LED lamp spectrum intelligent dynamic adjustment healthy illumination system based on human body rhythm

Abstract

The invention belongs to the technical field of healthy lighting, and particularly relates to a human body rhythm-based LED lamp spectrum intelligent dynamic adjustment healthy lighting system, which comprises a rhythm state sensing module, a rhythm phase estimation engine, a spectrum strategy generation unit, an LED driving execution module and a system coordination controller, multi-source physiological and environmental data are fused, continuous robust estimation of rhythm phase is realized by using an improved limit cycle oscillator model and extended Kalman filtering, a target spectrum is adaptively generated according to a phase-spectrum mapping table and confidence, and long-term stability of an output spectrum is ensured through multi-channel LED driving and closed-loop calibration. According to the invention, rhythm synchronous healthy lighting with high fidelity, low delay and privacy security can be realized.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of healthy lighting, in particular to a LED light spectrum intelligent dynamic adjustment healthy lighting system based on human rhythm. BACKGROUND

[0002] The concept of healthy lighting has been deeply developed, and the intelligent lighting system based on human circadian rhythm regulation of light environment has become an important research direction. Human biological clock is sensitive to light intensity, color temperature and spectrum, especially 460-490nm blue light can inhibit the secretion of melatonin, regulate alertness and sleep cycle. LED light source is widely used in healthy lighting scene due to its advantages of adjustable spectrum, high energy efficiency and fast response, etc. It promotes the evolution of people-oriented lighting paradigm from static illumination control to dynamic spectrum adaptation. The intelligent lighting system integrating physiological perception and environmental feedback is gradually productized, and the core goal is to actively support rhythm homeostasis by matching the user's physiological state and light environment parameters in real time.

[0003] The current mainstream rhythm lighting system adopts a preset time table or a simple environment sensor (illuminance meter, clock module) to drive the change of LED color temperature and brightness. Some high-end solutions introduce physiological signals such as heart rate variability to assist in judging the user's wakefulness level. Such systems follow a fixed strategy of high color temperature and high brightness during the day and low color temperature and low brightness at night, are designed based on the average rhythm model of the group, and rely on the empirical fitting of the typical circadian rhythm curve. In office, nursing home and other scenes, it can improve the subjective comfort of users or enhance daytime concentration, and has rationality and engineering feasibility for early healthy lighting practice.

[0004] In related technologies, the real-time requirement of dynamic adjustment and the fundamental mismatch of individual rhythm phase non-steady state characteristics, the endogenous period of human circadian rhythm is not strictly 24 hours, and is disturbed by multiple factors such as light history and social activities, showing significant individual differences and time-varying drift; while the existing system simplifies the rhythm state as a time function or a transient mapping of static physiological indicators, lacks the ability to continuously track and model the rhythm phase, and cannot distinguish between rhythm abnormalities and normal states, which is easy to exacerbate the biological clock disorder due to phase judgment errors. This is due to the lack of explicit modeling of rhythm dynamics in the system architecture, and the adjustment logic is open-loop static mapping rather than closed-loop state estimation driven; if the perception dimension is increased to improve accuracy, it will also cause data heterogeneity and a surge in computing load, which exceeds the carrying capacity of embedded platforms, and simply optimizing local parameters cannot bridge the semantic gap between adjustment actions and real rhythm requirements.

[0005] Therefore, the present application provides a LED light spectrum intelligent dynamic adjustment healthy lighting system based on human rhythm. SUMMARY

[0006] In order to make up for the deficiencies of the prior art and solve at least one technical problem proposed in the background art.

[0007] The technical scheme adopted by the present application to solve its technical problems is: the LED light spectrum intelligent dynamic adjustment health lighting system based on human rhythm comprises: A rhythm state perception module, comprising a non-invasive physiological signal acquisition subsystem and an environmental parameter monitoring subsystem, is used for synchronously acquiring individual physiological states and environmental illumination information; A rhythm phase estimation engine is disposed in a secure microcontroller with a trusted execution environment, receives a standardized feature vector output by the rhythm state perception module, continuously outputs a current rhythm phase angle and a confidence interval thereof based on an improved limit cycle oscillator model and in combination with an extended Kalman filter algorithm; A spectrum strategy generation unit calls a corresponding preset spectrum template according to a rhythm function interval into which the rhythm phase angle falls, and dynamically adjusts spectrum sharpness according to the confidence interval to generate a target spectrum power distribution curve; An LED drive execution module contains a current control channel, drives different peak wavelength narrowband LED chip groups, converts the target spectrum power distribution into channel drive current instructions through a reverse spectrum decomposition algorithm, and is configured with a feedback calibration unit to close-loop correct deviations between actual output spectrum and target spectrum; A system coordination controller is used for scheduling task timing of each module, and performing communication data integrity verification and firmware security monitoring; Wherein, a phase anomaly detection mechanism is arranged between the rhythm phase estimation engine and the spectrum strategy generation unit, when the rhythm phase angle change rate exceeds a physiological reasonable range or the confidence interval continuously expands to above a critical threshold, the spectrum strategy update is frozen and a calibration prompt is sent to the user terminal.

[0008] As a preferred mode of the present embodiment, the non-invasive physiological signal acquisition subsystem comprises an optical plethysmogram sensor, a skin conductivity detection electrode and an inertial measurement unit; The environmental parameter monitoring subsystem comprises a visible spectrum analyzer, an infrared pyroelectric sensor and an environmental illuminometer, wherein the visible spectrum analyzer adopts a combination structure of a diffraction grating and a linear photodiode array.

[0009] As a preferred mode of the present embodiment, the improved limit cycle oscillator model adopted by the rhythm dynamics modeling unit in the rhythm phase estimation engine introduces a light history weighting factor and a social activity disturbance term as external input variables, wherein the light history weighting factor is obtained by exponentially decaying and weightedly integrating the environmental blue light energy density, and the social activity disturbance term is generated by joint determination of body activity intensity and personnel presence state.

[0010] As a preferred mode of the embodiment, the rhythm-spectrum response mapping table followed by the spectrum strategy generation unit divides the complete rhythm cycle into wakefulness promotion zone, alertness maintenance zone, relaxation transition zone and sleep preparation zone; Each zone corresponds to a group of spectral templates with specific blue light peak intensity, color temperature gradient and red light compensation ratio; When the input phase angle falls into any functional zone, the corresponding template is called as the base spectrum, and the spectral sharpness adjustment factor is calculated according to the confidence interval width Δθ: Wherein is the reference confidence width, and the final target spectrum is the weighted sum of η times the base spectrum and times the smoothed spectrum.

[0011] As a preferred mode of the embodiment, the multi-channel constant current driving circuit in the LED driving execution module adopts a hybrid control strategy of digital pulse width modulation and analog current regulation: High-frequency digital pulse width modulation is enabled to realize fast spectrum switching in the rhythm phase rapid change stage, and analog current regulation is switched to eliminate human eye perceptible flicker in the phase stable stage.

[0012] As a preferred mode of the embodiment, the inverse spectrum decomposition algorithm in the LED driving execution module introduces an LED aging attenuation model to dynamically correct the spectrum response matrix when solving the driving current of each channel, and the model is expressed as: ; Wherein τⱼ is the measured attenuation time constant of the jth channel LED.

[0013] As a preferred mode of the embodiment, the feedback calibration unit collects the actual output spectrum every fixed working period, calculates the root mean square error between it and the target spectrum, and if the error exceeds the first preset tolerance, the fine tuning mechanism is started. Under the premise of keeping the total luminous flux unchanged, the driving current is iteratively optimized by gradient descent method until the error converges to below the second preset tolerance.

[0014] As a preferred mode of the embodiment, the photoplethysmogram sensor and the inertial measurement unit in the physiological signal acquisition subsystem are co-located in the same physical installation position, so that the acceleration signal output by the inertial measurement unit is directly used as the reference input of the adaptive noise cancellation algorithm for filtering out motion artifacts in the pulse wave signal.

[0015] As a preferred mode of the embodiment, the system coordination controller is interconnected with the building energy management system through a power line carrier communication mode, receives global work schedule instructions to fine-tune the social disturbance term in the rhythm model, and uploads the rhythm phase state after local encryption to the cloud health service platform, but the original physiological signal data is always retained in the device end trusted execution environment and is not transmitted externally.

[0016] At the method level: a human rhythm-based LED light spectrum intelligent dynamic adjustment health lighting method, comprising the following steps: S1. Synchronize the acquisition of multi-source physiological signals and environmental parameters, generate standardized feature vectors after time alignment and preprocessing; S2. Based on the improved limit cycle oscillator model and the extended Kalman filter algorithm, recursively estimate the current rhythm phase angle and its confidence interval; S3. Determine the rhythm function interval to which the phase angle belongs, call the corresponding spectrum template, and dynamically adjust the spectrum sharpness according to the confidence interval to generate the target spectrum power distribution; S4. Perform illumination normalization in combination with the environmental background illumination to obtain the final target spectrum; S5. Calculate the LED channel driving current through the reverse spectrum decomposition algorithm, and drive the multi-channel constant current circuit to output; S6. Regularly collect the actual output spectrum, and if the deviation from the target spectrum is out of limit, trigger the closed-loop fine-tuning mechanism; S7. Real-time monitor the stability of rhythm phase estimation, if abnormality is detected, freeze the spectrum update and guide the user to complete the calibration process.

[0017] The beneficial effects of the present application are as follows: The human rhythm-based LED light spectrum intelligent dynamic adjustment health lighting system according to the present application realizes continuous and robust estimation of individual circadian rhythm phase through the synergistic effect of rhythm dynamics model and extended Kalman filter, avoiding phase misjudgment caused by simplifying rhythm state as a time function or instantaneous physiological indicator; By using the phase-spectrum response mapping table and the confidence adaptive mechanism, the spectrum intervention intensity is dynamically adjusted while ensuring the real-time adjustment, preventing overstimulation caused by model uncertainty; The fusion processing of multi-source physiological and environmental data under a unified time reference is realized, solving the problem of heterogeneous signal space-time alignment; ensuring that the LED spectrum output remains highly consistent with the target distribution in long-term operation, offsetting the performance drift caused by device aging through a closed-loop feedback calibration mechanism; and through the dual-core heterogeneous architecture and secure communication protocol of the system coordination controller, a health lighting control platform with real-time, security and privacy protection capabilities is built under limited embedded resources. BRIEF DESCRIPTION OF DRAWINGS

[0018] The invention will now be further described with reference to the accompanying drawings.

[0019] Fig. 1 This is a block diagram of the overall structure of the intelligent dynamic adjustment health lighting system for LED light spectrum based on human circadian rhythm as described in this invention. Fig. 2 This is a schematic diagram of the data flow between the rhythm phase estimation engine and the spectral strategy generation unit in this invention. Detailed Implementation

[0020] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0021] like Figs. 1-2 As shown in the figure, an embodiment of the present invention provides an intelligent dynamic adjustment health lighting system for LED lamp spectrum based on human circadian rhythm, including a circadian rhythm state sensing module, a circadian rhythm phase estimation engine, a spectral strategy generation unit, an LED drive execution module, and a system coordination controller; the modules are connected through a deterministic communication bus to form a closed-loop control path from sensing to execution.

[0022] The rhythm state sensing module consists of a non-invasive physiological signal acquisition subsystem and an environmental parameter monitoring subsystem; The physiological signal acquisition subsystem includes a photoplethysmography (PPG) sensor, a skin conductivity detection electrode, and an inertial measurement unit, which are used to acquire heart rate variability characteristics, sympathetic nerve activation levels, and body activity status, respectively. All of the above sensors are integrated into the wearable device or the seat embedded interface, and their sampling clocks are uniformly timed by the system coordinator to ensure that the timestamp alignment accuracy of each physiological signal meets the minimum time resolution requirements for rhythm modeling. The environmental parameter monitoring subsystem includes a visible spectrum analyzer, an infrared pyroelectric sensor, and an ambient illuminance meter, used to capture real-time information on the existing light composition, personnel presence, and background brightness within the space. The visible spectrum analyzer employs a combination of a diffraction grating and a linear photodiode array, possessing independent energy density resolution capabilities for the 460-490 nm blue light band. All sensed data is converted from analog to digital and transmitted to the rhythmic phase estimation engine via an SPI bus.

[0023] The rhythm phase estimation engine is deployed in a secure microcontroller with a trusted execution environment. Its internal structure is divided into three logical functional areas: a data preprocessing unit, a rhythm dynamics modeling unit, and a phase state update unit. The data preprocessing unit receives the raw data stream from the rhythm state perception module, and sequentially performs baseline drift correction, motion artifact filtering, and time-frequency domain feature extraction operations, outputting a standardized physiological-environment feature vector sequence. The circadian dynamics modeling unit constructs an individualized circadian evolution equation based on a modified limit cycle oscillator model, which introduces a light history weighting factor and a social activity disturbance term as external input variables. The initial phase offset is automatically initialized by the user's first use of the sleep-wake habit questionnaire and continuously corrected in subsequent runs. The phase state updating unit uses an extended Kalman filter algorithm to recursively estimate the state variables of the circadian dynamics model by taking the preprocessed feature vector as the observation input, and outputs the current circadian phase angle and its confidence interval. The phase angle is expressed in radians, corresponding to a specific physiological stage in the endogenous melatonin secretion cycle.

[0024] The spectrum strategy generation unit receives the phase angle and its confidence interval output by the circadian phase estimation engine, and generates the target spectral power distribution curve in combination with the pre-set rhythm-spectrum response mapping table. The mapping table is constructed according to the standard definition of the Circadian Action Metric by the International Commission on Illumination, dividing the complete circadian cycle into four functional intervals: wake-promoting, alertness-maintaining, relaxation-transitioning, and sleep-preparing. Each interval is associated with a set of spectral templates with specific blue peak intensity, color temperature gradient, and red light compensation ratio. When the input phase angle falls into a certain functional interval, the spectrum strategy generation unit calls the corresponding template as the base spectrum and dynamically adjusts the spectral sharpness parameter according to the confidence interval width: The higher the confidence, the closer the spectrum to the ideal template; The lower the confidence, the higher the spectral smoothing factor to reduce the intervention intensity. In addition, the spectrum strategy generation unit also receives the background illuminance information provided by the environmental parameter monitoring subsystem, performs illuminance normalization processing to ensure that the absolute luminous flux of the final output meets the human eye visual comfort constraints.

[0025] The LED drive execution module consists of a multi-channel constant current drive circuit, a spectrum synthesis optical assembly, and a feedback calibration unit. The multi-channel constant current drive circuit contains at least four independent current control channels, corresponding to narrow-band LED chip groups with peak wavelengths at 450 nm, 470 nm, 530 nm, and 630 nm. The drive current values of each channel are calculated by the target spectral power distribution output by the spectrum strategy generation unit using an inverse spectral decomposition algorithm, which takes into account the aging attenuation coefficient and temperature drift characteristics of the LED chips to ensure spectral stability over a long period of operation. The spectrum synthesis optical assembly uses a light mixing cavity structure with a high reflectivity diffuse coating inside, allowing the monochromatic light emitted by each LED chip to be fully spatially mixed before exiting, forming a uniform and continuous target spectrum. The feedback calibration unit includes a miniature spectral sensor and closed-loop adjustment logic. It collects the actual output spectrum at fixed working intervals and compares it with the target spectrum. If the deviation exceeds the preset tolerance, it triggers the drive current fine-tuning mechanism until the error converges to the allowable range.

[0026] The system coordinator and controller adopt a dual-core heterogeneous architecture to achieve task isolation and real-time assurance; The main core runs a lightweight real-time operating system, which is responsible for managing the data acquisition scheduling of the rhythm state perception module, the calculation triggering timing of the rhythm phase estimation engine, and the status monitoring of the LED driver execution module. The collaborative core runs a security monitoring program to continuously verify the integrity and legality of communication data between modules, preventing rhythm misjudgments caused by man-in-the-middle attacks or firmware tampering. The system coordinator controller also has a built-in non-volatile memory area to save users' rhythm model parameters, historical phase trajectories, and personalized preference settings, which can be quickly restored to the most recent valid state after the device is powered off and restarted. All interactions between modules follow a time-triggered communication protocol to ensure that critical control commands are transmitted within a deterministic delay window.

[0027] A phase anomaly detection mechanism is set up between the rhythm phase estimation engine and the spectral strategy generation unit; When the rate of change of the rhythm phase angle exceeds the physiologically reasonable range within multiple consecutive estimation cycles, or when the confidence interval continues to expand above the critical threshold, the mechanism determines that the current rhythm model is in an unstable state, freezes the spectral strategy update, and sends a calibration prompt signal to the user terminal. In calibration mode, the system guides the user to complete a brief subjective state assessment questionnaire and reinitializes the rhythm dynamics model parameters using high-density physiological data from the following 24 hours.

[0028] The inertial measurement unit in the rhythm state sensing module shares the same physical installation location with the photoplethysmography pulse wave sensor, which allows body motion signals to be directly used for motion artifact compensation of pulse wave signals, improving the reliability of heart rate variability feature extraction. The infrared pyroelectric sensor and the visible spectrum analyzer in the environmental parameter monitoring subsystem share the same field of view optical window, ensuring spatial consistency between personnel presence judgment and light composition analysis.

[0029] The multi-channel constant current drive circuit in the LED driver execution module adopts a hybrid control strategy of digital pulse width modulation and analog current regulation: For high-frequency dynamic adjustment needs, digital pulse width modulation is preferred to achieve fast response; for low-frequency steady-state maintenance needs, analog current regulation is switched to eliminate flicker effect, balancing adjustment accuracy and visual comfort.

[0030] The system coordinating controller is interconnected with the building energy management system via power line carrier communication, receives global work schedule instructions (such as office area closing time), and uploads the local rhythm phase status to the cloud health service platform for group rhythm trend analysis. However, all information involving individual identity is encrypted locally before being uploaded, and the core calculations of the rhythm phase estimation engine are always completed in a trusted execution environment on the device, ensuring that users' physiological privacy is not leaked.

[0031] At the methodological level, the following steps are included: S1. Simultaneously acquire multi-source physiological signals and environmental parameters, and generate standardized feature vectors after time alignment and preprocessing; S2. Based on the improved limit cycle oscillator model and the extended Kalman filter algorithm, the current rhythm phase angle and its confidence interval are recursively estimated; S3. Determine the rhythmic functional interval to which the phase angle belongs, call the corresponding spectral template, and dynamically adjust the spectral sharpness according to the confidence interval to generate the target spectral power distribution; S4. Perform illuminance normalization based on ambient background illuminance to obtain the final target spectrum; S5. Calculate the driving current of each LED channel using the reverse spectral decomposition algorithm, and drive the multi-channel constant current circuit output; S6. Periodically collect the actual output spectrum. If the deviation from the target spectrum exceeds the limit, trigger the closed-loop fine-tuning mechanism. S7. Monitor the stability of rhythm phase estimation in real time. If an anomaly is detected, freeze the spectral update and guide the user to complete the calibration process.

[0032] At the system physical deployment level, the rhythm state sensing module is divided into a non-invasive physiological signal acquisition subsystem and an environmental parameter monitoring subsystem; The physiological signal acquisition subsystem is integrated into a wearable wristband or an embedded interface in an office chair armrest, and includes a photoplethysmography (PPG) sensor, a pair of skin conductivity (EDA) detection electrodes, and a six-axis inertial measurement unit (IMU). The PPG sensor uses a dual-wavelength LED light source (525nm green light and 660nm red light) in conjunction with a high-sensitivity photodiode to continuously acquire transmitted or reflected pulse wave signals at a sampling rate of 100Hz; the EDA electrode surface is covered with an Ag / AgCl coating, with a sampling rate of 32Hz, to monitor changes in sympathetic nerve-mediated sweat gland activity; the IMU includes a triaxial accelerometer and a triaxial gyroscope, with an output data rate set to 50Hz, to capture minute body movements and posture transitions; The analog-to-digital converters (ADCs) of the above three types of sensors are all timed uniformly by the system coordinator through the I²C bus to ensure that the timestamp alignment error of each signal does not exceed ±2ms, which meets the minimum time resolution requirements for subsequent rhythm modeling.

[0033] The environmental parameter monitoring subsystem is fixedly installed on the lighting fixture body or the near-field area of ​​the wall, and includes a visible spectrum analyzer, a passive pyroelectric infrared (PIR) sensor and a silicon photodiode illuminance meter; The visible spectrum analyzer employs a combination of a planar diffraction grating (1200 lines / mm) and a 32-channel linear photodiode array (50 μm pixel pitch), covering a spectral response range of 400-700 nm. Its energy density resolution in the 460-490 nm blue light band reaches ±0.5 μW / cm² / nm. The PIR sensor has a field of view set to 110° for detecting the presence of a human body, with a response time of less than 1 second. The illuminance meter is calibrated by V(λ), with a range of 0.1-100,000 lux and an accuracy of ±3%. All environmental sensor data is converted by a 16-bit Σ-Δ ADC and then uploaded to the rhythm phase estimation engine via the SPI bus at a frequency of 10Hz.

[0034] The rhythm phase estimation engine is deployed in a trusted execution environment (TEE) of a secure microcontroller (such as NXPi.MXRT1170) with ARM TrustZone technology. Its internal logic is divided into a data preprocessing unit, a rhythm dynamics modeling unit, and a phase state update unit. The data preprocessing unit first performs baseline drift correction on the raw PPG signal: A sliding window mid-range filter (window length = 5s) is used to eliminate low-frequency drift caused by breathing modulation. An adaptive noise cancellation (ANC) algorithm is used to filter out motion artifacts by using the acceleration signal output by the IMU as a reference input. The purified PPG signal is sent to the time-frequency analysis module to extract the time-domain indicators (such as SDNN, RMSSD) and frequency-domain indicators (LF / HF ratio) of heart rate variability (HRV) to form an 8-dimensional physiological feature vector. Meanwhile, after the EDA signal is filtered by a high-pass filter (cutoff frequency 0.05Hz) to remove the trend term, the moving average of its first-order difference absolute value (window = 30s) is calculated as an indicator of the sympathetic activation level. The environmental data directly extracts blue light energy density, background illuminance, and human presence markers to form a 5-dimensional environmental feature vector; The two types of feature vectors are concatenated into a 13-dimensional standardized input vector, which is then fed into the rhythm dynamics modeling unit.

[0035] The rhythm dynamics modeling unit constructs individualized rhythm evolution equations based on an improved Limit Cycle Oscillator (LCO). This model abstracts the endogenous circadian rhythm as a stable limit cycle in a two-dimensional phase plane, with its state variables represented by the phase angle θ(t) and amplitude r(t). The dynamic equations are as follows: Wherein, represents the individual's intrinsic angular frequency (unit: rad / h), and the initial value is estimated from the user's daily routine questionnaire filled out upon first use (for example, if the user usually goes to sleep at 23:00 and wakes up at 7:00, then their intrinsic cycle is estimated to be approximately 24.2h). ; The weighted integral term for historical illumination input is τ, where τ is the light input delay (typically 15 min), and the weighting function adopts an exponential decay form. ( The integration interval is the past 4 hours; The social activity disturbance term is determined jointly by the intensity of physical activity (from IMU) and the presence status of personnel (from PIR), and its value ranges from [0,1]. α and β are the light sensitivity coefficient and social perturbation gain, respectively, initially set to 0.02 rad / (h·klux) and 0.01 rad / h, and subsequently adjusted through an online learning mechanism; γ is the amplitude recovery rate constant, which is fixed at 1 / 2. This ensures the system has disturbance-resistant stability.

[0036] The phase state update unit employs the Extended Kalman Filter (EKF) algorithm, using the aforementioned 13-dimensional eigenvectors as observation input to recursively estimate θ(t). The observation model is established as follows: in, Let k be the observation vector at time k. The noise is zero-mean Gaussian white noise; the nonlinear observation function h(θ) is implemented by an offline trained multilayer perceptron (MLP), which takes the phase angle θ as input and outputs the predicted physiological response features such as HRV, EDA and blue light sensitivity. Both the state transition Jacobian matrix and the observation Jacobian matrix of the EKF are calculated online using linearization. The filter is updated every 5 minutes, outputting the current rhythm phase angle θ̂ (in rad) and its covariance P, the latter being used to calculate the confidence interval width. (Corresponding to a 95% confidence level).

[0037] The spectral strategy generation unit receives θ̂ and Δθ, and combines them with a preset rhythm-spectral response mapping table to generate the target spectral power distribution Starget(λ). This mapping table is constructed based on the definition of circadian stimulus (CS) in the CIES026:2018 standard, dividing the complete 2π-radian rhythmic cycle into four functional regions: Awakening promotion region: θ∈[0,π / 2), corresponding to the early morning period, the target spectrum includes high-intensity 470nm blue light (peak irradiance ≥8μW / cm² / nm), color temperature 6500K, and red light (630nm) accounting for 15%; Alertness maintenance zone: θ∈[π / 2,π), corresponding to the period from morning to afternoon, with a moderate decrease in blue light intensity (peak value 5μW / cm² / nm), a color temperature of 5500K, and red light accounting for 10%; Relaxed transition zone: θ∈[π,3π / 2), corresponding to evening, the blue light peak drops to 2μW / cm² / nm, the color temperature is 3000K, and the red light proportion increases to 25%; Sleep preparation zone: θ∈[3π / 2,2π), corresponding to nighttime, blue light is suppressed to below 0.5μW / cm² / nm, color temperature is 2200K, and red light accounts for 40%.

[0038] When θ̂ falls within a certain interval, the corresponding template is used as the base spectrum Sbase(λ). Then, a spectral sharpness adjustment factor η is introduced, calculated as follows: Where Δθref is the reference confidence width, set to 0.3 rad (approximately corresponding to a time uncertainty of 1.7 hours). The final target spectrum is: Ssmooth(λ) is a smooth spectrum across the entire wavelength range (color temperature 4000K, no significant peaks), used to reduce the intensity of intervention when model uncertainty is high; The spectral strategy generation unit reads the ambient illuminance Lenv (in lux) and performs illuminance normalization according to the following formula: Lset is the preset working surface illuminance (e.g., 500 lux) to ensure that the total luminous flux meets the requirements for visual comfort.

[0039] The LED driver execution module consists of a four-channel constant current drive circuit, a mixing cavity type spectral synthesis optical component, and a feedback calibration unit. Four channels drive InGaN / AlInGaP narrowband LED chipsets with peak wavelengths of 450nm (15nm half-width), 470nm (12nm), 530nm (20nm) and 630nm (25nm) respectively. The maximum output current of each channel is 350mA, and the current regulation resolution is 1mA. The inverse spectral decomposition algorithm projects Sfinal(λ) onto the four-dimensional LED substrate space to solve the following optimization problem: Where i = [i1, i2, i3, i4]ᵀ is the driving current vector of each channel; sfinal is the discrete vector (61 dimensions in total) of Sfinal(λ) sampled in 5nm steps within the range of 400-700nm. M is a 4×61-dimensional spectral response matrix, whose elements Mⱼ k This indicates that the j-th LED is at wavelength λ k The relative radiation intensity at a given location is obtained by factory calibration using an integrating sphere and is periodically updated based on an aging model. Mⱼ k (t)=Mⱼ k (0)·exp(−t / τⱼ); τⱼ is the decay time constant of each LED channel (measured values: τ=15,000h for 450nm channel, τ=18,000h for 470nm channel, τ=20,000h for 530nm channel, and τ=25,000h for 630nm channel); λreg is the Tikhonov regularization parameter, set to 0.01 to prevent overfitting of the current solution.

[0040] The mixing cavity adopts a cylindrical aluminum cavity (80mm inner diameter, 60mm depth), with a PTFE-based high-reflection diffuse coating (reflectivity >98%, Lambertian properties) sprayed on the inner wall. Four sets of LED chips are arranged in a ring symmetrical arrangement at the bottom of the cavity. After multiple reflections, the light forms a uniform mixed light spot at the light outlet, with a spectral spatial non-uniformity of <3%. The feedback calibration unit integrates a miniature CMOS spectral sensor (such as Hamamatsu C12880MA), which acquires the actual output spectrum Sactual(λ) every 30 minutes and calculates its root mean square error (RMSE) compared to Sfinal(λ). If RMSE > 0.8 μW / cm² / nm, then the fine-tuning mechanism is activated: while keeping the total luminous flux constant, a small perturbation Δi is applied to the current vector i, and the optimization is iteratively performed using the gradient descent method until RMSE ≤ 0.5 μW / cm² / nm.

[0041] The system coordinator uses a dual-core heterogeneous architecture. The main core (Cortex-M7, 480MHz) runs the FreeRTOS real-time operating system and is responsible for scheduling the following tasks: Physiological signal acquisition is triggered every 100ms, the rhythm phase estimation engine is started every 5 minutes, the LED driving status is checked every 30 seconds, and the non-volatile memory (FRAM, capacity 64KB) is written every 10 minutes to save the θ̂ trajectory, model parameters and user preferences. The co-core (Cortex-M4, 240MHz) runs an independent security monitoring program to perform HMAC-SHA256 integrity checks on data packets on the SPI / I²C bus and verify firmware signatures. Once an anomaly is detected, the LED driver enable signal is cut off. All inter-module communication follows the Time Triggered Protocol (TTP), and critical commands (such as phase updates and spectrum switching) are guaranteed to be transmitted end-to-end within 50ms.

[0042] Furthermore, the system has a built-in phase anomaly detection mechanism; If |dθ̂ / dt|>0.5rad / h (exceeding the normal rhythmic drift rate) for three consecutive estimation periods, or Δθ continuously>0.8rad for more than 1 hour, the model is determined to be unstable, the spectral strategy update is frozen, the current spectral output is maintained, and a calibration prompt is pushed to the user's mobile APP via Bluetooth Low Energy (BLE). In calibration mode, the system guides users to complete a subjective questionnaire that includes five Likert scale items such as current alertness and recent sleep quality. Physiological data is collected at twice the frequency over the next 24 hours to reinitialize the ω0, α, and β parameters of the LCO model.

[0043] The PPG sensor and IMU are co-located on the same flexible PCB substrate, with a center-to-center distance of less than 5 mm. This allows the triaxial acceleration a(t) output by the IMU to be directly used as the reference noise input for the PPG signal xppg(t). The ANC algorithm uses a normalized least mean square (NLMS) filter, and the update equation is: in, It is a 15-dimensional motion feature vector. The error signal is given, μ = 0.01 is the step size factor, and ε = 10⁻. 6 To prevent the removal of zero constants, this design improves the signal-to-noise ratio of HRV feature extraction by approximately 6 dB.

[0044] The LED driver circuit adopts a hybrid strategy of digital pulse width modulation (PWM) and analog current regulation. For dynamic spectral switching (such as abrupt transition from the alertness maintenance region to the relaxation transition region), the 20kHz PWM mode is preferentially enabled with a duty cycle resolution of 0.1%, achieving a response speed of <100ms. During the steady-state maintenance phase (phase change rate < 0.1 rad / h), it automatically switches to the analog current mode controlled by a 12-bit DAC to completely eliminate flicker perceptible to the human eye (fluctuation depth < 1%).

[0045] The system coordinating controller integrates a power line carrier (PLC) communication module (compliant with the G3-PLC standard), which can receive global work and rest instructions issued by the building energy management system (such as entering the off-get off work mode at 18:00) and adjust the social disturbance term Asocial(t) of the adjustment law model accordingly. Meanwhile, the local encryption module (AES-256) packages and encrypts θ̂, user ID hash value and timestamp, and uploads them to the cloud health service platform via Wi-Fi for population rhythm epidemiological research; however, all data involving raw physiological signals are kept in the device's TEE and are not transmitted externally.

[0046] In summary, by combining the synergistic effect of the circadian rhythm dynamics model and the extended Kalman filter, continuous and robust estimation of the phase of an individual's diurnal rhythm can be achieved, avoiding phase misjudgment caused by simplifying the rhythm state to a time function or instantaneous physiological index. By utilizing the phase-spectral response mapping table and confidence adaptive mechanism, the intensity of spectral intervention is dynamically adjusted while ensuring real-time regulation, thus preventing overstimulation caused by model uncertainty. It achieves the fusion processing of multi-source physiological and environmental data under a unified time reference, solving the problem of spatiotemporal alignment of heterogeneous signals; it ensures that the LED spectral output remains highly consistent with the target distribution during long-term operation, and offsets the performance drift caused by device aging through a closed-loop feedback calibration mechanism; and through the dual-core heterogeneous architecture and secure communication protocol of the system coordinating controller, it builds a health lighting control platform with real-time performance, security and privacy protection capabilities under limited embedded resources.

[0047] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A smart dynamic adjustment health lighting system for LED lamp spectrum based on human circadian rhythms, characterized in that, include: The rhythm state perception module includes a non-invasive physiological signal acquisition subsystem and an environmental parameter monitoring subsystem, which are used to simultaneously acquire individual physiological state and ambient light information; The rhythm phase estimation engine is deployed in a secure microcontroller with a trusted execution environment. It receives the standardized feature vector output by the rhythm state perception module and continuously outputs the current rhythm phase angle and its confidence interval based on an improved limit cycle oscillator model and combined with an extended Kalman filter algorithm. The spectral strategy generation unit calls the corresponding preset spectral template according to the rhythmic functional interval into which the rhythmic phase angle falls, and dynamically adjusts the spectral sharpness according to the confidence interval to generate the target spectral power distribution curve. The LED driver execution module includes a current control channel that drives narrowband LED chip groups with different peak wavelengths. It converts the target spectral power distribution into channel drive current commands through an inverse spectral decomposition algorithm and is equipped with a feedback calibration unit to correct the deviation between the actual output spectrum and the target spectrum in a closed loop. The system coordination controller is used to schedule the timing of tasks in each module, as well as to perform communication data integrity verification and firmware security monitoring. The rhythm phase estimation engine and the spectral strategy generation unit are equipped with a phase anomaly detection mechanism. When the rate of change of the rhythm phase angle exceeds the physiologically reasonable range or the confidence interval continues to expand to above the critical threshold, the spectral strategy update is frozen and a calibration prompt is sent to the user terminal.

2. The intelligent dynamic adjustment health lighting system for LED lamp spectrum based on human circadian rhythm according to claim 1, characterized in that, The non-invasive physiological signal acquisition subsystem includes a photoplethysmography (PPG) sensor, a skin conductivity detection electrode, and an inertial measurement unit. The environmental parameter monitoring subsystem includes a visible spectrum analyzer, an infrared pyroelectric sensor, and an ambient illuminance meter. The visible spectrum analyzer adopts a combination structure of a diffraction grating and a linear photodiode array.

3. The intelligent dynamic adjustment health lighting system for LED lamp spectrum based on human circadian rhythm according to claim 1, characterized in that, The improved limit cycle oscillator model used in the rhythm dynamics modeling unit of the rhythm phase estimation engine introduces an illumination history weighting factor and a social activity disturbance term as external input variables. The illumination history weighting factor is obtained by exponentially decaying weighted integral of the ambient blue light energy density, and the social activity disturbance term is generated by jointly determining the intensity of physical activity and the presence status of people.

4. The intelligent dynamic adjustment health lighting system for LED lamp spectrum based on human circadian rhythm according to claim 1, characterized in that, The spectral strategy generation unit uses a rhythm-spectral response mapping table to divide the complete rhythmic cycle into an arousal promotion zone, an alertness maintenance zone, a relaxation transition zone, and a sleep preparation zone. Each interval corresponds to a set of spectral templates with specific blue light peak intensity, color temperature gradient, and red light compensation ratio; When the input phase corner falls into any functional interval, the corresponding template is called as the base spectrum, and the spectral sharpness adjustment factor is calculated based on the confidence interval width Δθ. η=exp(−Δθ / ); in As a reference confidence width, the final target spectrum is a weighted sum of η times the base spectrum and (1−η) times the smoothed spectrum.

5. A smart dynamic adjustment health lighting system for LED light spectrum based on human circadian rhythms according to claim 1, characterized in that, The multi-channel constant current drive circuit in the LED driver execution module adopts a hybrid control strategy of digital pulse width modulation and analog current regulation: High-frequency digital pulse width modulation is enabled during the rapid phase change phase of the rhythm to achieve rapid spectral switching, and analog current regulation is switched to eliminate flicker perceptible to the human eye during the phase stabilization phase.

6. The intelligent dynamic adjustment health lighting system for LED lamp spectrum based on human circadian rhythm according to claim 1, characterized in that, The inverse spectral decomposition algorithm in the LED driver execution module introduces an LED aging decay model to dynamically correct the spectral response matrix when solving for the driving current of each channel. This model is expressed as follows: ； Where τⱼ is the measured decay time constant of the j-th LED channel.

7. A smart dynamic adjustment health lighting system for LED light spectrum based on human circadian rhythms according to claim 1, characterized in that, The feedback calibration unit collects the actual output spectrum at fixed working intervals and calculates the root mean square error between the actual output spectrum and the target spectrum. If the error exceeds the first preset tolerance, the fine-tuning mechanism is activated. The driving current is iteratively optimized by gradient descent while keeping the total light flux constant until the error converges to below the second preset tolerance.

8. A smart dynamic adjustment health lighting system for LED lamp spectrum based on human circadian rhythm according to claim 1, characterized in that, The photoplethysmography (PPG) sensor and the inertial measurement unit in the physiological signal acquisition subsystem are placed in the same physical installation location, so that the acceleration signal output by the inertial measurement unit can be directly used as the reference input of the adaptive noise cancellation algorithm to filter out motion artifacts in the pulse wave signal.

9. A smart dynamic adjustment health lighting system for LED lamp spectrum based on human circadian rhythm according to claim 1, characterized in that, The system coordination controller is interconnected with the building energy management system via power line carrier communication. It receives global work and rest schedule instructions to fine-tune the social disturbance terms in the rhythm model and uploads the locally encrypted rhythm phase state to the cloud health service platform. However, the original physiological signal data is always kept in the trusted execution environment on the device and is not transmitted to the outside.

10. A method for intelligent dynamic adjustment of LED lamp spectrum for healthy lighting based on human circadian rhythms, applicable to the intelligent dynamic adjustment of LED lamp spectrum for healthy lighting system based on human circadian rhythms as described in any one of claims 1-9, characterized in that, Includes the following steps: S1. Simultaneously acquire multi-source physiological signals and environmental parameters, and generate standardized feature vectors after time alignment and preprocessing; S2. Based on the improved limit cycle oscillator model and the extended Kalman filter algorithm, the current rhythm phase angle and its confidence interval are recursively estimated; S3. Determine the rhythmic functional interval to which the phase angle belongs, call the corresponding spectral template, and dynamically adjust the spectral sharpness according to the confidence interval to generate the target spectral power distribution; S4. Perform illuminance normalization based on ambient background illuminance to obtain the final target spectrum; S5. Calculate the driving current of each LED channel using the reverse spectral decomposition algorithm, and drive the multi-channel constant current circuit output; S6. Periodically collect the actual output spectrum. If the deviation from the target spectrum exceeds the limit, trigger the closed-loop fine-tuning mechanism. S7. Monitor the stability of rhythm phase estimation in real time. If an anomaly is detected, freeze the spectral update and guide the user to complete the calibration process.

Images