Self-adaptive illumination system and method for promoting cognitive efficiency based on task cognitive type

Abstract

The invention discloses a self-adaptive illumination system and method for promoting cognitive efficiency based on task cognitive types, and relates to the technical field of intelligent illumination, and the method comprises the steps that a multi-mode task recognition module obtains user behavior data in an intelligent self-adaptive mode, extracts behavior characteristics of the user behavior data, and judges the task cognitive types of the user based on the behavior characteristics; the central processing and control module obtains an optimal lighting parameter through a reinforcement learning algorithm based on a task cognition type and user basic information in an intelligent self-adaptive mode, and sends a control instruction to the dimmable lighting module based on the optimal lighting parameter; the dimmable lighting module receives the control instruction and adjusts lighting parameters of the lamp according to the control instruction; the on-line efficiency verification and strategy optimization module quantifies the working memory efficiency of a user in the current light environment based on a micro-task in an intelligent adaptive mode, feeds back a working memory efficiency evaluation result to the central processing and control module, carries out algorithm optimization, and dynamically provides an optimal light environment capable of directly improving the working memory efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent lighting technology, and in particular to an adaptive lighting system and method for improving cognitive performance based on task cognition type. Background Technology

[0002] The statements in this section are merely background information relating to this disclosure and do not necessarily constitute prior art.

[0003] Indoor artificial lighting is an indispensable element in modern work and study environments. In existing technologies, lighting systems mainly focus on meeting basic visual function requirements (such as sufficient illuminance and no glare) and creating a comfortable atmosphere (such as adjustable color temperature and brightness).

[0004] Some advanced intelligent lighting systems can automatically adjust their lighting based on ambient light, user presence, time, or simple user commands (such as "reading mode" or "rest mode") and simple control strategies. For example, one existing technology proposes a lighting control system for health and improved task efficiency that dynamically adjusts the spectrum by sensing user presence and time progression, primarily aiming to protect eye health and maintain basic task focus. However, this system has significant limitations: first, it only responds to coarse-grained information such as "user presence" and "duration," failing to perceive the cognitive attributes of the task the user is performing (e.g., whether it's a memory-storage task or an executive function task); second, its adjustment strategy is unidirectional and fixed (e.g., reducing blue light over time), unable to provide personalized and precise adjustments based on the inherent cognitive needs of different tasks and individual user differences (e.g., gender). Another existing technology focuses on improving the environmental adaptability and stability of color temperature adjustment by assessing the complexity of ambient light. However, its core technology lies in enabling artificial light and ambient light to coexist harmoniously, rather than optimizing the user's inherent cognitive efficiency. The user instructions it receives (such as scene modes) are actively set by the user and are static. The system itself does not have the ability to actively identify the user's cognitive state and provide optimal cognitive support. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, this invention provides an adaptive lighting system and method for improving cognitive efficiency based on task cognition type. The aim is to enable the lighting system to actively identify the cognitive processing type of the user's task, and based on this and individual characteristics such as the user's gender, dynamically match the optimal lighting parameters through reinforcement learning algorithms, thereby specifically improving working memory efficiency.

[0006] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: In a first aspect, the present invention provides an adaptive lighting system for promoting cognitive effectiveness based on task cognitive type, comprising: The user selection and authorization module is used to receive the lighting mode selected by the user, including intelligent adaptive mode and manual setting mode; The multimodal task recognition module is used to acquire user behavior data and extract its behavioral features in intelligent adaptive mode, and to determine the user's task cognition type based on the behavioral features. The user feature input module is used to obtain basic user information; The central processing and control module is used to obtain the optimal lighting parameters based on the task cognition type and user basic information in intelligent adaptive mode through reinforcement learning algorithm, and then issue control commands to the dimmable lighting module based on the optimal lighting parameters. A dimmable lighting module is used to receive control commands and adjust the lighting parameters of the lamps according to the control commands; The online performance verification and strategy optimization module is used to quantify the user's working memory efficiency under the current light environment based on micro-tasks in intelligent adaptive mode, and feed the working memory efficiency evaluation results back to the central processing and control module for algorithm optimization.

[0007] In a further technical solution, the multimodal task recognition module includes a software behavior analysis subunit and a behavior sensing subunit, and the user behavior data includes software behavior and keyboard behavior.

[0008] In a further technical solution, the software behavior analysis subunit is used to monitor the user's window activities and input device behavior, convert the window activities and input device behavior into quantitative features, and determine the task cognition type based on the quantitative features.

[0009] In a further technical solution, the behavior sensing subunit is used to acquire an audio stream, extract multi-dimensional acoustic features from the audio stream, and determine the task cognition type based on the multi-dimensional acoustic features.

[0010] A further technical solution involves inputting the judgment results output by the software behavior analysis subunit and the behavior sensing subunit into the decision engine, and assigning a confidence level to each judgment result using a weighted voting method to comprehensively derive the final task cognition type.

[0011] In a further technical solution, the state in the reinforcement learning algorithm is task cognitive type-user gender, the action is a combination of illuminance and color temperature, and the reward function is composed of a weighted average of efficiency change, comfort, and stability penalties. The efficiency change is expressed as:

[0012] in, For the accuracy of micro-tasks, For the reaction time of microtasks, This represents the historical efficiency baseline value for the corresponding user under the same cognitive state.

[0013] A further technical solution is that the online performance verification and strategy optimization module includes a micro-task implantation and evaluation sub-unit. This sub-unit periodically pops up cognitive micro-tasks and records the accuracy and reaction time of the user in completing the cognitive micro-tasks. Based on the accuracy and reaction time, the user's working memory efficiency under the current light environment is quantified, and an efficiency score is calculated.

[0014] Secondly, this invention provides an adaptive illumination method for promoting cognitive efficacy based on task cognitive type, including: It receives the user's selected lighting mode, including intelligent adaptive mode and manual setting mode; In intelligent adaptive mode, user behavior data is acquired and its behavioral features are extracted, and the user's task cognition type is determined based on the behavioral features. Obtain basic user information; In intelligent adaptive mode, based on task cognitive type and user basic information, the optimal lighting parameters are obtained through reinforcement learning algorithm, and control commands are sent to the dimmable lighting module based on the optimal lighting parameters. Receive control commands and adjust the lighting parameters of the lamps accordingly; In intelligent adaptive mode, the user's working memory efficiency is quantified based on micro-tasks under the current light environment, and the working memory efficiency evaluation results are fed back to the central processing and control module for algorithm optimization.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the adaptive illumination method for promoting cognitive efficacy based on task cognitive type as described in the second aspect.

[0016] Fourthly, the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the adaptive illumination method for promoting cognitive efficacy based on task cognitive type as described in the second aspect.

[0017] The above one or more technical solutions have the following beneficial effects: The system of this invention actively identifies the cognitive processing type of the user's task through a multimodal task recognition module, and based on this and individual characteristics such as the user's gender, dynamically matches the optimal lighting parameters in the central processing and control module through a reinforcement learning algorithm, thereby specifically improving working memory efficiency.

[0018] This invention provides multiple lighting modes, including intelligent adaptive mode and manual setting mode, allowing users to flexibly select the lighting mode according to the actual activity type, thus improving the flexibility of lighting.

[0019] This invention achieves both precision and seamlessness in cognitive perception. Through multimodal information fusion (software behavior, acoustic features), it can accurately and seamlessly identify the cognitive attributes of user tasks, overcoming the limitations of existing technologies that only perceive the presence of the user.

[0020] This invention achieves dynamic and personalized control strategies. By employing reinforcement learning algorithms and combining them with online performance verification, the system can dynamically learn and self-optimize the optimal lighting formula for each user, surpassing the static strategies based on fixed scene patterns or ambient light compensation in existing technologies, and achieving true personalized adaptation.

[0021] This invention achieves both scientific validity and verifiability in performance enhancement. By embedding microtasks and conducting online evaluations, it provides objective, real-time data support for the technological effects that improve working memory efficiency. This transforms the technological effects from subjective claims into a quantifiable and verifiable closed-loop process within the system. This significantly enhances the persuasiveness and ingenuity of the solution.

[0022] The system of this invention has self-evolving capabilities. It is no longer a fixed product from the factory, but an "intelligent partner" that can continuously adapt to the user's individual cognitive characteristics and learning / work habits over time, thus possessing long-term value. Attached Figure Description

[0023] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0024] Figure 1 This is a schematic diagram of the architecture of an adaptive lighting system for improving cognitive effectiveness based on task cognition type, according to an embodiment of the present invention. Figure 2 This is a flowchart of an adaptive lighting method for improving cognitive efficacy based on task cognition type, according to an embodiment of the present invention. Figure 3 This refers to the accuracy of working memory for subjects of different genders under different lighting conditions when presented simultaneously in embodiments of the present invention. Figure 4 This refers to the accuracy of working memory of subjects of different genders under different lighting conditions during sequence presentation in embodiments of the present invention. Detailed Implementation

[0025] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0026] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0027] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0028] The specific technical issues are: to solve the problem that existing lighting systems cannot seamlessly and accurately identify the cognitive processing type of a user's current task (e.g., simple tasks that rely primarily on memory storage vs. complex tasks that require central execution functions); and to solve the problem that existing systems lack a dynamic optimization and online verification mechanism based on the combination of "task type - user gender" when adjusting lighting, thus failing to ensure that the provided lighting environment can indeed improve the user's current working memory efficiency.

[0029] Example 1 like Figure 1 As shown, this embodiment discloses an adaptive lighting system for improving cognitive effectiveness based on task cognition type, including: The user selection and authorization module is used to receive the lighting mode selected by the user, including intelligent adaptive mode and manual setting mode.

[0030] In this embodiment, the user can select the lighting mode during the initialization phase. The user selection and authorization module provides the user with a clear interface for mode selection and data authorization to ensure that the system complies with relevant laws and regulations on personal information protection and to ensure the user's right to choose independently.

[0031] When users first use the system, they can choose between two lighting modes: Intelligent Adaptive Mode and Manual Setting Mode. Intelligent Adaptive Mode: In this mode, the system will enable intelligent control functions throughout the entire process, including multimodal task recognition, online performance verification, and strategy optimization. Choosing Intelligent Adaptive Mode indicates that the user acknowledges and agrees to the system collecting and analyzing relevant behavioral and environmental data (such as software usage, keyboard sounds, etc.) within necessary limits to provide personalized lighting optimization services. Manual Setting Mode: Users can completely skip the intelligent recognition and adaptive adjustment process and directly set lighting parameters (illuminance, color temperature, etc.) manually according to their personal preferences. In Manual Setting Mode, the system does not collect any user behavior, sound, or other data that may involve personal privacy; it is used only as a regular dimmable lighting device.

[0032] Users can switch between the two lighting modes at any time in the system settings. In intelligent adaptive mode, the system will prompt for authorization again before the first use of each data collection function (such as window monitoring and audio feature analysis) to ensure the user's informed consent. All data is processed locally on the device and is not uploaded to the cloud. Furthermore, the locally cached behavioral data will be cleared when the user switches to manual mode or exits the system.

[0033] The system's multiple lighting modes make it suitable not only for office and study scenarios involving electronic devices, but also for scenarios without electronic devices (such as reading paper books or writing notes). Users can freely choose the lighting mode according to the type of activity: when performing tasks related to electronic devices, they can choose the intelligent adaptive mode to obtain cognitively optimized lighting; when performing tasks without electronic devices or when privacy is a concern, they can choose the manual setting mode to adjust the lighting environment themselves.

[0034] The multimodal task recognition module is used to acquire user behavior data and extract its behavioral features in intelligent adaptive mode, and determine the user's task cognition type based on the behavioral features; user behavior data includes software behavior and keyboard behavior. This module infers the current cognitive process type by jointly using multiple data sources.

[0035] In this embodiment, the multimodal task recognition module includes a software behavior analysis subunit and a behavior sensing subunit. Task recognition types include performance-oriented tasks and memory-based tasks.

[0036] The software behavior analysis subunit monitors the window titles, process names, and micro-patterns of user interactions with active applications. For example, if it detects that a user is frequently switching windows, using a programming IDE or complex data analysis tools, combined with frequent keyboard shortcut usage and sporadic mouse click patterns, it can be identified as "performing a functional task"; while when a user spends a long time continuously typing in a document editor or spreadsheet, it is identified as "a memory-based task".

[0037] Furthermore, based on software behavior analysis, the user's current task cognition type is inferred. The specific steps are as follows: (1) Monitor window activity. On Windows, macOS, and Linux, the user32 API, CGWindowList, and X11 library can be used to listen for events such as window focus switching, title changes, and process names. The core is to record the timestamp, window title, and process ID / name.

[0038] (2) Monitor input devices. Use cross-platform libraries (such as uiohook) to listen for global keyboard key presses (key codes, press / release) and mouse events (click, move).

[0039] (3) Transform the original event stream into quantitative features and calculate window features and input features respectively. Window features include switching frequency: count the number of times the window focus changes per unit time (e.g., per minute); application type identification: determine the current application category based on process name and window title keywords (e.g., "IntelliJ IDEA", "Visual Studio", "Excel", "WPS Writer"); dwell time: calculate the continuous dwell time on a specific application or window. Input features include keyboard activity: count the number of keystrokes per unit time; shortcut key identification: match key sequences with a predefined list of commonly used shortcut keys (especially combination keys unique to IDE and office software); input continuity: determine whether the input is continuous or sporadic and bursty by analyzing the time interval distribution between keystroke events; mouse activity ratio: calculate the ratio of mouse events (clicks, moves) to keyboard events.

[0040] (4) Input the above window features and input features into the software behavior classification model for processing, and output the task cognition type. Collect and label a large amount of training data in advance, that is, record user events and label their task cognition types. Train the software behavior classification model based on the window features and input features extracted from the training data to obtain the trained software behavior classification model. The software behavior classification model can be selected from random forest or neural network, and can be flexibly selected according to the actual situation without specific limitations.

[0041] The behavior sensing subunit analyzes ambient sound and determines the task cognition type based on the acoustic characteristics of the ambient sound. It analyzes ambient sound using a built-in high-precision microphone (not for recording, but only for analyzing acoustic characteristics). Continuous, rhythmic keyboard tapping may correspond to a "memory storage task," while intermittent, scattered tapping sounds accompanied by pauses in thought, or frequent page scrolling sounds, may correspond to an "executive function task."

[0042] Furthermore, based on keyboard behavior, the user's current task cognition type is inferred. The specific steps are as follows: (1) The audio stream is acquired in real time through a microphone and preprocessed to obtain the preprocessed audio stream. The preprocessing involves cutting the continuous audio stream into short frames to analyze instantaneous features and applying a high-pass filter to the audio stream to filter out background noise as much as possible.

[0043] (2) Extract multidimensional acoustic features from the preprocessed audio stream. Multidimensional acoustic features include temporal and rhythmic features, spectral features, and perceptual features. The audio stream is discarded directly after feature extraction and is not stored or transmitted.

[0044] When extracting temporal and rhythmic features, first detect valid sound events (such as a crisp tap), calculate the time interval between consecutive events. The intervals for memory storage tasks will show a short and uniform distribution, while the intervals for execution function tasks will be long and fluctuate greatly (thinking pauses). Analyze the energy envelope of the audio signal. Rhythmic continuous input will produce an approximately periodic envelope, while scattered input will show irregular bursts of energy peaks.

[0045] Analyze the frequency components of sound events. Different operations (such as keystrokes, mouse clicks, scrolling, picking up a water glass) have subtle differences in their frequency spectrum. Through continuous frequency spectrum analysis, we can indirectly infer complex operations such as frequent page scrolling.

[0046] Perceptual features, namely Mel-frequency cepstral coefficients (MFCCs), are key features in speech recognition and are also applicable to distinguishing the texture of sound. They can effectively capture attributes such as the crispness of keyboard keystrokes and serve as auxiliary features.

[0047] (3) Input the multidimensional acoustic features into the keyboard behavior classification model for processing and output the task cognition type. A large amount of training data is collected and labeled in advance, that is, audio is recorded and its task cognition type is labeled. The keyboard behavior classification model is trained based on the acoustic features extracted from the training data to obtain the trained keyboard behavior classification model. The keyboard behavior classification model uses a model suitable for time-series classification. Random forest or recurrent neural network can be selected flexibly according to the actual situation without specific limitations. The keyboard and mouse sounds of different users are very different. The system provides a calibration mode to allow users to perform fixed operations (such as continuous typing and scrolling pages) for a few minutes in a quiet environment to learn the unique acoustic features of their devices.

[0048] In some implementations, a physiological signal sensing subunit may be optionally included in the multimodal task recognition module, depending on the specific circumstances. This subunit is used to determine the task cognition type based on the user's physiological signals. Facial movement unit analysis is performed using a non-contact camera, or a simple forehead-mounted EEG sensor is used to monitor the user's attention index, blink frequency, and prefrontal cortex EEG load characteristics. Higher cognitive load is associated with specific EEG bands (such as...). (Wave) Related to this can serve as supplementary evidence for determining "performance-oriented tasks".

[0049] Furthermore, based on the user's physiological signals, the task cognitive type is determined through the following steps: (1) Obtain the facial video stream from the camera, run the facial key point detection and action unit encoding model (such as the algorithm based on OpenFace or MediaPipe) on the local device in real time, and directly output the intensity value of the action unit (AU) (such as AU4 eyebrow drop, AU45 blink), without storing the original video image.

[0050] Alternatively, the user's raw brainwave waveform can be obtained using a simple forehead-mounted EEG sensor.

[0051] (2) Extract facial features or EEG features. Facial features include attention level, blink frequency, and head posture. EEG features include attention index and cognitive load characteristics (the raw EEG signal is subjected to Fast Fourier Transform to calculate specific bands (especially the prefrontal cortex)). The relative power of the wave (approximately 4-8 Hz).

[0052] (3) Establish a personalized baseline for each user. During system initialization, allow users to undergo a brief calibration in a relaxed state, recording their baseline blink frequency, baseline attention index, and baseline EEG band power. All subsequent analyses will use this baseline as a reference. Match facial features or EEG features with task cognitive types.

[0053] It should be noted that facial expressions or EEG characteristics are greatly influenced by personal habits, and a single biometric feature cannot be used as a reliable basis for judgment, but is only an auxiliary reference.

[0054] The judgment results output by the software behavior analysis subunit and the behavior sensing subunit are input into the decision engine. A weighted voting method is used to assign confidence to each judgment result, and the final task cognition type judgment is obtained by combining the results.

[0055] It should be noted that in practical applications, user privacy and compliance must be fully considered. Monitoring and listening operations in the multimodal task recognition module must be performed with the user's knowledge and explicit consent (user authorization), clearly informing them of the scope, purpose, and storage method of data collection. Ideally, data should be processed on the user's local device and not uploaded.

[0056] In some implementations, the addition of physiological data can further improve accuracy. In situations where physiological sensors are inconvenient or unnecessary, the system can determine the task type solely by relying on the fusion of software behavior analysis and behavior sensing subunits (keyboard, mouse, acoustic features). A high recognition accuracy is maintained through a multi-feature fusion judgment algorithm.

[0057] Furthermore, in this embodiment, the multi-feature fusion judgment algorithm is specifically as follows: First, high-order dynamics and temporal pattern features are extracted from the original keyboard, mouse, and acoustic data; second, a hierarchical fusion model based on attention mechanism is adopted to dynamically integrate multimodal behavior sequences and couple prior information from the software context; finally, through online self-learning and user feedback closed loop, the fusion judgment weight is continuously optimized to achieve highly robust cognitive state recognition under conditions without physiological sensing.

[0058] In some implementations, the multimodal task recognition module can be linked with the user's smart wearable device (such as a smartwatch or smart ring) to obtain data such as the user's heart rate variability (HRV) and electrical activity of the skin (EDA), which can be used as auxiliary signals to assess the user's cognitive load and stress, further enriching the dimensions of state recognition.

[0059] The user feature input module is used to obtain basic user information; In this embodiment, the user feature input module receives basic information input by the user, such as gender and age. This can be a simple user profile settings interface, or it can be obtained through user registration information when the system is first used.

[0060] The central processing and control module is used to obtain the optimal lighting parameters based on task recognition type and user basic information in intelligent adaptive mode; the lighting parameters include illuminance and color temperature.

[0061] In this embodiment, the central processing and control module includes a dynamic lighting strategy library. This library is not a simple static database, but rather integrates an intelligent decision-making algorithm based on reinforcement learning. The intelligent decision-making algorithm uses "task cognitive type - user gender" as the state and the "illuminance - color temperature" combination as the action. The algorithm's goal is no longer simply to query a fixed value, but to explore and utilize multiple effective parameter combinations (such as high illuminance and low color temperature, low illuminance and high color temperature) that have been preliminarily validated in the literature to find the "optimal action" for a specific user in the current state.

[0062] In intelligent adaptive mode, a reinforcement learning-based intelligent decision-making algorithm constructs an intelligent agent capable of autonomously learning and optimizing decisions. The agent continuously interacts with the environment to find the optimal task-user-lighting parameter matching strategy. The core objective is to improve the user's cognitive work efficiency by optimizing lighting parameters, specifically: (1) Define the state, action and reward functions.

[0063] The state is task cognition type - user gender, and the action is a discrete or continuous illuminance-color temperature combination. To avoid combinatorial explosion and utilize existing knowledge, we can start from the effective combinations validated in the literature and define a set of "basic actions".

[0064] The reward function is constructed around the objective efficiency metrics provided by the micro-task implantation and evaluation sub-units. Specifically, the reward... It consists of three weighted factors: efficiency reward, comfort reward, and stability penalty.

[0065] in, For the change in efficiency, For comfort, As a stability penalty, , , These are the weights corresponding to the changes in efficiency, comfort, and stability penalties, respectively.

[0066] Furthermore, the core efficiency changes It can be calculated using the following formula:

[0067] in, For the accuracy of micro-tasks, For the reaction time of microtasks, (The ratio of baseline accuracy to baseline reaction time in the individual efficiency baseline) represents the historical efficiency baseline value of the corresponding user under the same cognitive state. The change in efficiency directly and objectively quantifies the relative improvement rate of the user's working memory efficiency after performing the lighting adjustment action, thus providing a clear and robust optimization direction for the reinforcement learning agent.

[0068] Furthermore, comfort This system is used to assess a user's subjective acceptance of the current lighting environment. To quantify comfort, the calculation integrates three data sources: explicit feedback, implicit behavioral analysis, and personalized preference matching, aiming to balance automatic system optimization with individual user preferences. The specific calculation is achieved using the following formula:

[0069] in, , , For the weighting coefficients, satisfying , representing the relative importance of explicit feedback, behavioral analysis, and preference matching in comfort assessment, respectively; This is a direct reward based on user-initiated feedback; For indirect rewards based on implicit behavior analysis, comfort is inferred through user interaction patterns; Rewards are based on personalized preference matching.

[0070] In this embodiment, the initial default value of the weighting coefficient can be set to: , , =0.2, and can be dynamically adjusted based on data accumulation.

[0071] When a user expresses their preference through a preset interface (such as a keyboard shortcut or gesture), a display feedback is triggered: if the user accepts the current lighting (e.g., by clicking "Keep" or remaining inactive for more than 30 minutes), then... If the user refuses (e.g., actively adjusts or clicks something they find inappropriate), then ,in For constant reward magnitude (e.g.) ), used to control the strength of a single feedback.

[0072] The formula for calculating indirect rewards based on implicit behavior analysis is as follows:

[0073]

[0074] in, For behavioral deviation, This represents the rate of change of the standard deviation of mouse movement speed relative to an individual's baseline. This represents the rate of change of the coefficient of variation of the key press interval. , For normalization constants, , The weights for mouse and keyboard actions are respectively. This is used to balance the effects of mouse and keyboard behavior.

[0075] The system maintains a dynamic preference matrix for each user. , where each entry Indicates in task type and time period Illumination based on user's historical preferences Color temperature and confidence level ( ).

[0076] Define matching distance :

[0077] in, and Given the current illuminance and current color temperature, and This is a normalization factor (e.g., 1000 lux and 2000 k respectively).

[0078] but The calculation formula is:

[0079] in, The maximum allowed distance threshold (e.g.) This ensures that high rewards are obtained when the parameters are close to high confidence preferences.

[0080] Furthermore, stability penalty This is a negative reward item used to prevent frequent or drastic changes in lighting parameters from interfering with users, ensuring a smooth transition in the lighting environment while also saving energy. Its calculation covers four dimensions: the magnitude of change, frequency, oscillation mode, and energy consumption. The specific formula is as follows:

[0081]

[0082]

[0083]

[0084] in, Penalty for the magnitude of change; This represents the absolute change in illuminance between adjacent time points. , for Illuminance at any time for Illuminance at any given time; This represents the absolute change in color temperature between adjacent time points. , for Color temperature at any moment for Color temperature at any given moment; and The maximum allowable single-step change (e.g.) =200 lux =500k), to prevent visual discomfort; and The penalty coefficient (e.g., all set to 0.4) controls the penalty intensity for changes in illuminance and color temperature; Penalty for frequency of change; This indicates the number of lighting adjustments made within the past time window; The maximum number of adjustments allowed within the time window (e.g.) =5 times); This is a frequency penalty coefficient used to suppress excessively frequent adjustments; The mode oscillation penalty is designed to detect and punish the behavior of parameters switching back and forth between values; To mitigate energy-saving and stability penalties, the use of highly efficient and stable lighting setups is encouraged; The minimum illuminance required for the current task type (e.g., 500 lux for a reading task); As an energy-saving factor, this penalty term causes the system to prefer lower illuminance while meeting demand, in order to reduce energy consumption.

[0085] Define the parameter sequence for the most recent K adjustments (e.g., K=5). If a continuous oscillation pattern of "A→B→A" exists, then calculate the oscillation score. :

[0086] in, This represents the average value of the oscillation amplitude.

[0087] Oscillation penalty is calculated based on the oscillation scoring model.

[0088] in, The oscillation penalty coefficient (e.g.) =0.4), used to reduce invalid oscillations.

[0089] The parameters in the reward function were determined experimentally and will not be elaborated further here.

[0090] (2) Construct a policy network.

[0091] A deep neural network is used as the policy network. Its input layer corresponds to the state vector, and its output layer corresponds to the probability distribution of each action. The optimal combination of lighting parameters is found by using a deep deterministic policy gradient or proximal policy optimization algorithm.

[0092] (3) System interaction and online learning.

[0093] Before deploying the network, the policy network is pre-trained using historical data to obtain a pre-trained policy network. Online interaction and learning include observing the state, making decisions (the policy network outputs actions based on the state), executing, evaluating (calculating reward values), and learning.

[0094] To avoid instability or ineffective exploration in the early stages of reinforcement learning, the system sets reasonable initial weight values ​​through experiments. The initial values ​​of each coefficient are based on: Cognitive efficacy priority principle: such as efficiency weight The high initial value reflects that the core optimization goal of the system is to improve working memory efficiency.

[0095] User experience assurance principles: weighting of comfort and stability ( , The initialization of the system is to avoid user discomfort or interference caused by frequent and drastic changes in lighting while pursuing efficiency. This "exploitation-exploitation" balance is a common design paradigm in the field of reinforcement learning, aiming to ensure the basic usability and security of the system's behavior.

[0096] Signal reliability estimation: The initial confidence weights of each sub-module (software behavior, acoustic features) in multimodal fusion can be allocated based on their historical recognition accuracy on the training data. Modalities with higher accuracy receive higher initial weights, which is a rational initialization based on data.

[0097] In subsequent use, through reinforcement learning loop, the system uses interaction data with specific users to continuously optimize their decision-making strategies, thereby dynamically generating and iterating the optimal combination of weight parameters specific to that user.

[0098] In manual setting mode, the central processing and control module receives the lighting parameters manually set by the user according to personal preferences and generates corresponding control commands.

[0099] A dimmable lighting module is used to receive control commands and adjust the lighting parameters of the lamps according to the control commands.

[0100] In this embodiment, in intelligent adaptive mode, after obtaining the user's current optimal lighting parameters, the central processing and control module sends control commands to the dimmable lighting module; in manual setting mode, after obtaining the user's manually set lighting parameters, the central processing and control module sends corresponding control commands to the dimmable lighting module. The dimmable lighting module includes one or more LED luminaires with adjustable spectrum and brightness. The luminaires receive control commands and can precisely and steplessly adjust their output illuminance (lx) and correlated color temperature (K).

[0101] The online performance verification and strategy optimization module is used to quantify the user's working memory efficiency under the current light environment based on micro-tasks in intelligent adaptive mode, and feed the working memory efficiency evaluation results back to the central processing and control module for algorithm optimization.

[0102] In this embodiment, the online performance verification and strategy optimization module is the core of achieving closed-loop control and intelligence, implemented in an intelligent adaptive mode. Because it involves micro-task testing and feedback, users can choose to use the system at appropriate times to customize the most suitable personalized lighting conditions. Specifically, it includes: Micro-task implantation and evaluation sub-unit: With the user's consent, the system will periodically (e.g., every 30 minutes) pop up minimalist cognitive micro-tasks on the edge of the user's screen without significantly interfering with the user's main task. For example: A 6-second change detection task (quickly determine whether two images are the same).

[0103] A 1-back task (quickly determine if the currently appearing letter is the same as the previous one).

[0104] This sub-unit records the user's accuracy and reaction time in completing these micro-tasks, serving as a real-time quantitative indicator of the user's working memory efficiency under the current lighting conditions. In other words, working memory efficiency is used to measure improvements in cognitive performance. Efficient working memory is the ability to maintain high accuracy within a reasonable timeframe, and an individual efficiency baseline (including baseline accuracy and baseline reaction time) is determined based on historical data.

[0105] An efficiency score is calculated based on the user's accuracy and reaction time in completing micro-tasks. , represented as:

[0106] in, This represents the current accuracy of the microtasks. This represents the current response time of the microtask. This represents the baseline accuracy for microtasks. This represents the baseline reaction time for microtasks. An efficiency score greater than 1 indicates that the current efficiency is higher than the personal efficiency baseline, while a score less than 1 indicates that the current efficiency is lower than the personal efficiency baseline. It directly reflects the relative ability to respond correctly per unit of time.

[0107] Feedback learning mechanism: A reinforcement learning algorithm that uses the evaluation results of micro-tasks (the user's accuracy and reaction time in completing the micro-tasks) as reward signals to feed back to the central processing module. If the user's micro-task performance improves after a certain "illuminance-color temperature" action, then that action is reinforced (by the change in efficiency). The weight of a light source is determined by its relative position in the current state; conversely, it is reduced if the weight is lowered. Thus, the system can dynamically and individually learn and approximate the optimal lighting configuration for each user, rather than relying on a fixed preset.

[0108] Through the above technical solutions, a personalized lighting strategy library capable of self-learning and evolution has been constructed, transcending fixed preset mapping relationships.

[0109] The weight parameters involved in the above strategy optimization process are essentially an adaptive parameter system based on domain knowledge initialization and dynamically learned and optimized through user interaction data. The aim is to start from general group patterns and ultimately achieve individual-specific adaptation. Weight coefficients are as follows: , , , , , , , These variables, etc., are all defined as adjustable parameter variables within the system. Together, they constitute part of the policy space for the system to perform reinforcement learning or multimodal decision-making, and are part of the optimization objective.

[0110] The specific workflow of the system is as follows: (1) Initialization: The user sets the gender and selects the lighting mode. In the intelligent adaptive mode, the system starts with the default or literature-recommended lighting strategy. In the manual setting mode, the system provides lighting with the lighting parameters manually set by the user.

[0111] (2) Continuous perception and recognition: In intelligent adaptive mode, the multimodal task recognition module runs continuously and judges the user's task cognition type in real time.

[0112] (3) Intelligent decision-making: In intelligent adaptive mode, the central processing and control module selects an "illuminance-color temperature" action instruction based on the currently identified "task type-user gender" status through reinforcement learning algorithm.

[0113] (4) Execution: The dimmable lighting module executes control commands.

[0114] (5) Online verification and optimization: In intelligent adaptive mode, the online performance verification and strategy optimization module periodically executes micro-tasks, collects performance data, and feeds it back to the decision-making algorithm as a reward signal to update the strategy library. This step forms a closed loop of "execution-verification-learning", which is the core difference from existing technologies.

[0115] (6) Dynamic adjustment: In intelligent adaptive mode, the system outputs better illumination parameters under the same or similar conditions in the future based on the learned experience, and continuously optimizes the user's working memory efficiency.

[0116] In some implementations, cognitive microtasks are not limited to change detection or N-back. For example, they may employ: Numerical memory breadth task: Quickly flash a string of numbers and ask the user to recall it.

[0117] Stroop task: Evaluate the user's executive function and inhibitory control.

[0118] A simple visual search task: evaluating visual processing speed.

[0119] Different micro-tasks can assess cognitive function from different perspectives, and the system can intelligently select the most relevant micro-task type based on the characteristics of the main task.

[0120] In some implementations and scenarios, to avoid any form of interference, the system can employ more implicit performance evaluation metrics. For example, when a user performs a specific, quantifiable main task (such as data entry or code writing), changes in task completion speed and error rate can be directly analyzed as reward signals.

[0121] In some implementations, the reinforcement learning algorithms in the dynamic lighting policy library can be classic Q-Learning, SARSA, or more advanced deep reinforcement learning (DRL) networks, such as Deep Q-Network (DQN), which is more advantageous in dealing with high-dimensional state spaces.

[0122] In some implementations, the system employs a cloud-edge-device collaborative architecture. User terminals (desk lamps, room lights) are responsible for data collection and execution, while complex task recognition models and reinforcement learning algorithms run on edge servers or in the cloud. This reduces terminal costs and allows for the training of more powerful general-purpose models using massive amounts of user data (under anonymity and authorization).

[0123] In some implementations, the core functionality of the system can be encapsulated as a software development kit (SDK) or plugin, enabling existing operating systems (such as Windows, macOS) or office software suites to call smart lights that conform to the system's standards, thus achieving widespread and convenient deployment.

[0124] All the above alternative solutions can achieve the core objective of this invention, namely, to improve the user's working memory efficiency through closed-loop intelligent control. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this invention should be included within the protection scope of this invention.

[0125] Experimental description: Subsequent rigorously controlled experiments confirmed that specific indoor artificial lighting parameters (a combination of illuminance and correlated color temperature) can significantly improve performance on working memory tasks, and this effect is modulated by the type of cognitive processing involved in the task and by the individual's gender. Key experimental evidence is as follows: Experiment 1: Simultaneous presentation of working memory tasks (the completion of this task mainly relies on information storage function, with a low execution load).

[0126] This experiment examined various combinations of different illuminance (100 lx vs 1000 lx) and associated color temperature (3000 K vs 6500 K), finding that individuals in low color temperature (3000 K) and high illuminance (1000 lx) environments showed significantly improved working memory performance. Specifically, female participants showed a significantly higher accuracy rate (83.25%) under low color temperature but higher accuracy rate (81.33%) (p < 0.05). Male participants also showed a significantly higher accuracy rate (80.58%) under low color temperature but higher accuracy rate (77.50%) (p < 0.01). Specific results are as follows... Figure 3 As shown in Table 1: Table 1. Accuracy of working memory efficiency under different lighting conditions for participants of different genders

[0127] Figure 3 In the diagram, * indicates a significant difference at the 0.05 level, ** indicates a significant difference at the 0.01 level, and the T-line represents the standard deviation.

[0128] Experimental results demonstrate that for working memory tasks that primarily involve storage and processing, a "high illuminance-low color temperature" lighting environment has a clear promoting effect on working memory function.

[0129] Experiment 2: Sequence Presentation Working Memory Task (involves more executive functions, such as updating, inhibition, and temporal processing, resulting in a higher executive load).

[0130] The control task difficulty in this experiment was the same as in Experiment 1. To eliminate the confounding factor of "task difficulty," the study adjusted the presentation time to ensure that there was no statistically significant difference in baseline task difficulty between Experiment 1 (simultaneous presentation) and Experiment 2 (sequential presentation) (t(94) = -0.18, p = 0.86). Only the presentation method of the memory items was changed to examine different cognitive processing processes. Figure 4 As shown in Table 2, the illumination effect exhibited significant gender specificity. Female participants performed in a pattern consistent with Experiment 1; under low color temperature (3000 K), the accuracy rate for high illumination (1000 lx) (74.71%) was still significantly higher than that for low illumination (71.96%), p < 0.05. Male participants followed a "matching principle": under low color temperature (3000 K), the accuracy rate for low illumination (100 lx) (77.63%) was significantly higher than that for high illumination (74.83%), p < 0.05. Under high color temperature (6500 K), the accuracy rate for high illumination (1000 lx) (78.96%) was significantly higher than that for low illumination (76.21%), p < 0.05.

[0131] Table 2. Working memory accuracy of participants of different genders under different lighting conditions

[0132] Figure 4 In the diagram, * indicates a significant difference at the 0.05 level, and the T-line represents the standard deviation.

[0133] The experimental results demonstrate that for sequential tasks requiring more execution functions, the enhancement effect of illumination still exists, but the optimal parameter combination varies depending on the individual (gender) and the nature of the task, proving that illumination regulation needs to be personalized.

[0134] Example 2 This embodiment discloses an adaptive lighting method for improving cognitive efficacy based on task cognition type, including: It receives the user's selected lighting mode, including intelligent adaptive mode and manual setting mode; In intelligent adaptive mode, user behavior data is acquired and its behavioral features are extracted, and the user's task cognition type is determined based on the behavioral features. Obtain basic user information; In intelligent adaptive mode, based on task cognitive type and user basic information, the optimal lighting parameters are obtained through reinforcement learning algorithm, and control commands are sent to the dimmable lighting module based on the optimal lighting parameters. Receive control commands and adjust the lighting parameters of the lamps accordingly; In intelligent adaptive mode, the user's working memory efficiency is quantified based on micro-tasks under the current light environment, and the working memory efficiency evaluation results are fed back to the central processing and control module for algorithm optimization.

[0135] Example 3 The purpose of this embodiment is to provide a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method of Embodiment 2.

[0136] Example 4 The purpose of this embodiment is to provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the method of Embodiment 2.

[0137] The steps and methods involved in the apparatuses of Embodiments 3 and 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0138] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0139] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0140] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. An adaptive illumination system for improving cognitive efficacy based on task-based cognitive types, characterized in that, include: The user selection and authorization module is used to receive the lighting mode selected by the user, including intelligent adaptive mode and manual setting mode; The multimodal task recognition module is used to acquire user behavior data and extract its behavioral features in intelligent adaptive mode, and to determine the user's task cognition type based on the behavioral features. The user feature input module is used to obtain basic user information; The central processing and control module is used to obtain the optimal lighting parameters based on the task cognition type and user basic information in intelligent adaptive mode through reinforcement learning algorithm, and then issue control commands to the dimmable lighting module based on the optimal lighting parameters. A dimmable lighting module is used to receive control commands and adjust the lighting parameters of the lamps according to the control commands; The online performance verification and strategy optimization module is used to quantify the user's working memory efficiency under the current light environment based on micro-tasks in intelligent adaptive mode, and feed the working memory efficiency evaluation results back to the central processing and control module for algorithm optimization.

2. The adaptive illumination system for promoting cognitive efficacy based on task cognition type as described in claim 1, characterized in that, The multimodal task recognition module includes a software behavior analysis subunit and a behavior sensing subunit, and the user behavior data includes software behavior and keyboard behavior.

3. The adaptive illumination system for promoting cognitive efficacy based on task cognition type as described in claim 2, characterized in that, The software behavior analysis subunit is used to monitor the user's window activities and input device behavior, convert window activities and input device behavior into quantitative features, and determine the task cognition type based on the quantitative features.

4. The adaptive illumination system for promoting cognitive efficacy based on task cognition type as described in claim 2, characterized in that, The behavior sensing subunit is used to acquire audio streams, extract multi-dimensional acoustic features from the audio streams, and determine the task cognition type based on the multi-dimensional acoustic features.

5. The adaptive illumination system for promoting cognitive efficacy based on task cognition type as described in claim 2, characterized in that, The judgment results output by the software behavior analysis subunit and the behavior sensing subunit are input into the decision engine. A weighted voting method is used to assign a confidence level to each judgment result, and the final task cognition type is obtained by combining the results.

6. The adaptive illumination system for promoting cognitive efficacy based on task cognition type as described in claim 1, characterized in that, In the reinforcement learning algorithm, the state is task cognition type - user gender, the action is illuminance - color temperature combination, and the reward function is a weighted average of efficiency change, comfort, and stability penalties. The efficiency change is expressed as: in, For the accuracy of micro-tasks, For the reaction time of microtasks, This represents the historical efficiency baseline value for the corresponding user under the same cognitive state.

7. The adaptive illumination system for promoting cognitive efficacy based on task cognition type as described in claim 1, characterized in that, The online performance verification and strategy optimization module includes a micro-task implantation and evaluation sub-unit. This sub-unit periodically pops up cognitive micro-tasks and records the accuracy and reaction time of the user in completing the cognitive micro-tasks. Based on the accuracy and reaction time, the user's working memory efficiency under the current light environment is quantified, and an efficiency score is calculated.

8. An adaptive illumination method for improving cognitive efficacy based on task-based cognitive types, characterized in that, include: It receives the user's selected lighting mode, including intelligent adaptive mode and manual setting mode; In intelligent adaptive mode, user behavior data is acquired and its behavioral features are extracted, and the user's task cognition type is determined based on the behavioral features. Obtain basic user information; In intelligent adaptive mode, based on task cognitive type and user basic information, the optimal lighting parameters are obtained through reinforcement learning algorithm, and control commands are sent to the dimmable lighting module based on the optimal lighting parameters. Receive control commands and adjust the lighting parameters of the lamps accordingly; In intelligent adaptive mode, the user's working memory efficiency is quantified based on micro-tasks under the current light environment, and the working memory efficiency evaluation results are fed back to the central processing and control module for algorithm optimization.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the adaptive illumination method for promoting cognitive efficacy based on task cognitive type as described in claim 8.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the adaptive illumination method for promoting cognitive efficacy based on task cognitive type as described in claim 8.

Images