Intelligent control methods, devices, and electronic equipment for phototherapy devices

Abstract

This invention discloses an intelligent control method, device, and electronic device for a phototherapy device. The invention relates to the technical field of intelligent control methods. Based on the current phototherapy mode of the phototherapy device, multiple motion images of the user, and the user's past sleep events, the current sleep-aiding event of the phototherapy device is determined. Based on this current sleep-aiding event, the user's facial expressions, and multiple environmental parameters, multiple key illumination factors of the phototherapy device are determined, improving the accuracy of these key illumination factors. Therefore, based on multiple content to be optimized, the current phototherapy mode of the phototherapy device, and the user's sleep posture change events, intelligent control events of the phototherapy device are determined. Based on the recognition of intelligent control events, corresponding intelligent illumination content is determined. Based on the intelligent illumination content, the user's surrounding environmental events, and the corresponding sleep duration, the rhythmic control system of the phototherapy device is determined, improving the accuracy of the rhythmic control system of the phototherapy device.

Description

Technical Field

[0001] This invention relates to the technical field of intelligent control methods, and more particularly to an intelligent control method, device, and electronic device for phototherapy equipment. Background Technology

[0002] With the development of technology, phototherapy equipment has been gradually applied to people's lives. Phototherapy equipment is an intelligent device that uses light of specific wavelengths, intensities and doses to regulate physiological functions. It is not an ordinary lighting fixture, but outputs corresponding therapeutic light.

[0003] In recent years, with the rise of health consumer technology, phototherapy devices have gradually expanded from professional medical settings to the field of home sleep management. These devices aim to correct the user's circadian rhythm by simulating the spectrum and intensity changes of natural sunlight, thereby improving problems such as difficulty falling asleep, light sleep, and difficulty waking up in the morning, and have become an important branch of sleep technology. Currently, there are several representative phototherapy sleep aids on the market. For example, Philips' SmartSleep series Sleep & Wake-up Light product guides breathing rhythms with light and combines a sunset simulation function to help users relax and fall asleep. The light intensity gradually decreases to simulate the sunset process, reducing melatonin secretion inhibition and promoting a sleep environment. In addition, there is Hatch Baby's Hatch Restore product, which integrates programmable color temperature and brightness changes to simulate sunrise and sunset, providing users with a standardized bedtime ritual. By simulating changes in natural light, it helps users improve sleep quality. Its corresponding patent application is US2024 / 0389935A1.

[0004] However, existing phototherapy devices illuminate the user along the first direction, neglecting to consider the illumination area of ​​the device relative to the user, and failing to take into account the current sleep-aiding event and the user's facial expressions. This affects the accuracy of several key illumination factors of the phototherapy device, resulting in low accuracy of the device's rhythm control system. Its core technical bottlenecks lie in "static light output," "unidirectional program control," and "lack of closed-loop biofeedback." These factors together lead to insufficient accuracy of the device's key phototherapy parameters (such as dosage, timing, and spectrum), ultimately significantly reducing the effectiveness, robustness, and user experience of the entire rhythm control system.

[0005] Therefore, there is an urgent need in this field for an intelligent solution that can sense the environment, understand users, and dynamically adjust. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides an intelligent control method, device, and electronic device for phototherapy equipment.

[0007] This invention provides an intelligent control method for a phototherapy device, comprising:

[0008] The illumination status of the phototherapy device is determined based on multiple illumination data and the corresponding usage scenario; the corresponding illumination-based interaction status is determined based on the illumination status of the phototherapy device and the interaction data of the phototherapy device relative to the user.

[0009] In this light-based interactive state, the light-illuminating area of ​​the phototherapy device relative to the user is marked, and the current phototherapy mode of the phototherapy device is determined based on the light-illuminating area and multiple working data of the phototherapy device;

[0010] The current sleep-inducing event of the phototherapy device is determined based on the current phototherapy mode of the phototherapy device, multiple motion images of the user, and the user's past sleep events; multiple key light factors of the phototherapy device are determined based on the current sleep-inducing event, the user's facial expressions, and multiple environmental parameters.

[0011] The sleep-aiding effect coefficient of the phototherapy device is determined based on multiple key light factors, their corresponding priorities, and the current circadian rhythm of the phototherapy device; multiple areas to be optimized are determined based on the user's sleep state, the sleep-aiding effect coefficient of the phototherapy device, and the corresponding multiple light data.

[0012] The system collects user sleep posture change events, determines intelligent control events for the phototherapy device based on multiple content to be optimized, the current phototherapy mode of the phototherapy device, and the user's sleep posture change events, determines the corresponding intelligent light content based on the identification of intelligent control events, and determines the rhythm control system of the phototherapy device based on the intelligent light content, the user's surrounding environmental events, and the corresponding sleep duration.

[0013] This invention provides an intelligent control device for a phototherapy device. The intelligent control device is applied to the aforementioned intelligent control method for a phototherapy device. The intelligent control device includes:

[0014] The light-based interactive state module is used to determine the light state of the light therapy device based on multiple light data of the light therapy device and the corresponding usage scenario; and to determine the corresponding light-based interactive state based on the light state of the light therapy device and the interaction data of the light therapy device relative to the user.

[0015] The current phototherapy module is used to mark the light area of ​​the phototherapy device relative to the user in this light-based interactive state, and determine the current phototherapy mode of the phototherapy device based on the light area and multiple working data of the phototherapy device;

[0016] The key illumination factor module is used to determine the current sleep aid event of the phototherapy device based on the current phototherapy mode of the phototherapy device, multiple motion images of the user, and the user's past sleep events; and to determine multiple key illumination factors of the phototherapy device based on the current sleep aid event, the user's facial expressions, and multiple environmental parameters.

[0017] The "Content to be Optimized" module is used to determine the sleep-aiding effect coefficient of the phototherapy device based on multiple key light factors, their corresponding priorities, and the current circadian rhythm of the phototherapy device; and to determine multiple contents to be optimized based on the user's sleep state, the sleep-aiding effect coefficient of the phototherapy device, and the corresponding multiple light data.

[0018] The intelligent control module is used to collect the user's sleep posture change events, determine the intelligent control events of the phototherapy device based on multiple content to be optimized, the current phototherapy mode of the phototherapy device, and the user's sleep posture change events, determine the corresponding intelligent light content based on the identification of the intelligent control events, and determine the rhythm control system of the phototherapy device based on the intelligent light content, the user's surrounding environmental events, and the corresponding sleep duration.

[0019] This invention provides an electronic device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the intelligent control method for the phototherapy device described above.

[0020] Compared with the prior art, the beneficial effects of the present invention are:

[0021] (1) This invention integrates the user's real-time physiological data (movement, respiration, heart rate), emotional state (facial expression), long-term sleep history, and current environment to dynamically diagnose the user's "sleep-aiding events" (such as "anxious difficulty falling asleep") and generate a unique light therapy formula accordingly. The system no longer provides static light therapy, but provides a continuously optimized "dynamic light therapy process" that changes with the user's state, thus achieving truly personalized intervention;

[0022] (2) By constructing a "light field model" and aligning it with the user's "digital body model," this invention can accurately calculate the "dose" of light irradiating specific areas of the user's body (such as the face). This transforms phototherapy from a subjective and vague experience into a quantifiable, calculable, and targeted precision medical approach, greatly enhancing its scientific validity and reliability.

[0023] (3) This invention constructs a complete closed loop of "perception-decision-execution-re-perception". It not only executes preset instructions, but also evaluates the theoretical effect by calculating the "sleep aid effect coefficient" and compares it with the actual "sleep state"; when an "optimization gap" is found, the system can autonomously diagnose the problem and determine the "content to be optimized" in the next step; this enables the device to have the ability to continuously learn and improve itself, and to better adapt to the ever-changing needs of users and environmental interference;

[0024] (4) This invention incorporates environmental parameters (temperature, noise, ambient light) into the decision-making system. It can identify whether environmental factors are “cooperative interference” or “primary contradiction” and make the light make corresponding “compensation strategies”. This makes the phototherapy device no longer an isolated instrument, but an intelligent agent that can understand and adapt to complex real-life scenarios. It can cope with complex scenarios and has strong anti-interference capabilities.

[0025] Therefore, the sleep-aiding effect coefficient of the phototherapy device is determined based on multiple key light factors, their corresponding priorities, and the current circadian rhythm of the phototherapy device. Multiple optimization elements are identified based on the user's sleep state, the sleep-aiding effect coefficient of the phototherapy device, and corresponding light data. User sleep posture change events are collected, and intelligent control events of the phototherapy device are determined based on these optimization elements, the current phototherapy mode of the device, and the user's sleep posture change events. The corresponding intelligent light content is determined based on the identification of these intelligent control events. The circadian rhythm control system of the phototherapy device is determined based on the intelligent light content, the user's surrounding environmental events, and the corresponding sleep duration. By introducing multiple optimization elements, the intelligent light content is further controlled, achieving greater accuracy in the intelligent light content, the user's surrounding environmental events, and the corresponding sleep duration, thus improving the accuracy of the circadian rhythm control system of the phototherapy device. Attached Figure Description

[0026] Figure 1 This is a flowchart illustrating the intelligent control method for the phototherapy device in an embodiment of the present invention;

[0027] Figure 2 This is a flowchart illustrating step S11 in the intelligent control method of the phototherapy device in this embodiment of the invention.

[0028] Figure 3 This is a flowchart illustrating step S12 in the intelligent control method of the phototherapy device in this embodiment of the invention.

[0029] Figure 4 This is a flowchart illustrating step S13 in the intelligent control method of the phototherapy device in this embodiment of the invention.

[0030] Figure 5This is a flowchart illustrating step S14 of the intelligent control method for the phototherapy device in this embodiment of the invention.

[0031] Figure 6 This is a flowchart illustrating step S15 of the intelligent control method for the phototherapy device in this embodiment of the invention.

[0032] Figure 7 This is a schematic diagram of the structure of the intelligent control device of the phototherapy equipment in an embodiment of the present invention. Detailed Implementation

[0033] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0034] Please see Figures 1 to 7 A smart control method for a phototherapy device, applied in a smart control scenario; the smart control method for the phototherapy device includes:

[0035] Step S11: Determine the illumination state of the phototherapy device based on multiple illumination data and the corresponding usage scenario; determine the corresponding illumination-based interaction state based on the illumination state of the phototherapy device and the interaction data of the phototherapy device relative to the user;

[0036] Step S12: In this light-based interactive state, mark the light area of ​​the phototherapy device relative to the user, and determine the current phototherapy mode of the phototherapy device based on the light area and multiple working data of the phototherapy device;

[0037] Step S13: Determine the current sleep-aiding event of the phototherapy device based on the current phototherapy mode of the phototherapy device, multiple motion images of the user, and the user's past sleep events; determine multiple key light factors of the phototherapy device based on the current sleep-aiding event, the user's facial expressions, and multiple environmental parameters;

[0038] Step S14: Determine the sleep-aiding effect coefficient of the phototherapy device based on multiple key light factors, their corresponding priorities, and the current circadian rhythm of the phototherapy device; determine multiple items to be optimized based on the user's sleep state, the sleep-aiding effect coefficient of the phototherapy device, and the corresponding multiple light data.

[0039] Step S15: Collect the user's sleep posture change events, determine the intelligent control events of the phototherapy device based on multiple content to be optimized, the current phototherapy mode of the phototherapy device, and the user's sleep posture change events, determine the corresponding intelligent light content based on the identification of the intelligent control events, and determine the rhythm control system of the phototherapy device according to the intelligent light content, the user's surrounding environmental events, and the corresponding sleep duration.

[0040] refer to Figure 2 In step S11, the specific steps are as follows:

[0041] S111: Real-time monitoring of phototherapy equipment and collection of multiple light data points from the equipment; determination of the usage scenario of the phototherapy equipment based on its current location, surrounding brightness, and the user's current state; determination of the light status of the phototherapy equipment based on multiple light data points, the corresponding usage scenario, and the equipment's lifespan.

[0042] S112: When the phototherapy device outputs phototherapy light to the user, collect multiple working data of the phototherapy device, the corresponding light data, and the changes in the surrounding brightness to determine the interaction data of the phototherapy device relative to the user. Based on the light status of the phototherapy device, the interaction data of the phototherapy device relative to the user, and the continuous light duration of the phototherapy device, determine the corresponding light-type interaction status.

[0043] In the embodiments of this application, the phototherapy device is monitored in real time, and multiple light data of the phototherapy device are collected; the usage scenario of the phototherapy device is determined according to the current position of the phototherapy device, the corresponding ambient brightness, and the current state of the user; the light status of the phototherapy device is determined based on the multiple light data of the phototherapy device, the corresponding usage scenario, and the lifespan of the phototherapy device, which takes into account the overall consideration of the multiple light data of the phototherapy device, the corresponding usage scenario, and the lifespan of the phototherapy device, and ensures the accuracy of the light status of the phototherapy device.

[0044] At this time, through the built-in sensor array and control feedback loop of the device, multi-dimensional data describing the light output characteristics are collected at a high frequency (>10Hz); these data not only cover macroscopic visual parameters, but also go deeper into the core spectral level that determines physiological effects.

[0045] Spectral Power Distribution (SPD): The light energy distribution per nanometer in the 400-700nm range is obtained in real time through a miniature spectrometer or based on multi-channel LED drive current and a known spectral model. This is the most basic data for calculating the circadian rhythm factor (CS). Six-channel drive parameters: The PWM duty cycle and constant current value of ultracool white light (CH1), warm yellow light (CH2), 480nm blue light (CH3), amber light (CH4), and red light (CH5) are obtained by directly reading the drive IC register. Comprehensive optical parameters: Correlated color temperature (CCT), color rendering index (CRI), luminous flux (Φ), and retinal melanopsin equivalent illuminance are calculated in real time based on SPD. Device physical status: The ripple coefficient of the drive power supply, LED module temperature (via NTC thermistor), and cumulative working time are collected.

[0046] The system uses Wi-Fi SSID / BSSID fingerprint positioning, Bluetooth beacon, or GPS signals to determine whether the device is located in "home - master bedroom", "hotel - guest room", or "office"; it continuously monitors ambient light levels using an HDR ambient light sensor to distinguish between "complete darkness (<1lx)", "weak ambient light (1-5lx)" and "indoor lights on (>50lx)"; it obtains heart rate variability (HRV) and body movement recorder data by connecting to wearable devices via BLE, or obtains "bedtime reminder" events via the mobile phone calendar API.

[0047] The scenario decision logic uses a rule-based or lightweight machine learning classifier, for example: IF(location="home-master bedroom") AND(ambient light < 1 lx) AND(user state="sitting") AND(time > 22:00) THEN scenario="sleep preparation-sleep aid mode activated".

[0048] The CS value and CCT calculated by the current SPD are compared with the target parameters of the scene to calculate the "performance matching degree" score; based on the cumulative working time and temperature history of the LED, the light decay model is called to predict the current maximum luminous flux attenuation rate and evaluate the "device load health"; finally, a multi-dimensional state vector is generated, such as: {Scene ID: "Sleep_Prep", Current CS: 0.12, Target CS: 0.15, Matching degree: 80%, Device load: 65%, Light decay compensation coefficient: 1.1, State confidence: "High"}.

[0049] Furthermore, when the phototherapy device outputs phototherapy light to the user, multiple working data of the phototherapy device, corresponding illumination data, and changes in ambient brightness are collected to determine the interaction data between the phototherapy device and the user. Based on the illumination status of the phototherapy device, the interaction data between the phototherapy device and the user, and the continuous illumination duration of the phototherapy device, the corresponding light-based interaction state is determined. This takes into account the overall consideration of the illumination status of the phototherapy device, the interaction data between the phototherapy device and the user, and the continuous illumination duration of the phototherapy device, ensuring the accuracy of the corresponding light-based interaction state.

[0050] At this point, the system achieves interactive perception through multi-dimensional data acquisition, including real-time monitoring of working data, illumination data, and changes in ambient brightness; working data acquisition covers user-initiated operations (such as touch adjustment and voice commands) and automatic system adjustments (such as dynamic light recipe execution); illumination data acquisition monitors dynamic changes in illumination components through high-frequency SPD calculations, especially the power changes and circadian rhythm factor change rates of the 480nm blue light channel; ambient brightness change acquisition detects step changes in illumination intensity through an ambient light sensor.

[0051] The system inputs three types of changing data into an event-driven fusion model, and transforms the raw data stream into interactive events with clear semantics through pattern recognition and correlation analysis. For example, when the system detects that the user manually reduces the brightness and the CS value decreases, it determines it as a "user actively reduces rhythmic stimulation" event; when the ambient light increases and the system automatically increases the output, it determines it as a "system counteracts environmental interference" event.

[0052] The system uses the "lighting state" determined by S111 as the evaluation benchmark, performs time-series analysis on the interaction event sequence, and evaluates the cumulative effect over time through the circadian rhythm integral model (CRIM). Finally, it generates a state vector containing the full picture of the current interaction, covering multiple dimensions such as the baseline state ID, dominant interaction event, cumulative light duration, cumulative CS dose, user tolerance assessment, and interaction confidence.

[0053] refer to Figure 3 In step S12, the specific steps are as follows:

[0054] S121: Real-time monitoring of the light irradiation of the phototherapy device on the user, determining the corresponding light irradiation range based on the relative position of the phototherapy device to the user, the corresponding light irradiation posture, and the light irradiation data of the phototherapy device, and determining the light irradiation area of ​​the phototherapy device relative to the user based on the light irradiation range, the user's current posture, and overall shape;

[0055] S122: Collect multiple working data from the phototherapy device, determine multiple phototherapy data combinations based on the matching of the illumination area and the multiple working data of the phototherapy device, determine the corresponding phototherapy features based on the identification of each phototherapy data combination, and determine the current phototherapy mode of the phototherapy device based on the feature position, corresponding feature shape and overall shape of the phototherapy device of the multiple phototherapy features.

[0056] In the embodiments of this application, the illumination of the phototherapy device on the user is monitored in real time. The corresponding illumination range is determined based on the relative position of the phototherapy device to the user, the corresponding illumination posture, and the illumination data of the phototherapy device. The illumination area of ​​the phototherapy device relative to the user is determined based on the illumination range, the user's current posture, and the overall shape. This approach takes into account the overall consideration of the illumination range, the user's current posture, and the overall shape, ensuring the accuracy of the illumination area of ​​the phototherapy device relative to the user.

[0057] At this point, the light emitted by the device is transformed from abstract electrical parameters into a geometric entity with clear boundaries and intensity distribution in three-dimensional space. The system uses millimeter-wave radar or ToF sensors to measure the distance, azimuth, and pitch angles between the device and the user in real time with sub-centimeter accuracy, and constructs the position of the device in a spherical coordinate system with the user as the origin. At the same time, the three-axis attitude angles are obtained through the device's internal IMU to define the precise direction of the beam's main axis in space.

[0058] In terms of optical parameter extraction, the system reads the driving current of the current six-channel LED and calculates the total luminous flux and luminous intensity distribution in real time by combining the optical characteristics of each channel LED. Subsequently, the system uses ray tracing or point cloud diffusion algorithms to construct a dynamic "light field model" by combining optical parameters and device attitude, which describes the path and intensity attenuation of each major ray emitted from the light source. The system presets an "effective illumination threshold" (such as 1 lx) and performs isosurface clipping on the three-dimensional light field model. The closed three-dimensional geometry formed by the set of points with illumination intensity equal to or higher than the threshold is the "illumination range".

[0059] The system precisely aligns the physical "lighting range" with the user's "digital body model," identifies the specific areas affected by the light, and assigns them physiological significance. The system generates a real-time point cloud data stream containing information about the user's body reflection points using millimeter-wave radar, and processes the point cloud data using a deep learning model to achieve key point detection, pose estimation, and morphology segmentation.

[0060] During the illumination area determination stage, the system precisely aligns the "illumination range" coordinate system with the user's "digital body model" coordinate system and performs a three-dimensional Boolean intersection operation. The result of the operation is the "illumination area". The system performs fine marking and quantification of this area, including area identification (e.g., 95% head area coverage) and dose calculation (average illuminance, cumulative light dose, rhythmic stimulation factor input value, etc.).

[0061] Specifically, in the scenario where the phototherapy device helps users sleep, at 22:50, the phototherapy device has confirmed that it is currently in "sleep aid mode"; the millimeter-wave radar module of the phototherapy device is activated, and a strong reflected signal is detected 0.75 meters directly in front of the device, with an azimuth angle of 0 degrees and a pitch angle of -15 degrees. The system determines that the user is located at this position; the IMU data shows that the current pitch angle of the device is 20 degrees, ensuring that the beam can be projected onto the user's face at a suitable angle.

[0062] The system reads the current working data: CH2 (warm yellow light) 2500mA, CH5 (red light) 1000mA, and other channels are closed. Based on the SPD and beam angle data of these channels, the system constructs an ellipsoidal light field model with the device as the vertex, the main axis pointing towards the user's face, and a half-peak angle of about 45 degrees. The system trims the light field model with a threshold of 0.5lx, and finally determines an ellipsoidal "illuminance range" with a major axis of about 1.2 meters and a minor axis of about 0.8 meters in three-dimensional space.

[0063] Radar point cloud data is fed into the attitude estimation model in real time. The model successfully identifies the user's head and shoulder contours and determines that the user is currently in a "supine" posture. The system segments the user point cloud model into "head region", "torso region", etc.

[0064] The system performs a Boolean intersection operation between the "illumination range" ellipsoid and the user's "digital body model." The results show that the ellipsoid overlaps with the "head region" by as much as 98% and with the "torso region" by about 20%. Therefore, the phototherapy device ultimately determines its illumination area to "mainly cover the user's face and a small amount of the upper chest," and calculates that the average illuminance received by the user's facial area is 35 lx, with a cumulative CS dose of 0.12. This precise area and dose information becomes the core basis for subsequent personalized phototherapy adjustments.

[0065] Furthermore, multiple working data points from the phototherapy device are collected. Based on the matching of the illumination area and the multiple working data points of the phototherapy device, multiple phototherapy data combinations are determined. Among the multiple phototherapy data combinations, the corresponding phototherapy features are determined based on the identification of each phototherapy data combination. The current phototherapy mode of the phototherapy device is determined based on the feature positions, corresponding feature shapes, and overall shape of the phototherapy device of the multiple phototherapy features. This comprehensive consideration of the feature positions, corresponding feature shapes, and overall shape of the phototherapy device ensures the accuracy of the current phototherapy mode of the phototherapy device.

[0066] At this point, the abstract device operating parameters are bound to specific physical action areas to form a "data combination" with a clear direction. The system reads the registers of the six-channel LED driver in real time to obtain the precise constant current value and PWM dimming duty cycle of each channel. At the same time, it collects dynamic formula parameters such as the slope of the brightness decay curve and the color temperature switching transition time, as well as system status parameters including the ripple coefficient of the driver power supply, the junction temperature of the LED module, and the total power consumption of the system.

[0067] During the data combination generation stage, the system uses the "illuminated area" determined in S121 as a label, and encapsulates all the working data collected at the current moment with the label to form a "phototherapy data combination" under the timestamp. This process continues at a high frequency (e.g., 1Hz) to generate a queue of "phototherapy data combinations" arranged in time sequence. Each combination describes "at a specific time point, the light acting on a specific area is generated by which working parameters".

[0068] From the original "data combination", "phototherapy features" that can directly describe the physiological effects and visual characteristics of light are extracted through physical models and algorithms; the system is based on six-channel drive current and SPD of each channel, and synthesizes the current spectral power distribution in real time through Grassmann's law, extracting key features such as correlated color temperature, color rendering index, R9 value and specific band radiation power.

[0069] Simultaneously, the system calculates the illuminance, luminous flux, and rhythmic stimulation factors of the target area based on SPD and photometric functions, and calculates the rate of change of illumination parameters, including the rate of change of illuminance, the rate of change of color temperature, and the period and frequency of dynamic light formulation, by analyzing a continuous "data combination" queue; finally, all features are transformed into a high-dimensional numerical vector, which is the "phototherapy feature" vector.

[0070] The system compares the real-time extracted "phototherapy feature" vectors with the system's built-in "phototherapy mode library" and uses algorithms to identify the currently executing phototherapy mode. The system analyzes the coordinate position and distribution of the feature vectors in the high-dimensional feature space. For example, extremely low CS and CCT values ​​constitute a "sleep aid" mode, while high CS and CCT values ​​constitute a "wake-up" mode.

[0071] The system has a built-in feature library containing multiple phototherapy modes. Each mode has a typical "feature pattern" as a template. For example, the sleep aid mode requires CCT < 2500K, CS < 0.2, and illuminance < 50lx. The system uses a machine learning classifier or a rule-based similarity algorithm to compare the real-time feature vector with the mode library template and outputs the mode label with the highest matching degree and confidence score. When the confidence score exceeds the preset threshold, the system finally determines the "current phototherapy mode".

[0072] Specifically, the phototherapy device has been designed to "primarily cover the user's face, with limited coverage of the upper chest"; the device collects working data at a frequency of 1Hz: CH2 (warm yellow light) 2500mA, CH5 (red light) 1000mA, with other channels closed; the dynamic formula display shows that the brightness is linearly decreasing at a rate of 0.5lx per minute; the system binds this data to the "lighting area: face and neck" label to form a "phototherapy data combination".

[0073] The system calculates the "phototherapy features" in real time based on this data combination: after synthesizing SPD, the calculated CCT is 2050K, CRI is 97, and the 480nm blue light radiation power is close to 0; the average illuminance of the facial area is 35lx, and the CS value is 0.11; the illuminance change rate dE / dt is -0.5lx / min; the system generates the feature vector: [2050K,97,0.11,35lx,-0.5lx / min,“Face”).

[0074] The system inputs the real-time feature vector into the phototherapy mode classifier. The calculation shows that the similarity with the "sleep aid mode" template is as high as 98%, the similarity with the "reading mode" is less than 20%, and the similarity with the "wake-up mode" is less than 5%. Since the matching confidence with the "sleep aid mode" is far higher than the threshold, the phototherapy device finally determines its current phototherapy mode as the "sleep aid mode". This accurate pattern recognition result provides the most direct contextual basis for judging the "sleep aid event" in the subsequent S13.

[0075] refer to Figure 4 In step S13, the specific steps are as follows:

[0076] S131: Collect multiple motion images of the user, determine the corresponding motion change content based on the recognition of the multiple motion images of the user, determine the first sleep aid content based on the motion change content and the user's previous sleep events, determine the second sleep aid content based on the current light therapy mode of the light therapy device and the motion change content, and determine the current sleep aid event of the light therapy device based on the first sleep aid content and the second sleep aid content.

[0077] S132: Collect the user's facial image, determine the user's facial expression based on the recognition of the user's facial image, and determine the first light effect combination based on the current sleep aid event and the user's facial expression;

[0078] S133: Mark the environmental space constructed by the user and the phototherapy device, determine multiple environmental parameters based on the detection of the environmental space, determine a second light influence combination based on the current sleep aid event and the multiple environmental parameters, and determine multiple key light factors of the phototherapy device based on the first light influence combination and the second light influence combination.

[0079] In the embodiments of this application, multiple motion images of the user are collected, and corresponding motion change content is determined based on the recognition of the multiple motion images of the user. First sleep aid content is determined based on the motion change content and the user's previous sleep events. Second sleep aid content is determined based on the current light therapy mode of the light therapy device and the motion change content. The current sleep aid event of the light therapy device is determined based on the first sleep aid content and the second sleep aid content. This approach takes into account both the first and second sleep aid content as a whole, ensuring the accuracy of the current sleep aid event of the light therapy device.

[0080] At this point, the system captures the user's dynamic behavior through non-contact sensors and transforms it into machine-understandable quantitative data. The system uses millimeter-wave radar as the core sensor, which generates a point cloud data stream containing distance, speed, and angle information by transmitting and receiving electromagnetic waves. This continuous point cloud sequence constitutes a "motion image" describing the macroscopic and microscopic movements of the human body. The raw point cloud data is then subjected to noise reduction and trajectory smoothing through algorithms such as Kalman filtering to eliminate environmental noise and sensor measurement errors.

[0081] In terms of action change content recognition, the system uses a sliding window mechanism to analyze continuous point cloud sequences. It extracts temporal features by calculating the centroid displacement, Doppler spectrum distribution, and motion trajectory of key reflection points within the window. These features are then input into a pre-trained deep learning model (such as LSTM or GRU), which is adept at processing time series data and can recognize complex action patterns. Finally, the model outputs a standardized "action change content" vector, which objectively describes the user's current action state.

[0082] Based on the changes in the action and the user's past sleep events, the system determines the first sleep aid content and correlates the user's real-time behavior with their long-term sleep patterns to infer the underlying reasons for the behavior. The system accesses the user's personal health database through an encrypted API, integrates long-term data from wearable devices, mobile apps, and medical institutions, and queries and extracts key historical sleep events, such as average sleep latency, number of nighttime awakenings, sleep efficiency, and sleep quality scores related to specific times.

[0083] The system matches real-time "action changes" with patterns in the historical database. If it finds that a user's sleep quality has often declined after a certain behavior in the past, a strong correlation is established. Based on this correlation, the system infers the "first sleep aid content," which is a personalized interpretation of the current behavior. For example, the user's current frequent turning over behavior is a typical precursor to difficulty falling asleep.

[0084] The system determines the second sleep aid content based on the current light therapy mode of the light therapy device and the change in movement. The system obtains the "current light therapy mode" and its specific parameters determined in S12, and uses the built-in knowledge base to define the ideal behavioral response of the user under a specific light therapy mode. The system compares the real-time "change in movement" with the ideal response. When it finds that there is a significant deviation between the actual behavior and the ideal response, it infers the "second sleep aid content", which is an evaluation of the effect of the current light therapy plan. For example, even though the device is in sleep aid mode, the user still shows restlessness, indicating that the sedative effect of the current light formula has not reached the expected level.

[0085] The system assigns different weights to the "first sleep aid content" and the "second sleep aid content," or uses a Bayesian network to use the "first content" as the prior probability and the "second content" as the current observation evidence, and calculates the posterior probability to obtain a probabilistic "sleep aid event" judgment. The fused result is a structured and semantically clear description of the "current sleep aid event," which not only describes the phenomenon but also includes the cause and effect evaluation. For example, if the user is in the early stages of falling asleep and shows difficulty falling asleep due to historical anxiety tendencies, and the current light therapy plan has insufficient sedative effect and needs adjustment.

[0086] Specifically, at 23:15, the phototherapy device confirmed that it was currently in "sleep aid mode" and was being applied to the user's face; the millimeter-wave radar of the phototherapy device continuously collected point cloud data in the past 5 minutes, and after analysis by the built-in LSTM model, it output the "movement change content" vector: [turning frequency: 4 times / minute, body movement amplitude: moderate to high, respiratory rate: 18 times / minute, respiratory stability: 0.4].

[0087] The system queried the user's sleep database and found that the user had exhibited similar "frequent tossing and turning" behavior 70% of the nights in the past month between 23:00 and 23:30, and the average sleep latency on these nights exceeded 30 minutes; based on this, the system determined the first sleep aid content to be: "The user's current behavior is consistent with their typical sleep difficulty pattern".

[0088] The system obtains the ideal response model for the current "sleep aid mode": In this mode, the user should reduce the amplitude of body movements and the respiratory rate should drop to below 15 breaths / minute within 10 minutes; Comparing the real-time "movement changes" with the ideal model, it is found that the user's behavior is far from meeting expectations; The system determines the second sleep aid content as: "The current light therapy plan has failed to effectively inhibit the user's nerve excitability."

[0089] The system uses a weighted fusion algorithm, combining the "first content" and the "second content", to finally determine the current sleep aid event as: "The user has entered their typical sleep-falling cycle, the current light therapy program is not strong enough, and the sedative effect needs to be enhanced to break this cycle"; this accurate event diagnosis will directly drive the effect coefficient calculation and optimized content generation in S14.

[0090] Furthermore, the system collects the user's facial images, determines the user's facial expressions based on the recognition of the facial images, and determines the first layer of lighting influence combination based on the current sleep aid event and the user's facial expressions. This comprehensive consideration of the current sleep aid event and the user's facial expressions ensures the accuracy of the first layer of lighting influence combination.

[0091] At this point, by analyzing subtle changes in the user's face, an intuitive emotional dimension is added to the evidence for the "sleep-aiding event"; with the user's explicit authorization, the device uses a low-resolution near-infrared (NIR) or thermal imaging camera, both of which can work in complete darkness and will not emit visible light to disturb the user's sleep; the camera captures images or thermal data streams of the user's face at a low frame rate (such as 1-2 fps) to form a sequence of "facial images".

[0092] In terms of facial expression recognition, all image processing is performed on the device's local embedded neural network accelerator (NPU). The raw image data is destroyed immediately after analysis and is never uploaded to the cloud to ensure user privacy. The system runs a lightweight convolutional neural network (CNN) model, which is specifically designed to detect facial key points in low-light or thermal imaging data and further identify facial action units, such as frowning (AU4) representing "discomfort" and relaxing the corners of the mouth (AU12+25) representing "relaxation". The model maps the detected AU combinations to quantized expression vectors, such as [tension: 0.8, relaxation: 0.1, pain: 0.4]. Then, the classifier converts the vector into discrete expression labels, such as "anxiety", "discomfort", "relaxation" or "calm".

[0093] The macroscopic "sleep-aiding event" inferred in S131 is correlated with the microscopic "facial expression" identified in S132 to form a closed-loop hypothesis about how light affects the user's emotions. The system receives two key inputs: the "current sleep-aiding event" determined in S131 and the "facial expression" identified in S132. The consistency between the two inputs is verified by the built-in rule engine or a simple Bayesian network.

[0094] When the consistency is "high", the system strongly binds these two pieces of information to form the "first layer of light influence combination", the core conclusion of which is that "the physiological regulatory effect of light is strongly correlated with the user's emotional state". This combination is not only descriptive, but also instructive, indicating the primary goal of the current light regulation. The output is a structured instruction set: {goal: "relieve anxiety", priority: "highest", evidence: "behavioral agitation + anxious expression"}. This combination will serve as the key input for determining the priority of "key light factors" in S14.

[0095] Specifically, at 23:20, the phototherapy device determined the current sleep-aiding event to be "the user has entered their typical sleep-falling difficulty cycle, the current phototherapy intensity is insufficient, and the sedative effect needs to be enhanced to break this cycle"; the near-infrared camera of the phototherapy device captured a facial image of the user with the user's authorization; the CNN model on the local NPU analyzed the image and detected that the user's frown lines were deepened and the corners of the mouth were slightly downturned; the model calculated the expression vector as [tension: 0.75, relaxation: 0.2], and finally determined the user's facial expression as "anxiety".

[0096] The system compares the keyword "anxious difficulty falling asleep" in the "current sleep aid event" with the "anxiety" tag in "facial expression". The cross-validation logic determines that the two are highly consistent, with a confidence level of 92%. Based on this, the system forms the first layer of light influence combination, the core conclusion of which is: "The user's anxiety is the core driving factor of the current difficulty falling asleep, and light regulation must prioritize intervention against this emotion". This combination is transformed into a specific optimization goal and given the highest priority.

[0097] Therefore, the environmental space constructed by the user and the phototherapy device is marked, and multiple environmental parameters are determined based on the detection of this environmental space. A second light influence combination is determined based on the current sleep-aiding event and multiple environmental parameters. Based on the first and second light influence combinations, multiple key light factors of the phototherapy device are determined. This approach is compatible with the overall consideration of the first and second light influence combinations, ensuring the accuracy of multiple key light factors of the phototherapy device. At the same time, the current phototherapy mode of the phototherapy device is introduced, enabling further control over the current sleep-aiding event and improving the accuracy of multiple key light factors of the phototherapy device.

[0098] At this point, the scope of light therapy is expanded from the binary system of "human-light" to the ternary system of "human-light-environment", aiming to identify physical environmental factors that affect sleep. The system uses point cloud data from millimeter-wave radar to outline the main objects in the room, such as the bed, wardrobe, and windows, to construct a simplified two-dimensional or three-dimensional "environmental space model" and mark key areas such as the "sleep zone", "window zone", and "equipment zone".

[0099] In terms of environmental parameter detection, the device has a built-in integrated environmental sensor module that communicates with the main control MCU via I²C or SPI bus to collect and quantify multiple environmental parameters in real time. These include thermal environment (air temperature and relative humidity), acoustic environment (sound pressure level is collected through MEMS microphone and background noise level is obtained through FFT analysis), light environment (light intensity and color temperature of non-device light are detected through a high-sensitivity ambient light sensor), and air quality (particulate matter concentration is detected through a miniature PM2.5 laser sensor).

[0100] The system performs correlation analysis between environmental factors and sleep state, assesses whether environmental parameters are synergistic factors or sources of interference in the current "sleep-aiding event", and defines how light should respond to these environmental challenges. The system has a built-in knowledge base based on sleep science and environmental psychology, which includes rules such as high temperature leading to decreased sleep quality and increased noise increasing the risk of difficulty falling asleep.

[0101] The system maps and analyzes the "current sleep aid event" determined in S131 with the environmental parameters detected in real time to identify synergistic or interfering factors. For example, if the room temperature is detected to be 27°C, the system will identify it as "thermal interference," which forms a synergistic negative effect with "anxious difficulty falling asleep." Based on this identification, the system forms a "second light influence combination" and defines light compensation strategies, such as using visual and physiological means to help users achieve psychological "cooling" and relaxation.

[0102] By taking into account the dual influences of the user's internal emotions and the external environment, a set of specific, actionable, and prioritized "key lighting factors" is determined through a multi-objective optimization algorithm. The system receives two "lighting influence combinations" as input. The first combination defines the core physiological goal, and the second combination defines the auxiliary environmental goal. Weights are assigned to different goals based on the severity and urgency of the event.

[0103] The system maps the weighted objectives to specific lighting parameters. For example, to "alleviate anxiety," it is necessary to lower the color temperature, increase the proportion of amber light, and slow down the rate of change in brightness; to "compensate for thermal interference," it is necessary to enhance the visual coolness and maintain low illuminance. Finally, a multi-objective optimization algorithm is used to optimize the adjustment of all lighting parameters, find the solution that best satisfies both objectives at the same time, and output a list of "key lighting factors" sorted by priority.

[0104] Specifically, at 23:25, the phototherapy device had formed the first layer of light influence combination: "The user's anxiety is the core driving factor for current difficulty falling asleep, and light regulation must prioritize intervention against this emotion"; the environmental sensor module of the phototherapy device detected: air temperature of 27°C, relative humidity of 65%, background noise of 38dB (from the air conditioner outdoor unit), and ambient light of 0.2lx (very dark); the system integrated these data into multiple environmental parameters.

[0105] The system associates the "anxious difficulty falling asleep" event with a room temperature of 27°C and determines from the knowledge base that "thermal interference" is an important factor that exacerbates anxiety. The system then forms a second layer of light influence combination: "Light needs to help users achieve psychological relaxation in a slightly warm environment and counteract the irritability caused by heat discomfort."

[0106] The system integrates the effects of the first layer (anxiety relief) and the second layer (compensation for thermal interference), and through multi-objective optimization, finally determines several key lighting factors and prioritizes them: Spectral composition (highest priority): While maintaining extremely low rhythmic stimulation (CS<0.1), the weight of amber light (CH4) is increased by 10%, while the weight of red light (CH5) is slightly adjusted by 5%. This spectral combination is more "cool" and "calming" visually and psychologically; Dynamic changes (second highest priority): The brightness decay rate is further reduced from -0.25lx / min to -0.1lx / min, providing an extremely smooth transition and avoiding any dynamic changes that may cause irritation; Color temperature (medium priority): The target color temperature is maintained at 1900K and will not be lowered further, because under this spectral composition, 1900K can provide the best visual comfort.

[0107] refer to Figure 5 In step S14, the specific steps are as follows:

[0108] S141: Collect multiple sub-rhythm data of the phototherapy device, determine the current rhythm content of the phototherapy device based on the multiple sub-rhythm data and the corresponding light state, determine the multiple light influence content based on the current rhythm content of the phototherapy device and multiple key light factors, and determine the sleep aid effect coefficient of the phototherapy device based on the priority of the multiple light influence content and multiple key light factors.

[0109] S142: Collect multiple sub-sleep state data of the user, determine the user's sleep state based on the identification of multiple sub-sleep state data, and determine the first level of optimization combination based on the user's sleep state and the sleep-aiding effect coefficient of the light therapy device;

[0110] S143: Determine a second set of optimization combinations based on the user's sleep state and multiple light data from the phototherapy device, and determine multiple optimization items based on the first set of optimization combinations, the second set of optimization combinations, and the current sleep aid events of the phototherapy device.

[0111] In the embodiments of this application, multiple sub-rhythm data of the phototherapy device are collected. The current rhythm content of the phototherapy device is determined based on the multiple sub-rhythm data and the corresponding light state. The multiple light influence content is determined based on the current rhythm content of the phototherapy device and multiple key light factors. The sleep-aiding effect coefficient of the phototherapy device is determined based on the priority of the multiple light influence content and multiple key light factors. This approach takes into account the overall consideration of the multiple light influence content and the priority of multiple key light factors, ensuring the accuracy of the sleep-aiding effect coefficient of the phototherapy device.

[0112] At this point, the physical light parameters output by the device are converted into quantitative indicators with clear physiological significance, constructing a "digital profile" describing the impact of light on the user's physiological system; the system calculates multiple key "sub-rhythm data" in real time at high frequency (e.g., >1Hz), including the rhythm factor (CS) calculated based on the circadian rhythm integral model (CRIM) and nonlinear light response function (NLRF), the retinal melanopsin equivalent illuminance (Emel) calculated according to CIE standards, and the cumulative light dose obtained by integrating the CS value or Emel value over time.

[0113] During the current rhythm content determination stage, the system performs spatiotemporal alignment and fusion of the real-time calculated "sub-rhythm data" with the "illuminance status" (such as color temperature and illuminance) determined in S11, and finally generates a multi-dimensional "current rhythm content" vector that accurately describes the physiological characteristics of the current illumination, for example: {CS:0.09, Emel:8.5lx, cumulative dose:0.15lx·h, CCT:1900K, illuminance:28lx}.

[0114] The "key lighting factors" identified in S13 are correlated with the "current rhythm content" quantified in S141.1 to explain how each factor specifically affects physiological indicators. The system establishes a dynamic model to describe how changes in lighting parameters affect rhythm indicators. For example, the model knows that increasing the amber light weight will slightly reduce the CS value, but has a positive impact on visual comfort.

[0115] The system cross-analyzes "current rhythm content" and "key lighting factors" to generate a series of "multiple lighting effects content". Each content is a specific causal description, such as: {Influence 1: "Amber light weight increased by 15%, resulting in a decrease in CS value from 0.11 to 0.09 and a 20% increase in visual calmness"}, or {Influence 2: "The rate of brightness decay slowed down, resulting in a reduction in negative fluctuations in the user's heart rate variability (HRV) and a decrease in physiological stress level"}.

[0116] All the scattered "light-affecting content" are weighted and aggregated according to their importance to form a single, comparable value that measures the overall sleep-aiding efficacy of the current light therapy program; the system assigns weights to the corresponding "light-affecting content" according to the priority of the "key light factors" determined in S13.

[0117] The system employs the Analytic Hierarchy Process (AHP) or a simple weighted summation model. Each "lighting effect content" has not only a physiological effect value but also a psychological effect evaluation value. The system sums the weighted scores of all effect content to obtain the original total score, which is then normalized to the interval [0,1] using a Sigmoid function or linear mapping. This final value is the "sleep-aiding effect coefficient." For example, a coefficient of 0.85 indicates that the current lighting scheme, after comprehensively considering all key factors and their priorities, has achieved 85% of the ideal sleep-aiding efficacy.

[0118] Specifically, at 23:30, the phototherapy device adjusted its illumination parameters based on the conclusion of S133 (increasing amber light and slowing down dynamic changes); the real-time calculation module of the phototherapy device continued to work and collected the latest "sub-rhythm data": CS value 0.09, Emel 8.5 lx, cumulative dose 0.15 lx·h; combined with the current "illumination status" (CCT 1900K, illuminance 28 lx), the system generated the current rhythm content vector: {CS:0.09, Emel:8.5 lx, cumulative dose:0.15 lx·h, CCT:1900K, illuminance:28 lx}.

[0119] The system analyzes these changes and generates multiple lighting effects: {Impact 1: "The weight of amber light (CH4) increases from 800mA to 1200mA, causing the CS value to drop from 0.11 to 0.09, significantly enhancing the visual 'warmth' and 'envelopment' feeling"}; {Impact 2: "The brightness decay rate decreases from -0.25lx / min to -0.1lx / min, causing the lighting changes to no longer be subconsciously perceived by users, reducing the risk of micro-awakening caused by environmental changes"}.

[0120] Based on the priority of S133 ("anxiety relief" is the highest), the system assigns a weight of 0.7 to "Influence 1" and a weight of 0.3 to "Influence 2". Through internal model calculations, the system finds that the combined score of these two influences is very high. After normalization, the system finally determines that the sleep-aiding effect coefficient of the current light therapy program is 0.88. This high score indicates that, from the perspective of theory and physiological models, the current light therapy program is very close to the ideal sleep-aiding state.

[0121] Furthermore, multiple sub-sleep state data of the user are collected, and the user's sleep state is determined based on the identification of the multiple sub-sleep state data. The first layer of optimization combination is determined based on the user's sleep state and the sleep-aiding effect coefficient of the light therapy device. This takes into account both the user's sleep state and the sleep-aiding effect coefficient of the light therapy device, ensuring the accuracy of the first layer of optimization combination.

[0122] At this point, transforming the user's original physiological signals into standardized sleep stage classifications through multi-source data fusion and machine learning models is a key step in evaluating the actual effect of phototherapy. The system synchronously collects multiple "sub-sleep state data" from the user's wearable device at a frequency of 1Hz via Bluetooth Low Energy (BLE) 5.0 protocol, including heart rate (HR), RMSSD index of heart rate variability (HRV), respiratory rate (RR), and body movement records based on accelerometer.

[0123] Meanwhile, the millimeter-wave radar of the phototherapy device provides independent non-contact respiratory rate and body movement data to cross-validate the accuracy of the wearable device data and handle abnormal situations such as the device not being worn or signal loss. In the sleep state recognition stage, the system preprocesses and extracts features from the collected time series data, inputs the feature vectors into a pre-trained sleep stage classification model, which is usually based on random forest, SVM or lightweight LSTM, and finally outputs a discrete "sleep state" label, such as "awake (W)", "light sleep (N1 / N2)", "deep sleep (N3)" or "rapid eye movement (REM)", with a confidence score.

[0124] The system compares the "theoretical effect" (sleep-aid effect coefficient) of the light therapy program with the user's "actual state" (sleep state) to identify the limitations of the current program and define the optimization direction for the next stage. The system establishes a state-effect expectation mapping table. For example, when the "sleep-aid effect coefficient" is greater than 0.8, the expected "sleep state" should be "light sleep" or deeper.

[0125] The system compares the "sleep-aiding effect coefficient" calculated by S141 with the "sleep state" identified by S142.1. When the actual state does not reach the expected state, the system identifies an "optimization gap." Based on this gap, the system forms a structured "first-level optimization combination," the core of which is a qualitative description of the current problem and implies the direction of the optimization strategy. For example, if the effect coefficient is high but the user is still awake, it means that the problem is not in the physiological intensity of light, but in other dimensions. The final generated combination is a binary tuple containing "problem description" and "strategy direction," such as: {Problem description: "The physiological effect of light therapy has reached the standard, but the user is still not asleep," Strategy direction: "It is necessary to break through the current physiological stimulation framework and explore new sleep-inducing pathways"}.

[0126] Specifically, at 23:35, the phototherapy device calculated the current sleep aid effect coefficient to be 0.88 and continued to monitor the user. The phototherapy device collected data from the user's smart bracelet via BLE: heart rate 72 bpm, HRV RMSSD value 25 ms (low, indicating sympathetic dominance), respiratory rate 16 breaths / minute, and 3 small body movements in the past 5 minutes. This data was input into the sleep staging model on the device, and the model output the result: sleep state "awake" with a confidence level of 95%.

[0127] The system compares the "awake" state with a "sleep-aiding effect coefficient" of 0.88. According to the internal expectation mapping table, a coefficient of 0.88 should guide the user into a "light sleep" state. The system detects a significant "optimization gap": the theoretical effect is good, but the actual effect is poor. Based on this, the system determines the first optimization combination to be: {Problem description: "The physiological sedative effect of the current light therapy program is close to maximizing, but it fails to effectively guide the user from wakefulness to sleep", strategy direction: "The optimization focus should shift from enhancing physiological stimulation to adjusting the stimulation mode or introducing new sensory covariates"}. This combination clearly indicates that simply "strengthening" the existing program is ineffective and a "change" strategy is needed.

[0128] Therefore, a second set of optimization combinations is determined based on the user's sleep state and multiple light irradiation data from the phototherapy device. Based on the first set of optimization combinations, the second set of optimization combinations, and the current sleep-aiding events of the phototherapy device, multiple optimization items are determined. This approach takes into account the overall consideration of the first set of optimization combinations, the second set of optimization combinations, and the current sleep-aiding events of the phototherapy device, ensuring the accuracy of the multiple optimization items.

[0129] At this point, the system accurately correlates the user's "sleep state" (e.g., "awake") with the device's current "light data" (e.g., six-channel drive current, PWM duty cycle, dynamic recipe parameters), and generates one or more "adaptive hypotheses" based on this correlation. For example, the system hypothesizes that the user has adapted to the currently dominant amber light spectrum, resulting in a weakened sedative effect; or it hypothesizes that the current constant illuminance lacks variation and cannot effectively guide the user's autonomic nervous system into a synchronized state. Based on these hypotheses, the system forms a "second-level optimization combination," the core of which is to define specific parameter adjustment strategies, focusing more on specific, executable parameter exploration directions, such as: {Strategy direction: "Fine-tuning within the existing spectral framework to break sensory adaptation; or introducing ultra-low frequency dynamic modulation to attempt physiological rhythm synchronization"}.

[0130] The decision engine receives three core inputs: the macro-level problem defined in S142 (the first layer of combinations to be optimized), the micro-level entry point defined in S143 (the second layer of combinations to be optimized), and the fundamental background provided in S131 (the current sleep-aiding event). The decision engine uses these inputs as constraints and objective functions, and solves them through a rule-based or reinforcement learning model. For example, the rule could be: IF event CONTAINS "anxiety" AND policy IS "change" THEN prioritize "synchronization of physiological rhythms". The engine output is a series of structured "content to be optimized", each of which is an instruction containing "operation object", "target value", and "execution conditions", and is sorted according to its expected effect, risk, and relevance to the core problem.

[0131] Specifically, at 23:40, the phototherapy device has determined the first optimization combination (requiring a strategy change) and the second optimization combination (fine-tuning the spectrum or introducing dynamic modulation), and knows that the user is currently in a "awake" state. System analysis found that the user, in the "awake" state, is continuously exposed to a light environment with relatively stable spectrum and illuminance. The system generates the hypothesis that this stable state leads to sensory adaptation, or a lack of a "rhythmic anchor point" to guide the user into sleep. Based on this, the system determines the second optimization combination as: {Strategy direction: "Attempt to introduce ultra-low frequency dynamic modulation, using the user's breathing rhythm as an anchor point, to synchronously guide physiological rhythms"}.

[0132] The decision engine begins to integrate all information: Input 1 (first level): The strategy needs to be changed; simply strengthening sedation is no longer sufficient; Input 2 (second level): Attempt circadian rhythm synchronization; Input 3 (sleep-inducing event): The user experiences anxiety, and circadian rhythm synchronization has been proven effective in alleviating anxiety and relaxing the mind and body; The decision engine reasones and concludes that "circadian rhythm synchronization" is the optimal path to simultaneously satisfy all inputs; The system identifies several areas for optimization and prioritizes them:

[0133] Optimization Item 1 (Highest Priority): Activate the breathing synchronization function; the system accurately locks the user's breathing rate (16 breaths / minute, i.e., 0.267Hz) via radar and generates an instruction: on the basis of the current illuminance of 28lx, superimpose a sinusoidal brightness modulation with a frequency of 0.267Hz and an amplitude of ±2%; Optimization Item 2 (Backup Plan): If breathing synchronization is ineffective within 5 minutes, perform spectral fine-tuning; Instruction: without increasing the CS value, increase the driving current of the CH3 (480nm blue light) channel from 0mA to 20mA, introduce a trace amount of "skylight" component, and attempt to break the spectral adaptability; Optimization Item 3 (Basic Guarantee): Continue to execute the original slow voltage reduction strategy as background optimization; Instruction: maintain an illuminance decay rate of -0.05lx / min.

[0134] refer to Figure 6 In step S15, the specific steps are as follows:

[0135] S151: Real-time monitoring of multiple sleep postures of the user at different times, determining the user's sleep posture change events based on multiple sleep postures, and determining the first level of intelligent control content of the phototherapy device based on multiple content to be optimized and the user's sleep posture change events.

[0136] S152: Determine the second level of intelligent control content of the phototherapy device based on multiple content to be optimized and the current phototherapy mode of the phototherapy device, and determine the intelligent control event of the phototherapy device based on the first level of intelligent control content and the second level of intelligent control content;

[0137] S153: Identify intelligent control events and determine the corresponding intelligent lighting content. At the same time, collect the user's surrounding environment events. Based on the intelligent lighting content and the user's surrounding environment events, determine the first rhythm control content. Based on the intelligent lighting content and the user's sleep duration, determine the second rhythm control content. Based on the training of the first and second rhythm control content, determine the rhythm control system of the phototherapy device.

[0138] In the embodiments of this application, multiple sleep postures of the user at different times are monitored in real time, and sleep posture change events of the user are determined based on multiple sleep postures. At the same time, the first level of intelligent control content of the phototherapy device is determined based on multiple content to be optimized and the user's sleep posture change events. This takes into account the overall consideration of multiple content to be optimized and the user's sleep posture change events, and ensures the accuracy of the first level of intelligent control content of the phototherapy device.

[0139] At this point, an event-driven triggering mechanism is established to provide precise execution timing for intelligent control by accurately identifying changes in user behavior during sleep. The system uses millimeter-wave radar as the core sensor to generate a four-dimensional point cloud data stream containing distance, speed, and angle information at a high frequency (e.g., 10-20Hz). The point cloud data stream is input into a real-time attitude estimation algorithm based on a deep learning model, which can extract key points of the human skeleton from the sparse point cloud and infer the orientation of the torso.

[0140] The model maps the inferred skeleton information to discrete sleep posture categories, such as "supine", "side-lying_left", "side-lying_right", "prone", etc., and outputs a time-series posture label stream at a high frequency (e.g., 5Hz). The system maintains a finite state machine (FSM). When the posture label stream stably transitions from one state to another within N consecutive time windows (e.g., about 600ms), the state machine triggers a "sleep posture change event". This event is encapsulated into a structured data object containing event type, source posture, target posture, and timestamp.

[0141] The detected user behavior events are matched with the optimization strategies generated in S14 that are waiting to be executed, thereby transforming the static strategies that are "to be optimized" into dynamic control content that is "about to be executed". The system has a built-in dynamic rule base, which is based on sleep science and human factors engineering knowledge and defines which content to be optimized should be triggered by specific events.

[0142] When a "sleep posture change event" occurs, the system will match its attributes with all rules in the rule base. Once a match is successful, the system will extract the corresponding "content to be optimized" and transform it into a specific, executable "first-level intelligent control content". This content is usually a high-level control instruction, which includes the operation type and target parameters. It is encapsulated into a structured data object and placed in the execution queue, waiting for the next step of scheduling.

[0143] Furthermore, based on multiple optimization criteria and the current phototherapy mode of the phototherapy device, the second level of intelligent control content of the phototherapy device is determined. Based on the first and second levels of intelligent control content, the intelligent control events of the phototherapy device are determined, which takes into account the overall consideration of the first and second levels of intelligent control content and ensures the accuracy of the intelligent control events of the phototherapy device.

[0144] At this point, a proactive, pattern-based control strategy is established to complement the event-driven control of S151, ensuring that the system can proactively adjust according to the current pattern and optimization goals even without specific behavior triggers. The "current phototherapy mode" in the system is not a static label, but a dynamic behavior model that includes a state machine, time axis, and parameter thresholds, defining the goals at different time stages.

[0145] The system runs an evaluation task at a low frequency (e.g., every minute) to check which of the "multiple items to be optimized" meet the goals of the current mode stage. When a certain "item to be optimized" highly matches the goals of the current mode stage, the system will proactively convert it into "second-level intelligent control content," which is a preventative or maintenance control. This control content usually comes with execution conditions, and its priority is usually lower than the event-driven first-level control content. For example: {Control ID: "CTRL_101", Operation: "Spectral fine-tuning", Parameter: {Goal: "Increase the proportion of 660nm red light by 5%"}, Condition: "Expected to enter the deep sleep stage 5 minutes before"}.

[0146] The system integrates control requests from event-driven (first level) and pattern-driven (second level) approaches, resolves potential conflicts, and ultimately forms a unified and unambiguous set of execution instructions. The system maintains a control instruction queue, in which the control content generated by S151 and S152.1 is placed.

[0147] The system has a built-in conflict resolution module that processes conflicts based on preset arbitration rules: the first level of event-driven control usually has the highest priority; if two control operations require opposite adjustments to the same parameter, the higher-priority instruction will override the lower-priority instruction; some control operations require exclusive use of hardware resources, and when a higher-priority instruction is executed, other lower-priority instructions of the same type will be queued or ignored; after fusion and arbitration, the system encapsulates all instructions that are finally determined to be executed into a single "intelligent control event," which is a highly structured data packet containing all actions that need to be completed in the next execution cycle and their metadata.

[0148] Therefore, intelligent control events are identified, and corresponding intelligent lighting content is determined. Simultaneously, events related to the user's surrounding environment are collected. Based on the intelligent lighting content and these events, the first rhythmic control content is determined. Based on the intelligent lighting content and the user's sleep duration, the second rhythmic control content is determined. The rhythmic control system of the phototherapy device is then determined through training on both the first and second rhythmic control contents. This comprehensive approach ensures the accuracy of the phototherapy device's rhythmic control system. Furthermore, several optimization elements are introduced to further control the intelligent lighting content, achieving greater accuracy in intelligent lighting content, user's surrounding environmental events, and corresponding sleep duration, thus improving the precision of the phototherapy device's rhythmic control system.

[0149] At this point, the abstract "control event" is deconstructed into physical parameters that the driver can directly understand and execute; the system receives the "intelligent control event" data packet output by S152 and parses its internal "instruction set", and puts the instructions into different execution queues according to the type and priority of the instructions through a real-time scheduler.

[0150] For each instruction to be executed, the system translates it into a specific hardware control value through a parameter mapping table; for example, the instruction "start breathing synchronization" will be mapped to the dynamic modulation parameters of the PWM controller; the system writes these final control values ​​into the corresponding registers of the six-channel LED driver through the hardware abstraction layer (HAL), forming "intelligent lighting content" that directly determines the physical characteristics of LED light emission.

[0151] A feedforward control loop was established to enable the phototherapy system to actively counteract external environmental interference and maintain a stable sleep microenvironment. The system's environmental sensor module works continuously, and when it detects a parameter change that exceeds a preset threshold, it triggers a "surrounding environment event." For example, if the ambient light sensor detects a sudden increase in illuminance, it triggers a "light pollution event."

[0152] The system has a built-in feedforward compensation model. When an "environmental event" occurs, the model immediately calculates a compensation amount to offset its potential impact on the user's sleep. The system then adds this compensation amount to the current "intelligent lighting content" to form "first rhythm control content." This is a real-time, fine-tuning control mechanism, such as instantly increasing the light intensity to cope with "light pollution events" in order to maintain a constant total illuminance perceived by the user's face.

[0153] The introduction of a time dimension enables the phototherapy strategy to dynamically evolve according to the different stages of the user's sleep cycle, in order to match the physiological needs of different stages; when the system first confirms that the user has entered a "light sleep" state, it starts a high-precision timer to continuously track the user's "sleep duration".

[0154] The system has a strategy map based on sleep science that associates different sleep duration intervals with different light targets. When the timer enters a new duration interval, the system automatically generates a new "second rhythm control content" and makes baseline adjustments to the "intelligent light content". For example, when the sleep duration reaches 90 minutes, the proportion of specific wavelength red light is increased to promote deep sleep.

[0155] Through closed-loop feedback, each control process is transformed into training data, continuously optimizing its internal decision-making model, and ultimately forming a highly personalized, dynamically evolving "rhythm control system". The system uses each control action, triggering event, and subsequent changes in the user's sleep state as input features and output labels to form training samples.

[0156] These (input, output) data pairs are used to train a reinforcement learning (RL) model or an online gradient descent model, with the goal of learning an optimal policy, that is, choosing the optimal control action given the current state to maximize long-term rewards (such as total deep sleep duration). This continuously trained and optimized model constitutes a "rhythm control system", which is no longer a fixed set of rules, but an intelligent decision-making core that can self-adjust according to the user's unique physiological response patterns.

[0157] Please see Figure 7 , Figure 7 This is a schematic diagram of the structural composition of the intelligent control device of the phototherapy equipment in this embodiment of the invention; the intelligent control device of the phototherapy equipment includes:

[0158] The light-based interactive state module 21 is used to determine the light state of the light therapy device based on multiple light data of the light therapy device and the corresponding usage scenario; and to determine the corresponding light-based interactive state based on the light state of the light therapy device and the interaction data of the light therapy device relative to the user.

[0159] The current phototherapy module 22 is used to mark the light area of ​​the phototherapy device relative to the user in the light-interactive state, and determine the current phototherapy mode of the phototherapy device based on the light area and multiple working data of the phototherapy device.

[0160] The key illumination factor module 23 is used to determine the current sleep aid event of the phototherapy device based on the current phototherapy mode of the phototherapy device, multiple motion images of the user, and the user's previous sleep events; and to determine multiple key illumination factors of the phototherapy device based on the current sleep aid event, the user's facial expressions, and multiple environmental parameters.

[0161] Module 24, which is used to optimize content, determines the sleep-aiding effect coefficient of the phototherapy device based on multiple key light factors, their corresponding priorities, and the current circadian rhythm of the phototherapy device; and determines multiple contents to be optimized based on the user's sleep state, the sleep-aiding effect coefficient of the phototherapy device, and the corresponding multiple light data.

[0162] The intelligent control module 25 is used to collect the user's sleep posture change events, determine the intelligent control events of the phototherapy device based on multiple content to be optimized, the current phototherapy mode of the phototherapy device and the user's sleep posture change events, determine the corresponding intelligent light content based on the identification of the intelligent control events, and determine the rhythm control system of the phototherapy device according to the intelligent light content, the user's surrounding environmental events and the corresponding sleep duration.

[0163] This invention provides an electronic device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the intelligent control method for the phototherapy device described above.

[0164] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. An intelligent control device for a phototherapy equipment, characterized in that, include: The light-based interactive status module is used to determine the light status of the phototherapy device based on multiple light data from the phototherapy device and the corresponding usage scenario. The corresponding light-based interaction state is determined based on the light state of the phototherapy device and the interaction data of the phototherapy device relative to the user. The current phototherapy module is used to mark the light area of ​​the phototherapy device relative to the user in this light-based interactive state, and determine the current phototherapy mode of the phototherapy device based on the light area and multiple working data of the phototherapy device; The key light factor module is used to determine the current sleep-aiding event of the light therapy device based on the current light therapy mode of the light therapy device, multiple motion images of the user, and the user's previous sleep events. Based on the current sleep aid event, the user's facial expressions, and multiple environmental parameters, several key light factors for the phototherapy device were determined; The "Content to be Optimized" module is used to determine the sleep-aiding effect coefficient of the phototherapy device based on multiple key light factors, their corresponding priorities, and the current circadian rhythm of the phototherapy device; and to determine multiple contents to be optimized based on the user's sleep state, the sleep-aiding effect coefficient of the phototherapy device, and the corresponding multiple light data. The intelligent control module is used to collect the user's sleep posture change events, determine the intelligent control events of the phototherapy device based on multiple content to be optimized, the current phototherapy mode of the phototherapy device, and the user's sleep posture change events, determine the corresponding intelligent light content based on the identification of the intelligent control events, and determine the rhythm control system of the phototherapy device based on the intelligent light content, the user's surrounding environmental events, and the corresponding sleep duration.

2. The intelligent control device for the phototherapy equipment according to claim 1, characterized in that, The illumination status of the phototherapy device is determined based on multiple illumination data of the phototherapy device and the corresponding usage scenario; The corresponding light-based interaction state is determined based on the light illumination state of the phototherapy device and the interaction data between the phototherapy device and the user, including: Real-time monitoring of phototherapy equipment and collection of multiple light irradiation data points; determination of the usage scenario of the phototherapy equipment based on its current location, surrounding brightness, and the user's current state; determination of the light irradiation status of the phototherapy equipment based on multiple light irradiation data points, the corresponding usage scenario, and the lifespan of the phototherapy equipment. When the phototherapy device outputs phototherapy light to the user, multiple working data of the phototherapy device, the corresponding light data, and the changes in the surrounding brightness are collected to determine the interaction data between the phototherapy device and the user. Based on the light status of the phototherapy device, the interaction data between the phototherapy device and the user, and the continuous light exposure duration of the phototherapy device, the corresponding light-based interaction status is determined.

3. The intelligent control device for phototherapy equipment according to claim 1, characterized in that, In this light-based interactive state, the light-illuminating area of ​​the phototherapy device relative to the user is marked, and the current phototherapy mode of the phototherapy device is determined based on the light-illuminating area and multiple operating data of the phototherapy device, including: Real-time monitoring of the light irradiation of the phototherapy device on the user; determination of the corresponding light range based on the relative position of the phototherapy device to the user, the corresponding light irradiation posture, and the light irradiation data of the phototherapy device; determination of the light area of ​​the phototherapy device relative to the user based on the light range, the user's current posture, and overall shape. Multiple working data points of the phototherapy device are collected. Multiple phototherapy data combinations are determined based on the matching of the illumination area and the multiple working data points of the phototherapy device. Among the multiple phototherapy data combinations, the corresponding phototherapy features are determined based on the identification of each phototherapy data combination. The current phototherapy mode of the phototherapy device is determined based on the feature position of the multiple phototherapy features, the corresponding feature shape, and the overall shape of the phototherapy device.

4. The intelligent control device for the phototherapy equipment according to claim 1, characterized in that, The current sleep-aiding event of the phototherapy device is determined based on the current phototherapy mode of the phototherapy device, multiple motion images of the user, and the user's previous sleep events. Based on the current sleep-aiding event, the user's facial expressions, and multiple environmental parameters, several key light factors for the phototherapy device were determined, including: Multiple motion images of the user are collected, and the corresponding motion changes are determined based on the recognition of the multiple motion images. Based on the motion changes and the user's past sleep events, a first sleep aid content is determined. Based on the current light therapy mode of the light therapy device and the motion changes, a second sleep aid content is determined. Based on the first and second sleep aid content, the current sleep aid event of the light therapy device is determined.

5. The intelligent control device for the phototherapy equipment according to claim 4, characterized in that, The current sleep-aiding event of the phototherapy device is determined based on the current phototherapy mode of the phototherapy device, multiple motion images of the user, and the user's previous sleep events. Based on the current sleep-aiding event, the user's facial expressions, and multiple environmental parameters, several key illumination factors for the phototherapy device were determined, including: Collect the user's facial image, determine the user's facial expression based on the recognition of the user's facial image, and determine the first layer of light influence combination based on the current sleep aid event and the user's facial expression; The system identifies the environmental space created by the user and the phototherapy device, determines multiple environmental parameters based on the detection of this environmental space, determines a second light influence combination based on the current sleep aid event and the multiple environmental parameters, and determines multiple key light factors of the phototherapy device based on the first light influence combination and the second light influence combination.

6. The intelligent control device for the phototherapy equipment according to claim 1, characterized in that, The sleep-aiding effect coefficient of the phototherapy device is determined based on multiple key light factors, their corresponding priorities, and the current circadian rhythm of the phototherapy device. Based on the user's sleep status, the sleep-aiding effect coefficient of the light therapy device, and corresponding light illumination data, several areas for optimization were identified, including: Multiple sub-rhythm data of the phototherapy device are collected. The current rhythm content of the phototherapy device is determined based on the multiple sub-rhythm data and the corresponding light conditions. The multiple light influence content is determined based on the current rhythm content of the phototherapy device and multiple key light factors. The sleep-aiding effect coefficient of the phototherapy device is determined based on the priority of the multiple light influence content and multiple key light factors.

7. The intelligent control device for phototherapy equipment according to claim 6, characterized in that, The sleep-aiding effect coefficient of the phototherapy device is determined based on multiple key light factors, their corresponding priorities, and the current circadian rhythm of the phototherapy device. Based on the user's sleep status, the sleep-aiding effect coefficient of the light therapy device, and corresponding light intensity data, several areas for optimization were identified, including: Collect multiple sub-sleep state data of the user, determine the user's sleep state based on the identification of multiple sub-sleep state data, and determine the first level of optimization combination based on the user's sleep state and the sleep-aiding effect coefficient of the light therapy device; Based on the user's sleep state and multiple light data from the phototherapy device, a second set of optimization combinations is determined. Based on the first set of optimization combinations, the second set of optimization combinations, and the current sleep-aiding events of the phototherapy device, multiple optimization items are determined.

8. The intelligent control device for the phototherapy equipment according to claim 1, characterized in that, The process involves collecting user sleep posture change events, determining intelligent control events for the phototherapy device based on multiple optimization criteria, the current phototherapy mode of the device, and the user's sleep posture change events, identifying corresponding intelligent light content based on the recognition of these intelligent control events, and determining the circadian rhythm control system of the phototherapy device based on the intelligent light content, the user's surrounding environmental events, and the corresponding sleep duration. This system includes: The system monitors multiple sleep postures of the user at different times in real time, determines the user's sleep posture change events based on multiple sleep postures, and determines the first level of intelligent control content of the phototherapy device based on multiple content to be optimized and the user's sleep posture change events. The second level of intelligent control content of the phototherapy device is determined based on multiple content to be optimized and the current phototherapy mode of the phototherapy device. The intelligent control events of the phototherapy device are determined based on the first level of intelligent control content and the second level of intelligent control content. The system identifies intelligent control events and determines the corresponding intelligent lighting content. Simultaneously, it collects events from the user's surrounding environment. Based on the intelligent lighting content and the user's surrounding environment events, it determines the first rhythmic control content. Based on the intelligent lighting content and the user's sleep duration, it determines the second rhythmic control content. Based on the training of the first and second rhythmic control content, it determines the rhythmic control system of the phototherapy device.

9. An electronic device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, controls the phototherapy device according to any one of claims 1 to 8.

Images