Eye fatigue real-time evaluation method when digital equipment is used

Abstract

The invention discloses a method for evaluating eye fatigue in real time when digital equipment is used. The method comprises the following steps: acquiring data in the process of using the digital equipment by a user in real time, and dividing eye using situations; calculating the key visual load indexes, and performing linear combination on the key visual load indexes by adopting different weights according to eye use scenes to obtain an eye fatigue comprehensive coefficient; dynamically calibrating the comprehensive eyestrain coefficient in combination with twelve-hour and five-element theories of traditional Chinese medicine; according to a set fatigue threshold value, formulating a reminding intervention strategy: according to the collected historical eye use data, adopting a multiple regression or machine learning model to adjust the weight of each key visual load index under different eye use situations of each user; and dynamically adjusting the eye fatigue comprehensive coefficient. According to the invention, through multi-dimensional data fusion, evaluation is more objective and accurate; the method has a machine learning adaptive capability, can continuously optimize model parameters according to historical data of each user, and realizes real personalized evaluation.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary technical field of digital health management and vision care, and relates to the field of eye health monitoring and protection technology when using digital devices, specifically to a method for real-time assessment of eye fatigue when using digital devices. Background Technology

[0002] With the widespread use of digital devices such as smartphones and tablets, the problem of digital eye strain (DES) caused by prolonged screen time has become increasingly prominent. Excessive screen time can lead to symptoms such as dry eyes, soreness, blurred vision, decreased concentration, and even headaches. Especially during the COVID-19 pandemic, increased online classes, remote work, and entertainment activities at home have significantly increased daily screen time across all age groups. Surveys indicate that 50-60% of children and adolescents reported experiencing eye discomfort during the pandemic, and a considerable proportion of adults also experienced varying degrees of digital eye strain.

[0003] The causes of digital eye strain are complex, involving a combination of factors that contribute to poor eye use. One primary factor is the mismatch between screen brightness and ambient light: prolonged viewing of an overly bright screen in a dim environment causes frequent pupil constriction and intense light stimulation to the eyes; conversely, staring at a dimly lit screen in bright sunlight requires squinting and strain to see clearly, exacerbating visual tension. Furthermore, glare reflections from digital device screens or flickering ambient light can also increase eye discomfort. Second, prolonged close-up viewing is a significant contributing factor: digital devices are typically held very close, and prolonged close-up work (more than 30 minutes) causes prolonged tension in the ciliary muscles, leading to eye strain and temporary vision loss. Third, when using digital devices, people often focus excessively on the screen, reducing blinking frequency. Normally, a person blinks 15-20 times per minute, but when staring at a screen, the blinking frequency may decrease by half or more, leading to excessive tear film evaporation and dry eye symptoms. Fourth, the spectral composition and dynamic content of the screen also have an impact. For example, high-intensity blue light at night can suppress melatonin secretion and disrupt sleep rhythms, thereby aggravating eye and body fatigue; rapidly flashing or highly dynamic content (such as certain short videos and games) keeps the eyes constantly focused on tracking, which can also easily lead to fatigue.

[0004] Currently, there are some smart devices or applications on the market that offer eye health features, but most only focus on reminding or controlling a single factor, failing to comprehensively reflect the actual eye strain. For example, many digital devices offer "eye protection mode" or night mode, reducing the blue light component of the screen and adjusting color temperature and brightness to reduce the impact of short-wavelength blue light on retinal cells and circadian rhythms. These modes alleviate discomfort from nighttime screen use to some extent, but only involve adjusting the spectrum and brightness, without considering differences in ambient brightness or specific user behavior. Similarly, many systems have "usage time limits" or "timed rest reminders," prompting users to take a break when continuous use reaches a preset threshold (e.g., 1 hour). This sedentary reminder approach can encourage users to put down their digital devices periodically, but it lacks assessment of the current environment and specific eye strain: in very dim or shaky environments, eyes may already be very tired even before the time threshold is reached; while in well-lit, comfortable conditions, prolonged use may not cause significant discomfort. Furthermore, some innovative applications attempt to use cameras to monitor users' blinking frequency or eye condition to provide eye protection reminders. For example, the Eyeblink software monitors blink intervals using the front-facing camera of a laptop or mobile phone, issuing a reminder when the user doesn't blink for an extended period. Apple also introduced the "Screen Distance" feature in iOS 17, using the front-facing depth sensor to detect the distance between the eyes and the screen, prompting a pop-up reminder to move the device further away when the user holds it too close for an extended period (e.g., less than 30 cm). These features intervene in blinking habits and viewing distance respectively, playing a positive role in preventing dry eyes and reducing the harm of close-range eye strain. However, overall, current solutions are often fragmented and lack a comprehensive assessment method that integrates multiple key factors affecting eye fatigue.

[0005] The shortcomings of existing technologies are mainly reflected in the following aspects: The evaluation indicators are too simplistic and lack a comprehensive index: Most applications only monitor a single parameter (such as cumulative usage time, screen brightness, or eye distance), failing to comprehensively consider multi-dimensional factors such as ambient light, screen dynamics, user interaction behavior, and individual tolerance differences, making it difficult to accurately reflect the actual degree of eye fatigue.

[0006] Lack of adaptability to different visual scenarios: Current methods rarely distinguish between different scenarios in which users are reading static text, watching videos, or playing games. The visual load patterns in these scenarios vary greatly, but existing systems usually use a one-size-fits-all fixed threshold and strategy, which is prone to misjudgment or missed judgment.

[0007] Ignoring special environmental conditions: In reality, users may use digital devices in various complex environments, such as screen shaking due to bumpy travel, flickering indoor lighting, and screen glare under strong outdoor light. Eye fatigue will be exacerbated in these situations, but currently commonly used eye protection functions do not provide additional protection or adjustment for these scenarios.

[0008] Lack of personalization: Different users have different sensitivities to eye strain; some people feel fine even after prolonged use of their eyes, while others are highly sensitive to changes in brightness. Currently, most systems lack learning mechanisms that utilize personal historical data, making it impossible to dynamically adjust the assessment model to match individual differences.

[0009] In summary, there is a need for a method to overcome the shortcomings of existing technologies, help users develop healthy eye habits, and protect eye health in the digital age. Traditional Chinese medicine believes that "the liver opens into the eyes," and excessive eye strain and staying up late damage the liver, leading to insufficient liver blood, which in turn causes eye fatigue and dry eyes. Traditional Chinese medicine emphasizes regulating the body according to its biological rhythms and advocates "prevention before disease." For example, reminder strategies can be adjusted based on differences in the body's energy and visual system state at different times. This invention proposes a method for real-time and comprehensive monitoring of eye fatigue levels, and incorporates traditional Chinese medicine prevention concepts to provide personalized intervention suggestions in the early stages. This method integrates data from multiple built-in sensors in digital devices and screen parameters, combined with ambient light information and the traditional Chinese medicine concept of "treating disease before it occurs," to quantitatively assess the user's eye load and fatigue status. Summary of the Invention

[0010] To address the aforementioned problems, this invention proposes a real-time method for assessing eye fatigue when using digital devices. This method integrates multi-sensor data and traditional Chinese medicine (TCM) rhythm theory to achieve precise quantification of eye load and personalized preventative tips. An eye fatigue assessment model is constructed by comprehensively utilizing multi-source data, and the TCM concept of "prevention before disease occurs" is incorporated to calibrate and regulate the results.

[0011] The present invention specifically includes the following steps: Step 1: Multi-sensor data acquisition and eye usage scenario recognition; It acquires real-time data on device motion, ambient light intensity, screen brightness, and user interaction during user interaction with digital devices; and classifies eye-use scenarios by combining the type of application running in the foreground, the changing patterns of screen content, and user interaction characteristics; at the same time, it uses the front-facing camera to capture the user's facial image and uses the front-facing camera or a distance sensor to estimate the distance between the eyes and the screen.

[0012] The device motion data includes linear acceleration data collected by the device's built-in accelerometer and angular velocity data collected by the device's built-in gyroscope; the ambient light intensity data includes ambient illuminance measured by the device's built-in ambient light sensor. Screen brightness parameters include the current device screen brightness setting and screen dimming mode; user interaction data includes the number of times the user interacts with the screen per unit time, the speed and frequency of scrolling operations, application switching, and reflects the user's pace and level of engagement when viewing information.

[0013] Eye-use scenarios include reading static text, browsing short videos, playing games, social communication, and video calls. If a scenario cannot be clearly categorized, it will be marked as general browsing / idle time.

[0014] Step 2: Calculation of key visual load indicators; The relative motion instability of the screen is calculated based on the relative angular velocity and relative acceleration between the screen and the user's head, as well as the rate of content change. Calculate the brightness adaptation error based on screen brightness and ambient brightness. Calculate user interaction intensity based on user interaction data The continuous eye use duration factor is calculated using a piecewise function. ; Step 3: Comprehensive coefficient of eye fatigue calculate: Based on the eye use scenario, the key visual load indicators calculated in step two are linearly combined using different weights to obtain the comprehensive coefficient of eye fatigue. Step 4: Dynamic calibration of the comprehensive eye fatigue coefficient: The comprehensive coefficient of eye fatigue calculated in step three is dynamically calibrated by combining the twelve two-hour periods of the day and the Five Elements theory of Traditional Chinese Medicine. Different time-based correction factors are assigned to different time periods based on the corresponding meridian and Five Elements attributes of the 24 hours of the day. This is to reflect the influence of human physiological rhythms on eye tolerance.

[0015] Step 5: Set fatigue thresholds and develop reminder and intervention strategies based on the Twelve Time Periods and Five Elements theory of Traditional Chinese Medicine. Set a warning threshold and a high fatigue threshold for the comprehensive eye fatigue coefficient. When the comprehensive eye fatigue coefficient, after calibration in step four, rises continuously and exceeds the warning threshold for the first time, an eye care reminder is triggered. If the comprehensive eye fatigue coefficient continues to climb and exceeds the high fatigue threshold, the intervention intensity is increased.

[0016] When the ambient light is severely insufficient and the screen is too bright, when the user stares at the screen for a long time without blinking, or when the user holds the digital device too close to the screen, the system will issue a reminder and provide corresponding suggestions even if the overall eye fatigue coefficient has not exceeded the threshold.

[0017] Step Six: Based on the collected historical eye use data, use multiple regression or machine learning models to adjust the weights of key visual load indicators for each user under different eye use scenarios; dynamically adjust the comprehensive eye fatigue coefficient according to the eye use habits of different users to achieve personalized eye fatigue assessment.

[0018] Through the synergistic effect of steps one through six above, real-time assessment of eye fatigue when using digital devices is achieved.

[0019] Furthermore, high-frequency noise in the data collected by each sensor in step one is removed using a filtering algorithm.

[0020] Furthermore, in step one, the front-facing camera of the digital device is used to periodically sample the environmental scene and correct the ambient light sensor readings to obtain a more accurate ambient brightness; the screen uses PWM dimming mode.

[0021] Furthermore, in step one, when user authorization is obtained and the application scenario requires it, the front-facing camera is periodically activated to briefly capture user facial images. Locally running visual algorithms detect facial key points, extract blink frequency, average eye-opening degree, pupil diameter changes, and estimate or measure head posture and the approximate distance between the eyes and the screen.

[0022] Furthermore, in step two: screen relative motion instability This measure assesses the stability of a digital device's screen image within the user's field of vision. When the device moves violently or the environment shakes, the eyes need to frequently adjust to track the screen, leading to increased eye strain. Brightness adaptation error. Used to assess the match between screen brightness and ambient brightness. Screen brightness that is too bright or too dim relative to ambient brightness increases eye strain. User interaction intensity. This factor is used to characterize the eye strain caused by user actions. Frequent and rapid interactions mean that users are browsing a large amount of information in a short period of time, requiring the eyes to move quickly and focus on different areas, which easily leads to visual fatigue. (Continuous eye use duration factor) It is used to reflect the cumulative degree of eye fatigue caused by continuous use of the eyes.

[0023] Furthermore, in step three, the comprehensive coefficient of eye fatigue is normalized or mapped to obtain an intuitive and unified fatigue index value.

[0024] Furthermore, the time correction factor in step four... The value ranges from [-0.1, +0.1]; the final eye fatigue index is obtained through the effect of the time correction factor. .

[0025] Its value range [-0.1, +0.1] comes from the following experiments: The statistical distribution of blink frequency, interocular distance, and pupillary response at different time periods; the sensitivity assessment of the parameter adjustment experiment (±5%–±20%); and finally, ±10% was selected as the correction range that balances stability and effectiveness.

[0026] Furthermore, in scenarios where flickering exists (low-frequency flickering in ambient lighting or screen dimming), although Although the threshold was not exceeded, the flicker penalty will increase the overall visual load assessment; flicker penalty, distance penalty, and dry eye factor are introduced when calculating the comprehensive eye fatigue coefficient in step four to increase the contribution of brightness discomfort, usage distance, and dry eye symptoms to fatigue. The flicker is detected by high-speed sampling of ambient light or by observing the flicker frequency of the light source using a camera.

[0027] Furthermore, in step five, the warning threshold and the threshold exceeding the high fatigue level are determined through ROC curve analysis of labeled data, so as to minimize false alarms while ensuring timely reminders.

[0028] Furthermore, during model learning, sensitive raw physiological data is processed only locally on the user's device, while the data uploaded to the server is anonymized and aggregated statistical information. Users can also choose to opt out of data sharing and use only the local model, ensuring that user privacy is not violated while improving model intelligence.

[0029] Compared with the prior art, the present invention has the following beneficial effects: Multi-dimensional data fusion for more objective and accurate assessment: This invention simultaneously collects and analyzes multi-source data such as ambient light, screen brightness, device and user movement, interaction behavior, and screen time, minimizing biases caused by the limitations of single indicators. In particular, it innovatively introduces indicators such as screen relative motion and brightness adaptation error, making the assessment results closer to real-world visual load conditions. For example, in environments with bumpy rides or drastic changes in lighting, traditional assessments based solely on duration or brightness may underestimate environmental factors, while the comprehensive index of this invention can keenly capture these additional loads, thus providing greater accuracy and reliability.

[0030] Scene-adaptive and comprehensive in its coverage: This invention dynamically adjusts model parameters based on the user's visual environment, differentiating between different usage scenarios such as reading, watching videos, and playing games, thus avoiding the inaccuracies of a one-size-fits-all assessment. Simultaneously, the algorithm considers the impact of various special environmental factors (such as bumps, flicker, and strong glare) on eye strain and incorporates targeted processing into the model. Therefore, whether the user is reading in a quiet indoor environment, watching videos in a moving vehicle, or using digital devices in poorly lit places, the system can provide realistic assessments and suggestions, making it far more applicable than existing single-function systems.

[0031] Incorporating Traditional Chinese Medicine (TCM) concepts and emphasizing prevention: This invention integrates the TCM concept of "treating disease before it occurs" into digital health management, using circadian rhythm adjustments to make the assessment more aligned with human physiology. For example, it increases the weight of fatigue during midnight hours to remind users not to overuse their eyes against their biological clock. This combination of modern technology and traditional medicine, validated through data and small-scale application, ensures both scientific rigor and practicality.

[0032] Personalized Dynamic Optimization: This invention possesses machine learning adaptive capabilities, continuously optimizing model parameters based on each user's historical data to achieve truly personalized assessments. Over long-term use, the system adjusts to users' sensitivities (e.g., increasing the weight of brightness indicators for those sensitive to brightness, and reducing the impact of duration for those with lower tolerance), and fine-tunes the reminder threshold to suit individual habits. Compared to a generic model with fixed parameters, this personalized optimization significantly improves assessment accuracy and intervention effectiveness, avoiding situations where some find the reminders too harsh while others find them insufficient.

[0033] Low cost and high compatibility: The data required by this invention comes from conventional sensors (accelerometers, gyroscopes, light sensors, front cameras, etc.) and software interfaces of smart digital devices, requiring no additional hardware. Therefore, it is easy to deploy on various brands and models of mobile phones and tablets. For devices lacking certain sensing capabilities (such as those without distance detection cameras), the model can automatically downgrade to using available data sources, maintaining basic functionality. This means that this method has good industrialization prospects and can be integrated as a general eye protection function into operating systems or health applications, providing practical and effective eye health protection for a wide range of users in the digital age. Attached Figure Description

[0034] To more clearly illustrate the method flow and key modules of the present invention, the accompanying drawings show schematic diagrams of several embodiments of the present invention, wherein: Figure 1 This is a schematic diagram of the module architecture and a system flowchart used in the method described in this invention; Figure 2 This is a schematic diagram illustrating the calculation of the screen's relative motion instability SI; Figure 3 It relates to the relationship between screen brightness and ambient light, as well as brightness adaptation error. The diagram shows the ambient light intensity on the horizontal axis. (Unit: lux), vertical axis represents screen brightness. (Unit: cd / m²).

[0035] Figure 4 The fatigue saturation curve represents the duration of continuous eye use, with rapid accumulation in the initial stage and gradual saturation in the middle and later stages. Figure 5 The curve showing the recovery of eye fatigue index after cessation of eye use; Figure 6 This is a schematic diagram illustrating the performance response under conditions of bumpy environments and light flickering. Figure 7 A chart showing the fatigue index of meridians according to the time of day in Traditional Chinese Medicine; Figure 8 A schematic diagram of the ROC curve for detecting fatigue state using the EFI index; Figure 9 This is a bar chart showing the distribution of personalized fatigue thresholds for different users in the example. Detailed Implementation

[0036] To better understand the working principle and beneficial effects of the method of this invention, the evaluation process and results are illustrated below through specific scenario examples. It should be noted that these examples are primarily based on theoretical models and simulation parameters, aiming to illustrate the logic of the method's application; they can be calibrated and optimized in actual products based on user data.

[0037] like Figure 1 As shown, the method of the present invention is implemented based on a data acquisition module, a situation recognition module, a sub-index calculation module, and a fatigue index extreme module.

[0038] The data acquisition module is equipped with built-in sensors and cameras in digital devices to acquire device motion data, ambient light intensity data, screen brightness parameters, and user interaction data in real time during the user's use of the digital device. At the same time, it uses the front-facing camera to capture the user's facial image and uses the front-facing camera or distance sensor to estimate the distance between the eyes and the screen, and acquires the user's facial features (such as blink frequency, head posture, eye-screen distance, etc.).

[0039] The context recognition module is used to classify eye-use contexts by combining the type of application running in the foreground, the changing patterns of screen content, and user interaction characteristics. The sub-index calculation module is used to calculate screen relative motion instability (SI), brightness difference, interaction intensity, and duration load. Screen relative motion instability: By integrating the digital device's own sensors and the user's head movement information, the actual motion amplitude of the screen relative to the eyes is calculated, which is used to measure the extra burden that the eyes have to put on tracking the screen in bumpy or shaking environments.

[0040] Brightness difference: Based on the relationship between ambient illuminance and screen light emission characteristics, this defines the degree of mismatch between screen brightness and ambient light, quantifies the impact of "brightness-darkness contrast" on eye strain, and encourages maintaining harmony between screen brightness and the surrounding environment.

[0041] User interaction intensity: This analyzes the frequency and rhythm of touchscreen clicks, scrolling, and other operations, combined with the rate of change of screen content, to assess the level of eye strain on users. Frequent and rapid operations will increase the interaction intensity index, indicating that users are using their eyes intensively.

[0042] Continuous eye use duration load: The fatigue accumulation caused by prolonged continuous eye use is modeled by nonlinear function, and the recovery effect of rest time is superimposed to make the contribution of time factor to fatigue more in line with the actual physiological law (that is, fatigue tends to saturate after prolonged eye use, and short rest can partially relieve it).

[0043] The fatigue index calculation module integrates the calculation results from the sub-index calculation modules and, in conjunction with the biological rhythms of the twelve two-hour periods in traditional Chinese medicine and the principles of mutual generation and restraint of the five elements, makes minor corrections to the fatigue index, generating a comprehensive eye fatigue index that reflects the level of eye load in real time.

[0044] Based on the above modules, a method for real-time assessment of eye fatigue when using digital devices is proposed, which specifically includes the following steps: Step 1: Multi-sensor data acquisition and eye use scenario recognition.

[0045] During the user's use of digital devices, multiple sensor data are acquired in real time and the current eye-use scenario is identified: Device motion data: The linear acceleration and angular velocity data of the device are collected at a certain frequency (e.g., 50 Hz, which can be down-processed to 10 Hz internally) using the accelerometer and gyroscope built into the digital device, and high-frequency noise is removed by filtering algorithm.

[0046] Ambient light intensity data: Measured by an ambient light sensor (ALS) to measure the surrounding ambient illuminance. (Unit: lux). When necessary, the front-facing camera of a digital device can be used to sample the environmental scene (e.g., briefly capturing the environment every few minutes) to estimate factors such as light source direction and reflection, and to correct sensor readings to obtain a more accurate and effective ambient brightness. .

[0047] Screen brightness parameter: Reads the current brightness setting value of the digital device screen. (Typically a normalized value or backlight level between 0 and 1). Based on a pre-defined screen brightness calibration curve, Converted to actual screen brightness (Unit: cd / m²). Simultaneously, acquire the screen's dimming mode (e.g., whether PWM dimming is used and its frequency) to assess the impact of screen flicker on visual perception.

[0048] User interaction data: Monitors the number of user interactions with the screen per unit of time, including touch screen clicks. Number of scrolling / sliding times Record the speed and frequency of scrolling operations, application switching, etc. This data reflects the user's pace and level of engagement when viewing information.

[0049] User facial feature data (optional): When authorized by the user and required by the application scenario, periodically activate the front-facing camera to briefly capture user facial images (e.g., capture 10 seconds of video every 5 minutes). Detect facial key points using locally running visual algorithms, extract physiological parameters such as blink frequency, average eye opening degree, and pupil diameter changes, and estimate head posture (e.g., tilting or tilting angle) and the approximate distance between the eyes and the screen. If the device has infrared camera / depth sensing capabilities, it can also measure directly. To protect user privacy, the raw images collected are used only for real-time analysis on the client side and will not be stored or uploaded. Only statistical features (such as blink rate per minute, average pupil size, etc.) will be uploaded.

[0050] Based on the above data, combined with the types of applications running in the foreground (such as e-book readers, video players, games, chat applications, etc.), the changing patterns of screen content, and user interaction characteristics, the usage scenarios are divided into several categories: such as reading static text. Browsing short videos Playing games Social communication Video calls If it cannot be clearly categorized, mark it as general browsing / idle. Context recognition helps to apply differentiated model parameters to different use cases in subsequent evaluations, thereby improving the accuracy of the evaluation.

[0051] Step 2: Calculation of key visual load indicators.

[0052] Using the collected data, several intermediate indicators that can quantify visual load are calculated: Screen relative motion instability This method measures the stability of a digital device screen within the user's field of vision. When a digital device moves violently or the environment is turbulent, the eyes need to frequently adjust to track the screen, leading to increased eye strain. This method differs from traditional methods that only consider the device's own motion; it estimates the actual degree of screen motion relative to the eyes by fusing motion information from both the digital device and the user's head. Let the angular velocity of the digital device in the three axes be... The linear acceleration is Simultaneously, the angular velocity of the user's head is inferred through facial pose estimation using the front-facing camera or other sensors. and linear acceleration The motion of the screen relative to the head can then be represented as: , .

[0053] The root mean square (RMS) values ​​of the relative angular velocity and relative acceleration are calculated within a short time window (e.g., 1–3 seconds) to obtain the amplitude of the digital device's motion relative to the head. Simultaneously, the motion intensity of the screen content itself is acquired. For example, in video and animation scenes, the rate of change of content can be obtained through image frame differencing or interfaces provided by the operating system. Taking these three factors into account, the image instability index is defined as follows: ; in , , These are weighting coefficients, determined through experimental calibration (initial values) or model training (dynamically adjusted values).

[0054] like Figure 2 As shown in the figure, this plot displays the gyroscope angular velocity signals under different conditions. and head angular velocity estimated via the front camera The curve of change over time, and the difference between the two. The instantaneous value. The graph also overlays the amplitude of the screen content's motion. The changes are as follows. It can be seen that when the phone is shaken violently while the head remains relatively stable, Significantly increased; at the same time, if the screen displays rapidly scrolling / changing content, The combined effect of these three factors leads to an increase in the SI index. A higher SI value indicates more intense screen movement relative to the eyes, a more unstable image, and requires the eyes to constantly refocus and track the screen, increasing eye strain. Conversely, if both the user and the device are stationary and the screen content is static, the SI is close to 0. For example, SI ≈ 0 when reading while seated, but SI increases significantly when watching videos on digital devices while walking or riding in a bumpy vehicle.

[0055] Brightness adaptation error This is used to evaluate the matching degree between screen brightness and ambient light. A screen that is too bright or too dark increases the burden on the eyes' accommodation relative to the environment. For comfortable viewing, the screen brightness and ambient light should generally be in a certain proportion. By measuring the relationship between screen brightness and ambient light, the comfortable brightness ratio of a device in different environments can be calibrated. r (i.e., screen brightness) Effective brightness of the environment The ratio, for example, in low indoor light r It's probably around 0.3, but under strong outdoor sunlight, the screen reflection can affect the result. r (Smaller). In this embodiment, the brightness adaptation error is defined as: ; in This represents the effective ambient brightness (in lux) after fusion correction between the camera and the light sensor, but has been converted to equivalent brightness for consistency with... Compare), To prevent the denominator from being a tiny constant, this ratio reflects the degree to which the current screen brightness deviates from the ideal, comfortable brightness. A larger screen brightness indicates a mismatch between the screen's brightness and the environment, requiring the eyes to frequently adjust to changes in light and darkness, leading to eye strain; conversely, a smaller screen brightness indicates a smaller screen brightness. A smaller value (closer to 0) indicates a better match between the screen and ambient brightness, resulting in higher visual comfort. For example, using a high-brightness screen in a dim environment or a low-brightness screen in bright light will both cause discomfort. Increase. During subsequent intervention, it can be based on... Adjust screen brightness: Reduce screen brightness in dark environments and increase brightness in bright environments to reduce the contrast between light and dark.

[0056] like Figure 3 As shown in the diagram, the shaded area represents the recommended comfortable brightness range, and the curve represents the ideal relationship between screen brightness and ambient light. If the actual screen brightness deviates from this ideal curve, the brightness error is indicated by an arrow. Size: The longer the arrow, the larger the value. The larger the screen brightness, the more pronounced the visual discomfort. This diagram illustrates how to select the appropriate screen brightness to minimize eye strain in different lighting conditions.

[0057] User interaction intensity This is used to characterize the eye strain caused by user actions. Frequent and rapid interactions mean that users browse a large amount of information in a short period of time, requiring the eyes to move quickly and focus on different areas, which easily leads to visual fatigue. This invention integrates multiple behavioral characteristics to calculate... For example: touch screen click frequency per unit time (times / minute), scrolling speed (Speed ​​of screen content movement, e.g., pixels per second), degree of change in interaction intervals (The variance or kurtosis of the touch operation time interval indicates whether the operation rhythm is uneven,) and the frequency of gaze switching. (This can be approximated by switching interface elements or scrolling through paragraphs). The above features are then linearly combined according to certain weights: ; in – Weighting coefficients are set based on experience or optimized through learning. A higher number indicates more frequent user activity and greater eye strain; conversely, a lower number indicates prolonged static viewing of the same content with almost no user interaction. Approximately 0. It should be noted that the impact of interaction intensity on fatigue varies depending on the content context; therefore, this invention dynamically adjusts the intensity based on the context. The weighting of the calculation. For example, in the scenario of browsing short videos, user interaction is less, but visual stimulation mainly comes from content switching. In this case, the weight of clicks can be reduced and the impact of content switching on fatigue can be increased. In the scenario of reading, user interaction frequency is low, but long-term staring at monotonous content can also cause fatigue. The weight of the time factor can be appropriately increased in this scenario.

[0058] Continuous eye use duration factor This is used to reflect the cumulative degree of eye fatigue caused by continuous close-range eye use. Prolonged close-range eye use gradually increases eye fatigue, but the accumulation of fatigue is not linear and endless; rather, it tends to saturate after a certain period. This invention models this phenomenon using a nonlinear model and incorporates the effect of rest and recovery. Let the duration of continuous use of digital devices be denoted as . t minutes. When t is small, its contribution to fatigue is relatively small; as t increases... t As the body grows, fatigue accumulates more rapidly, but after a certain period of time (such as several hours), it approaches an upper limit and will not grow indefinitely.

[0059] like Figure 4 As shown, the curve illustrates the increasing trend of fatigue load with continuous eye use time t. It can be seen that the curve is steep in the initial stage (rapid accumulation of fatigue), and then gradually flattens out (approaching saturation). Figure 5 As shown, the effect of rest on fatigue levels is illustrated: the curves show the decrease in the fatigue index after stopping eye use at a certain fatigue level; as the rest time increases, the index gradually decreases to a lower level. This demonstrates the working principle of the saturation function and the continuous eye use duration factor in the duration model of this invention.

[0060] We choose an empirical piecewise function f(t) to describe this saturation tendency: for example, in t The increase in fatigue slowed down after 20, 40, and 60 minutes, gradually approaching its maximum after 60 minutes. Furthermore, it's assumed that users took short breaks during prolonged screen time (such as locking the screen for a short rest); in these cases, partial recovery from fatigue needs to be considered. The time interval between the most recent break is defined as... (minutes), then a recovery factor can be set: ; in This is the fatigue recovery time constant, set according to research (e.g., 10–20 minutes, meaning that resting for ten minutes or so can significantly restore some eye condition). The continuous eye use duration factor is defined as: ; When users do not rest ,but ,at this time Mainly composed of continuous duration The decision is made and gradually saturates; if the user has just rested for a period of time, then... This reduces the accumulated fatigue and decreases the effective continuous visual load. This saturation + recovery model more accurately reflects the cumulative effect of prolonged eye use and the relieving effect of intermittent rest, avoiding overestimation or underestimation caused by simple linear extrapolation.

[0061] Step 3: Calculate the comprehensive eye fatigue coefficient. The sub-indicators calculated in Step 2 are linearly combined according to appropriate weights. Additionally, when significant flickering is detected in the ambient light, it can cause further eye strain. Flickering can be detected by high-speed sampling of ambient light or by observing the flicker frequency of the light source using a camera. If low-frequency flickering exists in the ambient lighting or screen dimming (e.g., some LED lights or screen PWM frequencies are low), [flickering is introduced]. The penalty factor increases the contribution of brightness discomfort to fatigue. In scenarios where flicker is present, although... It may not be significant, but the flicker penalty will improve the overall visual load assessment to alert users to lighting quality issues.

[0062] like Figure 6 As shown, the left figure is the acceleration spectrum of the device in a bumpy vehicle environment. The peak appears at approximately 5 Hz (vehicle vibration frequency), corresponding to an increase in the SI index at this frequency band. The spectral peak and the corresponding SI increase are highlighted in the figure. The right figure is a schematic diagram of detecting ambient light flicker using a camera: alternating bright and dark stripes appear in the image, indicating that the light has a flickering effect (e.g., flickering at a power frequency of 100 Hz). In this embodiment, this phenomenon is detected and triggered. It adjusts the fatigue index and generates reminders and suggestions.

[0063] Other supplementary indicators: The following supplementary factors are also introduced in this embodiment to cover more details affecting eye strain: Distance penalty factor: The distance d between the user's eyes and the screen is estimated using a camera or dedicated sensor. If the system detects that the user is consistently holding the digital device too close (e.g., d < 30 cm), a distance penalty is added. The optimal viewing distance for the human eye is typically 30–50 cm. Viewing too closely increases the strain on the eye's accommodation significantly. Quantifying the distance between the human eye and the device... .when d When the distance drops below the threshold (close to the country), The screen moves rapidly to remind the user to move the device further away. Apple's iOS "screen distance" reminder feature uses a similar principle.

[0064] Blinking and Dry Eye Factors: Analysis of forward-facing images reveals that if a user's blinking frequency is significantly lower than normal (e.g., <5 times / minute) and persists for a period of time, or if the average eye-opening time is too long, leading to prolonged blink intervals, it indicates that the tear film may be thinning, causing dry eye discomfort. The user's blinking frequency or average eye-opening time is quantified as... Index (dry eye factor), when When the fatigue index rises (due to increased eye usage frequency or prolonged eye-opening time), a prompt can be triggered in advance (e.g., prompting the user to consciously blink more or use eye drops). This factor is particularly important for users who frequently wear contact lenses or are prone to dry eyes; the system can appropriately increase this factor in personalized settings for such users. Weights.

[0065] Obtain the comprehensive coefficient of eye fatigue without TCM correction. :

[0066] in, , , , These are a set of weighting coefficients selected for the current content context. These weights may have different emphases in different scenarios (for example, in a reading scenario, more emphasis is placed on duration). Its role is to emphasize the dynamics of the visuals in game scenarios. (e.g., its function). η and η are fixed coefficients for flicker penalty and distance penalty, This is a fixed coefficient for the dry eye factor.

[0067] To obtain a clear and consistent fatigue index value, one can... Perform normalization or mapping. For example, determine a reasonable range of values ​​based on historical data, and then... The score is compressed to a range of 0–100 using linear transformation or Logistic mapping, where 0 represents almost no fatigue and 100 represents extreme fatigue. This processing facilitates setting a uniform alert threshold and provides a clear visual representation to the user. When no special distinction is needed, this normalized value will be referred to as [the standard value]. (Physical model fatigue index).

[0068] Step 4: Traditional Chinese Medicine Rhythm Correction and Determination of the Final Fatigue Index (EFI).

[0069] Calculated from the physical model Based on this, the present invention combines the twelve-hour system of traditional Chinese medicine and the Five Elements theory to dynamically calibrate the fatigue index. A time-based correction factor is set. The value range is [-0.1, +0.1] (meaning it can fluctuate by a maximum of 10%). For example... Figure 7 As shown, based on the time of day and the corresponding meridian and five-element attributes, different time periods are assigned... Different values ​​are used to reflect the influence of human physiological rhythms on eye tolerance. In this embodiment, the dynamic adjustment of traditional Chinese medicine time is as follows: The hours of Zi (11:00 PM) and Chou (3:00 AM): These correspond to the peak of the Gallbladder and Liver meridians (belonging to the Wood element). Traditional Chinese medicine believes that "when a person lies down, blood returns to the liver," meaning the liver and gallbladder should rest and recuperate during this time. Forcing oneself to stay up late and strain one's eyes will further damage liver blood. Therefore, it is advisable to appropriately increase the fatigue index during this period. (+10%). This means that the same objective visual load will be perceived as a higher level of fatigue at midnight, serving as a reminder of the potential harm of nighttime eye strain.

[0070] Early morning (3:00-7:00 AM): Corresponding to the Lung and Large Intestine meridians being at their peak (belonging to the Metal element). After a night's rest, the body's Yang energy rises in the early morning, and the visual system has largely recovered. Furthermore, the Metal element in the meridians suppresses the Wood element, helping to alleviate the overactivity of the Liver. Therefore, during this period, fatigue levels can be slightly reduced. (-5%) indicates that the fatigue level in the morning is slightly lower than usual under the same load.

[0071] The afternoon hours of Shen Shi (3:00 PM – 7:00 PM) correspond to the Bladder and Kidney meridians being at their peak (belonging to the Water element). Kidney Water nourishes Liver Wood, and during this period, the body's energy is relatively abundant, with better tolerance for eye strain. Furthermore, most people experience peak daytime eye strain during this time, so short-term fatigue and discomfort are not significant. Therefore, the fatigue index can be slightly reduced, and settings can be adjusted accordingly. (-3%) indicates that eye strain was slightly reduced during this period.

[0072] Other time periods (Chen, Si, Wu, Xu, Hai, etc.): During these periods, the meridians belong to the Earth or Fire element, and the impact of fatigue is relatively neutral. No additional adjustments are made to the index.

[0073] The above values ​​are examples and can be further adjusted and optimized based on clinical and user data. The final Eye Fatigue Index (EFI) is obtained by applying a time-correction factor: .

[0074] To enhance early warning, the model's sensitivity is increased during critical periods such as nighttime, while indicators are appropriately relaxed in the early morning or afternoon to align with physiological rhythms. This adjustment is also used in subsequent steps to develop personalized intervention strategies.

[0075] It is important to emphasize that TCM adjustments are only a secondary calibration, with their magnitude strictly controlled within ±10%. The aim is to provide preventative reminders in accordance with physiological rhythms, and not to mask or reverse the main trends indicated by the physical model. In other words, the EFI baseline remains primarily driven by objective data, while the time-related factors provide subtle adjustments and changes in alerting strategies. Furthermore, The module can also incorporate machine learning methods to validate and optimize the above rules by accumulating a large amount of user data, so as to ensure that the corrections conform to the actual experience of most users.

[0076] Step 5: Setting fatigue thresholds and implementing reminder intervention strategies.

[0077] To translate EFI values ​​into practical interventions, appropriate thresholds need to be set. Based on research and experimental results, the eye fatigue index is divided into multiple levels, and corresponding thresholds are set. For example, EFI = 70 can be defined as the fatigue warning threshold, and EFI = 85 can be defined as the high fatigue threshold (severe fatigue). The specific threshold selection can be determined through ROC curve analysis of labeled data, ensuring timely alerts while minimizing false alarms.

[0078] When a user's EFI (Eye Fatigue Index) rises continuously and exceeds the warning threshold (e.g., 70) for the first time, the system triggers an eye protection reminder. The reminder content will comprehensively consider the current time, scenario, and fatigue factors. For example, in a typical daytime scenario, it may prompt, "Your eye fatigue index is high. We recommend following the 20-20-20 rule for rest: look at something 20 feet away for 20 seconds every 20 minutes." In the late night, the prompt will emphasize the importance of rest, such as, "It is past midnight. Prolonged use of your eyes while staying up late can damage your liver. Please stop using your eyes and rest as soon as possible." When the EFI continues to climb above the high fatigue threshold (e.g., 85), the system may increase intervention efforts, including: popping up a full-screen strong reminder, temporarily reducing screen brightness, and even, in extreme cases, suggesting that the user take a mandatory break for a few minutes before continuing to use the device (this mandatory measure can be enabled in advance by the user). The reminder can take the form of a notification, pop-up window, voice broadcast, or vibration, striving to draw sufficient attention without interfering with the user's normal use.

[0079] In addition to numerical thresholds, the reminder strategy of this invention also combines specific factor triggers and content-based suggestions: When a certain sub-indicator remains in an extremely unfavorable state, targeted suggestions can be given in advance even if the EFI has not exceeded the threshold. For example, when the ambient light is severely insufficient and the screen is too bright, a message can be given in advance: "The environment is too dark. Please turn on the lights or lower the screen brightness." Or, if it is detected that the user has been staring for a long time without blinking, a message can be given: "You blink too infrequently. It is recommended that you consciously blink to keep your eyes moist."

[0080] The reminders integrate modern ophthalmological knowledge with traditional Chinese medicine health advice. For example, in addition to the usual advice to rest and look into the distance, it may suggest "applying warm compresses to the eyes or massaging the Jingming acupoint to help relieve eye fatigue," and remind users to drink more water to nourish the liver and kidneys during the late afternoon and late evening hours. These suggestions are dynamically matched according to the time of day and user habits, achieving truly personalized health guidance.

[0081] Step Six: Based on the collected historical eye use data, use multiple regression or machine learning models to adjust the weights of key visual load indicators for each user under different eye use scenarios; dynamically adjust the comprehensive eye fatigue coefficient according to the eye use habits of different users to achieve personalized eye fatigue assessment.

[0082] To balance applicability to the general population with individual differences, this method provides online model learning and dynamic optimization mechanisms, including: Adaptive weight adjustment: initial weights of the model ( , The weights (such as EFI) are derived from general user experience. However, each user's eye usage characteristics are different, and the system can adaptively adjust these weights based on user feedback. For example, every few hours, the system can pop up a simple survey asking the user about their current eye comfort (e.g., a 1-5 rating, and asking if they experience dryness, soreness, etc.). Simultaneously, it combines objectively collected parameters such as blink frequency and pupil size changes at that moment to establish a relationship between the user's subjective feelings and the EFI index. Using accumulated data, multiple regression or machine learning models are employed to adjust the weights of each component, making the calculated EFI more consistent with the user's actual experience. After a period of learning, each user will have a personalized set of parameters, making the assessment results more accurate for that individual.

[0083] Personalized Thresholds and Strategies: Different users have varying tolerances and preferences for fatigue. By recording user feedback on reminders (such as whether they feel the reminders are too early or too late, and whether they accept the suggestions), the system can appropriately adjust the user's thresholds and reminder strategies. For example, if a user frequently experiences eye strain when their EFI reaches 65, their personal warning threshold can be lowered to 65; conversely, if a user tends to continue using their eyes when experiencing mild fatigue, the threshold can be increased or the reminder frequency reduced. The reminder content and intervention measures can also be optimized on a personalized basis. For example, providing more scientific data explanations to enhance persuasiveness for some users, while using a gentler tone to avoid offending others.

[0084] like Figure 8 The figure shows a schematic diagram of the ROC curve for detecting fatigue state using the EFI index. The area under the shaded area, AUC value, represents the model accuracy. Figure 9 The chart shows the distribution of personalized fatigue thresholds for different users. This chart indicates that after the model of the present invention is adjusted to be personalized, the reminder thresholds for different users may be different, but all can effectively identify high fatigue moments.

[0085] Model adaptive learning: By regularly collecting subjective user feedback (such as eye comfort scores) and objective physiological parameters (blinking frequency, pupil changes, etc.), the weights and thresholds of each indicator in the model are continuously adjusted to achieve precise adaptation to different users, becoming smarter the more it is used.

[0086] Data security and privacy: User privacy and data security are fully considered during model optimization. Sensitive raw physiological data is processed only locally on the user's device; only anonymized and aggregated statistical information is uploaded to the server. Users can also choose to opt out of data sharing and use only the local model. This ensures that user privacy is not violated while improving model intelligence.

[0087] Through the synergistic effect of step six in step one above, this invention achieves comprehensive assessment and intelligent management of eye fatigue when using digital devices. Compared with traditional single-factor alerts, this method can more accurately quantify the degree of fatigue, adapt more flexibly to changes in scenarios, and proactively help users prevent eye strain.

[0088] Through the above innovative combination, this invention can continuously assess the user's eye fatigue level during normal use of digital devices. When the fatigue index is detected to be rising and approaching a threshold, the system will promptly issue easy-to-understand eye care tips and suggestions to the user based on the current time and scenario (such as following the "20-20-20" rule to rest the eyes, looking into the distance to relax, applying warm compresses to the eyes, or performing eye exercises, etc.), helping users preventively relieve eye fatigue before discomfort becomes obvious. This fully embodies the traditional Chinese medicine concept of "prevention before disease," proactively protecting the user's eye health using smart terminal technology.

[0089] The effectiveness of the method described in this invention is demonstrated through the following application examples in this embodiment: Application Example 1: Fatigue Assessment of Sedentary Indoor Reading Scenario Description: A user is sitting quietly in a softly lit indoor environment, reading an e-book on their phone. The environment is stable and motionless, with moderate illumination. The reading session lasts for a relatively long time (approximately 1 hour), and the scenario takes place around 11:30 PM.

[0090] Data characteristics: The phone remained stationary and was held steadily. The accelerometer and gyroscope readings showed very little change; after filtering, the phone's motion relative to the head can be considered almost zero, therefore SI ≈ 0. The ambient light sensor measured the indoor illuminance. The user enabled the phone's eye protection mode and lowered the brightness to 30%. The screen brightness was then calculated using calibration. Based on the device's comfortable brightness ratio in this environment. r (Assuming) r ≈ 0.8 cd / (m²·lux)), ideal screen brightness is approximately The brightness error was calculated. This means the screen brightness is almost perfectly matched to the ambient light (the warmer color temperature in eye protection mode also reduces visual stimulation). In terms of interaction, reading mainly involves occasional page turning and scrolling, with a swipe rate of [number] ms per minute. Click time Secondly, the operation rhythm is slow. This leads to the interaction intensity. Very low, let's assume it's 5 (using a standardized score of 0–100). Continuous screen time t = 60 minutes. According to the duration model of the present invention, when t After 60 minutes, fatigue accumulation will gradually approach saturation (assuming saturation corresponds to 100 points, at which point it will reach approximately 80 points). If the user takes a short break of 1 minute during this period, the recovery factor will... This indicates that 10% of the previously accumulated 80 points of fatigue can be recovered, representing an effective fatigue duration factor. The score dropped to approximately 72 points.

[0091] Calculation process: Since this scenario is a reading context, the corresponding model weights are assumed to be... , , , (Emphasis on duration). 23:30 at night belongs to the sub-period, a period influenced by Traditional Chinese Medicine correction factors. Substitute the data into the comprehensive model: , , , (Consider a short rest and recovery period); Then the fatigue coefficient of the physical layer:

[0092] Will The normalized mapping is to the standard score (assuming a full scale of 100 corresponds to the physical model at full load, here 33.4 corresponds to approximately 33.4 points); Traditional Chinese Medicine Correction: point; The results showed that after one hour of reading, the EFI (Eye Fatigue Index) rose to approximately 37 (not exceeding the warning threshold of 75, but significantly higher than the initial value). Because the user was reading late at night, the rate of increase was slightly higher than during the day. The system recorded the upward trend of the EFI curve in the background. When the EFI approached the threshold (e.g., around 70), a reminder would pop up to notify the user to rest. In this example, assuming the user continued reading for about 90 minutes, the EFI gradually climbed and approached 75. At this point, the system might issue a prompt: "You have been using your eyes continuously for 1.5 hours, and your eye fatigue index is high. It is recommended to stop and rest for 5-10 minutes, and look into the distance to relax your eyes." If the user ignored the reminder and continued using their eyes, around 1 a.m. (the beginning of the "Chou" hour), the EFI might further rise to over 80, triggering a stronger reminder: "It is now the early morning; your liver and gallbladder need rest. It is recommended that you stop using your eyes as soon as possible and ensure you get some sleep." This scenario demonstrates that this method provides a reasonable assessment of the cumulative fatigue from prolonged static eye use and utilizes traditional Chinese medicine's rhythms to specifically emphasize the effects of staying up late on eye use, making the reminders preventative. If users take appropriate rest as advised, their fatigue index will decrease, thus preventing further deterioration.

[0093] Application Example 2: Fatigue Assessment of Watching Videos Inside a Moving Train Car Scenario Description: A user is watching a streaming video on their mobile phone while riding a bus for 30 minutes. The moving vehicle causes the environment to shake, and the lighting fluctuates. The user is standing and holding the phone in one hand. The scene takes place around 6:00 PM.

[0094] Data characteristics: The vehicle's bumps and swaying cause the phone to constantly move relative to the user's line of sight. Accelerometer / gyroscope data fluctuates dramatically; for example, the acceleration signal shows a significant peak in the spectrum (approximately 5 Hz, corresponding to vehicle vibration). This data is processed to obtain data within a certain time window. , ,and The indicators are also high (the video content changes rapidly). This results in a significant increase in SI, several orders of magnitude higher than in a static environment. Regarding ambient light, the streetlights and shadows outside the bus alternate, and the ambient light intensity inside the bus... (For example, brightness can reach 100 lux in well-lit areas and drop to 20 lux in dimly lit areas). The phone's automatic brightness adjustment is enabled but exhibits a lag, resulting in an overall... It remains at approximately 150 cd / m². Therefore, when the interior dims while the screen remains bright, the instantaneous brightness error can reach [a certain value]. Assuming r=1 at that time (fully compensated environment), ideally... ,but This represents a very high brightness contrast; even if a slightly higher brightness is preferable (for example, a screen brightness of 50 cd / m² is desired). A value above 0.67 indicates the screen is significantly too bright and glaring. Users rarely touch the screen while watching videos, except for occasional adjustments to volume or brightness; this can be considered... Continuous usage time t = 30 minutes, converted according to the duration model. Approximately 50 minutes (not yet saturated). 18:00 in the evening falls within the You hour (the kidney meridian is at its peak, belonging to the water element), adjusted according to Traditional Chinese Medicine. (Slightly reduced).

[0095] Calculation process: Since it is a short-video scene, the weights are assumed to be... , , , (Weighting brightness and motion relative to reading). A typical value at a given moment: , , , ,but: Physical layer fatigue coefficient: ; The standard score mapped to this is approximately 48 points. Traditional Chinese Medicine Correction: EFI = 48 × (1 - 0.03) ≈ 46.6 points; As can be seen, the EFI (Effective Visual Index) approached 50 after only 30 minutes of video viewing. Although the duration is short, the rate of fatigue accumulation is much faster than in quiet reading in Example 1 due to shaking and poor brightness. This indicates that eye strain can be triggered significantly in a short time when using the eyes in complex environments. In this scenario, the present invention will intervene earlier with a reminder: when SI (Self-Impact Index) and... If the EFI remains high for an extended period, even if it hasn't reached the 75 threshold, a warning will be triggered in advance according to the rules. For example, after a user has watched for about 15–20 minutes and experienced multiple shaking and strong light changes, the system may display a prompt: "The environment is shaking and the light is unstable. Please minimize your screen time. If possible, pause watching or listen to audio instead." This allows the user to temporarily stop watching or close their eyes to rest, waiting for the environment to stabilize before continuing, thus preventing the fatigue index from continuing to rise rapidly. If the user persists in watching for 30 minutes or even longer, the EFI is likely to exceed the 75 threshold at 40–50 minutes, at which point the system will again remind and strongly recommend stopping screen time. This application example demonstrates the sensitive response and proactive warning capability of the model of this invention to adverse visual environments. It can provide early intervention based on environmental load when traditional calculations based on screen time are insufficient, which aligns with the spirit of "prevention is better than cure."

[0096] Application Example 3: Fatigue Assessment of Nighttime Close-Range Mobile Phone Chatting Scene description: Late at night (around 01:00, the period of the Ox in Chinese lunar hours), the user lies on the bed, browsing social media and chatting messages on the mobile phone sideways. The room lights are turned off and the surrounding environment is almost completely dark. To see the small words on the screen clearly, the user holds the mobile phone very close to the eyes (about 25 cm). The continuous usage time is short, about 15 minutes.

[0097] Data characteristics: The mobile phone and the user's body are almost stationary, SI≈0. The ambient illuminance is very low (<5 lux). Even if the mobile phone screen is adjusted to a low brightness, it still appears very bright in the dark. It is estimated that (the mobile phone may have automatically dropped to the lowest brightness level but is still dazzling in a completely dark environment). Assume that the comfortable brightness ratio r of the device at this time is approximately 0.2 (a weak brightness can be maintained in an extremely dark environment to see clearly), and the ideal is about 1 cd / m², but the actual brightness of 80 cd / m² is much higher than this value. Therefore, the brightness error , close to the maximum value, indicating that the screen is severely over-bright. The user is browsing social content, and the interaction behaviors include swiping up and down the message list and typing replies, with a medium frequency: 5 swipes per minute and 30 - 40 characters typed. Considering the scrolling and input rhythm and assigning corresponding weights, a rough estimate . The continuous eye usage time t = 15 minutes. According to the duration model is about 20 (the fatigue accumulation is still in the early stage). However, due to the extremely close distance between the eyes and the screen (25 cm), the system will trigger a distance penalty . Assume that when the distance is less than 30 cm, a penalty of 1 point is added for every 1 cm reduction (simplified model). Then, when the distance is reduced from 30 cm to 25 cm, 5 points are added, (or calculated according to a more complex model, an approximate value is taken in this example). 1:00 am belongs to the period of the Ox in Chinese lunar hours, .

[0098] Calculation process: In the chat / social scenario, the model may pay more attention to the brightness and distance factors. Let , , , , 0 considering other factors such as distance penalty. Substitute the above data: , , , , , then: Physical layer fatigue coefficient: ; After normalization, it is about 45.7 points; Traditional Chinese medicine correction: EFI = 45.65 × (1 + 0.1) ≈ 50.2 points; Within just 15 minutes, the EFI exceeded 50, demonstrating the rapid accumulation of visual fatigue under extremely unfavorable conditions (complete darkness + close proximity). Although the EFI had not yet reached the threshold of 75, several sub-indicators were already in dangerous states (almost at maximum). , plus The system will issue a warning very promptly. It might pop up a message after only 5 minutes of use: "You are currently in a dark environment and the screen is too close to your eyes, causing extreme eye strain. Please turn on the lights or move the device away as soon as possible, and stop using it if necessary." This is proactive intervention to prevent users from continuing to use their eyes in such a poor environment. If the user adjusts their posture or environment, such as turning on a night light to reduce eye strain... Holding the phone further away reduces the distance penalty, significantly slowing the rate of increase in the fatigue index and potentially preventing further alerts. Conversely, if the user ignores this and continues close-range, dark use, the EFI (Eye Fatigue Index) may exceed the warning threshold of 75 within approximately 30 minutes. At this point, the system will issue another alert and strongly recommend stopping. Once the user complies with the advice to stop using the phone and rest with their eyes closed for a few minutes, the fatigue index will rapidly decrease (resting in the dark effectively relieves pressure on the retina and ciliary muscles). This application example demonstrates the invention's sensitive detection capability under high-risk eye-use behaviors, enabling it to detect adverse factors and provide clear corrective suggestions in a very short time, preventing problems before they arise.

[0099] Application Example 4: Assessment of eye fatigue in low-light flicker environments Scenario Description: A user is working overtime in the office at 7:30 PM, using their phone to research information for about 20 minutes. Only one old-fashioned fluorescent light is on in the room, providing dim lighting with noticeable flickering. For comfort, the user lowers the phone's brightness and activates eye protection mode; simultaneously, due to fatigue, they recline on the office sofa to rest while using their phone.

[0100] Data characteristics: Low ambient light and flickering light. The average value measured by the ambient light sensor... However, the fluctuations are significant, with the camera's high-speed sampling frames showing approximately 100 Hz flickering in the lights. The phone's eye protection mode filters out some blue light and reduces screen brightness. Approximately 30%, converted Due to the dark environment, the ideal brightness should be even lower (assuming r=0.5, then the expected brightness is...). ²), The actual brightness of the current screen is higher than expected; the brightness error is roughly estimated. If you're considering the slight improvement in visual comfort offered by the eye protection mode, then it's effective. (Still relatively high). Users scroll through webpage content at a slower speed, 3 swipes per minute. Usage time t = 20 minutes, (Not yet at fatigue saturation level). This is the You period (Kidney meridian at its peak). What's special about this scenario is the presence of flickering and a side-lying posture: the flickering is triggered by an algorithm detecting the frequency and modulation depth. Sleeping on your side may cause one eye to receive more screen light and result in uneven tear distribution, making it prone to dry eyes. In addition, users may blink less when fatigued. The index may rise. The system will respond to these factors.

[0101] Calculation process: Assuming webpage browsing (belonging to a general reading scenario) uses weights. , , , , 0.1, The system detected flashing lights at this moment. A score was given based on the user's low blink rate. Additional value = 5 points (total additional 15). Substitute the data: , , (No obvious relative motion) ,but Physical layer fatigue coefficient: ; The normalized score is approximately 26.5. Traditional Chinese Medicine correction: EFI = 26.5 × (1 - 0.03) ≈ 25.7 points; The result was only around 25 points, far below the warning threshold. This is because the continuous screen time was short and the overall pace was slow; despite the unfavorable environmental factors, the fatigue index remained within a safe range. However, this invention monitors for flicker and posture issues in the background and provides timely improvement suggestions. A prompt might pop up after a few minutes of use: "Flickering ambient light has been detected. Prolonged use under this light may cause eye discomfort. We recommend improving lighting conditions (turning on other light sources or replacing with more stable lamps)." If the user follows the suggestion and turns on another desk lamp, the ambient brightness increases and the flickering effect decreases. and Both will decrease, and the EFI (Eye Fatigue Index) will rise more slowly. On the other hand, if the user continues to view the phone while lying on their side, the system may later prompt: "Prolonged viewing while lying on your side can lead to uneven light exposure in the eyes, causing dryness. It is recommended to sit up straight or blink occasionally to rest your eyes." These prompts are given in the early stages of fatigue accumulation, helping users optimize their eye environment and posture, thereby preventing a rapid increase in EFI later. This application example demonstrates that the invention not only focuses on the absolute value of the fatigue index but also detects and provides feedback on potential factors affecting fatigue, embodying a comprehensive and proactive eye protection concept.

[0102] As can be seen from the above embodiments, the method of the present invention can flexibly adapt to different usage scenarios and environmental conditions, comprehensively assess various factors causing visual fatigue, and provide timely feedback. Whether it is cumulative fatigue in a quiet environment or instantaneous load in a complex environment, this method can accurately quantify and provide targeted suggestions, truly realizing scientific and intelligent eye health management on digital devices.

Claims

1. A method for real-time assessment of eye fatigue when using digital devices, characterized in that: Specifically, the steps include the following: Step 1: Multi-sensor data acquisition and eye usage scenario recognition; It acquires real-time data on device motion, ambient light intensity, screen brightness, and user interaction during the user's use of digital devices; and classifies eye-use scenarios by combining the type of application running in the foreground, the changing patterns of screen content, and user interaction characteristics; at the same time, it uses the front-facing camera to capture the user's facial image and uses the front-facing camera or a distance sensor to estimate the distance between the eyes and the screen. The device motion data includes linear acceleration data collected by the device's built-in accelerometer and angular velocity data collected by the device's built-in gyroscope; the ambient light intensity data includes ambient illuminance measured by the device's built-in ambient light sensor. Screen brightness parameters include the current device screen brightness setting and screen dimming mode; user interaction data includes the number of times the user interacts with the screen per unit time, records the speed and frequency of scrolling operations, application switching, and reflects the user's pace and level of engagement when viewing information. Eye-use scenarios include reading static text, browsing short videos, playing games, social communication, and video calls. If they cannot be clearly categorized, they are marked as general browsing / idle time. Step 2: Calculation of key visual load indicators; The relative motion instability of the screen is calculated based on the relative angular velocity and relative acceleration between the screen and the user's head, as well as the rate of content change. Calculate the brightness adaptation error based on screen brightness and ambient brightness. Calculate user interaction intensity based on user interaction data The continuous eye use duration factor is calculated using a piecewise function. ; Step 3: Comprehensive coefficient of eye fatigue calculate: Based on the eye use scenario, the key visual load indicators calculated in step two are linearly combined using different weights to obtain the comprehensive coefficient of eye fatigue. Step 4: Dynamic calibration of the comprehensive eye fatigue coefficient: The comprehensive coefficient of eye fatigue calculated in step three is dynamically calibrated by combining the twelve two-hour periods of traditional Chinese medicine and the Five Elements theory; different time-based correction factors are assigned to different time periods according to the two-hour periods of a day and their corresponding meridian Five Elements attributes. To reflect the influence of human physiological rhythms on eye tolerance; Step 5: Set fatigue thresholds and develop reminder and intervention strategies based on the Twelve Time Periods and Five Elements theory of Traditional Chinese Medicine. Set an early warning threshold and a threshold for exceeding the high fatigue level for the comprehensive eye fatigue coefficient. When the comprehensive eye fatigue coefficient, after calibration in step four, rises continuously and exceeds the early warning threshold for the first time, an eye protection reminder is triggered. When the comprehensive eye fatigue coefficient continues to rise and exceeds the high fatigue threshold, the intervention intensity should be increased. When the ambient light is severely insufficient and the screen is too bright, the user stares at the screen for a long time without blinking, or the user holds the digital device too close to the screen, a reminder will be given and corresponding suggestions will be provided even if the overall eye fatigue coefficient has not exceeded the threshold. Step 6: Based on the collected historical eye use data, use multiple regression or machine learning models to adjust the weights of key visual load indicators for each user under different eye use scenarios; dynamically adjust the comprehensive eye fatigue coefficient according to the eye use habits of different users to achieve personalized eye fatigue assessment. Through the synergistic effect of steps one through six above, real-time assessment of eye fatigue when using digital devices is achieved.

2. The method for real-time assessment of eye fatigue when using digital devices as described in claim 1, characterized in that: High-frequency noise in the data collected by each sensor in step one is removed by a filtering algorithm.

3. The method for real-time assessment of eye fatigue when using digital devices as described in claim 1, characterized in that: In step one, the front-facing camera of the digital device is used to periodically sample the environmental scene and correct the ambient light sensor readings to obtain a more accurate ambient brightness; the screen uses PWM dimming mode.

4. The method for real-time assessment of eye fatigue when using digital devices as described in claim 1, characterized in that: In step one, user facial feature data is also collected. When user authorization is obtained and the application scenario requires it, the front-facing camera is periodically activated to collect user facial images for a short period of time. Facial key points are detected through locally running visual algorithms, blink frequency, average eye opening degree, pupil diameter changes are extracted, and head posture and approximate distance between the eyes and the screen are estimated or measured.

5. The method for real-time assessment of eye fatigue when using digital devices as described in claim 1, characterized in that: In step two: screen relative motion instability This measure assesses the stability of a digital device's screen image within the user's field of vision. When the device moves violently or the environment shakes, the eye needs to frequently adjust to track the screen, leading to increased eye strain; brightness adaptation error... Used to assess the degree of match between screen brightness and ambient brightness; screen brightness that is too bright or too dim relative to ambient brightness will increase the burden on eye accommodation; user interaction intensity. This factor is used to characterize the eye strain caused by user actions. Frequent and rapid interactions mean that users browse a large amount of information in a short period of time, requiring the eyes to move quickly and focus on different areas, which easily leads to visual fatigue; continuous eye use duration factor. It is used to reflect the cumulative degree of eye fatigue caused by continuous use of the eyes.

6. The method for real-time assessment of eye fatigue when using digital devices as described in claim 1, characterized in that: In step three, the comprehensive coefficient of eye fatigue is normalized or mapped to obtain an intuitive and unified fatigue index value.

7. The method for real-time assessment of eye fatigue when using digital devices as described in claim 1, characterized in that: The time correction factor in step four The value ranges from [-0.1, +0.1]; the final eye fatigue index is obtained through the effect of the time correction factor. .

8. The method for real-time assessment of eye fatigue when using digital devices as described in claim 1, characterized in that: In scenarios where flickering is present, although Although the threshold is not exceeded, the flicker penalty will increase the overall visual load assessment; when calculating the comprehensive eye fatigue coefficient in step four, flicker penalty, distance penalty and dry eye factor are introduced to increase the contribution of brightness discomfort, usage distance and dry eye symptoms to fatigue; the flicker is detected by high-speed sampling of ambient light or by observing the flicker frequency of the light source using a camera.

9. The method for real-time assessment of eye fatigue when using digital devices as described in claim 1, characterized in that: In step five, the warning threshold and the threshold for exceeding the high fatigue level are determined through ROC curve analysis of labeled data, so as to minimize false alarms while ensuring timely reminders.

10. The method for real-time assessment of eye fatigue when using digital devices as described in claim 1, characterized in that: During the model learning process, sensitive raw physiological data is processed only locally on the user's device, while the data uploaded to the server is anonymized and aggregated statistical information; users can also choose to opt out of data sharing and use only the local model; ensuring that user privacy is not violated while improving model intelligence.

Images