timestamp calibration to ensure that brain-computer interface signals, physiological signals, task behavior data, lighting environment data, and light source control data are within the same time frame. Furthermore, bandpass filtering, power frequency suppression, electrooculography artifact removal, motion artifact correction, and baseline correction can be applied to EEG signals; coordinate calibration, drift correction, blink interpolation, outlier removal, and pupil baseline correction can be applied to eye movement and pupil signals; filtering, smoothing, peak detection, and abnormal waveform removal can be applied to heart rate, PPG, or electrodermal signaling; abnormal reaction time removal, missed answer marking, accuracy statistics, and task stage division can be applied to learning task behavior data; and sensor calibration, waveform smoothing, drift correction, and missing value processing can be applied to lighting environment data.

[0043] In operation S120, the light-specific transient response characteristics and stable cognitive state characteristics are input into the visual cognitive performance evaluation model to obtain the visual cognitive performance evaluation results of the target object.

[0044] After obtaining the optical specific response window features and stable cognitive window features, the optical specific response window features and stable cognitive window features can be input into a pre-trained Lightweight Gradient Boosting Machine (LightGBM) visual cognitive performance estimation model to obtain the visual cognitive performance evaluation results of the target object, so as to estimate the current visual cognitive performance state of the target object in real time.

[0045] Specifically, the feature vectors within the light-specific response window include pupillary light reflex amplitude, pupillary constriction latency, pupillary constriction speed, pupillary recovery time, visual evoked response amplitude, and visual evoked response latency. The feature vectors within the stable cognitive window include EEG data. ratio, Inhibition index, Relative power, mean fixation duration, number of retrospectives, blink rate, task-related pupillary dilation, HRV, reaction time coefficient of variation, moving average of accuracy, desktop illuminance, vertical illuminance of the eye, correlated color temperature, proportion of short-wavelength components, and screen brightness.

[0046] Individual baseline correction and standardization were performed on the above features. For the target object... In learning task types Current window The first in The baseline correction value for each feature can be expressed as:

[0047] (1).

[0048] in, Indicates the number of elements in the current window. The original statistical values ​​of each feature, and These represent the target object under the reference lighting environment and the corresponding task type, respectively. The baseline mean and baseline standard deviation of each feature. To prevent tiny constants with a denominator of zero.

[0049] The system combines the individual baseline-corrected light-specific response features, stable cognitive features, current light environment exposure parameters, current learning task type encoding, and signal quality label into a model input vector. And input the LightGBM visual cognitive performance estimation model:

[0050] (2).

[0051] in, It includes at least the visual attention index, alertness level index, reading efficiency or visual search efficiency index, short-term memory performance index, cognitive load index, visual fatigue risk index, visual discomfort risk index, and comprehensive visual cognitive performance index.

[0052] Among them, the comprehensive visual cognition performance index It can be obtained by weighting factors such as task accuracy, reaction speed, attention stability, short-term memory performance, and fatigue risk, for example:

[0053] (3).

[0054] in, This represents the comprehensive visual cognitive performance index. Indicates the level of visual attention. Indicating short-term memory performance, This indicates reaction speed, reading efficiency, or visual search efficiency. Indicates the risk of visual fatigue. This indicates a risk of visual discomfort. to This refers to the weighting coefficient. The weighting coefficient can be adjusted based on the learning task type, individual baseline, current time period, and historical control effects.

[0055] In operation S130, multiple candidate combinations of optical environment parameters are determined based on the set of security constraints related to the learning task.

[0056] The set of safety constraints can be denoted as The light environment parameters to be optimized should meet the following requirements: .in, Indicates the first A candidate combination of light environment parameters. The set of safety constraints includes at least the following: illuminance safety constraints, eye vertical illuminance constraints, brightness abrupt change constraints, color temperature change amplitude constraints, color temperature change rate constraints, short-wavelength component ratio constraints, flicker risk constraints, modulation depth constraints, PWM parameter constraints, light field switching frequency constraints, screen brightness coordination constraints, nighttime rhythm protection constraints, learning duration constraints, and individual sensitivity constraints.

[0057] In operation S140, for each candidate light environment parameter combination, the candidate light environment parameter combination, the current light environment parameter combination, the visual cognitive performance evaluation result, and the task type of the learning task are input into the light environment response prediction model to obtain the visual cognitive performance prediction result related to the candidate light environment parameter combination.

[0058] Light environment response prediction model The model can be trained using pre-collected group sample data and individual closed-loop feedback data during system operation. Specifically, within safety constraints, the system records the visual cognitive state before each light environment adjustment, the current light environment parameters, the actual light environment parameters executed, the amount of light parameter change, the learning task type, the individualized baseline of the target object, and the changes in visual cognitive performance after adjustment, and constructs these as training samples. The model learns "how a change in a certain light environment parameter will lead to changes in visual cognitive performance under a specific learner, a specific task, and a specific initial state." Therefore, the light environment response prediction model is used to predict the changes in visual cognitive performance, visual fatigue risk, visual comfort, and visual discomfort risk that a candidate solution may cause within a future feedback window before executing the candidate light parameters. This model can be pre-trained using group data and then calibrated online or periodically updated based on the actual feedback data of a specific target object, thereby achieving individualized light response prediction.

[0059] In operation S150, based on the visual cognitive performance prediction results associated with each of the multiple candidate light environment parameter combinations, a target light environment parameter combination is selected from the multiple candidate light environment parameter combinations so as to adjust the light source of the learning space using the target light environment parameter combination.

[0060] After selecting the target lighting environment parameter combination, lighting environment parameter control instructions can be generated based on this combination to control and change the lighting environment parameters. These instructions include, but are not limited to, target desktop illuminance, target eye vertical illuminance, target brightness, target correlated color temperature, target spectral composition, target short-wavelength component ratio, target spatial light distribution, target screen brightness, target flicker safety control parameters, target PWM safety control parameters, rate of change, fading time, execution area, and execution priority. Through this step, the system converts the visual cognitive state output by the LightGBM visual cognitive performance estimation model into executable lighting environment parameter control instructions.

[0061] Furthermore, the light environment execution module can receive the generated light environment parameter control commands to drive dimmable LED lights, adjustable color temperature lights, adjustable spectrum lights, desktop study lights, background lights, auxiliary side lights, overhead lights, multi-source light source arrays, and display terminals to perform corresponding adjustments. These adjustments include illuminance adjustment, brightness adjustment, correlated color temperature adjustment, spectral composition adjustment, short-wavelength component ratio adjustment, spatial light distribution adjustment, display terminal brightness adjustment, and flicker or PWM parameter adjustment. For different learning areas, desktop accent lighting, background fill light, side fill light, and ambient light can be adjusted separately. For example, in paper reading tasks, desktop lighting uniformity can be improved and shadows reduced; in screen learning tasks, the background fill light ratio can be increased to reduce the contrast between the screen and the surrounding environment; and in rest and recovery tasks, the intensity of desktop accent lighting can be reduced and switched to softer background lighting.

[0062] Furthermore, to avoid interference from sudden changes in the light environment, a smooth transition strategy is adopted, causing illuminance, brightness, color temperature, or spectral parameters to change gradually according to a preset curve. The curve is preferably an S-shaped smooth curve. The amount of change in illuminance, brightness, color temperature, and spectral parameters per unit time can also be limited to prevent sudden brightening, sudden darkening, rapid color temperature jumps, or frequent switching of the light field.

[0063] During the execution of the lighting environment execution module, the actual light output detection unit can collect data such as adjusted desktop illuminance, vertical eye illuminance, brightness, color temperature, spectrum, flicker parameters, PWM parameters, and screen brightness, and determine whether the actual output meets the target value. If there is a deviation between the actual output and the target control command, the system can perform compensatory control or feed back the deviation information to the lighting environment multi-parameter optimization decision module to regenerate the control command.

[0064] Through the embodiments of the present invention, measured light output waveforms are collected from the learning space, and multimodal physiological signals are collected from the target object, thereby obtaining the visual cognitive performance evaluation results of the target object. Based on the visual cognitive performance evaluation results, a target light environment parameter combination is selected from multiple candidate light environment parameter combinations determined from a set of safety constraints related to the learning task. This allows for the adjustment of the light source in the learning space using the target light environment parameter combination, enabling light environment control guided by visual cognitive performance and improving the individualized adaptation capability of the learning space lighting.

[0065] In one embodiment of this application, to obtain light-specific transient response features and stable cognitive state features related to the learning task, the following operations can be performed: extracting event feature points corresponding to light environment regulation events from the measured light output waveform to obtain the event feature point time; performing optical link phase correction on the event feature point time using the trigger time of the light environment regulation event to obtain the corrected light stimulation event time; determining a light-specific response window based on the corrected light stimulation event time; extracting light-specific transient response features related to the learning task from multimodal physiological signals based on the light-specific response window; determining at least one stable cognitive window based on the stable start time that satisfies the stability criterion in the measured light output waveform; and extracting at least one steady-state sub-feature related to the learning task from multimodal physiological signals based on at least one stable cognitive window, wherein the stable cognitive state features include at least one steady-state sub-feature.

[0066] Specifically, determining the optical specific response window and stable cognitive window using phase correction may include the following steps.

[0067] The first step is to establish a unified timestamp and event number. Edge computing processors or main control processors can be used to provide a unified time reference to the light source controller, light environment detection unit, brain-computer interface acquisition device, eye-tracking pupil acquisition device, physiological signal acquisition device, and learning task terminal, and generate a unique event number for each light environment adjustment command. The event number is used to associate the control trigger time, actual light output waveform, brain-computer interface data, physiological data, and task behavior data within the same dimming event.

[0068] The second step is to obtain the trigger time for light source control. This can be achieved by using the light source controller or lamp driver to detect the trigger time when the light source controller receives the trigger time from the edge computing processor. After receiving the ambient light control command, record the actual moment when the driver output starts updating, as the trigger moment for light source control. This moment could be the moment the driver writes to the dimming register, the moment the PWM duty cycle or drive frequency begins to update, or the moment the luminaire controller confirms the start of executing the dimming command. The light source controller will number the event. Target light environment parameters and Data is sent to the data synchronization and preprocessing module in the edge computing processor.

[0069] The third step is to acquire the actual light output waveform and the timing of event feature points. This can be done using a light environment detection unit to acquire the actual light output waveform at the desktop area, the eye-level vertical illuminance detection position, or the near-end detection position of the lamp. The actual light output waveform can be an illuminance change curve, a brightness change curve, a color temperature change process, a spectral change process, or a flicker modulation waveform. The light environment detection unit determines the output waveform based on the event number. The actual optical output waveform corresponding to this dimming event is sent to the data synchronization and preprocessing module, or the characteristic point of the optical output event is extracted locally. Then it is sent to the data synchronization and preprocessing module. Event feature point time. It can be determined based on the threshold crossing point, half-slope point, maximum slope point, or stable starting point of the actual optical output waveform. For example, suppose the actual optical output before dimming is... The target light output is This allows the actual light output to achieve the target change. The moment is taken as the characteristic moment of the event:

[0070] (4).

[0071] in, The preferred ratio is the preset ratio. That is, the moment when the actual light output reaches 50% of the target change is taken as the event characteristic point moment.

[0072] Alternatively, it can be determined based on the maximum point of the first derivative of the actual optical output waveform:

[0073] (5).

[0074] When focusing more on the starting point of the stable cognitive window, the actual stable moment of light output can be further determined. That is, the earliest moment when the actual light output deviates from the target value by less than a preset threshold within a continuous preset time period:

[0075] (6).

[0076] in, Indicates the stability error threshold. This indicates the time required to determine continuous stability.

[0077] The fourth step is to calculate the phase offset of the optical output link. This can be done using the data synchronization and preprocessing module based on the information provided by the light source controller. The light environment monitoring unit provides Calculate the first Phase offset of the optical output link during the dimming event:

[0078] (7).

[0079] in, This represents the delay between the start of the dimming command execution by the light source driver and the actual light output reaching the preset characteristic point. This delay includes the response delay of the luminaire driver, as well as the combined effects of factors such as the rise or fall of light emission from the LED, control link delay, and sensor detection delay.

[0080] The fifth step is to generate the corrected photostimulation event timestamps. This can be done using the data synchronization and preprocessing module based on the phase shift.

[0081] (8).

[0082] because ,therefore In other words, the corrected light stimulation event time actually corresponds to the moment when the actual light output waveform reaches the preset characteristic point, rather than the trigger moment of the light source control command. Through this processing, subsequent window segmentation is based on the actual light exposure event, rather than the control command time.

[0083] The sixth step is to send the phase correction results to the feature segmentation unit. The event number can be obtained using the data synchronization and preprocessing module. Phase offset Corrected photostimulation event timing Actual light output stabilization time and the target light environment parameters corresponding to this dimming event. The data is then sent to the dual-timescale feature parsing module. If the dual-timescale feature parsing module and the data synchronization and preprocessing module are located in the same edge computing processor, then this sending process is a data transfer between modules within the processor; if they are located in different devices, the phase correction results can be transmitted via wired communication, local area network, Bluetooth, Wi-Fi, or other communication methods.

[0084] Step 7: Window segmentation based on the corrected event timestamps. After data preprocessing and temporal correction, the multimodal data can be segmented into dual-timescale windows based on the corrected light-stimulated event timestamps to obtain the light-specific response window and the stable cognitive window. This can be achieved using the dual-timescale feature analysis module. and After that, with As the starting point of the photosensitive response window, with As the starting point for a stable cognitive window.

[0085] Specifically, let the first The corrected photostimulation event time after the adjustment of secondary light environment parameters is: The actual time when the light output enters a stable state is The current learning task type is ,in It can represent task types such as paper reading, writing assignments, memorization, mathematical calculations, visual search, screen learning, exam preparation, or rest and recovery.

[0086] The first type of window is the photosensitive response window. This window uses the corrected photostimulation event time. Starting from a value between 0 and 2 seconds or other preset short time range, the data is used to extract the first... A set of optically specific transient response feature vectors corresponding to the secondary optical parameter adjustment event. The optically specific response window can be represented as:

[0087] (9).

[0088] The light-specific response window is used to extract transient visual physiological responses directly caused by changes in light environment parameters, including pupillary light reflex, visual evoked responses, event-related potentials (ERPs), and short-term EEG responses induced by light changes. Corresponding features include, but are not limited to, pupillary light reflex amplitude, pupillary constriction latency, pupillary recovery time, visual evoked response amplitude, ERP latency, short-term EEG energy changes, and response features related to the actual light output waveform. These features are primarily used to characterize the direct visual physiological responses triggered by changes in light parameters themselves and to avoid misinterpreting transient light responses as changes in stable cognitive states.

[0089] The second type of window is the stable cognition window. This window starts after the actual light output enters a stable state, preferably using a sliding window of 10 seconds or other preset lengths, and is continuously updated in steps of 1 second, 2 seconds, or other preset increments. The stable cognition window can be represented as:

[0090] (10).

[0091] in, Indicates the first A stable cognitive window, This indicates the initial moment when the actual light output enters a stable state. Indicates the sliding step size. Indicates the window length. Preferably, , Desirable , Alternatively, a preset step size can be used. If the current stable cognitive window spans the task switching time, the system will terminate the window or create a new window with a new task type to avoid mixing cognitive state features from different learning tasks.

[0092] The stable cognitive window is used to extract visual cognitive-related features during the continuous learning phase, including but not limited to EEG theta, alpha, and beta band power and ratios, attention stability indicators, alertness level indicators, fixation duration, scan rate, regression count, blink rate, task-related pupillary dilation, heart rate variability, skin conductance changes, reading speed, visual search efficiency, reaction time fluctuations, answer accuracy, missed answer rate, incorrect answer rate, and short-term memory task performance. Through the above window segmentation, EEG, pupillary, eye movement, heart rate, skin conductance, and task behavior data are all truncated and feature extracted with reference to real light exposure events.

[0093] Figure 2 A schematic diagram illustrating the principle of optical link phase correction and dual-window segmentation according to an embodiment of the present invention is shown.

[0094] like Figure 2 As shown, optical link phase correction is achieved based on the correction event time of the actual output light waveform. The optical specific response window is 0 to 2 seconds after the correction event time or a preset window, and the stable cognition window is a 10-second sliding window after the actual stable light output time.

[0095] Furthermore, based on the segmented light-specific response window and stable cognitive window, dual-timescale feature extraction can be performed to obtain light-specific transient response features related to the learning task and at least one steady-state sub-feature related to the learning task, wherein the stable cognitive state feature includes at least one steady-state sub-feature.

[0096] Learning task types The system participates in the feature extraction and interpretation process. For paper-based reading tasks, it focuses on extracting features such as reading speed, fixation duration, number of regressions, blink rate, and task-related pupil dilation within a stable cognitive window. For writing or calculation tasks, it focuses on extracting features related to reaction time fluctuations, moving average accuracy, posture stability, and visual fatigue. For visual search or exam training tasks, it focuses on extracting features related to reaction time, missed answer rate, incorrect answer rate, and visual search efficiency. For screen learning tasks, it further incorporates features such as screen brightness, ambient brightness, background illumination, and eye-tracking fixation stability. For rest and recovery tasks, it focuses on extracting recovery-related features such as blink rate, pupil recovery status, heart rate variability, and risk of visual discomfort.

[0097] By extracting features at both time scales, the system can distinguish between "direct physiological responses triggered by changes in light" and "changes in stable cognitive states formed during continuous learning." For example, if the system detects rapid pupil constriction shortly after increased light intensity, but attention indicators, task accuracy, and reaction time do not deteriorate within the stable cognitive window, the system will preferentially interpret this change as pupillary light reflex rather than directly classifying it as visual fatigue or decreased attention.

[0098] Furthermore, the following explains how phase correction and feature extraction are implemented from a hardware perspective. The main devices used for phase correction may include edge computing processors, light source controllers, ambient light detection units, brain-computer interface acquisition devices, eye-tracking pupil acquisition devices, physiological signal acquisition devices, and learning task terminals.

[0099] The edge computing processor, acting as the master control device, is responsible for generating optical environment control commands, assigning event numbers, receiving timestamps and sensor data from various devices, and performing phase correction calculations.

[0100] The light source controller is responsible for receiving control commands and driving the lamps to perform dimming, while also recording the trigger time of the light source control. And then send it back to the edge computing processor.

[0101] The light environment detection unit is responsible for acquiring the actual light output waveform. And send the actual optical output feature point moments to the edge computing processor. Or the original optical output waveform. The edge computing processor, based on... and Calculate phase offset Generate corrected photostimulation event timestamps and will , and event number Send to the dual-timescale feature parsing module.

[0102] Brain-computer interface acquisition devices, eye-tracking and pupil acquisition devices, and physiological signal acquisition devices do not need to directly calculate phase offset. Their main function is to continuously acquire and upload time-stamped data such as EEG, eye movement, pupil, PPG, heart rate, and skin conductance according to a unified time reference.

[0103] The dual-timescale feature parsing module can be used to analyze the data sent by the edge computing processor. and The system extracts corresponding windows from the continuous physiological data stream and extracts light-specific response features and stable cognitive state features. If the dual-timescale feature analysis module is deployed within the same edge computing processor, the phase correction results are transmitted within the same processor. If the feature analysis module is deployed in an independent computing device or a cloud server, the edge computing processor sends the event number, phase offset, corrected event time, and stable time to that device via a communication interface. Thus, there is a clear information exchange between the devices: the light source controller provides the control trigger time, the light environment detection unit provides the actual light emission feature time, the edge computing processor calculates the phase offset and issues a correction event marker, and the feature analysis module completes window segmentation and feature extraction based on the correction event marker.

[0104] The technical solution of this invention can also be achieved by introducing the concepts of a first device and a second device. The first device can be a data synchronization and preprocessing module in an edge computing processor or a main control processor. The first device obtains the light source control trigger time from the light source controller or lamp driver. The actual light output waveform or the actual light output event feature point time is obtained from the light environment detection unit. The first device is based on Calculate phase offset And generate corrected light-stimulated event timestamps. The first device then numbered the event. Phase offset Corrected photostimulation event timing and the actual time when the light output stabilizes The data is sent to a second device. The second device can be a dual-timescale feature parsing module or an edge computing unit where this module is deployed. The second device utilizes... Determine the photospecific response window and utilize A stable cognitive window is determined, and a corresponding time period is extracted from continuously collected EEG, pupil, eye movement, heart rate, skin conductance, and task behavior data, and features are extracted. If the first and second devices are deployed on the same edge processor, the above transmission process is a data transfer between internal modules; if they are deployed on different hardware devices, information exchange is completed through local wired or wireless communication interfaces.

[0105] Furthermore, when learning tasks are presented through a display terminal, the system can synchronously correct task presentation events and map task events, light stimulus events, and human response events onto the same timeline. To prevent the display interface itself from becoming an additional source of light stimulation, the system can lock screen brightness, screen color temperature, background grayscale, and contrast parameters, or incorporate display terminal brightness as part of the light environment parameters for subsequent coordinated control.

[0106] The system acquires the actual emitted light waveform through a real light output detection unit and corrects for light stimulation events based on the phase shift between the light source control trigger time and the actual light emission characteristic point time, reducing the time error between control commands and actual light exposure. Furthermore, the system sets up a light-specific response window and a stable cognitive window to identify transient light responses such as pupillary light reflection and visual evoked responses, as well as stable cognitive states such as attention, reading efficiency, memory, and fatigue, avoiding misinterpreting simple light reflection as decreased attention or visual fatigue.

[0107] After obtaining the visual cognitive performance evaluation results of the target object, the current light environment regulation target can be generated by combining the current learning task type, individualized baseline, historical regulation effects, and learning time period, and a set of safety constraints for the personal learning space can be established: based on the task type of the learning task, the safety boundaries of the light environment parameter regulation behavior in multiple dimensions are determined, resulting in multiple sub-constraint sets; based on the multiple sub-constraint sets, a set of safety constraints related to the learning task is obtained. The determined set of safety constraints is used to determine the regulation target, weight configuration, and safety boundaries required for subsequent evaluation of candidate light environment parameter combinations, but does not directly perform a comprehensive benefit score on the candidate light environment parameter combinations.

[0108] After obtaining the set of safety constraints related to the learning task using the above method, multiple candidate combinations of optical environment parameters can be determined based on the set of safety constraints. These multiple sub-constraint sets include a flicker safety constraint subset. The specific method is as follows: Initialize the multiple optical environment parameters to be optimized to obtain an initial combination of optical environment parameters. This initial combination of optical environment parameters includes multiple flicker safety control parameters related to the flicker safety constraint subset. If the multiple flicker safety control parameters satisfy the multiple flicker safety constraints included in the flicker safety constraint subset, match the initial combination of optical environment parameters with the multiple safety constraints included in the safety constraint set to obtain matching results. If the matching results indicate that the initial combination of optical environment parameters matches all the multiple safety constraints included in the safety constraint set, then the initial combination of optical environment parameters is determined as a candidate combination of optical environment parameters.

[0109] For paper-based reading tasks, control objectives may include improving reading speed, enhancing fixation stability, reducing the number of regressions, and minimizing the risk of abnormal blinking and visual fatigue. For writing assignments, control objectives may include maintaining appropriate desktop illumination, improving attentional stability, and reducing the risk of fatigue caused by prolonged close-range visual activity. For memorization tasks, control objectives may include enhancing alertness, maintaining short-term memory performance, and reducing the risk of excessive cognitive load. For visual search or test preparation tasks, control objectives may include improving visual search efficiency, reducing reaction time fluctuations, and reducing missed answers. For screen-based learning tasks, control objectives may include coordinating screen brightness with ambient brightness, reducing glare and contrast, and minimizing visual fatigue. For rest and recovery tasks, control objectives may include reducing stimulation, improving visual comfort, and promoting fatigue recovery.

[0110] Furthermore, it can be based on the type of learning task. Weight configuration for generating task objectives:

[0111] (11).

[0112] in, Indicates the type of the current learning task. The corresponding target weight vector, to These are used to characterize the weights of target items such as visual cognitive performance, visual comfort, visual fatigue risk, visual discomfort risk, and prediction uncertainty in the subsequent evaluation of candidate solutions.

[0113] Weighting can be adjusted based on the learning task type, individualized baseline, current time period, historical control effects, and target audience preferences. For example, in reading tasks, improving visual cognitive performance and reducing fatigue risk have higher weights; in screen learning tasks, reducing the contrast between screen and ambient brightness and the risk of visual discomfort have higher weights; and in rest and recovery tasks, visual comfort and fatigue recovery-related factors have higher weights.

[0114] Candidate combinations of lighting environment parameters can be represented by lighting environment parameter vectors. express:

[0115] (12).

[0116] in, Indicates desktop illuminance. Indicates vertical illuminance at the eye level. Indicates brightness. Indicates the correlated color temperature. Indicates spectral composition, Indicates the proportion of shortwave components. Indicates the spatial distribution of the light field. Indicates the direction of incidence. Represents a dynamically changing curve. This indicates the flicker frequency in the actual light output or drive output. This indicates the depth of actual light output brightness or illuminance fluctuation. Indicates the modulation depth of the light source driving signal. This indicates the PWM duty cycle, PWM drive frequency, or other PWM control parameters. This indicates the brightness of the display terminal.

[0117] also, , , and It is not used as an independent stimulus to improve visual cognitive performance, but rather as a parameter related to flicker safety control in safety constraint judgment. When generating light environment parameter control commands, the system limits the light source driving frequency, PWM duty cycle, modulation depth, and measured light output fluctuation depth to ensure that the actual output light flicker frequency, fluctuation depth, and modulation depth meet preset safety limits, avoiding the risk of visual discomfort or fatigue to learners caused by low-frequency flicker, excessively high fluctuation depth, or inappropriate PWM modulation.

[0118] Furthermore, the set of security constraints can be represented as the intersection of multiple sub-constraint sets:

[0119] (13).

[0120] in, This indicates illuminance and vertical illuminance constraints at the eye level. This represents constraints on the rate of change of brightness abruptly and the rate of change of the light environment. This indicates the correlated color temperature and the constraint on the rate of color temperature change. This indicates a constraint on the proportion of shortwave components. Indicates flicker safety constraints. This indicates a screen brightness coordination constraint. This indicates a protective constraint on nighttime circadian rhythms. This refers to individualized safety constraints based on individual sensitivity, historical adverse feedback, and boundaries set by parents or teachers.

[0121] Furthermore, the flicker safety constraint can be expressed as: .in, This represents the set of flicker safety constraints. For any candidate combination of optical environment parameters... The system determines whether the following conditions are met based on the actual light output detection results and drive control parameters:

[0122] (14).

[0123] (15).

[0124] (16).

[0125] (17).

[0126] in, Indicates the preset safe flicker frequency range. Indicates the maximum permissible depth of light output fluctuation. Indicates the maximum allowed modulation depth. This represents the preset set of safe PWM control parameters. If a candidate parameter combination does not meet any of the above constraints, the system will directly eliminate the candidate scheme, or reduce the dimming amplitude, adjust the driving frequency, reduce the modulation depth, and regenerate a candidate scheme.

[0127] While optimizing cognitive performance, eye safety constraints are introduced, including upper limit of illuminance, vertical illuminance to the eye, sudden brightness changes, flicker risk, modulation depth, color temperature change range, proportion of short-wavelength components, screen light coordination, nighttime rhythm protection, and continuous learning duration. These constraints are combined with individual baselines and historical feedback to form personalized learning lighting strategies, avoiding excessively strong, cold, flickering, or rapidly changing light stimulation caused by short-term attention enhancement.

[0128] After obtaining candidate light environment parameter combinations, it is necessary to further obtain visual cognitive performance prediction results related to the candidate light environment parameter combinations using a light environment response prediction model. The visual cognitive performance evaluation results include the current states of multiple control targets, and the predicted response changes include the state changes of multiple control targets. The specific method is as follows: Based on the visual cognitive performance evaluation results, the current light environment parameter combination, candidate light environment parameter combinations, the changes between the candidate and current light environment parameter combinations, the task type encoding of the learning task, the individualized baseline of the target object, and the historical control records of the learning space, a candidate scheme input vector is generated. The candidate scheme input vector is then input into the light environment response prediction model to obtain the predicted response changes. The current states and state changes of multiple control targets are then superimposed to obtain the predicted states of each control target. Based on the predicted states of multiple control targets, the visual cognitive performance prediction results related to the candidate light environment parameter combinations are obtained.

[0129] After generating multiple candidate combinations of lighting environment parameters, it is necessary to call the lighting environment response prediction model to predict the changes in visual cognitive performance that each candidate combination of lighting environment parameters may cause within the future feedback window.

[0130] Let the learner's state vector at the current moment be... The current lighting environment parameters are: , No. The candidate light environment parameter combinations are: The learning task type is The individualized baseline of the target object is Historical regulation records are The light environment response prediction model can then be expressed as:

[0131] (18).

[0132] in, Indicates the combination of candidate lighting environment parameters. Then, in the future feedback window The amount of change in visual cognitive response obtained from internal prediction; This represents a light environment response prediction model; the changes in visual cognitive response include, but are not limited to, changes in overall visual cognitive performance, changes in visual attention, changes in reading efficiency, changes in short-term memory performance, changes in visual fatigue risk, changes in visual discomfort risk, and changes in visual comfort.

[0133] Furthermore, the predicted state corresponding to the candidate parameter combination can be obtained based on the current state and the predicted change:

[0134] (19).

[0135] in, This indicates the current state of visual cognitive performance. Indicates the combination of candidate lighting environment parameters. The future visual cognitive performance state predicted afterward.

[0136] Therefore, the state of visual cognitive performance no longer depends on the overall visual cognitive performance. Visual comfort visual fatigue risk Risk of visual discomfort The current state value is not determined by the current state value, but by the predicted value of the candidate solution output by the light environment response prediction model, that is, by the predicted comprehensive visual cognitive performance. Predicting visual comfort Predicting the risk of visual fatigue and predicting the risk of visual discomfort A comprehensive decision.

[0137] Specifically, the aforementioned predicted quantities can be obtained from the output of the LightGBM ambient light response prediction model or calculated from its output. Let the current visual cognitive state be... The current lighting environment parameters are: The candidate light environment parameter combinations are The change in optical parameters is The learning task type is coded as The individualized baseline of the target object is Historical regulation records are Then the input vector of the candidate solution can be expressed as:

[0138] (20).

[0139] The light environment response prediction model outputs candidate solutions in the future feedback window. The predicted change in response within:

[0140] (twenty one).

[0141] in, This can include changes in visual attention, short-term memory, reading efficiency or visual search efficiency, cognitive load, risk of visual fatigue, risk of visual discomfort, and visual comfort. The system overlays these predicted changes with the current state to obtain the predicted future state corresponding to the candidate solution.

[0142] (twenty two).

[0143] For example, if the current state includes the visual attention index Short-term memory performance index Reading efficiency index Visual fatigue risk index and visual discomfort risk index Then the model can predict the candidate solutions respectively. Changes in the above indicators:

[0144] (twenty three).

[0145] (twenty four).

[0146] (25).

[0147] (26).

[0148] (27).

[0149] Based on this, candidate parameter combinations The corresponding predicted overall visual cognitive performance can be obtained by weighting the predicted basic indicators:

[0150] (28).

[0151] in, to The weighting coefficients for the visual cognitive performance index can be set or updated based on the type of learning task, individualized baseline, and historical control effects.

[0152] Predicting visual comfort The visual comfort level can be directly output from the light environment response prediction model, or it can be calculated based on the predicted visual discomfort risk corresponding to the candidate solution, the degree of matching between screen and ambient brightness, the magnitude of changes in the light environment, the flicker safety margin, and the historical comfort feedback of the target object. For example, the system can express visual comfort as:

[0153] (29).

[0154] in, This represents the comfort calculation function. This indicates the degree of mismatch between screen brightness and ambient brightness. This indicates the magnitude of change in the candidate optical parameter combination relative to the current optical environment. Indicates the safety margin for flicker. This indicates the target object's historical comfort feedback records.

[0155] Predicting the risk of visual fatigue and predicting the risk of visual discomfort It can be used as a direct output of the LightGBM light environment response prediction model, or it can be calculated by comprehensively considering indicators such as predicted EEG, eye movement, pupil size, blink rate, reaction time fluctuation, and subjective feedback trend. Preferably, in the embodiments of the present invention, the system uses both as output items of the light environment response prediction model so that they can be directly used in the evaluation of the benefits of candidate solutions.

[0156] Forecast uncertainty Used to represent the model's response to candidate solutions The reliability of the predictions. The prediction uncertainty can be obtained through the ensemble prediction variance of multiple LightGBM sub-models, the model's historical prediction error, the distance between the candidate scheme and the training sample distribution, or the prediction interval width. For example, the system can be trained... Given three LightGBM response prediction sub-models based on different training subsets, each outputting prediction results for candidate solutions, the prediction uncertainty can be expressed as:

[0157] (30).

[0158] in, Indicates the first The sub-model for the first The output of each predictive indicator, This represents the prediction variance between models. This represents the uncertainty weights corresponding to different forecasting indicators. If... If the score exceeds a preset threshold, the system may lower the benefit score of the candidate solution, select a more conservative combination of candidate parameters, or trigger manual confirmation or protection strategies.

[0159] Therefore, the predicted comprehensive visual cognitive performance, predicted visual comfort, predicted visual fatigue risk, and predicted visual discomfort risk of the candidate solutions are not preset values, but are calculated by the light environment response prediction model based on historical closed-loop control data, current state, candidate light environment parameters, and individualized baseline, and are continuously updated in subsequent execution feedback.

[0160] Furthermore, when the system detects an increased risk of visual fatigue, abnormal blink rate, abnormal pupil recovery, excessively close head-down distance, excessively long continuous learning time, insufficient model confidence, or discomfort actively reported by the target object, it prioritizes triggering protective strategies. These protective strategies include reducing brightness stimulation, switching to a softer color temperature, increasing background illumination, reducing the rate of change in the light environment, pausing enhanced light parameter adjustments, prompting for rest, or requesting manual confirmation.

[0161] After obtaining multiple candidate light environment parameter combinations, it is necessary to select the target light environment parameter combination from these combinations. The visual cognitive performance prediction result includes the prediction state of each of the multiple control targets. The specific method is as follows: Based on the task type of the learning task, determine the control weight of each of the multiple control targets; using the control weight of each of the multiple control targets, perform a weighted summation of the prediction states of each of the multiple control targets to obtain the control score related to the candidate light environment parameter combination; based on the maximum control score among the control scores related to each of the multiple candidate light environment parameter combinations, select the target light environment parameter combination from the multiple candidate light environment parameter combinations.

[0162] In embodiments of the present invention, instead of directly mapping the current visual cognitive state to a single dimming strategy, a set of candidate light environment parameters is first generated:

[0163] (31).

[0164] in, This represents the set of candidate light environment parameters generated at the current moment. Indicates the first A candidate combination of light environment parameters.

[0165] The system ultimately selects the solution with the highest overall benefit from the set of candidate parameters that satisfy the safety constraints.

[0166] (32).

[0167] in, The optimal combination of light environment parameters for the target. This represents the set of eye safety constraints.

[0168] The system performs a comprehensive benefit evaluation on candidate solutions based on the predicted state. Candidate solutions The overall return function can be expressed as:

[0169] (33).

[0170] in, Indicates candidate parameter combinations Predictive comprehensive visual cognitive performance within the future feedback window Indicates predicted visual comfort. Indicates the predicted risk of visual fatigue. Indicates the predicted risk of visual discomfort. This indicates the uncertainty in the prediction of the light environment response model for this candidate solution. to This represents the weights of different objectives.

[0171] Through the embodiments of the present invention, instead of directly mapping a fixed dimming strategy based on the current state, after generating multiple candidate combinations of light environment parameters, the light environment response prediction model predicts the changes in visual cognitive performance, visual fatigue risk, visual comfort, and discomfort risk that each candidate scheme may cause within the future feedback window, and evaluates and ranks the candidate schemes accordingly, thereby selecting the optimal or suboptimal combination of light environment parameters.

[0172] In a specific embodiment of the present invention, a closed-loop feedback adjustment can be performed on the combination of light environment parameters. That is, when the closed-loop feedback evaluation result indicates a decrease in visual cognitive performance after adjustment, or an increase in the risk of visual fatigue after adjustment, a new combination of light environment parameters is generated. The specific method is as follows: In response to triggering a target light environment adjustment event, based on the target measured light output waveform and target multimodal physiological signals corresponding to the target light environment adjustment event, and the task behavior data of the target object, the impact of triggering the target light environment adjustment event on visual cognitive performance and the risk of visual fatigue is determined, and a closed-loop feedback evaluation result is obtained. The target light environment adjustment event is triggered when the light source of the learning space is adjusted using the target light environment parameter combination. When the closed-loop feedback evaluation result indicates a decrease in visual cognitive performance after adjustment, or an increase in the risk of visual fatigue after adjustment, a new combination of light environment parameters is generated.

[0173] After the light environment adjustment is completed, brain-computer interface signals, multimodal physiological signals, task behavior data, and actual light output data of the target subjects can be collected to evaluate the impact of this adjustment on visual cognitive performance and visual fatigue risk. Feedback evaluation indicators include changes in the comprehensive visual cognitive performance index, visual attention index, alertness level, reading speed, visual search efficiency, reaction time, reaction time fluctuation, short-term memory performance, task accuracy, visual fatigue risk, visual discomfort risk, blink rate, pupil recovery status, subjective discomfort feedback, and actual light output deviation.

[0174] If the overall visual cognitive performance index increases after adjustment, and the risk of visual fatigue and visual discomfort does not increase, the system can mark the current combination of light environment parameters as an effective strategy and increase its recommended weight in the same or similar tasks. If attention continues to decline, reaction time fluctuations increase, task accuracy decreases, reading efficiency decreases, or the risk of visual fatigue increases after adjustment, the system determines that the current strategy is ineffective and regenerates the light environment adjustment scheme. If there is obvious visual discomfort, excessively long continuous learning time, excessively close head-down distance, abnormal blinking, abnormal vertical illuminance of the eyes, or insufficient model confidence, the system will prioritize the protection strategy or the manual confirmation strategy.

[0175] Furthermore, the individualized baseline, brain-computer interface model parameters, and light environment control strategies for the target subjects can be updated based on long-term usage data. The individualized baseline can be corrected using moving averages, exponential smoothing, or other online update methods. For example, the individualized baseline for a certain state indicator... You can update it as follows:

[0176] (34).

[0177] in, This represents the updated individual baseline. This represents the individual baseline before the update. This represents the state statistics collected under the current stable state. This represents the smoothing coefficient.

[0178] Furthermore, the system can update the light environment response prediction model based on feedback data after actual execution. Specifically, for each executed combination of light environment parameters, the system records the state before adjustment. Current lighting environment parameters Light environment parameters have been implemented. Learning task types Individualized baseline for target objects Predicted response results and the actual observation results after adjustment The system performs online fine-tuning, periodic retraining, or individualized calibration of the light environment response prediction model based on the error between the predicted results and the actual observation results.

[0179] The prediction error can be expressed as:

[0180] (35).

[0181] in, Indicates the prediction error. This represents the actual observed visual cognitive performance after adjustment. This represents the visual cognitive performance predicted by the model. By continuously accumulating prediction errors and feedback results, the system can continuously correct the response models of different individuals to changes in lighting environment parameters, thereby improving the accuracy of subsequent candidate lighting parameter evaluation and multi-objective optimization.

[0182] Simultaneously, the system can establish a personalized learning lighting strategy library, recording effective combinations of light environment parameters for specific target objects under different tasks, time periods, and states. For example, the system can record that the target object responds better to neutral to cool light and higher desktop illuminance when reading on paper in the morning, responds better to low-stimulation background lighting and lower color temperature when studying on a screen at night, and responds better to soft color temperature, low rate of change, and rest prompts when fatigue risk increases. Through continuous feedback and strategy iteration, the system can form a long-term adaptive personalized closed-loop light environment control method.

[0183] The system employs a combined optimization approach, considering various parameters including desktop illuminance, vertical eye illuminance, brightness, correlated color temperature, spectral composition, proportion of short-wavelength components, spatial distribution of the light field, dynamic variation curves, flicker parameters, PWM parameters, and display terminal brightness. Following the illumination process, the system continues to collect brain-computer interface signals, physiological signals, task behavior data, and actual light output data. Based on the control effect, it updates the individual baseline, brain-computer interface model, light environment response prediction model, and control strategy, forming a long-term adaptive closed-loop control system.

[0184] Figure 3 A schematic diagram of a brain-computer interface optical environment parameter regulation method for learning spaces according to an embodiment of the present invention is shown.

[0185] like Figure 3 As shown, closed-loop adaptive control is achieved by continuously optimizing the light environment parameters based on input and feedback in each control cycle.

[0186] In a specific embodiment of the present invention, a safety protection and fault tolerance mechanism can also be introduced to evaluate the combination of light environment parameters. The specific method is as follows: Based on the visual cognitive performance evaluation results of the target object, the model confidence score is obtained, wherein the model confidence score represents the credibility of the visual cognitive performance evaluation results; if the visual cognitive performance evaluation results meet the protection conditions, or if the model confidence score is lower than the confidence score threshold, the light source of the learning space is adjusted based on the protection strategy, and the adjustment of the light environment parameters is suspended; wherein, the adjustment of the light source of the learning space based on the protection strategy includes at least one of the following: reducing brightness stimulation, switching to a soft color temperature, increasing background fill light, and reducing the rate of change of light environment parameters.

[0187] Furthermore, the system can output model confidence scores to determine the reliability of the current visual cognitive performance estimation results. When there are too many EEG signal artifacts, eye tracking failures, missing pupil data, abnormal heart rate signals, insufficient task data, or model confidence scores below a preset threshold, the system can reduce the automatic adjustment amplitude, enter a conservative dimming strategy, or prompt the learner to put on the sensing device again, adjust their posture, or request manual confirmation.

[0188] Furthermore, when the confidence level of the light environment response prediction model is lower than a preset threshold, or when the prediction uncertainty of the candidate parameter combination is high, the system can reduce the automatic adjustment amplitude, select a conservative candidate parameter combination, or revert to a rule-based safe dimming strategy. Thus, the system can utilize the prediction model for multi-objective optimization while avoiding over-dimming or erroneous dimming when the model is unreliable.

[0189] Optimization of lighting environment parameters is achieved through a candidate set scoring and ranking. The system selects from the candidate set... Eliminate solutions that do not meet eye safety constraints, and select the parameter combination with the highest overall benefit from the remaining candidate solutions. If the model confidence is insufficient or the predicted risk of the candidate solution is high, the system will select a conservative candidate solution or trigger a protection strategy. The system can first generate multiple candidate combinations of light environment parameters based on the current state, then eliminate candidate solutions that do not meet the eye safety constraints, and finally comprehensively compare the predicted values ​​of visual cognitive performance, visual comfort, visual fatigue risk, and safety margin among the remaining candidate solutions to select the optimal or suboptimal control strategy.

[0190] For example, when the system detects a decrease in the learner's visual attention index and an increase in reaction time, but a low risk of visual fatigue, it can appropriately increase the desktop illuminance, increase the neutral to cool light component, or enhance the desktop accent lighting; when the system detects an increased risk of visual fatigue, an increased blink rate, abnormal pupil recovery, or unstable posture, it can reduce brightness stimulation, decrease the proportion of short-wavelength components, increase soft background light, or prompt a short break; when the system detects screen learning and there is a large contrast between the screen and the surrounding environment brightness, it can coordinate and adjust the screen brightness and background lighting to match the screen light output with the ambient light; when the system is in a nighttime learning scenario, it can prioritize limiting enhancement strategies with high color temperature, high illuminance, or high short-wavelength components.

[0191] Furthermore, during the execution of the above methods, the target object baseline, raw acquisition data, preprocessed data, feature vectors, brain-computer interface model parameters, light environment parameters, control commands, execution records, feedback results, and individualized strategy library can also be stored and managed.

[0192] For the target audience's brain-computer interface data, physiological data, learning behavior data, and identity information, privacy protection can be achieved through localized processing, anonymized storage, de-identification, access control, and encrypted transmission. Preferably, core physiological data and brain-computer interface data undergo primary processing locally, with anonymized statistical results, model update amounts, or control records only uploaded upon authorization. The system can also provide different levels of data viewing and policy setting functions based on the permissions of the target audience, parents, teachers, or administrators.

[0193] The system output includes, but is not limited to, current visual cognitive performance status, current visual fatigue risk, current lighting environment parameters, recommended or executed lighting environment control commands, changes in status before and after adjustment, strategy effectiveness markers, individualized strategy update results, and rest or protection prompts. These outputs can be displayed in real-time on mobile terminals, learning lamp control interfaces, display terminals, or backend management platforms, and can also serve as a basis for subsequent learning space lighting optimization, individualized learning environment configuration, and eye health management.

[0194] The above methods enable a complete closed-loop control process, from "brain-computer interface state recognition" to "visual cognitive performance regulation target construction" and then to "multi-parameter optimization of the light environment under safety constraints, lighting execution and feedback update". This allows the light environment of the learning space to no longer rely on fixed patterns or simple manual adjustments, but to be personalized, safe and cognitive performance-oriented adaptively adjusted according to the learner's real-time visual cognitive state and long-term feedback results.

[0195] Figure 4 A schematic diagram of a brain-computer interface optical environment parameter regulation method for learning spaces according to another embodiment of the present invention is shown.

[0196] like Figure 4 As shown, through learning space modeling, user baseline establishment and learning task initialization, multi-source data acquisition of humans, light and tasks, data synchronization preprocessing and actual light output phase correction, etc., the final result of personalized, safe and visually cognitive performance-oriented closed-loop control of the light environment was achieved.

[0197] In one specific embodiment of the present invention, a closed-loop control system for brain-computer interface light environment parameters oriented towards learning spaces is also provided. This system is used to collect learners' brain-computer interface signals, multimodal physiological signals, learning task behavior data, and actual light environment parameters during personal learning tasks such as reading, writing, memorizing, visual search, screen learning, exam training, or rest and recovery. It estimates the learner's visual cognitive performance in real time and generates light environment parameter control commands under eye safety constraints, driving devices such as adjustable light, adjustable color temperature, and adjustable spectrum lamps and display terminals to perform light environment adjustment. After adjustment, the system continues to collect learner state changes and actual light output results, evaluates the control effect, and updates the individual baseline, model parameters, and control strategy, thereby forming a personalized, safe, and visually cognitive performance-oriented closed-loop light environment control system for learning spaces.

[0198] A brain-computer interface optical environment parameter closed-loop control system for learning spaces may include: a human-light-task joint perception and baseline management module, a synchronous preprocessing and dual time-scale feature analysis module, a visual cognitive performance estimation and control target generation module, a safety constraint and optical environment multi-parameter optimization decision module, and an optical environment execution feedback and individualized update module.

[0199] The Human-Light-Task Joint Perception and Baseline Management Module is used to collect basic data on learners, learning tasks, and the light environment of the learning space, and to establish individualized baselines, providing input for subsequent state recognition and closed-loop regulation. The module includes a learning space information acquisition unit, a target individual baseline acquisition unit, a brain-computer interface and multimodal physiological acquisition unit, a light environment measurement acquisition unit, and a learning task behavior acquisition unit.

[0200] The learning space information acquisition unit is used to obtain basic environmental information about an individual's learning space, including the location of the desk, lighting arrangement, display terminal location, natural light entry point, desktop reflection characteristics, wall reflection characteristics, main reading area, writing area, screen learning area, and background supplementary lighting area. Based on this information, the system can establish a light field zoning model for the learning space, dividing it into a desktop accent lighting area, an eye incident light area, a background supplementary lighting area, a screen display area, and a surrounding environment area, facilitating subsequent zoned lighting control.

[0201] The target subject individual baseline acquisition unit is used to collect learners' initial state data under a reference lighting environment and establish individualized baseline profiles. Individualized baselines include, but are not limited to, attention baseline, alertness baseline, reading speed baseline, reaction time baseline, short-term memory baseline, EEG frequency band baseline, pupillary baseline, blinking baseline, heart rate or heart rate variability baseline, and visual fatigue baseline. These baselines are used to distinguish between inherent individual differences and state changes resulting from light environment adjustment, avoiding the system directly using group average standards to judge the state of individual learners. Furthermore, the target subject individual baseline acquisition unit can establish a target subject-task dual-index baseline for each type of feature used for visual cognitive state recognition. Let the target subject be... The learning task type is , No. The first feature is in the reference light environment. The statistics for each valid window are Then the individual baseline mean and standard deviation of this feature are respectively:

[0202] (36).

[0203] (37).

[0204] in, Indicates the first task type in the reference light environment The number of effective windows for each feature. (The above) and The data will then be used to perform individual baseline correction on features such as EEG, eye movement, pupil size, heart rate, and task behavior extracted during the formal learning phase.

[0205] The brain-computer interface (BCI) and multimodal physiological acquisition unit are used to collect BCI signals and related physiological signals from learners during the learning process. The BCI signals are preferably non-invasive EEG signals, which can be acquired by a wearable EEG headband, dry electrode EEG device, or lightweight BCI device. The related physiological signals include eye movement trajectory, fixation point, saccade behavior, blink information, pupil diameter and pupillary change rate, heart rate, heart rate variability (HRV), PPG signal, EDA signal, head movement, sitting stability, and head-down distance. These signals are used to reflect changes in visual attention, alertness level, cognitive load, visual fatigue, and eye-use behavior in adolescent learners.

[0206] The light environment measurement and acquisition unit is used to collect actual light environment parameters in the learning space, including desktop illuminance, vertical eye illuminance, field of view brightness, background brightness, correlated color temperature, spectral composition, short-wavelength component ratio, flicker frequency, modulation depth, PWM duty cycle, natural light input, screen brightness, and screen color temperature. Furthermore, this unit can be configured with an actual light output detection function to record the actual light output waveform after the luminaire receives control commands, including illuminance change curves, brightness change curves, color temperature change processes, spectral change processes, and flicker modulation waveforms, thereby determining the actual response of the luminaire or display terminal.

[0207] The learning task behavior collection unit is used to collect learners' current task type and performance data. The learning task types include paper reading, written assignments, memorization, mathematical calculations, visual search, screen learning, exam practice, and rest / recovery. The task performance data includes reading speed, answer accuracy, reaction time, reaction time fluctuation, missed answer rate, incorrect answer rate, short-term memory task performance, input rhythm, and screen learning behavior. Through this unit, the system can combine the learner's physiological state with specific learning task performance, avoiding reliance on a single physiological indicator for dimming judgment.

[0208] The synchronous preprocessing and dual-timescale feature analysis module is used to perform unified time synchronization, signal preprocessing, and dual-window feature extraction on three types of data: human, light, and task data. This distinguishes between transient reactions caused by light changes and cognitive state changes during stable learning phases. The synchronous preprocessing and dual-timescale feature analysis module includes a unified time synchronization unit, a data preprocessing unit, an actual light output phase correction unit, a light-specific response window feature extraction unit, and a stable cognitive window feature extraction unit.

[0209] The unified time synchronization unit maps the brain-computer interface device, eye-tracking pupil acquisition device, physiological sensing device, light environment sensor, light source controller, display terminal, and learning task terminal to a unified time axis. This unified time axis can be achieved through a common clock, hardware triggering, network time synchronization, or software timestamp calibration. The system records event information such as the start time of the learning task, the time of task stimulus presentation, the time of target object response, the time of light source control command issuance, the actual light emission time, the start time of light parameter changes, and the time of light parameter stabilization, providing a temporal basis for subsequent feature extraction and closed-loop feedback.

[0210] The data preprocessing unit is used to preprocess brain-computer interface signals, multimodal physiological signals, task behavior data, and light environment data. Specifically, it performs bandpass filtering, power frequency suppression, electrooculography artifact removal, motion artifact correction, and baseline correction on EEG signals; coordinate calibration, blink interpolation, outlier removal, and pupil baseline correction on eye movement and pupil signals; filtering and smoothing, peak detection, and abnormal waveform removal on heart rate, PPG, or electrodermal transfer signal (EDS) signals; abnormal reaction time removal, missed answer marking, and accuracy statistics on task behavior data; and sensor calibration, waveform smoothing, drift correction, and missing value processing on light environment data.

[0211] The actual light output phase correction unit is used to perform time correction on the light stimulation event based on the trigger time of the light source control and the time of characteristic points in the actual light output waveform. The characteristic points in the actual light output waveform can be determined based on threshold crossing points, maximum slope points, half-stroke points, stable cycle start points, or other preset characteristic points. Through this correction method, the system can reduce timing errors caused by lamp response delay, control link delay, sensor sampling delay, and changes in environmental reflection, thus establishing a more accurate correspondence between changes in the light environment and the learner's brainwaves, pupillary movements, eye movements, heart rate, and other human responses.

[0212] The light-specific response window feature extraction unit is used to extract transient physiological response features directly caused by light changes within a short time window following a change in light parameters. These transient physiological response features include, but are not limited to, pupillary light reflex amplitude, pupillary constriction latency, pupillary recovery time, visual evoked response amplitude, event-related potential latency, and short-term EEG response features induced by light changes. These features are primarily used to characterize the direct impact of light changes themselves on the visual pathway and neural responses.

[0213] The stable cognitive window feature extraction unit is used to extract visual cognitive-related features during the continuous learning phase after the actual light output enters a stable state. These stable cognitive state features include, but are not limited to, EEG theta, alpha, and beta band power and ratios, attention stability indices, alertness level indices, reading speed, visual search efficiency, fixation duration, number of retrospectives, blink rate, task-related pupillary dilation, heart rate variability, reaction time fluctuations, correct answer rate, missed answer rate, incorrect answer rate, and short-term memory task performance. Through this dual-timescale feature analysis, the system can avoid misjudging pupillary light reflection after brightness changes as visual fatigue, and also avoid directly equating short-term EEG responses caused by light stimulation with attentional changes during the stable learning phase.

[0214] The Visual Cognitive Performance Estimation and Regulation Target Generation Module is used to estimate learners' visual cognitive performance in real time based on a brain-computer interface model and generate a light environment regulation objective function according to the learning task type. The module includes a brain-computer interface model inference unit, a visual cognitive index output unit, a model confidence assessment unit, and a regulation target generation unit.

[0215] The brain-computer interface model inference unit receives the multimodal feature vectors output by the synchronous preprocessing and dual-timescale feature parsing modules and inputs them into the pre-trained or online-updated brain-computer interface model. The preferred brain-computer interface model is the LightGBM visual cognitive performance estimation model based on gradient boosting decision trees, which outputs the current visual cognitive performance state of the adolescent learner based on the multimodal features within the light-specific response window and the stable cognitive window.

[0216] Specifically, the system constructs feature vectors for the photosensitive response window and stable cognitive window, respectively. The feature vector for the photosensitive response window can be represented as:

[0217] (38).

[0218] in, Indicates the amplitude of pupillary light reflection. Indicates the pupillary constriction latency period. Indicates the speed of pupil constriction. Indicates pupil recovery time. Indicates the amplitude of the visually evoked response. This indicates the latency period of the visually evoked response.

[0219] The feature vector of a stable cognitive window can be represented as:

[0220] (39).

[0221] in, Indicates brainwave ratio, express Inhibition index express relative power of frequency band This represents the mean fixation duration. Indicates the number of times the view was viewed. Indicates blink rate, This indicates task-related pupil dilation. Indicates heart rate variability. Indicates the coefficient of variation during the reaction. This represents the moving average of accuracy. Indicates desktop illuminance. Indicates vertical illuminance at the eye level. Indicates the correlated color temperature. Indicates the proportion of shortwave components. This indicates the brightness of the display terminal.

[0222] Among the stable cognitive window features, standardized values ​​after individual baseline correction are preferred for human-related or behavioral features such as EEG, eye movement, pupil size, heart rate, and task behavior. For the first... The baseline correction value for each feature can be expressed as:

[0223] (40).

[0224] in, Represents the target object In learning task types Next The standardized change of each feature relative to the individual baseline; Indicates the number of elements in the current window. The original statistical values ​​of each feature; and These represent the target object under the reference lighting environment and the corresponding task type, respectively. The baseline mean and baseline standard deviation of each feature; To prevent tiny constants with a denominator of zero.

[0225] For lighting environment exposure characteristics such as desktop illuminance, eye-level vertical illuminance, correlated color temperature, short-wavelength component ratio, and display terminal brightness, the system can use these as contextual features input to the model, and represent them using sensor calibration range normalized values ​​or changes relative to the reference lighting environment. The final multimodal feature vector input to the LightGBM visual cognitive performance estimation model can be expressed as:

[0226] (41).

[0227] in, This indicates the transient response characteristics within the optical specific response window after baseline correction. This represents baseline-corrected human and behavioral characteristics within a stable cognitive window. This indicates the current light environment exposure parameters. This indicates the change in current lighting environment parameters relative to reference lighting environment parameters. The encoding vector representing the type of the current learning task. These are markers indicating signal quality, sensor effectiveness, or model confidence.

[0228] The output of the LightGBM visual cognitive performance estimation model can be represented as:

[0229] (42).

[0230] in, Indicates the visual attention index. An index indicating alertness level. Indicators of short-term memory performance This index represents reading efficiency, visual search efficiency, or reaction speed. Indicates cognitive load index, Indicates the risk index of visual fatigue. Indicates the risk index of visual discomfort. This represents the comprehensive visual cognitive performance index.

[0231] The visual cognitive index output unit is used to receive the output of the brain-computer interface model inference unit. These indicators are then analyzed into metrics related to the learner's current visual cognitive performance, including but not limited to visual attention index, alertness level index, reading efficiency index, visual search efficiency index, short-term memory performance index, cognitive load index, visual fatigue risk index, visual discomfort risk index, and comprehensive visual cognitive performance index. These indicators characterize the learner's learning-related visual cognitive state under specific learning tasks and lighting conditions, and serve as the basis for generating subsequent lighting environment control targets.

[0232] The model confidence assessment unit is used to determine the reliability of the current state estimation results. When there are too many EEG signal artifacts, eye tracking failures, missing pupil data, abnormal heart rate signals, insufficient task data, or the model output confidence is lower than a preset threshold, the system can reduce the automatic adjustment amplitude, enter a conservative dimming strategy, or prompt the learner to put on the sensor device again, adjust their posture, or request manual confirmation. By setting the model confidence mechanism, the system can avoid over-dimming or incorrect dimming when the input data quality is insufficient.

[0233] The control target generation unit generates light environment control targets based on visual cognitive performance indicators and the current learning task type. Different learning tasks correspond to different target weights. For example, in reading tasks, improving reading efficiency, gaze stability, and reducing the risk of visual fatigue have higher weights; in memorization tasks, improving alertness and short-term memory performance have higher weights; in screen learning tasks, reducing the contrast between screen and ambient brightness, reducing glare, and reducing the risk of visual fatigue have higher weights; and in rest and recovery tasks, reducing stimulation, improving visual comfort, and promoting fatigue recovery have higher weights. Through this unit, the system can transform the "detected learning state" into "control targets that can be used for light parameter optimization."

[0234] The safety constraint and lighting environment multi-parameter optimization decision-making module is used to evaluate the expected impact of different candidate light parameter combinations on visual cognitive performance within the eye safety boundary, based on a lighting environment response prediction model, and to generate optimal or suboptimal lighting environment parameter control instructions based on the multi-objective benefit evaluation results. The safety constraint and lighting environment multi-parameter optimization decision-making module includes a safety constraint management unit, a candidate light parameter generation unit, a lighting environment response prediction unit, a multi-objective benefit evaluation unit, a control strategy selection unit, and a control instruction generation unit.

[0235] The safety constraint management unit is used to establish light safety boundaries in the personal learning space of teenagers. These light safety boundaries include, but are not limited to, upper limit constraints on illuminance, vertical illuminance constraints for the eyes, constraints on abrupt changes in brightness, constraints on the amplitude of color temperature changes, constraints on the rate of color temperature changes, constraints on the proportion of short-wavelength components, constraints on flicker frequency, constraints on modulation depth, constraints on PWM parameters, constraints on light field switching frequency, constraints on screen brightness coordination, and constraints on nighttime rhythm protection. Through these constraints, the system can avoid generating excessively strong, cold, flickering, or rapidly changing light stimuli in order to temporarily enhance attention or alertness.

[0236] The candidate light parameter generation unit generates multiple combinations of candidate light environment parameters based on the current light environment status, learning task type, individual baseline, and historical control effects. These candidate light environment parameter combinations may include different desktop illuminance, vertical eye illuminance, correlated color temperature, spectral composition, short-wavelength component ratio, background light ratio, screen brightness, spatial light distribution, gradation time, and flicker control parameters. Candidate parameters that do not meet the safety constraints for adolescent eye use, such as excessively high illuminance, excessively rapid color temperature changes, excessively high short-wavelength components, high flicker risk, excessively large brightness abrupt changes, or excessive nighttime stimulation, are directly eliminated by the system.

[0237] The light environment response prediction unit is used to predict changes in visual cognitive state after the execution of candidate light environment parameter combinations. Specifically, this unit receives information such as the current visual cognitive performance state, current light environment parameters, candidate light environment parameter combinations, learning task type, individualized baseline, historical adjustment records, and the current time period. It inputs this information into a pre-trained or online-updated light environment response prediction model and outputs the prediction results for each candidate light environment parameter combination within a future feedback window. The prediction results include, but are not limited to, predicting the comprehensive visual cognitive performance index, predicting changes in visual attention, predicting changes in reading efficiency, predicting changes in short-term memory performance, predicting visual fatigue risk, predicting visual discomfort risk, predicting visual comfort, and predicting model confidence. The future feedback window can be 3 minutes, 5 minutes, or other preset durations after the light environment adjustment is completed. The light environment response prediction model preferably adopts a regression prediction model based on LightGBM. The inputs of this model include the current visual cognitive state, current light environment parameters, candidate light environment parameter combinations, changes in light parameters, learning task type, individualized baseline, and historical adjustment records; the outputs are the changes in visual cognitive performance, visual fatigue risk, visual comfort, and visual discomfort risk of the candidate scheme within the future feedback window. Through this light environment response prediction unit, the system can evaluate the expected effects of multiple candidate combinations of illuminance, color temperature, spectrum, short-wavelength component ratio, spatial light distribution, and screen brightness before actual dimming is performed, thus providing a clear basis for multi-objective optimization.

[0238] The multi-objective benefit evaluation unit is used to comprehensively evaluate candidate solutions that meet safety constraints based on factors such as visual cognitive performance, visual comfort, risk of visual fatigue, and risk of visual discomfort, and select the optimal or suboptimal combination of light environment parameters. In embodiments of the present invention, the system adopts an optimization method of "candidate set generation—safety constraint screening—LightGBM response prediction—objective function scoring and ranking," rather than relying on fixed dimming rules. Through this unit, the system can balance improving learning-related visual cognitive performance with ensuring eye safety, rather than simply adjusting a single illuminance or color temperature parameter.

[0239] The control strategy selection unit is used to select a control strategy based on the visual cognitive performance estimation results, model confidence, and the safety status of the target object. When the learner's attention decreases but the risk of visual fatigue is low, the system can moderately increase the desktop illuminance or increase the neutral to cool light component; when the risk of visual fatigue increases, the blink rate increases, or the pupil recovery is abnormal, the system can reduce the brightness stimulation, adjust to a soft color temperature, increase background illumination, or prompt a short rest; when screen learning is detected and the contrast between the screen and the ambient brightness is large, the system can coordinately adjust the screen brightness and the ambient brightness to reduce glare and contrast; when the model confidence is insufficient or the risk of discomfort increases, the system prioritizes a conservative gradual change strategy or a protection strategy.

[0240] The control command generation unit encapsulates the optimization results into light environment parameter control commands. These control commands include at least the target desktop illuminance, target eye vertical illuminance, target brightness, target correlated color temperature, target spectral composition, target short-wavelength component ratio, target spatial light distribution, target screen brightness, target flicker safety control parameters, target PWM safety control parameters, rate of change, fading time, execution area, and execution priority. Through this module, the system can convert the visual cognitive state output by the LightGBM visual cognitive performance estimation model into executable light environment control commands.

[0241] The optical environment execution feedback and individualized update module executes optical environment control commands and performs closed-loop feedback, individualized updates, and data security management based on the adjusted learner state and actual light output results. This module includes an optical environment execution unit, a smooth transition control unit, an actual execution verification unit, a closed-loop feedback evaluation unit, an individualized model update unit, and a storage, communication, and privacy protection unit.

[0242] The lighting environment execution unit receives lighting environment parameter control commands and drives dimmable LED lights, adjustable color temperature lights, adjustable spectrum lights, desktop study lights, background lights, auxiliary side lights, overhead lights, multi-source light source arrays, and display terminals to perform corresponding adjustments. These adjustments include illuminance adjustment, brightness adjustment, correlated color temperature adjustment, spectral composition adjustment, short-wavelength component ratio adjustment, spatial light distribution adjustment, display terminal brightness adjustment, and flicker or PWM parameter adjustment. For different learning areas, the system can adjust desktop accent lighting, background fill light, side fill light, and ambient light separately to create a light field distribution suitable for the current task.

[0243] The smooth transition control unit controls the rate and manner of change of illuminance, brightness, color temperature, and spectral parameters, ensuring that the lighting environment changes gradually according to linear, exponential, S-shaped, or other preset curves, avoiding sudden brightening, sudden darkening, rapid color temperature jumps, or frequent switching of the light field. This unit is particularly suitable for study spaces for teenagers, reducing the interference of sudden changes in the lighting environment on visual comfort and the continuity of learning.

[0244] The actual execution verification unit receives the actual light output data fed back by the light environment measurement and acquisition unit, and determines whether the actual output of the lamps or display terminals has reached the target value. If there are deviations between the actual desktop illuminance, eye vertical illuminance, brightness, color temperature, spectrum, flicker parameters, or screen brightness and the target value, the system can perform compensation control, or feed back the deviation information to the light environment multi-parameter optimization decision module to regenerate control commands.

[0245] The closed-loop feedback evaluation unit is used to continue collecting brain-computer interface signals, multimodal physiological signals, task performance data, and actual light output data after the light environment adjustment is completed, and to compare the changes in visual cognitive performance before and after the adjustment. The evaluation indicators include changes in the comprehensive visual cognitive performance index, changes in the visual attention index, changes in reading speed, changes in reaction time, changes in short-term memory performance, changes in visual fatigue risk, changes in blink rate, subjective discomfort feedback, and changes in task accuracy. If the comprehensive visual cognitive performance index increases after adjustment, and the risks of visual fatigue and discomfort do not increase, the system marks the current light environment parameters as an effective strategy; if attention continues to decline, reaction time fluctuations increase, task accuracy decreases, or the risk of visual fatigue increases after adjustment, the system determines that the current strategy is ineffective and regenerates a light environment adjustment plan; if obvious discomfort occurs, the continuous learning time is too long, or the head is too close, the system prioritizes triggering a rest reminder or soothing the light environment.

[0246] The individualized model update unit is used to update the individual baselines, model parameters, and strategy weights of adolescents under different time periods, task types, and lighting conditions based on long-term usage data. The system can gradually develop a library of learning lighting strategies suitable for specific adolescents. For example, a lower-stimulation background lighting strategy is more suitable for the target individual when studying on a screen at night; a higher desktop illuminance and neutral-to-cool color temperature strategy is more suitable when reading on paper in the morning; and a warm, soft, and low-rate-of-change recovery strategy is more suitable when fatigue risk increases. Through this unit, the system can be expanded from a one-time dimming system to a long-term adaptive, personalized closed-loop lighting environment control system.

[0247] The storage, communication, and privacy protection unit stores the target object's baseline, raw acquisition data, preprocessed data, feature vectors, brain-computer interface model parameters, lighting environment parameters, control commands, execution records, feedback results, and a personalized strategy library. It also enables data communication between the brain-computer interface device, lighting environment sensor, light source controller, display terminal, processor, mobile application, and cloud server. The communication methods may include wired communication, Bluetooth, Wi-Fi, ZigBee, Matter, local area network, or other IoT communication methods. For the target object's brain-computer interface data, physiological data, and learning behavior data, the system can employ localized processing, anonymized storage, de-identification processing, access control, and encrypted transmission for privacy protection. Preferably, core physiological data and brain-computer interface data undergo primary processing locally, with anonymized statistical results, model update amounts, or control records only uploaded upon authorization.

[0248] Furthermore, the closed-loop control system for the light environment parameters of the brain-computer interface for learning spaces can also include hardware and basic components to support the coordinated operation of the above five modules. The hardware and basic components include at least a non-invasive brain-computer interface acquisition device, eye-tracking and pupil-tracking devices, heart rate or PPG acquisition devices, skin conductance acquisition devices, posture acquisition devices, a light environment sensor, a flicker detection sensor, a screen brightness detection device, a light source controller, dimmable lamps, adjustable color temperature lamps, adjustable spectrum lamps, a desktop study lamp, a background lamp, a display terminal, a processor, a memory, a synchronization clock, and a communication interface. The processor is used to perform data preprocessing, time synchronization, dual-timescale feature extraction, brain-computer interface model inference, visual cognitive performance calculation, safety constraint judgment, light environment parameter optimization, and control command generation. The memory is used to store the target object baseline, raw data, preprocessed data, feature vectors, model parameters, light environment parameters, control records, feedback results, and an individualized strategy library.

[0249] In embodiments of the present invention, a closed-loop control system for brain-computer interface optical environment parameters for learning spaces can be implemented based on control flow. The control flow includes "joint perception—synchronous parsing—LightGBM state estimation—candidate optical parameter generation—LightGBM response prediction—safety constraint screening and benefit ranking—illumination execution—feedback update". First, the human-light-task joint perception and baseline management module collects learner brain-computer interface signals, multimodal physiological signals, task behavior data, and current optical environment data, and reads the individualized baseline. Second, the synchronous preprocessing and dual-timescale feature parsing module unifies the above data onto the same timeline and obtains the actual light stimulus event time through actual light output phase correction, thereby extracting light-specific transient response features and stable cognitive state features respectively. Subsequently, the visual cognitive performance estimation and control target generation module outputs indicators such as visual attention, reading efficiency, short-term memory, reaction speed, cognitive load, and visual fatigue risk based on the LightGBM visual cognitive performance estimation model, and generates control targets according to the learning task type. After obtaining the control targets, the adolescent safety constraint and optical environment multi-parameter optimization decision module calculates or filters target optical environment parameter combinations within the safety constraint range and generates control instructions. Finally, the light environment execution feedback and individualized update module drives the lighting equipment and display terminal to perform smooth adjustments, while continuously collecting the learner's state and actual light output data after adjustment to determine whether the adjustment is effective and update the individual baseline, model parameters, and control strategy. These five steps together constitute a closed-loop control chain for the light environment parameters of a brain-computer interface for learning spaces.

[0250] Figure 5 A schematic diagram of the architecture of a closed-loop control system for optical environment parameters of a brain-computer interface for learning spaces, according to an embodiment of the present invention, is shown.

[0251] like Figure 5 As shown, through the coordinated cooperation among the human-light-task joint perception and baseline management module, the synchronous preprocessing and dual time-scale feature analysis module, the visual cognitive performance estimation and regulation target generation module, the safety constraint and light environment multi-parameter optimization decision module, and the light environment execution feedback and individualized update module, the closed-loop regulation of light environment parameters for brain-computer interface for learning space is finally realized.

[0252] In one specific embodiment of the present invention, a computer-readable storage medium is also provided, on which a computer program is stored. When the computer program is executed by one or more processors, it implements the above-described method for regulating the optical environment parameters of a brain-computer interface for learning spaces.

[0253] The computer-readable storage medium can be a non-transitory computer-readable storage medium, including but not limited to read-only memory, random access memory, flash memory, solid-state drive, hard disk, optical disk, embedded memory, built-in memory of the learning lamp controller, edge computing terminal memory, mobile terminal memory, or server memory. The processor can be a learning lamp controller processor, edge computing processor, mobile terminal processor, display terminal processor, server processor, or a combination of the above processors.

[0254] When a computer program is executed by a processor, it must perform at least the following functions.

[0255] First, obtain basic environmental information of the learning space, learning task information, and individualized baseline information of the target object.

[0256] Second, it collects learners' brain-computer interface signals, multimodal physiological signals, task behavior data, and light environment parameters in the learning space.

[0257] Third, the collected data is synchronized in a unified time, preprocessed, corrected for actual light output phase, and extracted with dual time scale features to obtain multimodal features for visual cognitive state recognition.

[0258] Fourth, the LightGBM-based brain-computer interface model is invoked to estimate learners' visual attention, alertness level, reading or visual search efficiency, short-term memory performance, cognitive load, visual fatigue risk, and overall visual cognitive performance.

[0259] Fifth, based on the visual cognitive performance estimation results, the current learning task type, the individualized baseline, and eye safety constraints, candidate light environment parameter combinations are generated, and the LightGBM-based light environment response prediction model is called to predict and evaluate the candidate schemes, thereby generating light environment parameter control instructions.

[0260] Sixth, the light environment parameter control command is sent to the light source controller, lamp driver, display terminal control interface or multi-light source array controller to adjust one or more of the following: illuminance, brightness, correlated color temperature, spectral composition, short-wavelength component ratio, spatial light distribution, flicker control parameters, PWM parameters and display terminal brightness.

[0261] Seventh, after the light environment adjustment is completed, continue to collect brain-computer interface signals, multimodal physiological signals, task behavior data and actual light output data to evaluate the regulation effect, and update the individualized baseline of the target object, brain-computer interface model parameters, light environment response prediction model parameters and light environment control strategy based on the feedback results.

[0262] Eighth, the baseline, model parameters, light environment parameters, control commands, execution records and feedback results of the target object are stored, and the brain-computer interface data, physiological data and learning behavior data involving the target object are localized, anonymized, and subject to access control or encrypted transmission.

[0263] Furthermore, when abnormalities are detected in the acquisition equipment, insufficient brain-computer interface signal quality, missing eye movement or pupil data, abnormal data from the light environment sensor, failure of the lighting fixture, model confidence level below a preset threshold, or discomfort reported by the learner, the computer program can pause the enhanced automatic dimming strategy, switch to a preset safe light environment, or trigger a rest prompt, a re-acquisition prompt, or a manual confirmation process.

[0264] Through the aforementioned storage medium scheme, when the computer program is executed by the processor, it can realize a closed-loop control process from "human-light-task data acquisition" to "visual cognitive state estimation", and then to "light environment parameter control, execution feedback and individualized update under safety constraints", so that the light environment of the learning space can be personalized, safe and adaptively adjusted according to the learner's real-time visual cognitive performance.

[0265] In summary, the embodiments of the present invention have the following advantages.

[0266] First, it achieves closed-loop control of the lighting environment guided by visual cognitive performance. This invention no longer adjusts lighting solely based on ambient illuminance, time of day, preset scene modes, or manual user selection. Instead, it uses a brain-computer interface model to estimate in real-time the visual attention, alertness level, reading efficiency, visual search efficiency, short-term memory performance, cognitive load, and risk of visual fatigue of adolescent learners, and translates these indicators into lighting environment parameter control commands. Thus, the lighting environment of the learning space can transform from traditional "fixed-pattern lighting" to a closed-loop regulation of "state recognition—parameter decision-making—lighting execution—feedback update," enabling the lighting environment to proactively serve the adolescent's reading, writing, memory, screen learning, and attention maintenance processes.

[0267] Secondly, this invention enhances the individualized adaptability of lighting in learning spaces for teenagers. By establishing individualized baseline profiles for target individuals and combining brain-computer interface signals, multimodal physiological signals, task behavior data, and historical control results, the system identifies differences in learners' responses to variations in contrast, color temperature, spectrum, brightness, and screen lighting environment. The system can adjust control strategies based on age group, learning task type, learning time period, susceptibility to visual fatigue, and historical feedback effects, thereby creating personalized learning lighting solutions for different teenagers and avoiding insufficient adaptation caused by using uniform, fixed parameters.

[0268] Third, it balances improved learning performance with eye safety. While optimizing visual cognitive performance, this invention introduces eye safety constraints, incorporating factors such as upper illuminance limits, vertical illuminance to the eyes, sudden brightness changes, color temperature variations, short-wavelength component ratios, flicker risk, modulation depth, screen brightness coordination, and nighttime circadian rhythm protection into the control boundaries. Therefore, the system avoids generating excessively strong, cold, flickering, or rapidly changing light stimuli to temporarily enhance alertness or attention, thus reducing visual fatigue, visual discomfort, and the risk of inappropriate light exposure while improving learning-related performance.

[0269] Fourth, this invention improves the accuracy and stability of dynamic light environment control. Through actual light output detection and phase correction mechanisms, it can correct the deviation between the timing of the light source control command issuance and the actual time of light exposure received by the learner, reducing the impact of lamp response delay, control link delay, sensor sampling delay, and changes in environmental reflection on state recognition. Simultaneously, this invention employs a dual-timescale feature analysis method using a light-specific response window and a stable cognitive window. This allows it to distinguish between transient reactions such as pupillary light reflection and visual evoked responses shortly after light changes and continuous cognitive state changes during the stable learning phase. This avoids misinterpreting simple light reflection as decreased attention or visual fatigue, thereby improving the accuracy and stability of closed-loop control decisions.

[0270] Fifth, it achieves joint optimization control of multiple lighting environment parameters. This invention does not merely adjust illuminance or color temperature individually, but rather jointly optimizes multiple parameters, including desktop illuminance, vertical eye illuminance, brightness, correlated color temperature, spectral composition, short-wavelength component ratio, spatial distribution of the light field, incident direction, dynamic change curve, flicker parameters, PWM parameters, and display terminal brightness. The system can make comprehensive decisions based on different learning tasks and user states, considering visual cognitive performance, visual comfort, fatigue risk, and eye safety, thereby achieving a more refined and stable lighting environment control effect for the learning space.

[0271] Sixth, it supports long-term feedback updates and continuous optimization. After each adjustment of the light environment, this invention continues to collect brain-computer interface signals, multimodal physiological signals, task behavior data, and actual light output data, and judges the effectiveness of the current strategy based on changes in visual cognitive performance before and after the adjustment. For effective combinations of light environment parameters, the system can increase their recommended weight in the same or similar tasks; for strategies that are ineffective or cause visual fatigue or discomfort, the system can reduce their recommended weight or trigger protective strategies. Through long-term feedback accumulation, this invention can continuously update the user's individual baseline, brain-computer interface model parameters, and light environment control strategies, enabling the lighting system to have long-term adaptive capabilities.

[0272] Seventh, enhancing its engineering application value in real-world learning scenarios. This invention can be deployed in various settings such as home study desks, dormitory study areas, personal study pods, study rooms, libraries, school classrooms, training spaces, smart campuses, and youth health lighting systems. The system can work collaboratively with dimmable, color-temperature, and spectrum-adjustable lamps, desktop study lights, background lights, multi-source arrays, and display terminals, adapting to various learning tasks such as paper reading, writing assignments, memorization, screen learning, exam preparation, and rest / recovery. Therefore, this invention has good engineering feasibility, scenario compatibility, and widespread application value.

[0273] Eighth, improve the security and controllability of data use. This invention incorporates storage, communication, and privacy protection mechanisms within the system, enabling localized processing, anonymized storage, de-identification, access control, and encrypted transmission of the target object's brain-computer interface data, physiological data, learning behavior data, and individual baseline data. This achieves closed-loop control of the personalized lighting environment while reducing the risk of sensitive data leakage and enhancing the data security and reliability of the intelligent lighting system for personal learning spaces.

[0274] The following describes embodiments of the present invention using a personal learning space as an application scenario, but the scope of protection of the present invention is not limited to the following embodiments. In the description of the single closed-loop control process, an example is taken of a teenager completing paper reading, memorization, mathematical calculations, visual search, and screen learning tasks at a home study desk. In the prototype verification section, test data from 15 learners in a home learning scenario are further used as a verification example. In the 15 prototype verifications, each learner completed individual baseline establishment, formal learning tasks, light environment control, closed-loop feedback recording, and comparative tests of different lighting modes according to the same system configuration and methodology.

[0275] The first step is to set up the home learning space and configure the hardware.

[0276] In this embodiment of the invention, the home learning space is an independent learning area of ​​approximately 10 m². The study desk measures 1200 mm × 600 mm. Above the desk is a set of adjustable LED study lights with adjustable brightness, color temperature, and spectrum, positioned approximately 650 mm from the desk. The lights allow for continuous adjustment of the desk illuminance from 200 lx to 800 lx, the correlated color temperature from 3000 K to 6500 K, and the short-wavelength component ratio from 0.15 to 0.32. A soft background fill light is installed behind the study desk, with an adjustable background illuminance from 50 lx to 250 lx, used to reduce the contrast between the desk, screen, and surrounding environment.

[0277] A desktop illuminance sensor is placed at the front edge of the study desk to detect the actual desktop illuminance in the area where books or exercise books are located. A brightness and color temperature sensor is placed near the display terminal to detect the field of view brightness, background brightness, and screen brightness. A small ambient light detection unit is placed near the learner's line of sight to collect the vertical illuminance at the eyes and the actual light output waveform. The illuminance sampling rate of the ambient light detection unit is set to 100 Hz, the flicker detection sampling rate is set to 2 kHz, and the color temperature and spectrum detection period is set to 1 s.

[0278] Learners wear a lightweight, non-invasive EEG headband to collect EEG signals from the frontal, parietal, or occipital regions at a sampling rate of 250Hz. An eye-tracking / pupil monitoring device collects fixation point, fixation duration, saccades, regressions, blink information, and pupil diameter at a sampling rate of 60Hz. A wristband collects heart rate, PPG, and electrodermal signaling (EDS) signals, with a PPG sampling rate of 64Hz and an EDS sampling rate of 32Hz. A learning task terminal presents reading comprehension, short-term memory, mathematical calculation, or visual search tasks and records task start time, task presentation time, accuracy rate, reaction time, missed answer rate, and incorrect answer rate. The sensors, learning light, display terminal, and edge computing processor synchronize data using a unified time synchronization protocol.

[0279] The second step is to establish a reference lighting environment and individual baselines.

[0280] When the system is used for the first time, it first enters the individual baseline establishment stage. In the embodiments of the present invention, the reference light environment is set as follows: desktop illuminance 500 lx, vertical eye illuminance 230 lx, correlated color temperature 4000 K, background supplemental lighting illuminance 120 lx, display terminal brightness 120 cd / m², screen color temperature 5000 K, short-wavelength component ratio 0.22, and flicker modulation depth not exceeding 3%. Under this reference light environment, the system collects 10 minutes of baseline data from learners, including 2 minutes of resting fixation, 4 minutes of paper reading, 2 minutes of visual search, and 2 minutes of short-term memory task. In the validation of 15 adolescent prototypes, the system performed the above baseline collection process on each learner and established an independent individualized baseline profile for each; the following values, such as reading speed, reaction time, blink rate, pupil diameter, EEG theta / β power ratio, visual fatigue risk index, and visual attention index, are examples of the individual baseline of one representative learner.

[0281] In an embodiment of the present invention, the individual baseline data obtained by the system are as follows: reading speed baseline is 420 words / minute, average reaction time baseline is 780ms, task accuracy baseline is 88%, blink rate baseline is 14 times / minute, average pupil diameter baseline is 3.2mm, pupillary light reflex recovery time baseline is 0.62s, EEG theta / β power ratio baseline is 0.78, visual fatigue risk index baseline is 0.20, and visual attention index baseline is 0.72. The above baseline data is written into the individualized baseline profile of the target subject for subsequent state estimation, light environment parameter optimization, and closed-loop feedback updates. Furthermore, the system does not directly use the group average threshold to judge the learner's state, but establishes an individual baseline for each target subject. Through this individual baseline correction, the system can determine the degree of deviation of current indicators such as EEG, eye movement, pupil size, heart rate, reaction time, or reading efficiency from the adolescent's normal state.

[0282] The system also sets eye safety constraints for teenagers. In embodiments of the invention, the adjustable range of desktop illuminance is limited to 300 lx to 750 lx, and the vertical illuminance to the eyes does not exceed 350 lx; ​​the daytime range of correlated color temperature is 3500 K to 6000 K, and after 20:30 at night it is limited to 3000 K to 4500 K; the single change in desktop illuminance does not exceed 150 lx, the single change in color temperature does not exceed 800 K, and the smooth transition time of the light environment is not less than 30 s; the proportion of short-wavelength components does not exceed 0.30 during the day and 0.24 at night; the brightness of the display terminal is limited to the range of 90 cd / m² to 160 cd / m², and is matched with the ambient brightness. When an increased risk of visual fatigue is detected or the learner actively reports discomfort, the system prioritizes the protection strategy instead of continuing to implement the enhanced light stimulation strategy.

[0283] The third step is to perform data synchronization and phase correction for the formal learning phase.

[0284] In the formal learning phase, learners first complete a 20-minute paper-based reading task. The system continuously collects data on EEG, eye movement, pupil size, heart rate, skin conductance, task performance, and light environment parameters, and records the time of the light source control command issuance, the actual light emission time, and the task event time. In the verification of 15 prototypes, each learner completed paper-based reading, screen learning, and corresponding task performance recording according to the same task flow; the system performs actual light emission phase correction, dual-timescale window segmentation, and visual cognitive performance estimation for each learner. To ensure an accurate correspondence between changes in the light environment and the learner's physiological response, the system performs phase correction based on the actual light output waveform. For example, in an embodiment of the present invention, if the trigger time of a dimming command is 19:32:10.000, and the light environment detection unit detects that the actual light output waveform reaches 50% of the target change at 19:32:10.180, then: The system thus determines the corrected light stimulation event time as 19:32:10.180, instead of directly using the control command issuance time as the physiological response analysis anchor point. This method reduces the impact of lamp response delay and control link delay on state recognition. The system further performs dual-timescale window segmentation based on the corrected light stimulation event time to obtain a light-specific response window and a stable cognitive window. In embodiments of the present invention, , The light-specific response window is used to extract transient features such as pupillary light reflex, visual evoked response, and short-term EEG response; the stable cognition window is used to extract features related to attention, reading efficiency, reaction time, blink rate, and visual fatigue during the continuous learning phase.

[0285] The fourth step is to estimate visual cognitive performance and perform enhanced dimming under reading tasks.

[0286] The following example, using a representative learner from 15 prototype verifications, illustrates the state estimation, candidate light parameter prediction and evaluation, safety constraint screening, and illumination execution processes in a single learning cycle of this invention. Other learners follow the same process, but their individual baselines, model input features, candidate parameter gains, and final control instructions may differ due to individual variations. Within the first 8 minutes after the start of the paper reading task, the system detects that the learner's state is relatively stable, with desktop illuminance maintained at 500 lx, correlated color temperature maintained at 4000 K, and background supplemental lighting illuminance at 120 lx. At this time, the LightGBM visual cognitive performance estimation model outputs a visual attention index. Short-term memory performance index Reading efficiency index Visual fatigue risk index Visual discomfort risk index For paper-based reading tasks, embodiments of the present invention are configured as follows: , , , , Based on this, the overall visual cognitive performance index in the initial stage of the reading task was approximately 0.55.

[0287] When reading reached the 12-minute mark, the system detected that the learner's visual attention index had decreased to 0.58, reading efficiency index to 0.62, average fixation duration increased from 250 ms to 285 ms, and the number of re-views increased from 14 times / minute to 18 times / minute. However, the visual fatigue risk index was only 0.26, still below the preset fatigue protection threshold of 0.40. The system determined that the current state was "decreased attention but low fatigue risk," and therefore entered the candidate lighting environment parameter generation and lighting environment response prediction process, instead of directly executing dimming according to fixed rules. The system first generated multiple candidate lighting environment parameter combinations that met the safety constraints for teenagers, then called the LightGBM lighting environment response prediction model to predict the impact of each candidate scheme on visual cognitive performance, visual fatigue risk, and visual comfort within the future feedback window. Finally, the candidate scheme with the highest overall benefit and sufficient safety margin was selected for execution.

[0288] After evaluating candidate solutions, the system selected the option with higher overall benefits and lower fatigue risk as the execution plan. The desktop illuminance was adjusted from 500 lx to 620 lx, the correlated color temperature from 4000 K to 4800 K, the short-wavelength component ratio from 0.22 to 0.26, and the background illumination from 120 lx to 150 lx. These adjustments were completed within 60 seconds using a smooth S-shaped curve to avoid sudden brightening or rapid color temperature changes. Five minutes after the adjustments, the system detected an increase in reading speed from 420 words / minute to 455 words / minute, a decrease in average fixation duration to 248 ms, a decrease in the number of regressions to 15 times / minute, a recovery in the visual attention index to 0.66, and a visual fatigue risk index of 0.28, which did not exceed the safety threshold. Therefore, the system marked this combination of lighting parameters as an effective strategy for the current reading task and maintained this parameter combination in operation.

[0289] The fifth step is to implement protective dimming when the risk of fatigue increases.

[0290] After approximately 45 minutes of continuous learning, the system detected an increase in blink rate from 14 times / minute to 24 times / minute, pupillary light reflex recovery time from 0.62 s to 0.95 s, eye movement regression frequency to 23 times / minute, and EEG theta / β power ratio from 0.78 to 1.05. Simultaneously, reading speed decreased to 395 words / minute, and task accuracy decreased to 84%. The LightGBM visual cognitive performance estimation model outputs a visual fatigue risk index. Visual discomfort risk index The model confidence level was 0.86. The system determined that the learner had entered a state of increased fatigue risk. In this state, protective strategies take precedence over enhanced cognitive performance optimization strategies; even if certain candidate light parameter combinations might predict a short-term increase in attention, the system will discard the candidate scheme if its predicted fatigue or discomfort risk exceeds the safety threshold. At this point, the system will no longer execute the enhanced lighting strategy, but will trigger protective light environment control. Specifically, the system will smoothly reduce the desktop illuminance from 620 lx to 430 lx, reduce the correlated color temperature from 4800 K to 3500 K, reduce the proportion of short-wavelength components from 0.26 to 0.18, and increase the background supplemental lighting illuminance from 150 lx to 200 lx to reduce the brightness contrast between the desktop and the surrounding environment. This adjustment process lasts for 90 seconds, and a "Recommended rest for 5 minutes" prompt is simultaneously output on the learning lamp control interface.

[0291] Within 3 minutes of adjustment, the system continued to collect feedback data. The feedback showed that the blink rate decreased to 19 times / minute, the visual fatigue risk index decreased to 0.40, the visual discomfort risk index decreased to 0.25, and the subjective discomfort feedback changed from "mild discomfort" to "acceptable." Therefore, the system recorded this strategy as an effective protective strategy under fatigue conditions and increased its recommended weight in subsequent similar conditions.

[0292] The sixth step is to perform coordinated control of ambient light and display terminal during the screen learning task.

[0293] When learners switch from paper-based reading to screen-based learning tasks, the learning task terminal detects that the display terminal has entered a continuous use state. At this time, the system detects that the original brightness of the display terminal is 170 cd / m², which exceeds the preset screen brightness collaborative control range, and the background brightness is low, resulting in a large contrast between the screen and the surrounding environment. If the lighting environment of the paper-based reading mode is maintained, learners may experience screen glare, unstable gaze, or an increased risk of eye discomfort. Based on this, the system switches to the screen-based learning task objective, shifting the focus of lighting environment control from simply increasing desktop illuminance to the collaborative control of "screen brightness—background supplementary lighting—risk of visual fatigue." The system adjusts the display terminal brightness from 170 cd / m² to 130 cd / m², adjusts the screen color temperature from 5000 K to 4600 K, increases the background supplementary lighting illuminance from 150 lx to 220 lx, maintains the desktop illuminance within the range of 430 lx to 480 lx, and limits the correlated color temperature of the ambient lighting to no more than 4200 K. After adjustment, the brightness contrast between the screen and the background decreased, and the phenomena of prolonged staring and frequent blinking in the eye-tracking data were reduced. The system detected that the visual discomfort risk index decreased from 0.31 to 0.22 within 10 minutes of screen learning, and the task accuracy remained at around 90%.

[0294] This process demonstrates that the technical solution of the present invention can not only adjust the illuminance and color temperature of the learning lamp, but also coordinately control the brightness of the display terminal as part of the overall light environment parameters, thereby adapting to the scenario of frequent switching between paper materials and electronic screens in real learning processes.

[0295] The seventh step involves predicting the light environment response, optimizing multiple parameters, and generating control commands.

[0296] When generating dimming commands, the system constrains the light source driving frequency, PWM duty cycle, modulation depth, and measured light output fluctuation depth to ensure that the flicker frequency, fluctuation depth, and modulation depth of the actual output light meet preset safety limits. In embodiments of this invention, the system does not directly map the current visual cognition state to a single dimming strategy, but first generates a set of candidate light environment parameters. For each candidate parameter combination... The system first determines whether it meets the set of constraints for safe eye use by teenagers. For candidate solutions that meet the safety constraints, the system calls the LightGBM ambient light response prediction model to predict the visual cognitive response within the future feedback window, and finally selects the solution with the highest overall benefit from the set of candidate parameters that meet the safety constraints.

[0297] For example, when a decrease in attention but a low risk of fatigue is detected at the 12-minute mark of a reading task, the system generates three candidate solutions: Candidate Solutions For a desktop illuminance of 560 lx, a correlated color temperature of 4400 K, and a short-wavelength component ratio of 0.24; candidate schemes For desktop illuminance of 620 lx, correlated color temperature of 4800 K, and short-wavelength component ratio of 0.26; candidate schemes The desktop illuminance is 680 lx, correlated color temperature is 5200 K, and short-wavelength component ratio is 0.29. The system inputs the current visual cognitive state, current lighting environment parameters, candidate parameter combinations, reading task type, and individual baseline of the target object into the LightGBM lighting environment response prediction model. It then predicts the comprehensive visual cognitive performance, visual fatigue risk, and visual comfort of the three candidate solutions within a 5-minute feedback window. For example, the model predicts... It can improve the visual cognitive performance index by 0.04 and increase the risk of fatigue by 0.01; It can improve the visual cognitive performance index by 0.08 and increase the risk of fatigue by 0.03; It can improve the visual cognitive performance index by 0.10, but increases the fatigue risk by 0.09, and is close to the safety boundary of the short-wavelength component. After a comprehensive evaluation of the system's benefits, it was selected... As an optimal or near-optimal balance, rather than choosing the option that offers the highest improvement in cognitive performance prediction but carries a higher risk of fatigue. This demonstrates the process by which the present invention achieves candidate parameter evaluation and multi-objective optimization under safety constraints through a light environment response prediction model. In the validation of 15 prototypes, the system independently performed the above-mentioned candidate light environment parameter generation, LightGBM light environment response prediction, safety constraint screening, and comprehensive benefit ranking processes for each learner; the candidate schemes and their predicted benefits were determined based on each learner's individual baseline, current state, and historical control records.

[0298] The eighth step involves closed-loop feedback, individualized updates, and outputting prototype verification data.

[0299] After each light environment adjustment, the system continues to collect brain-computer interface signals, multimodal physiological signals, task behavior data, and actual light output data for 3 to 5 minutes to determine whether the adjustment was effective. If the overall visual cognitive performance improves after the adjustment, and the risk of visual fatigue and visual discomfort does not increase, the system marks the current parameter combination as an effective strategy; if the risk of fatigue increases, attention continues to decline, or the target object reports discomfort after the adjustment, the system reduces the weight of the strategy and regenerates a control plan.

[0300] Furthermore, the system can update the LightGBM light environment response prediction model based on feedback data after actual execution. For each executed combination of light environment parameters, the system records the visual cognitive state before adjustment. Current lighting environment parameters Light environment parameters have been implemented. Learning task type, individualized baseline of target audience, and predicted response results and the actual observation results after adjustment The system obtains the prediction error. Based on this prediction error, the system periodically retrains, fine-tunes online, or performs individualized calibration on the light environment response prediction model to improve the accuracy of subsequent candidate light parameter evaluation. The system can also update individual baselines, where baseline updates can be applied to one or more of the following: attention baseline, reading speed baseline, reaction time baseline, pupil baseline, blinking baseline, EEG frequency band baseline, or visual fatigue baseline. In embodiments of the present invention, the following is taken... For example, learner attention baseline The steady-state statistics after three consecutive effective readings of regulation The updated attention baseline is In the prototype verification, 15 learners aged 12-16 years with normal or corrected vision were selected for a controlled test in a home learning scenario. Each learner established an independent baseline and completed the test for 5 consecutive days; each day included 30 minutes of paper reading and 20 minutes of screen learning. Each learner experienced a fixed lighting environment mode, a preset reading mode, and the closed-loop control mode of this invention. The testing order of the three lighting methods could be arranged in a balanced or random order to reduce the influence of the sequence effect. The fixed lighting environment mode was set to a desktop illuminance of 500 lx and a correlated color temperature of 4000 K; the preset reading mode was set to a desktop illuminance of 550 lx and a correlated color temperature of 4500 K; the closed-loop control mode of this invention dynamically adjusted the light intensity within the range of 380 lx-650 lx, 3500 K-5000 K, and a short-wavelength component ratio of 0.18-0.27 based on the real-time status.

[0301] Test results are expressed as the within-group average of 15 learners. Under fixed lighting conditions, the average reading speed was [missing information]. Words per minute, task accuracy rate The average reaction time is The increase in the risk index of visual fatigue after 50 minutes of study is ms. In the preset reading mode, the average reading speed is: Words per minute, task accuracy rate The average reaction time is The visual fatigue risk index increased by ms. In the closed-loop control mode of this invention, the average reading speed is: Words per minute, task accuracy rate The average reaction time is The visual fatigue risk index increased by ms. The above results demonstrate that, under the conditions of the embodiments of the present invention, the present invention can improve reading efficiency, task accuracy, and reaction speed stability while significantly reducing the increase in the risk of visual fatigue.

[0302] Furthermore, to verify the effectiveness of actual light output phase correction and the dual-timescale window mechanism, the embodiments of this invention set two control conditions on a sample of 15 learners: First, without actual light output phase correction, the event time was directly taken as the trigger time of the light source control; second, without distinguishing between the light-specific response window and the stable cognitive window, a unified sliding window was directly used for state recognition. The control results showed that without phase correction, the model's recognition consistency rate for the "attention decline / fatigue increase" state was an average of 82.1%; without the dual-timescale window mechanism, the recognition consistency rate was an average of 80.6%; and with the complete scheme of this invention, the recognition consistency rate increased to an average of 89.4%, and the number of times pupil light reflection was misjudged as visual fatigue decreased from an average of 6 times per 50 minutes to an average of 2 times. This demonstrates that actual light output phase correction and dual-timescale feature analysis can improve the accuracy and stability of closed-loop control.

[0303] Furthermore, embodiments of the present invention can also be applied to multi-learner scenarios such as study rooms, libraries, or classrooms. In this case, the system can divide the space into multiple learning areas, such as a front-row reading area, a central screen learning area, a back-row study area, and a rest and recovery area. Each area is equipped with an independent light environment sensor and a zoned lighting control unit, and can generate a local light environment adjustment strategy based on the visual cognitive state of representative learners or student groups within that area.

[0304] For example, in a study room, the system can divide every four study seats into a control zone, with each zone equipped with a set of dimmable and color-temperature adjustable lights and a desktop illuminance detection node. If most learners in a zone are engaged in paper-based reading, the system can maintain the desktop illuminance in that area at 500 lx–650 lx and the correlated color temperature at 4000 K–5000 K. If a zone enters screen-based learning mode, the system can appropriately increase the background lighting ratio and simultaneously reduce the brightness of the display terminal or remind the target to adjust the screen brightness. If the overall fatigue risk in a zone increases, the system can reduce the stimulating light parameters and output a rest reminder. In this extended scenario, the system can achieve zoned, grouped, or representative individual-driven learning lighting optimization while ensuring overall spatial lighting uniformity, safety, and management convenience.

[0305] The embodiments of this invention illustrate that the method and system described herein can, in a real personal learning space, complete LightGBM visual cognitive performance estimation, light environment response prediction, candidate light parameter scoring and ranking under adolescent safety constraints, illumination execution, feedback evaluation, and individualized updates based on the learner's brain-computer interface state, multimodal physiological state, learning task performance, and actual light environment feedback. Therefore, this invention enables the learning space light environment to transform from a fixed lighting pattern into a closed-loop adaptive control system oriented towards adolescent visual cognitive performance, possessing good practicality, feasibility, and promotional value.

[0306] Those skilled in the art will understand that the features described in the various embodiments of the present invention can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, the features described in the various embodiments of the present invention can be combined and / or combined in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or combinations fall within the scope of the present invention.

[0307] The embodiments of the present invention have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of the invention. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of the invention, and all such substitutions and modifications should fall within the scope of the invention.

Claims

1. A method for regulating the optical environment parameters of a brain-computer interface for learning spaces, characterized in that, include: During the learning process of the target object, in response to the light environment adjustment event triggered in the learning space where the target object is located, feature extraction is performed based on the measured light output waveform collected from the learning space and the multimodal physiological signals collected from the target object to obtain light-specific transient response features and stable cognitive state features related to the learning task. The light environment adjustment event is triggered when the light source of the learning space is adjusted using the current combination of light environment parameters. The light-specific transient response characteristics and the stable cognitive state characteristics are input into the visual cognitive performance evaluation model to obtain the visual cognitive performance evaluation results of the target object. Based on the set of security constraints associated with the learning task, multiple candidate combinations of optical environment parameters are determined. For each candidate light environment parameter combination, the candidate light environment parameter combination, the current light environment parameter combination, the visual cognitive performance evaluation result, and the task type of the learning task are input into the light environment response prediction model to obtain the visual cognitive performance prediction result related to the candidate light environment parameter combination. Based on the visual cognitive performance prediction results associated with each of the multiple candidate light environment parameter combinations, a target light environment parameter combination is selected from the multiple candidate light environment parameter combinations so as to adjust the light source of the learning space using the target light environment parameter combination.

2. The method according to claim 1, characterized in that, The feature extraction based on the measured light output waveform acquired from the learning space and the multimodal physiological signals acquired from the target object yields light-specific transient response features and stable cognitive state features related to the learning task, including: The event feature points corresponding to the light environment adjustment event are extracted from the measured light output waveform to obtain the event feature point time. The optical link phase is corrected by using the triggering time of the light environment conditioning event to obtain the corrected light stimulation event time. Based on the corrected photostimulation event time, a photospecific response window is determined, and based on the photospecific response window, photospecific transient response features related to the learning task are extracted from the multimodal physiological signals. At least one stable cognitive window is determined based on the stable start time that satisfies the stability criterion in the measured light output waveform, and at least one steady-state sub-feature related to the learning task is extracted from the multimodal physiological signal based on at least one of the stable cognitive windows, wherein the stable cognitive state feature includes at least one of the steady-state sub-features.

3. The method according to claim 1, characterized in that, Also includes: Based on the task type of the learning task, the safety boundaries of the light environment parameter regulation behavior in multiple dimensions are determined, resulting in multiple sub-constraint sets. Based on multiple sets of sub-constraints, a set of security constraints related to the learning task is obtained.

4. The method according to claim 3, characterized in that, The plurality of sub-constraint sets includes a strobe safety constraint subset; The step of determining multiple candidate combinations of optical environment parameters based on a set of security constraints related to the learning task includes: Multiple optical environment parameters to be optimized are initialized to obtain an initial optical environment parameter combination, wherein the initial optical environment parameter combination includes multiple flicker safety control parameters related to the flicker safety constraint subset; When the multiple flicker safety control parameters satisfy the multiple flicker safety constraint conditions included in the flicker safety constraint subset, the initial light environment parameter combination is matched with the multiple safety constraint conditions included in the safety constraint set to obtain the matching result; If the matching result indicates that the initial combination of light environment parameters matches all of the multiple security constraints included in the security constraint set, then the initial combination of light environment parameters is determined as the candidate combination of light environment parameters.

5. The method according to claim 1, characterized in that, The step of inputting the candidate light environment parameter combinations, the current light environment parameter combinations, the visual cognitive performance evaluation results, and the task type of the learning task into the light environment response prediction model to obtain visual cognitive performance prediction results related to the candidate light environment parameter combinations includes: Based on the visual cognitive performance evaluation results, the current light environment parameter combination, the candidate light environment parameter combination, the change between the candidate light environment parameter combination and the current light environment parameter combination, the task type encoding of the learning task, the individualized baseline of the target object, and the historical control records of the learning space, a candidate scheme input vector is generated. The candidate scheme input vector is input into the light environment response prediction model to obtain the predicted response change. The visual cognitive performance evaluation result is superimposed with the predicted response change to obtain the visual cognitive performance prediction result related to the candidate light environment parameter combination.

6. The method according to claim 5, characterized in that, The visual cognitive performance evaluation result includes the current state of each of the multiple control targets, and the predicted response change includes the state change of each of the multiple control targets. The step of superimposing the visual cognitive performance evaluation result with the predicted response change to obtain the visual cognitive performance prediction result related to the candidate light environment parameter combination includes: The current state and state change of each of the multiple control targets are superimposed to obtain the predicted state of each of the multiple control targets; Based on the predicted states of each of the multiple control targets, a visual cognitive performance prediction result related to the candidate light environment parameter combination is obtained.

7. The method according to claim 1, characterized in that, The step of selecting a target light environment parameter combination from multiple candidate light environment parameter combinations based on visual cognitive performance prediction results associated with each of the multiple candidate light environment parameter combinations includes: Based on the task type of the learning task and the visual cognitive performance prediction results related to the candidate light environment parameter combination, the regulation score related to the candidate light environment parameter combination is obtained. The target light environment parameter combination is obtained by screening from the multiple candidate light environment parameter combinations based on the maximum control score among their respective control scores.

8. The method according to claim 7, characterized in that, The visual cognitive performance prediction results include the prediction states of multiple control targets; The process of obtaining a control score related to the candidate light environment parameter combination based on the task type of the learning task and the visual cognitive performance prediction results related to the candidate light environment parameter combination includes: Based on the task type of the learning task, determine the respective control weights of the multiple control targets; By utilizing the respective control weights of the multiple control targets, the predicted states of the multiple control targets are weighted and summed to obtain the control score related to the candidate light environment parameter combination.

9. The method according to claim 1, characterized in that, Also includes: In response to a target light environment adjustment event, based on the target measured light output waveform and target multimodal physiological signals corresponding to the target light environment adjustment event, and the task behavior data of the target object, the impact of the triggering of the target light environment adjustment event on visual cognitive performance and visual fatigue risk is determined, and a closed-loop feedback evaluation result is obtained. The target light environment adjustment event is triggered when the light source of the learning space is adjusted using the target light environment parameter combination. If the closed-loop feedback evaluation result indicates a decline in visual cognitive performance after regulation, or an increase in the risk of visual fatigue after regulation, a new combination of light environment parameters is generated.

10. The method according to claim 1, characterized in that, Also includes: Based on the visual cognitive performance evaluation results of the target object, the model confidence score is obtained, wherein the model confidence score represents the degree of credibility of the visual cognitive performance evaluation results; If the visual cognitive performance evaluation results meet the protection conditions, or if the model confidence is lower than the confidence threshold, the light source of the learning space is adjusted based on the protection strategy, and the adjustment of the light environment parameters is suspended. The adjustment of the light source in the learning space based on the protection strategy includes at least one of the following: Reduce brightness stimulation, switch to a softer color temperature, increase background fill light, and reduce the rate of change of lighting environment parameters.

Images